text
stringlengths
254
1.16M
--- title: Acid–base imbalance as a risk factor for mortality among COVID-19 hospitalized patients authors: - Nosayba Al-Azzam - Basheer Khassawneh - Sayer Al-Azzam - Reema A. Karasneh - Mamoon A. Aldeyab journal: Bioscience Reports year: 2023 pmcid: PMC10037419 doi: 10.1042/BSR20222362 license: CC BY 4.0 --- # Acid–base imbalance as a risk factor for mortality among COVID-19 hospitalized patients ## Abstract Severe coronavirus disease 2019 (COVID-19) infection can lead to extensive lung infiltrate, a significant increase in the respiratory rate, and respiratory failure, which can affect the acid–base balance. No research in the Middle East has previously examined acid–base imbalance in COVID-19 patients. The present study aimed to describe the acid–base imbalance in hospitalized COVID-19 patients, determine its causes, and assess its impact on mortality in a Jordanian hospital. The study divided patients into 11 groups based on arterial blood gas data. Patients in normal group were defined as having a pH of 7.35–7.45, PaCO2 of 35–45 mmHg, and HCO3− of 21–27 mEq/L. Other patients were divided into 10 additional groups: mixed acidosis and alkalosis, respiratory and metabolic acidosis with or without compensation, and respiratory and metabolic alkalosis with or without compensation. This is the first study to categorize patients in this way. The results showed that acid–base imbalance was a significant risk factor for mortality ($P \leq 0.0001$). Mixed acidosis nearly quadruples the risk of death when compared with those with normal levels (OR = 3.61, $$P \leq 0.05$$). Furthermore, the risk of death was twice as high (OR = 2) for metabolic acidosis with respiratory compensation ($$P \leq 0.002$$), respiratory alkalosis with metabolic compensation ($$P \leq 0.002$$), or respiratory acidosis with no compensation ($$P \leq 0.002$$). In conclusion, acid–base abnormalities, particularly mixed metabolic and respiratory acidosis, were associated with increased mortality in hospitalized COVID-19 patients. Clinicians should be aware of the significance of these abnormalities and address their underlying causes. ## Introduction The new strain of the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) disease, which is known as COVID-19, spreads through sneezing and coughing droplets [1]. Coronavirus disease 2019 (COVID-19) spread worldwide after emerging in Wuhan, China. COVID-19 affects many systems of the human being, mainly the respiratory system, through acting on angiotensin-converting enzyme 2 (ACE-2) receptors that are located at the surface of respiratory cells [2,3]. COVID-19 has a huge impact on other systems that harbor ACE-2 receptors on their surfaces, such as the cardiovascular, neurological, gastrointestinal, and musculoskeletal systems [4]. SARS-CoV-2 is a highly contagious virus [5], and infected individuals are infectious before symptoms appear [4]. However, estimates differ on the significance of asymptomatic people spreading the virus [6]. SARS-CoV-2 can cause a variety of symptoms, including asymptomatic infection and severe pneumonia-induced death [5]. The clinical range of the disease manifests as mild, moderate, or severe illness [7,8]. Intensive therapy, which includes non-invasive and invasive ventilation, antipyretics, antivirals, antibiotics, and steroids, is necessary for moderate and severe cases, which also call for hospitalization. Plasma exchange therapy and immunomodulatory drugs may be necessary for the treatment of complicated cases [7]. A COVID-19 infection can lead to a deterioration of lung capacity brought on by pulmonary interstitial fibrosis [9]. SARS-CoV-2, on the other hand, causes excessive and prolonged cytokine and chemokine responses in some infected individuals, a phenomenon known as the ‘cytokine storm’. Cytokine storms cause acute respiratory distress syndrome (ARDS), or multiple-organ dysfunction, which leads to physiological deterioration and death [10]. Several predictors of COVID-19 severity and mortality have been identified [11]. In a retrospective, observational cohort study of 3988 consecutive critically ill patients with laboratory-confirmed COVID-19 referred for ICU admission to the coordinating center in Italy, it was discovered that independent risk factors associated with mortality included older age, male sex, and histories of chronic obstructive pulmonary disease, hypercholesterolemia, and Type 2 diabetes mellitus [12]. Another study conducted in Jordan confirmed that several risk factors, such as older age, smoking, admission severity status, comorbidities, and lab test results, were linked to COVID-19 mortality [13]. A normal acid–base balance is necessary to ensure appropriate physiology and cell activity. The occurrence of any acid–base abnormalities increases the probability of experiencing a negative consequence [14,15]. Acid–base disorders are classified as either respiratory or metabolic, depending on the carbon dioxide (CO2) tension and the bicarbonate ions (HCO3−) levels in physiological fluids [16,17]. Acid–base changes are caused by a number of illnesses, including respiratory failure, shock, renal failure, and hepatic failure [18]. An early and correct identification of an acid–base imbalance is required to improve the outcome because severe acid–base derangements can be life-threatening [15]. As a common acid–base disorder in COVID-19 patients, respiratory alkalosis was linked to an increased risk of severe events [19]. Understanding normal physiological function is the first step in interpreting acid–base disorders. To maintain acid–base balance, the body employs buffering processes, ventilation rate, and renal mechanisms. The use of pH, pCO2, and HCO3− as coordinates can help to identify compensation and mixed acid–base disorders [17,20]. In a retrospective study of 112 COVID-19 patients who were hospitalized at the University Hospital of Modena, it was discovered that $79.7\%$ of the patients had abnormal acid–base balances. Metabolic alkalosis was the most significant change, and it was followed by respiratory alkalosis, combined alkalosis, respiratory acidosis, metabolic acidosis, and other compensated acid–base disturbances, in that order [21]. Another study that included 105 COVID-19 patients with ARDS within the first 48 h of needing noninvasive respiratory support discovered that the majority of the patients had respiratory alkalosis. Metabolic alkalosis, the second more common acid–base disorder, was mentioned. Only a small percentage of the patients had respiratory acidosis, and none of the patients had metabolic acidosis [22]. Although there have been some studies about acid–base imbalance in COVID-19 patients, there have been none from the Middle East. Consequently, the present study’s objectives were to describe the acid–base imbalance in hospitalized COVID-19 patients, identify the risk factors for this imbalance, and assess how it affected in-hospital mortality for COVID-19 patients in Jordan. In addition, the present study aimed to describe the clinical characteristics of patients with and without acid–base imbalances and examine the association between distinct acid–base imbalances and the outcome of adult COVID-19 patients. ## Data sources and research design This retrospective study was carried out at King Abdullah University Hospital (KAUH), a tertiary hospital and one of the largest medical structures in Jordan. Patients with COVID-19 who were admitted to KAUH between September 20, 2020, and August 8, 2021, whose arterial blood gas (ABG) measurements upon admission were reported, and whose SARS-CoV-2 nasopharyngeal swab polymerase chain reaction (PCR) confirmed positivity, were included in the study. Patients under the age of 18 and those with missing ABG parameters on admission or who were asymptomatic were excluded from the study. KAUH is a university teaching hospital, so all data are consented to be used in cohort studies once the patient agreed to be treated in the hospital and the IRB committee approved the study. ## Research variables Electronic hospital records were used to identify patients’ clinical data, such as vital signs, comorbidities, and hospitalization course and outcomes. Laboratory results were also included in the study alongside age, gender, smoking status, height, and weight. The body mass index (BMI) was computed using the formula BMI = weight (kg)/height2 (m2) and categorized according to World Health Organization (WHO) guidelines [23]. Comorbidities were identified using related International Classification of Diseases (ICD) codes, and laboratory results at admission were interpreted using hospital laboratory reference values. According to the National Institutes of Health’s (NIH) Clinical Spectrum of SARS-CoV-2 Infection [24], the severity of the patient’s condition upon admission was classified. Patients with a positive test but no symptoms were classified as ‘asymptomatic’, and those exhibiting multiple symptoms but no respiratory distress were classified as having ‘mild illness’. Patients with lower respiratory disease on clinical assessment or imaging and an oxygen saturation (SpO2) of $94\%$ or more on room air at sea level were classified as having ‘moderate illness’. Patients were classified as having ‘severe illness’ when they exhibited characteristics such as a SpO2 of less than $94\%$, a PaO2/FiO2 of less than 300 mmHg, a respiratory rate of more than 30 breaths per minute, or lung infiltrates of more than $50\%$. Patients with respiratory failure, septic shock, and/or multiple organ dysfunction were considered ‘critically ill’. ## Patient classification ABG values at admission were used to categorize patients. Table 1 displays the ranges used to categorize the patients. Patients were deemed normal only if their laboratory arterial blood pH was between 7.35 and 7.45, PaCO2 was between 35 and 45 mmHg, and HCO3− level was between 21 and 27 mEq/L. Patients with respiratory acidosis had a high CO2 concentration (>45) and a pH less than 7.4, whereas patients with respiratory alkalosis had a low CO2 concentration (<35) and a pH greater than 7.4. On the other hand, metabolic alkalosis is considered when HCO3− exceeds 27 mEq/L and pH exceeds 7.4, whereas metabolic acidosis occurs when pH falls below 7.4 and HCO3− falls below 21 mEq/L. The compensatory mechanisms from the respiratory or metabolic sides were also considered in categorizing the patients, as illustrated in Table 1. **Table 1** | Patients’ classification | pH | CO2 | HCO3− | | --- | --- | --- | --- | | Normal | 7.35–7.45 | 35–45 | 21–27 | | Mixed acidosis | <7.35 | >45 | <21 | | Mixed alkalosis | >7.45 | <35 | >27 | | Respiratory acidosis with compensation (RAC) | <7.4 | >45 | >27 | | Respiratory acidosis with no compensation (RANC) | <7.4 | >45 | 21–27 | | Respiratory alkalosis with compensation (RAlkC) | >7.4 | <35 | <21 | | Respiratory alkalosis with no compensation (RAlkNC) | >7.4 | <35 | 21–27 | | Metabolic acidosis with compensation (MAC) | <7.4 | <35 | <21 | | Metabolic acidosis with no compensation (MANC) | <7.4 | 35-45 | <21 | | Metabolic alkalosis with compensation (MAlkC) | >7.4 | >45 | >27 | | Metabolic alkalosis with no compensation (MAlkNC) | >7.4 | 35-45 | >27 | ## Statistical analysis The analysis began with categorizing the patients based on their ABG values on admission. Then, a distributional study of patient features was done across all acid–base classification groups. Summary tables were generated to investigate the proportion of COVID-19 inpatients by age group, gender, and clinical features within each acid–base group. To investigate statistical differences in the frequencies of the categorical groupings, the acid–base status linked with each attribute was studied using χ2 tests. Furthermore, mortality rates, mechanical breathing requirements, and disease severity were assessed using χ2 tests to assess statistical differences across acid–base-categorized groups. In addition to ratio testing, we utilized a nominal logistic regression model to calculate the odds ratio and statistical significance for each of the investigated components and the acid-base status. A two-sided P ≤ 0.05 was regarded as statistically significant. ## Demographic data of the enrolled patients A total of 1233 patients admitted to KAUH with confirmed COVID-19 met the inclusion criteria and were included in the present study during the study period. Based on their arterial blood pH, PaCO2, and HCO3− levels, we classified the patients into 11 groups. Table 2 shows the patients’ blood acid–base state based on their characteristics, comorbidities, and laboratory potassium test. A total of 613 patients ($49.7\%$) were above the age of 65, 718 ($58.2\%$) were men, 527 ($42.7\%$) were obese, 781 ($63.3\%$) had hypertension, and 654 ($53.0\%$) were diabetic. **Table 2** | Count row% | Total (column%) | Normal | Mixed acidosis | Mixed alkalosis | RAC | RANC | RAlkC | RAlkNC | MAC | MANC | MAlkC | MAlkNC | P-value | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | Age | | 18–40 | 100 (8.1) | 25 (25.0) | 0 (0.0) | 0 (0.0) | 7 (7.0) | 10 (10.0) | 18 (18.0) | 17 (17.0) | 13 (13.0) | 4 (4.0) | 3 (3.0) | 3 (3.0) | 0.150 | | 41–65 | 520 (42.2) | 141 (27.1) | 8 (1.5) | 3 (0.6) | 30 (5.8) | 52 (10.0) | 77 (14.8) | 105 (20.2) | 46 (8.9) | 26 (5.0) | 13 (2.5) | 19 (3.7) | | | >65 | 613 (49.7) | 169 (27.6) | 10 (1.6) | 0 (0.0) | 36 (5.9) | 50 (8.2) | 77 (12.6) | 91 (14.9) | 81 (13.2) | 51 (8.3) | 20 (3.3) | 28 (4.6) | | | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | | Male | 718 (58.2) | 201 (28.0) | 8 (1.1) | 1 (0.1) | 41 (5.7) | 61 (8.5) | 105 (14.6) | 130 (18.1) | 77 (10.7) | 43 (6.0) | 23 (3.2) | 28 (3.9) | 0.754 | | Female | 515 (41.8) | 134 (26.0) | 10 (1.9) | 2 (0.4) | 32 (6.2) | 51 (9.9) | 67 (13.0) | 83 (16.1) | 63 (12.2) | 38 (7.4) | 13 (2.5) | 22 (4.3) | | | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | BMI | | Underweight | 2 (0.2) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 1 (50.0) | 1 (50.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0.723 | | Normal | 175 (14.2) | 39 (22.3) | 4 (2.3) | 0 (0.0) | 8 (4.6) | 22 (12.6) | 29 (16.6) | 24 (13.7) | 27 (15.4) | 11 (6.3) | 7 (4.0) | 4 (2.3) | | | Obese | 527 (42.7) | 141 (26.8) | 5 (1.0) | 1 (0.2) | 37 (7.0) | 48 (9.1) | 72 (13.7) | 93 (17.7) | 52 (9.9) | 34 (6.5) | 18 (3.4) | 26 (4.9) | | | Overweight | 418 (33.9) | 121 (29.0) | 8 (1.9) | 2 (0.5) | 19 (4.5) | 34 (8.1) | 59 (14.1) | 71 (17.0) | 52 (12.4) | 28 (6.7) | 9 (2.2) | 15 (3.6) | | | Missing | 111 (9.0) | 34 (30.6) | 1 (0.9) | 0 (0.0) | 9 (8.1) | 8 (7.2) | 12 (10.8) | 24 (21.6) | 8 (7.2) | 8 (7.2) | 2 (1.8) | 5 (4.5) | | | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | Smoking status | | Non-smoker | 908 (73.6) | 259 (28.5) | 14 (1.5) | 2 (0.2) | 53 (5.8) | 79 (8.7) | 119 (13.1) | 151 (16.6) | 99 (10.9) | 64 (7.1) | 29 (3.2) | 39 (4.3) | 0.505 | | Ex-smoker | 157 (12.7) | 44 (28.0) | 1 (0.6) | 0 (0.0) | 7 (4.5) | 16 (10.2) | 26 (16.6) | 24 (15.3) | 18 (11.5) | 10 (6.4) | 4 (2.6) | 7 (4.5) | | | Active smoker | 168 (13.6) | 32 (19.0) | 3 (1.8) | 1 (0.6) | 13 (7.7) | 17 (10.1) | 27 (16.1) | 38 (22.6) | 23 (13.7) | 7 (4.2) | 3 (1.8) | 4 (2.4) | | | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | Comorbidities | | HTN | 781 (63.3) | 207 (26.5) | 17 (2.2) | 3 (0.4) | 49 (6.3) | 70 (9.0) | 99 (12.7) | 124 (15.9) | 101 (12.9) | 63 (8.1) | 21 (2.7) | 27 (3.5) | 0.001 | | DM | 654 (53.0) | 169 (25.8) | 14 (2.1) | 3 (0.5) | 39 (6.0) | 60 (9.2) | 78 (11.9) | 93 (14.2) | 96 (14.7) | 59 (9.0) | 21 (3.2) | 22 (3.4) | 0.0001 | | Dyslipidemia | 69 (5.6) | 16 (23.2) | 5 (7.2) | 0 (0.0) | 4 (5.8) | 5 (7.3) | 8 (11.6) | 12 (17.4) | 6 (8.7) | 9 (13.0) | 3 (4.3) | 1 (1.4) | 0.005 | | IHD | 243 (19.7) | 69 (28.4) | 6 (2.5) | 0 (0.0) | 13 (5.4) | 20 (8.2) | 33 (13.6) | 38 (15.6) | 34 (14.0) | 16 (6.6) | 5 (2.1) | 9 (3.7) | 0.759 | | Atrial fibrillation | 53 (4.3) | 15 (28.3) | 2 (3.8) | 0 (0.0) | 2 (3.8) | 6 (11.3) | 7 (13.2) | 9 (17.0) | 6 (11.3) | 1 (1.9) | 4 (7.6) | 1 (1.9) | 0.485 | | Heart failure | 121 (9.8) | 30 (24.8) | 3 (2.5) | 1 (0.8) | 11 (9.1) | 10 (8.3) | 15 (12.4) | 15 (12.4) | 12 (9.9) | 13 (10.7) | 4 (3.3) | 7 (5.8) | 0.257 | | Asthma | 44 (3.6) | 9 (20.4) | 0 (0.0) | 0 (0.0) | 5 (11.4) | 9 (20.5) | 6 (13.7) | 5 (11.4) | 2 (4.5) | 1 (2.3) | 2 (4.6) | 5 (11.4) | 0.023 | | COPD | 15 (1.2) | 3 (20.0) | 0 (0.0) | 0 (0.0) | 1 (6.7) | 5 (33.3) | 1 (6.7) | 0 (0.0) | 2 (13.3) | 0 (0.0) | 0 (0.0) | 3 (20.0) | 0.006 | | Chronic kidney disease | 118 (9.6) | 20 (16.9) | 5 (4.2) | 0 (0.0) | 7 (5.9) | 9 (7.6) | 12 (10.2) | 6 (5.1) | 35 (29.7) | 19 (16.1) | 3 (2.5) | 2 (1.7) | <.0001 | | ESRD | 32 (2.6) | 6 (18.8) | 1 (3.1) | 0 (0.0) | 1 (3.1) | 5 (15.6) | 2 (6.2) | 1 (3.1) | 12 (37.5) | 3 (9.4) | 0 (0.0) | 1 (3.1) | 0.001 | | Immunocompromised | 52 (4.2) | 11 (21.1) | 2 (3.9) | 1 (1.9) | 3 (5.8) | 3 (5.8) | 6 (11.5) | 10 (19.2) | 9 (17.3) | 4 (7.7) | 2 (3.8) | 1 (1.9) | 0.240 | | Malignancy | 96 (7.8) | 24 (25.0) | 3 (3.1) | 1 (1.0) | 3 (3.1) | 8 (8.3) | 6 (6.3) | 20 (20.8) | 11 (11.5) | 11 (11.5) | 3 (3.1) | 6 (6.3) | 0.085 | | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | Potassium laboratory test | | Low | 105 (8.5) | 12 (11.4) | 1 (0.95 | 1 (0.95) | 6 (5.7) | 6 (5.7) | 14 (13.3) | 28 (26.7) | 8 (7.6) | 5 (4.8) | 7 (6.7) | 17 (16.2) | 0.0001 | | Normal | 879 (71.3) | 261 (29.7) | 12 (1.4) | 2 (0.2) | 50 (5.7) | 75 (8.5) | 131 (14.9) | 154 (17.5) | 94 (10.7) | 49 (5.6) | 25 (2.8) | 26 (3.0) | | | High | 97 (7.9) | 21 (21.7) | 5 (5.2) | 0 (0.0) | 5 (5.2) | 14 (14.4) | 6 (6.2) | 5 (5.2) | 20 (20.6) | 19 (19.6) | 1 (1.0) | 1 (1.0) | | | Missing | 152 (12.3) | 41 (27.0) | 0 (0.0) | 0 (0.0) | 12 (7.9) | 17 (11.2) | 21 (13.8) | 26 (17.1) | 18 (11.8) | 8 (5.3) | 3 (2.0) | 6 (4.0) | | | Total (%) | 1233 | 335 (27.2) | 18 (1.5) | 3 (0.2) | 73 (5.9) | 112 (9.1) | 172 (14.0) | 213 (17.3) | 140 (11.4) | 81 (6.6) | 36 (2.9) | 50 (4.1) | | The distribution of acid–base status among patients hospitalized due to COVID-19 is shown in Figure 1. Among all acid–base classified groups, the highest percent ($27.2\%$) was for patients who had normal ranges for all three measures, followed by respiratory alkalosis with no compensation (RAlkNC) ($17.3\%$), respiratory alkalosis with compensation (RAlkC) ($14.0\%$), and metabolic acidosis with compensation (MAC) ($11.4\%$). **Figure 1:** *The distribution of acid–base status among patients hospitalized due to COVID-19The pie chart represents the distribution of the hospitalized COVID-19 patients according to their acid–base status. Most patients (27.2%) had normal ABG levels. The distribution of patients with abnormal ABG values was as follows: mixed acidosis (1.5%), mixed alkalosis (0.2%), RAC (5.9%), RANC (9.1%), RAlkC (14.0%), RAlkNC (17.3%), MAC (11.4%), MANC (6.6%), MAlkC (2.9%), and MAlkNC (4.1%).* Normal-parameter patients were predominantly elderly (>65 years old, $$n = 169$$), males ($$n = 201$$), obese ($$n = 141$$), and nonsmokers ($$n = 259$$). Most patients among all different variable groups (highest percent in a row), according to the parameters analyzed, are shown to have normal acid–base values, with the exception of those who were active smokers, had asthma, chronic obstructive pulmonary disease (COPD), chronic kidney disease (CKD), end-stage renal disease (ESRD), or have low potassium levels. The bulk of the patients who were underweight were not in the normal category either. As the numbers show in Table 2, RAlkNC was found to have the highest percent of acid–base imbalance among active smokers ($22.6\%$). Among asthmatics, the dominant classification was respiratory acidosis with no compensation (RANC) ($20.5\%$) or normal ($20.4\%$) acid–base levels, whereas the main acid–base imbalance among COPD patients was RANC ($33.3\%$). While MAC was the major acid–base classification among patients with chronic renal disease ($29.7\%$) and ESRD ($37.5\%$). Further, hypertension, diabetes, dyslipidemia, asthma, COPD, CKD, and ESRD comorbidities, as well as potassium levels, were the only factors found to have a significant impact on acid–base status (Table 2). Among the laboratory findings (Supplementary Table S1) from COVID-19 patients, only the potassium test was a significant factor, as shown in Table 2. The majority of patients have normal kidney function test (KFT) potassium testing ($$n = 879$$). The highest percent of patients with low potassium levels had RAlkNC ($26.7\%$) and RAlkC ($13.3\%$). Not surprisingly, the highest percentages of patients with normal potassium KFT also have normal ABG values. In descending order, the highest precents of patients with increased potassium levels had normal ABG levels ($21.7\%$), MAC ($20.6\%$), and metabolic acidosis with no compensation (MANC) ($19.6\%$). Then, we investigated the impact of acid–base status on illness severity. The severity of the patients is shown in Table 3 by their acid–base status. The severity of the condition has no significant bearing on the outcome (acid–base balance) ($$P \leq 0.3913$$). The majority of patients ($57\%$ of total) were classified as critical cases, while the minority ($1.1\%$ of total) were classified as mild. Critical cases were found in $62.9\%$ of RAlkNC, $61.1\%$ of mixed acidosis as well as RAlkC, $60.3\%$ of respiratory acidosis with compensation (RAC), $58.9\%$ of RANC, and $58.3\%$ of metabolic alkalosis with compensation (MAlkC) cases. **Table 3** | Count row (%) | Total | Severity | Severity.1 | Severity.2 | Severity.3 | P value | Invasive mechanical ventilators need | Invasive mechanical ventilators need.1 | P value.1 | Survival | Survival.1 | P value.2 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | | Mild | Moderate | Severe | Critical | | No | Yes | | Survived | Deceased | | | Normal | 335 (27.2) | 4 (1.2) | 40 (11.9) | 106 (31.6) | 185 (55.2) | 0.3913 | 259 (77.3) | 76 (22.7) | <0.0001 | 215 (64.2) | 120 (35.8) | <0.0001 | | Mixed acidosis | 18 (1.5) | 0 (0.0) | 3 (16.7) | 4 (22.2) | 11 (61.1) | | 3 (16.7) | 15 (83.3) | | 4 (22.2) | 14 (77.8) | | | Mixed alkalosis | 3 (0.24) | 0 (0.0) | 0 (0.0) | 3 (100) | 0 (0.0) | | 2 (66.7) | 1 (33.3) | | 1 (33.3) | 2 (66.7) | | | RAC | 73 (5.9) | 0 (0.0) | 9 (12.3) | 20 (27.4) | 44 (60.3) | | 52 (71.2) | 21 (28.8) | | 44 (60.3) | 29 (39.7) | | | RANC | 112 (9.1) | 3 (2.7) | 19 (17.0) | 24 (21.4) | 66 (58.9) | | 82 (73.2) | 30 (26.8) | | 59 (52.7) | 53 (47.3) | | | RAlkC | 172 (14.0) | 0 (0.0) | 16 (9.3) | 51 (29.7) | 105 (61.1) | | 117 (68.0) | 55 (32.0) | | 80 (46.5) | 92 (53.5) | | | RAlkNC | 213 (17.3) | 3 (1.4) | 20 (9.4) | 56 (26.3) | 134 (62.9) | | 164 (77.0) | 49 (23.0) | | 124 (58.2) | 89 (41.8) | | | MAC | 140 (11.4) | 2 (1.4) | 22 (15.7) | 42 (30.0) | 74 (52.9) | | 95 (67.9) | 45 (32.1) | | 56 (40.0) | 84 (60.0) | | | MANC | 81 (6.6) | 2 (2.5) | 12 (14.8) | 27 (33.3) | 40 (49.4) | | 55 (67.9) | 26 (32.1) | | 38 (46.9) | 43 (53.1) | | | MAlkC | 36 (2.9) | 0 (0.0) | 5 (13.9) | 10 (27.8) | 21 (58.3) | | 27 (75.0) | 9 (25.0) | | 19 (52.8) | 17 (47.2) | | | MAlkNC | 50 (4.1) | 0 (0.0) | 4 (8.0) | 18 (36.0) | 28 (56.0) | | 37 (74.0) | 13 (26.0) | | 33 (66.0) | 17 (34.0) | | | Total | 1233 | 14 (1.1) | 150 (12.2) | 361 (29.3) | 708 (57.4) | | 893 (72.4) | 340 (27.6) | | 673 (54.6) | 560 (45.4) | | We also investigated the requirement for an invasive mechanical ventilator (Table 3). In total, 340 patients ($27.6\%$) required invasive mechanical ventilation. The demand for an invasive mechanical ventilator is substantially influenced by acid–base status ($P \leq 000.1$). Among the acid–base categories, the normal group had the lowest percentage of those who needed mechanical ventilation. Mixed acidosis was the only category that had a larger proportion of patients who needed mechanical ventilation ($83.3\%$) than those who did not. Then, we explored the impact of the patients’ acid–base status on their survival (Table 3). About $54.6\%$ of all enrolled patients survived, while $45.4\%$ perished. The acid–base state has a substantial effect on survival ($P \leq 0.0001$). The metabolic alkalosis with no compensation (MAlkNC) ($66.0\%$) and normal ($64.2\%$) groups had the highest survival percentages among the acid–base classed groups. Among the acid–base classed groups, MAC ($60\%$), MANC ($53.1\%$), mixed acidosis ($77.8\%$), mixed alkalosis ($66.7\%$), RAlkC ($53.5\%$), and MANC ($53.1\%$) were more common in deceased patients than in survivors. Since demographics, comorbidities, and laboratory tests influenced acid-base status and could be a risk factor for mortality, the effect of acid–base status on the death could be attributable to the other investigated factors. To confirm the link between acid–base status and mortality, a regression model that included all analyzed covariates (demographics, comorbidities, and laboratory tests) along with acid–base status was utilized. As numbers show in Table 4, being an older patient (OR = 2.16, CI = 1.248–3.728), an active smoker (OR = 1.8, CI = 1.215–2.667), or having CKD (OR = 1.99, CI = 1.243–3.195) all raise the likelihood of death. In addition, having high levels of sodium, C-reactive protein, AST, or low levels of albumin also raise the risk of death (OR > 1, $P \leq 0.05$). **Table 4** | Level | Odds ratio | P-value | Lower 95% | Upper 95% | | --- | --- | --- | --- | --- | | Age (18–40 age group is a reference) | Age (18–40 age group is a reference) | Age (18–40 age group is a reference) | Age (18–40 age group is a reference) | Age (18–40 age group is a reference) | | 41–65 | 1.25 | 0.414 | 0.733 | 2.129 | | >65 | 2.16 | 0.006 | 1.248 | 3.728 | | Smoking status (Non-smoker is a reference) | Smoking status (Non-smoker is a reference) | Smoking status (Non-smoker is a reference) | Smoking status (Non-smoker is a reference) | Smoking status (Non-smoker is a reference) | | Active smoker | 1.80 | 0.003 | 1.215 | 2.667 | | Ex-smoker | 1.30 | 0.210 | 0.864 | 1.950 | | CKD (Yes vs. no) | 1.99 | 0.004 | 1.243 | 3.195 | | CRP (High vs. low) | 4.31 | 0.031 | 1.142 | 16.292 | | Sodium (High vs. Normal) | 7.75 | <0.0001 | 2.911 | 20.616 | | Albumin (Low vs. Normal) | 1.68 | 0.0003 | 1.267 | 2.2404 | | AST (High vs. Normal) | 2.17 | <0.0001 | 1.589 | 2.9660 | | Acid-base status (Normal is a reference) | | | | | | MAC | 2.12 | 0.002 | 1.332 | 3.359 | | MAlkC | 1.42 | 0.370 | 0.660 | 3.056 | | MAlkNC | 0.66 | 0.258 | 0.318 | 1.359 | | MANC | 1.42 | 0.222 | 0.809 | 2.492 | | Mixed acidosis | 3.61 | 0.049 | 1.004 | 12.975 | | Mixed alkalosis | 3.06 | 0.390 | 0.239 | 39.157 | | RAC | 1.00 | 0.999 | 0.560 | 1.786 | | RAlkC | 1.94 | 0.002 | 1.283 | 2.925 | | RAlkNC | 1.18 | 0.404 | 0.796 | 1.760 | | RANC | 1.76 | 0.022 | 1.087 | 2.862 | Even so, the acid–base status remains a substantial risk factor. When compared with those with normal levels, mixed acidosis increases the fatality risk almost four times (OR = 3.61, $$P \leq 0.049$$). Furthermore, MAC, RAlkC, or RANC doubles the mortality risk when compared with normal. Finally, a nominal logistic regression analysis was performed to determine factors associated with the occurrence of acid–base imbalance (Table 5). Variables like current smoking (OR: 1.85, $$P \leq 0.007$$), CKD (OR: 2.01, $$P \leq 0.012$$), high LDH (OR: 2.55, $$P \leq 0.0425$$), low potassium (OR: 4.27, $P \leq 0.0001$), and high AST (OR: 1.72, $$P \leq 0.0017$$) were associated with higher odds of acquiring an acid–base imbalance. **Table 5** | Variable | OR | P-value | Lower 95% | Upper 95% | | --- | --- | --- | --- | --- | | Current smoker | 1.85 | 0.007 | 1.182 | 2.881 | | CKD | 2.01 | 0.012 | 1.166 | 3.457 | | High LDH | 2.55 | 0.0425 | 1.032 | 6.318 | | Low potassium | 4.27 | <.0001 | 2.191 | 8.307 | | High AST | 1.72 | 0.0017 | 1.226 | 2.408 | ## Discussion The present study is the first comprehensive report of the range of acid–base changes as measured by ABG levels among hospitalized COVID-19 patients at a Jordanian hospital. COVID-19 is a highly contagious virus that causes pneumonia [25], and pneumonia is one of the leading causes of respiratory alkalosis [17]. As a result, we decided to investigate acid–base imbalance in COVID-19 patients. Surprisingly, the group that had the highest number of hospitalized patients was the normal group ($$n = 335$$, $27.2\%$). A separate study discovered that just $25\%$ of COVID-19 patients admitted to the ICU had normal acid–base balance, whereas $61\%$ had alkalosis. However, the categorization process used in our investigation and the previously cited study differed [26]. The most common acid–base abnormalities were RAlkNC ($17.3\%$), RAlkC ($14.0\%$), and MAC ($11.4\%$). Similarly, respiratory alkalosis was reported to be the most common acid–base disorder in patients hospitalized due to COVID-19 infection, accounting for $40\%$ of the study population [22]. This study found that the rate of respiratory alkalosis is approximately 2-fold higher than respiratory acidosis ($31.3\%$ vs. $15\%$ of the study populations). Similar findings were observed in two previous studies [21,22]. One study found that the occurrence of respiratory alkalosis in COVID-19 patients is due to hyperventilation induced by hypoxia (PO2 = 70.1 ± 32.9 mmHg). Moreover, patients with respiratory failure with hypercapnia (type 2 respiratory failure) were found to experience respiratory acidosis (PCO2 = 62.2 ± 13.4 mmHg) [21]. In the present study, hypertension, diabetes mellitus, dyslipidemia, asthma, COPD, CKD, and ESRD comorbidities, as well as potassium levels, were shown in the results to have a significant impact on acid-base status. In a study by Alfano et al., it was demonstrated that the presence of CKD and the levels of potassium have a significant effect on the acid-base status of hospitalized COVID-19 patients, similar to our findings [21]. In the present study, the majority of diabetic individuals had normal ABG readings, and metabolic acidosis with compensation coming in second. This can be understood since diabetes is a common cause of ketoacidosis, which is a kind of metabolic acidosis [27]. Asthma and COPD can both produce hypoventilation and, as a result, respiratory acidosis [28,29]. In line with this, the highest percent of asthmatic patients in our study exhibited uncompensated respiratory acidosis. The kidneys are responsible for maintaining the body’s acid–base balance by reabsorbing bicarbonate that has been filtered by the glomeruli and excreting titratable acids and ammonia through the urine. Acid retention and metabolic acidosis occur in CKD when renal function declines [30]. The highest percent of CKD patients recruited in the present study were additionally compensated for metabolic acidosis. Patients with ESRD undergoing hemodialysis are at risk of interdialytic acid buildup and chronic acidosis-related consequences, including mortality [31]. In line with this, the majority of ESRD patients who participated in this trial had MAC. Despite the fact that only 18 patients had mixed acidosis, 14 of those patients died. Furthermore, when compared with normal, mixed acidosis quadruples the fatality risk. The most frequently reported acid–base derangement in non-COVID-19 critically ill patients is severe mixed acidemia. This was observed in $6\%$ of critically ill patients and was associated with a $57\%$ ICU mortality rate [32]. Our results showed that $83.3\%$ of mixed acidosis patients needed mechanical ventilation during their hospital stay, and $77.8\%$ of them passed away. RAlkC is found in $14.0\%$ of patients and has been shown to be a key factor in mortality. When compared with normal patients, it doubled the death risk. In a Chinese study of 230 adult COVID-19 patients, 66 patients ($28.7\%$) showed respiratory alkalosis (pH > 7.45, PaCO2 < 35 mmHg) on admission, and those patients have an increased chance of developing severe cases [19]. In our study, $31\%$ exhibited respiratory alkalosis, with $14.0\%$ compensated and $17.3\%$ uncompensated instances. Although severity was not a statistically significant factor in our study, approximately $60\%$ of respiratory acidosis cases were critical. Another study found that 24 out of 32 ($75\%$) individuals with CO2 levels <35 mm Hg died [33]. However, this percent may include cases of pulmonary alkalosis and mixed alkalosis. When compared to patients with normal ABG readings, RANC and MAC doubled the risk of fatality. Lung-protective ventilation in COVID-19 respiratory failure may result in severe respiratory acidosis without considerable hypoxemia [34]. The higher mortality rate in patients with respiratory acidosis may be due to the binding of protein S to angiotensin-converting enzyme 2 receptors, which penetrate alveolar epithelial cells, resulting in direct toxic effects and an overactive immune response. This causes a systemic inflammatory response, resulting in a cytokine storm and lung tissue harm. Acute respiratory distress syndrome and metabolic acidosis can occur in severe cases [35]. Approximately $52.9\%$ of MAC cases in our study were critical cases, and $60\%$ of them deceased. The significant association between the presence of acid–base derangements and increased mortality among hospitalized patients with severe COVID-19 infection should alert treating clinicians to the importance of addressing these abnormalities. Understanding the underlying etiology for acid–base disorders and early intervention for metabolic and respiratory acidosis may help to improve the outcome of these patients. Moreover, early identification of major acid–base disorders like mixed respiratory and metabolic acidosis may help to identify patients with poor prognoses. The present study has a few limitations. It utilized inpatient record data from a single tertiary hospital in the north of Jordan, and data collection was retrospective. In addition, our study was limited to the acid–base status at the time of admission. Future prospective research on the dynamic acid–base status may be required. ## Conclusion Acid–base imbalance in hospitalized patients with COVID-19 is a risk factor for mortality, especially mixed respiratory and metabolic acidosis. Rigorous acid–base monitoring during COVID-19 hospitalization should be required in order to identify patients at increased risk of death. Early detection of such derangements may help to prevent future clinical derangements and hence improve survival. ## Data Availability Authors agree to make any materials, data, code, and associated protocols available upon request. ## Competing Interests The authors declare that there are no competing interests associated with the manuscript. ## Funding The authors declare that there are no sources of funding to be acknowledged. ## CRediT Author Contribution Nosayba Al-Azzam: Conceptualization, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review & editing. Basheer Khassawneh: Conceptualization, Validation, Investigation, Visualization, Methodology, Writing—review & editing. Sayer Al-Azzam: Conceptualization, Validation, Investigation, Visualization, Methodology, Writing—review & editing. Reema A. Karasneh: Conceptualization, Validation, Investigation, Visualization, Methodology, Writing—review & editing. Mamoon A. Aldeyab: Conceptualization, Supervision, Validation, Investigation, Methodology, Writing—review & editing. ## Ethics Approval This study was approved by Jordan University of Science and Technology’s Institutional Review Board (IRB) committee ($\frac{27}{137}$/2021). ## References 1. Abd El-Aziz T.M., Stockand J.D.. **Recent progress and challenges in drug development against COVID-19 coronavirus (SARS-CoV-2) - an update on the status**. *Infect. Genet. Evol.* (2020) **83** 104327. DOI: 10.1016/j.meegid.2020.104327 2. Islam K.U., Iqbal J.. **An update on molecular diagnostics for COVID-19**. *Front Cell Infect. Microbiol.* (2020) **10** 560616. DOI: 10.3389/fcimb.2020.560616 3. Aleem A., Akbar Samad A.B., Slenker A.K.. *Emerging Variants of SARS-CoV-2 And Novel Therapeutics Against Coronavirus (COVID-19)* (2022) 4. Crook H., Raza S., Nowell J., Young M., Edison P.. **Long covid-mechanisms, risk factors, and management**. *BMJ* (2021) **374** n1648. DOI: 10.1136/bmj.n1648 5. Al-Azzam N., Elsalem L., Gombedza F.. **A cross-sectional study to determine factors affecting dental and medical students' preference for virtual learning during the COVID-19 outbreak**. *Heliyon* (2020) **6** e05704. DOI: 10.1016/j.heliyon.2020.e05704 6. Long M.J.C., Aye Y.. **Science’s response to CoVID-19**. *Chem. Med. Chem.* (2021) **16** 2288-2314. DOI: 10.1002/cmdc.202100079 7. Parasher A.. **COVID-19: current understanding of its pathophysiology, clinical presentation and treatment**. *Postgrad. Med. J.* (2021) **97** 312-320. DOI: 10.1136/postgradmedj-2020-138577 8. Dharmarajan S., Bharathi M., Sivapuram K., Prakash B., Madhan S., Madhu A.. **Hearing loss-a camouflaged manifestation of COVID 19 infection**. *Indian J. Otolaryngol. Head Neck Surg.* (2021) **73** 494-498. DOI: 10.1007/s12070-021-02581-1 9. Tian S., Xiong Y., Liu H., Niu L., Guo J., Liao M.. **Pathological study of the 2019 novel coronavirus disease (COVID-19) through postmortem core biopsies**. *Mod. Pathol.* (2020) **33** 1007-1014. DOI: 10.1038/s41379-020-0536-x 10. Ye Q., Wang B., Mao J.. **The pathogenesis and treatment of the `Cytokine Storm' in COVID-19**. *J. Infect.* (2020) **80** 607-613. DOI: 10.1016/j.jinf.2020.03.037 11. Pandita A., Gillani F.S., Shi Y., Hardesty A., McCarthy M., Aridi J.. **Predictors of severity and mortality among patients hospitalized with COVID-19 in Rhode Island**. *PloS ONE* (2021) **16** e0252411. DOI: 10.1371/journal.pone.0252411 12. Grasselli G., Greco M., Zanella A., Albano G., Antonelli M., Bellani G.. **Risk factors associated with mortality among patients with COVID-19 in intensive care units in Lombardy, Italy**. *JAMA Intern Med.* (2020) **180** 1345-1355. DOI: 10.1001/jamainternmed.2020.3539 13. Karasneh R.A., Khassawneh B.Y., Al-Azzam S., Al-Mistarehi A.H., Lattyak W.J., Aldiab M.. **Risk factors associated with mortality in COVID-19 hospitalized patients: data from the Middle East**. *Int. J. Clin. Pract.* (2022) **2022** 9617319. DOI: 10.1155/2022/9617319 14. Mitchell J.H., Wildenthal K., Johnson R.L.. **The effects of acid-base disturbances on cardiovascular and pulmonary function**. *Kidney Int.* (1972) **1** 375-389. DOI: 10.1038/ki.1972.48 15. Raphael K.L., Zhang Y., Wei G., Greene T., Cheung A.K., Beddhu S.. **Serum bicarbonate and mortality in adults in NHANES III**. *Nephrol. Dial. Transplant.* (2013) **28** 1207-1213. DOI: 10.1093/ndt/gfs609 16. Tinawi M.. **Respiratory acid-base disorders: respiratory acidosis and respiratory alkalosis**. *Arch. Clin. Biomed. Res.* (2021) **5** 158-168. DOI: 10.26502/acbr.50170157 17. Ayers P., Dixon C., Mays A.. **Acid-base disorders: learning the basics**. *Nutr. Clin. Pract.* (2015) **30** 14-20. DOI: 10.1177/0884533614562842 18. Berend K., de Vries A.P., Gans R.O.. **Physiological approach to assessment of acid-base disturbances**. *N. Engl. J. Med.* (2014) **371** 1434-1445. DOI: 10.1056/NEJMra1003327 19. Wu C., Wang G., Zhang Q., Yu B., Lv J., Zhang S.. **Association between respiratory alkalosis and the prognosis of COVID-19 patients**. *Front. Med. (Lausanne)* (2021) **8** 564635. DOI: 10.3389/fmed.2021.564635 20. Ayers P., Warrington L.. **Diagnosis and treatment of simple acid-base disorders**. *Nutr. Clin. Pract.* (2008) **23** 122-127. DOI: 10.1177/0884533608314534 21. Alfano G., Fontana F., Mori G., Giaroni F., Ferrari A., Giovanella S.. **Acid base disorders in patients with COVID-19**. *Int. Urol. Nephrol.* (2022) **54** 405-410. DOI: 10.1007/s11255-021-02855-1 22. Chiumello D., Pozzi T., Fratti I., Modafferi L., Montante M., Papa G.F.S.. **Acid-base disorders in COVID-19 patients with acute respiratory distress syndrome**. *J. Clin. Med.* (2022) **11** 2093. DOI: 10.3390/jcm11082093 23. **Nutrition—body mass index**. (2022) 24. **Clinical spectrum of SARS-CoV-2 infection**. (2022) 25. Rahimi B., Vesal A., Edalatifard M.. **Coronavirus and its effect on the respiratory system: is there any association between pneumonia and immune cells**. *J. Family Med. Prim. Care* (2020) **9** 4729-4735. DOI: 10.4103/jfmpc.jfmpc_763_20 26. Zemlin A.E., Sigwadhi L.N., Wiese O.J., Jalavu T.P., Chapanduka Z.C., Allwood B.W.. **The association between acid-base status and clinical outcome in critically ill COVID-19 patients admitted to intensive care unit with an emphasis on high anion gap metabolic acidosis**. *Ann. Clin. Biochem.* (2022) 45632221134687. DOI: 10.1177/00045632221134687 27. Yoshimi T.. **Acid-base balance disorder in various diseases–diabetes mellitus**. *Nihon Rinsho Japanese J. Clin. Med.* (1992) **50** 2206-2212. PMID: 1434013 28. Epstein S.K., Singh N.. **Respiratory acidosis**. *Respir. Care* (2001) **46** 366-383. PMID: 11262556 29. Lamba T.S., Sharara R.S., Singh A.C., Balaan M.. **Pathophysiology and classification of respiratory failure**. *Crit. Care Nurs. Q.* (2016) **39** 85-93. DOI: 10.1097/CNQ.0000000000000102 30. Nagami G.T., Hamm L.L.. **Regulation of acid-base balance in chronic kidney disease**. *Adv. Chronic Kidney Dis.* (2017) **24** 274-279. DOI: 10.1053/j.ackd.2017.07.004 31. Qian Q.. **Acid-base alterations in ESRD and effects of hemodialysis**. *Semin. Dial.* (2018) **31** 226-235. DOI: 10.1111/sdi.12659 32. Jung B., Rimmele T., Le Goff C., Chanques G., Corne P., Jonquet O.. **Severe metabolic or mixed acidemia on intensive care unit admission: incidence, prognosis and administration of buffer therapy. A prospective, multiple-center study**. *Crit. Care* (2011) **15** R238. DOI: 10.1186/cc10487 33. Chen T., Wu D., Chen H., Yan W., Yang D., Chen G.. **Clinical characteristics of 113 deceased patients with coronavirus disease 2019: retrospective study**. *BMJ* (2020) **368** m1091. DOI: 10.1136/bmj.m1091 34. Akkanti B., Jagpal S., Darwish R., Saavedra Romero R., Scott L.K., Dinh K.. **Physiologic improvement in respiratory acidosis using extracorporeal Co2 removal with hemolung respiratory assist system in the management of severe respiratory failure from coronavirus disease 2019**. *Crit. Care Explor.* (2021) **3** e0372. DOI: 10.1097/CCE.0000000000000372 35. Xie P., Ma W., Tang H., Liu D.. **Severe COVID-19: A review of recent progress with a look toward the future**. *Front. Public Health* (2020) **8** 189. DOI: 10.3389/fpubh.2020.00189
--- title: 'Residual Lung Abnormalities after COVID-19 Hospitalization: Interim Analysis of the UKILD Post–COVID-19 Study' authors: - Iain Stewart - Joseph Jacob - Peter M. George - Philip L. Molyneaux - Joanna C. Porter - Richard J. Allen - Shahab Aslani - J. Kenneth Baillie - Shaney L. Barratt - Paul Beirne - Stephen M. Bianchi - John F. Blaikley - James D. Chalmers - Rachel C. Chambers - Nazia Chadhuri - Christopher Coleman - Guilhem Collier - Emma K. Denneny - Annemarie Docherty - Omer Elneima - Rachael A. Evans - Laura Fabbri - Michael A. Gibbons - Fergus V. Gleeson - Bibek Gooptu - Neil J. Greening - Beatriz Guillen Guio - Ian P. Hall - Neil A. Hanley - Victoria Harris - Ewen M. Harrison - Melissa Heightman - Toby E. Hillman - Alex Horsley - Linzy Houchen-Wolloff - Ian Jarrold - Simon R. Johnson - Mark G. Jones - Fasihul Khan - Rod Lawson - Olivia Leavy - Nazir Lone - Michael Marks - Hamish McAuley - Puja Mehta - Dhruv Parekh - Karen Piper Hanley - Manuela Platé - John Pearl - Krisnah Poinasamy - Jennifer K. Quint - Betty Raman - Matthew Richardson - Pilar Rivera-Ortega - Laura Saunders - Ruth Saunders - Malcolm G. Semple - Marco Sereno - Aarti Shikotra - A. John Simpson - Amisha Singapuri - David J. F. Smith - Mark Spears - Lisa G. Spencer - Stefan Stanel - David R. Thickett - A. A. Roger Thompson - Mathew Thorpe - Simon L. F. Walsh - Samantha Walker - Nicholas David Weatherley - Mark E. Weeks - Jim M. Wild - Dan G. Wootton - Chris E. Brightling - Ling-Pei Ho - Louise V. Wain - Gisli R. Jenkins journal: American Journal of Respiratory and Critical Care Medicine year: 2022 pmcid: PMC10037479 doi: 10.1164/rccm.202203-0564OC license: CC BY 4.0 --- # Residual Lung Abnormalities after COVID-19 Hospitalization: Interim Analysis of the UKILD Post–COVID-19 Study ## Body Long-term symptoms of coronavirus disease (COVID-19) have been widely reported and can have a severe impact on quality of life, frequently characterized by chronic breathlessness (1–3). Postmortem studies on patients with COVID-19 have highlighted diffuse parenchymal alterations, including alveolar damage, exudation, and the development of pulmonary fibrosis, which may explain chronic respiratory symptoms in survivors (4–6). A number of studies have identified similarities between severe COVID-19 and idiopathic pulmonary fibrosis, an archetypal interstitial lung disease (ILD). These include shared genetic etiology [7, 8], circulating biomarkers [9, 10], similarities in pulmonary function, and radiological features [11]. Viral injury may promote lung fibrosis, and chronic viral infection has been shown to be associated with developing idiopathic pulmonary fibrosis [12]. Consequently, survivors of COVID-19 may develop parenchymal abnormalities consistent with ILD, including radiological patterns of ground-glass opacities and reticulations. To understand the potential risk of COVID-19 leading to the development of longer-term ILD and fibrosis, the UK Interstitial Lung Disease Consortium (UKILD) post–COVID-19 study aims to investigate the risk factors and nature of long-term lung damage from COVID-19 in a longitudinal observational study. To support clinical and research management, this planned interim analysis of the UKILD post–COVID-19 study addresses the extent of residual lung abnormalities after hospitalization after completion of an early follow-up visit of the prospective PHOSP–Covid-19 (Post-Hospitalization COVID-19) study [13]. Some of the results of these studies have been previously reported in the form of a preprint (medRxiv, 16 March 2022; https://www.medrxiv.org/content/$\frac{10.1101}{2022.03.10.22272081}$v2). ## Abstract ### Rationale Shared symptoms and genetic architecture between coronavirus disease (COVID-19) and lung fibrosis suggest severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection may lead to progressive lung damage. ### Objectives The UK Interstitial Lung Disease Consortium (UKILD) post–COVID-19 study interim analysis was planned to estimate the prevalence of residual lung abnormalities in people hospitalized with COVID-19 on the basis of risk strata. ### Methods The PHOSP–COVID-19 (Post-Hospitalization COVID-19) study was used to capture routine and research follow-up within 240 days from discharge. Thoracic computed tomography linked by PHOSP–COVID-19 identifiers was scored for the percentage of residual lung abnormalities (ground-glass opacities and reticulations). Risk factors in linked computed tomography were estimated with Bayesian binomial regression, and risk strata were generated. Numbers within strata were used to estimate posthospitalization prevalence using Bayesian binomial distributions. Sensitivity analysis was restricted to participants with protocol-driven research follow-up. ### Measurements and Main Results The interim cohort comprised 3,700 people. Of 209 subjects with linked computed tomography (median, 119 d; interquartile range, 83–155), 166 people ($79.4\%$) had more than $10\%$ involvement of residual lung abnormalities. Risk factors included abnormal chest X-ray (risk ratio [RR], 1.21; $95\%$ credible interval [CrI], 1.05–1.40), percent predicted DlCO less than $80\%$ (RR, 1.25; $95\%$ CrI, 1.00–1.56), and severe admission requiring ventilation support (RR, 1.27; $95\%$ CrI, 1.07–1.55). In the remaining 3,491 people, moderate to very high risk of residual lung abnormalities was classified at $7.8\%$, and posthospitalization prevalence was estimated at $8.5\%$ ($95\%$ CrI, 7.6–9.5), rising to $11.7\%$ ($95\%$ CrI, 10.3–13.1) in the sensitivity analysis. ### Conclusions Residual lung abnormalities were estimated in up to $11\%$ of people discharged after COVID-19–related hospitalization. Health services should monitor at-risk individuals to elucidate long-term functional implications. ## Scientific Knowledge on the Subject Current studies highlight persistent breathlessness and radiological patterns suggestive of lung fibrosis, as well as shared genetic architecture with idiopathic pulmonary fibrosis, in people discharged after severe coronavirus disease (COVID-19) hospitalization. Survivors of COVID-19 may develop parenchymal abnormalities consistent with lung fibrosis. ## What This Study Adds to the Field This study assesses the risk factors for residual lung abnormalities, provides evidence of persistent abnormalities within 1 year of discharge from over 200 computed tomography scans, and estimates the prevalence of lung abnormalities after discharge to be up to $11\%$ in a broad range of COVID-19 severity. The findings emphasize the importance for health services to undertake active radiological and physiological monitoring to assess progression or resolution over time. ## Participants This interim analysis was restricted to participants of the PHOSP–COVID-19 study, a prospective longitudinal cohort study of adults discharged from National Health Service hospitals across the United Kingdom after admission for confirmed or clinically diagnosed COVID-19, previously described in detail [14]. Individuals withdrawing consent from PHOSP–COVID-19 were excluded. Individuals being managed for an a priori diagnosed ILD or pulmonary fibrosis, as recorded by site teams using hospital notes, were identified by hand searches of comorbidities and subsequently excluded. ## Interim Study Design Interim participants were discharged by the end of March 2021, representing wave one of the pandemic; interim data were collected up to October 2021 and were restricted to within 240 days of discharge. Analyses were performed with data recorded through routine follow-up (PHOSP–COVID-19 Tier 1) and those with completed early research follow-up visits (PHOSP–COVID-19 Tier 2). Clinically indicated thoracic computed tomography (CT) scans were identified through the PHOSP–COVID-19 study via linkage to a radiological database; linked CT scans were requested at clinical discretion. The presence of residual lung abnormalities on volumetric CTs was scored on a lobar basis; the percentage involvement of ground-glass opacities, reticulations, or the sum of involvement was averaged across lobes to quantify residual abnormality [15]. The primary outcome was visually scored residual abnormalities greater than $10\%$ lung involvement on CT [15]. Risk factors implicated in worse outcomes after COVID-19 hospitalization of individuals with ILD were described [16]. These included sex, age, ethnicity, body mass index, and IMD (Index of Multiple Deprivation). A modified WHO (World Health Organization) clinical progression scale was used to define the severity of admission: 1) no supplemental oxygen; 2) supplemental oxygen only; 3) continuous positive airway pressure (CPAP); and 4) invasive mechanical ventilation (IMV) or extracorporeal membrane oxygenation (ECMO). Abnormal chest X-ray (CXR) reports were classified at follow-up, defined as “suggestive of lung fibrosis,” “extensive, persistent changes greater than one-third of lung involvement,” and “indeterminate,” compared with “other” or “normal.” Breathless and cough symptoms were recorded at follow-up with the patient symptom questionnaire developed for the PHOSP–COVID-19 study [14]. Percent predicted values for FVC (ppFVC) and DlCO (ppDlCO) were obtained at follow-up visits and calculated using global lung initiative reference equations. ## Statistical Analysis Risk factor data were presented descriptively overall, according to PHOSP–COVID-19 tier, and within the sample of linked and scored CTs. Chi-square tests were performed on nonmissing categories. Residual abnormalities on paired CT scans were tested with paired t test; changes in scored residual lung abnormalities over time were estimated using linear mixed effect models, with random effects of timing at the level of the individual, adjusted for sex and IMD. A random sample of 70 CT scans was tested for interrater agreement by Cohen’s kappa (κ), with a second radiologist blinded to scores. Univariate relative risk ratios of threshold greater than $10\%$ residual abnormalities and difference in involvement on CT were modeled with dichotomized exposure variables. Bayesian binomial and linear associations were estimated using 12,500 Markov Chain Monte Carlo (MCMC) iterations, including a burn-in of 2,500 and 10,000 subsequent simulations using random-walk Metropolis-Hastings sampling. Noninformative, flat priors were selected, and estimates were reported with a $95\%$ CrI. Linear associations were additionally adjusted for demographics of sex and IMD. Clinical risk factors with consistent significant effects were selected to develop risk strata of suspected residual lung abnormalities after COVID-19 hospitalization. For the indexing of risk strata, missing data on clinical indicators were imputed to the reference (lowest risk) category. The percentage of participants within moderate- to very-high–risk strata and no CT scored were defined as at-risk. Hospital admissions were compared between at-risk groups using chi-square, and 15 index admission variables were selected from 61 by least absolute shrinkage and selection operator. Bayesian inference with a binomial distribution of at-risk cases and noncases [17] was used to estimate the prevalence of suspected residual lung abnormalities after COVID-19 hospitalization within 240 days of discharge, reported with a $95\%$ CrI. MCMC simulations were run as described above. Noninformative, uniform β priors were used and compared in sensitivity analyses with uniform Jeffrey’s priors, as well as skeptical and power priors informed by published population studies of ILD [18, 19]. Sensitivity analyses were performed in PHOSP–COVID-19 Tier 2 research follow-up participants in which data completeness was greater. Analyses were performed in Stata SE16.0 within the Scottish National Safe Haven Trusted Research Environment. ## Cohort Demographics and Patterns of Lung Damage A total of 3,700 PHOSP–COVID-19 participants reached the criteria for inclusion in the interim UKILD post–COVID-19 study cohort. This included 1,304 patients with data available through routine clinical care (Tier 1) and 2,396 who had completed an early follow-up research visit within 240 days of discharge (Tier 2) (Figure 1). We observed that 255 of 3,700 ($6.9\%$) participants in the interim cohort had a linkable thoracic CT scan performed, 220 of 2,396 ($9.2\%$) Tier 2 participants had CT scans performed, and 35 of 1,304 ($2.7\%$) of Tier 1 participants had CT scans performed ($P \leq 0.001$). Of 255 participants with linked CT scans within 240 days of discharge (median, 113 days; interquartile range [IQR], 69–166) (Figure E1 in the online supplement), a total of 209 ($82.0\%$) were visually scored with interrater agreement on $70\%$ of scans (Cohen’s κ, 0.33). Participants with a CT scored were majority male ($68.4\%$), White ($68.9\%$), had a median age of 58 (52–67), and had a median time to early follow-up visit of 140 days (IQR, 106–170) (Table 1). **Figure 1.:** *CONSORT flow diagram of UKILD post–COVID-19 study interim cohort definition. White boxes derived from the PHOSP-COVID database. Blue boxes represent computed tomography samples linked with Post-Hospitalization COVID (PHOSP-COVID) identifiers in a radiological database. CONSORT = Consolidated Standards of Reporting Trials; CT = computed tomography; ILD = interstitial lung disease; UKILD = UK interstitial lung disease consortium post–COVID-19 study.* TABLE_PLACEHOLDER:Table 1. Residual lung abnormalities greater than $10\%$ were observed in 166 of 209 ($79.4\%$) participants. Visual scoring of involvement indicated ground-glass opacities affected a mean of 25.5 ± $5.9\%$ of the lung, reticulation a mean of 15.1 ± $11.0\%$, with residual abnormalities involved a mean of 40.6 ± $20.8\%$ of the lung (Figure 2A). A total of 33 people had a repeat CT visually scored after a minimum of 90 days (median, 161 d; IQR, 109–187), 28 of 33 ($84.8\%$) of whom were classified with residual abnormalities greater than $10\%$ on the initial scan, with 26 of 28 ($92.9\%$) observed to have greater than $10\%$ involvement in subsequent scans. In paired analysis, the overall change in residual lung abnormalities was −$3.62\%$ ($95\%$ confidence interval [CI], −6.10 to −1.13; $$P \leq 0.006$$) (Figure 2B). The involvement of lung reticulations and ground-glass opacities did not significantly change with a mean difference of −$2.08\%$ ($95\%$ CI, −4.66 to 0.51; $$P \leq 0.112$$) and −$1.54\%$ ($95\%$ CI, −4.74 to 1.39; $$P \leq 0.293$$), respectively (Figures 2C and 2D). Using all scored CT scans, the mean weekly change in lung involvement was estimated at −$0.13\%$ per week ($95\%$ CI, −0.20 to −0.05) for reticulations and −$0.13\%$ per week ($95\%$ CI, −0.22 to −0.04) for ground-glass opacities (Figure 2E). The weekly change in residual lung abnormalities was −$0.20\%$ per week ($95\%$ CI, −0.28 to −0.11) (Figure 2F). Representative CT images of residual lung abnormality demonstrated persistent involvement more than 100 days after discharge (Figure 3). **Figure 2.:** *Extent of residual lung abnormalities on linked computed tomography. (A) Mean percentage lung involvement of reticulations, ground-glass opacities, and residual abnormalities within 240 days of discharge with visually scored involvement greater than 10%, presented with standard deviation (n = 166). Percentage lung involvement of (B) residual abnormalities, (C) reticulations, and (D) ground-glass opacities at initial and repeat computed tomography scans with greater than 90 days between (n = 33), with P values from paired t test. (E) Estimated percent lung involvement of ground-glass opacities (top, blue) and reticulations (bottom, red) from linear mixed effects by weeks after discharge (n = 209, scans = 242). (F) Estimated percent lung involvement of residual abnormalities from linear mixed effects by weeks after discharge, presented with mean weekly effect and 95% confidence intervals (n = 209, scans = 242). CT = computed tomography.* **Figure 3.:** *Representative computed tomography (CT) images of residual lung abnormalities. Representative (A) coronal and (B) axial noncontrast CT imaging from the same individual performed 137 days after discharge after a coronavirus disease (COVID-19) admission scored with 52.5% total lung involvement of residual lung abnormality, of which 18.3% was reticulation and 34.2% ground-glass opacity. Peripheral reticulation (arrows) is evident, surrounded by faint areas of ground-glass opacity. Representative coronal CT images from the same individual at (C) 114 days after discharge scored 56.8% lung involvement (28.5% reticulation and 28.3% ground-glass opacity), and (D) 239 days after discharge scored 49.2% total lung involvement (20.0% reticulation and 29.2% ground-glass opacity). Peripheral areas of reticulation (black arrow) and ground-glass opacity (white arrow) in the right lung.* Overall, the median time to follow-up in the UKILD interim cohort ($$n = 3$$,700) was 127 days (IQR, 91–173), the median age was 59 (IQR, 50–68), and the cohort was majority male ($60.7\%$). Tier 1 participants ($$n = 1$$,304) had a median time to follow-up of 101 days (IQR, 82–138), a median age of 60 (IQR, 51–70), and the majority were male ($58.9\%$); demographics were similar in Tier 2 participants ($$n = 2$$,396) with a median time to research visit of 141 days (IQR, 100–180), a median age of 59 (IQR, 50–67), and the majority male ($61.7\%$) (Table 1). There was minimal evidence of systematic bias in the characteristics between Tier 2 and Tier 1 participants in nonmissing data (Table 1), although the representation of people aged below 60 was greater in Tier 2 participants ($52.5\%$ vs. $48.8\%$; $$P \leq 0.027$$); similarly, there were small differences in the representation of ethnicity ($P \leq 0.001$), greater representation of the lowest deprivation quintile ($19.1\%$ vs. $16.1\%$; $$P \leq 0.031$$), as well as lower representation of normal CXR ($32.5\%$ vs. $39.2\%$; $$P \leq 0.004$$). Tier 2 participants had a median ppFVC at research follow-up of $90.2\%$ (IQR, 78.6–101.6) with missing records at $55.5\%$, whereas median ppDlCO was $87.5\%$ (IQR, 74.0–101.3) with missing records at $78.8\%$; lung function was largely missing in the routine follow-up of Tier 1 participants. We observed $34.6\%$ of people reported worsening cough or dyspnea since discharge in Tier 2. ILD diagnostic criteria of lung function (ppFVC and ppDlCO), CXR, and symptoms were frequently missing, particularly in Tier 1 of clinical follow-up (Figure E2). In Tier 1, 578 of 1,304 ($44.3\%$) participants were missing data on all four characteristics at interim analysis, whereas in Tier 2, 362 of 2,396 ($15.1\%$) participants were missing data on all four characteristics. In contrast, a total of 202 ($8.4\%$) Tier 2 participants had complete data on all, and no Tier 1 participants had complete lung function, CXR, or symptom data. In the subsample of participants with scored CTs, data was missing at a rate similar to Tier 2 for lung function (ppDlCO, $70.3\%$ and ppFVC, $60.8\%$), CXR ($47.4\%$), and Patient Symptom Questionnaire ($43.1\%$) (Table 1). ## Risk of Residual Lung Abnormalities and Persistence Over Time Univariate risk ratios were calculated to assess the risk of residual lung abnormalities greater than $10\%$ on CT. A greater risk was observed in males (risk ratio [RR], 1.42; $95\%$ CrI, 1.17–1.77) and in those over 60 years of age (RR, 1.22; $95\%$ CrI, 1.06–1.40). Clinical indicators, including severe illness on admission requiring CPAP, IMV, or ECMO (RR, 1.40; $95\%$ CrI, 1.23–1.63), abnormal CXR findings (RR, 1.40; $95\%$ CrI, 1.22–1.61), and ppDlCO less than $80\%$ (RR, 1.26; $95\%$ CrI, 1.02–1.58) were also associated with greater risk, with consistent effects for the relative mean difference of percent involvement after adjustment for sex and deprivation quintile (Table 2). **Table 2.** | Characteristic | Risk Factor Present, % | Risk Factor Absent, % | Univariate Risk Ratio | 95% Credible Interval | Estimated Mean Difference, % | 95% Credible Interval.1 | Adjusted Mean Difference, % | 95% Credible Interval.2 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Male | 87.4 | 62.1 | 1.42 | (1.17 to 1.77) | 12.46 | (5.76 to 19.59) | 11.26 | (4.24 to 18.04) | | Age ⩾60, yr | 87.9 | 71.8 | 1.22 | (1.06 to 1.40) | 8.29 | (2.11 to 14.44) | 8.57 | (3.61 to 6.16) | | Non-White | 78.5 | 79.9 | 0.97 | (0.84 to 1.12) | 3.48 | (−3.78 to 10.88) | 3.84 | (−4.95 to 9.37) | | IMD (Q1 and 2) | 87.2 | 74.4 | 1.17 | (1.02 to 1.34) | 6.91 | (0.38 to 13.33) | 6.28 | (−0.31 to 12.91) | | BMI >30 | 87.3 | 71.6 | 1.22 | (1.04 to 1.45) | 3.93 | (−3.70 to 11.52) | 4.54 | (−2.40 to 11.65) | | CPAP/IMV | 93.8 | 67.0 | 1.4 | (1.23 to 1.63) | 20.56 | (14.80 to 26.36) | 20.14 | (14.34 to 25.69) | | aCXR | 100.0 | 73.0 | 1.4 | (1.22 to 1.61) | 14.96 | (3.89 to 25.78) | 11.54 | (0.53 to 21.59) | | ppFVC <80 | 86.2 | 79.3 | 1.07 | (0.85 to 1.31) | 10.4 | (−0.90 to 22.00) | 11.99 | (−0.14 to 23.52) | | ppDlCO <80 | 96.0 | 75.7 | 1.26 | (1.02 to 1.58) | 19.04 | (7.65 to 30.71) | 15.31 | (2.84 to 28.06) | | PSQ worse | 78.4 | 80.0 | 0.99 | (0.81 to 1.21) | 4.49 | (−4.58 to 13.54) | 4.71 | (−4.31 to 13.87) | Three significant clinical indicators were selected to index the risk of residual lung abnormalities after COVID-19 in the remaining cohort ($$n = 3$$,491) on the basis of combined thresholds: ppDlCO less than $80\%$; abnormal CXR; and severe illness on admission. Individuals were considered to be at very high risk when reaching the defined thresholds in all three indicators (risk index four), high risk when two thresholds were reached (risk index three), or moderate risk if reaching ppDlCO or CXR thresholds alone (risk index two). Individuals reaching the threshold of the severity of illness on admission alone were considered low risk in the absence of other indicators (risk index one). Those who did not reach any threshold were considered very low risk (risk index zero). A total of 14 of 3,419 ($0.4\%$) participants were considered very high risk, 143 of 3,419 ($4.1\%$) high risk, 116 of 3,419 ($3.3\%$) moderate risk, 1,256 of 3,419 ($36.0\%$) low risk, and 1,962 of 3,419 ($56.2\%$) very low risk (Table 3). Combined, 273 of 3,419 ($7.8\%$) participants in strata of moderate to very high risk were defined as at-risk, and 8 of 46 ($17.4\%$) participants with an unscored clinically indicated CT were at-risk. In sensitivity analyses applying risk stratification to Tier 2 alone, 231 of 2,219 ($10.4\%$) participants were at moderate to very high risk, including $20\%$ of those with an unscored clinically indicated CT (Table 3). **Table 3.** | Interim Cohort | Interim Cohort.1 | Interim Cohort.2 | Interim Cohort.3 | Interim Cohort.4 | | --- | --- | --- | --- | --- | | Strata | Unscored (n = 3,491), n | % | Sensitivity (n = 2,219), n | % | | Very high | 14 | 0.4 | 14 | 0.6 | | High | 143 | 4.1 | 123 | 5.5 | | Moderate | 116 | 3.3 | 94 | 4.2 | | Low | 1256 | 36.0 | 767 | 34.6 | | Very low | 1962 | 56.2 | 1221 | 55.0 | No differences were observed between at-risk participants ($$n = 273$$) and participants with greater than $10\%$ residual abnormalities on CT ($$n = 166$$) according to a representation of males, older age, ethnicity, deprivation, body mass index, severity of admission, ppFVC less than $80\%$, or Patient Symptom Questionnaire (Table E1). There was a lower representation of normal CXR in the at-risk group ($14.7\%$ vs. $30.1\%$; $P \leq 0.001$) and more representation of ppDlCO less than $80\%$ ($55.3\%$ vs. $14.5\%$; $P \leq 0.001$). The percentage of people who did not have a severe admission requiring CPAP, ECMO, or IMV was similar in both groups ($44.3\%$ vs. $45.2\%$), whereas CXR was missing in $26.0\%$ of the at-risk group and $48.2\%$ of people with residual abnormalities scored. Comparing at-risk participants to low-risk participants, there were more records of immunosuppressant ($18.3\%$ vs. $9.9\%$; $$P \leq 0.001$$) and corticosteroid treatment ($35.3\%$ vs. $26.5\%$; $$P \leq 0.019$$) preadmission, intensive care unit stays ($50.0\%$ vs. $33.4\%$; $P \leq 0.001$), and complications of acute respiratory distress syndrome (ARDS; $25.0\%$ vs. $13.7\%$; $P \leq 0.001$) (Table E2). In addition, there were more recorded unscheduled emergency visits after discharge ($34.8\%$ vs. $25.2\%$; $$P \leq 0.001$$), with a greater representation of visits in which patients presented with symptoms of shortness of breath ($33.7\%$ vs. $24.3\%$; $$P \leq 0.046$$). Findings were similar in comparisons of CT-scored residual lung abnormalities greater than $10\%$ compared with those not reaching this threshold, although statistical significance was not always met (Table E2). On the basis of the distribution of at-risk cases, the prevalence of residual lung abnormalities after COVID-19 hospitalization was estimated at $8.51\%$ ($95\%$ CrI, 7.56–9.51) using noninformative priors, or $6.49\%$ ($95\%$ CrI, 5.75–7.27) with skeptical priors on the basis of ILD population prevalence estimated at 1 in 1,000 (Table 4 and Figure E3) [18, 19]. In sensitivity analyses on the basis of Tier 2 distribution, the prevalence of residual lung abnormalities after COVID-19 hospitalization was estimated at $11.67\%$ ($95\%$ CrI, 10.28–13.14) using noninformative priors, or $7.74\%$ ($95\%$ CrI, 6.79–8.72) using skeptical priors. **Table 4.** | Model | Prevalence, % | 95% CrI | Prior | a | b | DIC | | --- | --- | --- | --- | --- | --- | --- | | 1 | 8.51 | (7.56–9.51) | Uniform | 1.0 | 1.0 | 9.38 | | 1-i | 8.48 | (7.52–9.49) | Jeffreys | 0.5 | 0.5 | 9.45 | | 1-ii | 6.49 | (5.75–7.27) | Skeptical | 1.0 | 1000.0 | 28.67 | | 1-iii | 7.37 | (6.53–8.24) | Power | 1.0 | 1000.0 | 14.99 | | 2 | 11.67 | (10.28–13.14) | Uniform | 1.0 | 1.0 | 9.2 | | 2-i | 11.61 | (10.19–13.04) | Jeffreys | 0.5 | 0.5 | 9.27 | | 2-ii | 7.74 | (6.79–8.72) | Skeptical | 1.0 | 1000.0 | 45.97 | | 2-iii | 9.32 | (8.17–10.54) | Power | 1.0 | 1000.0 | 20.91 | ## Discussion These data demonstrate that residual lung abnormalities were visually identifiable on clinically indicated thoracic follow-up CT imaging in a substantial proportion of patients within 8 months of discharge after COVID-19 hospitalization. The involvement of scored residual lung abnormalities minimally declined per week after discharge, and minimal resolution was observed in paired subsequent scans at least 90 days apart. Key clinical risk factors associated with residual abnormalities in the early follow-up period included abnormal CXR, ppDlCO less than $80\%$, and severe admissions requiring invasive support (IMV, CPAP, or ECMO). In those without a scored CT, $0.4\%$ were in very-high–risk strata (all three indicators present), $4.1\%$ in high-risk strata (any two indicators present), and $3.3\%$ in moderate-risk strata (presence of either ppDlCO less than $80\%$ or abnormal CXR, alone). Combining these risk strata, $7.8\%$ of the interim cohort had suspected residual lung abnormalities after COVID-19 hospitalization, which increased to $10.4\%$ in sensitivity analysis on those with planned research follow-up. On the basis of Bayesian modeling, we estimate the prevalence of suspected residual lung abnormalities with greater than $10\%$ lung involvement to be up to $11.7\%$ in people hospitalized with acute COVID-19 infections before March 2021. This UKILD Post–COVID-19 Study interim analysis of residual abnormalities in patients hospitalized for COVID-19 offers the largest assessment of prevalence in hospitalized individuals to date and is consistent with findings from a number of smaller studies that demonstrate persistent radiological patterns and impaired gas transfer during extended follow-up of patients with COVID-19 (20–23). At the time of this interim analysis, it is not possible to determine whether the observed residual lung abnormalities represent early ILD with potential for progression or whether they reflect residual pneumonitis that may be stable or resolve over time [24]. The $10\%$ threshold used was determined to support the distinction of interstitial lung damage from interstitial lung abnormalities [15]. Longer-term follow-up and mechanistic studies will be required to determine the clinical trajectory of these observations. When linked longitudinal scans were available, most patients did not show evidence of substantial improvement, although such clinically requested CTs may be overrepresented by those with slower recovery. However, approximately half the people with visually scored residual abnormalities above the $10\%$ threshold did not require CPAP, IMV, or ECMO during their admission and less than one quarter had ARDS recorded as a complication, suggesting medium- and longer-term disability consequent to severe COVID-19 infection, consistent with prior studies [18]. The risk factors for a residual abnormality scored in the CT subsample (abnormal CXR, ppDlCO less than $80\%$, and severe admissions requiring invasive support) were applied to the remaining hospitalized cohort to generate clinically applicable risk strata. For participants in receipt of a clinically indicated but unscored CT, $17.4\%$ of people were in moderate- to very-high–risk strata for residual lung abnormalities (sensitivity, $20.0\%$). These rates were similar to meta-analysis estimates of the percentage of clinically indicated CT scans with radiological patterns suggestive of fibrosis ($29\%$; $95\%$ CI, 22–$37\%$) and people with impaired gas transfer ($17\%$; $95\%$ CI, 13–23), neither of which were associated with the timing of follow-up within the first year after COVID-19 [25]. In paired CT scans greater than 90 days apart, we demonstrate no significant difference in the mean change for percent involvement of reticulations and ground-glass opacities, whereas the scored involvement of reticulations and ground-glass opacities on the basis of all CT scans declined by $0.13\%$ per week of study from discharge, suggesting persistence over time in at-risk groups. Differences between individuals at moderate to very high risk and those at lower risk suggested more immunosuppressant and corticosteroid treatment preadmission, ICU stays, and ARDS complications, as well as further unscheduled emergency visits after discharge both overall and including a presentation with breathlessness. Classification of at-risk participants using clinically applicable strata identified those who may have had a more severe viral injury and inflammatory response during acute infection, as well as subsequent respiratory exacerbations after COVID-19. Recent analysis identified a hyperinflammatory phenotype of COVID-19–related ARDS was associated with worse outcomes, with better survival linked to corticosteroid treatment [26]. Surviving a hyperinflammatory response to COVID-19 may be consistent with residual lung abnormalities, including fibrosing nonspecific interstitial pneumonia and alveolar damage [27]. Residual lung abnormalities after COVID-19 were not uncommon in this hospitalized population and may persist long-term, but indicators that could support diagnosis and clinical management of lung disease were frequently unavailable. Considering approximately 280,000 people were discharged after confirmed COVID-19 admission in the United Kingdom National Health Service by the end of March 2021 [28], these results emphasize the importance for health services to undertake active radiological and physiological monitoring, especially in people at moderate or above risk [15]. ## Strengths and Limitations The UKILD long–COVID-19 Study interim cohort excluded participants with any evidence of ILD before hospitalization, and we used informative skeptical priors and power priors for more conservative estimates of prevalence, which continued to suggest a substantial burden of residual lung abnormalities after COVID-19 hospitalization. The approach we report can be reasonably applied to other cohorts and time points, with current findings used as informative priors for updating Bayesian inference. Although included CTs were assumed to be representative of clinically indicated radiology, this is limited by local management protocols, the timing of services, and changes to healthcare service prioritization during the COVID-19 pandemic, which increases the chances of selection and ascertainment bias. Furthermore, individuals with linked CT may have unrecorded preexisting disease or present with radiological patterns other than reticulation and ground-glass opacities. Fair interrater agreement (κ, 0.33) of CT scoring was observed, representing agreement in $70\%$ of scans. We recognize these interim findings may also be limited by misclassification. Descriptive analyses identified substantial missing data in clinical risk factors, limiting multiple imputation techniques. We used dichotomized thresholds with single data imputation at the reference category to support risk strata classification, maintain denominators, and provide conservative estimates. In contrast, lung involvement of reticulation and ground-glass opacities was frequently scored on CTs that were clinically indicated, contributing to selection bias. It is similarly likely that repeat CT scans reflect a sample of individuals that did not experience clinical improvement over time. We report estimates from multilevel models to support the interpretation of residual lung abnormalities over time. Although our findings are on the basis of people hospitalized with mixed severity of COVID-19 infection, we recognize that they may not be generalizable to all populations, especially those not admitted to the hospital. Severe admissions requiring CPAP or IMV were overrepresented in the PHOSP–COVID-19 dataset relative to hospitalized survivors of COVID-19 [14]. Linked clinical admission data suggested $50\%$ of at-risk individuals and those scored with residual abnormalities attended ICUs during admission, and up to $25\%$ had complications of anemia and ARDS. Furthermore, these data reflect people who were discharged before the end of March 2021 and do not represent later severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants in fully vaccinated populations that more frequently led to milder infections. ## Conclusions Thresholds of ppDlCO, CXR, and severity of admission can stratify the risk of residual abnormalities on CT involving more than $10\%$ of the lung, informing clinical management, particularly of individuals meeting moderate- to very-high–risk strata. Longitudinal analysis of CT scans suggested the persistence of abnormalities over study time, although the longer-term functional consequence is unknown and may be limited by clinical indication. These findings highlight the importance of radiological and physiological monitoring of patients at both early and later follow-ups and suggest up to $11\%$ of people discharged from an acute COVID-19 admission are at risk of residual lung abnormalities. Further study is required to elucidate the progressive development of radiological patterning or resolution over time. ## Acknowledgment The authors would like to acknowledge the support of the electronic Data Research and Innovation Service team (Public Health Scotland) for their involvement in obtaining approvals, provisioning, and linking data, and the use of the secure analytical platform within the National Safe Haven. This study would not be possible without all the participants who have given their time and support. The authors thank all the participants and their families. The authors thank the many research administrators, healthcare and socialcare professionals who contributed to setting up and delivering the study at all of the 65 NHS trust health boards and 25 research institutions across the United Kingdom, as well as all the supporting staff at the NIHR Clinical Research Network, Health Research Authority, Research Ethics Committee, Department of Health and Social Care, Public Health Scotland, and the United Kingdom Health Security Agency, and support from the ISARIC Coronavirus Clinical Characterisation Consortium (ISARIC4C). The authors thank Kate Holmes at the NIHR Office for Clinical Research Infrastructure (NOCRI) for her support in coordinating the charities group. The PHOSP–COVID-19 industry framework was formed to provide advice and support in commercial discussions, and we thank the Association of the British Pharmaceutical Industry, as well as Ivana Poparic and Peter Sargent at NOCRI, for coordinating this. The authors are very grateful to all the charities that have provided insight into the study: Action Pulmonary Fibrosis, Alzheimer’s Research UK, Asthma + Lung UK, British Heart Foundation, Diabetes UK, Cystic Fibrosis Trust, Kidney Research UK, MQ Mental Health, Muscular Dystrophy UK, Stroke Association Blood Cancer UK, McPin Foundations, and Versus Arthritis. The authors thank the NIHR Leicester Biomedical Research Centre patient and public involvement group and the Long Covid Support Group. ## References 1. Arnold DT, Hamilton FW, Milne A, Morley AJ, Viner J, Attwood M. **Patient outcomes after hospitalisation with COVID-19 and implications for follow-up: results from a prospective UK cohort**. (2021) **76** 399-401. PMID: 33273026 2. Carfì A, Bernabei R, Landi F. **Persistent symptoms in patients after acute COVID-19**. (2020) **324** 603-605. PMID: 32644129 3. Mandal S, Barnett J, Brill SE, Brown JS, Denneny EK, Hare SS. **‘Long-COVID’: a cross-sectional study of persisting symptoms, biomarker and imaging abnormalities following hospitalisation for COVID-19**. (2021) **76** 396-398. PMID: 33172844 4. Carsana L, Sonzogni A, Nasr A, Rossi RS, Pellegrinelli A, Zerbi P. **Pulmonary post-mortem findings in a series of COVID-19 cases from northern Italy: a two-centre descriptive study**. (2020) **20** 1135-1140. PMID: 32526193 5. Ducloyer M, Gaborit B, Toquet C, Castain L, Bal A, Arrigoni PP. **Complete post-mortem data in a fatal case of COVID-19: clinical, radiological and pathological correlations**. (2020) **134** 2209-2214. PMID: 32767018 6. Zhao L, Wang X, Xiong Y, Fan Y, Zhou Y, Zhu W. **Correlation of autopsy pathological findings and imaging features from 9 fatal cases of COVID-19 pneumonia**. (2021) **100** e25232. PMID: 33761714 7. Fadista J, Kraven LM, Karjalainen J, Andrews SJ, Geller F, Baillie JK. **Shared genetic etiology between idiopathic pulmonary fibrosis and COVID-19 severity**. (2021) **65** 103277. PMID: 33714028 8. Allen RJ, Guillen-Guio B, Croot E, Kraven LM, Moss S, Stewart I. **Genetic overlap between idiopathic pulmonary fibrosis and COVID-19**. (2022) **60** 2103132. PMID: 35595312 9. Nasr El-Din A, Ata KAE, Abdel-Gawad AR, Fahmy NF. **Impact of high serum levels of MMP-7, MMP-9, TGF-β and PDGF macrophage activation markers on severity of COVID-19 in obese-diabetic patients**. (2021) **14** 4015-4025. PMID: 34611417 10. Moin ASM, Sathyapalan T, Diboun I, Atkin SL, Butler AE. **Identification of macrophage activation-related biomarkers in obese type 2 diabetes that may be indicative of enhanced respiratory risk in COVID-19**. (2021) **11** 6428. PMID: 33742062 11. Guler SA, Ebner L, Aubry-Beigelman C, Bridevaux PO, Brutsche M, Clarenbach C. **Pulmonary function and radiological features 4 months after COVID-19: first results from the national prospective observational Swiss COVID-19 lung study**. (2021) **57** 2003690. PMID: 33419891 12. Sheng G, Chen P, Wei Y, Yue H, Chu J, Zhao J. **Viral infection increases the risk of idiopathic pulmonary fibrosis: a meta-analysis**. (2020) **157** 1175-1187. PMID: 31730835 13. Wild JM, Porter JC, Molyneaux PL, George PM, Stewart I, Allen RJ. **Understanding the burden of interstitial lung disease post-COVID-19: the UK interstitial lung disease-long COVID study (UKILD-Long COVID)**. (2021) **8** e001049 14. Evans RA, McAuley H, Harrison EM, Shikotra A, Singapuri A, Sereno M. **Physical, cognitive, and mental health impacts of COVID-19 after hospitalisation (PHOSP-COVID): a UK multicentre, prospective cohort study**. (2021) **9** 1275-1287. PMID: 34627560 15. Hatabu H, Hunninghake GM, Richeldi L, Brown KK, Wells AU, Remy-Jardin M. **Interstitial lung abnormalities detected incidentally on CT: a position paper from the Fleischner Society**. (2020) **8** 726-737. PMID: 32649920 16. Drake TM, Docherty AB, Harrison EM, Quint JK, Adamali H, Agnew S. **Outcome of hospitalization for COVID-19 in patients with interstitial lung disease. An international multicenter study**. (2020) **202** 1656-1665. PMID: 33007173 17. Hoff P. *A first course in Bayesian statistical methods* (2009) 18. Duchemann B, Annesi-Maesano I, Jacobe de Naurois C, Sanyal S, Brillet PY, Brauner M. **Prevalence and incidence of interstitial lung diseases in a multi-ethnic county of greater Paris**. (2017) **50** 1602419. PMID: 28775045 19. Maher TM, Bendstrup E, Dron L, Langley J, Smith G, Khalid JM. **Global incidence and prevalence of idiopathic pulmonary fibrosis**. (2021) **22** 197. PMID: 34233665 20. Wu X, Liu X, Zhou Y, Yu H, Li R, Zhan Q. **3-month, 6-month, 9-month, and 12-month respiratory outcomes in patients following COVID-19-related hospitalisation: a prospective study**. (2021) **9** 747-754. PMID: 33964245 21. Han X, Fan Y, Alwalid O, Li N, Jia X, Yuan M. **Six-month follow-up chest CT findings after severe COVID-19 pneumonia**. (2021) **299** E177-E186. PMID: 33497317 22. Han X, Fan Y, Alwalid O, Zhang X, Jia X, Zheng Y. **Fibrotic interstitial lung abnormalities at 1-year follow-up CT after severe COVID-19**. (2021) **301** E438-E440. PMID: 34313470 23. Zou JN, Sun L, Wang BR, Zou Y, Xu S, Ding YJ. **The characteristics and evolution of pulmonary fibrosis in COVID-19 patients as assessed by AI-assisted chest HRCT**. (2021) **16** e0248957. PMID: 33755708 24. Gupta VK, Alkandari BM, Mohammed W, Tobar AM, Abdelmohsen MA. **Ventilator associated lung injury in severe COVID-19 pneumonia patients - case reports: ventilator associated lung injury in COVID-19**. (2020) **8** 100310. PMID: 33364262 25. Fabbri L, Moss S, Khan F, Chi W, Xia J, Robinson KA. (2021) 26. Sinha P, Furfaro D, Cummings MJ, Abrams D, Delucchi K, Maddali MV. **Latent class analysis reveals COVID-19-related acute respiratory distress syndrome subgroups with differential responses to corticosteroids**. (2021) **204** 1274-1285. PMID: 34543591 27. Ravaglia C, Doglioni C, Chilosi M, Piciucchi S, Dubini A, Rossi G. **Clinical, radiological and pathological findings in patients with persistent lung disease following SARS-CoV-2 infection**. (2022) **60** 2102411. PMID: 35301248 28. 28. NHS England 2021 https://www.england.nhs.uk/statistics/statistical-work-areas/covid-19-hospital-activity/. (2021)
--- title: 'Macrophage Inactivation by Small Molecule Wedelolactone via Targeting sEH for the Treatment of LPS-Induced Acute Lung Injury' authors: - Juan Zhang - Min Zhang - Xiao-Kui Huo - Jing Ning - Zhen-Long Yu - Christophe Morisseau - Cheng-Peng Sun - Bruce D. Hammock - Xiao-Chi Ma journal: ACS Central Science year: 2023 pmcid: PMC10037491 doi: 10.1021/acscentsci.2c01424 license: CC BY 4.0 --- # Macrophage Inactivation by Small Molecule Wedelolactone via Targeting sEH for the Treatment of LPS-Induced Acute Lung Injury ## Abstract Soluble epoxide hydrolase (sEH) plays a critical role in inflammation by modulating levels of epoxyeicosatrienoic acids (EETs) and other epoxy fatty acids (EpFAs). Here, we investigate the possible role of sEH in lipopolysaccharide (LPS)-mediated macrophage activation and acute lung injury (ALI). In this study, we found that a small molecule, wedelolactone (WED), targeted sEH and led to macrophage inactivation. Through the molecular interaction with amino acids Phe362 and Gln384, WED suppressed sEH activity to enhance levels of EETs, thus attenuating inflammation and oxidative stress by regulating glycogen synthase kinase 3beta (GSK3β)-mediated nuclear factor-kappa B (NF-κB) and nuclear factor E2-related factor 2 (Nrf2) pathways in vitro. In an LPS-stimulated ALI animal model, pharmacological sEH inhibition by WED or sEH knockout (KO) alleviated pulmonary damage, such as the increase in the alveolar wall thickness and collapse. Additionally, WED or sEH genetic KO both suppressed macrophage activation and attenuated inflammation and oxidative stress in vivo. These findings provided the broader prospects for ALI treatment by targeting sEH to alleviate inflammation and oxidative stress and suggested WED as a natural lead candidate for the development of novel synthetic sEH inhibitors. Targeting sEH with WED to enhance levels of EETs suppressed the macrophage activation via regulating GSK3β-mediated NF-κB and Nrf2 pathways, which resulted in the attenuation of LPS-induced ALI. ## Introduction Macrophages are the first defense line of the innate immune system. They are distributed in many tissues of the human body, particularly in the respiratory tract.1 Macrophages function as scavengers to phagocytize bacteria, pathogens, and inhaled particulates and remove necrotic cells from the pulmonary environment to maintain homeostasis.2,3 As the phagocytosis of macrophages becomes dysfunctional, there is an increase in pathogens and apoptotic cells, which results in secondary necrosis and inflammation and aggravates the pathological course of diseases.4 Meanwhile, the presence of apoptotic cells in the lung maintains the proinflammatory phenotype of macrophages to decrease the ability for resolving the inflammation, which further drives the inflammatory profile in many pulmonary conditions.3 In the respiratory system, activated macrophages, whether caused by pathogens or exogenous substrates, e.g. lipopolysaccharide (LPS) and particulate matter, produce abundant proinflammatory factors, such as tumor necrosis factor-alpha (TNF-α) and interleukin-1 beta and 6 (IL-1β and IL-6), which contributes to the organ damage in inflamed tissues.1,5,6 Furthermore, the activation of macrophages induces the NADPH oxidase 2 (NOX2) and inducible nitric oxide synthase (iNOS) expression to promote the release of reactive oxygen and nitrogen species (ROS and RNS), thereby leading to mitochondrial damage and, finally, to the organ injury.3 Additionally, inflammation and oxidative stress mutually accelerate the damage.7,8 Therefore, blocking inflammation and oxidative stress serves as a vital strategy for the treatment of diseases of the respiratory system.9,10 Polyunsaturated fatty acids (PUFAs) belong to a number of ω-3 and ω-6 families and play an important role for maintaining the physiological health through their bioactive metabolites. The best studied of the polyunsaturated fatty acids is arachidonic acid (AA) and other PUFAs.11−13 AA exists in the form of phospholipids located in membranes and is released to the cytoplasm principally by the action of phospholipase A2.14 *It is* metabolized by three main pathways—cytochrome P450s (CYPs, e.g., CYP3A and CYP2J), cyclooxygenases (COXs, e.g., COX-2), and lipoxygenases (LOXs, e.g., 12-LOX), into bioactive derivatives.14 Among bioactive derivatives of AA, epoxy fatty acids (EpFAs) represented by epoxyeicosatrienoic acids (EETs), produced by several CYP450 oxidases (e.g., CYP2J and CYP2C), have received great attention from scientists because of their outstanding physiological effects, especially anti-inflammatory, antioxidant, and antalgesic activities.14−16 However, EETs are rapidly metabolized in the presence of epoxide hydrolases (EHs) represented by soluble epoxide hydrolase (sEH), which causes the loss of their multiple effects.12,17−22 sEH, encoded by Ephx2, is a bifunctional enzyme with 555 amino acid residues, and its C-terminal mediates the hydrolysis of bioactive epoxy fatty acids (EpFAs), such as EETs and epoxydocosapentaenoic acids (EDPs),14,23 while the role of the N-terminal phosphatase is poorly understood. Recently, sEH inhibition to enhance levels of EETs has become an attractive research strategy to treat diseases related to inflammation, such as lung injury and diabetes.24−26 An effective approach to discover innovative drugs involves probing natural products for possible drugs or leads because of their complex and changeable structures and multiple biological activities.27,28 Increasing evidence supports the therapeutic effects of natural products and traditional Chinese medicines, such as kuraninone, alisol B, (2S,3S)-britanicafanin A, 3β-hydroxy-25-anhydro-alisol F, and Inula japonica.24,25,29−31 Wedelolactone (WED), first isolated from *Wedelia calendulacea* by Govindachari and co-workers in 1956, is a polyphenol sharing a coumarin skeleton with a benzofuran moiety at C-3 and C-4.32 Emerging evidence demonstrates multiple pharmacological responses of WED, including anti-inflammatory, anticancer, and hepatoprotective effects, as well as the remission of Parkinson’s disease (PD) and kidney injury.33−37 Recent research focused on WED has indicated that it attenuates lung collagen deposition and fibrotic pathology in bleomycin-mediated pulmonary fibrosis.38 Similarly, WED protects bronchial epithelial cells from cigarette smoke extract-induced damage, as well.39 However, the protective mechanism and molecular target of WED in macrophage-mediated respiratory diseases, especially acute lung injury (ALI), still remains to be elucidated. Herein, we investigated the ability of WED to reduce inflammation and oxidative stress in LPS-activated macrophages. To understand how it exerted the effect, we performed the target fishing experiments on the basis of the affinity chromatography to identify tentatively the cellular direct target of WED. We found that targeting sEH with WED enhanced levels of EETs to trigger the glycogen synthase kinase 3beta (GSK3β) inhibition, thereby leading to modulation of the nuclear factor-kappa B (NF-κB) and nuclear factor E2-related factor 2 (Nrf2) pathways. Meanwhile, pharmacological sEH inhibition by WED or sEH knockout (KO) exerted a significant therapeutic effect in the ALI animal model treated with LPS. Collectively, this study revealed that sEH served as a valuable target for the treatment of the inflammatory response and oxidative stress of ALI. ## WED Alleviated the Inflammatory Response In Vitro through the NF-κB Pathway We investigated the inflammation resolving effect of WED by first determining the concentration of WED (5, 10, and 20 μM) in LPS-induced RAW264.7 macrophages as the cytotoxicity of 40 μM of WED (cell viability less than $80\%$, Figure S1). As described in Figure 1A, WED displayed a significantly anti-inflammatory effect in LPS-induced macrophages because it dose-dependently suppressed the release of inflammatory factors (e.g., TNF-α, IL-6, and NO; Figure S1). Furthermore, WED reversed the upregulation of LPS-induced inflammatory genes [e.g., TNF-α, iNOS, IL-6, COX-2, monocyte chemoattractant protein-1 (MCP-1), and intercellular cell adhesion molecule-1 (ICAM-1)], and proteins (e.g., TNF-α, COX-2, iNOS, IL-6, and MCP-1) through the NF-κB pathway in a dose-dependent manner (Figure 1B,C). The confocal microscopy results supported the observation that LPS exposure promoted the translocation of the transcript factor p65 to the nucleus, while WED administration suppressed this effect by WED administration (5, 10, and 20 μM; Figure 1D), which was further supported by Western blot to detect the level of p65 in the cytoplasmic and nuclear fractions (Figure 1E). These results supported the anti-inflammatory effect of WED in vitro. **Figure 1:** *WED alleviated inflammatory responses in vitro through the NF-κB pathway. (A) WED suppressed the release of LPS-induced TNF-α (p < 0.0001; df = 4, 10; F-value = 143.3) and IL-6 (p < 0.0001; df = 4, 10; F-value = 238.1) in RAW264.7 cells (mean ± SEM, n = 3, one-way ANOVA). (B) WED downregulated mRNA levels of TNF-α (p < 0.0001; df = 4, 10; F-value = 71.86), IL-6 (p < 0.0001; df = 4, 10; F-value = 147.1), COX-2 (p < 0.0001; df = 4, 10; F-value = 79.7), iNOS (p < 0.0001; df = 4, 10; F-value = 33.1), MCP-1 (p < 0.0001; df = 4, 10; F-value = 21.1), and ICAM-1 (p = 0.0019; df = 4, 10; F-value = 9.6) in LPS-induced RAW264.7 cells (mean ± SEM, n = 3, one-way ANOVA). (C) WED inactivated the NF-κB pathway (p < 0.0001; df = 4, 10; F-value = 28.2) to downregulate expression levels of its target proteins TNF-α (p < 0.0001; df = 4, 10; F-value = 29.6), IL-6 (p < 0.0001; df = 4, 10; F-value = 31.1), COX-2 (p < 0.0001; df = 4, 10; F-value = 105.5), iNOS (p < 0.0001; df = 4, 10; F-value = 41.8), and MCP-1 (p < 0.0001; df = 4, 10; F-value = 23.7) in LPS-induced RAW264.7 cells (mean ± SEM, n = 3, one-way ANOVA). (D,E) WED suppressed the translocation of p65 to the nucleus analyzed by confocal microscopy (D) and Western blot (E) (mean ± SEM, n = 3, one-way ANOVA. For p65 in nucleus, p < 0.0001; df = 4, 10; F-value = 30.3. For p65 in cytoplasm, p = 0.0015; df = 4, 10; F-value = 10.2).* ## WED Alleviated the Oxidative Stress In Vitro through the Nrf2 Pathway Inflammatory response and oxidative stress always coexist in diseases, and they mutually promote and accelerate the progression of diseases. This knowledge stimulated us to examine the antioxidant effect of WED in LPS-induced macrophages. Described in Figure 2A, LPS treatment reduced the glutathione (GSH) level and superoxide dismutase (SOD) activity when compared with the control group, while WED (5, 10, and 20 μM) administration reversed these changes. The result of flow cytometry described in Figure 2B,C also suggested that WED suppressed the increase of LPS-induced ROS-positive cells, which was supported by a fluorescence experiment stained by 2′,7′-dichlorofluorescein diacetate (DCFH-DA) (Figure 2E). The production of ROS usually originates from mitochondrial damage. Thus, we subsequently detected the mitochondrial DNA (mtDNA) content, mitochondrial membrane potential, and genes and proteins responsible for the fusion and fission in the mitochondria. The mtDNA and JC-1 staining revealed that LPS resulted in the increase of the mtDNA copy number and mitochondrial membrane potential (Figure 2D,F), whereas these changes were reversed after administration of WED (5, 10, and 20 μM) in combination with effects of WED on the regulation of mitofusin 1 and 2 (Mfn1 and 2) and optic atrophy 1 (Opa1), which are responsible for the mitochondrial fusion, dynamin-related protein 1 (Drp1), and fission 1 (Fis1), involved in the mitochondrial fission (Figures 2G and S2), thereby requiring the protective effect of WED toward LPS-mediated mitochondrial damage. Nrf2 is the key transcript factor involved in the mitochondrial redox system; thus, we also assayed for the regulatory effect of WED toward this pathway. Pretreatment of WED (5, 10, and 20 μM) upregulated expressions of heme oxygenase-1 (HO-1), glutamate–cysteine ligase catalytic subunit (GCLC), glutamate–cysteine ligase modifier subunit (GCLM), NAD(P)H/quinone oxidoreductase 1 (NQO-1), and Nrf2; downregulated the kelch-like ECH-associated protein l (Keap1) expression; and regulated mRNA levels of their corresponding genes, such as HO-1, Keap1, NQO-1, and Nrf2, in LPS-induced macrophages (Figures 2H,J and S3). Additionally, WED treatment promoted the traffic of Nrf2 to the nucleus in LPS-induced macrophages (Figure 2I,K). This observation was supported by the results of a luciferase experiment showing the agonistic effect of WED toward the Nrf2 receptor (Figure 2L). These findings are consistent with the antioxidant effects of WED in macrophages treated with LPS. **Figure 2:** *WED alleviated the oxidative reduction in vitro through the Nrf2 pathway. (A) WED enhanced levels of GSH (p = 0.0018; df = 4, 10; F-value = 9.8) and SOD (p = 0.0002; df = 4, 10; F-value = 17.1) in LPS-induced RAW264.7 cells (mean ± SEM, n = 3, one-way ANOVA). (B) The flow cytometry demonstrated that WED reduced the LPS-induced ROS-positive cells. (C) Quantitative data of ROS-positive cells in LPS-stimulated RAW264.7 cells treated with WED (mean ± SEM; n = 3; one-way ANOVA; p < 0.0001; df = 4, 10; F-value = 24.0). (D) WED reversed the increase of mtDNA content in LPS-induced RAW264.7 cells (mean ± SEM; n = 3; one-way ANOVA; p < 0.0001; df = 4, 10; F-value = 27.6). (E) The fluorometric analysis indicated that WED suppressed the ROS production. (F) WED reversed the effect of LPS-mediated mitochondrial membrane potential. (G) WED regulated expressions of proteins Mfn1, Mfn2, Op1, Drp1, and Fis1 involved in the mitochondrial fusion and fission. (H) WED activated the Nrf2 pathway to regulate expression levels of HO-1, NQO-1, GCLC, GCLM, Nrf2, and Keap1. (I,K) WED suppressed the translocation of Nrf2 to the nucleus analyzed by Western blot (I) (mean ± SEM, n = 3, one-way ANOVA. For Nrf2 in nucleus, p < 0.0001; df = 4, 10; F-value = 38.9. For Nrf2 in cytoplasm, p = 0.0001; df = 4, 10; F-value = 18.67) and the confocal microscopy (K). (J) WED regulated mRNA levels of genes Keap1 (p < 0.0001; df = 4, 10; F-value = 202.5), HO-1 (p = 0.0002; df = 4, 10; F-value = 16.5), NQO-1 (p < 0.0001; df = 4, 10; F-value = 84.5), and Nrf2 (p < 0.0001; df = 4, 10; F-value = 30.0) involved in the Nrf2 pathway (mean ± SEM, n = 3, one-way ANOVA). (L) The luciferase assay demonstrated the activation of WED against the Nrf2 receptor (mean ± SEM; n = 5; one-way ANOVA; p < 0.0001; df = 3, 16; F-value = 65.6).* ## WED Attenuated the Pulmonary Damage in LPS-Induced ALI Mice The intratracheal instillation of LPS was used to construct the ALI mouse model for investigating the protective effect of WED in vivo. As described in Figure 3A–C, LPS treatment (5 mg/kg) increased the alveolar wall thickness and collapse. It also resulted in the activation of macrophages and the infiltration of neutrophils because of an increase in clusters of differentiation 68 (CD68)- and granulocyte-differentiation antigen-1 (Gr-1)-positive cells compared with the control group, while these changes were reversed in LPS-induced ALI mice after WED (5, 10, and 20 mg/kg) treatment. Furthermore, we found that WED treatment significantly reduced the number of white blood cells (WBCs), polymorphonuclear leukocytes (PMNs), and mononuclear leukocytes (MNs) in the bronchoalveolar lavage fluid (BALF) (Figure 3D), as well as levels of TNF-α and IL-6 and the activities of myeloperoxidase (MPO) and lactate dehydrogenase (LDH) in the BALF and lungs of LPS-induced ALI mice (Figure 3D,E). Collectively, these data reveal that WED attenuates the pathological lung injury. **Figure 3:** *WED attenuated the course of ALI in LPS-induced ALI mice. (A) Representative H&E staining plots. (B) Representative CD68 staining plots. (C) Representative Gr-1 staining plots. (D) WED attenuated the infiltration of proteins (p < 0.0001; df = 5, 42; F-value = 11.3), WBCs (p < 0.0001; df = 5, 42; F-value = 36.2), PMNs (p < 0.0001; df = 5, 42; F-value = 57.8), and MNs (p < 0.0001; df = 5, 42; F-value = 32.1) to the pulmonary alveoli and reduced the production of IL-6 (p < 0.0001; df = 5, 42; F-value = 31.8) and TNF-α (p < 0.0001; df = 5, 42; F-value = 41.2) and the activity of MPO (p < 0.0001; df = 5, 42; F-value = 9.8) and LDH (p < 0.0001; df = 5, 42; F-value = 27.7) in LPS-induced ALI mice (mean ± SEM, n = 8, one-way ANOVA). (E) WED reduced the production of IL-6 (p < 0.0001; df = 5, 42; F-value = 23.8) and TNF-α (p < 0.0001; df = 5, 42; F-value = 14.2) and the activity of MPO (p < 0.0001; df = 5, 42; F-value = 27.6) and LDH (p < 0.0001; df = 5, 42; F-value = 7.9) in the lung of LPS-induced ALI mice (mean ± SEM, n = 8, one-way ANOVA).* ## WED Attenuated LPS-Stimulated Inflammation and Oxidative Stress In Vivo The TNF-α staining results suggest that WED pretreatment (5, 10, and 20 mg/kg) suppressed the increase of TNF-α positive cells in lungs of LPS-induced ALI mice (Figure 4A), as well as the TNF-α mRNA level (Figure 4B). Moreover, WED treatment downregulated mRNA and protein levels of IL-6, iNOS, ICAM-1, COX-2, and MCP-1 through the inactivation of the NF-κB pathway because of inhibition of the phosphorylation of p65 (Figures 4B,C and S4A). **Figure 4:** *WED attenuated the inflammation and oxidative stress in vivo. (A) Representative TNF-α staining plots. (B) Effects of WED against inflammatory genes TNF-α (p < 0.0001; df = 5, 24; F-value = 30.5), IL-6 (p < 0.0001; df = 5, 24; F-value = 56.7), COX-2 (p < 0.0001; df = 5, 24; F-value = 12.8), iNOS (p < 0.0001; df = 5, 24; F-value = 21.0), ICAM-1 (p < 0.0001; df = 5, 24; F-value = 21.4), and MCP-1 (p < 0.0001; df = 5, 24; F-value = 25.5) in LPS-induced ALI mice (mean ± SEM, n = 5, one-way ANOVA). (C) WED suppressed the NF-κB pathway to downregulate expression levels of its target proteins IL-6, COX-2, and MCP-1. (D) Representative scanning electron microscope plots. (E) Effects of WED toward the MDA level (p < 0.0001; df = 5, 42; F-value = 13.4) and the SOD activity (p < 0.0001; df = 5, 42; F-value = 15.4) in LPS-induced ALI mice (mean ± SEM, n = 8, one-way ANOVA). (F) WED reversed the increase of the mtDNA content in LPS-induced ALI mice (mean ± SEM; n = 8; one-way ANOVA; p < 0.0001; df = 5, 24; F-value = 20.3). (G) WED regulated expressions of proteins Mfn1, Mfn2, Op1, Drp1, and Fis1 involved in the mitochondrial fusion and fission in LPS-induced ALI mice. (H) Representative Nrf2 staining plots. (I) WED activated the Nrf2 pathway to regulate expression levels of HO-1, NQO-1, GCLC, GCLM, Nrf2, and Keap1 in LPS-induced ALI mice.* The ROS imbalance leads to oxidative stress, which contributes to the mtDNA damage in the ALI and leads to the abnormal mitochondrial ultrastructure. Figures 4E,F and S4B revealed that the LPS challenge led to a striking increase in the malondialdehyde (MDA) level and the mtDNA content and a remarkable decrease in the levels of GSH and SOD, which were significantly suppressed after WED treatment (5, 10, and 20 mg/kg). The mitochondrial architecture in the control group (Figure 4D) was characterized by isolated, small, and rounded mitochondria with a clear ridge, while the LPS challenge led to mitochondrial swelling and the disappearance of a clear ridge (Figure 4D). The appearance of ridges returned to normal with WED treatment (5, 10, and 20 mg/kg). Meanwhile, we found that WED treatment regulated the expression of proteins and genes involved in mitochondrial fusion (e.g., Mfn1, Op1, and Mfn2) and fission (e.g., Drp1 and Fis1) in LPS-induced ALI, thus allowing the recovery of the mitochondrial function (Figures 4G and S5). Additionally, WED treatment reversed LPS-mediated oxidative stress in vivo by activating the Nrf2 pathway, which was supported by the experiments on the basis of the Nrf2 staining, Western blot, and PCR (Figures 4H,I and S6). These aforementioned results revealed anti-inflammatory and antioxidant effects of WED in vivo. ## sEH Served as the Direct Cellular Target of WED in the Anti-Inflammation and Antioxidation To discover the direct cellular target of WED, we used epoxy-activated Sepharose beads coupled with WED to perform the target fishing experiment. A distinct protein band appeared at ∼63 kDa after the sliver staining (Figure 5A) and was identified as sEH on the basis of the LC-MS/MS analysis (Figure 5B), which was further verified by Western blot using its corresponding antibody (Figure 5C). Meanwhile, the immunofluorescent colocalization analysis demonstrated the direct binding of WED and sEH (Figure 5H,I), which was supported by immunoprecipitation (IP)-MS, cellular thermal shift assay (CETSA), and drug affinity responsive target stability (DARTS) results (Figure 5D–F). The MST and fluorescence-based binding assay results revealed the outstanding binding affinity of WED and sEH with Kd values of 87.2 and 68.7 nM (Figure 5G,K), respectively. sEH as the functional protein with a hydrolase activity in the C-terminal plays a critical role in the AA metabolism (Figure 5J); we found that WED significantly suppressed the human and mouse sEH activity in the enzyme level (Figures 5L and S9). In addition, WED enhanced the level of the sEH substrate 14,15-EET and reduced its diol level in LPS-induced macrophages (Figure 5M). The ratio of 14,15-EET and 14,15-DHET also reflected the inhibitory effect of WED in the cell level (Figure 5M). These results suggest the direct binding effect of WED with sEH. **Figure 5:** *sEH served as the direct target of WED in anti-inflammation and antioxidation. (A) Identification of the cellular target of WED using the pull-down technology on the basis of WED-coupled Sepharose 6B beads and LC–MS/MS analysis. (B) The LC-MS/MS plot of sEH. (C) The binding protein was detected by Western blot. (D) The IP-MS analysis indicated the interaction of WED with sEH. (E) CETSA and DARTS results. (F) Quantitative data of CETSA and DARTS (mean ± SEM; n = 3; one-way ANOVA; p < 0.0001; df = 5, 12; F-value = 100.2). (G) The MST result of WED with sEH (mean ± SEM, n = 3). (H) The scheme of Bio-WED. (I) The colocation of WED with sEH detected by the fluorescence microscope. (J) The CYP/sEH-mediated AA metabolism pathway. (K) The binding capability of WED with sEH detected by the fluorescence-based binding assay (mean ± SEM, n = 3). (L) The inhibitory effect of WED against the human sEH activity detected by the system of human recombinant sEH-mediated hydrolysis of the substrate PHOME (mean ± SEM, n = 3). (M) WED suppressed the sEH activity (14,15-EET/14,15-DHET, p < 0.0001; df = 4, 10; F-value = 55.89) analyzed by levels of 14,15-EET (p = 0.0003; df = 4, 10; F-value = 15.1) and its corresponding diol (p = 0.0034; df = 4, 10; F-value = 8.2) in LPS-mediated RAW264.7 cells (mean ± SEM, n = 3, one-way ANOVA).* ## sEH Knockdown and Rescue Abolished Anti-Inflammatory and Antioxidant Effects of WED In Vitro Next, we performed sEH knockdown and rescue experiments to explore the role of sEH in anti-inflammatory and antioxidant effects of WED in LPS-stimulated macrophages. As shown in Figures S7 and S8, sEH knockdown and rescue could regulate the NF-κB and Nrf2 pathways, which are responsible for inflammatory response and oxidative stress, in macrophages. Moreover, sEH knockdown suppressed the increase of LPS-induced TNF-α, IL-6, COX-2, iNOS, and Keap1 mRNA levels and reversed the decrease of LPS-induced NQO-1 and Nrf2 (Figure 6A,C). Knockdown also reversed effects of LPS toward transcript factors p65 and Nrf2 (Figure 6B), whereas the anti-inflammatory and antioxidant effects of WED were abolished in LPS-mediated sEH knockdown cells (Figures 6B and S10). In addition, sEH rescue aggravated LPS-induced inflammation and oxidative stress. This was illustrated by expressions of phosphorylated p65, Nrf2, and their downstream genes (Figures 6C,D and S10), while the protective effect of WED in vitro after sEH rescue was significantly weakened (Figures 6C,D and S10). These results demonstrate the effects of WED depend on its interaction with the sEH target. **Figure 6:** *sEH knockdown and rescue abolished anti-inflammatory and antioxidant effects of WED in vitro. (A) sEH knockdown abolished the effects of WED toward inflammatory and antioxidant genes TNF-α (p = 0.0028; df = 1, 8; F-value = 18.0), IL-6 (p < 0.0001; df = 1, 8; F-value = 355.1), COX-2 (p = 0.0058; df = 1, 8; F-value = 13.9), iNOS (p < 0.0001; df = 1, 8; F-value = 133.8), NQO-1 (p = 0.0037; df = 1, 8; F-value = 16.4), Nrf2 (p = 0.0194; df = 1, 8; F-value = 8.5), and Keap1 (p = 0.0006; df = 1, 8; F-value = 30.0) in LPS-induced RAW264.7 cells (mean ± SEM, n = 3, two-way ANOVA). (B) sEH knockdown abolished the effects of WED toward the NF-κB and Nrf2 pathways measured by Western blot. (C) sEH rescue weakened the effects of WED toward inflammatory and antioxidant genes TNF-α (p = 0.0006; df = 1, 8; F-value = 29.7), IL-6 (p = 0.0921; df = 1, 8; F-value = 3.7), COX-2 (p = 0.0489; df = 1, 8; F-value = 5.4), iNOS (p = 0.4830; df = 1, 8; F-value = 0.5), NQO-1 (p = 0.6083; df = 1, 8; F-value = 0.3), HO-1 (p = 0.8389; df = 1, 8; F-value = 0.04), and Nrf2 (p = 0.8812; df = 1, 8; F-value = 0.02) in LPS-induced RAW264.7 cells (mean ± SEM, n = 3, two-way ANOVA). (D) sEH rescue weakened the effects of WED toward the NF-κB and Nrf2 pathways measured by Western blot.* ## Phe362 and Gln384 Played Roles in the Binding of WED with sEHfacs To explore the WED-binding sites on sEH, molecular dynamics stimulations were performed (Figures 7 and S11). The root-mean-square deviation (RMSD) of the WED–sEH complex remained about 1.5 Å (Figure 7A) with the binding energy of −108.51 kJ/mol (Figure S11) during the 40 ns of the molecular dynamic stimulation, revealing the stability of the WED–sEH complex. The protein trajectories of apo-sEH and the WED–sEH complex revealed that the binding of WED with sEH reduced the volume of the catalytic cavity of sEH (Figures 7B and S11E,F). There were still about 1–4 hydrogen bonds between WED and Phe362, Ile363, Ser374, Asn378, Tyr383, Gln384, Gln502, and Met503 in the molecular dynamic stimulation (Figure 7C,D). The result of 40 ns of molecular dynamic stimulation was plotted in Figure 7E, which shows that WED interacted with MD2 through two hydrogen bond interactions of amino acid residue Phe362 and Gln384 with distances of 1.6 and 2.6 Å, respectively, thereby demonstrating the role of Phe362 and Gln384 in the binding of WED with sEH. Next, we mutated Phe362 and Gln384 into Phe362Ala and Gln384Ala, respectively, for exploration of their roles. In the pull-down experiment based on the biotin–avidin system of the probe WED coupled with biotin (Bio-WED), the Phe362Ala and Gln384Ala mutations abolished the binding of WED with sEH (Figure 7F), and similar results were afforded from the experiments of CETSA and DARTS (Figure 7F). The Phe362 and Gln384 mutations abolished the anti-inflammatory and antioxidative effects of WED in LPS-stimulated RAW264.7 cells, as well (Figure 7G). These findings illustrate that amino acid residues Phe362 and Gln384 are important sites for WED to bind to sEH. **Figure 7:** *Phe362 and Gln384 played roles in the binding of WED with sEH. (A) The RMSD plot analyzed by molecular dynamics. (B) Effect of WED against the volume of the pocket. (C) The trajectories of Phe362, Ile363, Ser374, and Asn378 with WED. (D) The trajectories of Tyr363, Gln384, Gln502, and Met503 with WED. (E) The interactions of WED with sEH through hydrogen bonds of Phe362 and Gln384. (F) Phe362Ala and Gln384Ala mutations abolished the binding of WED with sEH. (G) Phe362Ala and Gln384Ala mutations abolished the anti-inflammatory [IL-6, p = 0.0014, df = (2, 12), F-value = 12.0; TNF-α, p < 0.0001, df = (2, 12), F-value = 37.4; COX-2, p = 0.0005, df = (2, 12), F-value = 15.1] and antioxidative effects [HO-1, p = 0.0001, df = (2, 12), F-value = 20.6; Nrf2, p < 0.0001, df = (2, 12), F-value = 30.5] of WED in LPS-stimulated RAW264.7 cells (mean ± SEM, n = 3, two-way ANOVA).* ## GSK3β Was the Downstream Key Pathway of sEH in Anti-Inflammatory and Antioxidant Effects of WED Accumulating evidence has demonstrated that the inhibition of sEH suppresses the GSK3β regulation of the NF-κB and Nrf2 pathways in the nervous system.40 First, we investigated the effect of sEH toward GSK3β in macrophages and found that sEH knockdown significantly promoted the phosphorylation of GSK3β (Ser9), thereby allowing the inhibition of the GSK3β activity (Figures 8A and S12A). Similarly, inhibition of sEH by WED could decrease the GSK3β activity in macrophages after LPS exposure (Figures 8C and S12C) while remarkably increasing the GSK3β activity after sEH rescue (Figures 8B and S12B). Therefore, we further investigated whether the downstream key pathway involving the sEH was involved in the protective effect of WED using the GSK3β inhibitor LiCl. Inhibition of GSK3β by LiCl (5 mM) attenuated macrophage-activation-mediated inflammation and oxidative stress, such as expressions of phosphorylated p65, Nrf2, and their downstream genes (Figure 8D,E). It is worth noting that the LiCl plus WED group did not display further additive or synergetic protective effects in inflammation and oxidative stress when compared with the WED group in LPS-induced macrophages (Figure 8D,E). Next, we investigated whether the inhibition of sEH regulated the GSK3β activity through EETs using the sEH substrate 14,15-EET (5 μM). Notably, 14,15-EET reduced the increase of LPS-induced GSK3β activity because of the upregulation of the phosphorylated-GSK3β level and abolished the effect of the inhibition of sEH by WED toward the GSK3β activity (Figure S13). These results suggest that the anti-inflammatory and antioxidant effects of WED depend on sEH to regulate the GSK3β-mediated NF-κB and Nrf2 pathways through EETs in macrophages. **Figure 8:** *GSK3β is the downstream key pathway of sEH in the anti-inflammatory and antioxidant effects of WED. (A) sEH knockdown suppressed the GSK3β activity. (B) sEH rescue activated the GSK3β activity. (C) Inhibition of sEH by WED suppressed the GSK3β activity. (D) Inhibition of GSK3β by LiCl abolished effects of WED toward inflammatory and antioxidant genes TNF-α (p = 0.0316; df = 1, 8; F-value = 6.8), IL-6 (p = 0.0006; df = 1, 8; F-value = 29.8), COX-2 (p < 0.0001; df = 1, 8; F-value = 52.5), iNOS (p = 0.0001; df = 1, 8; F-value = 50.2), HO-1 (p = 0.0097; df = 1, 8; F-value = 11.4), NQO-1 (p = 0.0005; df = 1, 8; F-value = 32.2), and Nrf2 (p < 0.0001; df = 1, 8; F-value = 55.7) in vitro (mean ± SEM, n = 3, two-way ANOVA). (E) Inhibition of GSK3β by LiCl abolished effects of WED toward the NF-κB (p = 0.0154; df = 1, 8; F-value = 9.4) and Nrf2 (p < 0.0001; df = 1, 8; F-value = 69.8) pathways measured by Western blot (mean ± SEM, n = 3, two-way ANOVA).* ## WED Inhibited the sEH Catalytic Activity and Led to the Inhibition of GSK3β In Vivo In the LPS-induced ALI mouse model, we detected the sEH level and demonstrated the sEH overexpression in ALI mice (Figure S14). Next, we found that WED treatment significantly inhibited the decrease of levels of sEH substrates, such as 11,12-EET and 14,15-EET, except for in the case of low dose WED treatment (5 mg/kg, Figures 9A and S15). Conversely, the effect of WED toward 8,9-EET was not significant in LPS-exposed ALI mice except for the high dose of WED (20 mg/kg, Figures 9A and S15). As expected, administration of WED remarkably decreased levels of the corresponding diols, including 8,9-DHET, 11,12-DHET, and 14,15-DHET. Notably, WED treatment suppressed the sEH catalytic activity on the basis of the ratio of EETs and their corresponding DHETs (Figure 9B). Additionally, inhibition of sEH by WED treatment significantly suppressed the GSK3β pathway by promoting its phosphorylation at Ser9 (Figure 9C). All of the abovementioned results demonstrate the effect of sEH inhibition by WED on the GSK3β pathway in LPS-induced ALI mice. **Figure 9:** *WED suppressed the sEH activity to allow the inhibition of GSK3β in vivo. (A) Heat map of sEH substrates (e.g., 8,9-EET, 11,12-EET, and 14,15-EET) and their corresponding diols (e.g., 8,9-DHET, 11,12-DHET, and 14,15-DHET). (B) WED reduced the ratio of 8,9-DHET/8,9-EET (p < 0.0001; df = 5, 30; F-value = 20.5), 11,12-DHET/11,12-EET (p < 0.0001; df = 5, 30; F-value = 37.5), and 14,15-DHET/14,15-EET (p < 0.0001; df = 5, 30; F-value = 22.9) in vivo to suppress the sEH activity (mean ± SEM, n = 6, one-way ANOVA). (C) Inhibition of sEH by WED led to the inactivation of the GSK3β activity (p < 0.0001; df = 5, 12; F-value = 25.3) in LPS-induced ALI mice (mean ± SEM, n = 3, one-way ANOVA).* ## sEH Genetic KO Abolished the Pulmonary Protective Effect of WED In order to explain whether the in vivo pulmonary protective effect of WED depended on the sEH target, we constructed sEH knockout (KO) mice through the deletion of four exons of sEH from exon 2 to exon 5, which were supported by the genotyping and Western blot analysis (Figure 10A). *Ephx2* genetic deletion attenuated pulmonary structural changes (Figure 10B), and the infiltration of macrophages and neutrophils (Figure 10C) contributed to a decrease in the MPO activity and levels of TNF-α and IL-6 in LPS-induced ALI (Figure 10D). Notably, the sEH KO plus WED group did not show further protective effects in LPS-induced ALI mice (Figure 10B–D). Meanwhile, suppression of sEH by WED via contribution to the GSK3β inhibition was not observed in LPS-mediated ALI Ephx2–/– mice (Figure 10E). These results reveal that the pulmonary protective effect of WED depends on sEH. **Figure 10:** *sEH genetic KO abolished the pulmonary protective effect of WED. (A) Genotyping and confirmation of sEH knockout mice. (B,C) Representative plots of H&E (B) and CD68 and Gr-1 (C) staining in LPS-induced ALI Ephx2+/+ and Ephx2–/– mice treated with WED. (D) Measurement of pulmonary MPO (p < 0.0001; df = 1, 20; F-value = 32.7), TNF-α (p = 0.0323; df = 1, 20; F-value = 5.3), and IL-6 (p = 0.0002; df = 1, 20; F-value = 20.5) from LPS-induced ALI Ephx2+/+ and Ephx2–/– mice treated with WED (mean ± SEM, n = 6, two-way ANOVA). (E) Ephx2 KO abolished the effect of WED on the inhibition of GSK3β (p = 0.0019; df = 1, 8; F-value = 20.4) in LPS-induced ALI (mean ± SEM, n = 3, two-way ANOVA).* ## sEH Genetic KO Abolished Anti-Inflammatory and Antioxidant Effects of WED Lastly, the effects of WED on LPS-induced inflammation and oxidative stress were investigated in Ephx2+/+ and Ephx2–/–mice. sEH genetic deletion led to a decrease in TNF-α positive cells (Figure 11A) and in levels of COX-2; phosphorylated p65; and the NF-κB target genes COX-2, IL-6, iNOS, TNF-α, and ICAM-1 in comparison with the LPS-induced ALI Ephx2+/+ mice (Figures 11B,C and S16A). In addition, *Ephx2* genetic KO resulted in an increase in HO-1 and Nrf2 positive cells (Figure 11D) and in the levels of GSH and SOD (Figure 11E). The decreased level of MDA (Figure 11E) depended on the Nrf2 pathway, which contributed to the regulation of Mfn1, Drp1, Fis1, NQO-1, and Nrf2 (Figures 11F,G and S16B). Further effects were not observed in LPS-induced ALI Ephx2–/– mice treated with WED. These results are consistent with anti-inflammatory and antioxidant effects of WED being dependent on sEH inhibition. **Figure 11:** *sEH genetic KO abolished the anti-inflammatory and antioxidant effects of WED. (A) Representative TNF-α staining plots in LPS-induced ALI Ephx2+/+ and Ephx2–/–mice treated with WED. (B) Ephx2 KO abolished the effect of WED on expressions of COX-2 and p-p65/p65 in LPS-induced ALI. (C) Ephx2 KO abolished the effect of WED on mRNA levels of COX-2 (p = 0.0267; df = 1, 16; F-value = 6.0), iNOS (p = 0.0033; df = 1, 16; F-value = 11.9), TNF-α (p = 0.0058; df = 1, 16; F-value = 10.1), IL-6 (p = 0.0003; df = 1, 16; F-value = 21.6), and ICAM-1 (p = 0.0842; df = 1, 16; F-value = 3.4) in LPS-induced ALI (mean ± SEM, n = 5, two-way ANOVA). (D) Representative HO-1 and Nrf2 staining plots in LPS-induced ALI Ephx2+/+ and Ephx2–/– mice treated with WED. (E) Measurement of pulmonary MDA (p = 0.0012; df = 1, 20; F-value = 14.3), GSH (p = 0.0008; df = 1, 20; F-value = 15.5), and SOD (p = 0.0012; df = 1, 20; F-value = 14.4) from LPS-induced ALI Ephx2+/+ and Ephx2–/– mice treated with WED (mean ± SEM, n = 6, two-way ANOVA). (F) Ephx2 KO abolished the effect of WED on expressions of Mfn1, Drp1, HO-1, and Nrf2 in LPS-induced ALI. (G) Ephx2 KO abolished the effect of WED on mRNA levels of NQO-1 (p = 0.0047; df = 1, 16; F-value = 10.8) and Nrf2 (p = 0.0041; df = 1, 16; F-value = 11.2) in LPS-induced ALI (mean ± SEM, n = 5, two-way ANOVA).* ## Discussion The C-terminal hydrolase activity of sEH is in a family of α/β hydrolase proteins. The inhibition of this C-terminal domain stabilizes levels of EETs to regulate various biological processes; therefore, sEH functions as a promising therapeutic target for diseases related to inflammation.14,24 *In this* study, sEH was identified as a key target for ALI by enhancing levels of EETs to suppress the macrophage activation. Moreover, we described the mechanistic insights into the sEH inhibition by the small natural molecule WED by targeting amino acid residues Phe362 and Gln384. Additionally, we demonstrated sEH as an important modulator of the GSK3β-mediated NF-κB and Nrf2 pathways in macrophages for inflammation and oxidative stress. These findings provided broader prospects for the treatment of ALI by targeting sEH to inhibit the macrophage activation and suggested WED as a potential agent for the development of sEH inhibitors. Macrophages are cellular components of the innate immune system that reside in virtually all tissues and possess the capacity for cleaning apoptotic cells and releasing growth factors, thereby contributing to inflammation, innate immune response, and homeostasis.2,3 Depending on the pathogen recognition receptors, macrophages can phagocytize bacteria and other pathogens and inhaled particulates, especially in the respiratory system, meanwhile producing abundant inflammatory mediators,1,41,42 such as TNF-α and IL-6. This brings about mitochondria dysfunction via effects on the expression of proteins involved in the fusion and fission,2,3,8 such as Mfn1, Opa1, Drp1, and Fis1. Furthermore, the mitochondria release an amount of ROS, which contributes to oxidative stress through the inactivation of the Nrf2 pathway. Previous studies have demonstrated the effect of WED toward inflammation and oxidative stress in Parkinson’s disease and kidney injury.33−37 Furthermore, WED has attenuated bleomycin-mediated pulmonary fibrosis and protected bronchial epithelial cells from cigarette-smoke-extract-induced damage,38,39 which suggests the potential of WED against inflammation and oxidative stress in ALI. As our expected, these phenomena were all observed by regulating the NF-κB and Nrf2 pathways in LPS-induced macrophages and ALI animals after WED treatment. Although the ability to define drug target protein recognition technology is rapidly developing, the discovery of drug targets still faces many challenges.43 Recent researches have reported the application of the affinity chromatography, biotin–avidin, and ligand-induced stability shift-dependent proteomics technologies for discovering the direct target of natural products, such as kurarinone, alisol B, eupalinolide B, and handelin.24,44,45 *In this* study, we uncovered that sEH was the direct cellular target of WED on the basis of technologies employing affinity chromatography. The sEH is a bifunctional enzyme with the C-terminal hydrolase and N-terminal phosphatase activities. The sEH is responsible for the hydrolysis of the bioactive metabolites EETs transformed by CYP2J, CYP2C, and other CYP enzymes from AA to their corresponding epoxides.14,23 Increasing evidence has revealed the role of sEH in inflammation and oxidative stress-mediated lung diseases.46−48 For example, chemical inhibition of sEH by 1-trifluoromethoxyphenyl-3-(1-propionylpiperidin-4-yl) urea (TPPU) suppressed LPS-mediated macrophage activation in vitro to regulate the proinflammatory cytokine level,49−51 increased the survival rate of LPS-induced ALI, and attenuated the neutrophil infiltration and alveolar capillary leakage in the ALI animal model.49,50 Furthermore, *Ephx2* genetic KO attenuated the pulmonary inflammation and edema in hyperoxia- or cigarette-smoke-mediated lung injury mice,46−48 thereby suggesting the potential of sEH in macrophage-activation-induced ALI. As expected, sEH knockdown alleviated the inflammation and oxidative stress, and the sEH rescue magnified the abovementioned changes in vitro. Similarly, sEH KO remarkably attenuated the course of ALI in an LPS-induced mouse model. Furthermore, the protective effect of WED was abolished by sEH knockdown, KO, and rescue. Several sEH inhibitors, such as the ureas AR9281 (UC1153), EC5026 (a TPPU analog), and GSK2188931B (an amide), have entered human safety trials.17,52 They are all slow, tightly binding competitive inhibitors based on their urea or amide moieties because of their hydrogen bond interaction with amino acid residues Asp333, Tyr381, and Tyr465, in charge of opening and fixating on the epoxide ring, respectively.17 Recently, we have reported an uncompetitive sEH inhibitor kurarinone with an anti-PD effect and revealed that it interacted with Tyr343, Ile363, Gln384, and Asn472 in the surrounding of the catalytic cavity according to its cocrystal with sEH.24 Similarly, WED suppressed the sEH activity through hydrogen bonds with Phe362 and Gln384, which was supported by the results of the pull-down, CETSA, and DARTS experiments through Phe362Ala and Gln384Ala mutations. Collectively, we first reported that sEH inhibition served as a target for the treatment of ALI through enhancing the EETs level to regulate GSK3β-mediated NF-κB and Nrf2 pathways, which resulted in the inactivation of macrophages in vitro and in vivo. Furthermore, we identified small molecule WED as an inhibitor for targeting sEH through the interaction with amino acid residues Phe362 and Gln384. These findings provid broader prospects for the ALI treatment by targeting sEH to alleviate the inflammation and oxidative stress and suggest WED as a natural product drug and a leading candidate for the development of new sEH inhibitors. ## Chemicals and Reagents WED was isolated and identified from *Inula britannica* on the basis of the 1H and 13C NMR spectra (Figures S17 and S18) with support of a check of purity by thin-layer chromatography and purity and structure by LC-MS. The primary antibodies for sEH (10833-1-AP), NQO-1 (A1518), GSK3β (A2081), p-GSK3β [5558], TNF-α (17590-1-AP), IL-6 (21865-1-AP), iNOS [13120], MCP-1 (25542-1-AP), COX-2 [12282], p65 [8242], p-p65 [3033], Mfn1 (13798-1-AP), Mfn2 (12186-1-AP), HO-1 (A19062), Drp1 (12957-1-AP), GCLC (12601-1-AP), Opa1 (27733-1-AP), GCLM (14241-1-AP), Fis1 (10956-1-AP), Nrf2 (16396-1-AP), and Keap1 (10503-2-AP) were purchased from Cell Signaling Technology (CST, Danvers, MA, USA), Abclonal (Wuhan, China), Proteintech (Wuhan, China), Abcam (Massachusetts, USA), and Affinity (Cincinnati, OH, USA). Recombinant human and mouse sEH afforded from Prof. Bruce D. Hammock (University of California) were gifts. ## Cell Culture and Treatment RAW264.7 macrophages were seeded into 96-well plates and cultured in DMEM with $10\%$ fetal bovine serum (FBS) at 37 °C in humidified air containing $5\%$ CO2 at 37 °C. Overnight, cells were treated with WED. After the incubation for 24 h, the cells were collected for the cellular viability using the CCK-8 kit. The cells were seeded into the 96-well plate overnight and pretreated with WED (5, 10, or 20 μM) for 1 h before the challenge with LPS (500 ng/mL). After 24 h, the supernatants were collected and analyzed for the anti-inflammatory and antioxidant effects of WED. For the sEH rescue experiment, cells were pretreated with sEH (5, 10, or 20 ng/mL) for 1 h, and incubated with WED (20 μM). After 24 h of LPS exposure, the cells were harvested for PCR and Western blot analyses. For the inhibition experiment of GSK3β, cells were pretreated with the GSK3β inhibitor LiCl (5 mM) or 14,15-EET (5 μM) for 1 h before WED (20 μM) treatment. Then cells were administrated LPS for 24 h before being harvested for PCR and Western blot analyses. ## Transient Transfection Small interfering RNA of Ephx2 (siEphx2) and negative control siRNA (siCtrl) were designed and synthesized by GenePharma (Shanghai, China). siCtrl (5 μL) and siEphx2 (5 μL) were transiently transfected into RAW264.7 cells using the transfection reagent (5 μL). After 6 h of transfection, fresh DMEM was added when replacing the medium, and the cells were continuously incubated for 36 h. The cells were harvested for both PCR and Western blot analyses to evaluate the efficiency of siEphx2 silencing. For the confirmatory experiment of targeting sEH with WED, wild-type (WT) and siEphx2 cells were pretreated with WED for 1 h before the LPS challenge (500 ng/mL). Cells were harvested for PCR and Western blot analyses after 24 h of incubation. ## ROS Detection by Flow Cytometry and Fluorescent Microscopic Analysis Cells were pretreated with WED (5, 10, or 20 μM) for 1 h before the challenge with LPS. After 24 h, the ROS were measured by using probe DCFH-DA for 30 min at 37 °C. The cells were collected and used for the analysis of ROS-positive cells through the flow cytometry and a Leica DM4B microscope (Leica, Germany). ## Target Protein Fishing Assay Epoxy-activated Sepharose 6B beads (GE Healthcare, Chicago, USA) coupled with WED were conducted according to the manufacturer’s protocol, as previously described.44 Cell lysates were incubated with WED beads or WED overnight at 4 °C, and PBS was used to elute the nonspecific binding proteins three times. The bead-captured proteins were analyzed by silver staining, Western blot, and a LC-MS/MS system. WED coupled with biotin (Bio-WED) was synthesized and identified by NMR (Figures S19–S22), as previously described,53 for the colocalization of WED and sEH. ## Cellular Thermal Shift Assay (CETSA) Cell lysates were incubated with WED (50 μM) for 30 min at 4 °C. The variable temperature experiments were performed at 46, 50, 54, 58, and 62 °C for 3 min, and the supernatants, afforded by the centrifugation, were analyzed by Western blot using the sEH antibody. The vehicle was used as the control group. ## Drug Affinity Responsive Target Stability (DARTS) Assay Cell lysates pretreated with WED (5, 10, and 20 μM) for 1 h were treated with Pronase (7.5 μg/mL) and incubated for 15 min. After the centrifugation, the supernatants were analyzed by Western blot using the sEH antibody. The vehicle was used as the control group. ## Fluorescence-Based Binding Assay WED (0.1 μM) was incubated with different concentrations of human recombinant sEH in PBS for 2 min, and then, the fluorescence signal was recorded on a microplate reader. The dissociation constant (Kd) of WED with sEH was fitted as previously reported.54 ## Microscale Thermophoresis (MST) Assay First, using the Monolith NT kit to label the sEH protein,55 different concentrations of WED were added in the standard buffer with the labeled sEH protein (200 nM) for the incubation (15 min) and then analyzed on a Monolith NT.115 instrument (NanoTemper Technologies, München, Germany) at room temperature. All the data were analyzed by the NT analysis software to afford the Kd value of WED with sEH. ## Animal and Treatment WT (Ephx2+/+) and sEH KO (Ephx2–/–) C57BL/6 mice (8 weeks, 22–24 g) were obtained from the Experimental Animal Center of Dalian Medical University (Dalian, China) and Cyagen Biosciences Inc. (Guangzhou, China), respectively, and kept under 12 h of light and 12 h of dark environment with a controlled temperature (22–24 °C) and humidity (50–$60\%$). First, WED (5, 10, or 20 mg/kg) and LPS (5 mg/kg) were dissolved in $10\%$ hydroxypropyl β-cyclodextrin and saline, respectively, and then stored at 4 °C. Mice were randomly classified into six groups, including the control, WED (20 mg/kg), LPS, LPS + WED (5 mg/kg), LPS + WED (10 mg/kg), and LPS + WED (20 mg/kg) groups. WED (5, 10, or 20 mg/kg) was administrated intragastrically to mice in the WED and LPS + WED groups for a week by intragastric infusion, followed by administration of LPS (50 μL) through the intratracheal instillation after 1 h of the last administration of WED. Mice in the control and LPS groups were conducted with the corresponding vehicle or LPS (50 μL), according to the abovementioned protocol. After 24 h of LPS administration, BALF was collected as follows: the lungs were lavaged with 0.5 mL saline thrice, and the recovered fluid was used to analyze the levels of proteins IL-6 and TNF-α; activities of MPO and LDH; and the number of WBCs, PMNs, and MNs. Second, Ephx2+/+ and Ephx2–/– mice were classified into four groups (10 mice/group): the LPS (5 mg/kg)-treated Ephx2+/+ group, the LPS (5 mg/kg) + WED (20 mg/kg)-treated Ephx2+/+ group, the LPS (5 mg/kg)-treated Ephx2–/– group, and the LPS (5 mg/kg) + WED (20 mg/kg)-treated Ephx2–/– group. Ephx2+/+ and Ephx2–/– mice were evaluated using the abovementioned protocol, and the lungs were collected. ## Pulmonary Pathological Assessments After being fixed with $4\%$ paraformaldehyde for 24 h and embedded with paraffin, the lungs were sliced into 4 μm thick sections, which were then used for the pulmonary pathological analysis according to the Hematoxylin and Eosin (H&E, Beyotime, Shanghai, China) kit. ## LC-MS/MS Analysis The supernatant of the lung sample was collected after homogenization and centrifugation (20000 g) for 20 min at 4 °C, and then, levels of 8,9-EET, 11,12-EET, 14,15-EET, 8,9-DHET, 11,12-DHET, and 14,15-DHET were calculated on the basis of their standard curves and peak areas afforded from an AB Sciex Qtrap 5500 LC-MS/MS system (Foster City, CA, USA), respectively, as previously reported.24 ## Immunohistochemical and Immunofluorescent Staining For cell samples, cells were incubated with WED or the vehicle in 6-well plates before the LPS (500 ng/mL) challenge. After fixation, the cells were successively incubated with primary antibodies p65, Keap1, or Nrf2 at 4 °C overnight and the fluorescent secondary antibody for 1 h and then analyzed with a Leica DM4B microscope (Leica, Germany). For the colocalization of WED with sEH, the cells were incubated with Bio-WED or the vehicle in 6-well plates for 1 h. After fixation, the cells were incubated with primary sEH antibody at 4 °C overnight and the fluorescent secondary antibody for 1 h and then analyzed by a Leica DM4B microscope (Leica, Germany). For the lung samples, $10\%$ goat serum was used for blocking microwaved lung sections for 20 min, followed by incubation with a cluster of CD68, Gr-1, TNF-α, HO-1, and Nrf2 antibodies overnight at 4 °C. The sections were washed with PBS three times before the incubation with normal or fluorescent secondary antibodies. Finally, the sections were used for the immunohistochemical and immunofluorescent analyses. ## Measurement of MPO, LDH, TNF-α, IL-6, MDA, GSH, and SOD The concentrations of protein in cell, BALF, and lung samples were measured using the BCA method. The activities of MPO, LDH, and SOD and the contents of MDA and GSH in BALF and lung were measured using the appropriate kits (Jiancheng Bioengineering Institute, Nanjing, China). The levels of TNF-α and IL-6 were determined using their corresponding ELISA kits (Elabscience Biotechnology Co., Ltd., Wuhan, China). ## Real-Time Quantitative PCR The total RNA in the cells and lungs was extracted with a TRIzol reagent, and its quantity and purity were analyzed in a Nanodrop spectrophotometer (Thermo, Waltham, MA, USA). The primers of COX-2, Mfn1, Mfn2, Opa1, Drp1, MCP-1, ICAM-1, Fis1, HO-1, TNF-α, NQO-1, GCLC, IL-6, GCLM, Keap1, iNOS, and Nrf2 were added in the cDNA afforded after the reverse transcription, together with TransStart Tip Green qPCR SuperMix (Transgen, Beijing, China), respectively, and then real-time qPCR was performed on an Applied Biosystem 7500 Real-time PCR System (Thermo, Waltham, MA, USA). The copy number of every gene was normalized to the reference gene β-actin, and its relative mRNA expression was determined on the basis of the 2–ΔΔCt method. For the copy number of mtDNA, the DNA was determined using the Universal Genomic DNA Purification Mini Spin Kit (Beyotime, Shanghai, China) and assayed using the primers of mtDNA by real-time qPCR on the basis of the 2–ΔΔCt method. ## Western Blot Cell and lung samples were added in the lysis buffer with cocktail, homogenized, and centrifuged (15 000 rpm, 15 min, 4 °C) to obtain total proteins. Protein (20–30 μg) was loaded on $8\%$–$12\%$ SDS-PAGE for electrophoresis, and the polyvinylidene difluoride (PVDF) membranes transferred with the target protein were incubated with the corresponding primary antibodies overnight at 4 °C after blocking with $5\%$ skim milk for 2 h. After the incubation with a horseradish peroxidase-conjugated secondary antibody, the membranes were incubated with the ECL reagent, followed by detection on the Tanon 5200-ECL detection system. ## Immunoprecipitation (IP) Cell lysates were treated with WED (20 μM) at 4 °C for 30 min and then incubated with sEH or IgG antibody and protein A/G magnetic beads at 4 °C overnight. The resulting protein was analyzed by Western blot and the LC-MS/MS system. ## Soluble Epoxide Hydrolase Activity Assay The inhibition of human and mouse sEH by WED was evaluated using PHOME as the substrate, as previously described.24,56,57 ## Molecular Dynamics Stimulation The interaction of WED with sEH (PDB: 4OCZ) was analyzed using the GROMACS package, according to previous methods,24,56,57 and plotted by PyMOL 2.4 software. ## Statistical Analysis All the data are presented as means ± standard error of the mean (SEM) and analyzed on the basis of one-way ANOVA followed by Tukey’s test in Prism GraphPad Prism 8.0, except for the data in the experiments of sEH knockdown, sEH rescue, the inhibition of GSK3β by LiCl, and sEH KO experiment (two-way ANOVA followed by Sidak’s test). If the p-value was less than 0.05, the result was considered significant. ## References 1. Robinson N., Ganesan R., Hegedus C., Kovacs K., Kufer T. A., Virag L.. **Programmed necrotic cell death of macrophages: Focus on pyroptosis, necroptosis, and parthanatos**. *Redox Biol.* (2019) **26** 101239. DOI: 10.1016/j.redox.2019.101239 2. Kopf M., Schneider C., Nobs S. P.. **The development and function of lung-resident macrophages and dendritic cells**. *Nat. Immunol.* (2015) **16** 36-44. DOI: 10.1038/ni.3052 3. Belchamber K. B. R., Donnelly L. E.. **Macrophage dysfunction in respiratory disease**. *Results Probl. Cell Differ.* (2017) **62** 299-313. DOI: 10.1007/978-3-319-54090-0_12 4. Henson P. M., Tuder R. M.. **Apoptosis in the lung: induction, clearance and detection**. *Am. J. Physiol. Lung Cell. Mol. Physiol.* (2008) **294** L601-11. DOI: 10.1152/ajplung.00320.2007 5. Qin X., Zhou Y., Jia C., Chao Z., Qin H., Liang J., Liu X., Liu Z., Sun T., Yuan Y., Zhang H.. **Caspase-1-mediated extracellular vesicles derived from pyroptotic alveolar macrophages promote inflammation in acute lung injury**. *Int. J. Biol. Sci.* (2022) **18** 1521-1538. DOI: 10.7150/ijbs.66477 6. Zhang H., Liu J., Zhou Y., Qu M., Wang Y., Guo K., Shen R., Sun Z., Cata J. P., Yang S., Chen W., Miao C.. **Neutrophil extracellular traps mediate m(6)A modification and regulates sepsis-associated acute lung injury by activating ferroptosis in alveolar epithelial cells**. *Int. J. Biol. Sci.* (2022) **18** 3337-3357. DOI: 10.7150/ijbs.69141 7. Liu Y., Xiang D., Zhang H., Yao H., Wang Y.. **Hypoxia-inducible factor-1: A potential target to treat acute lung injury**. *Oxid. Med. Cell. Longev.* (2020) **2020** 8871476. DOI: 10.1155/2020/8871476 8. Shi J., Yu T., Song K., Du S., He S., Hu X., Li X., Li H., Dong S., Zhang Y., Xie Z., Li C., Yu J.. **Dexmedetomidine ameliorates endotoxin-induced acute lung injury in vivo and in vitro by preserving mitochondrial dynamic equilibrium through the HIF-1a/HO-1 signaling pathway**. *Redox Biol.* (2021) **41** 101954. DOI: 10.1016/j.redox.2021.101954 9. Ho M. S., Mei S. H., Stewart D. J.. **The immunomodulatory and therapeutic effects of mesenchymal stromal cells for acute lung injury and sepsis**. *J. Cell. Physiol.* (2015) **230** 2606-17. DOI: 10.1002/jcp.25028 10. Gustot T.. **Multiple organ failure in sepsis: prognosis and role of systemic inflammatory response**. *Curr. Opin. Crit. Care* (2011) **17** 153-9. DOI: 10.1097/MCC.0b013e328344b446 11. Laye S., Nadjar A., Joffre C., Bazinet R. P.. **Anti-inflammatory effects of omega-3 fatty acids in the brain: Physiological mechanisms and relevance to pharmacology**. *Pharmacol. Rev.* (2018) **70** 12-38. DOI: 10.1124/pr.117.014092 12. Pallas M., Vazquez S., Sanfeliu C., Galdeano C., Grinan-Ferre C.. **Soluble epoxide hydrolase inhibition to face neuroinflammation in Parkinson’s disease: A new therapeutic strategy**. *Biomolecules* (2020) **10** 703. DOI: 10.3390/biom10050703 13. Feuerer N., Marzi J., Brauchle E. M., Carvajal Berrio D. A., Billing F., Weiss M., Jakobi M., Schneiderhan-Marra N., Shipp C., Schenke-Layland K.. **Lipidome profiling with Raman microspectroscopy identifies macrophage response to surface topographies of implant materials**. *Proc. Natl. Acad. Sci. U. S. A.* (2021) **118** e2113694118. DOI: 10.1073/pnas.2113694118 14. Sun C. P., Zhang X. Y., Morisseau C., Hwang S. H., Zhang Z. J., Hammock B. D., Ma X. C.. **Discovery of soluble epoxide hydrolase inhibitors from chemical synthesis and natural products**. *J. Med. Chem.* (2021) **64** 184-215. DOI: 10.1021/acs.jmedchem.0c01507 15. Wang Y., Wagner K. M., Morisseau C., Hammock B. D.. **Inhibition of the soluble epoxide hydrolase as an analgesic strategy: A review of preclinical evidence**. *J. Pain Res.* (2021) **14** 61-72. DOI: 10.2147/JPR.S241893 16. An G., Lee K. S. S., Yang J., Hammock B. D.. **Target-mediated drug disposition-a class effect of soluble epoxide hydrolase inhibitors**. *J. Clin. Pharmacol.* (2021) **61** 531-537. DOI: 10.1002/jcph.1763 17. Fuller-Pace F. V., Southern P. J.. **Temporal analysis of transcription and replication during acute infection with lymphocytic choriomeningitis virus**. *Virology* (1988) **162** 260-3. DOI: 10.1016/0042-6822(88)90419-9 18. Deng J., Yang H., Haak V. M., Yang J., Kipper F. C., Barksdale C., Hwang S. H., Gartung A., Bielenberg D. R., Subbian S., Ho K. K., Ye X., Fan D., Sun Y., Hammock B. D., Panigrahy D.. **Eicosanoid regulation of debris-stimulated metastasis**. *Proc. Natl. Acad. Sci. U. S. A.* (2021) **118** e2107771118. DOI: 10.1073/pnas.2107771118 19. Pu Y., Yang J., Chang L., Qu Y., Wang S., Zhang K., Xiong Z., Zhang J., Tan Y., Wang X., Fujita Y., Ishima T., Wan D., Hwang S. H., Hammock B. D., Hashimoto K.. **Maternal glyphosate exposure causes autism-like behaviors in offspring through increased expression of soluble epoxide hydrolase**. *Proc. Natl. Acad. Sci. U. S. A.* (2020) **117** 11753-11759. DOI: 10.1073/pnas.1922287117 20. Fishbein A., Wang W., Yang H., Yang J., Hallisey V. M., Deng J., Verheul S. M. L., Hwang S. H., Gartung A., Wang Y., Bielenberg D. R., Huang S., Kieran M. W., Hammock B. D., Panigrahy D.. **Resolution of eicosanoid/cytokine storm prevents carcinogen and inflammation-initiated hepatocellular cancer progression**. *Proc. Natl. Acad. Sci. U. S. A.* (2020) **117** 21576-21587. DOI: 10.1073/pnas.2007412117 21. Bergmann C. B., McReynolds C. B., Wan D., Singh N., Goetzman H., Caldwell C. C., Supp D. M., Hammock B. D.. **sEH-derived metabolites of linoleic acid drive pathologic inflammation while impairing key innate immune cell function in burn injury**. *Proc. Natl. Acad. Sci. U. S. A.* (2022) **119** e2120691119. DOI: 10.1073/pnas.2120691119 22. Wang Y., Yang J., Wang W., Sanidad K. Z., Cinelli M. A., Wan D., Hwang S. H., Kim D., Lee K. S. S., Xiao H., Hammock B. D., Zhang G.. **Soluble epoxide hydrolase is an endogenous regulator of obesity-induced intestinal barrier dysfunction and bacterial translocation**. *Proc. Natl. Acad. Sci. U. S. A.* (2020) **117** 8431-8436. DOI: 10.1073/pnas.1916189117 23. Gomez G. A., Morisseau C., Hammock B. D., Christianson D. W.. **Structure of human epoxide hydrolase reveals mechanistic inferences on bifunctional catalysis in epoxide and phosphate ester hydrolysis**. *Biochemistry* (2004) **43** 4716-23. DOI: 10.1021/bi036189j 24. Sun C. P., Zhou J. J., Yu Z. L., Huo X. K., Zhang J., Morisseau C., Hammock B. D., Ma X. C.. **Kurarinone alleviated Parkinson’s disease via stabilization of epoxyeicosatrienoic acids in animal model**. *Proc. Natl. Acad. Sci. U. S. A.* (2022) **119** e2118818119. DOI: 10.1073/pnas.2118818119 25. Zhao W. Y., Yan J. J., Zhang M., Wang C., Feng L., Lv X., Huo X. K., Sun C. P., Chen L. X., Ma X. C.. **Natural soluble epoxide hydrolase inhibitors from**. *Chem. Biol. Interact.* (2021) **345** 109571. DOI: 10.1016/j.cbi.2021.109571 26. Imig J. D., Hammock B. D.. **Soluble epoxide hydrolase as a therapeutic target for cardiovascular diseases**. *Nat. Rev. Drug Discov.* (2009) **8** 794-805. DOI: 10.1038/nrd2875 27. He Y. Q., Zhou C. C., Yu L. Y., Wang L., Deng J. L., Tao Y. L., Zhang F., Chen W. S.. **Natural product derived phytochemicals in managing acute lung injury by multiple mechanisms**. *Pharmacol. Res.* (2021) **163** 105224. DOI: 10.1016/j.phrs.2020.105224 28. Thomford N. E., Senthebane D. A., Rowe A., Munro D., Seele P., Maroyi A., Dzobo K.. **Natural products for drug discovery in the 21st century: Innovations for novel drug discovery**. *Int. J. Mol. Sci.* (2018) **19** 1578. DOI: 10.3390/ijms19061578 29. Zhao W. Y., Zhang X. Y., Zhou M. R., Tian X. G., Lv X., Zhang H. L., Deng S., Zhang B. J., Sun C. P., Ma X. C.. **Natural soluble epoxide hydrolase inhibitors from**. *Int. J. Biol. Macromol.* (2021) **183** 811-817. DOI: 10.1016/j.ijbiomac.2021.04.187 30. Zhang J., Zhang M., Zhang W. H., Zhu Q. M., Huo X. K., Sun C. P., Ma X. C., Xiao H. T.. **Total flavonoids of**. *Phytomedicine* (2022) **107** 154380. DOI: 10.1016/j.phymed.2022.154380 31. Zhang J., Luan Z. L., Huo X. K., Zhang M., Morisseau C., Sun C. P., Hammock B. D., Ma X. C.. **Direct targeting of sEH with alisol B alleviated the apoptosis, inflammation, and oxidative stress in cisplatin-induced acute kidney injury**. *Int. J. Biol. Sci.* (2023) **19** 294-310. DOI: 10.7150/ijbs.78097 32. Govindachari T. R., Nagarajan K., Pai B. R.. **Chemical esamination of**. *J. Chem. Sci.* (1956) 629-632. DOI: 10.1039/jr9560000629 33. Luo Q., Ding J., Zhu L., Chen F., Xu L.. **Hepatoprotective effect of wedelolactone against concanavalin A-induced liver injury in mice**. *Am. J. Chin. Med.* (2018) **46** 819-833. DOI: 10.1142/S0192415X1850043X 34. Zhu M. M., Wang L., Yang D., Li C., Pang S. T., Li X. H., Li R., Yang B., Lian Y. P., Ma L., Lv Q. L., Jia X. B., Feng L.. **Wedelolactone alleviates doxorubicin-induced inflammation and oxidative stress damage of podocytes by IkappaK/IkappaB/NF-kappaB pathway**. *Biomed. Pharmacother.* (2019) **117** 109088. DOI: 10.1016/j.biopha.2019.109088 35. Pan H., Lin Y., Dou J., Fu Z., Yao Y., Ye S., Zhang S., Wang N., Liu A., Li X., Zhang F., Chen D.. **Wedelolactone facilitates Ser/Thr phosphorylation of NLRP3 dependent on PKA signalling to block inflammasome activation and pyroptosis**. *Cell Prolif.* (2020) **53** e12868. DOI: 10.1111/cpr.12868 36. Sharma S., Trivedi S., Pandey T., Ranjan S., Trivedi M., Pandey R.. **Wedelolactone mitigates parkinsonism via alleviating oxidative stress and mitochondrial dysfunction through NRF2/SKN-1**. *Mol. Neurobiol.* (2021) **58** 65-77. DOI: 10.1007/s12035-020-02080-4 37. Wang G., Bi Y., Xiong H., Bo T., Han L., Zhou L., Zhang C., Zhang Y.. **Wedelolactone protects against cisplatin-induced nephrotoxicity in mice via inhibition of organic cation transporter 2**. *Hum. Exp. Toxicol.* (2021) **40** S447-S459. DOI: 10.1177/09603271211047915 38. Yang J. Y., Tao L. J., Liu B., You X. Y., Zhang C. F., Xie H. F., Li R. S.. **Wedelolactone attenuates pulmonary fibrosis partly through activating AMPK and regulating Raf-MAPKs signaling pathway**. *Front. Pharmacol.* (2019) **10** 151. DOI: 10.3389/fphar.2019.00151 39. Ding S., Hou X., Yuan J., Tan X., Chen J., Yang N., Luo Y., Jiang Z., Jin P., Dong Z., Feng L., Jia X.. **Wedelolactone protects human bronchial epithelial cell injury against cigarette smoke extract-induced oxidant stress and inflammation responses through Nrf2 pathway**. *Int. Immunopharmacol.* (2015) **29** 648-655. DOI: 10.1016/j.intimp.2015.09.015 40. Sun C. P., Zhang X. Y., Zhou J. J., Huo X. K., Yu Z. L., Morisseau C., Hammock B. D., Ma X. C.. **Inhibition of sEH via stabilizing the level of EETs alleviated Alzheimer’s disease through GSK3beta signaling pathway**. *Food Chem. Toxicol.* (2021) **156** 112516. DOI: 10.1016/j.fct.2021.112516 41. Yuan R., Li Y., Han S., Chen X., Chen J., He J., Gao H., Yang Y., Yang S., Yang Y.. **Fe-curcumin nanozyme-mediated reactive oxygen species scavenging and anti-inflammation for acute lung injury**. *ACS Cent. Sci.* (2022) **8** 10-21. DOI: 10.1021/acscentsci.1c00866 42. Duan X., Liu B.. **A nanozymatic solution to acute lung injury**. *ACS Cent. Sci.* (2022) **8** 7-9. DOI: 10.1021/acscentsci.1c01494 43. Rinschen M. M., Ivanisevic J., Giera M., Siuzdak G.. **Identification of bioactive metabolites using activity metabolomics**. *Nat. Rev. Mol. Cell Biol.* (2019) **20** 353-367. DOI: 10.1038/s41580-019-0108-4 44. Wang L. C., Liao L. X., Lv H. N., Liu D., Dong W., Zhu J., Chen J. F., Shi M. L., Fu G., Song X. M., Jiang Y., Zeng K. W., Tu P. F.. **Highly selective activation of heat shock protein 70 by allosteric regulation provides an insight into efficient neuroinflammation inhibition**. *EBioMedicine* (2017) **23** 160-172. DOI: 10.1016/j.ebiom.2017.08.011 45. Zhang X. W., Feng N., Liu Y. C., Guo Q., Wang J. K., Bai Y. Z., Ye X. M., Yang Z., Yang H., Liu Y., Yang M. M., Wang Y. H., Shi X. M., Liu D., Tu P. F., Zeng K. W.. **Neuroinflammation inhibition by small-molecule targeting USP7 noncatalytic domain for neurodegenerative disease therapy**. *Sci. Adv.* (2022) **8** eabo0789. DOI: 10.1126/sciadv.abo0789 46. Li Y., Yu G., Yuan S., Tan C., Lian P., Fu L., Hou Q., Xu B., Wang H.. **Cigarette smoke-induced pulmonary inflammation and autophagy are attenuated in**. *Inflammation* (2017) **40** 497-510. DOI: 10.1007/s10753-016-0495-z 47. Liu L. P., Li B., Shuai T. K., Zhu L., Li Y. M.. **Deletion of soluble epoxide hydrolase attenuates mice hyperoxic acute lung injury**. *BMC Anesthesiol.* (2018) **18** 48. DOI: 10.1186/s12871-018-0490-z 48. Li P. S., Tao W., Yang L. Q., Shu Y. S.. **Effect of soluble epoxide hydrolase in hyperoxic acute lung injury in mice**. *Inflammation* (2018) **41** 1065-1072. DOI: 10.1007/s10753-018-0758-y 49. Luo X. Q., Duan J. X., Yang H. H., Zhang C. Y., Sun C. C., Guan X. X., Xiong J. B., Zu C., Tao J. H., Zhou Y., Guan C. X.. **Epoxyeicosatrienoic acids inhibit the activation of NLRP3 inflammasome in murine macrophages**. *J. Cell. Physiol.* (2020) **235** 9910-9921. DOI: 10.1002/jcp.29806 50. Zhou Y., Liu T., Duan J. X., Li P., Sun G. Y., Liu Y. P., Zhang J., Dong L., Lee K. S. S., Hammock B. D., Jiang J. X., Guan C. X.. **Soluble Epoxide Hydrolase Inhibitor Attenuates Lipopolysaccharide-Induced Acute Lung Injury and Improves Survival in Mice**. *Shock* (2017) **47** 638-645. DOI: 10.1097/SHK.0000000000000767 51. Dong L., Zhou Y., Zhu Z. Q., Liu T., Duan J. X., Zhang J., Li P., Hammcok B. D., Guan C. X.. **Soluble epoxide hydrolase inhibitor suppresses the expression of triggering receptor expressed on myeloid cells-1 by inhibiting NF-kB activation in murine macrophage**. *Inflammation* (2017) **40** 13-20. DOI: 10.1007/s10753-016-0448-6 52. Lazaar A. L., Yang L., Boardley R. L., Goyal N. S., Robertson J., Baldwin S. J., Newby D. E., Wilkinson I. B., Tal-Singer R., Mayer R. J., Cheriyan J.. **Pharmacokinetics, pharmacodynamics and adverse event profile of GSK2256294, a novel soluble epoxide hydrolase inhibitor**. *Br. J. Clin. Pharmacol.* (2016) **81** 971-9. DOI: 10.1111/bcp.12855 53. Wu G., Zhu L., Yuan X., Chen H., Xiong R., Zhang S., Cheng H., Shen Y., An H., Li T., Li H., Zhang W.. **Britanin ameliorates cerebral ischemia-reperfusion injury by inducing the Nrf2 protective pathway**. *Antioxid. Redox Signal.* (2017) **27** 754-768. DOI: 10.1089/ars.2016.6885 54. Wang K. B., Dickerhoff J., Yang D.. **Solution structure of ternary complex of berberine bound to a dGMP-fill-in vacancy G-quadruplex formed in the PDGFR-beta promoter**. *J. Am. Chem. Soc.* (2021) **143** 16549-16555. DOI: 10.1021/jacs.1c06200 55. Zhao C., Wang D., Gao Z., Kan H., Qiu F., Chen L., Li H.. **Licocoumarone induces BxPC-3 pancreatic adenocarcinoma cell death by inhibiting DYRK1A**. *Chem. Biol. Interact.* (2020) **316** 108913. DOI: 10.1016/j.cbi.2019.108913 56. Sun C. P., Zhang J., Zhao W. Y., Yi J., Yan J. K., Wang Y. L., Morisseau C., Liu Z. B., Hammock B. D., Ma X. C.. **Protostane-type triterpenoids as natural soluble epoxide hydrolase inhibitors: Inhibition potentials and molecular dynamics**. *Bioorg. Chem.* (2020) **96** 103637. DOI: 10.1016/j.bioorg.2020.103637 57. Liu Z. B., Sun C. P., Xu J. X., Morisseau C., Hammock B. D., Qiu F.. **Phytochemical constituents from**. *Int. J. Biol. Macromol.* (2019) **133** 1187-1193. DOI: 10.1016/j.ijbiomac.2019.04.055
--- title: The cut-off value for HOMA-IR discriminating the insulin resistance based on the SHBG level in women with polycystic ovary syndrome authors: - Aleksandra Biernacka-Bartnik - Piotr Kocełak - Aleksander Jerzy Owczarek - Piotr Stanisław Choręza - Leszek Markuszewski - Paweł Madej - Monika Puzianowska-Kuźnicka - Jerzy Chudek - Magdalena Olszanecka-Glinianowicz journal: Frontiers in Medicine year: 2023 pmcid: PMC10037532 doi: 10.3389/fmed.2023.1100547 license: CC BY 4.0 --- # The cut-off value for HOMA-IR discriminating the insulin resistance based on the SHBG level in women with polycystic ovary syndrome ## Abstract ### Introduction The study aimed to estimate the cut-off value for homeostatic model assessment for insulin resistance (HOMA-IR) discriminating the insulin resistance based on the sex hormones binding globulin (SHBG) level in women with polycystic ovary syndrome (PCOS). ### Materials and methods Data from medical records of 854 Caucasian women diagnosed with PCOS were analyzed. Anthropometric data, fasting plasma glucose, insulin and SHBG levels were measured. HOMA-IR was calculated with a standard formula. The cut-off value was calculated using receiver-operating characteristics. ### Results Circulating SHBG levels below the normal range (26.1 nmol/L) were found in $25.4\%$ of study participants. This subgroup had a significantly higher BMI, fasting glucose and insulin concentrations and HOMA-IR values. Empirical optimal cut-off values for HOMA-IR corresponding to low SHBG levels was ≥2.1 [area under the curve (AUC) 0.73, accuracy 0.65, sensitivity $72.3\%$, specificity $63.1\%$, positive predictive value (PPV) $40.0\%$, negative predictive value (NPV) $87.0\%$]. ### Conclusions Our study suggests that the cut-off point for HOMA-IR discriminating the insulin resistance based on the SHBG level, in young Caucasian women with polycystic ovary syndrome is 2.1, and is consistent with the cut-off value adopted by the European Group for the Study of Insulin Resistance (above 2.0). ## Introduction Sex hormone binding globulin (SHBG) is a homodimer glycoprotein with a high affinity and specificity for androgens and estrogens [1]. It is produced mainly in the liver and its synthesis is regulated mostly by circulating sex hormones and hyperinsulinemia compensating insulin resistance (2–4). Thus, SHBG may be a useful marker of the severity of hepatic insulin resistance and fatty liver that is linked to hepatic insulin resistance. Numerous previously published studies demonstrated that low circulating SHBG levels may serve as a surrogate marker of fatty liver (5–7). It has also been shown that SHBG levels were inversely proportional to the severity of fatty liver, insulin levels and homeostatic model assessment for insulin resistance (HOMA-IR) values [8]. Moreover, the expression of SHBG mRNA correlated negatively with the accumulation of triglycerides in hepatocytes [9]. A meta-analysis confirmed these observations, showing that low SHBG levels correlate with non-alcoholic fatty liver disease (NAFLD) in both women and men [10]. One of the consequences of hepatic insulin resistance in NAFLD is increased gluconeogenesis resulting in the impaired fasting glucose level. Concurrently, the lower SHBG level is the predictor of type 2 diabetes [11]. During a 5 years follow-up, men with the lowest SHBG levels had a four-fold higher risk of type 2 diabetes [12]. This finding was corroborated by a meta-analysis of 13 prospective, observational studies [13]. In a large cohort study including 42,034 women, a higher risk of type 2 diabetes was associated with SHBG levels < 50 nmol/L [14]. The role of SHBG in type 2 diabetes development is supported by experimental studies performed with the insulin-resistant human trophoblast cells (HTR8-SVneo cell line) characterized by low expression of SHBG, GLUT-3 and GLUT-4 (glucose transporters type 3 and 4) as well as high expression of GLUT-1. Notably, overexpression of SHBG inhibited levels of GLUT-1 mRNA and promoted the expression of GLUT-3 and GLUT-4. This finding suggests that SHBG may affect glucose metabolism and induce insulin resistance by regulating the activity of glucose transporters [15]. In addition, incubation of macrophages and adipocytes with 20 nM SHBG significantly inhibited the synthesis of proinflammatory cytokines (monocyte chemoattractant protein-1, tumor necrosis factor and interleukin-6) induced by lipopolysaccharide treatment [16]. Polycystic ovary syndrome (PCOS) is defined as multiple endocrine and metabolic disturbances, among which the central position is ovarian dysfunction. Insulin resistance is one of the key factors in the pathogenesis of hormonal and metabolic disturbances observed in women with PCOS. However, it should be noted that insulin resistance is not a part of PCOS diagnosis. A gold standard for the assessment of insulin resistance is the hyperinsulinemic-euglycemic clamp technique. However, this method is very complicated and is not used in daily clinical practice. In clinical studies and daily practice, insulin resistance is assessed on the basis of a mathematical model named HOMA-IR, which probably reflects more hepatic than muscle insulin resistance [17]. However, there is a lack of a clearly defined cut-off point for HOMA-IR related to insulin resistance. Among many of the proposed values for the general population, the value of 2.5 and above is most often used [18]. Notwithstanding, studies performed in Caucasian and Thai women with PCOS suggested the HOMA-IR cut-off value of at least 2.0 [19, 20]. Also, the European Group for the Study of Insulin Resistance uses the same cut-off point (≥2.0) [21]. As mentioned above, compensatory hyperinsulinemia inhibits hepatic SHBG synthesis. Concordantly, we hypothesized that SHBG level may be a useful marker of the severity of hepatic insulin resistance. Contrary to the detectable cut-off point characterizing insulin resistance, the laboratory assays for SHBG have specified reference ranges and its lower limit may be used to establish a corresponding HOMA-IR cut-off point. Therefore, the aim of this study was to estimate the cut-off value for HOMA-IR discriminating the insulin resistance based on the SHBG level in women with PCOS. ## Materials and methods The retrospective study includes data from the medical records of 859 Caucasian women for the first time diagnosed with PCOS on the basis of the *Rotterdam criteria* [22], hospitalized at the Department of Gynecological Endocrinology from 2012 to 2019. The inclusion criteria included age 18–30 years and diagnosis of PCOS. The exclusion criteria were: diagnosis of type 2 diabetes and other endocrinological disturbances, any pharmacological therapy, treatment of obesity in the past and currently and the lack of necessary data in the medical records. The analyzed data set included: age, body mass, height and routine measurements of fasting glucose, insulin and SHBG levels, all performed in a single hospital laboratory using the same set of methods for all study subjects. Glucose concentration was measured using the colourimetric method (Roche reagents for Cobas e111). Insulin and SHBG levels were determined using the ECLIA method (Roche Diagnostic GmbH, Mannheim, Germany reagents for Cobas E411). Body mass index (BMI) and HOMA-IR values were calculated with standard formulas: As the retrospective analysis of patients' records does not meet the criteria of a medical experiment, the approval of the Bioethical Committee was not required. ## Data analysis Women with HOMA-IR values above 10 ($$n = 5$$)—data outliers, related to non-compliance and to the assessment of measured parameters in non-fasting subjects, were excluded from the analysis. The remaining women were divided according to the lower limit of the SHBG concentration laboratory's reference range for women aged 18–50 years (< 26.1 nmol/L) into a subgroup with concentrations above and below this limit [$$n = 637$$ ($74.6\%$) and $$n = 217$$ ($25.4\%$), respectively]. ## Statistical analysis Statistical analysis was performed using STATISTICA 13.0 PL (TIBCO Software Inc., Palo Alto, CA, US), StataSE 13.0 (StataCorp LP, TX, US) and R software [23]. Statistical significance was set at a p value below 0.05. All tests were two-tailed. Imputations were not done for missing data. Nominal and ordinal data were expressed as percentages. Interval data were expressed as median with lower and upper quartiles. The distribution of variables was evaluated by the W Shapiro-Wilk test and the quantile-quantile (Q-Q) plot. In order to compare two groups with SHBG ≥ 26.1 nmol/L and SHBG < 26.1 nmol/L, the t-Student test for independent data or the U Mann-Whitney test was used, according to data distribution. The homogeneity of variances was assessed by the F Fisher-Snedecor test. The nominal and ordinal data were compared with the χ2 test. Correlation between SHGB levels and other variables was assessed with the ρ Spearman rank correlation coefficient. Age adjustment was done with the Spearman rank partial correlation coefficient (package ppcor in R). In order to find a cut-off point discriminating the insulin resistance based on the SHBG level, parametric and non-parametric receiver-operating characteristic (ROC) curves were calculated with an area under the curve (AUC) and corresponding sensitivity, specificity, positive and negative predictive value as well as with accuracy of classification. In order to find an optimal, empirical cut-off point value for HOMA-IR, the Youden J statistic (index) was used. ## Results Study groups' characteristics' are listed in Table 1. Circulating SHBG levels below the reference lower limit of 26.1 nmol/L were found in $25.4\%$ of study participants. This subgroup was characterized by a significantly higher BMI, fasting glucose and insulin concentrations as well HOMA-IR values. Obesity and impaired fasting glucose (IGF) were more frequently diagnosed in a subgroup with SHBG below 26.1 nmol/L ($59.1\%$ vs. $18.6\%$; $p \leq 0.001$ and $17.2\%$ vs. $6.7\%$; $p \leq 0.001$, respectively). As expected, the median HOMA-IR value was significantly higher in a subgroup with low SHBG levels (2.8 vs. 1.7; $p \leq 0.001$). Figure 1 shows the ROC curve of HOMA-IR and SHBG levels below the lower limit of the laboratory reference range (< 26.1 nmol/L). An empirical optimal cut-off, based on the Youden index, for HOMA-IR discriminating the insulin resistance, was ≥2.1 (Table 2). Subjects with HOMA-IR values below the established cut-off had a very low risk of having impaired fasting glucose (OR = 0.035; $95\%$ CI: 0.013–0.097; $p \leq 0.001$) and decreased SHBG level (OR = 0.19; $95\%$ CI: 0.13–0.27; $p \leq 0.001$) (Table 3). There was a moderate negative correlation between HOMA-IR values and SHBG levels (crude: ρ = −0.50; $p \leq 0.001$, age-adjusted: ρ = −0.45; $p \leq 0.001$), as well as positive with BMI values (crude: ρ = −0.53; $p \leq 0.001$, age-adjusted: ρ = 0.60; $p \leq 0.001$). ## Discussion To the best of our knowledge, this is the first study estimating the cut-off value for HOMA-IR discriminating the insulin resistance based on the SHBG level in women with PCOS. It is established that HOMA-IR is a better measure of hepatic than muscle insulin resistance. In turn, compensatory hyperinsulinemia inhibits SHBG synthesis in the liver. In our study, $25.4\%$ of women with PCOS had circulating SHBG levels below the adopted lower limit of the laboratory reference range (26.1 nmol/L). This subgroup was characterized by a significantly more frequent occurrence of overweight and obesity diagnosed based on BMI values, according to the Word Health *Organization criteria* [24], compared to the subgroup with normal SHBG levels. As expected, impaired fasting glucose was also significantly more prevalent in this subgroup, corresponding to a significantly higher median HOMA-IR value (2.9 vs. 1.7). These results, as well as the negative correlations between SHBG levels and HOMA-IR values or insulin levels, once again confirm that low SHBG levels are associated with the occurrence of insulin resistance. These correlations indicate that hyperinsulinemia and insulin resistance explain nearly $50\%$ variability of SHBG concentrations. It is consistent with the results of a previous study analyzing the correlation between SHBG levels and insulin resistance in postmenopausal women [4]. Among factors not included in our analysis was hyperandrogenemia exerting a suppressive effect on SHBG secretion, mostly in men [2, 3]. However, a meta-analysis of 26 studies including 3,349 menopausal women showed that testosterone but not DHEA administration decreased SHBG levels [25]. Thus, hyperandrogenemia potentially may modulate the associations between SHBG levels and hyperinsulinemia also in women with PCOS. However, estradiol/testosterone and estradiol/androstenedione indexes are quite similar in both women with PCOS and obesity and women with PCOS and normal-weight [26]. Moreover, 12 months therapy with estrogens, which certainly affects the androgens/estrogens index, did not cause changes in insulin sensitivity in women with PCOS [27]. These data suggest that at least the androgens/estrogens ratio has a much less important role than the changes in BMI/fat depot in the modulation of insulin resistance. In our study, the empirically estimated HOMA-IR cut-off point discriminating the insulin resistance based on the SHBG level below the lower limit of the laboratory reference range (< 26.1 nmol/L) was 2.1. Thus, it is between the previously adopted cut-off points > 2.5 [28], > 2.0 [20, 21] and 1.67 [29]. Of note, the HOMA-IR cut-off point determined in our study was characterized by quite high sensitivity but low specificity. Therefore, in many cases the low SHBG level would not allow for the diagnosis of insulin resistance but, on the other hand, the likelihood of false positive results is low. Therefore we do not recommend using SHBG level to diagnose insulin resistance. However, it should be noted that in our subgroup with SHBG levels below 26.1 mmol/L, the prevalence of impaired fasting plasma glucose was about three times more frequent than in a subgroup with SHBG 26.1 mmol/L and above. Of note, the established HOMA-IR cut-off point in our study of 2.1 is very close to the value of 2.0 in Thai women with PCOS [20]. This discrepancy indicates a tightening circle in the search for the optimal HOMA-IR cut-off point for diagnosis of insulin resistance in the population of young women with PCOS. In our study, subjects with HOMA-IR values below the established here cut-off value had a very low risk of impaired fasting glucose. These results are in accordance with a previously published study [17] suggesting that our HOMA-IR cut-off point is a good marker of hepatic insulin resistance. Of note, the cut-off point of 2.1 established in our study is similar to the value determined in 833 Chinese women diagnosed with PCOS and components of metabolic syndrome [30]. In addition, the median SHBG concentration in this cohort was 27.9 nmol/L (lower quartile 18.8 nmol/L, upper quartile 45.5 nmol/L) [30], so it was close to the lower limit of the laboratory reference range used in our study. There are some confounders that should be considered when analyzing HOMA-IR values and corresponding cut-off points discriminating the insulin resistance based on the SHBG level. Borai et al. [ 31] indicated that studies determining the cut-off points for insulin resistance indicators should refer to the method of insulin assessment, because its concentrations may significantly differ depending on the type of used kit. This may be the effect of several factors, such as variable specificity, different calibration settings, and different formulas used to convert insulin units, as demonstrated by a comparison of 11 insulin determination methods by Manley et al. [ 32]. The same authors observed that the distribution of HOMA-IR values differed even twice, depending on the method of insulin assessment [33]. This fact can significantly affect the HOMA-IR cut-off point value estimated in different studies. The results of our and other studies cause reflection or the use of only one parameter in the assessment of insulin resistance with no precisely defined cut-off point, which is associated with a high risk of not recognizing this disturbance. As mentioned above, HOMA-IR calculation is highly variable; therefore, requiring a wider analysis of insulin resistance based on various indicators, perhaps including SHBG. This approach is also recommended by the authors of a study analyzing the advantages and disadvantages of various methods of insulin resistance assessment [33]. Our study has several limitations. The main limitation is its retrospective design. It also lacks hyperinsulinemic-euglycemic clamp, oral glucose tolerance test (OGTT), and HbA1c assessments, as well as body composition and visceral obesity (waist circumference) and fatty liver measures. However, the hyperinsulinemic-euglycemic clamp is still missing the reference values and, therefore, should not be used for the identification of subjects with hepatic insulin resistance. Moreover, both the hyperinsulinemic-euglycemic clamp and OGTT better characterize muscle insulin resistance, while HOMA-IR better assesses hepatic insulin resistance, which was the aim of our study [34]. Another limitation is not taking into account hyperandrogenemia as a factor influencing SHBG synthesis. However, it has been previously shown that the contribution of SHBG to the variation in HOMA-IR is not dependent on estrogen and androgens levels in postmenopausal women [35]. We hypothesize that this observation may also apply to premenopausal women, as recently published data show the similar predictive significance of SHBG levels for the development of insulin resistance in pre- and postmenopausal women [36]. The strength of our study relies on the large size of the study group and the inclusion of a homogenous cohort of young Caucasian women (between 20 and 30 years of age) with PCOS and a wide range of BMI. Of note, the established cut-off point for HOMA-IR may not be universal for all methods of insulin assessment. We think that the established here cut-off value for HOMA-IR, based on SHBG decline, could be useful for clinicians to identify women with PCOS that may benefit from the implementation of interventions such as an increase in physical activity and changes in eating habits to decrease visceral and liver fat accumulation and prevent the development of type 2 diabetes and cardiovascular disease. ## Conclusions Our study suggests that the cut-off point for HOMA-IR discriminating the insulin resistance based on the SHBG level in young Caucasian women with PCOS is 2.1 and is consistent with the cut-off value adopted by the European Group for the Study of Insulin Resistance (above 2.0). ## Data availability statement The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. ## Ethics statement Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the patients/participants or patients/participants legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements. ## Author contributions Concept and study design: AB-B, JC, and MO-G. Data collection: PK and PM. Analysis: AO and PC. Data interpretation and final approval and review: PM, MP-K, JC, and MO-G. Manuscript writing: AB-B and LM. All authors contributed to the article and approved the submitted version. ## Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ## Publisher's note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. ## References 1. Ding EL, Song Y, Malik VS, Liu S. **Sex differences of endogenous sex hormones and risk of type 2 diabetes: a systematic review and meta-analysis**. *JAMA.* (2006) **295** 1288-99. DOI: 10.1001/jama.295.11.1288 2. Rosner W, Hryb DJ, Khan MS, Nakla AM, Romas NA. **Androgens, estrogens, and second messengers**. *Steroids.* (1998) **63** 278-81. DOI: 10.1016/S0039-128X(98)00017-8 3. Rosner W, Hryb DJ, Kahn SM, Nakla AM, Romas NA. **Interactions of sex hormone-binding globulin with target cells**. *Mol Cell Endocrinol.* (2010) **316** 79-85. DOI: 10.1016/j.mce.2009.08.009 4. Akin F, Bastemir M, Alkiş E, Kaptanoglu B. **SHBG levels correlate with insulin resistance in postmenopausal women**. *Eur J Intern Med.* (2009) **20** 162-7. DOI: 10.1016/j.ejim.2007.09.023 5. Carreau AM, Pyle L, Garcia-Reyes Y, Rahat H, Vigeres T, Jensen T. **Clinical prediction score of nonalcoholic fatty liver disease in adolescent girls with polycystic ovary syndrome (PCOS-HS index)**. *Clin Endocrinol (Oxf).* (2019) **91** 544-52. DOI: 10.1111/cen.14062 6. Sarkar M, VanWagner LB, Terry JG, Carr JJ, Rinella M, Schreiner PJ. **Sex hormone-binding globulin levels in young men are associated with nonalcoholic fatty liver disease in midlife**. *Am J Gastroenterol.* (2019) **114** 758-63. DOI: 10.14309/ajg.0000000000000138 7. Seo IH, Lee HB, Kim S, Lee YJ, Jung DH. **Inverse relationship between hepatic steatosis and alanine aminotransferase with sex hormone-binding globulin in men**. *Yonsei Med J.* (2017) **58** 731-6. DOI: 10.3349/ymj.2017.58.4.731 8. Ye J, Yao Z, Tan A, Gao Y, Chen Y, Lin X. **Low serum sex hormone-binding globulin associated with insulin resistance in men with nonalcoholic fatty liver disease**. *Horm Metab Res.* (2017) **49** 359-64. DOI: 10.1055/s-0043-102690 9. Luo J, Chen Q, Shen T, Wang X, Fang W, Wu X. **Association of sex hormone-binding globulin with nonalcoholic fatty liver disease in Chinese adults**. *Nutr Metab (Lond).* (2018) **15** 79. DOI: 10.1186/s12986-018-0313-8 10. Jaruvongvanich V, Sanguankeo A, Riangwiwat T, Upala S. **Testosterone, sex hormone-binding globulin and nonalcoholic fatty liver disease: a systematic review and meta-analysis**. *Ann Hepatol.* (2017) **16** 382-94. DOI: 10.5604/01.3001.0009.8593 11. Zhu H, Wang N, Han B, Li Q, Chen Y, Zhu C. **Low sex hormone-binding globulin levels associate with prediabetes in Chinese men independent of total testosterone**. *PLoS ONE.* (2016) **11** e0162004. DOI: 10.1371/journal.pone.0162004 12. Hu J, Zhang A, Yang S, Wang Y, Goswami R, Zhu H. **Combined effects of sex hormone-binding globulin and sex hormones on risk of incident type 2 diabetes**. *J Diabetes.* (2016) **8** 508-15. DOI: 10.1111/1753-0407.12322 13. Muka T, Nano J, Jaspers L, Bramer WM, Hofman A, Dehghan A. **Associations of steroid sex hormones and sex hormone-binding globulin with the risk of type 2 diabetes in women: a population-based cohort study and meta-analysis**. *Diabetes.* (2017) **66** 577-86. DOI: 10.2337/db16-0473 14. O'Reilly MW, Glisic M, Kumarendran B, Subramanian A, Manolopoulos KN, Tahrani GN. **Serum testosterone, sex hormone-binding globulin and sex-specific risk of incident type 2 diabetes in a retrospective primary care cohort**. *Clin Endocrinol (Oxf).* (2019) **90** 145-54. DOI: 10.1111/cen.13862 15. Feng C, Jin Z, Sun L, Wang X, Chi X, Zhang X. **Endogenous SHBG levels correlate with that of glucose transporters in insulin resistance model cells**. *Mol Biol Rep.* (2019) **46** 4953-65. DOI: 10.1007/s11033-019-04946-w 16. Yamazaki H, Kushiyama A, Sakoda H, Fujishiro M, Yamamotoya T, Nakatsu Y. **Protective effect of sex hormone-binding globulin against metabolic syndrome: in vitro evidence showing anti-inflammatory and lipolytic effects on adipocytes and macrophages**. *Mediators Inflamm.* (2018) **2018** 3062319. DOI: 10.1155/2018/3062319 17. Tripathy D, Almgren P, Tuomi T, Groop L. **Contribution of insulin-stimulated glucose uptake and basal hepatic insulin sensitivity to surrogate measures of insulin sensitivity**. *Diabetes Care.* (2004) **27** 2204-10. DOI: 10.2337/diacare.27.9.2204 18. Matthews DR, Hosker JP, Rudenski AS, Naylor DR, Treacher DF, Turner RC. **Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man**. *Diabetologia.* (1985) **28** 412-9. DOI: 10.1007/BF00280883 19. Jensterle M, Weber M, Pfeifer M, Prezelj J, Pfutzner A, Janez A. **Assessment of insulin resistance in young women with polycystic ovary syndrome**. *Int J Gynaecol Obstet.* (2008) **102** 137-40. DOI: 10.1016/j.ijgo.2008.03.017 20. Wongwananuruk T, Rattanachaiyanont M, Leerasiri P, Indhavivadhana S, Techatraisak K, Angsuwathana S. **The usefulness of Homeostatic Measurement Assessment-Insulin Resistance (HOMA-IR) for detection of glucose intolerance in Thai women of reproductive age with Polycystic Ovary Syndrome**. *Int J Endocrinol.* (2012) **2012** 571035. DOI: 10.1155/2012/571035 21. Balkau B, Charles MA. **Comment on the provisional report from the WHO consultation. European Group for the Study of Insulin Resistance (EGIR)**. *Diab Med.* (1999) **16** 442-43. DOI: 10.1046/j.1464-5491.1999.00059.x 22. **Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome (PCOS)**. *Hum Reprod* (2004) **19** 41-7. DOI: 10.1093/humrep/deh098 23. 23.R Core Team. R: A Language Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing (2013). Available online at: http://www.R-project.org/~. *R: A Language Environment for Statistical Computing.* (2013) 24. 24.WHO. Obesity. Preventing and managing the global epidemic. Report of a WHO. Geneva: WHO (1998).. *Obesity. Preventing and managing the global epidemic. Report of a WHO* (1998) 25. Marina L, Sojat AS, Maseroli E, Spaggiari G, Pandurevic S, Santi D. **Hormonal profile of menopausal women receiving androgen replacement therapy: a meta-analysis**. *J Endocrinol Invest.* (2020) **43** 717-35. DOI: 10.1007/s40618-020-01192-x 26. Franik G, Maksym M, Owczarek AJ, Chudek J, Madej P, Olszanecka-Glinianowicz M. **Estradiol/testosterone and estradiol/androstenedione indexes and nutritional status in PCOS women—a pilot study**. *Eur J Obstet Gynecol Reprod Biol.* (2019) **242** 166-9. DOI: 10.1016/j.ejogrb.2019.05.045 27. Bhattacharya SM, Ghosh M, Basu R. **Effects of ethinyl estradiol and desogestrel on clinical and metabolic parameters in Indian patients with polycystic ovary syndrome**. *J Obstet Gynaecol Res.* (2012) **38** 285-90. DOI: 10.1111/j.1447-0756.2011.01682.x 28. Farias de Azevedo Salgado AL, de Carvalho L, Oliveira AC, dos Santos VN, Vieira JG, Parise ER. **Insulin resistance index (HOMA-IR) in the differentiation of patients with non-alcoholic fatty liver disease and healthy individuals**. *Arq Gastroenterol* (2010) **47** 165-9. DOI: 10.1590/S0004-28032010000200009 29. Conus F, Rabasa-Lhoret R, Péronnet F. **Characteristics of metabolically obese normal-weight (MONW) subjects**. *Appl Physiol Nutr Metab.* (2007) **32** 4-12. DOI: 10.1139/h06-092 30. Li R, Yu G, Yang D, Li S, Lu S, Wu X. **Prevalence and predictors of metabolic abnormalities in Chinese women with PCOS: a cross- sectional study**. *BMC Endocr Disord.* (2014) **14** 76. DOI: 10.1186/1472-6823-14-76 31. Borai A, Livingstone C, Kaddam I, Ferns G. **Selection of the appropriate method for the assessment of insulin resistance**. *BMC Med Res Methodol.* (2011) **11** 158. DOI: 10.1186/1471-2288-11-158 32. Manley SE, Stratton IM, Clark PM, Luzio SD. **Comparison of 11 human insulin assays: implications for clinical investigation and research**. *Clin Chem.* (2007) **53** 922-32. DOI: 10.1373/clinchem.2006.077784 33. Manley SE, Luzio SD, Stratton IM, Wallace TM, Clark PMS. **Preanalytical, analytical, and computational factors affect homeostasis model assessment estimates**. *Diabetes Care.* (2008) **31** 1877-83. DOI: 10.2337/dc08-0097 34. Hoffman RP. **Indices of insulin action calculated from fasting glucose and insulin reflect hepatic, not peripheral, insulin sensitivity in African-American and Caucasian adolescents**. *Pediatr Diab* (2008) **9** 57-61. DOI: 10.1111/j.1399-5448.2007.00350.x 35. Davis SR, Robinson PJ, Moufarege A, Bell RJ. **The contribution of SHBG to the variation in HOMA-IR is not dependent on endogenous oestrogen or androgen levels in postmenopausal women**. *Clin Endocrinol (Oxf).* (2012) **77** 541-7. DOI: 10.1111/j.1365-2265.2011.04301.x 36. Ottarsdottir K, Hellgren M, Bock D, Nilsson AG, Daka B. **Longitudinal associations between sex hormone-binding globulin and insulin resistance**. *Endocr Connect.* (2020) **9** 418-25. DOI: 10.1530/EC-20-0141
--- title: “There Is Method to This Madness” A Qualitative Investigation of Home Medication Management by Older Adults authors: - Olajide Fadare - Matthew Witry journal: Pharmacy year: 2023 pmcid: PMC10037564 doi: 10.3390/pharmacy11020042 license: CC BY 4.0 --- # “There Is Method to This Madness” A Qualitative Investigation of Home Medication Management by Older Adults ## Abstract Objectives: This paper explores [1] the systems and processes older adults use to manage medications at home, and [2] the well-being goals of personal interest that motivate them. Methods: Qualitative interviews were conducted in the homes of 12 older adults in a small city in the Midwest United States. Interviews were analyzed using inductive template analysis. Results: The average age of older adults in this study was 74.2 years (SD = 10.5), $66.7\%$ were women. The most prominent home medication management tools used were pill boxes, containers and vials, and medication lists. Routines were often aligned with activities of daily living such as teeth brushing and eating. Their medication management work occurred in contexts of other household members and budget constraints. Routines and practices were sometimes idiosyncratic adaptations and supported goals of maintaining control and decreasing vulnerability. Conclusion: In developing routines for home medication management, older adults developed systems and deliberate processes to make sense of their medication experiences in the context of their home environment and based on available resources. ## 1. Introduction Polypharmacy (taking multiple medications) to treat chronic conditions is more common as adults age [1]. On average, older adults take 4–5 medications at a time, mostly for cardiovascular diseases, depression, and pain [2,3]. While medications can have therapeutic benefits, polypharmacy is associated with a significant amount of morbidity, mortality [4,5], and potentially avoidable health care utilization [6,7,8]. For example, a five-year longitudinal study by Chang et al. [ 2020] observed over 2 million ($67\%$) polypharmacy-related hospitalizations among the sample of older adults studied [9]. Polypharmacy also increases the risk of accidental and intentional drug poisoning [1,10]. While medications are a mainstay of medical treatment for older adults with chronic conditions, when someone goes home from the hospital, clinic, and pharmacy, they are largely left to develop their own routine and system for medication management and organization. As such, decisions, practices, adaptations, and supplementations that shape patients’ home medication use and management often go unreported and undocumented [11,12,13]. Although doctors, pharmacists, and other health care professionals can have a positive impact on people taking their medications as prescribed, ultimately it is up to the patient to make decisions about what medications to take (or not), when, and if they modify or supplement their regimen. Moreover, prescriber and pharmacist guidance on communicating with patients about medicines typically includes inquiring about adherence but often does not include asking patients about their medication management and medication-taking routine, even though there is evidence that a patient’s ability to articulate their medication-taking routine has a positive association with adherence [14]. Some studies have examined the medication management systems of people in their homes using a work-system safety paradigm. The work-system safety paradigm conceptualizes the residential home as a system within which patients carry out the self-healthcare task of medication management [15,16]. For instance, Holden et al. [ 2017], using a human factors approach, conceptualized home medication management as patient work that is executed within a patient work system. These authors conceptualized the patient work system as comprising a microdomain of the patient interacting with tools and technology to carry out medication management tasks within the home, nested within a macrodomain of social, physical, and environmental factors [17]. According to Doucette and colleagues [2017], the systems approach to home medication management (SAHMM) model, the patient’s work system, comprises six components: (a) the patient, (b) medication management tasks, (c) medication management tools and technology, (d) the internal layout of the patient’s home, (e) availability and access to healthcare providers within the area, and (f) the social context or living arrangement within the home, that is, whether the patient lives with family and caregivers, or lives alone. These six components interact to shape the home medication management processes that influence patient outcomes [18]. Other studies have also suggested that patient outcomes such as medication adherence and medication safety are influenced by factors beyond the perimeters of conventional healthcare systems such as hospitals and pharmacies, and many of these influencing factors pertain to the strategies that patients use to manage their medications at home [15,16,18,19,20,21]. Strategies for home medication management include medication sorting and organization [22], establishing routines and cues for medication-taking such as associating it with teeth brushing and other activities of daily living [23], and requesting prescription refills [16]. Available evidence suggests that these strategies may facilitate patient adherence to treatment depending on their respective implementation contexts [24]. However, understanding why these strategies work, and the underlying cognitions that inform how patients perform medication management at home, is lacking. Further, as medication adherence and safety are goals that are typically defined and framed from the perspective of healthcare professionals, we propose that patients also have personally important well-being goals that drive patients’ engagement in home medication management. Exploring the lived experiences of community-dwelling medication users may offer new insights into their home medication management processes and cognitions, and the well-being goals of personal interest that motivate them. Such insights, together with other studies, may help primary care providers, pharmacists, and other stakeholders identify and develop potential service, product, and system targets for improving the quality and safety associated with using medications at home. The objectives of this study were to use the systems approach to home medication management model (SAHMM) (Figure 1) to [1] describe the home medication management system and processes for a sample of community-dwelling older adults, and [2] to elucidate the well-being goals of personal interest that motivate their engagement in home medication management. ## 2. Materials and Methods This qualitative analysis was part of an intervention study of a medication management questionnaire designed to help pharmacists make recommendations to patients about home medication use. The study was approved by the University of Iowa IRB. Participants were patients of a local independently-owned community pharmacy in a small city in the Midwest US. The pharmacy staff offered participation in the study during May and June of 2018 to patrons who were 55 years or older and took four or more medicines. The pharmacy primarily recruited during less-busy times when the study could be explained and the questionnaire could be administered, thus, convenience sampling. The rationale for these criteria was to target persons expected to have developed routines for taking their chronic medications who were either nearing retirement or retired but still managing their own medicines. Patients completing a questionnaire could mark if they were interested in being contacted for an in-home interview about how they manage their medications at home. Interviews were expected to last about 60 min and participants were paid $50 for their participation in the interview. Twenty-four patients agreed to participate in the questionnaire study and of these, 12 also agreed to be interviewed. Informed consent was obtained from each participant, including permission to share photographs of their medication organization system in published research reports. Interviewing 12 participants generally is sufficient for saturation for studies in health services research [25]. Interviews took place in participants’ homes and primarily occurred around the kitchen table, although participants also would lead the investigator around the home to show the various places where participants engaged in medication-related work such as medication storage. Most interviews were one-on-one, with a small number also including brief contributions from participants’ spouses or caregivers. Interviews were conducted by a Ph.D. investigator with training in qualitative research methods and 10 years of qualitative research experience. The investigator’s positionality included being a health services researcher and educator with expertise in patient-centered communication and medication adherence behaviors. The interview guide is included in Appendix A. *The* general interview topics included the medication management approach for a typical day, routines and cues, organizers, storage, difficulties, tools used, lists, and approach to refills. Other topics discussed were perspectives on physician and pharmacist roles in medication management. Participants consented to photos being taken of key medication organization setups, and names and other identifying information were blurred for privacy. Interviews were audio-recorded and edited to remove any identifying information, and transcribed by an online transcription service. Data were managed using MAXQDA v2018, Verbi Berlin. ## Data Analysis Interviews were de-identified and analyzed after all the interviews were conducted and transcribed. Transcripts were analyzed using template analysis for descriptive coding based on the SAHMM complemented with inductive qualitative analysis to capture insights into the responses that were beyond the descriptive codes of the template [26]. In this process of descriptive coding, the researchers sought to present the facts of observation as seen through the eyes of the study participants. This was carried out by retaining the study participants’ own words through data analysis and reporting, as evidenced by the representative quotes highlighted in the results section. The SAHMM was used as a template to provide initial code categories for data vignettes. Areas of congruence of discrepancy between the theoretical propositions of the SAHMM and the study data were explored during data analysis. The two authors worked iteratively and collaboratively to apply several rounds of descriptive coding using a constant comparison approach of reviewing and coding the transcripts independently, getting together to reconcile codes and themes, reviewing the emergent themes against the relevant literature, and going back to the transcripts. Themes were identified across transcripts based on the conceptualization by Boeije [2020] of being short interpretive statements that capture salient ideas that weave across subjects and dialogue [27]. Representative quotes were identified to demonstrate the voices of a variety of participants and their experiences. To support the trustworthiness of the analysis, the authors offer a positionality statement, used a conceptual model for template analysis, engaged in frequent discussions about interpreting transcripts, coding, and themes, and used pictures and representative quotes to support the overall narrative of the analysis. ## 3. Results Twelve older adults were interviewed in their homes. Nine were women and the sample had an average age of 74.2 years (SD = 10.5) (Table 1). The average number of medications used was 7.75 (SD = 3.5). Most medication storage and manipulation occurred in the kitchen, with some participants also storing medications in the bedroom and bathroom. Most participants also used various types of 7-day organizers. Two participants took their medication doses directly from the prescription vials. Template analysis of the study transcripts revealed three components for the work system of older adults engaging in home medication management. These were Person, Tools and Technology, and Household. Two dimensions of home medication management process—patient work and collaborative work—were identified. Two themes—maintaining a sense of control and avoiding vulnerability—were used to describe the personally important well-being goals of older adults managing their medications at home. A theme—“there is method to this madness”—was used to describe the cognitions that underpin home medication management. The latter theme emerged from the inductive analysis of the data as informed by the literature related to cognition and patient work. Themes and representative quotes are summarized in Table 2. ## 3.1. The Home Medication Management Work System Three components of the patient’s home medication management work system emerged from the qualitative interviews and were described using the SAHMM model. The three work-system components were: Person, Tools and Technologies, and Household. ## 3.1.1. Person The Person component of the home medication management work system refers to the individual who is responsible for medication management in the home. Most study participants reported living with a spouse and considered their health to be satisfactory. The older adults in this study were generally able to perform routine activities with little to no assistance and were actively involved in their home medication management, although some spouses assisted. Two characteristics were particularly relevant from the interview data under the Person domain—being a long-time medication user and admitting to occasional memory or forgetfulness issues (Table 2). ## 3.1.2. Tools and Technology The most prominent tools used by these older adults for home medication management were pill boxes and containers (Figure 2), medication lists, and pill cutters. The participants in the present study used a variety of pill boxes based on a 7-day compartment design, some used 7-day boxes with several dose containers per day (e.g., morning and evening) or multiples of the 7-day organizers. Some also used additional travel containers or multi-compartment vials for extra supplies. Some repurposed small dishes to set out the next dose or a special dose, and one participant’s spouse repurposed 2 days of a 7-day organizer for morning and evening supplements which were refilled daily. A common practice was to write on the containers with permanent markers, although these tended to fade or wear out over time. In addition to pill boxes, these adults often devised their own approach to creating and managing a medication list. Information managed with medication lists varied but included descriptions of the medications and doses, and some lists contained additional information such as pharmacy information, medical conditions, and allergies. Pill cutters were used to divide tablets in half to aid in swallowing Older adults used technology mostly for finding out things such as medication identity and side effects. Technologies used were the internet, computers and word processors, and stationery. ## 3.1.3. Household The household component of the home medication management work system refers to the physical, social, and economic attributes of the home that influence medication management. For the older adults in this study, these household attributes were physical locations and spaces within the home for medication storage, other household members (e.g., spouses, children, and siblings), and cost (economic) concerns. Medication storage spaces were distributed across multiple locations within the home and commonly included the kitchen, where many older adults filled their weekly pill boxes, stored medications in cupboards, and took their medications from their organizers; the bathroom, where many stored extra medication or as-needed medications in a medicine cabinet; and the bedroom, where some used drawers to store their medications or stored a large plastic tote with their extra medications and supplies). See Figure 3. It also was common to intermix medications with other household or personal items, whether next to glassware in the kitchen or a dresser drawer alongside personal items, sometimes the interviewee intentionally created a dedicated storage space, but that was not the norm. [ storage spaces]. Another finding under household was the role of other household members and how they influenced patient work of medication management. Family members often interfered in the patient’s home medication management by encouraging, but occasionally disrupting patient routines. Last under household were concerns about the cost of medications given the impact of cost on a household budget. This included medication co-pay amounts, out-of-pocket costs for provider visits, and the cost of procuring additional pharmacy services to aid home medication management. Cost concerns were sometimes substantial enough to make older adults anxious about maintaining their medication supply long term, although none admitted to an acutely critical situation, one did mention the hypothetical of choosing between paying for their medications or other essential household expenses. Moreover, purchasing supplemental pharmacy services to aid home medication management appeared to be limited by cost. ## 3.2. Home Medication Management Processes Home medication management processes refer to how patients managing medications at home allocate resources and effort toward the achievement of desired health outcomes. Home medication management processes emerge from patients considering all the components of the work system simultaneously, rather than as independent factors, in their judgment about how ‘best’ to manage their medications at home. Consistent with the SAHMM model, two dimensions of home medication management processes—patient work and collaborative work—were observed in this study. ## 3.2.1. Patient Work For this sample of older adults, patient work included several common activities such as procuring medications, organizing medications and filling pill boxes, taking medications, and monitoring their medications. Many of these activities were based on a routine they had established. For medication procurement, participants typically called the pharmacy or signed up for automatic refills. Participant 8W said “when I get down to five tablets, I call [the pharmacy] to refill”, while participant 4W said, “[the pharmacy] keeps me well supplied [medication procurement]”. Participant 7M used a combination of manual and automatic refills; “first of all, [the pharmacy] has got me on an automatic renewal… if it’s going to be a weekend or holiday, then I’ll check my medications especially the [insulin] and if it looks like I will run out, I’ll call [the pharmacy] and they’ll get me fixed up.” 7M [procurement] The sorting and filling of medication boxes was a prominent activity that built on their choice of tools, most commonly, some version of the 7-day pill organizer, but sometimes a less conventional approach. Some older adults also repacked their medications into other containers. See Figure 4. Once these adults had procured and organized their medications, they then move into their established routine of medication-taking. The routine for taking their medication often was aligned with activities of daily living such as teeth brushing and eating [routine cue], with some adjustments needing to be made for special circumstances such as recreation and travel [routine disruption]. Lastly, these individuals also monitored medications, especially for side effects, and to detect potential lack of fit between dosage recommendations and activities of daily living. They may then act on this information by changing how they take the medication or by seeking medical advice. Several participants also were savvy in verifying the name and physical characteristics (e.g., shape and color) of medications. ## 3.2.2. Collaborative Work Patients’ medication management in the home often involves collaborations between the patient and healthcare professionals. These relationships allow health care professionals to assist patients in sensemaking about their medications and conditions and monitoring their treatment regimens. The professionals most frequently consulted by the participants in this study were physicians and pharmacists. Overall, most of the participants had positive relationships with their physicians. Several participants expressed confidence in their physicians’ orders and adherence to their prescribed regimens. Some patients expressed a preference to be on fewer medications and appreciated efforts to deprescribe. Like interviewees’ positive feelings for their personal physicians, study participants who were also patrons of the participating independently owned pharmacy had developed valuable personal relationships with the pharmacy staff. Trust and familiarity allowed study participants to feel comfortable collaborating with the pharmacy for medication management in an ongoing fashion. Personable communication from the pharmacy staff also fostered patient–HCP collaboration. Overall, the generally positive experiences and perspectives of study participants toward their doctors and pharmacists likely served as a relevant backdrop for how older adults take their medications at home, and the systems and routines they have created for their medication management. ## 3.3. Outcomes: Personally Important Well-Being Goals of Home Medication Management The perspectives and behaviors of the older adults in this study suggest their medication management practices appear to be in pursuit of personally important well-being goals as described by two themes—maintaining a sense of control and avoiding vulnerability. This desire for control surrounds many of their decisions and routines related to how these older adults used and managed their medications and interacted with their health care providers. Patients also engaged in medication management as a way to avoid feeling vulnerable. Avoiding vulnerability informed patients’ adherence to treatment as medication-taking was viewed as a way to control events and avoid hospital visits, emergencies, or general downturns in health. ## 3.4. Interpretation across the SAHMM Domains: “There Is Method to This Madness” Older adults doing the work of home medication management did not necessarily follow a simple prescriptive set of activities. Rather, patients developed their own home medication management practices based on an appraisal of the resources that were available and accessible to them, and how these resources intersected with well-being goals that were important to them. Home medication management practices that may appear random, haphazard, and sometimes counterintuitive [madness], were clever adaptations of common structural elements of the work system described above in ways that allowed each patient to achieve their desired medication management outcomes [method]. Previous research suggests that patients engaging in home medication management use cognitions to understand their medication experiences, and to form connections between the attributes of their work systems, their activities of daily living, and their health. Given the available resources and their health and medication experiences, older adults used cognitions such as sensemaking to develop home medication management routines and to inform collaborative work with their primary care providers. This included making connections between past medication experiences and a current situation and considering the potential outcomes of different medication use actions. This relationship between medication experiences and outcomes informed how older adults developed their routines and adapted tools and technology to perform medication management tasks. ## 4. Discussion The study findings show that home medication management is a complex work, with deliberate, although sometimes idiosyncratic, aspects. These routines appear to be performed by patients motivated by well-being goals of maintaining control and decreasing vulnerability. The home is a medication management workplace where patients use various tools and technologies, combined with household social and economic considerations, to integrate medications into daily life. When community-dwelling older adults return to their homes with medications from the clinic or pharmacy, they assume responsibility for managing their medications and taking them as prescribed. However, this duty is fulfilled differently by every individual and there were few consistent approaches. This is consistent with other studies considering the complex sociotechnical system that shapes the lived experience of its residents [20,28]. Here, older adults managing medications at home worked to achieve a balance between the household contexts of other family members, social support, cost considerations, and others and the tools and technical aspects of medication management [16,18,28]. This balancing act may involve actions such as reorganizing the physical layout of places within the home (e.g., the kitchen, dining area, or bathroom medicine cabinet to allocate space for medication storage, redesigning routines of daily living to accommodate medication use, associating medication use with specific activities of daily living, and retrieving and organizing medication-related information, and others [16,18,23]. Sometimes this may include potentially unsafe practices such as putting medications and vitamins into dishes or other containers. These home medication management activities are not performed arbitrarily, rather, they appear to be conducted intentionally to integrate their medications into their home life, while taking into consideration the resources, constraints, preferences, and goals of home medication management. As shown by the method-to-madness theme, older adults were actively trying to figure out medication management systems that work. In doing this, patients appeared to use deliberate, cognitive processes such as sensemaking to understand and form connections between their medication experiences, the home, and their health to guide medication use. Older adults’ sensemaking of their medication experience at home appeared to be driven by their perceived outcomes of different medication use actions beyond clinically defined medication therapy goals. For instance, adherence to primary care provider recommendations when managing medications at home, unsupervised, may depend on whether the medication use action gives patients a sense of control or makes them feel less vulnerable concerning their health. As the psycho-emotional burden of managing chronic illnesses is associated with a sense of increased vulnerability and loss of control over one’s health and well-being, with negative consequences for patient health outcomes [29,30], it can be expected that patients will seek opportunities to engage in health behaviors that may restore a sense of control to them or allow them to minimize or avoid vulnerability. These health behaviors can include refusing to take certain medications or self-adjusting one’s dosage and dosing schedule. More so, patients tend to double down on taking their medications as prescribed, particularly when the consequences, real or perceived, of not taking their medications heighten their sense of vulnerability [31,32,33]. However, while self-adjustment of medication dosage may give the patient a sense of being in control or less vulnerable, it also raises the risk of harm when done without the professional guidance of health care providers. Nonetheless, having a sense of control and avoiding vulnerability appear to be well-being goals of personal importance that may influence how older adults manage their medication at home, and warrant more attention from primary care providers and researchers [29,31,33]. At the same time, several patients had a feeling of taking too many medications and engaged with their primary care provider to reduce the number of pills they were taking, which in some cases resulted in deprescribing, which those patients appreciated. Deprescribing is the systematic identification, reduction, or discontinuation of the number and dose of inappropriate medications by healthcare providers to manage polypharmacy, reduce the risk of medication-related adverse events, and improve clinical and well-being outcomes for the patient [34,35,36]. Some older adults acted on this perceived need for deprescribing by steering the prescriber–patient communication towards deprescribing, thus creating more opportunities for meaningful healthcare provider communication with patients. ## 4.1. Implications for Primary Care These study findings can inform health care stakeholders including family physicians, nurse practitioners, physician assistants, and pharmacists, that patients are not just passive recipients of healthcare, but are actively working as best able to manage their medications in pursuit of well-being, control, and decreasing their vulnerability associated with their medication burden [37,38]. This is underscored by older adults in this study who generally viewed themselves as ‘experts’ on home medication management [“I ask the pharmacist, “*Does this* interfere with anything else I’m doing?” So, I feel like I’m smart about taking [my] medicines.”] This ‘expert patient’ perception probably explains study participants’ hands-on approach to carrying out home medication management tasks and maintaining home medication management processes, especially when changes in the physiological measures of chronic disease management such as blood glucose and blood pressure prompt dosing changes that lead to alterations in-home medication management routines. Hence, providers who endeavor to engage patients in discussions about the fit between a treatment regimen and the patients’ routine of daily living, and whether the patient anticipates any difficulties in being able to effectively manage the prescribed medications at home may see greater patient buy-in. Such conversations can humanize the provider–patient relationship and may engender better patient adherence to treatment [38,39]. Further, study findings suggest that clinically defined medication therapy outcomes such as medication adherence, quality of life improvement, and avoiding adverse events are necessary but may not be sufficient to improve and maintain patients’ adherence to treatment. Making medication therapy more collaborative and focused on the outcomes of personal interest to the patient has been shown to be beneficial [37]. Patient–provider communication sessions provide ample opportunities for doing this. Beyond asking about a patient’s history of personal illness, primary care providers could increase their focus on elucidating the patient’s intrinsic motivations for engaging in home medication management and tailor their counseling accordingly [39]. Study participants also gave suggestions for human factors improvements to the design of home medication management tools and technologies. The tools and technologies that appeared to need the most improvement were medication lists and pill boxes. Most study participants used a medication list to organize information and aid in remembering. Medication lists were majorly prepared by the patients as they generally found similar artifacts (e.g., medication information sheets) received from the clinic or pharmacy confusing [“I don’t even look at the handouts. Sometimes I get screwed up. They don’t make sense.”]. Pharmacists could more routinely discuss medication lists with patients, and when possible, be involved in designing and maintaining comprehensive and up-to-date medication lists, as well-designed medications lists could facilitate prompt identification of risk of medication-related adverse events and prompt initiation of deprescribing, as may be necessary [40,41]. Patients also suggested using different colors for day and night compartments in the design of pill boxes. The concerns about medication lists and pillboxes suggest the need for more human factors considerations in designing medication management tools and technologies [42]. ## 4.2. Limitations This was a convenience sample of 12 older adults in a single Midwestern US town. While there appeared to be variation in income and education, all the study participants were white and used a single independently-owned pharmacy. Interviewing more adults, including those from other socioeconomic and demographic groups, may yield additional medication management practices, beliefs, and experiences. The present analysis may not have reached saturation of the diverse concept of medication management practices in the home but may serve as a start to further develop an understanding of this phenomenon. Additionally, the authors did not engage in member-checking. ## 4.3. Future Research In addition to purposefully interviewing more demographically and socioeconomically diverse participants, future work can test how best to collect information about home medication management practices and vulnerabilities, best practices for discussing medication management practices in an open and non-judgmental way, focused on finding solutions, and what role pharmacists home visits may have in supporting older adults with their home medication management. ## 5. Conclusions This sample of older adults had well-developed routines for medication management and use, although elements of their practices may put them at risk for medication misadventures. In developing routines for home medication management, older adults used cognitive processes to make sense of their medication experiences and available resources to guide medication use actions to avoid vulnerability and gain a sense of control in determining their well-being. ## References 1. da Silva A.F., de Paula Silva J.. **Polypharmacy, automedication, and the use of potentially inappropriate medications: Cause of intoxications in the elderly**. *Rev. Med. Minas Gerais* (2022.0) **32** e32101. DOI: 10.5935/2238-3182.2022e32101 2. Young E.H., Pan S., Yap A.G., Reveles K.R., Bhakta K.. **Polypharmacy prevalence in older adults seen in United States physician offices from 2009 to 2016**. *PLoS ONE* (2021.0) **16**. DOI: 10.1371/journal.pone.0255642 3. Charlesworth C.J., Smit E., Lee D.S.H., Alramadhan F., Odden M.C.. **Polypharmacy Among Adults Aged 65 Years and Older in the United States: 1988–2010**. *J. Gerontol. Ser. A* (2015.0) **70** 989-995. DOI: 10.1093/gerona/glv013 4. Schöttker B., Muhlack D.C., Hoppe L.K., Holleczek B., Brenner H.. **Updated analysis on polypharmacy and mortality from the ESTHER study**. *Eur. J. Clin. Pharmacol.* (2018.0) **74** 981-982. DOI: 10.1007/s00228-018-2445-1 5. Leelakanok N., Holcombe A.L., Lund B.C., Gu X., Schweizer M.L.. **Association between polypharmacy and death: A systematic review and meta-analysis**. *J. Am. Pharm. Assoc.* (2017.0) **57** 729-738.e10. DOI: 10.1016/j.japh.2017.06.002 6. Lu W.H., Wen Y.W., Chen L.K., Hsiao F.Y.. **Effect of polypharmacy, potentially inappropriate medications and anticholinergic burden on clinical out-comes: A retrospective cohort study**. *Can. Med. Assoc. J.* (2015.0) **187** E130. DOI: 10.1503/cmaj.141219 7. Sganga F., Landi F., Ruggiero C., Corsonello A., Vetrano D.L., Lattanzio F., Cherubini A., Bernabei R., Onder G.. **Polypharmacy and health outcomes among older adults discharged from hospital: Results from the CRIME study**. *Geriatr. Gerontol. Int.* (2015.0) **15** 141-146. DOI: 10.1111/ggi.12241 8. Payne R.A., Abel G., Avery A.J., Mercer S.W., Roland M.O.. **Is polypharmacy always hazardous? A retrospective cohort analysis using linked electronic health records from primary and secondary care**. *Br. J. Clin. Pharmacol.* (2014.0) **77** 1073-1082. DOI: 10.1111/bcp.12292 9. Chang T.I., Park H., Kim D.W., Jeon E.K., Rhee C.M., Kalantar-Zadeh K., Kang E.W., Kang S.-W., Han S.H.. **Polypharmacy, hospitalization, and mortality risk: A nationwide cohort study**. *Sci. Rep.* (2020.0) **10** 18964. DOI: 10.1038/s41598-020-75888-8 10. Lee E.H., Park J.O., Cho J.P., Lee C.A.. **Prioritising Risk Factors for Prescription Drug Poisoning Among Older Adults: A Multi-Method Study**. *Int. J. Environ. Res. Public Health* (2021.0) **18**. DOI: 10.3390/ijerph18115948 11. A Mohammed M., Moles R.J., Chen T.. **Medication-related burden and patients’ lived experience with medicine: A systematic review and metasynthesis of qualitative studies**. *BMJ Open* (2016.0) **6** e010035. DOI: 10.1136/bmjopen-2015-010035 12. Shoemaker S.J., Ramalho de Oliveira D.. **Understanding the meaning of medications for patients: The medication experience**. *Pharm. World Sci.* (2008.0) **30** 86-91. DOI: 10.1007/s11096-007-9148-5 13. Witry M., Parry R., McDonough R., Deninger M.. **Analysis of medication adherence-related notes from a service-oriented community pharmacy**. *Res. Soc. Adm. Pharm.* (2018.0) **14** 589-594. DOI: 10.1016/j.sapharm.2017.07.006 14. Phillips L.A., Duwe E.A.G.. **Prescribing Providers Estimate Patients’ Adherence to Hypertension and Type 2 Diabetes Medications from Patients’ Medication-Taking Routines: An Observational Study**. *J. Gen. Intern. Med.* (2019.0) **34** 1688-1690. DOI: 10.1007/s11606-019-05054-y 15. Holden R.J., Schubert C.C., Mickelson R.S.. **The patient work system: An analysis of self-care performance barriers among elderly heart failure patients and their informal caregivers**. *Appl. Ergon.* (2015.0) **47** 133-150. DOI: 10.1016/j.apergo.2014.09.009 16. Mickelson R.S., Holden R.J.. **Medication management strategies used by older adults with heart failure: A systems-based analysis**. *Eur. J. Cardiovasc. Nurs.* (2018.0) **17** 418-428. DOI: 10.1177/1474515117730704 17. Holden R.J., Valdez R.S., Schubert C.C., Thompson M.J., Hundt A.S.. **Macroergonomic factors in the patient work system: Examining the context of patients with chronic illness**. *Ergonomics* (2017.0) **60** 26-43. DOI: 10.1080/00140139.2016.1168529 18. Doucette W.R., Vinel S., Pennathur P.. **Initial development of the Systems Approach to Home Medication Management (SAHMM) model**. *Res. Soc. Adm. Pharm.* (2017.0) **13** 39-47. DOI: 10.1016/j.sapharm.2015.12.013 19. Carayon P., Karsh B.-T., Gurses A.P., Holden R.J., Hoonakker P., Hundt A.S., Montague E., Rodriguez A.J., Wetterneck T.B.. **Macroergonomics in Health Care Quality and Patient Safety**. *Rev. Hum. Factors Ergon.* (2013.0) **8** 4-54. DOI: 10.1177/1557234X13492976 20. Carayon P., Wetterneck T.B., Rivera-Rodriguez A.J., Hundt A.S., Hoonakker P., Holden R., Gurses A.P.. **Human factors systems approach to healthcare quality and patient safety**. *Appl. Ergon.* (2014.0) **45** 14-25. DOI: 10.1016/j.apergo.2013.04.023 21. Holden R.J., Valdez R.S.. **Conceptualizing the patient work system, Part B: Macroergonomic challenges**. *Proceedings of the 11th International Symposium on Human Factors in Organizational Design and Management-XI Nordic Ergonomics Society Annual Conference* 22. Schwartz J.K.. **Pillbox use, satisfaction, and effectiveness among persons with chronic health conditions**. *Assist. Technol.* (2017.0) **29** 181-187. DOI: 10.1080/10400435.2016.1219884 23. Tordoff J., Simonsen K., Thomson W.M., Norris P.T.. **It’s just routine.” A qualitative study of medicine-taking amongst older people in New Zealand**. *Pharm. World Sci.* (2010.0) **32** 154-161. DOI: 10.1007/s11096-009-9361-5 24. Sanders M.J., Van Oss T.. **Using Daily Routines to Promote Medication Adherence in Older Adults**. *Am. J. Occup. Ther.* (2013.0) **67** 91-99. DOI: 10.5014/ajot.2013.005033 25. Hennink M.M., Kaiser B.N., Marconi V.C.. **Code Saturation Versus Meaning Saturation: How Many Interviews Are Enough?**. *Qual. Health Res.* (2017.0) **27** 591-608. DOI: 10.1177/1049732316665344 26. Thomas D.R.. **A general inductive approach for analyzing qualitative evaluation data**. *Am. J. Eval.* (2006.0) **27** 237-246. DOI: 10.1177/1098214005283748 27. Boeije H.. **A Purposeful Approach to the Constant Comparative Method in the Analysis of Qualitative Interviews**. *Qual. Quant.* (2002.0) **36** 391-409. DOI: 10.1023/A:1020909529486 28. Carayon P.. **Sociotechnical systems approach to healthcare quality and patient safety**. *Work* (2012.0) **41** 3850-3854. DOI: 10.3233/WOR-2012-0091-3850 29. Burish T.G., Carey M.P., Wallston K.A., Stein M.J., Jamison R.N., Lyles J.N.. **Health Locus of Control and Chronic Disease: An External Orientation May Be Advantageous**. *J. Soc. Clin. Psychol.* (1984.0) **2** 326-332. DOI: 10.1521/jscp.1984.2.4.326 30. Notenboom K., Beers E., Riet-Nales D.A., Egberts T., Leufkens H.G.M., Jansen P.A.F., Bouvy M.. **Practical Problems with Medication Use that Older People Experience: A Qualitative Study**. *J. Am. Geriatr. Soc.* (2014.0) **62** 2339-2344. DOI: 10.1111/jgs.13126 31. Synnes O., Orøy A.J., Råheim M., Bachmann L., Ekra E.M.R., Gjengedal E., Høie M., Jørgensen E., Michaelsen R.K.A., Sundal H.. **Finding ways to carry on: Stories of vulnerability in chronic illness**. *Int. J. Qual. Stud. Health Well-Being* (2020.0) **15** 1819635. DOI: 10.1080/17482631.2020.1819635 32. Affleck G., Tennen H., Pfeiffer C., Fifield J.. **Appraisals of control and predictability in adapting to a chronic disease**. *J. Personal. Soc. Psychol.* (1987.0) **53** 273. DOI: 10.1037/0022-3514.53.2.273 33. Stanton A.L., Revenson T.A., Tennen H.. **Health Psychology: Psychological Adjustment to Chronic Disease**. *Annu. Rev. Psychol.* (2007.0) **58** 565-592. DOI: 10.1146/annurev.psych.58.110405.085615 34. Dills H., Shah K., Messinger-Rapport B., Bradford K., Syed Q.. **Deprescribing Medications for Chronic Diseases Management in Primary Care Settings: A Systematic Review of Randomized Controlled Trials**. *J. Am. Med. Dir. Assoc.* (2018.0) **19** 923-935.e2. DOI: 10.1016/j.jamda.2018.06.021 35. Dharmarajan T.S., Choi H., Hossain N., Munasinghe U., Lakhi F., Lourdusamy D., Onuoha S., Murakonda P., Skokowska-Lebelt A., Kanagala M.. **Deprescribing as a clinical improvement focus**. *J. Am. Med. Dir. Assoc. Assoc.* (2020.0) **21** 355-360. DOI: 10.1016/j.jamda.2019.08.031 36. Kruse W., Rampmaier J., Frauenrath-Volkers C., Volkert D., Wankmüller I., Micol W., Oster P., Schlierf G.. **Drug-prescribing patterns in old age**. *Eur. J. Clin. Pharmacol.* (1991.0) **41** 441-447. DOI: 10.1007/BF00626366 37. Kennedy I.. **Patients are experts in their own field: The interests of patients and healthcare professionals are intertwined**. *BMJ* (2003.0) **326** 1276-1277. DOI: 10.1136/bmj.326.7402.1276 38. Tattersall R.. **The expert patient: A new approach to chronic disease management for the twenty-first century**. *Clin. Med.* (2002.0) **2** 227-229. DOI: 10.7861/clinmedicine.2-3-227 39. Gallagher K.M., Updegraff J.A.. **Health Message Framing Effects on Attitudes, Intentions, and Behavior: A Meta-analytic Review**. *Ann. Behav. Med.* (2011.0) **43** 101-116. DOI: 10.1007/s12160-011-9308-7 40. Stuijt C., Bekker C., Bemt B.V.D., Karapinar F.. **Effect of medication reconciliation on patient reported potential adverse events after hospital discharge**. *Res. Soc. Adm. Pharm.* (2021.0) **17** 1426-1432. DOI: 10.1016/j.sapharm.2020.10.012 41. Hilmer S.N., Gnjidic D., Le Couteur D.G.. **Thinking through the medication list: Appropriate prescribing and depre-scribing in robust and frail older patients**. *Aust. Fam. Physician* (2012.0) **41** 924-928. PMID: 23210113 42. Wickens C.D., Gordon S.E., Liu Y., Lee J.. *An Introduction to Human Factors Engineering* (2004.0) **Volume 2**
--- title: 'Teaching Deprescribing and Combating Polypharmacy in the Pharmacy Curriculum: Educational Recommendations from Thematic Analysis of Focus Groups' authors: - Devin Scott - Alina Cernasev - Rachel E. Barenie - Sydney P. Springer - David R. Axon journal: Clinics and Practice year: 2023 pmcid: PMC10037566 doi: 10.3390/clinpract13020040 license: CC BY 4.0 --- # Teaching Deprescribing and Combating Polypharmacy in the Pharmacy Curriculum: Educational Recommendations from Thematic Analysis of Focus Groups ## Abstract In the last two decades in the United States (US), the previous research has focused on medication optimization, including polypharmacy. Polypharmacy is associated with several negative outcomes, which may be resolved by deprescribing medications that are no longer necessary. Although deprescribing is a critical aspect of a pharmacist’s role, some studies have demonstrated that student pharmacists are less familiar with their future role in deprescribing. Thus, this study aimed to explore student pharmacists’ perceptions of deprescribing in the pharmacy curriculum. This qualitative study was conducted with student pharmacists enrolled in three Doctor of Pharmacy (Pharm. D.) programs in the US. The participants, all student pharmacists at the time of the study, were identified via an email requesting their voluntary participation in a focus group study. The focus groups were conducted via an online platform over three months in 2022, and recruitment continued until thematic saturation was obtained. Using thematic analysis, the corpus of the transcribed data was imported into Dedoose®, a qualitative software that facilitated the analysis. Three themes emerged from the data: [1] the importance of deprescribing; [2] barriers to deprescribing; [3] education recommendations. The data highlight that the student pharmacists believe integrating deprescribing content into the clinical, didactic, and simulation education would help them overcome the identified obstacles. Colleges of pharmacy should consider emphasizing the importance of deprescribing in their curriculum, creating programs to assist future pharmacists in addressing the barriers to deprescribing, and adopting the suggested educational strategies to improve the deprescribing education that is offered. ## 1. Introduction Student pharmacists in the United States (US) study evidence-based medicine and the safety and effectiveness of medications to optimize therapy for their patients [1]. Medication optimization is increasingly important given the aging health status of the US population. For example, in 2012, approximately 60 million people in the US had two or more chronic health conditions that lead to increased morbidity [2]. Increased morbidity leads to an increase in the number of medications being used, as highlighted by a retrospective database analysis that showed $36.8\%$ of individuals were taking more than five medications (i.e., polypharmacy) to manage their chronic conditions between 2009 and 2016 [3]. Polypharmacy is associated with several negative outcomes, which may be mitigated with appropriate deprescribing efforts [4]. Although deprescribing is becoming an important role for pharmacists, many student pharmacists are unfamiliar with the concept of deprescribing and their role in it [5]. Deprescribing has been defined as “the process of withdrawal of an inappropriate medication, supervised by a health care professional with the goal of managing polypharmacy and improving outcomes” [6]. Previous studies have shown that the teaching of deprescribing in professional programs varies [7] and that student pharmacists are often unprepared to recommend or implement deprescribing in practice [8,9,10]. Another study found that student pharmacists felt more prepared to identify inappropriate medications than medical students [9]. Although symposiums have been held in recent years to discuss deprescribing education [11,12], student pharmacists still need opportunities in their professional curriculum to become equipped for deprescribing activities upon entering clinical practice. While clinical tools to assist with deprescribing efforts are available, challenges implementing these tools exist. For example, one of the tools focuses on certain drug classes instead of a comprehensive guide for all medications [6]. There is also the precedent of expanding healthcare professionals’ roles to formally include deprescribing. For instance, in the nursing community, there have been recent efforts, with a degree of success, to implement aspects of deprescribing into nurses’ practice [13]. If deprescribing is to be successful, however, new pharmacists require support from experienced healthcare colleagues [14]. Despite this progress, there is still a lack of knowledge regarding how student pharmacists perceive deprescribing. Thus, this study aimed to explore student pharmacists’ perceptions of the implementation of the place of deprescribing in the pharmacy curriculum. ## Subjects, Recruitment, Data Collection, and Analysis This qualitative study was conducted with student pharmacists enrolled in three different Doctor of Pharmacy (Pharm. D.) programs, including the University of Tennessee Health Science Center (UTHSC, IRB# 21-08234-XM), University of Arizona (IRB# 2021-015-PHPR), and University of New England (UNE, IRB# 0821). The subjects, who were all student pharmacists at the time of the study, were identified via an email that requested their voluntary participation in a focus group (FG) study. Focus groups enhance the potential for producing richer data and obtaining a collective opinion about the research question, because they facilitate brainstorming and discussion between four or more participants simultaneously [15]. Research team members with experience in qualitative data collection (D.S. and A.C.) led all focus groups. Focus groups were conducted via an online platform (Zoom), which enabled students from all three universities to attend [16]. The recruitment occurred simultaneously over three months in the fall semester of 2022 in all three colleges of pharmacy. Interested student pharmacists signed up on a list, and A.C. contacted them to find out the preferred date and time they would like to participate in the study. A semi-structured focus group strategy was used to allow the students to discuss their experiences with deprescribing in the pharmacy curriculum and during their Advanced Pharmacy Practice Experience (APPE) and Institutional Pharmacy Practice Experience (IPPE) [15]. This study used the theory of planned behavior to conceptualize the semi-structured focus groups [17]. Springer et al., 2022 provided additional information regarding the conceptualization of this study and how the theory of planned behavior guided the development of the questions [8]. The semi-structured strategy allowed the researchers to largely pose the same questions to each group while allowing for additional questions that were raised by earlier discussions to be included in focus groups conducted later. This strategy to incorporate additional questions enhanced the external validity of the study findings [15]. The questions were divided into three topics, which focused on deprescribing education [8]. The virtual focus group transcripts were audio recorded, professionally transcribed, and analyzed using thematic analysis that followed the six-step process as outlined by Braun and Clarke [18]. The steps included [1] familiarization with the corpus of data; [2] inductively coding the entire dataset; [3] identifying emerging themes; [4] reviewing themes with the research team; [5] defining and naming the themes; and [5] writing the analysis [18]. The analysis team (A.C., D.S., and D.R.A.) continued recruiting subjects until thematic saturation was achieved, at which point no new themes emerged with subsequent focus groups [15]. All focus group transcripts were uploaded to a qualitative analysis software, Dedoose® (Manhattan Beach, CA, USA), which was used for generating initial codes and developing and reviewing themes. Standards for Reporting Qualitative Research (SPQR) criteria for demonstrating the quality of qualitative research were met [19]. A previous manuscript described the methodology in more detail [8]. ## 3. Results Student pharmacists from three colleges of pharmacy in the US were invited to participate in this study. The total number of eligible student pharmacists was 1366 (UNE, $$n = 158$$; University of Arizona, $$n = 526$$; UTHSC, $$n = 682$$). Of these, 26 student pharmacists consented to participate in four focus group discussions. Most participants ($$n = 16$$) were enrolled in their fourth year, seven were enrolled in their third year, and three were enrolled in their second year. Participants ranged in age from 21 to 37 years old, with a mean age of 24 years old. A total of four focus groups were conducted over three months in 2021. The average time of the focus groups was 76 min. Three major themes were revealed by thematic analysis. The first theme centered on student assertions that deprescribing is vitally important for pharmacists, patients, and allied health professionals. The second theme encompassed perceptions that there are significant barriers to deprescribing. The third theme consisted of the education recommendations, including didactic, clinical, and simulation education, surrounding deprescribing. These themes spotlight the importance of deprescribing, outline the barriers to deprescribing for pharmacists, and offer educational solutions to overcome barriers and improve patient and population health by deprescribing. Theme 1: The Importance of Deprescribing: “It’s probably one of the forefront problems for pharmacies these days”. The student pharmacists repeatedly called attention to the importance of deprescribing for patient health and for their future careers. Student 2 (ST2) asserted that pharmacists play a central role in deprescribing, as they cannot rely on other healthcare professionals to deprescribe: ST3, from another FG echoed a similar sentiment: ST3 argued that pharmacists are crucial members of the healthcare team, especially in terms of deprescribing. ST5 also highlighted that deprescribing is particularly important for women’s health: In sum, ST5 indicated that deprescribing and deprescribing education could help to ameliorate the healthcare disparities faced by women Polypharmacy was a key concern for participants, who linked the prevalence of polypharmacy to the importance of deprescribing. ST1 stated: ST2 agreed and called attention to the unique role pharmacists play in the deprescribing process: In another FG, ST5 again noted that deprescribing is vitally important to reducing polypharmacy: In summation, ST5 linked deprescribing to polypharmacy and medication adherence. Finally, ST1 wrapped up their focus group’s discussion surrounding the importance of deprescribing by emphasizing pharmacist’s role in improving patient health: The student pharmacists considered deprescribing to be a crucial part of their role as future pharmacists and identified deprescribing as a potential solution to polypharmacy, an improved medication adherence, and a help to mitigating the healthcare disparities faced by women. Theme 2: Barriers to Deprescribing: “Another barrier would be…”. The student pharmacists commented that the biggest barriers to successful deprescribing are due to prescriber and patient resistance to deprescribing and trust gaps between pharmacists and prescribers and between pharmacists and patients. ST4 asserted that patients trust physicians more than pharmacists and that pharmacists need to focus on building trust with their patients: ST1 echoed that concern: In summary, ST1 argued that the trust disparity between patients and physicians and between patients and pharmacists is a major barrier to initiating deprescribing. ST10 spoke about an unsuccessful attempt to deprescribe a very large dose of an antipsychotic that was driven by the fear of the prescriber on the healthcare team: ST5 provided further examples of unsuccessful attempts to deprescribe related to the trust gap: In regard to hesitancy about approaching prescribers about deprescribing, ST3 reiterated this common theme: The student pharmacists did remain optimistic that they could successfully deprescribe by focusing on establishing trust with physicians, as indicated by ST4: In sum, ST1 again called attention to the vital role that pharmacists play in the healthcare team in pursuit of improving patient health. In the FGs, the student pharmacists repeatedly linked prescriber and patient resistance to deprescribing to trust gaps between pharmacists and prescribers and between pharmacists and patients. Ultimately, the student pharmacists argued that pharmacists have to overcome these barriers and build trusting relationships with prescribers and patients in order to successfully deprescribe. Theme 3: Education Recommendations: “[Deprescribing] needs to be taught soon, you know, at the start of the pharmacy curriculum rather than at the end” The student pharmacists were verbose when offering suggestions for improving deprescribing education, which included deprescribing simulations, integrating deprescribing throughout the didactic curriculum, and emphasizing deprescribing during clinical experiences. Table 1 provides an overview of the educational recommendations provided by the student pharmacists in the focus group interviews. The student pharmacists suggested integrating a variety of deprescribing simulations throughout pharmacy school. ST3 suggested integrating deprescribing simulations into the pharmacy curriculum: ST2 agreed stating: ST2 saw a need for simulations that include deprescribing discussions with prescribers: ST4 reflected on the professional communication simulations they experienced and called for frequent practice with the hard conversations related to deprescribing: ST3 expanded on the simulation discussion and suggested adding deprescribing education throughout the curriculum: In addition to deprescribing simulations, the student pharmacists called for more didactic deprescribing instruction. ST2 suggested adding deprescribing to lectures and asked faculty to “slip in like deprescribing info for each disease state or drug that you’re going over.” ( ST2, FG2) ST1 concurred: ST4 offered a similar suggestion: ST1 called for integrating deprescribing into the curriculum and suggested offering continuing education focused on deprescribing: In addition to simulation and didactic instruction, the student pharmacists called for preceptors to place a greater emphasis on deprescribing during clinical experiences. ST3 suggested adding objectives to clinical rotations with preceptors to increase exposure to deprescribing: ST5 concurred: ST2 echoed the sentiment: ST3 shared their experience with preceptors discussing deprescribing and argued that preceptor education on deprescribing was paramount to building the confidence to deprescribe: The deprescribing education recommendations centered around simulation, didactics, and clinical rotations. The student pharmacists proposed integrating deprescribing into the existing pharmacy curriculum, adding deprescribing-focused simulations, and providing guidance to preceptors surrounding deprescribing education during clinical experiences. ## 4. Discussion The three emergent themes offer insight into the importance of deprescribing for pharmacists and the barriers faced when initiating deprescribing and offer recommendations for improving pharmacy education on deprescribing. With the increased prevalence of polypharmacy in the US, understanding the barriers pharmacists face when deprescribing and recommendations to enhance deprescribing education in the pharmacy curriculum is needed [4]. Focus groups were conducted to better understand student pharmacists’ perspectives on deprescribing. During these focus groups, the participants were invited to share their experiences with and thoughts about deprescribing. The themes established from the student pharmacists’ responses during these interviews provide a justification for an increased emphasis on deprescribing education, call for contemplation on the barriers to deprescribing faced by pharmacists, and offer a roadmap for improving deprescribing education for student pharmacists. The student pharmacists identified deprescribing and polypharmacy as pressing, interrelated issues facing pharmacists today. Throughout the interviews, the student pharmacists repeatedly emphasized the importance of deprescribing to improve patient health [20]. The participants viewed deprescribing as a key function of their role as pharmacists. They called attention to the fact that pharmacists play a key role in deprescribing as they often have the most comprehensive view of their patient’s medication regimen, which is supported by Reeder and Mutnick, who “found that pharmacist-obtained medication histories resulted in less discrepancies and more-thorough medication histories than did physician-obtained medication histories” [21]. The student pharmacists suggested that deprescribing could be a possible solution to polypharmacy, an aid to increase medication adherence, and a step towards combating the healthcare challenges faced by women [22,23]. The assertion that deprescribing can reduce polypharmacy is supported by Sun et al., who stated: “The reduction of polypharmacy prevalence rates, along with a decrease in the associated [adverse drug reactions] can be accomplished through the process of deprescribing” [13]. While the student pharmacists argued for the importance of deprescribing, they also highlighted that initiating deprescribing faces significant barriers. During focus groups, the student pharmacists shared their unsuccessful deprescribing experiences, which they attributed to patient and prescriber resistance to deprescribing. These concerns align with Reeve’s work on deprescribing tools [6]. Reeve stated that “tools which highlight inappropriate medications may not be effective at increasing deprescribing as time is needed to assess the appropriateness in the individual, discuss it with the patient and/or caregivers and plan the deprescribing process” [6]. Ultimately, the student pharmacists linked their failed deprescribing experiences to the strong trust patients place in prescribers and their own limited relationships with patients. The participants called attention to the strong trust bond between prescribers and patients that often dwarfed their own trust bond with patients. They also spoke to the need for developing strong, trusting relationships between pharmacists and patients. To overcome patient and prescriber resistance to deprescribing, the student pharmacists suggested focusing on building strong relationships with patients and on building trust between pharmacists and prescribers. Emphasizing the importance of strengthening relationships between pharmacists and patients may be promising for deprescribing and limiting polypharmacy, as Waszyk-Nowacyzk et al. found: of the patients they surveyed at a single-center in a large Polish city, “$79.4\%$ of patients would like to benefit from medicines use reviews provided by a pharmacist” [24]. Additionally, pharmacists, in Poland, interviewed by Łucja Zielińska-Tomczak et al. on the topic of interprofessional collaboration “indicated that the younger generation of physicians seems more cooperative than older doctors”, while the physicians who were interviewed in Poland “supported these views and suggested a positive effect of informal relationships with pharmacists on doctors’ openness towards collaboration” [25]. To achieve the benefits of deprescribing, such as improving patient health, and to overcome the barriers to deprescribing, the participants proposed adopting new educational practices surrounding deprescribing simulation, didactics, and clinical experiences. Other researchers also concluded, based on their survey of trainees in pharmacy medicine and nursing, that “alterations to the current curricular content may be warranted to address lack of preparedness to deprescribe in clinical practice” [9]. The pharmacy students called for deprescribing simulations with prescribers and with patients so they could practice those difficult conversations and receive feedback and advice. This aligns with Palaganas, Epps, and Raemer, who argued that “given the increasing adoption of experiential learning and team-based learning, [healthcare simulation] has become a preferred vehicle for [interprofessional education]” [26]. Participants asked for deprescribing to be integrated within the didactic curriculum whenever discussing disease states or drugs. This desire for additional didactic instruction on deprescribing tracks with a survey of stunt pharmacists by Clark et al., who found that “less than half of students felt that their didactic training adequately prepared them for deprescribing in the clinical setting” [10]. In addition, the student pharmacists recommended that pharmacy schools give preceptors learning objectives based on deprescribing and to make deprescribing part of the clinical curriculum. These recommendations are supported by Raiman-Wilms et al., who argued that “Teaching health professional learners how to apply evidence into clinical shared decision-making is important in the teaching of prescribing and deprescribing” [12]. Deprescribing is vitally important to reducing polypharmacy and improving patient health [5,7,13]. While there are significant barriers to deprescribing, the educational recommendations offered here are a meaningful step towards integrating deprescribing into the everyday practice of pharmacists and student pharmacists alike. ## 4.1. Strengths, Limitations, and Future Studies This study offers a broad view on student pharmacists’ perceptions of deprescribing, as it was compromised of a heterogeneous sample size of student pharmacists from geographical locations throughout the US (East, South, and West). The student pharmacists sampled were predominately third- and fourth-year students. As with any study of this nature, it is possible that only those with an interest in this topic consented to participate in the study, which may be a source of sampling bias. While third- and fourth-year students can provide a comprehensive take on deprescribing education throughout the pharmacy curriculum, future research may focus on the recruitment of first- and second-year students pharmacists to better understand their unique perspectives on deprescribing. The use of TPB to guide the qualitative study design facilitated the recording of a heterogeneous sample of student pharmacist voices. Additionally, the use of video conferencing software aided in the recruitment of student pharmacists from across the US to offer their experiences with and perspectives on deprescribing. The sample characteristics and recruitment methods employed may, however, limit the generalizability of the findings. The findings from this study call for more longitudinal research on deprescribing and the collection of perspectives from across the healthcare education system. ## 4.2. Conclusions and Future Studies This study builds on previous work that discussed student pharmacists’ limited knowledge about deprescribing and their belief that deprescribing education is necessary [8]. This study expands upon that previous work by calling attention to the importance of deprescribing in pharmacy today and providing educational suggestions to overcome barriers to deprescribing through three emerging themes: [1] deprescribing improves patient care and is a necessary skill for pharmacists; [2] there are significant barriers to deprescribing; [3] barriers to deprescribing can be overcome by providing additional educational opportunities. With the recognized benefits of deprescribing for patients, the student pharmacists see the need to become proficient in deprescribing, yet there are significant barriers that pharmacists continue to face when deprescribing. The student pharmacists believe that the integration of deprescribing content in clinical, didactic, and simulation education will help them to overcome the barriers to deprescribing and to positively impact patient care. Colleges of pharmacy should consider emphasizing the importance of deprescribing, creating programs to assist future pharmacists in addressing barriers to deprescribing, and adopting the educational strategies suggested to improve deprescribing education. ## References 1. Medina M.S., Plaza C.M., Stowe C.D., Robinson E.T., DeLander G., Beck D.E., Melchert R.B., Supernaw R.B., Roche V.F., Gleason B.L.. **Center for the Advancement of Pharmacy Education 2013 educational outcomes**. *Am. J. Pharm. Educ.* (2013) **77** 162. DOI: 10.5688/ajpe778162 2. Ward B.W., Schiller J.S., Goodman R.A.. **Peer reviewed: Multiple chronic conditions among us adults: A 2012 update**. *Prev. Chronic Dis.* (2014) **11** E62. DOI: 10.5888/pcd11.130389 3. Young E.H., Pan S., Yap A.G., Reveles K.R., Bhakta K.. **Polypharmacy prevalence in older adults seen in United States physician offices from 2009 to 2016**. *PLoS ONE* (2021) **16**. DOI: 10.1371/journal.pone.0255642 4. Halli-Tierney A., Scarbrough C., Carroll D.G.. **Polypharmacy: Evaluating risks and deprescribing**. *Am. Fam. Physician* (2019) **100** 32-38. PMID: 31259501 5. Poots A.J., Jubraj B., Ward E., Wycoco A., Barnett N.. **Education around medication review and deprescribing: A survey of medical and pharmacy students’ perspectives**. *Ther. Adv. Drug Saf.* (2020) **11** 2042098620909610. DOI: 10.1177/2042098620909610 6. Reeve E., Gnjidic D., Long J., Hilmer S.. **A systematic review of the emerging definition of ‘deprescribing’with network analysis: Implications for future research and clinical practice**. *Br. J. Clin. Pharmacol.* (2015) **80** 1254-1268. DOI: 10.1111/bcp.12732 7. Barnett N.L., Jubraj B., Grant D., Reddy B., Stevenson J.M.. **Medication review, polypharmacy and deprescribing: Results of a pilot scoping exercise in undergraduate and postgraduate education**. *Pharm. Educ.* (2021) **21** 126-132. DOI: 10.46542/pe.2021.211.126132 8. Springer S.P., Cernasev A., Barenie R.E., Axon D.R., Scott D.. **“I Think Deprescribing as a Whole Is a Gap!”: A Qualitative Study of Student Pharmacist Perceptions about Deprescribing**. *Geriatrics* (2022) **7**. DOI: 10.3390/geriatrics7030060 9. Zimmerman K.M., Bell C.A., Donohoe K.L., Salgado T.M.. **Medicine, pharmacy and nursing trainees’ perceptions of curriculum preparation to deprescribe and interprofessional roles in the deprescribing process**. *Gerontol. Geriatr. Educ.* (2020) **41** 63-84. DOI: 10.1080/02701960.2019.1661840 10. Clark C.M., Hejna M., Shao E., Maerten-Rivera J.L., Monte S.V., Wahler R.G.. **Knowledge and Attitudes of Student Pharmacists Regarding Polypharmacy and Deprescribing: A Cross-Sectional Study**. *Pharmacy* (2020) **8**. DOI: 10.3390/pharmacy8040220 11. Oldenburg L.I.K., Dalin D., Drastrup A.M., Vermehren C.. **Deprescribing: What is the gold standard? Themes that characterized the discussions at the first Danish symposium on evidence-based deprescribing**. *Explor. Res. Clin. Soc. Pharm.* (2022) **5** 100102. DOI: 10.1016/j.rcsop.2022.100102 12. Raman-Wilms L., Farrell B., Sadowski C., Austin Z.. **Deprescribing: An educational imperative**. *Res. Soc. Adm. Pharm.* (2019) **15** 790-795. DOI: 10.1016/j.sapharm.2018.08.011 13. Sun W., Grabkowski M., Zou P., Ashtarieh B.. **The Development of a Deprescribing Competency Framework in Geriatric Nursing Education**. *West. J. Nurs. Res.* (2021) **43** 1043-1050. DOI: 10.1177/01939459211023805 14. Poots A.J., Jubraj B., Barnett N.L.. **Education around deprescribing: ‘spread and embed’the story so far**. *Br. Med. J. Publ. Group* (2017) **24** 7-9. DOI: 10.1136/ejhpharm-2016-001153 15. Krueger R.A.. **Analyzing focus group interviews**. *J. Wound Ostomy Cont. Nurs.* (2006) **33** 478-481. DOI: 10.1097/00152192-200609000-00004 16. Lobe B., Morgan D., Hoffman K.A.. **Qualitative data collection in an era of social distancing**. *Int. J. Qual. Methods* (2020) **19** 1609406920937875. DOI: 10.1177/1609406920937875 17. Ajzen I.. **The theory of planned behavior**. *Organ. Behav. Hum. Decis. Process.* (1991) **50** 179-211. DOI: 10.1016/0749-5978(91)90020-T 18. Braun V., Clarke V.. **Using thematic analysis in psychology**. *Qual. Res. Psychol.* (2006) **3** 77-101. DOI: 10.1191/1478088706qp063oa 19. O’Brien B.C., Harris I.B., Beckman T.J., Reed D.A., Cook D.A.. **Standards for reporting qualitative research: A synthesis of recommendations**. *Acad. Med.* (2014) **89** 1245-1251. DOI: 10.1097/ACM.0000000000000388 20. Woodward M.C.. **Deprescribing: Achieving better health outcomes for older people through reducing medications**. *J. Pharm. Pract. Res.* (2003) **33** 323-328. DOI: 10.1002/jppr2003334323 21. Reeder T.A., Mutnick A.. **Pharmacist-versus physician-obtained medication histories**. *Am. J. Health-Syst. Pharm.* (2008) **65** 857-860. DOI: 10.2146/ajhp070292 22. Reeve E., Wiese M.D.. *Benefits of Deprescribing on Patients’ Adherence to Medications* (2014) 23. Rochon P.A., Petrovic M., Cherubini A., Onder G., O’Mahony D., Sternberg S.A., Stall N.M., Gurwitz J.H.. **Polypharmacy, inappropriate prescribing, and deprescribing in older people: Through a sex and gender lens**. *Lancet Healthy Longev.* (2021) **2** e290-e300. DOI: 10.1016/S2666-7568(21)00054-4 24. Waszyk-Nowaczyk M., Guzenda W., Kamasa K., Zielińska-Tomczak Ł., Cerbin-Koczorowska M., Michalak M., Przymuszała P., Plewka B.. **Polish Patients’ Needs and Opinions about the Implementation of Pharmaceutical Care in Diabetes**. *Int. J. Environ. Res. Public Health* (2023) **20**. DOI: 10.3390/ijerph20020945 25. Zielińska-Tomczak Ł., Cerbin-Koczorowska M., Przymuszała P., Marciniak R.. **How to effectively promote interprofessional collaboration?—A qualitative study on physicians’ and pharmacists’ perspectives driven by the theory of planned behavior**. *BMC Health Serv. Res.* (2021) **21**. DOI: 10.1186/s12913-021-06903-5 26. Palaganas J.C., Epps C., Raemer D.B.. **A history of simulation-enhanced interprofessional education**. *J. Interprofessional Care* (2014) **28** 110-115. DOI: 10.3109/13561820.2013.869198
--- title: Identification of Prescribing Patterns in Hemodialysis Outpatients Taking Multiple Medications authors: - Hiroyuki Nagano - Koji Tomori - Mano Koiwa - Shotaro Kobayashi - Masahiro Takahashi - Hideki Makabe - Hirokazu Okada - Akifumi Kushiyama journal: Pharmacy year: 2023 pmcid: PMC10037568 doi: 10.3390/pharmacy11020043 license: CC BY 4.0 --- # Identification of Prescribing Patterns in Hemodialysis Outpatients Taking Multiple Medications ## Abstract We investigated the relationship between multidrug administration and the characteristics, pathophysiology, and drug class in outpatients with hemodialysis. A retrospective cross-sectional study was conducted at Saitama Medical University Hospital in October 2018. Multidrug administration was defined as receiving either more than six drugs or more than the median number of drugs. The drugs used were represented by their anatomical classification codes in the Anatomical Therapeutic Chemistry Classification System (ATC classification). A latent class analysis (LCA) was used to identify clusters at risk of receiving multiple medications. A stepwise logistic regression analysis was performed to select ATC classifications prone to being involved in multidrug administration. As of October 2018, 98 outpatients with hemodialysis were enrolled in the study. In the LCA, when diabetes was the main primary disease, oral hypoglycemic agents available to dialysis patients were limited, but the number of drugs administered was large. Old age, poor nourishment, a long history of dialysis, and chronic nephritis were associated with multidrug administration among nondiabetic patients. In the second level of the ATC classification, the drugs frequently used were coded A02 (drugs for acid-related disorders), A07 (antidiarrheal agents, intestinal anti-inflammatory/anti-infective agents), B01 (antithrombotic agents), and N05 (psycholeptics). The prescribing patterns for either diabetic patients or nondiabetic elderly patients were identified in outpatients with hemodialysis taking multiple medications, and drugs for acid-related disorders, antidiarrheal agents, intestinal anti-inflammatory/anti-infective agents, antithrombotic agents, and psycholeptics are frequently used in those patients. ## 1. Introduction Polypharmacy, reportedly, increases not only medical costs [1] but also health disorders, such as frailty and fall-related disadvantages [2,3]. The relationship between the drug count and adverse events in elderly patients has been shown to involve the risk of adverse drug events being markedly increased in patients receiving ≥5 drugs [4]. In Japan, polypharmacy is defined as receiving 5–6 drugs, and taking ≥5 drugs increases the risk of falls, while taking ≥6 drugs increases the risk of adverse drug events [5]. Multidrug administration is a risk factor for the progression of chronic kidney disease (CKD) [6], and $91\%$ of CKD patients who are prescribed multiple drugs have a prescription that requires attention to interaction [7]. As CKD progresses, it leads to end-stage kidney disease (ESKD), and hemodialysis is one of the options for renal replacement therapy. The number of dialysis patients (prevalence) per 1 million population is increasing year by year [8]. According to the 2018 United State Renal Data System (USRDS), the prevalence of dialysis patients in *Japan is* the second highest in the world after Taiwan [9]. In recent years, aging has become a major social problem in Japan, and the function of the kidney declines with age, so the proportion of patients with CKD increases in the elderly [10]. Therefore, it is expected that the number of dialysis patients will continue to increase. From reports in the United States and Japan, hemodialysis patients are prone to polypharmacy [11,12], and the factors that cause polypharmacy are related to hypertension, diabetes, cardiovascular disease, and dyslipidemia [6]. However, the pattern of drug selection for primary diseases and comorbidities among outpatients with hemodialysis leading to multidrug administration usage is unknown. For pharmacists to intervene in cases of multidrug administration, it is first necessary for them to understand the prescribing patterns. ## 2.1. Objectives The primary objective was to investigate the relationship between polypharmacy and the characteristics, pathophysiology, and drug classes in hemodialysis outpatients, as the association between drug choice and polypharmacy use in this population is unclear. ## 2.2. Patient Population and Setting The subjects were outpatient hemodialysis patients at our hospital as of October 2018. We conducted a retrospective cross-sectional study. ## 2.3. Data Collection, Definitions and Outcomes We collected information on patient background, blood tests, items related to dialysis efficiency and nutritional effects, and drugs used from electronic medical records. As the characteristics of the study population, the median and interquartile range of the number of drugs used, age, and dialysis history were investigated. The following items were investigated in terms of number and proportion: sex, online-hemodiafiltration (online-HDF), number of visits to other clinical departments, comorbidities (diabetes mellitus, cardiovascular disease, cerebrovascular disease, peripheral arterial disease, liver diseases), and primary diseases of renal failure (diabetic nephropathy, nephrosclerosis, chronic glomerular nephritis, polycystic kidney disease, IgA nephropathy, unknown, other.). The following items were investigated for the median and interquartile range: blood test values (serum albumin, corrected calcium, serum phosphorus, hemoglobin concentration, serum ferritin level, intact parathyroid hormone [i-PTH], β2-microglobulin [M]) and dialysis efficiency/nutritional effect (kt/V, normal Protein Catabolic Rate [nPCR], Geriatric Nutritional Risk Index [GNRI], cardiothoracic ratio [CTR], dry weight [DW]). Multidrug administration was defined as receiving either more than six drugs or more than the median number of drugs. The items obtained as continuous variables were categorized. Those for which control target values could be defined (corrected calcium level, serum phosphorus level, hemoglobin concentration, serum ferritin level, i-PTH, β2-M, kt/V, nPCR, GNRI) were classified as within or outside the target value. Items that were difficult to define, such as the age, dialysis history, ALB, and DW, were categorized as not less than or less than the median, and the CTR was categorized as not less than or less than $50\%$. The drugs used were classified using the first level (anatomical group) of the Anatomical Therapeutic Chemical Classification System (ATC classification). ## 2.4. Statistical Analysis A latent class analysis (LCA) was used to identify the hidden groups of patients with a high prevalence of multidrug administration usage. LCA is a post hoc grouping method [13], and patients with such data as their prescription drugs and primary diseases, indicated as binary data, were classified into three clusters. We next examined the trend in the risk of taking multiple drugs. An LCA assumes that there are latent subgroups (called classes) that affect categorical data that are actually observed and correlate with each other. It is used to analyze the structure by estimating the probability that a subject belongs to a certain class and the conditional probability of the item reaction when the subject belongs to a certain class using the latent class model [14]. A bivariate analysis was performed with the first level of the ATC classification as the explanatory variable for the objective variable, which identified the number of drugs used greater than or equal to the median number of drugs as multidrug use. For classifications that showed a significant difference in this analysis, a logistic regression analysis was performed using the stepwise method for all second-level classifications (treatment method subgroups). Univariate and multivariate logistic regression was performed to evaluate the association between polypharmacy and specific comorbidities (diabetes, cardiovascular diseases cerebrovascular diseases, peripheral arterial diseases, and liver diseases) The JMP Pro software program, ver. 15.2 (SAS Institute, Cary, NC, USA), was used for these statistical analyses. ## 3.1. Characteristics of the Study Population As of October 2018, we had collected 98 outpatients with hemodialysis from our hospital. The characteristics of the study population are shown in Table 1. The median number of drugs was nine. There were 74 and 52 patients who used ≥6 drugs or ≥9 drugs, respectively. The median age was 65 years old, and 52 patients were ≥65 years old. The median dialysis history was 42 months, and the number of patients who had been on dialysis for over 42 months was 49. There were 32 women, 46 online-HDFs, and 41 in other departments. Diabetes was the most common comorbidity (45 cases), followed by cardiovascular disease (36 cases), cerebrovascular disease (15 cases), liver diseases (14 cases) and peripheral arterial disease (11 cases). Forty-five patients had diabetes, thirty-five were taking diabetes medications, and nineteen were being treated with insulin. Seventeen patients were using one diabetes medication, thirteen patients were using two, four patients were using three, and one patient was using four. Regarding the primary disease of renal failure, diabetic nephropathy was the most common (38 cases), followed by nephrosclerosis (26 cases), with diabetic nephropathy and nephrosclerosis overlapping in four cases. Chronic glomerulonephritis was noted in 12 cases, polycystic kidney disease in 9, IgA nephropathy in 5, unknown primary disease in 6, and others in 10 cases. The median values for each blood test are shown: serum albumin 3.5 g/dL, corrected calcium 9.0 mg/dL, serum phosphorus 5.1 mg/dL, hemoglobin concentration 11.3 g/dL, serum ferritin level 87.0 ng/mL, i-PTH 153.4 pg/mL, and β2-M 24.1 mg/L. Serum albumin values were outside the control target range in 48 cases ($49.0\%$), corrected calcium in 15 cases ($15.3\%$), serum phosphorus in 20 cases ($20.4\%$), hemoglobin concentration in 30 cases ($30.9\%$), serum ferritin level in 64 cases ($66.7\%$), i-PTH in 24 cases ($24.5\%$), and β2-M in 21 cases ($21.9\%$). The median values for items related to dialysis efficiency and nutritional effect were kt/V 1.52, nPCR 0.82 g/kg/day, and GNRI 92.33. The kt/V was outside the control target range in 34 cases ($35.1\%$), nPCR in 67 cases ($69.1\%$), and GNRI in 38 cases ($40.9\%$). The CTR was ≥$50\%$ in 35 cases ($36.5\%$). The median DW was 59.1 kg, and there were 48 patients weighing ≥59.1 kg. Among first-level ATC classifications, codes A, C, and V were used by more than $80\%$ of patients. ## 3.2. The LCA The LCA was performed for 98 outpatients on dialysis, resulting in a 3-class model (Figure 1). In class 1 (C1), most patients were in the multidrug group of ≥6 drugs, and a substantial number were in the multidrug group of ≥9 drugs. C1 was characterized by diabetes as a comorbidity, and the primary disease was diabetic nephropathy in $90.2\%$ of cases. There were almost no cases of polycystic kidney disease or chronic nephritis. When logistic regression analysis was performed (Supplementary Table S1), diabetes was the only comorbidity that was independently and significantly associated with other diseases for multidrug use. Compared with other groups, C1 consulted other clinical departments more frequently, and was often hypocalcemic ($19.1\%$) and anemic ($16.0\%$), but ferritin levels were not decreased. In $90.2\%$ of cases, the nPCR did not meet management goals. The DW tended to be higher than the median. A total of $99.5\%$ of C1 patients used group A drugs (gastrointestinal tract and metabolism), while $87.1\%$ used group C drugs (circulatory system), $47.0\%$ used group N drugs (nervous system), and $47.0\%$ used group B drugs (blood and hematopoietic organs). In class 2 (C2), $91.4\%$ of the patients were in the multidrug group with ≥6 drugs. C2 was characterized by almost no patients having diabetes as a comorbidity, as these patients tended to have vascular diseases. The primary disease was chronic nephritis, followed by nephrosclerosis. Contrary to C1, high calcium was prominent in C2 ($15.2\%$). C2 also had intact PTH that is high ($22.5\%$) and low ($26.1\%$), falling on both sides of the control target. As in C1, anemia was common, but ferritin was above the control target ($15.3\%$) in many cases. C2 also had the highest percentage of patients with high β2-microglobulin level ($30.1\%$) in the three groups. Most of the drugs used were from groups A (gastrointestinal tract and metabolic action), C (circulatory system), and H (systemic In class 3 (C3), there were few patients who took ≥6 drugs, and almost none took ≥9 drugs. C3 was characterized by almost no cases of diabetes as a comorbidity or diabetic nephropathy as a primary disease. Renal sclerosis was present. Compared with other clusters, hyperphosphatemia was more common in C3 ($28.8\%$) than in the others. Anemia was rarely seen in C1 ($4.1\%$). C3 patients tended to use fewer drugs in groups B (blood and hematopoietic organs) and N (nervous system) and more in group M (musculoskeletal system). Almost all the patients in C3 were <65 years old and had never visited any other clinical departments. The GNRI was able to be managed to meet appropriate goals. ## 3.3. Results of a Bivariate Analysis (Fisher’s Exact Test) The relationship between multidrug administration (≥9 drugs per day) and each first-level ATC classification (anatomical group) was examined (Table 2). Significant differences were found among ATC classifications, with A (Alimentary Tract and Metabolism), B (Blood and Blood-forming Organs), C (Cardiovascular System), and N (Nervous System) drugs used significantly more frequently than others. ## 3.4. A Logistic Regression Analysis Using Stepwise Variable Selection A stepwise regression analysis with multidrug administration as the objective variable was performed using treatment subgroup codes (second category of A, B, C, and N codes), which were significant according to a bivariate analysis (Table 3). A02 (drugs for acid-related disorders), A07 (antidiarrheal agents, intestinal anti-inflammatory/anti-infective agents), B01 (antithrombotic agents), and N05 (psycholeptics) drugs were selected. Furthermore, the use of these drugs was associated with polypharmacy independent of each other as well as with additive polypharmacy. ## 4. Discussion The number of drugs used by dialysis patients tends to be large. A US study reported that the average number of drugs used per day was 11, and the number of tablets taken was 19 [11], while a Japanese study reported an average of 8.6 oral drugs taken per day, with 17.8 tablets [12]. The present results were similar to those of other studies in Japan. Such multidrug use is attributed to most dialysis patients having end-stage renal disease (ESRD) as well as other common chronic diseases, such as hypertension, diabetes, cardiovascular disease, and mineral and bone disorders [12]. Indeed, almost half of the patients in the present study visited multiple clinical departments, suggesting that they had a number of chronic diseases. Several aspects concerning the patient image leading to multidrug administration have been clarified. Old age and poor nutritional status have previously been reported as characteristics of patients with polypharmacy [15]. Patients with diabetes as the primary disease tend to show considerably more multidrug use than patients without diabetes. However, the number of oral hypoglycemic agents available for dialysis patients is limited. *In* general, SU agents, biguanides, and some glinides, which are characterized by renal excretion, are contraindicated. Furthermore, SGLT2 inhibitors are not usually effective [16]. In fact, the number of nonhypoglycemic agents likely contributed to the increase in polypharmacy usage in diabetic patients with dialysis. A cohort study of diabetics estimated that diabetic patients had a $13\%$ higher risk of upper gastrointestinal bleeding than nondiabetic patients [17], and diabetes was a risk factor for gastric ulcers in elderly dialysis patients [18]. Helicobacter pylori infection is more common in patients with type 2 diabetes than in nondiabetic patients [19], and it is possible that more drugs for treating gastrointestinal ulcers are used in diabetes patients than in nondiabetes patients as well. In addition, diabetes is a risk factor for the onset of constipation in dialysis patients [20], so laxatives are often used for constipation in dialysis patients with diabetes. Moreover, depressive subjects reportedly have a significantly higher prevalence of diabetes than nondepressive ones [21], so psycholeptics, including sleeping pills, may be used due to sleep disorders caused by depressive symptoms. One of the backgrounds of patients using antithrombotic drugs is coronary artery disease. About half ($53.3\%$) of CKD patients have significant coronary artery disease, even if they are asymptomatic at the time of dialysis induction, and among CKD patients with diabetes, $83.3\%$ have coronary artery disease [22]. Patients with CKD during the conservative period before dialysis have a significantly higher prevalence of peripheral arterial disease than the general population [23]. Aside from diabetic patients, we also noted that those with high frailty nature, such as the elderly and female [24], were likely to use ≥9 drugs in our study. In a cross-sectional study [15] of elderly people with frailty, the frequency of using ≥10 drugs was significantly higher than in those without frailty. There have also been several studies [25,26,27,28,29] describing the relationship between frailty and polypharmacy. In addition, the group of nondiabetic patients with multidrug administration seen in our study were underweight, had a high malnutritional risk, and had unregulated mineral bone metabolism The association between frail and undernourished patients and polypharmacy indicates that these patients are a key concern for polypharmacy, which is of a different nature than the diabetic group. We identified the drugs most often used in patients with polypharmacy. To improve or prevent multidrug administration, it is necessary to be aware of the fact that the number of drugs used may increase indirectly due to the medications used in patients with a variety of symptoms such as acid-related disorders, diarrhea, intestinal infection, anxiety, or insomina, even if those medicines are not themselves direct treatments for the primary diseases. The use of acid-related disease therapeutic drugs, such as proton pump inhibitors (PPIs), contributed to multidrug administration. Several limitations associated with the present study warrants mention. Since the number of target patients was relatively small, the analysis was divided into three arbitrary classes. Therefore, some patient characteristics that should have been extracted may have been overlooked. Another limitation is ethnicity, as our data were mainly obtained from Japanese patients, which may limit the extrapolation of our results to other populations. Finally, because this was a cross-sectional study, what interventions for drug use should be implemented to eliminate polypharmacy should be explored in future prospective studies. ## 5. Conclusions Outpatients with hemodialysis were found to use a large number of drugs. The prescribing patterns for diabetes patients and nondiabetic elderly patients were identified in outpatients with hemodialysis taking multiple medications, and drugs for acid-related disorders, antidiarrheal agents, intestinal anti-inflammatory/anti-infective agents, antithrombotic agents, and psycholeptics were found to be frequently used in those patients. ## References 1. Fulton M.M., Allen E.R.. **Polypharmacy in the elderly: A literature review**. *J. Am. Acad. Nurse Pract.* (2005.0) **17** 123-132. DOI: 10.1111/j.1041-2972.2005.0020.x 2. Gnjidic D., Hilmer S.N., Blyth F.M., Naganathan V., Cumming R.G., Handelsman D.J., McLachlan A.J., Abernethy D.R., Banks E., Le Couteur D.G.. **High-risk prescribing and incidence of Frailty among older community-dwelling men**. *Clin. Pharmacol. Ther.* (2012.0) **91** 521-528. DOI: 10.1038/clpt.2011.258 3. Kojima T., Akishita M., Nakamura T., Nomura K., Ogawa S., Iijima K., Eto M., Ouchi Y.. **Polypharmacy as a risk for fall occurrence in geriatric outpatients**. *Geriatr. Gerontol. Int.* (2011.0) **12** 425-430. DOI: 10.1111/j.1447-0594.2011.00783.x 4. Kojima T., Akishita M., Kameyama Y., Yamaguchi K., Yamamoto H., Eto M., Ouchi Y.. **High risk of adverse drug reactions in elderly patients taking six or more drugs: Analysis of inpatient database**. *Geriatr. Gerontol. Int.* (2012.0) **12** 761-762. DOI: 10.1111/j.1447-0594.2012.00868.x 5. **Guidelines for Medical Treatment and Its Safety in the Elderly. The Japan Geriatrics Society**. (2015.0) 6. Schmidt I.M., Hübner S., Nadal J., Titze S., Schmid M., Bärthlein B., Schlieper G., Dienemann T., Schultheiss U.T., Meiselbach H.. **Patterns of medication use and the burden of polypharmacy in patients with chronic kidney disease: The German Chronic Kidney Disease study**. *Clin. Kidney J.* (2019.0) **12** 663-672. DOI: 10.1093/ckj/sfz046 7. Santos-Díaz G., Pérez-Pico A., Suárez-Santisteban M., García-Bernalt V., Mayordomo R., Dorado P.. **Prevalence of Potential Drug–Drug Interaction Risk among Chronic Kidney Disease Patients in a Spanish Hospital**. *Pharmaceutics* (2020.0) **12**. DOI: 10.3390/pharmaceutics12080713 8. **JSDT Renal Data Registry 2019. Japanese: The Japanese Society for Dialysis Therapy** 9. **Chapter 11: International Comparison, the 2018 USRDS Annual Data Report. Washington: United States Renal Data System**. (2018.0) 10. Imai E., Horio M., Iseki K., Yamagata K., Watanabe T., Hara S., Ura N., Kiyohara Y., Hirakata H., Moriyama T.. **Prevalence of chronic kidney disease (CKD) in the Japanese general population predicted by the MDRD equation modified by a Japanese coefficient**. *Clin. Exp. Nephrol.* (2007.0) **11** 156-163. DOI: 10.1007/s10157-007-0463-x 11. Chiu Y.-W., Teitelbaum I., Misra M., de Leon E.M., Adzize T., Mehrotra R.. **Pill Burden, Adherence, Hyperphosphatemia, and Quality of Life in Maintenance Dialysis Patients**. *Clin. J. Am. Soc. Nephrol.* (2009.0) **4** 1089-1096. DOI: 10.2215/CJN.00290109 12. Nagano N., Ito K., Honda M., Sunaga S., Tagahara A., Nohara T., Nohara A., Hoshi A., Tamei N., Ando T.. **The magnesium included as a pharmaceutical excipient in phosphate binders might affect the serum magnesium levels of dialysis patients**. *Nihon Toseki Igakkai Zasshi* (2016.0) **49** 571-580. DOI: 10.4009/jsdt.49.571 13. Juul-Larsen H.G., Christensen L.D., Bandholm T., Andersen O., Kallemose T., Jørgensen L.M., Petersen J.. **Patterns of Multimorbidity and Differences in Healthcare Utilization and Complexity among Acutely Hospitalized Medical Patients (≥65 Years)—A Latent Class Approach**. *Clin. Epidemiol.* (2020.0) **12** 245-259. DOI: 10.2147/CLEP.S226586 14. Lanza S.T., Collins L.M., Lemmon D.R., Schafer J.L.. **PROC LCA: A SAS Procedure for Latent Class Analysis**. *Struct. Equ. Model. Multidiscip. J.* (2007.0) **14** 671-694. DOI: 10.1080/10705510701575602 15. Herr M., Robine J.-M., Pinot J., Arvieu J.-J., Ankri J.. **Polypharmacy and frailty: Prevalence, relationship, and impact on mortality in a French sample of 2350 old people**. *Pharmacoepidemiol. Drug Saf.* (2015.0) **24** 637-646. DOI: 10.1002/pds.3772 16. **Dosage Recommendations for Drugs That Require the Most Attention in Renal Impairment. The Japanese Society of Nephrology and Pharmacotherapy** 17. Wasse H., Gillen D.L., Ball A.M., Kestenbaum B.R., Seliger S.L., Sherrard D., Stehman-Breen C.O.. **Risk factors for upper gastrointes-tinal bleeding among end-stage renal disease patients**. *Kidney Int.* (2003.0) **64** 1455-1461. DOI: 10.1046/j.1523-1755.2003.00225.x 18. Lin X.-H., Lin C.-C., Wang Y.-J., Luo J.-C., Young S.-H., Chen P.-H., Hou M.-C., Lee F.-Y.. **Risk factors of the peptic ulcer bleeding in aging uremia patients under regular hemodialysis**. *J. Chin. Med. Assoc.* (2018.0) **81** 1027-1032. DOI: 10.1016/j.jcma.2018.03.007 19. Hosseininasab Nodoushan S.A., Nabavi A.. **The Interaction of Helicobacter pylori Infection and Type 2 Diabetes Mellitus**. *Adv. Biomed. Res.* (2019.0) **8** 15. DOI: 10.4103/abr.abr_37_18 20. Nishihara M., Hirata S., Izumi S., Furukubo T., Ota M., Fujita M., Yamakawa T., Tanaka K.. **A survey of constipation in dialysis patients**. *Nihon Toseki Igakkai Zasshi* (2004.0) **37** 1887-1892. DOI: 10.4009/jsdt.37.1887 21. Jong I.-C., Tsai H.-B., Lin C.-H., Ma T.-L., Guo H.-R., Hung P.-H., Hung K.-Y.. **Close correlation between the ankle-brachial index and symptoms of depression in hemodialysis patients**. *Int. Urol. Nephrol.* (2017.0) **49** 1463-1470. DOI: 10.1007/s11255-017-1598-7 22. Ohtake T., Kobayashi S., Moriya H., Negishi K., Okamoto K., Maesato K., Saito S.. **High Prevalence of Occult Coronary Artery Stenosis in Patients with Chronic Kidney Disease at the Initiation of Renal Replacement Therapy: An Angiographic Examination**. *J. Am. Soc. Nephrol.* (2005.0) **16** 1141-1148. DOI: 10.1681/ASN.2004090765 23. Leskinen Y., Salenius J.P., Lehtimäki T., Huhtala H., Saha H.. **The prevalence of peripheral arterial disease and medial arterial calcification in patients with chronic renal failure: Requirements for diagnostics**. *Am. J. Kidney Dis.* (2002.0) **40** 472-479. DOI: 10.1053/ajkd.2002.34885 24. Takeuchi H., Uchida H.A., Kakio Y., Okuyama Y., Okuyama M., Umebayashi R., Wada K., Sugiyama H., Sugimoto K., Rakugi H.. **The Prevalence of Frailty and its Associated Factors in Japanese Hemodialysis Patients**. *Aging Dis.* (2018.0) **9** 192-207. DOI: 10.14336/AD.2017.0429 25. Ballew S.H., Chen Y., Daya N.R., Godino J.G., Windham B.G., McAdams-DeMarco M., Coresh J., Selvin E., Grams M.E.. **Frailty, Kidney Function, and Polypharmacy: The Atherosclerosis Risk in Communities (ARIC) Study**. *Am. J. Kidney Dis.* (2017.0) **69** 228-236. DOI: 10.1053/j.ajkd.2016.08.034 26. Jung H.W., Yoo H.J., Park S.Y., Kim S.W., Choi J.Y., Yoon S.J., Kim C.H., Kim K.I.. **The Korean version of the FRAIL scale: Clinical fea-sibility and validity of assessing the frailty status of Korean elderly**. *Korean J. Intern. Med.* (2016.0) **31** 594-600. DOI: 10.3904/kjim.2014.331 27. Woo J., Leung J.. **Multi-morbidity, dependency, and frailty singly or in combination have different impact on health outcomes**. *Age* (2013.0) **36** 923-931. DOI: 10.1007/s11357-013-9590-3 28. Jamsen K.M., Bell J.S., Hilmer S.N., Kirkpatrick C.M.J., Ilomäki J., Le Couteur D., Blyth F.M., Handelsman D.J., Waite L., Naganathan V.. **Effects of Changes in Number of Medications and Drug Burden Index Exposure on Transitions Between Frailty States and Death: The Concord Health and Ageing in Men Project Cohort Study**. *J. Am. Geriatr. Soc.* (2016.0) **64** 89-95. DOI: 10.1111/jgs.13877 29. Saum K.-U., Schöttker B., Meid A.D., Holleczek B., Haefeli W.E., Hauer K., Brenner H.. **Is Polypharmacy Associated with Frailty in Older People? Results From the ESTHER Cohort Study**. *J. Am. Geriatr. Soc.* (2016.0) **65** e27-e32. DOI: 10.1111/jgs.14718
--- title: 'Interventions to Increase Physical Activity in Community-Dwelling Older Adults in Regional and Rural Areas: A Realist Synthesis Review Protocol' authors: - Stephen Cousins - Rebecca McKechnie - Patricia Jackman - Geoff Middleton - Tshepo Rasekaba - Irene Blackberry journal: Methods and Protocols year: 2023 pmcid: PMC10037574 doi: 10.3390/mps6020029 license: CC BY 4.0 --- # Interventions to Increase Physical Activity in Community-Dwelling Older Adults in Regional and Rural Areas: A Realist Synthesis Review Protocol ## Abstract The importance of physical activity (PA) for the health and wellbeing of older adults is well documented, yet many older adults are insufficiently active. This issue is more salient in regional and rural areas, where evidence of the most critical components of interventions that explain PA participation and maintenance in older populations is sparse. This realist review will [1] systematically identify and synthesise literature on PA interventions in community-dwelling older adults in regional and rural areas, and [2] explore how and why those interventions increase PA in that population. Using a realist synthesis framework and the behaviour change wheel (BCW), context–mechanism–outcome (C-M-O) patterns of PA interventions for older adults in regional and rural areas will be synthesised. Thematic analysis will be employed to compare, contrast, and refine emerging C-M-O patterns to understand how contextual factors trigger mechanisms that influence regional and rural community-dwelling older adults’ participation in PA interventions. This realist review will be the first to adopt a BCW analysis and a realist synthesis framework to explore PA interventions in community-dwelling older adults in regional and rural areas. This review will provide recommendations for evidence-based interventions to improve PA participation and adherence by revealing the important mechanisms apparent in this context. Systematic review registration: (PROSPERO CRD42023402499). ## 1. Introduction By 2050, the global population of adults over 60 years of age will increase beyond 2 billion [1]. Population aging is further accentuated in regional and rural areas, with patterns of migration showing an out-migration of the younger population to metropolitan areas and an in-migration of older adults to regional and rural areas [2]. For instance, $35\%$ of Australians aged ≥ 65-years live outside metropolitan centers [3] and $24.3\%$ of English adults aged ≥ 65-years reside in rural locations [4]. Advancing age heightens the risk of developing chronic, degenerative health issues, including musculoskeletal, cardiovascular, and respiratory diseases, along with cancer, diabetes, cognitive decline, and multimorbidity [5,6]. These conditions lead to a loss of functional independence, capability, and quality of life in older adults [7,8]. Health disparities within older populations are also evident based on one’s location of residence. People living in regional and rural areas report poorer health outcomes and higher rates of morbidity [9], lower rates of leisure-time physical activity [10], as well as higher rates of obesity [11] and lower quality of life and social functioning [12] compared to their counterparts residing in metropolitan areas. These health disparities are likely associated with some of the unique challenges faced by older adults in regional and rural areas, such as reduced access to health care services (exacerbated by the challenges of transport and distance), fewer local amenities and infrastructure, health workforce shortages, and greater likelihood of suffering from social isolation and loneliness [4,13]. Consequently, there is a recognisable need to improve access to health care services and provide sustained health promotion opportunities for older adults living in regional and rural areas. There is now considerable evidence demonstrating that regular and adequate participation in both active and passive forms of physical activity (PA) can help mitigate or manage health conditions in older adults and promote healthy aging [14,15]. Indeed, even a “low dose” of 75 min of moderate to vigorous intensity PA per week can improve health outcomes for those aged 60 years and above [16]. Engaging in regular PA provides additional benefits relevant to older adults, including numerous psychological benefits [17], improved cognitive function [18], and protective effects against the impacts of dementia and Alzheimer’s disease [19]. Similar positive effects have also been demonstrated in older adults in regional and rural community settings, with improvements in PA, physical function, and psychological indices reported [20]. Furthermore, PA participation can promote “social connectiveness” amongst older adults [21], which is particularly important for those living in regional and rural areas due to the greater risk of experiencing social isolation and loneliness because of their age and location. Although the potential benefits of PA are widely documented, older populations continue to be at greater risk of physical inactivity compared to middle-aged or younger adults [22]. For example, only $55\%$ of men and $41\%$ of women aged 75 years and above meet PA guideline recommendations in the UK [23], with a little over $25\%$ of Australians aged over 65 years meeting their respective guidelines [24]. These contrast with rates of 61–$82\%$ among people aged 18–49 years [25]. Despite the well-known benefits of PA for healthy aging and the promotion of specific PA guidelines for older adults [26], there is little certainty of the most critical characteristics or components that explain the efficacy of PA interventions for older adults, including those living in regional and rural areas, especially in terms of how and why they are effective. Furthermore, researchers have also highlighted the need for greater understanding of the complex environmental and contextual factors underlying PA uptake and maintenance in older populations [27]. Therefore, given the lack of clarity surrounding the key determinants of effective PA interventions in community-dwelling older adults in rural and regional areas, research is required to advance understanding of the contextual factors and mechanisms underlying PA engagement in this population to better inform policy and practice. ## Justification for This Review Realist synthesis is a methodology that extends the scope of a traditional narrative or systematic review. Traditional systematic review approaches to evaluating interventions, such as meta-analysis, are generally employed to determine whether an intervention has been effective, and to what extent, but can lack explanatory power [28]. Given the complex, dynamic, and multi-faceted nature of interventions, it is important to explain the contextual factors and mechanisms associated with specific outcomes [29]. Thus, a review methodology that seeks to understand and unravel the underlying complexities of effective interventions is required to develop greater understanding of PA interventions that appear to be efficacious for community-dwelling older adults in rural and regional areas. Realist synthesis has emerged as a widely used strategy for understanding complex health and social interventions [30,31]. The purpose of a realist review is to go beyond examining intervention effectiveness to develop a fine-grained understanding of how an intervention works, for whom, and in what contexts [32]. Thus, by adopting a realist synthesis methodology, this approach offers an opportunity to uncover the mechanisms behind reported improvements in PA levels for older adults living in regional and rural areas. By doing so, this will clarify how improvements occur, who benefits exclusively, and what contexts (i.e., circumstances) are particularly important for interventions to be effective. The objective of this review is to systematically identify and synthesise literature on PA interventions in community-dwelling older adults in regional and rural areas to explore how and why those interventions increase PA in that population. A realist synthesis will be conducted to address the following questions: [1] what are the contexts and mechanisms that increase physical activity among community-dwelling older adults in regional and rural areas; [2] what interventions are feasible, sustainable, and effective to be implemented for community-dwelling older adults residing in regional and rural areas; and [3] what support structures are required to improve the delivery of these interventions to this target population? ## 2. Experimental Design *In* general, PA interventions are underpinned by assumptions of how they work to bring about their intended outcome(s). A realist review uses a systematic and theory-driven approach to refine these assumptions into theories, which can then be empirically tested. In a realist synthesis, the theory of how a program “works” is structured according to the “context–mechanism–outcome” (C-M-O) approach [32]. That is, the program theory is explained as the contextual (C) factor in addition to a potential resource mechanism (Mresource), hypothesized to have triggered the relevant mechanism response (the underlying process or behavior) (Mresponse) to generate the outcome of interest (O) [33]. The process of our realist review is focused on identifying, explaining, and testing these semi-predictable C-M-O patterns (called demi-regularities). For example, a theory could be proposed that for community-dwelling older adults residing in a regional or rural area (C), participating in a PA intervention that has a dedicated expert delivering the program who educates and encourages the participants (Mresource) increases the participants’ confidence in their physical capabilities (Mresponse), thereby facilitating regular participation in PA (O). This realist review will be based on the approach of Pawson et al. [ 32] and will be consistent with publication standards for realist reviews (RAMESES criteria) [34]. The process will focus on identifying, explaining, and testing semi-predictable patterns or demi-regularities in C-M-O configurations. Stages will comprise literature search and screening, quality assessment, data extraction, data analysis and synthesis, and dissemination. An overview of the stages of the review is presented in Figure 1. ## 3.1. Stage 1: Systematic Literature Search and Screening A systematic literature search will be conducted to identify relevant studies of interventions aimed at increasing PA behaviours in community-dwelling older adults residing in regional and rural areas. Relevant literature will be obtained by the lead author (SC) by searching four electronic search databases: [1] CINAHL, [2] Embase, [3] Medline, and [4] SPORTDiscus. Systematic search strings will be designed using terms targeting three major constructs: [1] older adult, [2] physical activity/exercise/sport, and [3] community/regional/rural locations and will be combined with the AND operator. Similar key terms will be entered and separated by the term OR and truncation (*) used to capture all possible variations of selected key terms, as follows: (old* adult*/person*/people*/men*/women* OR old age OR aged OR aged, 80 and over OR senior* OR elder*), (physical activit* OR exercise* OR sport*), (rural* OR regional* OR community*/populat*/area*/town*/city*/location*). In addition to electronic database searches, the reference lists of included articles will be screened to identify any further articles that satisfy the eligibility criteria. Studies published up to July 2022, with no limit on earliest year of publication, will be included if they: [1] sample community-dwelling older adults (≥65 years), with or without diagnosed illness, living in regional/rural areas, regardless of previous experience or exposure to PA, exercise, or sport, of which at least $50\%$ of participants are ≥65 years; [2] evaluate the outcomes of PA programs/interventions/behaviours, including active forms of intermittent and work-related physical activity (i.e., housework, gardening, manual labour), exercise, and sport; [3] contain original data; and [4] are published in the English language. Articles will not be restricted by country, study design, or outcome measures. Articles will be excluded if: [1] no outcomes of PA programs/interventions/behaviours are included; [2] the geographic location of the intervention (urban vs. rural) is not specified; [3] the intervention only includes passive forms of exercise and/or is rehabilitation- or treatment-focused; [4] the study is not published in English or is not an original investigation; and [5] <$50\%$ of participants are ≥65-years, or terms including “seniors”, “elderly”, or “older adult” are not used. Identified article citations and abstracts will be uploaded to Covidence (SC), a web-based collaborative software platform that streamlines production of systematic and other literature reviews (Veritas Health Innovation, Melbourne, Australia—www.covidence.org, accessed on 1 June 2022). After removal of duplicates, articles will be screened in a two-stage process. Stages 1 and 2, respectively, will comprise title and abstract screening and full article review by two reviewers (SC and RM), with each article reviewed independently against the eligibility criteria by both reviewers. Articles will be accepted or rejected based on consensus. Discrepancies in decisions to include or exclude an article will be resolved by a third independent reviewer (TR), whereby the majority decision results in a study being included or excluded. A flow diagram of this systematic search process, recording excluded (with reasons) and included studies for data synthesis, will be included when presenting the results of this review. ## 3.2. Stage 2: Quality Assessment Each study will be assessed independently by two reviewers for rigour, risk of bias, and outcome quality, using the relevant Joanna Briggs Institute (JBI) appraisal tool based on the study design employed. JBI critical appraisal tools provide a systematic approach for assessing the methodological quality of a study and the extent to which it has addressed the potential for bias within its design, conduct, and analysis [35]. Results from each reviewer’s quality assessment will be discussed to assist with subsequent analysis, synthesis, and interpretation of findings. ## 3.3. Stage 3a: Data Extraction Each article will undergo independent data extraction by any two reviewers from the research team. This will be followed by a team meeting during which the reviewer pair presents extracted data to the rest of the team, after which discussion and refinement of the extracted data will take place. Extracted data will include the following: authors’ details, year of publication, country, participant details including comorbidity and functional status, study design, and intervention details (setting, mode of delivery, follow-up, comparison groups), outcome measures, main findings (including physical activity outcomes), and hypothesised contextual components and mechanisms (how the intervention may have “worked” to trigger change). A copy of the data extraction form template is provided as supplementary material (see Supplementary File S1). Collating this information will allow for studies to be grouped in terms of similar participants, interventions, or setting characteristics as required during data analysis. ## 3.4. Stage 3b: Data Analysis and Synthesis C-M-O configurations will be developed by each respective reviewer using the data extraction template, and will include:Context (intervention, setting, or participant characteristic to which a mechanism may be applied).Mechanism, resource (a strategy or approach applied to a given context).Mechanism, response (an intermediate outcome in direct response to the mechanism resource).Outcome (the desired/intended or measured output). C-M-O patterns arising will be synthesised to develop program theories using the behaviour change wheel (BCW) [36] and realist synthesis framework [36]. Program theories will provide a structure for exploring the complex relationships between health interventions and outcomes, often involving diagrams or flow charts that convey the relationships between contextual factors, mechanisms, and outcomes [37]. In accordance with the realist synthesis approach [38], thematic analysis will be employed to compare, contrast, and refine emerging C-M-O configurations within and across studies to identify demi-regularities to understand how contextual factors trigger the mechanisms that influence community-dwelling older adults’ participation in PA programs. In addition, iterative content assessment between members of the research team, all of whom bring different theoretical and applied expertise, will stimulate critical discussions that assist in reaching consensus and refuting or refining the proposed C-M-O theories, whilst offering an opportunity to explore new theoretical propositions that can be generated from the data. Findings will be interpreted within the context of the BCW to assist in characterising interventions and linking intervention components to changes in PA behaviour. ## 3.5. Dissemination of Findings Upon completion, findings of this review will be shared with academic and non-academic communities through peer-reviewed journal publications, conference presentations, presentations to relevant stakeholders and practitioners, and using the social media platforms of the authors’ affiliated institutions. It is important that the findings of the review, if indicating valuable information for health service delivery, are shared widely to enable the framework and recommendations for evidence-based interventions generated through this research to be translated into practice, with the ultimate goal of increasing PA in community-dwelling older adults in regional and rural areas. To the researchers’ knowledge, this review will also be the first to adopt a BCW analysis and realist synthesis framework to assess PA interventions in community-dwelling older adults in regional and rural areas; therefore, the review process, as well as the review findings, will be included in any presentation or dissemination process. ## Limitations Realist synthesis is not designed to report on the effectiveness of an intervention based solely on quantifiable outcomes, but rather is designed to contextualise success—or lack thereof—and the mechanisms that produce these outcomes. As such, frameworks for interventions may be developed, but definitive answers to research questions may not be achieved. Despite best efforts to describe the review process here, the iterative nature of a realist synthesis may see delineation from the review process as new themes are developed. As such, this BCW analysis and realist synthesis protocol is considered a flexible tool with which to initiate and guide the process, maintain the aim and scope, and establish the transparency of the process. Our adoption of a systematic literature search may increase reliance on published data above grey literature and, as such, this synthesis may be vulnerable to publication bias. Furthermore, by only including English-language publications, our review is also susceptible to language bias. ## 5. Summary This protocol paper provides an account of our intended process in undertaking a realist synthesis and BCW analysis to understand which interventions assist in improving PA behaviours among community-dwelling older adults residing in rural and regional areas, why, amongst whom, and in what contexts. Existing evidence has previously been limited to older adult populations or rural and regional dwellers and has focussed on the effectiveness of programs, with limited insight into the contexts and mechanisms through which they may work. As such, the findings of this realist synthesis will provide rich insight into the mechanisms that underpin successful (or unsuccessful) interventions and the contexts in which they work. By doing so, this will extend our understanding of facilitators and barriers to implementing PA interventions among older adults in rural and regional areas and will ultimately provide relevant and timely recommendations for evidence-based interventions to improve PA uptake and potential health outcomes among this target population. ## References 1. **World Population Ageing Report**. (2015) 2. Winterton R., Warburton J.. **Models of care for socially isolated older rural carers: Barriers and implications**. *Rural Remote Health* (2011) **11** 1678. DOI: 10.22605/RRH1678 3. 3. Australian Institute of Health and Welfare Older AustraliansAIHWCanberra, Australia2021Available online: https://www.aihw.gov.au/reports/older-people/older-australia-at-a-glance/contents/diverse-groups-of-older-australians/regional-remote-communities(accessed on 20 January 2022). *Older Australians* (2021) 4. **Rural Ageing**. (2018) 5. Prasad S., Sung B., Aggarwal B.B.. **Age-associated chronic diseases require age-old medicine: Role of chronic inflammation**. *Prev. Med.* (2012) **54** 29-37 6. Chatterji S., Byles J., Cutler D., Seeman T., Verdes E.. **Health, functioning, and disability in older adults- present status and future implications**. *Lancet* (2015) **385** 563-575. DOI: 10.1016/S0140-673661462-8 7. Marventano S., Ayala A., Gonzalez N., Rodríguez-Blázquez C., Garcia-Gutierrez S., Forjaz M.J.. **Multimorbidity and functional status in community-dwelling older adults**. *Eur. J. Intern. Med.* (2014) **25** 610-616. DOI: 10.1016/j.ejim.2014.06.018 8. Jindai K., Nielson C.M., Vorderstrasse B.A., Quiñones A.R.. **Multimorbidity and Functional Limitations Among Adults 65 or Older, NHANES 2005–2012**. *Prev. Chronic Dis.* (2016) **13** E151. DOI: 10.5888/pcd13.160174 9. Schofield D., Earnest A.. **Demographic change and the future demand for public hospital care in Australia, 2005 to 2050**. *Aust. Health Rev.* (2006) **30** 507. DOI: 10.1071/AH060507 10. Trivedi T., Liu J., Probst J.C., Merchant A., Jones S., Martin A.. **Obesity and obesity-related behaviors among rural and regional adults in the USA**. *Rural Remote Health* (2015) **15** 3267. PMID: 26458564 11. Cohen S.A., Greaney M.L., Sabik N.J.. **Assessment of dietary patterns, physical activity and obesity from a national survey: Rural-regional health disparities in older adults**. *PLoS ONE* (2018) **13**. DOI: 10.1371/journal.pone.0208268 12. Baernholdt M., Yan G., Hinton I., Rose K., Mattos M.. **Quality of life in rural and regional adults 65 years and older: Findings from the National Health and Nutrition Examination survey**. *J. Rural Health* (2012) **28** 339-347. DOI: 10.1111/j.1748-0361.2011.00403.x 13. **National Strategic Framework for Rural and Remote Health. Australia**. (2012) 14. Nelson M., Rejeski W., Blair S., Duncan P., Judge J.O., King A., Macera C., Castaneda-Sceppa C.. **Physical activity and public health in older adults: Recommendation from the American College of Sports Medicine and the American Heart Association**. *Med. Sci. Sports Exerc.* (2007) **39** 1435-1445. DOI: 10.1249/mss.0b013e3180616aa2 15. Bonanni R., Cariati I., Romagnoli C., D’Arcangelo G., Annino G., Tancredi V.. **Whole Body Vibration: A Valid Alternative Strategy to Exercise?**. *J. Funct. Morphol. Kinesiol.* (2022) **7**. DOI: 10.3390/jfmk7040099 16. Hupin D., Roche F., Gremeaux V., Chatard J.-C., Oriol M., Gaspoz J.-M., Barthélémy J.-C., Edouard P.. **Even a low-dose of moderate-to-vigorous physical activity reduces mortality by 22% in adult’s aged ≥60 years: A systematic review and meta-analysis**. *Br. J. Sports Med.* (2015) **49** 1262-1267. DOI: 10.1136/bjsports-2014-094306 17. Hamer M., Stamatakis E., Steptoe A.. **Dose-response relationship between physical activity and mental health: A Scottish Health Survey**. *Br. J. Sports Med.* (2009) **43** 1111-1114. DOI: 10.1136/bjsm.2008.046243 18. Falck R., Davis J., Best J., Crockett R., Liu-Ambrose T.. **Impact of exercise training on physical and cognitive function among older adults: A systematic review and meta-analysis**. *Neurobiol. Aging* (2019) **79** 119-130. DOI: 10.1016/j.neurobiolaging.2019.03.007 19. De la Rosa A., Olaso-Gonzalez G., Arc-Chagnaud C., Millan F., Salvador-Pascual A., García-Lucerga C., Blasco-Lafarga C., Garcia-Dominguez E., Carretero A., Correas A.G.. **Physical exercise in the prevention and treatment of Alzheimer’s disease**. *J. Sport Health Sci.* (2020) **9** 394-404. DOI: 10.1016/j.jshs.2020.01.004 20. Moore M., Warburton J., O’Halloran P.D., Shields N.. **Effective community-based physical activity interventions for older adults living in rural and regional areas: A systematic review**. *J. Aging Phys. Act.* (2016) **24** 158-167. DOI: 10.1123/japa.2014-0218 21. Schrempft S., Jackowska M., Hamer M., Steptoe A.. **Associations between social isolation, loneliness, and objective physical activity in older men and women**. *BMC Public Health* (2019) **19**. DOI: 10.1186/s12889-019-6424-y 22. Sallis J.F., Bull F., Guthold R., Heath G.W., Inoue S., Kelly P., Oyeyemi A.L., Perez L.G., Richards J., Hallal P.C.. **Progress in physical activity over the Olympic quadrennium**. *Lancet* (2016) **388** 1325-1336. DOI: 10.1016/S0140-673630581-5 23. **Health Survey for England. Adult’s Health-Related Behaviours**. (2018) 24. **4364.0.55.001—Australian Health Survey: First Results, 2017–2018. Canberra**. (2019) 25. Du Y., Liu B., Sun Y., Snetselaar L.G., Wallace R.B., Bao W.. **Trends in Adherence to the Physical Activity Guidelines for Americans for Aerobic Activity and Time Spent on Sedentary Behavior Among US Adults, 2007 to 2016**. *JAMA Netw. Open* (2019) **2** e197597. DOI: 10.1001/jamanetworkopen.2019.7597 26. **Physical Activity and Sedentary Behaviour: A Brief to Support Older People**. (2022) 27. Zubala A., MacGillivray S., Frost H., Kroll T., Skelton D.A., Gavine A., Gray N.M., Toma M., Morris J.. **Promotion of physical activity interventions for community dwelling older adults: A systematic review of reviews**. *PLoS ONE* (2017) **12**. DOI: 10.1371/journal.pone.0180902 28. Pawson R., Bellamy J., Popay J.. **Realist synthesis: An explanatory focus for systematic review**. *Moving Beyond Effectiveness in Evidence Synthesis: Methodological Issues in the Synthesis of Diverse Sources of Evidence* (2006) 83-94 29. Rycroft-Malone J., McCormack B., Hutchinson A.M., DeCorby K., Bucknall T.K., Kent B., Schultz A., Snelgrove-Clarke E., Stetler C.B., Titler M.. **Realist synthesis: Illustrating the method for implementation research**. *Implement. Sci.* (2012) **7** 33. DOI: 10.1186/1748-5908-7-33 30. Hewitt G., Sims S., Harris R.. **The realist approach to evaluation research: An introduction**. *Int. J. Ther. Rehabil.* (2013) **19** 250-259. DOI: 10.12968/ijtr.2012.19.5.250 31. Pawson R.. **Evidence-based policy: The promise of ‘realist synthesis’**. *Evaluation* (2002) **8** 340-358 32. Pawson R., Greenhalgh T., Harvey G., Walshe K.. **Realist review—A new method of systematic review designed for complex policy interventions**. *J. Health Serv. Res. Policy* (2005) **10** 21-34. DOI: 10.1258/1355819054308530 33. Pawson R.. **Realist synthesis: New protocols for systematic review**. *Evidence-Based Policy* (2006) 73-104 34. Wong G., Greenhalgh T., Westhorp G., Buckingham J., Pawson R.. **RAMESES publication standards: Realist synthesis**. *BMC Synth.* (2013) **11** 35. Munn Z., Moola S., Riitano D., Lisy K.. **The development of a critical appraisal tool for use in systematic reviews addressing questions of prevalence**. *Int. J. Health Policy Manag.* (2014) **3** 123-128. DOI: 10.15171/ijhpm.2014.71 36. Michie S., van Stralen M., West R.. **The behaviour change wheel: A new method for characterising and designing behaviour change interventions**. *Implement. Sci.* (2011) **6** 42. DOI: 10.1186/1748-5908-6-42 37. Pawson R., Tiley N.. *Realistic Evaluation* (1997) 38. Wong G., Westhorp G., Pawson R., Greenhalgh T.. *Realist Synthesis. RAMESES Training Material* (2013) 39. Wong G., Westhorp G., Manzano A., Greenhalgh J., Jagosh J., Greenhalgh T.. **RAMESES II reporting standards for realist evaluations**. *BMC Med.* (2016) **14**. PMID: 27342217 40. Wong G., Greenhalgh T., Westhorp G., Pawson R.. **Realist methods in medical education research: What are they and what can they contribute?**. *Med. Educ.* (2012) **46** 89-96. PMID: 22150200 41. Macfarlane F., Greenhalgh T., Humphrey C., Hughes J., Butler C., Pawson R.. **A new workforce in the making?: A case study of strategic human resource management in a whole-system change effort in healthcare**. *J. Health Organ. Manag.* (2011) **25** 55-72. PMID: 21542462 42. Lacouture A., Breton E., Guichard A., Ridde V.. **The concept of mechanism from a realist approach: A scoping review to facilitate its operationalization in public health program evaluation**. *Implement. Sci.* (2015) **10** 153. PMID: 26519291
--- title: Reassessing the Abundance of miRNAs in the Human Pancreas and Rodent Cell Lines and Its Implication authors: - Guihua Sun - Meirigeng Qi - Alexis S. Kim - Elizabeth M. Lizhar - Olivia W. Sun - Ismail H. Al-Abdullah - Arthur D. Riggs journal: Non-Coding RNA year: 2023 pmcid: PMC10037588 doi: 10.3390/ncrna9020020 license: CC BY 4.0 --- # Reassessing the Abundance of miRNAs in the Human Pancreas and Rodent Cell Lines and Its Implication ## Abstract miRNAs are critical for pancreas development and function. However, we found that there are discrepancies regarding pancreatic miRNA abundance in published datasets. To obtain a more relevant profile that is closer to the true profile, we profiled small RNAs from human islets cells, acini, and four rodent pancreatic cell lines routinely used in diabetes and pancreatic research using a bias reduction protocol for small RNA sequencing. In contrast to the previous notion that miR-375-3p is the most abundant pancreatic miRNA, we found that miR-148a-3p and miR-7-5p were also abundant in islets. In silico studies using predicted and validated targets of these three miRNAs revealed that they may work cooperatively in endocrine and exocrine cells. Our results also suggest, compared to the most-studied miR-375, that both miR-148a-3p and miR-7-5p may play more critical roles in the human pancreas. Moreover, according to in silico-predicted targets, we found that miR-375-3p had a much broader target spectrum by targeting the coding sequence and the 5′ untranslated region, rather than the conventional 3′ untranslated region, suggesting additional unexplored roles of miR-375-3p beyond the pancreas. Our study provides a valuable new resource for studying miRNAs in pancreata. ## 1. Introduction The pancreas is an essential metabolic organ composed of, among other cell types, endocrine and exocrine cells that undergo unique stages during differentiation. This process is controlled by the expression of specific transcription factors. Multipotent progenitor cells in the pancreatic bud first differentiate into tip progenitors which then develop into acinar cells and trunk progenitor cells. The latter further develop into ductal cells, another major type of exocrine cells, and endocrine progenitor cells. Next, endocrine progenitor cells differentiate into islets of Langerhans containing glucagon-producing alpha, insulin-producing beta, somatostatin-producing delta, ghrelin-producing epsilon, and pancreatic polypeptide-producing PP cells, which produce essential hormones to control glucose homeostasis. Endocrine cell differentiation is controlled by Neurogenin 3 (NEUROG3 or NGN3) and other islet cell-specific factors [1,2]. While many transcription factors essential to the process of pancreatic development and function have been identified (listed in Supplemental File S1) [2,3,4], pancreatic development and function can also be modulated by other regulatory molecules such as microRNAs (miRNAs). Long and short non-coding RNAs, in concert with coding RNAs, function to orchestrate gene regulation [5,6]. miRNAs are small non-coding RNAs with biological activities [7]. miRNAs are known to modulate important processes in the pancreas both in models of good health and disease [8,9,10]. However, an analysis of published high-throughput sequencing datasets of small RNAs from the pancreas has revealed inconsistent results in the population of pancreatic miRNAs and their expression levels, especially miRNAs that are highly expressed in pancreata. In this study (events were depicted in Figure 1A), using a previously developed bias-reduction small RNA deep sequencing (smRNAseq) protocol, we profiled small RNA from eight pairs of human acinar and islet cells and identified miR-375-3p (hereinafter referred to as miR-375), miR-148a-3p, and miR-7-5p as the most highly abundant miRNAs in human pancreatic cells, in contrast to the generally accepted idea that miR-375 was the most abundant pancreatic miRNA that has been most studied. Because the four rodent pancreatic cell lines, including mouse alpha-TC1, beta-TC-6, MIN6, and rat INS-1, have been routinely used in pancreatic research and diabetes studies, we also profiled small RNAs in these cells. Due to the difficulties in elucidating the potential roles of these three miRNAs in human pancreas cell development and differentiation in vivo, in the current research, we performed in silico studies of miRNA-target interactions between these three miRNAs and their predicted and validated targets in pancreatic genes. According to their predicted targets, we found that both miR-148a-3p and miR-7-5p had a broader target spectrum than miR-375-3p for target sites located in the traditional 3′ untranslated region (3′UTR) that are favored by most miRNAs. However, when taking the target sites in the coding sequence (CDS) or the 5′ untranslated region (5′UTR) into consideration, miR-375-3p has a much broader target spectrum than miR-148a-3p or miR-7-5p. Most predicted miR-375 target sites are in CDS or 5′UTR. This result suggests that miR-375 may have additional unexplored roles in the pancreas and beyond it. According to their validated targets in highly expressed pancreatic genes and essential pancreatic genes (major transcription factors and key products), all three miRNAs play a critical role in the pancreas. Pathway analysis using the above-validated targets showed that miR-7-5p plays a more significant role in insulin pathways than miR-148a-3p, and miR-375-3p plays a less important role among the three miRNAs. We hope the results from this study will provide a metric for future in vivo studies of the miRNA regulation of human pancreatic development, differentiation, and function. ## 2.1. The Results of Published miRNA Profiles Indicate Discrepancies for Both miRNAs Expressed in Pancreases and Their Abundance in These Datasets About two decades ago, a large number of miRNAs were identified and cloned. Since then, numerous miRNA profiling datasets in different species and organs have been generated by different techniques and documented. Among them, the small RNA deep sequencing results are highly appreciated, as the sequencing approach can simultaneously profile almost all species of small RNAs as well as different isoforms of a given small RNA. However, there are limited human pancreas small RNA profiling data generated by smRNAseq and other techniques in the GEO database, probably due to the difficulty of accessing fresh human tissues in general. We reanalyzed one dataset containing smRNAseq results from both human whole islets and sorted alpha and beta cells (GSE52314 by Kameswaran et al.) [ 11]. The reanalysis employed miRge [12,13] and the analyzed results were compared to the published results. Due to the low sequencing depth for the three islet datasets (SRR1028929, SRR1028930, and SRR1028931) in GSE52314, they were pooled and treated as the miRNA profile of islets (hereafter referred to as miRNAs of islets). Because refined profiling methodologies (including both the sequencing device and the library construction protocol for smRNAseq) were used for sorted alpha and beta cell samples, datasets from alpha (SRR1028924) and beta (SRR1028925) cells have much higher sequencing depths than all three islet samples. Because alpha and beta cells account for the majority of human islet cells, we treated the average of the datasets of alpha and beta cells as the estimated miRNA profile of islets (hereafter referred to as estimated miRNAs of islets). Although our reanalysis of islet smRNAseq data agreed with published data in general (Figure 1B, “miRge” versus “Ori ref” in “Islets”, “Alpha”, “Beta”), indicating that the approach of our smRNAseq data analysis performs similarly to the approach used in the data analysis in original publications, there are differences between the miRNAs of islets (Figure 1B, “Islets”) and the estimated miRNAs of islets (Figure 1B, “Alpha + Beta”). A profile of miRNAs of islets revealed that let-7 family members accounted for over $50\%$ of the total miRNAs and miR-375 was the second-most abundant miRNA, representing $13\%$ of the total miRNAs in the pancreas. In contrast, in the estimated miRNAs of islets, the most abundant miRNA was miR-375, accounting for ~$38\%$ of total miRNAs, while miR-7-5p was the second-most abundant miRNA, representing $7\%$ of total miRNAs in the islets. The percentage drop in let-7 family members in the estimated miRNAs of islets was accompanied by a percentage increase in several miRNAs, including miR-127-3p, miR-191-5p, miR-99b-5p, miR-26, miR-27, miR-125a-5p, and miR-148a-3p. While there is a possibility that the sorting of alpha and beta cells from islets may have an effect on their profiling results, deeper sequencing methodology used for sequencing alpha and beta cells likely accounts for these differences, and the estimated miRNAs of islets may be closer to the true miRNA profiles of islets. There are also differences between the published data and our reanalyzed data, mainly in the abundance of let-7 family members, as indicated by the ratio of miRNAs in alpha cells and beta cells (Figure 1B). These variations may have arisen from the use of different mapping algorithms with different mismatch requirements, which could generate different total read counts for an miRNA with multiple family members. To rule out potential problems that could have been caused by analysis tools for small RNA profiling, we reanalyzed another dataset that contained whole islets and sorted the beta cells (GSE47720 by van de Bunt et al.) [ 14]. Then, we compared our analyzed data with the curated data in the miRmine database that contains analyzed data for all six runs in GSE47720 that have been generated using different data analyzing tools [15]. We first compared runs with high sequence reads, specifically SRR873381 for islets and SRR873401 for beta cells, which are likely more reliable due to their higher sequencing depth. Both results analyzed by miRge and miRmine agreed well with each other, further indicating that our analyzed result is not biased by the approach we have used to analyze smRNAseq data. Although miR-375 is still shown to be the most abundant miRNA in the pancreas and covered more than $40\%$ of total miRNA populations in the pancreas, the profiling data showed several differences compared to GSE52314. Notably, miR-7-5p, a miRNA that covered ~$7\%$ of the total miRNA reads from the pancreas and was the second-most abundant miRNA in the estimated miRNAs of islets in GSE52314 (Figure 1A, “Alpha + Beta”), dropped to ~$0.03\%$ in the dataset of GSE47720, representing an over 200-fold reduction in abundance (Figure 1B, “Islets” and “Beta”). The low expression level of miR-7-5p in the dataset of GSE47720 also conflicts with the published qPCR result showing that miR-7-5p was an islet-enriched miRNA [16] and an islet-specific microRNA during human pancreatic development [17]. Next, we compared the average of datasets in GSE47720 from all three runs (SRR873381, SRR871609, and SRR871652) for islets to the average of datasets from all three runs (SRR873401, SRR873410, and SRR871601) for beta cells, regardless of the sequencing depth, using the analyzed data by miRmine (Figure 1C, “Islets-all” and “Beta-all”). In the data averaged from our combined runs, miR-7-5p accounted for 4 to $5\%$ of total miRNAs in the islets, which represents a greater than 100-fold increase in abundance than in the deeper runs alone, again highlighting the possible impact of different sequencing methodologies on the outcome of miRNA abundance. Furthermore, pooling data from runs using different sequencing protocols may complicate the profiling outcomes, as in the case of miR-143-3p, which was found to be the second-most abundant islet miRNA ($16.9\%$ of the total) in the pooled data. Published miRNA profile data acquired via arrays also showed that, while miR-375 was the most abundant miRNA in alpha cells and beta cells, miR-7-5p was not among the 10 most plentiful types [18], suggesting that result heterogeneity can also be found in analyses where miRNA microarray techniques were employed. We also looked at miRNA profiling data in a human tissue miRNA database (miRNATissueAtlas2) and found that the expression of miR-375 is ranked 5th at 16 thousand reads per million mapped reads (RPM), behind the 26.4 thousand RPM of let-7b-5p, 35.4 thousand RPM of miR-26a-5p, 37.3 thousand RPM of let-7a-5p, and 98 thousand RPM of miR-143-3p in the human pancreas, despite being considered pancreas-specific and the most highly expressed pancreatic miRNA [19]. Therefore, it is possible that other small RNA profiling datasets for the pancreas also contain inconsistent small RNA profiling data and this kind of inconsistency in data may also exist in datasets of other human tissues. Despite the inconsistent miRNA profiling data from GSE47720 and GSE52314, the two datasets were extensively used for several miRNA functional studies [8,11,14,20,21,22]. Therefore, to provide a better metric for miRNA abundance in the pancreas for future studies on miRNA function in the human pancreas, a less biased miRNA profile from high-quality tissues closer to the true profile and more relevant to related studies is needed. **Figure 1:** *Bias in published pancreatic miRNA profiling data. (A) Flowchart of events in our experiment; (B) Top expressed miRNAs (by percentage of total miRNA) in islets, alpha, and beta cells in the dataset of GSE47720/PRJNA193453 as described in the original publication and reanalyzed by miRge; (C) Top expressed miRNAs (by percentage of total miRNA) in islets and beta cells in the dataset of GSE52314/PRJNA227380 that were reanalyzed by miRge or miRmine. “Islets” and “Beta” represent data from runs with high sequencing depth and “Islets-all” and “Beta-all” represent average RPM from all three runs for islets and beta cells regardless of sequencing depth. All data in both tables A and B are in percentage, 1% represents RPM at 10,000. To permit comparison by different analysis tools and data sources, the RPM of an miRNA was defined as the reads of an miRNA per one million miRNA reads. This was not normalized to all reads mapped to the genome in a deep sequencing run. Abbreviations: GEO = Gene Expression Omnibus, GSE = GEO Series; SRA = Sequence Read Archive; SRR = SRA run number; ref = reference; miRge and miRmine are two computing tools that were used to analyze small RNA deep sequencing data.* ## 2.2. Less Biased Small RNA Profiles in Human Acinar and Islet Cells Variations in pancreatic miRNA profiling data can arise from numerous causes, such as the different extents of RNA degradation in different samples related to the freshness, cause of death, and heterogenicity of the samples, as well as the methods used to acquire and process samples and the robustness of the sequencing technique [23]. It is very challenging to get high-quality fresh human samples due to the nature of human samples in that they need to be from brain-dead cadaveric donors that have gone through many regulation and processing steps, including hospitalization, transportation, and isolation prior to arriving at research laboratories. Among all the confounding factors, smRNAseq methodologies were found to be the major cause of result bias in small RNA profiling. Possible reasons for smRNAseq data heterogeneity include ligase bias towards certain sequences, biased PCR amplification of ligated products, unknown effects of small RNA modification on ligation or reverse transcription, unexpected small RNA sequence interactions with cloning adaptor sequences, and the use of different algorithms for analyzing smRNAseq data [23,24,25]. Even bearing these aspects in mind, a 10 to 100-fold variation in miRNA levels is not acceptable as these variations could confound conclusions from studies using datasets from such miRNA profiling and mislead the designs of future experiments depending on results from such profiling. To this end, we sequenced small RNAs in acinar cells and corresponding islets from human pancreata of eight organ donors using a bias reduction protocol for smRNAseq that we previously developed (Table 1) [25,26]. Most importantly, we have taken advantage of the unique opportunity of having an islet transplantation program in our institute to obtain high-quality human samples as fresh as possible and at relatively large numbers for the current study (Figure S1). Historically, miR-375 was thought to be the most abundant pancreatic miRNA, which is supported by many of the published pancreatic miRNA profiles, and is the most studied pancreatic miRNA, especially using mouse models. However, the analysis of miRNA abundance in our smRNAseq dataset, as reads per million mapped reads (RPM), revealed that miR-148a-3p, miR-375, and miR-7-5p probably were the most abundant pancreatic miRNAs (Figure 2A). We found that miR-148a-3p was the most abundant miRNA in human acinar cells and miR-148a-3p, miR-375, and miR-7-5p were the most abundant miRNAs in human islets. miR-26a/b-5p and miR-27a/b-3p were equally well expressed in acinar cells and islets. In contrast, miR-148a-3p, miR-21, and miR-217 were significantly lower in islets compared to acinar cells. The abundant miRNAs, miR-148a-3p, miR-375, and miR-7-5p also showed great variation in their expression levels between acinar cells and islets. miR-148a-3p was reduced from $48.25\%$ in acinar cells to about $15.92\%$ in islets, while miR-375 was increased from $4.75\%$ in acinar cells to $16.91\%$ in islets, and miR-7-5p was increased from <$0.7\%$ in acinar cells to $20.18\%$ in islets (Figure 2B,C, Supplementary File S2). We validated the smRNAseq results using small RNA qRT-PCR (Figure 2D). The small RNA qRT-PCR results were normalized to the miR-26a-5p expression in each sample and were compared to normalized results in the smRNAseq data (ratio of RPM of a miRNA to RPM of miR-26a-5p). miR-26a-5p was chosen for normalization because it has a similar expression in acini and islets and is highly expressed in both tissues (Figure 2A). The qRT-PCR data agreed well with smRNAseq data for the relative expression level of each miRNA in acinar cells versus islets (Figure 2D,E). However, there are instances where the smRNAseq data and qRT-PCR data do not agree very well. For example, qRT-PCR showed that both miR-375 and miR-21-5p exhibited a relatively higher expression level compared to smRNAseq. It has been a known issue that small RNA qRT-PCR data does not always agree with deep sequencing data. While qRT-PCR data are usually normalized to an miRNA or several miRNAs, deep sequencing data have the power to be normalized to all miRNA reads or all small RNA reads in a deep sequencing run. In this regard, deep sequencing data will have the advantage of representing the relative abundance of an miRNA. The other reason might be due to the nature of smRNAseq technology which can detect multiple isoforms of an miRNA. While smRNAseq with high sequencing depth may survey many isoforms of an miRNA, the detection of a certain high expression level sequence may be saturated. Conversely, due to the short length of mature miRNA sequences, qRT-PCR primarily detects one specific isoform with higher detecting ranges. Since we have a relatively large small RNA profiling dataset, we also attempted to identify novel pancreatic miRNAs in these samples, especially those that are highly expressed. Several candidate novel pancreatic miRNAs were predicted using miRge2 (sequence library matches sequences in miRBase Release 22.1, May 2018) [12]. Some of these predicted miRNAs were specifically expressed in acini or islets (Supplementary File S3). We characterized four predicted pancreatic miRNAs that had high read counts (Supplementary Figure S2). We designated these miRNAs miR-P1 to P4. miR-P1 was found at a 3 to 1 ratio in acini versus islets, and miR-P4 was found at a 1 to 1 ratio in acini versus islets. Both miR-P2 and miR-P3 were found only in islets. We used their mature sequences to query miRNAs in the latest version of miRBase (Release 22.1, October 2018, June 2021) and noted that miR-P1 matched with miR-802-3p, miR-P2 matched with hsa-mir-7-3-5p, miR-P3 matched with hsa-mir-153-1-5p, and miR-P4 matched with hsa-mir-452-3p. These miRNAs were likely deemed novel by miRge due to their precursor sequences. Alternatively, these miRNAs may have been included in the latest version of miRBase that was released after the release of miRge2 and updated thereafter [12,27]. Nonetheless, our data indicate that the most highly expressed miRNAs in the pancreas have probably been identified already. Additionally, miR-P2 results support our finding that miR-7-5p was present in greater amounts in islets compared to acini. Combining smRNAseq data from acini and islets together and classifying them by miRNA families, we found that miR-148a-3p was the most abundant miRNA in the human pancreas, followed, from the most to the least, by miR-375, miR-7-5p, miR-26-5p, let-7-5p, miR-21-5p, miR-30-5p, miR-200-3p, miR-27-3p, miR-143-3p, miR-217, miR-99-5p, miR-215-5p/192-5p, and miR-101-5p. Together, these miRNAs formed the top 14 abundant miRNAs/miRNA families in the pancreas (Figure 3A). Differentially expressed gene analysis using DESeq2 showed that there were many other miRNAs differentially expressed between acini and islets (Figure 3B). Cluster and heatmap analyses showed that miR-148a-3p, miR-375, and miR-7-5p were clustered together as the miRNAs with the highest expression, followed by the second-most highly expressed miRNA cluster, which included miR-30a-5p, miR-217, miR-127-3p, miR-221-3p, miR-148a-5p, and miR-216b-5p (Figure 3C). Interestingly, miR-7-5p and miR-217 had the greatest change in expression level between acinar cells and islets in both qRT-PCR and smRNAseq, consistent with miRNA microarray expression data in rat acinar cells and islets [16]. This result suggests a potential role for these miRNAs in the separation of tip progenitors from trunk progenitors. ## 2.3. Comparing Small RNA Profiles in Human Acini and Islets with Small RNA Profiling Data from Rodent Acini and Islets Rodents have been widely used in pancreas and endocrine research despite differences in tissue cytoarchitecture, endocrine cell proportions, and signaling between human and murine cells and organs [28]. We compared our human miRNA profiling data with a set of published murine miRNA profiles summarized in a comprehensive review of miRNAs and their functions in the pancreas [8]. Specially, we compared our profiling results of human acinar cells and islets with mouse acinar miRNA profiling data in GSE81260 [29] and mouse islet miRNA profiling data (no raw dataset in GEO) [30]. As observed in our human acinar cell miRNA profiling, mouse miR-148a-3p (mmu-miR-148a-3p) was the most abundant miRNA in murine acini. However, while the hsa-miR-26 family is the second-most abundant miRNA in human acinar cells, mmu-miR-375 is the second-most abundant miRNA in murine acini (Figure S3A). As observed in our human islet miRNA profiling, miR-148a-3p, miR-375, and miR-7-5p are also the three most abundant miRNAs in murine islets (Figure S3B). Due to their higher rate of proliferation and insulin production, rodent cell lines were more routinely used in pancreas and diabetes research than human cell lines. Therefore, we also profiled small RNAs in four widely used rodent cell lines, murine pancreatic alpha cell lines alpha-TC1, murine beta cell line beta-TC-6 and MIN6, and rat beta-like INS1 cell line. Our profiling results showed that the three mouse cell lines, alpha-TC1, beta-TC-6, and MIN6, had similar miRNA profiles. However, there are major differences between our miRNA profiling data and the published MIN6 cell smRNAseq data in GEO (GSE44262), with the exception of miR-375 being shown to be the top expressed miRNA in our profiling results for all three murine cell lines and the results for MIN6 cells from GSE44262 [20] (Figure S3C,D). In our profiling data, mmu-miR-7-5p and mmu-miR-26a-5p were the second- and third-most abundant miRNAs in all three cell lines, whereas data for MIN6 cells from GSE44262 showed that mmu-miR-7-5p was expressed at very low levels compared to mmu-miR-375. Instead, mmu-miR-182-5p and mmu-miR-27b-3p were ranked the second and third most abundant miRNAs in data for MIN6 cells from GSE44262 (Figure S3C,D). Notably, mmu-miR-148a-3p was expressed at low levels in these murine insulinoma cell lines, consistent with its role as a tumor suppressor gene that is downregulated in cancers [31,32,33]. In the rat beta-like INS1 cell line, rat miR-375 (rno-miR-375), rno-miR-7-5p, rno-miR-26a-5p, and rno-miR-148-3p were the four most abundant miRNAs similar to their abundance in human pancreas samples (Figure S3E and Figure 3A). INS-1 is also an insulinoma cell line. However, unlike mmu-miR-148a-3p expression in the three murine cell lines, rno-miR-148-3p was expressed at a relatively high level (Figure S3C,E). Published array data also showed that rno-miR-217 is abundant in rat acinar cells versus islets and agrees with our data on human acinar cells and islets [16]. In summary, our human pancreatic miRNA profiling data is consistent with published profiling data from rodent cells regarding dominant miRNAs in both acinar cells and islets. The presence of mmu-miR-7-5p in our MIN6 profiling data, an miRNA that was discriminately profiled by different smRNAseq methodologies, as indicated in our reanalysis of published data, indicated that our rodent cell line profiling may also be closer to their true expression level. However, we acknowledge that we only performed one replicate of sequencing for each murine cell line and the two donated cell lines (MIN-6 and INS-1) by other laboratories were not characterized for their identification. ## 2.4. Potential Roles of miR-375, miR-7-5p, and miR-148a-3p in the Human Pancreas According to Their Predicted Targets Acinar cells and islets are derived from tip and trunk progenitor cells, respectively [1,2]. Conceivably, differentially expressed miRNAs between acini and islets, such as miR-148a-3p, miR-7-5p, and miR-217, may play a role in cell fate decisions during this differentiation process [9]. *In* general, miRNAs function by regulating specific target molecules, such as transcription factors, by targeting their 3′UTR [34]. A number of transcription factors that participate in pancreatic development and differentiation were identified and described [1,2,3,4]. Based on this data, we generated a list of pancreatic transcription factors and functional genes and designated them as pancreas-essential genes (Supplementary File S1). Our initial analysis included targets predicted by the TargetScan algorithm (version 7.2 and 8.0), which used the seed sequence of an miRNA or a family of miRNAs (nucleotides #2 to 8 from 5′ end) to locate conserved complementary sequences in the 3′UTR regions of genes [35]. Targeting conserved sequences that cross several species is an indication of a bona fide target of an miRNA. Unexpectedly, miR-375, which has been widely studied and is believed to be critical for and specific to the pancreas, was found to target only a few of the designated pancreas-essential genes [36,37]. TargetScan predicted more targets for miR-148a-3p and miR-7-5p than for miR-375 with several among the pancreas-essential genes (Figure 4A). Therefore, we looked at miRNA targets in miRDB which include both predicted and validated targets [38]. miRDB (version 2020) gave similar results in that the number of targets of miR-148a-3p and miR-7-5p were greater than the number of targets of miR-375 (Figure 4B). Next, upon employing miRWalk, which covers all seed matched sites in CDS, 5′UTR, and 3′UTR [39], we found that the target sites of miR-148a-3p and miR-7-5p predicted by miRWalk (new version 2022) compared to the target sites predicted by TargetScan (version 8.0) increased by about 10-fold (from about 1000 to about 10,000), but the number of miR-375 target sites increased more than 70-fold (from about 500 to about 36,000). The increase in miR-375 targets was presumably due to most new targeting genes having miR-375 target sites located in their CDS or 5′ UTR regions. When seed sequences in CDS, 5′UTR, and 3′UTR were counted, more pancreas-essential genes were noted to be potential targets of miR-375 compared to miR-148a-3p and miR-7-5p (Figure 4C). This result suggests that miR-375 may employ a different targeting mechanism than the other two miRNAs. In contrast to the other two miRNAs, which are also highly expressed in other organs or tissues, miR-375 is only highly expressed in the pancreas, suggesting that miR-375 may be the most important pancreatic miRNA even though it is not the sole dominant miRNA in the pancreas. It is also conceivable that circulating miRNA-375, usually found in serum and exosomes, may behave like insulin, is promiscuous, and may signal beyond its function in the pancreas [40,41,42]. The broad target spectrum of miR-375 may be the result of evolution selection between miRNA-375 and its broad targets in the pancreas and beyond. The Venn *Diagram analysis* did not yield a single essential gene targeted by all three miRNAs (Figure 4). To focus on genes with conserved miRNA target sites that have a better chance to have biological functions, TargetScan was employed to assess the predicted genes of miR-375, miR-148a-3p, and miR-7-5p that are also pancreatic cell type-specific genes identified by single cell RNA sequencing in each pancreatic cell group [3,4]. Coinciding with its greater abundance in the pancreas, TargetScan (version 7.2) predicted more pancreas-essential gene targets of miR-148a-3p than miR-375 or miR-7-5p in the nine major pancreas cell types (Supplementary File S6). Here, we also discovered several essential pancreatic genes targeted by more than one miRNA. For example, miR-375 and miR-7-5p were predicted to target GATA-6 in acinar and ductal cells and PAX6 in alpha and PP cells. miR-375 and miR-148a-3p were predicted to target klf5 in acinar and ductal cells and ELAVL4 in beta and delta cells. miR-148a-3p and miR-7-5p may target EGFR and ZNF704 in acinar and ductal cells and MAP1B in beta and delta cells. NGN3 and MAFB, which are known to promote endocrine cell differentiation, were predicted targets of miR-148a-3p in alpha, beta, and PP cells, supporting a potential role for this miRNA in the differentiation of exocrine and endocrine cells. ## 2.5. Potential Roles of miR-7-5p, miR-148-3p, and miR-375 in the Human Pancreas According to Their Experimentally Validated Targets Next, we looked at target genes of miR-7-5p, miR-148-3p, and miR-375 that are highly expressed in islets (islet-genes 5 RPKM and up, based on an unpublished RNA sequencing dataset of human islets, the list of islet-genes is provided in Supplementary Data File S1) and the aforementioned list of essential genes for the pancreas (regardless of their expression level). We used multiMiR to retrieve the interactions between these miRNAs and their targets from validated targets documented in the integrated multiple microRNA-target databases [43] that contain experimentally validated miRNA-target interactions curated by miRecords [44], miRTarBase [45], and TarBase [46]. There are 920, 506, and 368 validated targets for miR-7-5p, miR-148-3p, and miR-375, respectively, in the list of islet-genes (Figure 5A). In agreement with the results for predicted targets, miR-375 still has the least number of validated targets, but in contrast to only a few common targets of the three miRNAs, there are 111 islet-genes and two essential genes (IRS1 and SOX4) that are targeted by all three miRNAs (SOX4 is also in the list of islet-genes) in the validated targets. Among the targeted essential genes, PSIP1 is targeted by both miR-375 and miR-148a-3p, GATA6 is targeted by both miR-7-5p and miR-148a-3p, ERRFI1 is targeted by both miR-375 and miR-7-5p, and all three genes are also in the list of islet-genes. miR-375 has 9 validated targets that are essential genes and 6 of them are in the list of islet-genes; miR-7-5p has 10 validated targets that are essential genes and 9 of them are in the list of islet-genes; miR-148a-3p has 9 validated targets that are essential genes and 4 of them are in the list of islet-genes (Figure 5B,C). Therefore, the three miRNAs have an almost equal number of validated targets in the list of essential genes, and some essential genes are targeted by two or all three miRNAs. Next, using their validated targets in the list of islet-genes, we performed a comparison KEGG pathway analysis. It is not surprising to find that miR-7-5p regulates more pathways than miR-148a-3p, and that both miR-7-5p and miR-148a-3p regulate more pathways than miR-375 does, probably because the former two have more validated targets in the list of islet-genes. This analysis also revealed that miR-7-5p regulates the insulin signaling pathway more significantly than miR-148a-3p, and miR-375 has less regulation for this pathway, indicating that miR-7-5p plays a more important role in insulin regulation than miR-148a-3p, and miR-375 seems to participate less in the insulin signaling pathway (Figure S4). However, all three miRNAs regulate the FoxO signaling pathway that regulates apoptosis, cell-cycle control, glucose metabolism, the AEG-RAGE signaling pathway in diabetic complications, and the EGFR signaling pathway (Figure 5D and Figure 6). The KEGG pathway analysis also revealed that the validated targets of the three miRNAs are most involved in glucose uptake, glycogenesis, proliferation, and differentiation—especially proliferation and differentiation, as revealed by Pathview [2017] analysis—and support their important regulatory roles in pancreas development [47] (Figure 6 and Figure S4). Taken together, the three major pancreatic miRNAs may work together to regulate pancreatic cell development and cell differentiation by targeting different pancreatic genes and essential pancreatic genes. ## 3. Discussion Taking advantage of our unique opportunity to access high-quality human pancreas tissue and our bias reduction small RNA profiling methodology, we have profiled small RNAs from eight paired human samples of acinar cells and islets and four rodent cell lines. Our small RNA profiling data of human pancreatic tissues showed that miR-148a-3p was highly expressed in acini and islets, being the most abundant miRNA in acini, and miR-148a-3p, miR-375, and miR-7-5p were found to be the three most abundant human islets miRNAs. Our data also showed that miR-7-5p is the most abundant miRNA in human islets, a finding that is distinct from published miRNA sequencing data and microarray data [16]. Furthermore, in contrast to the very low expression level of miR-148a-3p in both datasets of GSE52314 and GSE47720, our profiles showed that miR-148a-3p is the most abundant pancreatic miRNA and present in near-equal abundance to miR-375 and miR-7-5p in human islets. Multiple factors that are essential for pancreatic cell differentiation or pancreatic development have been predicted and/or validated to be the targets of these miRNAs, supporting the idea that combined targeting by miR-148a-3p, miR-375, and miR-7-5p could play a role in pancreatic cell development and differentiation. The function of miRNAs closely correlates with their abundance in cells. Highly expressed miRNAs have more chances to find and bind their targets and higher potency to repress their targets [48]. The overall miRNA expression level is controlled by transcription, processing to mature forms, and miRNA half-life. miRNAs can be produced rapidly and persist for minutes to weeks [49]. However, due to the numerous types, small size, and variation in half-life, miRNA profiling is challenging using traditional cloning, RT-PCR, microarray, and deep sequencing technologies. Among them, cloning methodologies are labor-intensive, costly, and restricted to the identification of highly expressed miRNAs in the early era of miRNA studies. Both RT-PCR and microarray technologies have similar limitations in that they rely on known small RNA sequences to identify their presence in new samples. High-throughput sequencing technologies can profile both known sequences and novel sequences, simplify small RNA profiling, increase capacity, reveal different isoforms and their abundance, and reduce costs, and have become the preferred small RNA profiling methodology. However, potential bias in the sampling of small RNA sequences has resulted in inconsistent smRNAseq data. Variations in small RNA profiling data from tissue samples are unavoidable due to the multiple cell types that make up complex tissues and organs. Variation in sample handling and RNA isolation methods also contributes to data heterogeneity. Our smRNAseq data showed that miR-7-5p, which easily degrades with a short half-life [50], was the most abundant miRNA in human islets, which has been shown to be a low abundance pancreatic miRNA in some published datasets, implying that the RNA samples we used for sequencing were of high quality and our sequencing protocol performed well in surveying all members of small RNAs. The high expression level of miR-148a-3p in the pancreas, elevated levels of miR-7-5p, and decreased levels of miR-148a-3p in endocrine cells compared to exocrine cells are compatible with the distinct proliferative rates of exocrine and endocrine cells. The meaning of the high levels of miR-148a-3p in acini or pancreas remains to be experimentally determined. Presumably, it may play a critical role in the pancreas since the majority of the pancreatic mass consists of acini and ductal cells [51]. In summary, reanalysis of published datasets and de nova analysis of human pancreata and rodent pancreas cell lines revealed discrepancies in miRNA expression patterns. The high abundance of miR-375, miR-148a-3p, and miR-7-5p in the pancreas revealed by our less biased small RNA profiling methodology using high-quality human samples, and the special targeting mechanism of miR-375 and its ability to bind Toll-like receptors [52], suggests possible regulatory roles by the three miRNAs in pancreatic development, cell differentiation, and diabetes. There is a limitation in that we did not profile small RNA from all types of pancreatic cells, such as ductal cells and all members of islets (alpha, beta, delta, epsilon, and PP cells) in the current study due to the heterogeneity in pancreatic cells caused by subgroups in each cell type [53] and the technical challenges in single cell smRNAseq analysis [54,55]. There is another limitation in our study in that the acini and islets we used were of high quality but not as pure as sorted cells. Therefore, cross contamination of acini by islets, or islets by acini were unavoidable. However, fresh tissues have the advantage of reduced RNA degradation over decomposed tissues. In this regard, single cell smRNAseq analysis, which also needs islets to be dissociated into single cells, will also have the risk of RNA degradation. Nonetheless, either acini or islet samples must go through the process of collagenase digestion of the pancreata. Therefore, how we processed the samples may have impacted the profiling data we have obtained in the current study. Regarding the most highly expressed miRNAs in acini or islets, the high level of miR-375 in acini may have some carryover from islets and the high level of miR-148a-3p in islets may have some carryover from acini. However, the almost 30-fold expression level of miR-7-5p ($20.18\%$) in islets over in acini ($0.68\%$) indicates that the cross contamination between acini and islets in our samples is very low. Finally, we only analyzed eight pairs of samples with high islets quality, enough for statistical analysis, but we have not been able to consider other factors that could influence the outcome of the profiling, including gender, age, race, BMI, etc., and the outcome may also be dependent on the protocol used for tissue preparation. Based on the profiling data, we propose a model in which miRNA may control human pancreas development. Specifically, miR-375 may be specific for pancreas development, the expression of miR-148a-3p in the pancreas may guide lineage separation of exocrine cells from endocrine cells, and the expression of miR-7-5p may reenforce the fate of endocrine cells., Both miR-148a-3p and miR-7-5p or miR-7-5p alone may work with miR-375 to promote subtype cell differentiation in islets. ## 4. Materials and Methods All methods were carried out in accordance with relevant guidelines and regulations. ## 4.1. RNA Isolation Aliquots of islets and acini were frozen and stored at −80 °C for later RNA isolation. RNA was isolated from tissues and cultured cell lines using TRIzol (ThermoFisher, Waltham, MA, USA). The manufacturer’s instructions were followed throughout. RNA quality was determined, and amounts quantified using a Nanodrop and an Agilent Bioanalyzer. ## 4.2. Small RNA Deep Sequencing, Reads Processing, and Data Analysis Small RNA deep sequencing was carried out using a customized protocol [25,26]. Briefly, 100 ng to 1.0 µg of RNA was used to construct small RNA libraries for single reads, flow cell cluster generation, and 51 cycle (51-nt) sequencing on a HiSeq 2500 (Illumina, San Diego, CA, USA). The sequence depth was 30M reads per sample, 8 samples per lane (by barcoding), and was performed by the Integrative Genomics Core of City of Hope National Medical Center. To construct small RNA libraries, the 5′ adaptor used in the Illumina (San Diego, CA, USA) Truseq small RNA deep sequencing protocol was replaced with a customized 5′ adaptor by adding 3 random nucleotides (nts) at the 3′ end of the original 5′ adaptor to reduce bias [25,26]. The 3′ adaptor provided in the kit was used for 3′ end ligation and sample barcoding. Raw smRNAseq reads for each sample were separated by barcodes. Then the 5′ adaptor with the three random nts was removed prior to further processing and analyzing using miRge [12,13]. Unnormalized miRNA read counts generated by miRge were inputted into the Bioconductor package DESeq2 (version 1.30.0) in R (version 4.0.2) to generate the differential expressed miRNA results (islet versus acinus) at the default setting [the Wald-test was applied to assess the p value for differential gene expressions and the adjusted p value (p-adj) was obtained using the method of Benjamini and Hochberg] [56]. ## 4.3. Small RNA qRT-PCR We followed a published S-Poly(T) small RNA qRT-PCR detection protocol with some modification for multiplex in a reverse transcription step [26,57]. Briefly, 100 ng DNase I treated total RNA was poly-A tailed using the poly-A tailing kit from Epicentre (Madison, WI, USA). Since miR-26a-3p and miR-27a-3p had similar abundance in acini and islets, miRNA-26a-3p was selected as an RNA sample loading control to normalize relative expression of other miRNA, and miR-27a-3p was used to validate similar miRNA-26a-3p expression levels across different samples. Reverse transcription primers for all eight miRNAs were pooled and used in one reverse transcription reaction. Reverse transcription primers for miRNAs were designed according to the most abundant isoforms of a given miRNA in miRBase. All oligoes were from Integrated DNA Technologies (Coralville, IA, USA). Universal TaqMan probe and PCR mixtures were employed as published [57]. The ΔCq (quantitation cycle) value of an miNRA relative to miR-26a-3p in each sample was calculated as CqmiRNA-CqmiRNA-26a-3p. The value of 2ΔCt was calculated as the relative fold change of an miRNA to miRNA-26a-3p. miRNA qRT-PCR primers are list in Supplementary Table S1. ## 4.4. Isolation of Human Pancreatic Islets and Acinar Cells Brain-dead cadaveric donor pancreata were obtained from a local organ procurement organization. Organs were shipped after flushing with cold University of Wisconsin preservation solution. Pancreatic islet and acinar isolation protocol was followed to obtain these cells as described [58]. Briefly, pancreata were digested using collagenase supplemented with either thermolysin or neutral protease. Then, the digested pancreatic tissue was collected, washed, and purified in a cooled COBE 2991 cell processor (COBE Laboratories Inc., Lakewood, CA, USA). After islet purification, the acinar cells were collected from the remaining digested pancreatic tissue. As a standard procedure, following each isolation, immunofluorescent staining was used to check the purity of human islets and acinar cells. Islets were stained with insulin and acinar cells were stained with amylase to show the distinct cell population harvested during islets isolation (Figure S1). Acinar samples had <$1\%$ islets. Islets were cultured at 37 °C with $5\%$ CO2 for 24–72 h before assessing purity. The use of human pancreatic tissues was approved by the Institutional Review Board of City of Hope National Medical Center. Informed consent was obtained from the legal next of kin of each donor. Donor information is provided in Table 1. ## 4.5. Cell Lines and Culture Conditions Murine alpha-TC1 (ATCC #CRL-2934) and beta-TC-6 (ATCC # CRL-11506) cells were purchased from the American Type Culture Collection (Manassas, VA, USA). MIN6 cells were a kind gift from Dr. Jun-ichi Miyazaki (Osaka University Medical School, Osaka, Japan). Rat insulinoma INS-1 cells were a kind gift from Dr. Fouad Kandeel (City of Hope National Medical Center, Duarte, CA, USA). Both MIN-6 and INS-1 cell lines were not characterized for their identity. All cell lines were cultured at 37 °C in a humidified atmosphere containing $5\%$ CO2 in medium supplemented with 2 mM penicillin/streptomycin. Alpha-TC1 cells were cultured in DMEM supplemented with $10\%$ heat inactivated fetal bovine serum (FBS), 15 mM HEPES, 0.1 mM non-essential amino acids, $0.02\%$ BSA, 0.5 g/L Sodium Bicarbonate, and 2.0 g/L Glucose. Beta-TC-6 and MIN6 cells were cultured in DMEM supplemented with $15\%$ heat inactivated FBS. INS-1 cells were cultured in RPMI-1640 supplemented with $10\%$ FBS, 10 mM Hepes, 2 mM L-glutamine, 2 g/L Glucose, 1 mM sodium pyruvate, and 0.05 mM 2-mercaptoethanol. ## 4.6. Computational Resources R version 4.02, R studio, and many R or Bioconductor software packages were used for data analysis and visualization. ComplexHeatmap was used to generate the heatmaps [59], the EnhancedVolcano R package was used to generate the volcano plots [60], and clusterProfiler for KEGG pathway analysis [61]. Pathview was used to generate KEGG graphs [47]. Other smRNAseq analysis tools and miRNA targets prediction or document tools used included miRmine [15], miRDB [38], miRwalk [39], TargetScan [35], and multiMiR [43]. ## 4.7. Statistical Analyses Inferential statistics of differential miRNA expression in islets versus acinar cells were calculated using the default setting in DESeq2. Descriptive statistics in miRNA qPCR data or comparison of one miRNA in islets versus acini or in two different data sets were calculated by two tailed t-tests. Unless specified in the figure legends, p-adj < 0.05 for DESeq2 data or p-value < 0.05 for others was used as the cutoff for significance in gene expression. Error is expressed as SD. ## References 1. Murtaugh L.C.. **Pancreas and β-cell development: From the actual to the possible**. *Development* (2007.0) **134** 427-438. DOI: 10.1242/dev.02770 2. Al-Khawaga S., Memon B., Butler A.E., Taheri S., Abou-Samra A.B., Abdelalim E.M.. **Pathways governing development of stem cell-derived pancreatic β cells: Lessons from embryogenesis**. *Biol. Rev. Camb. Philos. Soc.* (2018.0) **93** 364-389. DOI: 10.1111/brv.12349 3. Muraro M.J., Dharmadhikari G., Grun D., Groen N., Dielen T., Jansen E., van Gurp L., Engelse M.A., Carlotti F., de Koning E.J.. **A Single-Cell Transcriptome Atlas of the Human Pancreas**. *Cell Syst.* (2016.0) **3** 385-394.e383. DOI: 10.1016/j.cels.2016.09.002 4. Segerstolpe A., Palasantza A., Eliasson P., Andersson E.M., Andreasson A.C., Sun X., Picelli S., Sabirsh A., Clausen M., Bjursell M.K.. **Single-Cell Transcriptome Profiling of Human Pancreatic Islets in Health and Type 2 Diabetes**. *Cell Metab.* (2016.0) **24** 593-607. DOI: 10.1016/j.cmet.2016.08.020 5. Mattick J.S., Makunin I.V.. **Non-coding RNA**. *Hum. Mol. Genet.* (2006.0) **15** R17-R29. DOI: 10.1093/hmg/ddl046 6. Amaral P.P., Dinger M.E., Mercer T.R., Mattick J.S.. **The eukaryotic genome as an RNA machine**. *Science* (2008.0) **319** 1787-1789. DOI: 10.1126/science.1155472 7. Bartel D.P.. **Metazoan MicroRNAs**. *Cell* (2018.0) **173** 20-51. DOI: 10.1016/j.cell.2018.03.006 8. Filios S.R., Shalev A.. **β-Cell MicroRNAs: Small but Powerful**. *Diabetes* (2015.0) **64** 3631-3644. DOI: 10.2337/db15-0831 9. Agbu P., Carthew R.W.. **MicroRNA-mediated regulation of glucose and lipid metabolism**. *Nat. Rev. Mol. Cell Biol.* (2021.0) **22** 425-438. DOI: 10.1038/s41580-021-00354-w 10. Kaviani M., Azarpira N., Karimi M.H., Al-Abdullah I.. **The role of microRNAs in islet β-cell development**. *Cell Biol. Int.* (2016.0) **40** 1248-1255. DOI: 10.1002/cbin.10691 11. Kameswaran V., Bramswig N.C., McKenna L.B., Penn M., Schug J., Hand N.J., Chen Y., Choi I., Vourekas A., Won K.J.. **Epigenetic regulation of the DLK1-MEG3 microRNA cluster in human type 2 diabetic islets**. *Cell Metab.* (2014.0) **19** 135-145. DOI: 10.1016/j.cmet.2013.11.016 12. Lu Y., Baras A.S., Halushka M.K.. **miRge 2.0 for comprehensive analysis of microRNA sequencing data**. *BMC Bioinform.* (2018.0) **19**. DOI: 10.1186/s12859-018-2287-y 13. Baras A.S., Mitchell C.J., Myers J.R., Gupta S., Weng L.C., Ashton J.M., Cornish T.C., Pandey A., Halushka M.K.. **miRge—A Multiplexed Method of Processing Small RNA-Seq Data to Determine MicroRNA Entropy**. *PLoS ONE* (2015.0) **10**. DOI: 10.1371/journal.pone.0143066 14. van de Bunt M., Gaulton K.J., Parts L., Moran I., Johnson P.R., Lindgren C.M., Ferrer J., Gloyn A.L., McCarthy M.I.. **The miRNA profile of human pancreatic islets and β-cells and relationship to type 2 diabetes pathogenesis**. *PLoS ONE* (2013.0) **8**. DOI: 10.1371/journal.pone.0055272 15. Panwar B., Omenn G.S., Guan Y.. **miRmine: A database of human miRNA expression profiles**. *Bioinformatics* (2017.0) **33** 1554-1560. DOI: 10.1093/bioinformatics/btx019 16. Bravo-Egana V., Rosero S., Molano R.D., Pileggi A., Ricordi C., Dominguez-Bendala J., Pastori R.L.. **Quantitative differential expression analysis reveals miR-7 as major islet microRNA**. *Biochem. Biophys. Res. Commun.* (2008.0) **366** 922-926. DOI: 10.1016/j.bbrc.2007.12.052 17. Joglekar M.V., Joglekar V.M., Hardikar A.A.. **Expression of islet-specific microRNAs during human pancreatic development**. *Gene Expr. Patterns* (2009.0) **9** 109-113. DOI: 10.1016/j.gep.2008.10.001 18. Klein D., Misawa R., Bravo-Egana V., Vargas N., Rosero S., Piroso J., Ichii H., Umland O., Zhijie J., Tsinoremas N.. **MicroRNA expression in alpha and β cells of human pancreatic islets**. *PLoS ONE* (2013.0) **8**. DOI: 10.1371/journal.pone.0055064 19. Keller A., Groger L., Tschernig T., Solomon J., Laham O., Schaum N., Wagner V., Kern F., Schmartz G.P., Li Y.. **miRNATissueAtlas2: An update to the human miRNA tissue atlas**. *Nucleic Acids Res.* (2022.0) **50** D211-D221. DOI: 10.1093/nar/gkab808 20. Baran-Gale J., Fannin E.E., Kurtz C.L., Sethupathy P.. **Beta cell 5′-shifted isomiRs are candidate regulatory hubs in type 2 diabetes**. *PLoS ONE* (2013.0) **8**. DOI: 10.1371/journal.pone.0073240 21. Jin W., Mulas F., Gaertner B., Sui Y., Wang J., Matta I., Zeng C., Vinckier N., Wang A., Nguyen-Ngoc K.V.. **A Network of microRNAs Acts to Promote Cell Cycle Exit and Differentiation of Human Pancreatic Endocrine Cells**. *iScience* (2019.0) **21** 681-694. DOI: 10.1016/j.isci.2019.10.063 22. Wong W.K.M., Joglekar M.V., Saini V., Jiang G., Dong C.X., Chaitarvornkit A., Maciag G.J., Gerace D., Farr R.J., Satoor S.N.. **Machine learning workflows identify a microRNA signature of insulin transcription in human tissues**. *iScience* (2021.0) **24** 102379. DOI: 10.1016/j.isci.2021.102379 23. Tian G., Yin X., Luo H., Xu X., Bolund L., Zhang X.. **Sequencing bias: Comparison of different protocols of microRNA library construction**. *BMC Biotechnol.* (2010.0) **10**. DOI: 10.1186/1472-6750-10-64 24. Fuchs R.T., Sun Z., Zhuang F., Robb G.B.. **Bias in ligation-based small RNA sequencing library construction is determined by adaptor and RNA structure**. *PLoS ONE* (2015.0) **10**. DOI: 10.1371/journal.pone.0126049 25. Sun G., Wu X., Wang J., Li H., Li X., Gao H., Rossi J., Yen Y.. **A bias-reducing strategy in profiling small RNAs using Solexa**. *RNA* (2011.0) **17** 2256-2262. DOI: 10.1261/rna.028621.111 26. Sun G., Cheng Y.W., Lai L., Huang T.C., Wang J., Wu X., Wang Y., Huang Y., Wang J., Zhang K.. **Signature miRNAs in colorectal cancers were revealed using a bias reduction small RNA deep sequencing protocol**. *Oncotarget* (2016.0) **7** 3857-3872. DOI: 10.18632/oncotarget.6460 27. Kozomara A., Birgaoanu M., Griffiths-Jones S.. **miRBase: From microRNA sequences to function**. *Nucleic Acids Res.* (2019.0) **47** D155-D162. DOI: 10.1093/nar/gky1141 28. Cabrera O., Berman D.M., Kenyon N.S., Ricordi C., Berggren P.O., Caicedo A.. **The unique cytoarchitecture of human pancreatic islets has implications for islet cell function**. *Proc. Natl. Acad. Sci. USA* (2006.0) **103** 2334-2339. DOI: 10.1073/pnas.0510790103 29. Dixit A.K., Sarver A.E., Yuan Z., George J., Barlass U., Cheema H., Sareen A., Banerjee S., Dudeja V., Dawra R.. **Comprehensive analysis of microRNA signature of mouse pancreatic acini: Overexpression of miR-21-3p in acute pancreatitis**. *Am. J. Physiol. Gastrointest. Liver Physiol.* (2016.0) **311** G974-G980. DOI: 10.1152/ajpgi.00191.2016 30. Tattikota S.G., Rathjen T., McAnulty S.J., Wessels H.H., Akerman I., van de Bunt M., Hausser J., Esguerra J.L., Musahl A., Pandey A.K.. **Argonaute2 mediates compensatory expansion of the pancreatic β cell**. *Cell Metab.* (2014.0) **19** 122-134. DOI: 10.1016/j.cmet.2013.11.015 31. Pan L., Huang S., He R., Rong M., Dang Y., Chen G.. **Decreased expression and clinical significance of miR-148a in hepatocellular carcinoma tissues**. *Eur. J. Med. Res.* (2014.0) **19** 68. DOI: 10.1186/s40001-014-0068-2 32. Zhu A., Xia J., Zuo J., Jin S., Zhou H., Yao L., Huang H., Han Z.. **MicroRNA-148a is silenced by hypermethylation and interacts with DNA methyltransferase 1 in gastric cancer**. *Med. Oncol.* (2012.0) **29** 2701-2709. DOI: 10.1007/s12032-011-0134-3 33. Wang W., Dong J., Wang M., Yao S., Tian X., Cui X., Fu S., Zhang S.. **miR-148a-3p suppresses epithelial ovarian cancer progression primarily by targeting c-Met**. *Oncol. Lett.* (2018.0) **15** 6131-6136. DOI: 10.3892/ol.2018.8110 34. Takamura K., Hirao M., Kato K., Taguchi K.. **Surgical cases of congenital heart diseases in the older age group (over 30 years of age)**. *Kyobu Geka* (1975.0) **28** 108-112. PMID: 1168803 35. Agarwal V., Bell G.W., Nam J.W., Bartel D.P.. **Predicting effective microRNA target sites in mammalian mRNAs**. *eLife* (2015.0) **4** e05005. DOI: 10.7554/eLife.05005 36. Poy M.N., Eliasson L., Krutzfeldt J., Kuwajima S., Ma X., Macdonald P.E., Pfeffer S., Tuschl T., Rajewsky N., Rorsman P.. **A pancreatic islet-specific microRNA regulates insulin secretion**. *Nature* (2004.0) **432** 226-230. DOI: 10.1038/nature03076 37. Kloosterman W.P., Lagendijk A.K., Ketting R.F., Moulton J.D., Plasterk R.H.. **Targeted inhibition of miRNA maturation with morpholinos reveals a role for miR-375 in pancreatic islet development**. *PLoS Biol.* (2007.0) **5**. DOI: 10.1371/journal.pbio.0050203 38. Chen Y., Wang X.. **miRDB: An online database for prediction of functional microRNA targets**. *Nucleic Acids Res.* (2020.0) **48** D127-D131. DOI: 10.1093/nar/gkz757 39. Sticht C., De La Torre C., Parveen A., Gretz N.. **miRWalk: An online resource for prediction of microRNA binding sites**. *PLoS ONE* (2018.0) **13**. DOI: 10.1371/journal.pone.0206239 40. Yan C., Hu J., Yang Y., Hu H., Zhou D., Ma M., Xu N.. **Plasma extracellular vesiclepackaged microRNAs as candidate diagnostic biomarkers for earlystage breast cancer**. *Mol. Med. Rep.* (2019.0) **20** 3991-4002. DOI: 10.3892/mmr.2019.10669 41. Lee J., Kwon M.H., Kim J.A., Rhee W.J.. **Detection of exosome miRNAs using molecular beacons for diagnosing prostate cancer**. *Artif. Cells Nanomed. Biotechnol.* (2018.0) **46** S52-S63. DOI: 10.1080/21691401.2018.1489263 42. Marchand L., Jalabert A., Meugnier E., Van den Hende K., Fabien N., Nicolino M., Madec A.M., Thivolet C., Rome S.. **miRNA-375 a Sensor of Glucotoxicity Is Altered in the Serum of Children with Newly Diagnosed Type 1 Diabetes**. *J. Diabetes Res.* (2016.0) **2016** 1869082. DOI: 10.1155/2016/1869082 43. Ru Y., Kechris K.J., Tabakoff B., Hoffman P., Radcliffe R.A., Bowler R., Mahaffey S., Rossi S., Calin G.A., Bemis L.. **The multiMiR R package and database: Integration of microRNA-target interactions along with their disease and drug associations**. *Nucleic Acids Res.* (2014.0) **42** e133. DOI: 10.1093/nar/gku631 44. Xiao F., Zuo Z., Cai G., Kang S., Gao X., Li T.. **miRecords: An integrated resource for microRNA-target interactions**. *Nucleic Acids Res.* (2009.0) **37** D105-D110. DOI: 10.1093/nar/gkn851 45. Huang H.Y., Lin Y.C., Cui S., Huang Y., Tang Y., Xu J., Bao J., Li Y., Wen J., Zuo H.. **miRTarBase update 2022: An informative resource for experimentally validated miRNA-target interactions**. *Nucleic Acids Res.* (2022.0) **50** D222-D230. DOI: 10.1093/nar/gkab1079 46. Karagkouni D., Paraskevopoulou M.D., Chatzopoulos S., Vlachos I.S., Tastsoglou S., Kanellos I., Papadimitriou D., Kavakiotis I., Maniou S., Skoufos G.. **DIANA-TarBase v8: A decade-long collection of experimentally supported miRNA-gene interactions**. *Nucleic Acids Res.* (2018.0) **46** D239-D245. DOI: 10.1093/nar/gkx1141 47. Luo W., Brouwer C.. **Pathview: An R/Bioconductor package for pathway-based data integration and visualization**. *Bioinformatics* (2013.0) **29** 1830-1831. DOI: 10.1093/bioinformatics/btt285 48. Mullokandov G., Baccarini A., Ruzo A., Jayaprakash A.D., Tung N., Israelow B., Evans M.J., Sachidanandam R., Brown B.D.. **High-throughput assessment of microRNA activity and function using microRNA sensor and decoy libraries**. *Nat. Methods* (2012.0) **9** 840-846. DOI: 10.1038/nmeth.2078 49. Reichholf B., Herzog V.A., Fasching N., Manzenreither R.A., Sowemimo I., Ameres S.L.. **Time-Resolved Small RNA Sequencing Unravels the Molecular Principles of MicroRNA Homeostasis**. *Mol. Cell* (2019.0) **75** 756-768.e757. DOI: 10.1016/j.molcel.2019.06.018 50. Kingston E.R., Bartel D.P.. **Global analyses of the dynamics of mammalian microRNA metabolism**. *Genome Res.* (2019.0) **29** 1777-1790. DOI: 10.1101/gr.251421.119 51. Wollny D., Zhao S., Everlien I., Lun X., Brunken J., Brune D., Ziebell F., Tabansky I., Weichert W., Marciniak-Czochra A.. **Single-Cell Analysis Uncovers Clonal Acinar Cell Heterogeneity in the Adult Pancreas**. *Dev. Cell* (2016.0) **39** 289-301. DOI: 10.1016/j.devcel.2016.10.002 52. Raden M., Wallach T., Miladi M., Zhai Y., Kruger C., Mossmann Z.J., Dembny P., Backofen R., Lehnardt S.. **Structure-aware machine learning identifies microRNAs operating as Toll-like receptor 7/8 ligands**. *RNA Biol.* (2021.0) **18** 268-277. DOI: 10.1080/15476286.2021.1940697 53. Wang Y.J., Kaestner K.H.. **Single-Cell RNA-Seq of the Pancreatic Islets--a Promise Not yet Fulfilled?**. *Cell Metab.* (2019.0) **29** 539-544. DOI: 10.1016/j.cmet.2018.11.016 54. Wang N., Zheng J., Chen Z., Liu Y., Dura B., Kwak M., Xavier-Ferrucio J., Lu Y.C., Zhang M., Roden C.. **Single-cell microRNA-mRNA co-sequencing reveals non-genetic heterogeneity and mechanisms of microRNA regulation**. *Nat. Commun.* (2019.0) **10** 95. DOI: 10.1038/s41467-018-07981-6 55. Alberti C., Manzenreither R.A., Sowemimo I., Burkard T.R., Wang J., Mahofsky K., Ameres S.L., Cochella L.. **Cell-type specific sequencing of microRNAs from complex animal tissues**. *Nat. Methods* (2018.0) **15** 283-289. DOI: 10.1038/nmeth.4610 56. Love M.I., Huber W., Anders S.. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol.* (2014.0) **15** 550. DOI: 10.1186/s13059-014-0550-8 57. Kang K., Zhang X., Liu H., Wang Z., Zhong J., Huang Z., Peng X., Zeng Y., Wang Y., Yang Y.. **A novel real-time PCR assay of microRNAs using S-Poly(T), a specific oligo(dT) reverse transcription primer with excellent sensitivity and specificity**. *PLoS ONE* (2012.0) **7**. DOI: 10.1371/journal.pone.0048536 58. Qi M., McFadden B., Valiente L., Omori K., Bilbao S., Juan J., Rawson J., Oancea A.R., Scott S., Nair I.. **Human Pancreatic Islets Isolated From Donors With Elevated HbA1c Levels: Islet Yield and Graft Efficacy**. *Cell Transpl.* (2015.0) **24** 1879-1886. DOI: 10.3727/096368914X683548 59. Gu Z., Eils R., Schlesner M.. **Complex heatmaps reveal patterns and correlations in multidimensional genomic data**. *Bioinformatics* (2016.0) **32** 2847-2849. DOI: 10.1093/bioinformatics/btw313 60. Kevin Blighe S.R., Turkes E., Ostendorf B., Grioni A., Lewis M.. **EnhancedVolcano: Publication-Ready Volcano Plots with Enhanced Colouring and Labeling** 61. Wu T., Hu E., Xu S., Chen M., Guo P., Dai Z., Feng T., Zhou L., Tang W., Zhan L.. **clusterProfiler 4.0: A universal enrichment tool for interpreting omics data**. *Innovation* (2021.0) **2** 100141. DOI: 10.1016/j.xinn.2021.100141
--- title: Prevalence and Related Factors of Lower Urinary Tract Infection in Frail Older Adults Undergoing Major Noncardiac Surgery authors: - Warin Thangrom - Inthira Roopsawang - Suparb Aree-Ue journal: Geriatrics year: 2023 pmcid: PMC10037596 doi: 10.3390/geriatrics8020033 license: CC BY 4.0 --- # Prevalence and Related Factors of Lower Urinary Tract Infection in Frail Older Adults Undergoing Major Noncardiac Surgery ## Abstract Urinary tract infections are the most common complication after surgery in older adults, resulting in poor operative outcomes and reduced quality of life after discharge. However, there is limited research investigating the relationships between urinary tract infection and associated factors in frail older surgical patients, particularly in Thailand. This retrospective study included 220 frail older patients aged ≥ 60 years who had undergone major noncardiac surgery at a tertiary care hospital in Thailand from January 2015 to December 2019. The sample was recruited using the criteria indicated in the modified Frailty Index-11 and having the blood glucose level determined within 2 h before surgery. The prevalence of lower urinary tract infections was $15\%$ post-surgery. Firth’s logistic regression analysis revealed that the equation could predict the accuracy of lower urinary tract infections by $88.5\%$. Frailty, blood glucose levels, complication during admission, and personal factors together predicted the variability of lower urinary tract infections. Adjusting for other variables, being an older adult with severe frailty and complications during hospital admission significantly increased the risk of developing lower urinary tract infections (odds ratio = 3.46, $p \leq 0.05$; odds ratio = 9.53, $p \leq 0.001$, respectively). ## 1. Introduction Global aging is a phenomenon resulting in an increased number of older adults requiring surgery to regain their function and enhance their quality of life. To date, the number of noncardiac surgical procedures has increased to treat declining conditions due to aging, including degenerative joint diseases, vascular diseases, gastrointestinal disorders, and malignancies; however, postoperative complications impacting the quality of life have been extensively documented [1,2,3]. Current evidence has revealed that approximately 300 million noncardiac procedures are performed annually worldwide [4]. Older patients undergoing noncardiac surgery are more likely to experience adverse cardiovascular events or myocardial injury/infarction [5]. Clearly, the risk of developing postoperative complications increases with advanced age. Moreover, atypical symptoms or asymptomatic urinary tract infections (UTIs) are common in older adults, leading to the development of complicated UTIs that are difficult to correctly diagnose, resulting in them being frequently missed, thereby potentially increasing the risk of developing serious infections [6]. Therefore, providing efficient care for promoting health and avoiding preventable complications is more challenging in the older surgical population. UTIs can cause minor complications in other populations but have significantly more severe consequences for older adults, particularly those with chronic health conditions. Recent evidence has demonstrated that UTIs, particularly those of the lower tract, are the most prevalent hospital infections, accounting for $25\%$ of hospitalizations, which not only leads to high care costs but also increases detrimental events in older hospitalizations [7]. Furthermore, following surgery, UTIs are a significant cause of developing sepsis, greatly increasing healthcare costs, and causing up to $40\%$ of in-hospital mortality [8,9,10]. In addition, recent evidence has proposed that the age-related decline condition known as frailty plays an essential role in increasing functional deterioration with a decreased response to stressors, making a person vulnerable to poor outcomes [2,11,12,13,14,15]. Frailty is a well-established measure of surgical outcomes, leading to increased risks for developing complicated UTIs, particularly in older adults with chronic illnesses, including diabetes mellitus [16,17,18]. Because of the effects of physiological changes due to aging processes and frailty, an immune system disturbance evolves, resulting in a poor response to foreign bodies and pathogens. Diminished phagocytosis by T-cells reduces their capacity to destroy pathogens, leading to ineffective infection control, thereby increasing the risk of older adults developing infections [16,17,19]. Concurrently, frailty may interfere with glucose homeostasis, leading to pancreatic impairment—beta-cell dysfunction and increased alpha cell and glucagon secretion—resulting in dysglycemia (higher or lower blood glucose levels) [20,21]. Reduced glucose metabolism is a significant issue in the aging process, known as age-related glucose dysregulation [20,22,23]. Glucose tolerance progressively declines with age, contributing to the high prevalence of impaired glucose tolerance and Type 2 diabetes among older adults [18,23,24]. Under normal circumstances, the human regulatory systems respond to surgical stress by inducing specific mechanisms that aim to regulate the homeostasis condition. Remarkably, recent evidence has underlined that surgery typically induces altered systemic glucose metabolism [25,26]. No matter which part of the body undergoes surgery, the injury or trauma induces disorder mechanisms in glucose metabolism and innate immune responses; however, these responses can become altered in older adults [25,26,27]. In doing so, during illness, complex interactions between counterregulatory hormones and cytokines lead to an increase in excessive glucose production, which is also associated with insulin resistance [28,29]. Therefore, hyperglycemia is frequently reported in individuals with acute medical or surgical situations associated with the increased circulation of counterregulatory hormones, resulting in metabolic and hormonal alterations [20,22,23]. Several products of oxidative stress, free fatty acids, and inflammatory mediators are increased during those situations. These pathophysiologic alterations contribute to direct cellular damage, vascular malfunction, and immunological dysfunction [20,23,29,30]. Therefore, in surgical settings, older adults with or without Type 2 diabetes are at risk of experiencing postoperative complications. Identifying such risks and thereby preventing complications is essential to promote health and quality of care in this population. Prior studies have noted that both frailty and uncontrolled glucose levels increase the risk of detrimental health outcomes in older adults undergoing surgical procedures, particularly in developing lower UTIs and adverse outcomes, a prolonged length of hospital stay, and experiencing postoperative complications or in-hospital mortality, leading to increased care costs [11,13,16,18]. Effective identification or risk stratification is essential for gerontological healthcare professionals to prevent UTIs. However, much uncertainty remains regarding the relationship between age-related decline in conditions and surgical outcomes in the noncardiac surgical population. Furthermore, the nexus linkage between frailty, diabetes, and lower UTIs was not fully explained in the context of tertiary care hospitals, where older patients with complex health conditions are more likely to have elective surgery than in primary care hospitals. As such, this study aimed to investigate the prevalence of postoperative lower UTIs and the associated factors between age, gender, diabetes mellitus (DM), preoperative levels of frailty, blood glucose levels, complications during hospital admission, and lower UTIs in older noncardiac surgical adults. The findings of the present study may not only provide a better understanding of the impact of age-related decline on postoperative complications but could also improve awareness regarding screening for frailty and blood glucose control as associated factors for lower UTIs. In addition, the modification of related factors might decrease the risk of developing lower UTIs, promote efficient care, and enhance the quality of life during treatment and after discharge in the older population. ## 2.1. Study Design The methodological approach taken in this present study was a retrospective study design. The data were extracted from the electronic medical records of in-patient older patients aged ≥60 years scheduled for major noncardiac surgery at a tertiary care hospital in Thailand during 1 January 2015–31 December 2019. ## 2.2. Participants The eligible participants were recruited based on randomly selected electronic medical records for independent review; the specific requirements targeting variables of interest were delineated for the review. The inclusion criteria focused on the individuals who presented with a frailty status evaluated using the modified Frailty Index-11 (mFI-11) [15] and had their blood glucose level determined within 2 h before surgery. The exclusion criteria were: [1] mortality during hospitalization, [2] referral to a different hospital for further treatment, [3] receipt of any palliative care or terminal treatment during hospital admission, and [4] incomplete medical records or missing clinical diagnoses for hospital discharge. The sample size calculation was based on that of the logistic analysis concept by Bujang et al. [ 31]. The calculation formula was $$n = 100$$ + 50 (i), where “i” refers to the number of independent variables of the study: levels of frailty and blood glucose levels. Therefore, the appropriate sample size was expected to be a minimum of 200. To improve the predictive analysis, $10\%$ of the cases were added; thus, the total number of appropriate cases was 220. ## 2.3.1. Modified Frailty Index-11 The mFI-11 is an assessment tool to evaluate frailty status that originated from combining 11 clinically relevant parameters to indicate frailty [15]. The score is calculated by dividing the number of variables by the total number of variables assessed (n/11). In the study, the mFI-11 scores are classified into three categories of frailty status: pre-frail or low level of frailty (scores ≥ 0.09), frailty (scores ≥ 0.27), and severe frailty (scores ≥ 0.45). The mFI-11 is an efficient technique for risk-staging surgical patients and has been correlated with all surgical specialties. It has demonstrated good predictability in terms of adverse health outcomes, including postoperative complications, readmission, reoperation, discharge to skilled care, more extended hospital stays, and a high mortality rate [13,15]. Owner permission was granted to allow the mFI-11 to be used in the present study. ## 2.3.2. Electronic Health Records (EMRs) Criteria for Hospital-Diagnosed Lower UTIs In the present study, evidence of developing lower UTIs during hospitalization is considered to be based on clinical diagnoses made by a physician using standard coding for diagnosing UTIs (International Classification of Diseases (ICD)-9 or ICD-10), together with any treatments associated with UTIs and positive urinalysis and/or urine culture extracted from completed medical records. In contrast, the absence of such evidence is considered to be evidence of the absence of lower UTIs. ## 2.4. Statistical Analysis Descriptive statistics were applied for demographic variables: gender, age, marital status, education, comorbidities, body mass index, activities of daily living, pre-operative diagnosis and operating procedure, blood glucose levels (before surgery), pre-operative urine examination, postoperative urine examination, length of stay, and postoperative complications, which were expressed as frequency, percentage, mean, and standard deviation (SD). The prevalence rate of lower UTIs is expressed as a percentage based on a five-year simple random sample. The factors associated with lower UTIs were assessed using the multiple logistic regression analysis model, controlling for age, gender, DM, and complications during hospital admission. The Firth logistic regression approach was used to eliminate bias and enhance prediction precision for rare events. The significance set at $5\%$ (p-value < 0.05), the odds ratio (OR), and the $95\%$ confidence interval were used to determine the strength of the association. Data analysis was performed using the licensed Social Sciences for Windows (SPSS) version 18 and RStudio version (4.2.2 for MAC). ## 2.5. Ethical Considerations This retrospective study was conducted after obtaining the approval of the Human Research Ethics Committee and Institutional Review Board (number ID: MURA$\frac{2021}{483}$). All procedures were performed under the Declaration of Helsinki-Ethical principle for human data and research. ## 3.1. Participant Characteristics The participants included 220 older adults undergoing noncardiac surgery. Of these, the majority were female ($62.3\%$) with a mean age of 72.95 years (SD = 72.53; range 60–94 years), married ($72.7\%$), and had completed primary education ($65.5\%$). Regarding health information, most of the participants were independent ($90\%$), with $49.5\%$ being overweight or obese and $71.9\%$ having multiple comorbidities. At the preoperative stage, $28.6\%$ had high blood glucose levels, and approximately half of the participants were frail or severely frail ($53.6\%$). Overall, the length of the hospital stay was 11.9 days (ranging from 2 to 132 days, SD = 14.7). For surgical treatments, gastrointestinal surgery was the most prevalent among the noncardiac surgical procedures ($40.9\%$). Postoperative complications occurred in $20.9\%$ of patients, of whom approximately $20\%$ developed lower UTIs post-surgery. The details are presented in Table 1. Considering lower UTIs and frailty status, the more severe the frailty status, a higher rate of postoperative lower UTIs was identified after undergoing major noncardiac surgery. Stratifying lower UTIs by blood glucose levels, the results demonstrated that the higher the preoperative blood glucose level, the greater risk for acquiring postoperative lower UTIs; details are shown in Table 2. ## 3.2. Relationship among Frailty and Selected Factors for Lower UTIs Firth’s logistic regression model was performed to ascertain the effect of preoperative frailty, DM, complications during hospital admission, and blood glucose levels on the likelihood that participants develop postoperative lower UTIs (Table 3). The model explained $33.9\%$ (Nagelkerke R2) of the variance in postoperative UTIs and correctly classified $88.5\%$ of cases. The unadjusted analysis revealed that complications during admission, a very high blood glucose level, and severe frailty were significantly associated with developing UTIs postoperatively ($p \leq 0.05$). Exploring variables in the model and adjusting for age, gender, and having DM, the findings demonstrated that the complication during admission and frailty significantly increased the risk of developing lower UTIs post-surgery ($p \leq 0.05$). Older surgical patients who developed complications during admission were at a 10.57-fold higher risk for developing postoperative lower UTIs ($p \leq 0.001$). Those with severe frailty preoperatively had a 3.46-fold greater risk ($p \leq 0.05$), and frailty was associated with a 1.14-fold risk ($$p \leq 0.817$$) for developing postoperative lower UTIs compared with being pre-frail. However, age, gender, having DM, a high blood glucose level and a very high blood glucose level were not significantly associated with the development of postoperative lower UTIs in older adults undergoing noncardiac surgery ($p \leq 0.05$). ## 4. Discussion This study assessed the prevalence of lower UTIs and the association among age, gender, DM, complications during hospital admission, preoperative frailty status, and blood glucose level, with lower UTIs in older noncardiac surgical adults. The findings of the current study revealed that lower UTIs after a noncardiac surgery were common in this population. Moreover, severe frailty and the presence of any complication during hospital admission were significantly associated with lower UTIs. The prevalence of lower UTIs postoperatively was approximately $15\%$ in the current investigation. However, the prevalence of lower UTIs following noncardiac surgery may vary depending on the clinical setting, bladder catheterizations, or populations. Previous research reported that older adults undergoing colon cancer surgery frequently developed postoperative UTIs (approximately $17\%$), which were a primary cause of more severe complications, including respiratory infections and sepsis [11]. In Thailand, an infection epidemiology study of hospitalized older persons in tertiary care hospitals revealed that UTIs were the primary cause of in-hospital mortality ($16.74\%$), with a prevalence of $43.9\%$ [32]. In orthopedic surgery, the prevalence of postoperative UTIs was found in hip-fracture surgery to be $28.2\%$ [33] and in spine surgery to be 8.2–$11.8\%$ [34,35], while the prevalence of UTIs was $11\%$ in urogynecological surgery [36], and a $26\%$ incidence rate was found in colorectal cancer surgery [37]. The disparity in UTI prevalence may be due to varying preoperative care protocols among hospitals and populations. Moreover, the complexity of complicated or asymptomatic UTIs in older adults may be problematic for a definite diagnosis [6], which may delay an effective response and appropriate interventions. The prevalence of lower UTIs consequently needs to be interpreted with caution. Further inquiry with a greater focus on investigating biological risk factors influencing lower UTIs or comparing the preoperative protocol and catheterization status of those patients is thus suggested. The most prominent finding from the current analysis was that patients with a severely frail status prior to surgery who developed any other postoperative complication displayed an increased risk for lower UTIs post-surgery. These findings corroborated those of a large number of previous studies of postoperative complications in older adults across multiple surgeries, which highlighted that preoperative frailty significantly predicted postoperative complications of either minor or life-threatening events [13,16,17,18,38,39,40]. Clearly, the more severe the frailty, the higher the risk for the occurrence of postoperative complications. Moreover, such personal characteristics as increased age, malnutrition, or multiple comorbidities were more likely to enhance the risk of developing frailty and postoperative problems in older surgical patients [19]. These findings raise intriguing questions regarding the nature and extent of the complexity of aging and the nexus that underpins age-related decline and surgical outcomes in older adults undergoing noncardiac surgery. A further study with a greater focus on intervention or advanced surgical planning to modify preoperative risk factors for preventing postoperative complications in noncardiac surgery will need to be undertaken. Considering the frailty status, the finding of the present study was in agreement with previous studies using mFI-11 in terms of the prevalence of frailty. Moreover, our findings also demonstrated that the more severe the frailty status of a person, the higher the risk of developing lower UTIs. These present findings broadly support the work of other studies in older surgical adults associating frailty and UTIs. The earlier study of Shahrestani et al. [ 35] investigated the prevalence of frailty and UTIs in older adults with lumbar spine surgery; the findings showed that $11.8\%$ of this population was frail with UTIs. Additionally, frail older adults were 3.97 times more likely to develop UTIs after surgery than non-frail older adults (OR: 3.97, $95\%$ confidence interval: 3.21–4.95, $p \leq 0.0001$). Frailty was also associated with an increased risk of developing hospital-acquired catheter-associated UTIs within 2 weeks of admission and a 1.40 times higher risk for in-hospital mortality in hospitalized older adults [41]. Other prior studies have found that such complications coexisted with frailty rather than UTIs. A number of studies revealed that preoperative frailty, which was assessed by mFI-11, was also associated with such adverse outcomes as UTIs, respiratory complications, and surgical site infections [11,13,35,42]. Recent evidence revealed that older adults with DM and displaying preoperative frailty are at risk for developing UTIs and increased hospital stays [17,18,38,40]; moreover, the greater the severity of frailty, the higher the risks for mortality, a 30-day ICU admission, and health care utilization [17,18,38]. Our findings have an important implication for implementing a preoperative frailty assessment to prevent postoperative complications, not only UTIs but also others for noncardiac surgery in the older population. Importantly, the preoperative frailty prevalence must be treated with caution because the concept of frailty applied in the present study involved multiple diseases in classifying the severity of frailty. Therefore, the prevalence of preoperative frailty and its severity may differ from other studies that applied different concepts and instruments to measure frailty; the notion of our study was congruent with a scoping review study of frailty-measuring instruments used in orthopedic surgery [43]. In order to acquire a more comprehensive understanding of preoperative frailty, additional research is required to compare the various approaches to frailty assessment in this population. Regarding the relationship between the levels of frailty and lower UTIs, when controlling for other variables, the findings of the present study demonstrated this relationship to be similar to several earlier studies, indicating that the preoperative frailty status is significantly associated with an increased risk for lower UTIs [16,17,18,33,36,37,40,41,44]. These notions further support the idea that frailty is an age-related decline, which occurs across multiple bodily systems, resulting in an increased risk of developing complications, dependency, disability, and poor quality of life [11,14,39,45]. Participants with frailty are vulnerable to many stressors; thus, exposure to external or internal stressors—hospital admission, surgical procedures, and imbalanced homeostasis—is more likely to increase adverse outcomes [23,24]. Moreover, a South Korean study revealed that frailty frequently coexists with hypertension, osteoporosis, or diabetes, which increases the risk for adverse outcomes [46]. Therefore, older surgical adults who present with preoperative frailty are more vulnerable to postoperative complications such as UTIs, particularly those with co-morbid diseases such as DM. Additionally, other known factors include the presence of multiple comorbidities, chronic opioid use, bleeding conditions, hemolytic conditions, iron deficiency, and B12 deficiency, altered plasma glucose, and glycated hemoglobin in older adults [17]. Notably, frailty-associated UTIs are prevalent in frail older persons, making them difficult to manage effectively. Older adults with severe frailty and those at risk of having a long-term catheter, dehydration, incontinence, or bladder outflow obstruction are more prone to developing complicated UTIs and sepsis [47]. Hence, it would be prudent to identify frailty early in order to avoid preventable complications, such as UTIs, in this population. Recent evidence has highlighted that integrating the notions of frailty and the level of frailty is essential in enhancing appropriate interventions, medical therapeutics, and glycemic management for older adults with DM [17]. Despite the favorable findings, our study discovered that managing glucose levels in older surgical adults with DM who presented with a frail status appears more challenging. In the present study, the findings revealed that most of the participants had normal blood glucose levels even if they displayed frailty. Although either high or very high blood glucose levels were highly likely to be associated with the development of postoperative lower UTIs [48], the present study did not find an association between high blood glucose levels and an increased risk for lower UTIs. Additionally, frail older adults demonstrated varying blood glucose levels, but not all participants with frailty developed lower UTIs. These findings contradict recent studies that suggested that blood glucose level alterations among frail older adults with DM are more likely to increase the risk for infections [17,18,38,40,41]. A possible explanation for these findings could be that the surgical stress response plays a significant role not only in influencing the alteration of the inflammatory–immune system, resulting in an increased risk for infection or excessive inflammation, but also in enhancing the neuroendocrine–metabolic response, resulting in increased blood glucose and diminished energy resources [25]. Moreover, one reason why increased age, sex, and having DM were not significantly associated with lower UTIs post-surgery is that frailty may be a substantial element in increasing dysregulated homeostasis across bodily systems. Therefore, the more severe the frailty, the more likely the risk for infection and increased blood glucose after receiving surgery [25,49]. Therefore, monitoring and controlling glucose levels have become increasingly critical, particularly in geriatric surgery with frailty. Current evidence suggests that frailty and DM share potential mechanism factors—low-graded chronic inflammation and impaired oxidative stress resistance [49]. Frailty and DM appear to be the two sides of the coin; thus, an inflammation biomarker investigation is also required. Moreover, applying standard preoperative care in high-level tertiary care hospitals may prevent uncontrolled blood glucose levels, resulting in diminished risks of developing lower UTIs. Because the level of frailty was significantly associated with postoperative UTIs, the early identification of frailty at the preoperative stage will improve care quality and prevent common postoperative complications such as lower UTIs in this population. ## 5. Conclusions The present study provided new insights into the factors associated with lower UTIs in a tertiary care hospital as well as enhancing the understanding of the complexity of aging and surgical outcomes. In summary, this study revealed that severe preoperative frailty and having any postoperative complications significantly predicted lower UTIs in older adults undergoing major noncardiac surgery when controlled for age, gender, and diabetes mellitus. Because the consequences of in-hospital UTIs may be far beyond expectation regarding profound health problems in older adults, proactive professional nursing care is required. Effective preoperative intervention that combines the notion of frailty may be more beneficial in avoiding preventable postoperative complications such as lower UTIs in this population. Moreover, the findings of the present study may shed light on strengthening professional nursing roles—gerontological nurse practitioners or advanced practice nurses—to prevent postoperative lower UTIs by implementing a practical frailty assessment into daily clinical practice to enhance better preoperative quality care, prevent postoperative complications, and promote quality of life. Notably, these findings have significant implications suggesting that developing a preoperative intervention that integrates frailty-specific glycemic management may be essential to strengthening patient-centered care in this population. ## References 1. Fowler A.J., Abbott T.E.F., Prowle J., Pearse R.M.. **Age of patients undergoing surgery**. *BJS* (2019) **106** 1012-1018. DOI: 10.1002/bjs.11148 2. McIsaac D.I., MacDonald D.B., Aucoin S.D.. **Frailty for perioperative clinicians: A narrative review**. *Anesth. Analg.* (2020) **130** 1450-1460. DOI: 10.1213/ANE.0000000000004602 3. McIsaac D.I., Huang A., Wong C.A., Wijeysundera D.N., Bryson G.L., van Walraven C.. **Effect of preoperative geriatric evaluation on outcomes after elective surgery: A population-based study**. *J. Am. Geriatr. Soc.* (2017) **65** 2665-2672. DOI: 10.1111/jgs.15100 4. Smilowitz N.R., Redel-Traub G., Hausvater A., Armanious A., Nicholson J., Puelacher C., Berger J.S.. **Myocardial injury after noncardiac surgery: A systematic review and meta-analysis**. *Cardiol. Rev.* (2019) **27** 267-273. DOI: 10.1097/CRD.0000000000000254 5. Gao L., Chen L., He J., Wang B., Liu C., Wang R., Fan L., Cheng R.. **Perioperative myocardial injury/infarction after non-cardiac surgery in elderly patients**. *Front. Cardiovasc. Med.* (2022) **9** 910879. DOI: 10.3389/fcvm.2022.910879 6. Martino F.K., Novara G.. **Asymptomatic bacteriuria or urinary tract infection? new and old biomarkers**. *Int. J. Transl. Med.* (2022) **2** 52-65. DOI: 10.3390/ijtm2010006 7. Deiner S., Westlake B., Dutton R.P.. **Patterns of surgical care and complications in elderly adults**. *J. Am. Geriatr. Soc.* (2014) **62** 829-835. DOI: 10.1111/jgs.12794 8. Chin B.S., Kim M.S., Han S.H., Shin S.Y., Choi H.K., Chae Y.T., Jin S.J., Baek J.H., Choi J.Y., Song Y.G.. **Risk factors of all-cause in-hospital mortality among Korean elderly bacteremic urinary tract infection (UTI) patients**. *Arch. Gerontol. Geriatr.* (2011) **52** e50-e55. DOI: 10.1016/j.archger.2010.05.011 9. Curns A.T., Holman R.C., Sejvar J.J., Owings M.F., Schonberger L.B.. **Infectious disease hospitalizations among older adults in the United States from 1990 through 2002**. *Arch. Intern. Med.* (2005) **165** 2514-2520. DOI: 10.1001/archinte.165.21.2514 10. Malik A.T., Jain N., Scharschmidt T.J., Mayerson J.L., Khan S.N.. **Factors associated with post-operative sepsis following surgery for spinal tumors: An analysis of the ACS-NSQIP Database**. *Clin. Neurol. Neurosurg.* (2018) **172** 1-7. DOI: 10.1016/j.clineuro.2018.06.019 11. Hewitt J., Long S., Carter B., Bach S., McCarthy K., Clegg A.. **The prevalence of frailty and its association with clinical outcomes in general surgery: A systematic review and meta-analysis**. *Age Ageing* (2018) **47** 793-800. DOI: 10.1093/ageing/afy110 12. Makary M.A., Segev D.L., Pronovost P.J., Syin D., Bandeen-Roche K., Patel P., Takenaga R., Devgan L., Holzmueller C.G., Tian J.. **Frailty as a predictor of surgical outcomes in older patients**. *J. Am. Coll. Surg.* (2010) **210** 901-908. DOI: 10.1016/j.jamcollsurg.2010.01.028 13. Panayi A.C., Orkaby A.R., Sakthivel D., Endo Y., Varon D., Roh D., Orgill D.P., Neppl R.L., Javedan H., Bhasin S.. **Impact of frailty on outcomes in surgical patients: A systematic review and meta-analysis**. *Am. J. Surg.* (2019) **218** 393-400. DOI: 10.1016/j.amjsurg.2018.11.020 14. Rockwood K., Howlett S.E.. **Age-related deficit accumulation and the diseases of ageing**. *Mech. Ageing Dev.* (2019) **180** 107-116. DOI: 10.1016/j.mad.2019.04.005 15. Velanovich V., Antoine H., Swartz A., Peters D., Rubinfeld I.. **Accumulating deficits model of frailty and postoperative mortality and morbidity: Its application to a national database**. *J. Surg. Res.* (2013) **183** 104-110. DOI: 10.1016/j.jss.2013.01.021 16. Chao C.T., Lee S.Y., Wang J., Chien K.L., Huang J.W.. **Frailty increases the risk for developing urinary tract infection among 79,887 patients with diabetic mellitus and chronic kidney disease**. *BMC Geriatr.* (2021) **21**. DOI: 10.1186/s12877-021-02299-3 17. Strain W.D., Down S., Brown P., Puttanna A., Sinclair A.. **Diabetes and frailty: An expert consensus statement on the management of older adults with type 2 diabetes**. *Diabetes Ther.* (2021) **12** 1227-1247. DOI: 10.1007/s13300-021-01035-9 18. Yanase T., Yanagita I., Muta K., Nawata H.. **Frailty in elderly diabetes patients**. *Endocr. J.* (2018) **65** 1-11. DOI: 10.1507/endocrj.EJ17-0390 19. Pawelec G.. **Immunity and ageing in man**. *Exp. Gerontol.* (2006) **41** 1239-1242. DOI: 10.1016/j.exger.2006.09.005 20. Abdelhafiz A.H., Sinclair A.J.. **Metabolic phenotypes explain the relationship between dysglycaemia and frailty in older people with type 2 diabetes**. *J. Diabetes Complicat.* (2022) **36** 108144. DOI: 10.1016/j.jdiacomp.2022.108144 21. Mooradian A.D., Chehade J.M.. **Diabetes mellitus in older adults**. *Am. J. Ther.* (2012) **19** 145-159. DOI: 10.1097/MJT.0b013e3181ff7e37 22. Chia C.W., Egan J.M., Ferrucci L.. **Age-related changes in glucose metabolism, hyperglycemia, and cardiovascular risk**. *Circ. Res.* (2018) **123** 886-904. DOI: 10.1161/CIRCRESAHA.118.312806 23. Lima J.E.B.F., Moreira N.C.S., Sakamoto-Hojo E.T.. **Mechanisms underlying the pathophysiology of type 2 diabetes: From risk factors to oxidative stress, metabolic dysfunction, and hyperglycemia**. *Mutat. Res. Genet. Toxicol. Environ. Mutagen* (2022) **874–875** 503437. DOI: 10.1016/j.mrgentox.2021.503437 24. Shou J., Chen P.J., Xiao W.H.. **Mechanism of increased risk of insulin resistance in aging skeletal muscle**. *Diabetol. Metab. Syndr.* (2020) **12** 14. DOI: 10.1186/s13098-020-0523-x 25. Cusack B., Buggy D.J.. **Anaesthesia, Analgesia, and the surgical stress response**. *BJA Educ.* (2020) **20** 321-328. DOI: 10.1016/j.bjae.2020.04.006 26. Marik P.E., Ronco C., Bellomo R., Kellum J.A., Ricci Z.. **Chapter 76—Endocrinology of the stress response during critical illness**. *Critical Care Nephrology* (2019) 446-454. DOI: 10.1016/B978-0-323-44942-7.00076-5 27. Thorell A., Nygren J., Ljungqvist O.. **Insulin resistance: A marker of surgical stress**. *Curr. Opin. Clin. Nutr. Metab. Care* (1999) **2** 69-78. DOI: 10.1097/00075197-199901000-00012 28. Farrokhi F., Smiley D., Umpierrez G.E.. **Glycemic control in non-diabetic critically ill patients**. *Best Pract. Res. Clin. Endocrinol. Metab.* (2011) **25** 813-824. DOI: 10.1016/j.beem.2011.05.004 29. Téblick A., Peeters B., Langouche L., Van den Berghe G.. **Adrenal function and dysfunction in critically ill patients**. *Nat. Rev. Endocrinol.* (2019) **15** 417-427. DOI: 10.1038/s41574-019-0185-7 30. Stout M.B., Justice J.N., Nicklas B.J., Kirkland J.L.. **Physiological aging: Links among adipose tissue dysfunction, diabetes, and frailty**. *Physiology* (2017) **32** 9-19. DOI: 10.1152/physiol.00012.2016 31. Bujang M.A., Sa’at N., Sidik T., Joo L.C.. **Sample size guidelines for logistic regression from observational studies with large population: Emphasis on the accuracy between statistics and parameters based on real life clinical data**. *Malays. J. Med. Sci.* (2018) **25** 122-130. DOI: 10.21315/mjms2018.25.4.12 32. Kaewmak J.. **Epidemiology and Impacts of Hospital—Associated Infections on Older Patients in Tertiary Care Hospitals**. *Master’s Thesis* (2016) 33. Mongkolpun S., Roopsawang I., Aree-Ue S.. **Prevalence of urinary tract infection and health outcomes in older adults undergoing hip surgery: A prospective study**. *Pac. Rim Int. J. Nurs. Res.* (2022) **27** 19-33 34. Korytárová N., Kramer S., Schnepf A., Kreienbrock L., Volk H.A.. **Rate of surgical site and urinary tract infections in dogs after cessation of antibiotics following spinal surgery**. *Vet. Rec.* (2022) **16** e2340. DOI: 10.1002/vetr.2340 35. Shahrestani S., Ton A., Chen X.T., Ballatori A.M., Wang J.C., Buser Z.. **The influence of frailty on postoperative complications in geriatric patients receiving single-level lumbar fusion surgery**. *Eur. Spine J.* (2021) **30** 3755-3762. DOI: 10.1007/s00586-021-06960-8 36. O’Shea M., Dillon J., Hendrickson W., Jelovsek J.. **External validation and updating of a model to predict urinary tract infection after urogynecologic surgery**. *Am. J. Obstet. Gynecol.* (2022) **226** S1292-S1293. DOI: 10.1016/j.ajog.2021.12.044 37. Chen C., Wen T., Zhao Q.. **Probiotics used for postoperative infections in patients undergoing colorectal cancer surgery**. *Biomed. Res. Int.* (2020) **2020** 5734718. DOI: 10.1155/2020/5734718 38. Flaatten H., De Lange D.W., Morandi A., Andersen F.H., Artigas A., Bertolini G., Boumendil A., Cecconi M., Christensen S., Faraldi L.. **The impact of frailty on ICU and 30-day mortality and the level of care in very elderly patients (≥80 years)**. *Intensive Care Med.* (2017) **43** 1820-1828. DOI: 10.1007/s00134-017-4940-8 39. Roopsawang I., Thompson H., Zaslavsky O., Belza B.. **Predicting hospital outcomes with the Reported Edmonton Frail Scale-Thai version in orthopaedic older patients**. *J. Clin. Nurs.* (2020) **29** 4708-4719. DOI: 10.1111/jocn.15512 40. Scott M.M., Liang S.Y.. **Infections in older adults**. *Emerg. Med. Clin. N. Am.* (2021) **39** 379-394. DOI: 10.1016/j.emc.2021.01.004 41. Gomila A., Carratalà J., Eliakim-Raz N., Shaw E., Tebé C., Wolkewitz M., Wiegand I., Grier S., Vank C., Cuperus N.. **Clinical outcomes of hospitalised patients with catheter-associated urinary tract infection in countries with a high rate of multidrug-resistance: The Combacte-Magnet Rescuing Study**. *Antimicrob. Resist. Infect. Control* (2019) **8** 198. DOI: 10.1186/s13756-019-0656-6 42. Phan K., Kim J.S., Lee N.J., Somani S., Di Capua J., Kothari P., Leven D., Cho S.K.. **Frailty is associated with morbidity in adults undergoing elective anterior lumbar interbody fusion (ALIF) surgery**. *Spine J.* (2017) **17** 538-544. DOI: 10.1016/j.spinee.2016.10.023 43. Roopsawang I., Zaslavsky O., Thompson H., Aree-Ue S., Kwan R.Y.C., Belza B.. **Frailty measurements in hospitalised orthopaedic populations age 65 and older: A scoping review**. *J. Clin. Nurs.* (2022) **31** 1149-1163. DOI: 10.1111/jocn.16035 44. Palacios-Ceña D., Florencio L.L., Hernández-Barrera V., Fernandez-de-Las-Peñas C., de Miguel-Diez J., Martínez-Hernández D., Carabantes-Alarcón D., Jimenez-García R., Lopez-de-Andres A., Lopez-Herranz M.. **Trends in incidence and outcomes of hospitalizations for urinary tract infection among older people in Spain (2001–2018)**. *J. Clin. Med.* (2021) **10**. DOI: 10.3390/jcm10112332 45. Alser O., Naar L., Christensen M.A., Saillant N., Parks J., Mendoza A., Fagenholz P., King D., Kaafarani H.M.A., Velmahos G.C.. **Preoperative frailty predicts postoperative outcomes in intestinal-cutaneous fistula repair**. *Surgery* (2021) **169** 1199-1205. DOI: 10.1016/j.surg.2020.11.018 46. Kim S., Jung H.-W., Won C.W.. **What are the illnesses associated with frailty in community-dwelling older adults: The Korean frailty and aging cohort study**. *Korean J. Intern. Med.* (2020) **35** 1004-1013. DOI: 10.3904/kjim.2019.097 47. Kostakopoulos N.A., Karakousis N.D., Moschotzopoulos D.. **Frailty Associated urinary tract infections (FaUTIs)**. *J. Frailty Sarcopenia* (2021) **6** 9-13. DOI: 10.22540/JFSF-06-009 48. Marchant M.H.J., Viens N.A., Cook C., Vail T.P., Bolognesi M.P.. **The impact of glycemic control and diabetes mellitus on perioperative outcomes after total joint arthroplasty**. *J. Bone Joint. Surg. Am.* (2009) **91** 1621-1629. DOI: 10.2106/JBJS.H.00116 49. O’Donovan M., Sezgin D., Moloney E., Rodríguez-Laso Á., Liew A., O’Caoimh R.. **Frailty and diabetes in older adults: Overview of current controversies and challenges in clinical practice**. *Front. Clin. Diabetes Healthc.* (2022) **3** 54. DOI: 10.3389/fcdhc.2022.895313
--- title: 'Patient-Reported Quality of Care for Osteoarthritis in General Practice in South Tyrol, Italy: Protocol for Translation, Validation and Assessment of the OsteoArthritis Quality Indicator Questionnaire (OA-QI)' authors: - Christian J. Wiedermann - Pasqualina Marino - Antje van der Zee-Neuen - Isabella Mastrobuono - Angelika Mahlknecht - Verena Barbieri - Sonja Wildburger - Julia Fuchs - Alessandra Capici - Giuliano Piccoliori - Adolf Engl - Nina Østerås - Markus Ritter journal: Methods and Protocols year: 2023 pmcid: PMC10037599 doi: 10.3390/mps6020028 license: CC BY 4.0 --- # Patient-Reported Quality of Care for Osteoarthritis in General Practice in South Tyrol, Italy: Protocol for Translation, Validation and Assessment of the OsteoArthritis Quality Indicator Questionnaire (OA-QI) ## Abstract Background: Evidence-based recommendations for the treatment of knee and hip osteoarthritis are similar internationally. Nevertheless, clinical practice varies across countries. Instruments for measuring quality have been developed to improve health care through targeted interventions. Studies on health service quality must consider the structural and cultural characteristics of countries, because each of their strengths and weaknesses differ. However, such instruments for health-related patient-reported outcomes for osteoarthritis have not yet been validated in German and Italian languages. Objectives: In order to be able to set targeted measures for the improvement of prevention and non-surgical treatment of osteoarthritis in South Tyrol, Italy, the quality of care must be recorded. Therefore, the aim of the project is to update, translate, and validate the OsteoArthritis Quality Indicator (OA-QI) questionnaire version 2, an established and validated questionnaire in Norwegian and English, for Germany and Italy. The second aim is to determine the quality of care for osteoarthritis of the hip and knee in a sample of patients who consult general practice in South Tyrol, and for comparison with patients who are admitted to rehabilitative spa-treatments for osteoarthritis in the state of Salzburg, Austria. Discussion: The results of this study will enable the identification and closure of gaps in osteoarthritis care. Although it is expected that body weight and exercise will play special roles, other areas of nonsurgical care might also be involved. ## 1. Introduction Osteoarthritis (OA) is a joint disease characterized by joint stiffness, pain, disability, and reduced quality of life and is a major cause of pain and disability in the adult population worldwide [1]. For disease diagnosis, radiologic imaging is required only in cases in which the diagnosis is unclear or surgical treatment is necessary; otherwise, the diagnosis is made according to clinical criteria [2,3]. The prevalence of OA increases with age; almost one in two people will develop symptomatic knee OA and one in four will develop symptomatic hip OA during their lifetime [4,5,6,7]. OA exacerbates physical inactivity, which is partly responsible for a number of physical and psychological consequences that increase the risk of morbidity and mortality [8]. With an aging population and obesity epidemic, the prevalence of OA is expected to increase substantially. Based on a recent projection, the overall increase in the total number of patients with OA from 2019 to 2080 is expected to be $38\%$ for both men and women. The most affected groups were those aged 70–79 and 80 or more years. The increases based on the assumed main scenario (mean fertility, rate of immigration, and life expectancy) are forecasted to be $45\%$ and $245\%$ for men and $28\%$ and $148\%$ for women. Assuming a more plausible population growth scenario (higher fertility and rate of immigration, longer life expectancy), these numbers are $74\%$ and $360\%$ (men) and $48\%$ and $209\%$ (women), respectively [6]. In light of this enormous increase in OA incidence, it is likely that this disease will lead to a substantial socioeconomic burden on healthcare systems in the near and far future. These findings will lead to the development of sustainable strategies for the treatment and prevention of OA. OA, especially of the knee and hip, affects patients’ quality of life and is a major challenge for the healthcare system. Guidelines for the prevention and treatment of OA recommend a biopsychosocial approach, in which general practice and rehabilitation play special roles. Evidence-based recommendations and standards for OA management have been defined and have remained essentially unchanged for over a decade [9,10,11,12]. These recommendations, which include (i) patient education, (ii) self-management, (iii) exercise, and (iv) weight reduction, are beneficial for reducing pain and improving functionality. National quality registers were established years ago in various countries [13]. Whether these have led to measurable improvements in the disease burden remains unclear in many cases. Despite these benefits and the relative ease of implementation, they are often not offered to patients with symptomatic OA or implemented in clinical care [14,15,16]. ## 1.1. Quality of Osteoarthritis Care in General Practice Studies on the treatment quality of OA in Northern European countries indicate that quality can be significantly improved in different care settings [17]. To maximize the benefits of OA care, it is important to implement evidence-based and cost-effective care and reduce the use of treatments with limited or no evidence, while reducing the use of resources. Previous research has shown that family physicians (GPs) are reluctant to talk to their patients about relevant psychosocial issues and body weight [18]. It hasalso been shown that GPs favor monitoring patients’ physical function, pain, and analgesia over body mass index (BMI), self-management plans, and exercise advice [19]. Indeed, some GPs feel that they have insufficient expertise to advise patients about exercise [20]. A small number of best-practice initiatives to improve the quality of OA care have been carried out with varying results [13]. However, patient information and exercise have been identified as core treatments for OA with the potential to be improved through self-management programs [21]. Therefore, it is important to assess the quality of OA care, including these two aspects of practice, before implementing regional programs [22]. A recent analysis of quality indicators (QIs) showed large heterogeneities between healthcare systems in terms of exercise therapy, weight counseling, and referrals for laboratory and imaging tests [23]. These differences highlight the need for healthcare systems to carefully select QIs for knee and hip OA to validate the quality of OA care. It is strongly recommended that QIs be reviewed against the most recent guidelines before they are implemented. ## 1.2. Quality Indicators (QIs) of Osteoarthritis Care QIs can be used to assess healthcare quality. These indicators can refer to measurable elements of the metrics of material and human resources of healthcare (i.e., the structures), the activities performed (i.e., the process), and the changes in health status resulting from the healthcare provided (i.e., the outcomes) [24]. QI sets developed from OA care recommendations can be used to monitor and assess the quality of care provided. A systematic review of QI studies on knee and hip OA treatment concluded that QI sets are heterogeneous, precluding cross-cultural use and international comparisons, and only a few studies have included patient perspectives [23]. Patient-reported quality of care showed a large variation in different quality indicators across four European countries, possibly reflecting differences in healthcare priorities [17]. ## OsteoArthritis Quality Indicator (OA-QI) Questionnaire The OsteoArthritis Quality Indicator (OA-QI) questionnaire was developed in 2010 to measure patient-reported health-related quality of OA care [25]. The items of the instrument were based on published QIs from the literature and further refined through expert panels and patient interviews. Content validity was assessed as satisfactory after the OA-QI items were rated as relevant by the patient research partners and expert panels. The OA-QI was revised in 2015 [26]. The concept of the construct is for the disease specificity of care for OA and no other rheumatic diseases. The OA-QI was originally published in Norwegian and is available in Dutch [27], Danish, English, and Portuguese [17]. It has been successfully used for quality assessment in various settings in Denmark [28], Norway [29,30], Australia [31], and the United Kingdom [26]. The OA-QI questionnaire was the first validated instrument to measure patient-reported outcomes as a quality indicator for person-centered OA care [25]. It was revised in 2015 based on feedback (OA-QI v2) [30]. The number of items in the OA-QI v2 reduced from 17 to 16. The revised questionnaire was then completed. The questionnaire took three minutes to complete. The achievement of the QI items (i.e., the success rate for the answer options yes/no/not very concerned in a value range between 0 and 100) was calculated as a percentage (the total number of items achieved divided by the number of items eligible for each participant). A score of 100 indicates the best quality of care rating. The questionnaire is easy to use and recommended for use in primary care and general practice. Internal consistency, inter-observer reliability, and measurement errors were not tested. Reliability (intra-observer and test-retest) was tested and the intraclass correlation coefficient (ICC) was 0.89 ($95\%$ CI 0.83 to 0.93). The instrument was tested on 13 individuals with OA, followed by a short interview to assess the comprehensibility of the questionnaire. Content validity was rated satisfactory. Structural validity was rated as acceptable based on six predefined hypotheses. To assess construct validity, hypothesis tests were conducted and all ten a priori hypotheses were confirmed. The cross-cultural validity of the translated OA-QI was also tested, and the instrument was used in national and international studies. The minimum significant difference (MSD) after participation in an OA patient education program was 20.4. This instrument can be used free of charge. The English version of the instrument is available in the original paper [30] but has not yet been validated in this language. The OA-QI v2 consists of 16 self-administered items rating the individually perceived quality of care with selected responses (i.e., yes/no/not severely troubled), resulting in a total score ranging from 0 to 100 [30]. Higher scores in this range represent better quality of care. The reliability of the OA-QI v2 was estimated to be higher than that of the OA-QI v1 and its validity was acceptable. Therefore, the new version was recommended for future use as an outcome measure in studies to improve OA care [30]. The reliability, responsiveness, and interpretability of the OA-QI v2 were tested using the COSMIN checklist, which focuses on assessing the methodological quality of studies on the measurement properties of patient-related health outcomes with repeated evaluations [32]. ## 1.3. Objectives The main purpose of this study is to update, translate, and validate the OsteoArthritis Quality Indicator (OA-QI) questionnaire version 2, an established and validated questionnaire in Norwagian, for Germann and Italian languages, to assess the extent to which evidence-based treatment recommendations for OA care are followed at the regional level in South Tyrol, Italy, and to compare the survey results with those of a selected group of patients with OA in spa treatment for rehabilitation in Salzburg, Austria. To evaluate the quality of OA care, patients who contact their GP or seek spa-treatment because of complaints caused by OA of the hip or knee will be asked to fill out the OA-QI v2 questionnaire [30] that allows the quality of previous medical care to be assessed. The rehabilitation patient group is expected to have received a higher degree of attention with respect to OA as a (possible) cause of their complaints prior to admission to treatment. Therefore, this group may serve to disclose an attainable “standard” for optimized OA patient care in general practice. As the OA-QI questionnaire is not yet available in German or Italian in a tested format, the questionnaire will be translated and culturally adapted to German- and Italian-speaking patients in the following steps: initial translations, synthesis of the translations, back translations, expert committee review, test of the pre-final versions, and development of the German and Italian versions of OA-QI v2 (G-OA-QI v2 and I-OA-QI v2, respectively). This phase will include testing by patient representatives. In accordance with the original definition of OA quality, the individual QI items were based on the 2015 recommendations of professional societies for the treatment of knee and hip OA. After relevant changes in the treatment recommendations of the guidelines occurred recently, the third aim of the study is to control and eventually update the individual QI items of the G- and I-OA-QI v2. The resulting versions will be tested for validity in patients with OA in South Tyrol and Salzburg and finally used for a cross-sectional prospective observational quality assessment study. ## 2.1. Institutional Settings Scientific collaboration between the Institute of General Practice and Public Health (IGPPH) at the College of Health Professions−Claudiana in Bolzano and the Paracelsus Medical University (PMU) is longstanding and focused on quality of care [33,34,35,36]. Indicators for assessing the quality of primary care for chronic diseases were compared between Salzburg and South Tyrol in a study performed by the IGPPH in Bolzano and the Institute of General Practice, Family Medicine, and Preventive Medicine of the PMU in Salzburg. *In* general practice, quality indicators were assessed for chronic conditions including diabetes mellitus type 2, hypertension, coronary heart disease, cerebrovascular disease, peripheral arterial disease, chronic heart failure, atrial fibrillation, and chronic obstructive pulmonary disease, but not knee or hip OA [37]. The Institute of Physiology and Pathophysiology of the PMU harbors the Gastein Research Institute and is a research unit at the Ludwig Boltzmann Institute for Arthritis and Rehabilitation. Recent studies have focused on the projection of the expected number of OA patients to provide a meaningful basis for policymakers when planning and budgeting efforts to treat and prevent OA [6]. ## 2.2.1. Stepwise Translation Process First, the authors of OA-QI v2 were contacted, and permission was obtained for translation into German and Italian. The authors of the OA-QI v2 also confirmed that a German or Italian version of the instrument has not yet been developed. The English version of OA-QI v2 (Table 1) will be assessed for the need of cross-cultural adaption by a professional translation company and translated into Italian and German following an established forward-backward translation procedure, with independent translations and back translations. If cross-cultural adaption is deemed necessary by the translation experts, cognitive interviews to assess after-translation content validity from a patient perspective will be conducted with a limited number of participants ($$n = 5$$). Internal validity will be assessed for test-retest reliability using the intraclass correlation coefficients, agreement between assessments with Bland–Altman plots, and construct validity with Spearman’s correlation coefficients. Construct validity analyses will be performed using predefined hypotheses, as described in [25]. The English version of the OA-QI v2 will be translated by a translation company specialized in healthcare. Members of the research teams at the Institute of General Practice and Public Health (IGPPH) in Bolzano and the Paracelsus Medical University (PMU) in Salzburg then review the Italian and German translations. To verify the accuracy of the translation and update the questionnaire items, the two documents will then be sent to two rheumatologists in Italy and Austria, whose suggestions for changes, if any, are incorporated. Ideally, sets of quality indicators should be updated frequently to reflect the current evidence-based treatment recommendations. The OA-QI v2 was updated in 2015–2016, then tested for measurement properties, and published in 2018 [30]. Since then, except for the European League Against Rheumatism (EULAR) [38], the Osteoarthritis Research Society International (OARSI) [11], American College of Rheumatology (ACR) [12], and National Institute for Health and Care Excellence (NICE) [39], treatment recommendations for paracetamol as first-line pharmacological treatment have changed. Hence, the items in the OA-QI v2 on this aspect, as well as topical or oral nonsteroidal anti-inflammatory drugs, should be updated. The proposal will be made to rheumatologists to update item #12 from ‘If you have joint pain, was paracetamol the first recommended medication?’ to ‘If you have joint pain, was paracetamol or a nonsteroidal anti-inflammatory drug the first medication that was recommended?’, and to update item #13 from ‘If you have prolonged severe joint pain, which is not relieved sufficiently by paracetamol, have you been offered stronger pain killing medications? ( e.g., co-codamol, codeine, tramadol, co-proxamol, co-dydramol, dihydrocodeine)’ to ‘If you have prolonged severe joint pain, which is not relieved sufficiently by a nonsteroidal anti-inflammatory drug or paracetamol, have you been offered stronger pain killing medications? ( e.g., co-codamol, codeine, tramadol, co-proxamol, co-dydramol, dihydrocodeine). The two translated and reviewed G-OA-QI v2 and I-OA-QI v2 documents are then back-translated into the English language by the certified translation company, and the back-translated questionnaire is compared with the original OA-QI v2 to identify any major discrepancies (Figure 1). ## 2.2.2. Pilot Survey of German and Italian OsteoArthritis Quality Indicator Version 2 Questionnaires in Patients with Knee and Hip Osteoarthritis To test for clarity of the translated German and Italian G-OA-QI v2 and I-OA-QI v2 questionnaires and their validity, a pilot study will be conducted with 25 German-speaking and 25 Italian-speaking knee or hip osteoarthritis patients, respectively (for patient selection, see below). For this purpose, the patients will answer the 16 items of the G-OA-QI v2 and I-OA-QI v2 and provide information regarding selected socio-demographic and clinical characteristics. To calculate the test-retest reliability, they will fill out the respective questionnaires twice (i.e., at baseline and two weeks later) under the prerequisite that they do not see health professionals in the interim. Prior to this, the questionnaire items will be discussed with a subgroup of five patients per language in the context of cognitive debriefing interviews. ## 2.3. Participants Recruitment and Eligibility Criteria Participants’ health-related reports on the quality of hip and knee OA care will be assessed in a cross-sectional survey using the validated Italian or German version of the updated OA-QI v2 questionnaire, according to the patient’s mother tongue, in a sample of 220 patients with OA in South *Tyrolean* general practices (50 for validation of the I- and G-OA-QI v2 questionnaires and 170 for quality assessment) and 150 patients with OA visiting rehabilitation facilities in the state of Salzburg for OA treatment (all for quality assessment). The sample size for the cohort was based on a pragmatic approach based on the number of referrals from patients with knee and hip OA during a 2-year inclusion period and is chosen to enable subgroup analyses above a minimal participant number of 50 each. Sample sizes for the validation phase were determined according to current scientific recommendation [41]. For the necessary number of cases for the quality assessment study, there are no uniform recommendations for the a priori calculation. We followed the subject-to-item ratio recommendation [42], which ranges from 1:5 to 1:30 in the literature. With 16 items of the used OA-QI v2 tools, the number of 170 patients in South Tyrol and 150 patients in Salzburg, which was set for pragmatic reasons (number of GPs and average number of patients in their outpatient clinics), corresponds to a ratio of about 1:10. ## 2.3.1. General Practice Subjects in South Tyrol are recruited in up to 25 GP practices of the Department of Basic Medical Services of the South Tyrolean Public Health Services. A pragmatic approach to the inclusion of participants based on the GPs’ diagnosis of knee or hip OA is applied, irrespective of the diagnostic criteria the GPs use. Patients presenting with unspecified symptoms or diagnoses, such as ‘knee or hip pain’ or ‘knee or hip problems’, will be considered for recruitment. The inclusion criteria defining OA diagnosis according to NICE are people who (i) are 45 years or older, (ii) have activity-related joint pain, and (iii) have either no morning joint-related stiffness or morning stiffness that lasts no longer than 30 min. Imaging to diagnose OA is not routinely used unless there are atypical features or features that suggest an alternative or additional diagnosis [39]. Current medications for OA (analgesics, nonsteroidal anti-inflammatory drugs, agents modifying the structure of connective tissue, and potentially disease-modifying OA drugs, intra-articular therapy, corticosteroids, visco-supplementation, and closed joint cleaning) will be documented. The exclusion criteria will be as follows: malignant illness, rheumatoid or other inflammatory arthritis, severe degeneration of the hip or knee joint (Kellgren and Lawrence Grade IV [43]), other inflammatory rheumatic diseases, mental or psychiatric disorders, inability to cooperate with the study requirements, and involvement in any other pharmaceutical or exercise studies at the moment. ## 2.3.2. Rehabilitation Facilities in the Austrian State of Salzburg Subjects in Austria are recruited in up to 25 GP practices and through spas and rehabilitation physicians prior to the initiation of treatment. The eligibility criteria are equal to those applied in the recruitment process in South Tyrol. In addition to direct recruitment by physicians as mentioned above, the Gastein Research Institute may collect relevant data by extending the already existing ‘Radon indication registry for the assessment of pain reduction, increase in quality of life, and improvement in body functionality through low-dose *Radon hyperthermia* therapy (RadReg)’ using the G-OA-QI v2 questionnaire [44]. This registry collects data from individuals visiting the valley of Gastein for spa-treatment, including low-dose radon for a variety of rheumatic diseases including OA. Registry subjects are recruited by physicians participating in treatment spa centers in the Gastein Valley. Therefore, these physicians are already trained in handling the RadReg questionnaires and will additionally receive free skills training to aid them in the recruitment of participants for the current study. No sample size calculation was performed, but post-hoc analyses will provide insights into the power of the study. ## 2.4. Quality Indicator, Demographic and Disease Charactieristic The G-OA-QI v2 and I-OA-QI v2 questionnaires will be tested to assess the quality of OA care in the respective samples of consecutive OA patients participating in general practices in South Tyrol and in participating general practices, health care, and spa/rehabilitation centers in the state of Salzburg. As described in [30], a QI item will be considered achieved if the participant has checked ‘Yes.’ An item was considered “eligible” if the participant responded ‘Yes’ or ‘No’ for that item, whereas items were considered ‘not eligible’ and excluded from analysis if there was a missing/ambiguous response or if the participant had responded ‘Don’t remember,’ ‘Not overweight,’ ‘No such problems,’ and so on. Hence, the total number of eligible items varied across participants. A total of 170 German or Italian speaking subjects and 150 German-speaking subjects will be tested in South Tyrol and Salzburg, respectively. In South Tyrol, patient responses to the questionnaires will be collected in general practice before the personal visit of the patient to the GP. In Austria, the assessment will be performed equally in the case of recruitment through GPs or immediately after the patients’ admission and before the start of their treatments in the case of recruitment through spa/rehabilitation centers. The OA-QI v2 will be supplemented with demographic and clinically relevant data (Table 2) and will include the severity of OA (Lequesne Index [45] in its German [46] or Italian [47] versions) and EQ-5D-5L with subscales on mobility, self-care, usual activities, pain/discomfort, and anxiety/depression [48] for German [49] and Italian [50] in addition to the duration of knee or hip problems, other affected joints, and any surgical joint interventions. The Western Ontario and McMaster Universities (WOMAC) OA index subscale will be used in its Italian [51] and German [52] versions to assess physical function. ## 2.5. Study Registry Entry and Ethics This study is registered in the ISRCTN registry [55]. The Scientific Ethics Committee of the Autonomous Province of Bolzano, Italy reviewed the study protocol and approved the study conduct on 20 October 2022 (No. 103-2022). Ethical approval will be obtained from the study center in Salzburg, Austria, according to the national regulations. The study will be conducted according to the standards of good clinical and scientific practice in compliance with the Declaration of Helsinki [56]. Furthermore, the guidelines for ‘Strengthening the Reporting of Observational studies in Epidemiology’ (STROBE) for the publication of observational studies will be followed [57]. ## 2.6.1. Primary Outcome Achievement of the QI items of the G-OA-QI v2 and I-OA-QI v2 tools by patients with knee and hip OA in general practice in South Tyrol and rehabilitative spa treatment will be the primary outcome parameters. The mean total pass rate will be calculated as a percentage, as described [30] for the whole sample, as well as for subgroups including type of OA, language, and treatment location. ## 2.6.2. Secondary Outcomes Differences within the same healthcare setting will be identified as secondary outcomes depending on demographic and clinical characteristics. ## 2.7. Trial and Data Management The development and implementation of the study will follow the principles of the Declaration of Helsinki [56]. This type of data collection will be implemented by IGPPH in Bolzano and conducted in a pseudonymous form. Data provided by participants from online and transcribed paper questionnaires will be collected centrally in the SoSci Survey Software, version 3.2.46 (SoSci Survey GmbH, Munich, Germany). The online questionnaires are programmed such that all items have to be answered. The data are stored by IGPPH in Bolzano, Italy, and made available to the research team upon request after the end of the data analysis period. Data backup is regularly performed. Standard operating procedures (SOPs) chordate study procedures for various study assistants with appropriate training to regulate parallel procedures. ## 2.8. Statistical Analyses Descriptive statistics will describe these data according to their metric properties, and regression analyses will be performed to explore the association between the questionnaire scores and predefined clinical outcomes. Statistical analyses will be performed using the software package IBM SPSS Statistics for Windows and STATA. The results of the study will have the ability to identify strengths and weaknesses in the quality of OA care in the two study cohorts of South Tyrol and Salzburg, and to determine the association between the quality of care and clinical outcomes. ## 3. Discussion General practice and primary care have become increasingly relevant in the care of OA patients. This study provides an overview of the quality of care for knee and hip OA after consulting a GP in South Tyrol and the state of Salzburg or spa/rehabilitation treatment in the state of Salzburg. The strength of this study is that patients are included consecutively from centers that represent both rural and urban areas of Northern Italy and the state of Salzburg, thus increasing the representativeness of the study population. The use of self-report questionnaires containing retrospective information about previous treatments carries the risk of recall bias [58]. Another limitation is that it may include a small number of patients who do not meet the diagnostic criteria for OA, as the study is based on referrals from general practitioners for knee or hip OA from patients with non-specific diagnoses such as “knee pain” or “knee problems” if their age is ≥45 years. However, the self-report approach is the only way to collect information on patient-reported quality of care. Risks for meeting preset milestones include the recruitment of GPs study sites and planned patient numbers in both general practice and participating health centers. As similar study protocols have been successfully completed in the past and various sites will participate in the recruitment of patients in the state of Salzburg, we are confident that the project goals will also be achieved. This study was approved by the Italian Regional Ethical Committee of the Province of Alto Adige (No. 103-2022). Data will be anonymized and handled in line with the General Data Protection Regulation and Italian Data Protection Act. The study results will be submitted to international, open-access, peer-reviewed journals and disseminated at conferences. The validated translation of the OA-QI v2 into Italian and *German is* expected to result in two open-access peer-reviewed publications. The original article on the quality of OA care will be published in a clinical rheumatology journal and is expected to propose specific interventions for quality improvement in both South Tyrol and Salzburg. ## 4. Conclusions The results of this study will enable the identification and closure of gaps in OA care. Although it is expected that body weight and exercise will play special roles, other areas of nonsurgical care might also be involved. ## References 1. **GBD 2016 Disease and Injury Incidence and Prevalence Collaborators Global, Regional, and National Incidence, Prevalence, and Years Lived with Disability for 328 Diseases and Injuries for 195 Countries, 1990–2016: A Systematic Analysis for the Global Burden of Disease Study 2016**. *Lancet* (2017.0) **390** 1211-1259. DOI: 10.1016/S0140-6736(17)32154-2 2. Altman R., Alarcón G., Appelrouth D., Bloch D., Borenstein D., Brandt K., Brown C., Cooke T.D., Daniel W., Feldman D.. **The American College of Rheumatology Criteria for the Classification and Reporting of Osteoarthritis of the Hip**. *Arthritis Rheum.* (1991.0) **34** 505-514. DOI: 10.1002/art.1780340502 3. Altman R., Asch E., Bloch D., Bole G., Borenstein D., Brandt K., Christy W., Cooke T.D., Greenwald R., Hochberg M.. **Development of Criteria for the Classification and Reporting of Osteoarthritis. Classification of Osteoarthritis of the Knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association**. *Arthritis Rheum.* (1986.0) **29** 1039-1049. DOI: 10.1002/art.1780290816 4. Murphy L.B., Helmick C.G., Schwartz T.A., Renner J.B., Tudor G., Koch G.G., Dragomir A.D., Kalsbeek W.D., Luta G., Jordan J.M.. **One in Four People May Develop Symptomatic Hip Osteoarthritis in His or Her Lifetime**. *Osteoarthr. Cartil.* (2010.0) **18** 1372-1379. DOI: 10.1016/j.joca.2010.08.005 5. Murphy L., Schwartz T.A., Helmick C.G., Renner J.B., Tudor G., Koch G., Dragomir A., Kalsbeek W.D., Luta G., Jordan J.M.. **Lifetime Risk of Symptomatic Knee Osteoarthritis**. *Arthritis Rheum.* (2008.0) **59** 1207-1213. DOI: 10.1002/art.24021 6. Hitzl W., Stamm T., Kloppenburg M., Ritter M., Gaisberger M., van der Zee-Neuen A.. **Projected Number of Osteoarthritis Patients in Austria for the next Decades—Quantifying the Necessity of Treatment and Prevention Strategies in Europe**. *BMC Musculoskelet. Disord.* (2022.0) **23**. DOI: 10.1186/s12891-022-05091-5 7. Bijlsma J.W.J., Berenbaum F., Lafeber F.P.J.G.. **Osteoarthritis: An Update with Relevance for Clinical Practice**. *Lancet* (2011.0) **377** 2115-2126. DOI: 10.1016/S0140-6736(11)60243-2 8. **World Health Organization Physical Activity** 9. 9. National Clinical Guideline Centre (UK) Osteoarthritis: Care and Management in AdultsNational Institute for Health and Clinical Excellence: GuidanceNational Institute for Health and Care ExcellenceLondon, UK2014. *Osteoarthritis: Care and Management in Adults* (2014.0) 10. Fernandes L., Hagen K.B., Bijlsma J.W.J., Andreassen O., Christensen P., Conaghan P.G., Doherty M., Geenen R., Hammond A., Kjeken I.. **EULAR Recommendations for the Non-Pharmacological Core Management of Hip and Knee Osteoarthritis**. *Ann. Rheum. Dis.* (2013.0) **72** 1125-1135. DOI: 10.1136/annrheumdis-2012-202745 11. Bannuru R.R., Osani M.C., Vaysbrot E.E., Arden N.K., Bennell K., Bierma-Zeinstra S.M.A., Kraus V.B., Lohmander L.S., Abbott J.H., Bhandari M.. **OARSI Guidelines for the Non-Surgical Management of Knee, Hip, and Polyarticular Osteoarthritis**. *Osteoarthr. Cartil.* (2019.0) **27** 1578-1589. DOI: 10.1016/j.joca.2019.06.011 12. Kolasinski S.L., Neogi T., Hochberg M.C., Oatis C., Guyatt G., Block J., Callahan L., Copenhaver C., Dodge C., Felson D.. **2019 American College of Rheumatology/Arthritis Foundation Guideline for the Management of Osteoarthritis of the Hand, Hip, and Knee**. *Arthritis Rheumatol.* (2020.0) **72** 220-233. DOI: 10.1002/art.41142 13. Thorstensson C.A., Garellick G., Rystedt H., Dahlberg L.E.. **Better Management of Patients with Osteoarthritis: Development and Nationwide Implementation of an Evidence-Based Supported Osteoarthritis Self-Management Programme**. *Musculoskelet. Care* (2015.0) **13** 67-75. DOI: 10.1002/msc.1085 14. Basedow M., Esterman A.. **Assessing Appropriateness of Osteoarthritis Care Using Quality Indicators: A Systematic Review**. *J. Eval. Clin. Pract.* (2015.0) **21** 782-789. DOI: 10.1111/jep.12402 15. Hagen K.B., Smedslund G., Østerås N., Jamtvedt G.. **Quality of Community-Based Osteoarthritis Care: A Systematic Review and Meta-Analysis**. *Arthritis Care Res.* (2016.0) **68** 1443-1452. DOI: 10.1002/acr.22891 16. Healey E.L., Afolabi E.K., Lewis M., Edwards J.J., Jordan K.P., Finney A., Jinks C., Hay E.M., Dziedzic K.S.. **Uptake of the NICE Osteoarthritis Guidelines in Primary Care: A Survey of Older Adults with Joint Pain**. *BMC Musculoskelet. Disord.* (2018.0) **19**. DOI: 10.1186/s12891-018-2196-2 17. Østerås N., Jordan K.P., Clausen B., Cordeiro C., Dziedzic K., Edwards J., Grønhaug G., Higginbottom A., Lund H., Pacheco G.. **Self-Reported Quality Care for Knee Osteoarthritis: Comparisons across Denmark, Norway, Portugal and the UK**. *RMD Open* (2015.0) **1** e000136. DOI: 10.1136/rmdopen-2015-000136 18. Finset A.. **One Size Does Not Fit All: How to Talk to Patients about Obesity**. *Patient. Educ. Couns.* (2009.0) **76** 147-148. DOI: 10.1016/j.pec.2009.07.003 19. Clarson L.E., Nicholl B.I., Bishop A., Edwards J.J., Daniel R., Mallen C.D.. **Monitoring Osteoarthritis: A Cross-Sectional Survey in General Practice**. *Clin. Med. Insights. Arthritis Musculoskelet. Disord.* (2013.0) **6** 85-91. DOI: 10.4137/CMAMD.S12606 20. Cottrell E., Foster N.E., Porcheret M., Rathod T., Roddy E.. **GPs’ Attitudes, Beliefs and Behaviours Regarding Exercise for Chronic Knee Pain: A Questionnaire Survey**. *BMJ Open* (2017.0) **7** e014999. DOI: 10.1136/bmjopen-2016-014999 21. Smink A.J., Dekker J., Vliet Vlieland T.P.M., Swierstra B.A., Kortland J.H., Bijlsma J.W.J., Teerenstra S., Voorn T.B., Bierma-Zeinstra S.M.A., Schers H.J.. **Health Care Use of Patients with Osteoarthritis of the Hip or Knee after Implementation of a Stepped-Care Strategy: An Observational Study**. *Arthritis Care Res.* (2014.0) **66** 817-827. DOI: 10.1002/acr.22222 22. Smink A.J., Bierma-Zeinstra S.M.A., Schers H.J., Swierstra B.A., Kortland J.H., Bijlsma J.W.J., Teerenstra S., Voorn T.B., Dekker J., Vliet Vlieland T.P.M.. **Non-Surgical Care in Patients with Hip or Knee Osteoarthritis Is Modestly Consistent with a Stepped Care Strategy after Its Implementation**. *Int. J. Qual. Health Care* (2014.0) **26** 490-498. DOI: 10.1093/intqhc/mzu058 23. Arslan I.G., Rozendaal R.M., van Middelkoop M., Stitzinger S.A.G., de Kerkhove M.-P.V., Voorbrood V.M.I., Bindels P.J.E., Bierma-Zeinstra S.M.A., Schiphof D.. **Quality Indicators for Knee and Hip Osteoarthritis Care: A Systematic Review**. *RMD Open* (2021.0) **7** e001590. DOI: 10.1136/rmdopen-2021-001590 24. Mainz J.. **Defining and Classifying Clinical Indicators for Quality Improvement**. *Int. J. Qual. Health Care* (2003.0) **15** 523-530. DOI: 10.1093/intqhc/mzg081 25. Østerås N., Garratt A., Grotle M., Natvig B., Kjeken I., Kvien T.K., Hagen K.B.. **Patient-Reported Quality of Care for Osteoarthritis: Development and Testing of the Osteoarthritis Quality Indicator Questionnaire**. *Arthritis Care Res.* (2013.0) **65** 1043-1051. DOI: 10.1002/acr.21976 26. Blackburn S., Higginbottom A., Taylor R., Bird J., Østerås N., Hagen K.B., Edwards J.J., Jordan K.P., Jinks C., Dziedzic K.. **Patient-Reported Quality Indicators for Osteoarthritis: A Patient and Public Generated Self-Report Measure for Primary Care**. *Res. Involv. Engagem.* (2016.0) **2** 5. DOI: 10.1186/s40900-016-0019-x 27. Oomen J.M.H., Peters Y., van den Ende C.H., Schers H.J., Assendelft W.J.J., Vriezekolk J.E., Koëter S.. **Quality of Knee Osteoarthritis Care in the Netherlands: A Survey on the Perspective of People with Osteoarthritis**. *BMC Health Serv. Res* (2022.0) **22**. DOI: 10.1186/s12913-022-08014-1 28. Ingelsrud L.H., Roos E.M., Gromov K., Jensen S.S., Troelsen A.. **Patients Report Inferior Quality of Care for Knee Osteoarthritis Prior to Assessment for Knee Replacement Surgery—A Cross-Sectional Study of 517 Patients in Denmark**. *Acta Orthop.* (2020.0) **91** 82-87. DOI: 10.1080/17453674.2019.1680180 29. Grønhaug G., Østerås N., Hagen K.B.. **Quality of Hip and Knee Osteoarthritis Management in Primary Health Care in a Norwegian County: A Cross-Sectional Survey**. *BMC Health Serv. Res.* (2014.0) **14**. DOI: 10.1186/s12913-014-0598-x 30. Østerås N., Tveter A.T., Garratt A.M., Svinøy O.E., Kjeken I., Natvig B., Grotle M., Hagen K.B.. **Measurement Properties for the Revised Patient-Reported OsteoArthritis Quality Indicator Questionnaire**. *Osteoarthr. Cartil.* (2018.0) **26** 1300-1310. DOI: 10.1016/j.joca.2018.06.007 31. Umapathy H., Bennell K., Dickson C., Dobson F., Fransen M., Jones G., Hunter D.J.. **The Web-Based Osteoarthritis Management Resource My Joint Pain Improves Quality of Care: A Quasi-Experimental Study**. *J. Med. Internet. Res.* (2015.0) **17** e167. DOI: 10.2196/jmir.4376 32. Mokkink L.B., Terwee C.B., Patrick D.L., Alonso J., Stratford P.W., Knol D.L., Bouter L.M., de Vet H.C.W.. **The COSMIN Checklist for Assessing the Methodological Quality of Studies on Measurement Properties of Health Status Measurement Instruments: An International Delphi Study**. *Qual. Life Res.* (2010.0) **19** 539-549. DOI: 10.1007/s11136-010-9606-8 33. Paier-Abuzahra M.E., Mahlknecht A., Piccoliori G., Engl A., Sönnichsen A.. **Quality of Chronic Care in General Practices in Salzburg, Austria, and South Tyrol, Italy: A Comparative Process of Care Intervention Study**. *Z. Evid. Qual. Gesundhwes.* (2022.0) **170** 14-20. DOI: 10.1016/j.zefq.2022.01.005 34. Mahlknecht A., Wiedermann C.J., Sandri M., Engl A., Valentini M., Vögele A., Schmid S., Deflorian F., Montalbano C., Koper D.. **Expert-Based Medication Reviews to Reduce Polypharmacy in Older Patients in Primary Care: A Northern-Italian Cluster-Randomised Controlled Trial**. *BMC Geriatr.* (2021.0) **21**. DOI: 10.1186/s12877-021-02612-0 35. Piccoliori G., Mahlknecht A., Abuzahra M.E., Engl A., Breitenberger V., Vögele A., Montalbano C., Sönnichsen A.. **Quality Improvement in Chronic Care by Self-Audit, Benchmarking and Networking in General Practices in South Tyrol, Italy: Results from an Interventional Study**. *Fam. Pract.* (2021.0) **38** 253-258. DOI: 10.1093/fampra/cmaa123 36. Piccoliori G., Mahlknecht A., Sandri M., Valentini M., Vögele A., Schmid S., Deflorian F., Engl A., Sönnichsen A., Wiedermann C.. **Epidemiology and Associated Factors of Polypharmacy in Older Patients in Primary Care: A Northern Italian Cross-Sectional Study**. *BMC Geriatr.* (2021.0) **21**. DOI: 10.1186/s12877-021-02141-w 37. Mahlknecht A., Abuzahra M.E., Piccoliori G., Enthaler N., Engl A., Sönnichsen A.. **Improving Quality of Care in General Practices by Self-Audit, Benchmarking and Quality Circles**. *Wien. Klin. Wochenschr.* (2016.0) **128** 706-718. DOI: 10.1007/s00508-016-1064-z 38. Pendleton A., Arden N., Dougados M., Doherty M., Bannwarth B., Bijlsma J.W., Cluzeau F., Cooper C., Dieppe P.A., Günther K.P.. **EULAR Recommendations for the Management of Knee Osteoarthritis: Report of a Task Force of the Standing Committee for International Clinical Studies Including Therapeutic Trials (ESCISIT)**. *Ann. Rheum. Dis.* (2000.0) **59** 936-944. DOI: 10.1136/ard.59.12.936 39. **National Institute for Health and Care Excellence Osteoarthritis in over 16s: Diagnosis and Management** 40. Omair M.A., Al Suwayeh F., Almashaan A., Alqurtas E., Bedaiwi M.K., Almaghlouth I., Alkahalaf A., Almalaq H.M.. **Cross-Cultural Validation of the 5-Item Compliance Questionnaire for Rheumatology to the Arabic Language in Patients with Rheumatoid Arthritis**. *Patient Prefer. Adherence* (2021.0) **15** 1461-1467. DOI: 10.2147/PPA.S316263 41. Perneger T.V., Courvoisier D.S., Hudelson P.M., Gayet-Ageron A.. **Sample Size for Pre-Tests of Questionnaires**. *Qual. Life Res.* (2015.0) **24** 147-151. DOI: 10.1007/s11136-014-0752-2 42. Anthoine E., Moret L., Regnault A., Sébille V., Hardouin J.-B.. **Sample Size Used to Validate a Scale: A Review of Publications on Newly-Developed Patient Reported Outcomes Measures**. *Health Qual. Life Outcomes* (2014.0) **12** 2. DOI: 10.1186/s12955-014-0176-2 43. Kellgren J.H., Lawrence J.S.. **Radiological Assessment of Osteo-Arthrosis**. *Ann. Rheum. Dis.* (1957.0) **16** 494-502. DOI: 10.1136/ard.16.4.494 44. **Radon Indication Registry for the Assessment of Pain Reduction, Increase of Quality of Life and Improvement in Body Functionality throughout Low-Dose Radon Hyperthermia Therapy** 45. Dawson J., Linsell L., Doll H., Zondervan K., Rose P., Carr A., Randall T., Fitzpatrick R.. **Assessment of the Lequesne Index of Severity for Osteoarthritis of the Hip in an Elderly Population**. *Osteoarthr. Cartil.* (2005.0) **13** 854-860. DOI: 10.1016/j.joca.2005.05.006 46. Ludwig F.J., Melzer C., Grimmig H., Daalmann H.H.. **Cross cultural adaptation of the lequesne algofunctional indices for german speaking patients with osteoarthritis of the hip and the knee**. *Die Rehabil.* (2002.0) **41** 249-257. DOI: 10.1055/s-2002-33273 47. Salaffi F., Stancati A.. *Scale Di Valutazione e Malattie Reumatiche* (2001.0) 143-144 48. Bilbao A., Martín-Fernández J., Arenaza J.C., García I., Tomás-García N., Trujillo-Martín E., García-Perez L.. **Validation of the EQ-5D-5L in Patients with Hip or Knee Osteoarthritis**. *Value Health* (2017.0) **20** A760. DOI: 10.1016/j.jval.2017.08.2151 49. Marten O., Greiner W.. **EQ-5D-5L Reference Values for the German General Elderly Population**. *Health Qual. Life Outcomes* (2021.0) **19** 76. DOI: 10.1186/s12955-021-01719-7 50. Meregaglia M., Malandrini F., Finch A.P., Ciani O., Jommi C.. **EQ-5D-5L Population Norms for Italy**. *Appl. Health Econ. Health Policy* (2022.0) **21** 289-303. DOI: 10.1007/s40258-022-00772-7 51. Salaffi F., Leardini G., Canesi B., Mannoni A., Fioravanti A., Caporali R., Lapadula G., Punzi L.. **GOnorthrosis and Quality Of Life Assessment (GOQOLA) Reliability and Validity of the Western Ontario and McMaster Universities (WOMAC) Osteoarthritis Index in Italian Patients with Osteoarthritis of the Knee**. *Osteoarthr. Cartil.* (2003.0) **11** 551-560. DOI: 10.1016/S1063-4584(03)00089-X 52. Stucki G., Meier D., Stucki S., Michel B.A., Tyndall A.G., Dick W., Theiler R.. **Evaluation of a German version of WOMAC (Western Ontario and McMaster Universities) Arthrosis Index**. *Z. Rheumatol.* (1996.0) **55** 40-49. PMID: 8868149 53. Darlow B., Abbott H., Bennell K., Briggs A.M., Brown M., Clark J., Dean S., French S., Hinman R.S., Krägeloh C.. **Knowledge about Osteoarthritis: Development of the Hip and Knee Osteoarthritis Knowledge Scales and Protocol for Testing Their Measurement Properties**. *Osteoarthr. Cartil. Open* (2021.0) **3** 100160. DOI: 10.1016/j.ocarto.2021.100160 54. **Organization for Economic Co-Operation and Development OECD International Network on Financial Education—Supplementary Questions: Optional Survey Questions** 55. **ISRCTN—ISRCTN93874734: Finding out What Patients Think about the Care They Receive for Osteoarthritis in General Practice in South Tyrol, Italy** 56. **World Medical Association World Medical Association Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Subjects**. *JAMA* (2013.0) **310** 2191-2194. DOI: 10.1001/jama.2013.281053 57. von Elm E., Altman D.G., Egger M., Pocock S.J., Gøtzsche P.C., Vandenbroucke J.P.. **The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting Observational Studies**. *Lancet* (2007.0) **370** 1453-1457. DOI: 10.1016/S0140-6736(07)61602-X 58. Althubaiti A.. **Information Bias in Health Research: Definition, Pitfalls, and Adjustment Methods**. *J. Multidiscip. Health* (2016.0) **9** 211-217. DOI: 10.2147/JMDH.S104807
--- title: 'Perceptions of and Preparedness for the Application of Pharmacoeconomics in Practice, among Final Year Bachelor of Pharmacy Students in South Africa: A National Cross-Sectional Study' authors: - Carlien Schmidt - Moliehi Matlala - Brian Godman - Amanj Kurdi - Johanna C. Meyer journal: Pharmacy year: 2023 pmcid: PMC10037603 doi: 10.3390/pharmacy11020054 license: CC BY 4.0 --- # Perceptions of and Preparedness for the Application of Pharmacoeconomics in Practice, among Final Year Bachelor of Pharmacy Students in South Africa: A National Cross-Sectional Study ## Abstract For the improvement of access to health, many countries including South Africa, have adopted universal healthcare. However, this requires skills to apply health technology assessments for the facilitation of investment decisions. This study aimed to ascertain final year Bachelor of Pharmacy (BPharm) students’ perceptions of the relevance of pharmacoeconomics in pharmacy practice, and their level of preparedness to apply pharmacoeconomic principles, using a quantitative, cross-sectional, and descriptive design. Data were collected using a self-administered questionnaire over 12 months, and included student demographics, knowledge about pharmacoeconomics and its applicability in practice, as well as students’ satisfaction with the appropriateness of the curriculum content. Five of nine universities offering pharmacy education took part. The overallstudent response rate was $38.1\%$ ($\frac{189}{496}$), with $26.2\%$ ($\frac{45}{172}$) of students signifying a good understanding of basic pharmacoeconomic concepts. Pharmacoeconomics application in South Africa was perceived to be relevant by $87.5\%$ ($\frac{140}{160}$); however, $47.0\%$ ($\frac{79}{168}$) felt they were not prepared to apply pharmacoeconomic principles in medicine management, and $86.7\%$ ($\frac{137}{158}$) wanted to acquire additional pharmacoeconomic knowledge. Whilst students’ perceptions of the relevance of pharmacoeconomics were positive, results indicated a gap in knowledge, understanding, and application. Addressing this gap may increase students’ preparedness to apply pharmacoeconomic principles and better equip them for the practical application of pharmacoeconomics post qualification. Consequently, we have started this process. ## 1. Introduction Pharmaceutical therapy-related expenditure has become an essential consideration to healthcare payers worldwide focusing on pharmacoeconomic analyses, with medicine expenditure in some low-and middle-income countries (LMICs) accounting for up to $70\%$ of total healthcare expenditure [1,2,3,4,5]. In high-income countries, there is an increasing focus on new medicines for cancer and orphan diseases as requested prices increase with often limited health gain coupled with the potential to overwhelm universal healthcare systems with growing expenditures [6,7,8]. These concerns have intensified the focus on the necessity for the scientific valuation of costs and consequences of pharmaceutical treatments, including vaccines to guide future investments and policy decisions [1,9]. Limited healthcare resources have also increased interest in assessing the value and feasibility of funding competing healthcare treatments and programmes by performing pharmacoeconomic evaluations, especially among LMICs [3,10,11,12,13,14]. The rise in pharmacoeconomic research application is expanding the need for qualified individuals, who are able to analyse and understand research findings and translate these into practice, especially among LMICs with resource and other concerns [3,10,15,16,17,18,19,20,21]. To date, South Africa has utilised dedicated methods in specific situations during reimbursement and pricing decision-making process for medicines rather than a broader use in priority setting, where there are competing demands across disease areas [9,22,23]. Pharmacoeconomic submissions to the South African National Department of Health (NDoH) Pricing Committee have taken place voluntarily and for selected medicines in the private health sector [22,23,24]. This is due to the fact that South Africa currently has an unequal two-tier healthcare system, with a public and private sector. The public sector, which is state funded caters for approximately $80\%$ of the population with the goal of universal healthcare (UHC) [25,26]. The private sector, which caters for approximately $20\%$ of the population, is largely funded through medical aid contributions or health insurance [27]. Consequently, medicines in the public sector are usually subject to tenders as they have typically lost their patents. [ 23,28]. This is not the case for possible new medicines in the private sector, with pharmacoeconomic guidelines in South Africa initially developed for the private sector, recommending a third-party payer perspective [22]. Nonetheless, pharmacoeconomic analyses are now emerging in the public sector to help appraising different treatment approaches, including different public health approaches, with competing demands for finite resources [29,30,31,32,33,34,35,36,37]. This inconsistent use of pharmacoeconomic submissions may imply that national pharmacoeconomic evaluations and education settings are still emerging, although there are moves to improve submissions through international comparisons [10,23,38]. In this regard, South Africa will soon follow in the footsteps of many international countries concerning pharmacoeconomic research application when the National Health Insurance (NHI) system, aimed at universal health coverage, is fully functional [25,26,38]. In the near future, it is believed that health technology assessment methodologies will be used to prioritise interventions in key areas, including health promotion, disease prevention and treatment, with the most cost-effective, evidence-based therapies and strategies being deployed and paid for under the NHI [25]. This is similar to situations in other LMICs [11,12,13,14]. We will continue to thoroughly monitor the situation with its implications for the necessity to increase pharmacoeconomic understanding among healthcare students, with the likeliness of strengthening the pharmacoeconomic guidelines in South Africa. With their knowledge of medicines and their costs, pharmacists are uniquely equipped to use pharmacoeconomic analyses to influence expenditure on medicines and the distribution of resources for medicines [1,3,15,23,25,26]. This builds on previous approaches, including ABC- and VEN-analyses of medicine use and expenditures in hospitals [15,39,40,41]. The South African National Drug Policy covers various activities contributing to effective medicines management. Within this policy, the pharmacist’s role is to ensure that the South African population receives the medicines they need at a cost that is affordable to them, and the healthcare system, is also clearly stipulated [24]. This role further enhances the need for appropriate pharmacoeconomic knowledge and skills among pharmacists going forward. The number of academic institutions providing education on pharmacoeconomic analyses has grown internationally over the past two decades, with many institutions also increasing the extent of health economics coursework amongst undergraduate pharmacy students [3,16,17,18,19,20]. Similarly, education on pharmacoeconomics is now incorporated into the South African Bachelor of Pharmacy (BPharm) programmes, complying with the South African Pharmacy Council’s exit level outcomes for entry-level pharmacists and the qualification standards of the South African Qualifications Authority [21,42]. The education of pharmacoeconomic principles in South Africa was briefly covered in a study published in 2005 [3]. While this study mentioned a requirement for increased education on pharmacoeconomics in developing countries, the individual perceptions of undergraduate pharmacy students in South Africa regarding pharmacoeconomics have yet to be robustly investigated [3]. A South African-based study focusing on pharmacoeconomics in the healthcare system in Gauteng province highlighted that pharmacoeconomics education is essential for increased awareness and understanding of the subject among healthcare decision-makers, with most respondents concurring that pharmacoeconomics education would aid them in their scope of practice [43]. However, this has not been considered further. A shortfall in healthcare professionals’ and students’ knowledge and understanding of pharmacoeconomics, and its application in medicines management, appears to be universal across countries, including LMICs [16,44,45,46]. The results of a South African study conducted in 2005 highlighted a lack of pharmacoeconomic knowledge among healthcare workers, epidemiologists, and trained staff, subsequently leading to an absence of measures to control resources in both the public and private healthcare sectors [43]. This is starting to be addressed in South Africa and other African countries with groups, such as HTAi having dedicated interest groups for developing countries (https://htai.org/hta-in-developing-countries/; accessed on 24 November 2022) as well as ISPOR with its African chapter (https://www.ispor.org/member-groups/global-groups/networks/africa-network/executive-committee; accessed on 24 November 2022). Applied pharmacoeconomics is often viewed as an important skill set for pharmacists internationally, and is acknowledged for improving health system performance across countries [3,43,45,47,48]. Despite this positive view on the role of pharmacoeconomics in maximising patients’ outcomes from the available healthcare resources, healthcare students and decision-makers commonly feel unprepared to apply pharmacoeconomic principles in practice [16,45,48]. Researchers universally suggest increased education regarding pharmacoeconomics during undergraduate and postgraduate health education programmes to address current gaps in knowledge and the application of pharmacoeconomic analyses [3,10,16,17,43,44,47]. The expected future use of pharmacoeconomic evidence in South Africa as part of the government’s NHI plan clearly indicates that pharmacists will increasingly need knowledge, skills, and capabilities for critical analyses and implementation of pharmacoeconomic research findings. One way to support this is through adequate education in this field among BPharm students in South Africa [3,16,20]. Consequently, this study was undertaken to determine final year BPharm students’ perceptions of the relevance of pharmacoeconomics in their future practice in South Africa and their level of preparedness to apply their knowledge in practice, and be able to undertake and critically review pharmacoeconomic studies in the future. ## 2.1. Study Design and Population This was a descriptive cross-sectional study among final-year BPharm students enrolled at the nine South African universities offering the BPharm programme. Therefore, all nine universities were invited to allow their final year BPharm students to participate in the study. Specific inclusion criteria for students were: (i) Final year students enrolled in a BPharm programme at a South African university; and (ii) students willing to participate in the study. ## 2.2. Data Collection Instrument and Procedure Data were collected through a structured, self-administered questionnaire available in English. The questionnaire content was based on a comprehensive review of literature sources, the current pharmacoeconomic curricular content for the South African BPharm programme, and pharmacoeconomic theoretical principles [3,16,17,19,42,44,45,47,49,50,51]. Three pharmacoeconomic experts initially reviewed the questionnaire, with their feedback subsequently incorporated into the revised questionnaire. A pre-test of the questionnaire was conducted among six pharmacist interns at the SMU School of Pharmacy to determine its face validity, length of completion, and relevance of the questions included. The final questionnaire consisted of 31 questions (Supplementary File S1), grouped into six sections to collect students’ demographic information, evaluate exposure to training on pharmacoeconomic principles and techniques during BPharm enrolment, and assess perceptions and preparedness pertaining to pharmacoeconomic analyses using a five-point Likert scale [44,45]. A Likert scale was considered the most appropriate measurement scale to assess respondents’ perceptions and preparedness, as it allows for the measurement of different levels of agreement and disagreement. As a result, providing good insights into respondents’ perceptions. Furthermore, the Likert scale has been used extensively in descriptive and quantitative studies across countries [52,53,54]. Students completed a paper-based version of the questionnaire, with completed questionnaires returned to the first author in a sealed envelope, or completed the questionnaire electronically using SurveyMonkey®, an online survey platform. Data were collected between November 2018 and December 2019, varying among universities with an average of 3 months per university. ## 2.3. Data Capture and Analysis Prior to analysis, participating universities were anonymised and recoded as “A”, “B”, “C”, “D”, and “E”. Captured data were proofread, cross-checked, and discrepancies resolved. Data analysis was descriptive and undertaken in custom formulated Microsoft Office Excel® spreadsheets. Categorical variables were summarised by frequency counts and percentages. Responses to the five-point Likert scale questions were condensed into three categories to facilitate the analysis and interpretation of results. Responses to open-ended questions were typed, categories were manually created, and responses were coded into these categories and counted where applicable. Even though comparing the individual universities was not the primary aim of this study, a sensitivity analysis was conducted to assess the effect of variation in response rates amongst the universities on the study outcomes. For this purpose, we grouped universities based on their response rates into “Low-“, “Medium-“, and “High-“ response rates. Subsequently, we compared the scores for two randomly selected study outcomes, namely “Level of understanding of pharmacoeconomics” and “Preparedness to apply pharmacoeconomics in practice” between the three groups, using one-way ANOVA and Fisher Exact tests, respectively. Herein, p-values < 0.05 were considered as statistically significant. ## 2.4. Ethical Considerations The Sefako Makgatho University Research Ethics Committee provided ethics clearance for the study (SMUREC/P/$\frac{97}{2018}$:PG), after which the nine universities were invited to participate in the study. Upon acceptance of the invitation by a university, permission to conduct the study was formally requested, which included submission of the protocol to the respective university’s research ethics committees. Only upon receipt of permission and ethics clearance from a particular university students were invited to participate in the survey. Participation was voluntary, responses were anonymous, and no personal, identifying information was collected. Students participating in the survey first provided informed consent before completing the questionnaire. Data were treated as highly confidential, with completed questionnaires stored under secure conditions. All data are stored securely for future reference and for a period of 5 years, after which it will be destroyed according to university policies. ## 3.1. Response Rate Eight of the nine invited universities offering the BPharm programme in South Africa responded positively to the invitation. Five of the eight universities were able to provide ethics clearance for their students to participate in the study during the allocated study period. The ethical clearance process at two of the three remaining universities was delayed considerably, with permission granted only after data collection for the study had been concluded. No further response was received from the one remaining university. The final target study population from the five universities included 496 final year BPharm students, from which an overall response rate of $38.1\%$ ($\frac{189}{496}$) was obtained, ranging from $18.1\%$ to $93.9\%$ at individual universities (Table 1). Four of the universities’ students responded using the paper-based version of the questionnaire, while the students at one of the universities responded using the electronic platform (Table 1). Overall, $48.1\%$ ($\frac{91}{189}$) of students answered all questions in the questionnaire. As a result of inconsistent responses, sample sizes varied between questions. Twenty-two of the 189 students ($11.6\%$) provided additional comments on pharmacoeconomics. ## 3.2. Respondent Demographics Table 2 demonstrates that the mean age of students was 24.3 years (SD = 2.34), ranging between 20.0 and 32.9 years, with the majority being female ($71.4\%$). A few students ($7.1\%$; $\frac{13}{184}$) held other degrees. Nearly half of the students surveyed ($49.2\%$) intended to complete their internships at public sector institutional pharmacies. ## 3.3. Pharmacoeconomics Education during the BPharm Programme Overall, $37.8\%$ ($$n = 62$$) of BPharm students in this study indicated that pharmacoeconomics was covered under “Hospital Pharmacy Practice”-related subjects (Table 3). Of the 178 students, 74 ($41.6\%$) indicated that pharmacoeconomics was presented during their fourth year of the programme. The majority of students ($87.3\%$; $\frac{151}{173}$) underwent a formal assessment of their knowledge of pharmacoeconomics during their BPharm programme. Table 3 shows that most students ($88.0\%$) recalled being taught pharmacoeconomics through lectures. The number of hours allocated to pharmacoeconomics in students’ timetables ranged from 0.1 to 40 h (mean = 4.4; SD = 4.52). Of the 22 additional comments at the end of the survey, five students ($22.7\%$) said that “pharmacoeconomics should be a subject/module/course on its own”. ## 3.4. Understanding of Pharmacoeconomic Concepts Table 4 demonstrates that, of the 172 students indicating their level of understanding of pharmacoeconomic concepts, 40 ($23.3\%$) signified an overall poor understanding of these concepts, whereas 83 ($48.3\%$) had a fair understanding. Only over a quarter of students signified an overall good understanding ($\frac{45}{172}$), and $66.2\%$ were able to correctly answer questions regarding the scope of pharmacoeconomics ($\frac{104}{157}$). However, $37.0\%$ of students ($\frac{57}{154}$) wrongly indicated that “pharmacoeconomics calculates the costs of medicines and treatments only”. Only $23.4\%$ of students ($\frac{33}{138}$) provided correct answers to each type of analysis, namely, cost-minimization analysis, cost-effectiveness analysis, cost-utility analysis, and cost-benefit analysis. ## 3.5. Relevance of Pharmacoeconomics in Practice Overall, the majority of students ($87.5\%$; $\frac{140}{160}$) perceived the application of pharmacoeconomics in South African medicines management as “relevant” (Table 5). Of the 22 additional comments, most students ($88.8\%$) felt that applying pharmacoeconomics in practice was an essential skill that pharmacists should possess, with $84.0\%$ indicating that pharmacists should be responsible for performing pharmacoeconomic evaluations in practice. ## 3.6. Preparedness for Application of Pharmacoeconomics in Practice Only over a third of students ($38.1\%$; $\frac{64}{168}$) felt that their undergraduate exposure to pharmacoeconomics was insufficient to understand basic principles (Table 6). Nearly half of the students ($54.2\%$; $\frac{91}{168}$) perceived pharmacoeconomics as “interesting” and “enjoyable”, with $47.0\%$ ($\frac{79}{168}$) who felt not adequately prepared to apply pharmacoeconomics in practice. Less than half ($45.7\%$; $\frac{75}{164}$) of the students thought of themselves as competent to perform basic pharmacoeconomic analyses. ## 3.7. Sensitivity Analysis Based on the universities’ response rates (see Table 1), Universities A and C were grouped into “Low response rate“, Universities B and E into “Medium response rate“, and University D labelled as “High response rate”. The mean scores for understanding of basic pharmacoeconomic concepts ($$p \leq 0.006$$) and advanced pharmacoeconomic concepts ($p \leq 0.001$) were statistically significantly different between the three groups (see Table 7). However, post-hoc analysis showed no significant difference between the high- and low-response rate groups for basic ($$p \leq 0.991$$) and advanced ($$p \leq 0.774$$) understanding of pharmacoeconomic analyses. In terms of “Preparedness to apply pharmacoeconomics in practice,” there was no statistically significant difference between the three groups for both being able to interpret the results of pharmacoeconomic analyses for decision-making ($$p \leq 0.810$$) and being adequately prepared to apply pharmacoeconomic concepts in practice to conduct the analyses ($$p \leq 0.792$$) (see Table 8). ## 3.8. Future Education in Pharmacoeconomics The vast majority of students ($93.8\%$; $\frac{152}{162}$) believed that future education regarding pharmacoeconomic studies and their application was essential to their role as pharmacists, while only five ($3.1\%$) perceived further education as “not necessary”. Similarly, most surveyed students ($84.4\%$; $\frac{135}{160}$) would have wanted more education on pharmacoeconomics during their BPharm training. Among the additional comments provided, $27.2\%$ ($\frac{6}{22}$) of the students thought that undergraduate pharmacoeconomics tuition should be increased. The majority of students ($86.7\%$; $\frac{137}{158}$) wanted to acquire further pharmacoeconomics knowledge, of whom two thirds ($66.4\%$; $\frac{91}{137}$) wanted to acquire further knowledge through continuous professional development (CPD) programmes, 36 ($26.3\%$) through self-study, and 40 ($29.2\%$) through postgraduate studies. Among the 22 additional comments at the end of the survey, three students ($13.6\%$) indicated that they would like to acquire more knowledge regarding pharmacoeconomics. ## 4. Discussion The overall response rate of $38.1\%$ is seen as acceptable for voluntary questionnaire surveys, and similar to other published studies in this area [44,45,47,55,56]. The study results principally highlighted two key issues for the future, which are the most important outcomes of this study. First, few pharmacy students had an overall good understanding of pharmacoeconomic concepts. However, only over half of those surveyed felt they received enough teaching exposure to pharmacoeconomics to understand the basic principles and concepts during their university training. However, the majority wanted to receive more undergraduate training and tuition. Second, less than half of the students participating in the survey felt competent to perform basic pharmacoeconomic analyses, with more students considering themselves as “not prepared” to conduct these studies compared with those who feel prepared. Most students in our study underwent education regarding pharmacoeconomics in their fourth (final) year. This is consistent with the study by Catić and Skrbo in Bosnia, in which most pharmacy students were taught pharmacoeconomics in their fifth (final) year [16]. We assume that students nearing the end of their BPharm studies have the necessary fundamental medicine-related knowledge to fully understand pharmacoeconomics and its application. However, this is not always the case. According to most students in this study, pharmacoeconomics was a mandatory subject/module/course, which concurs with the findings from similar studies [3,16,19,20,57,58,59,60]. However, the mean hours spent teaching pharmacoeconomics in South Africa varied significantly compared with similar studies outside of South Africa [3,17,19,47,50]. This might help in explaining why more than a third of students in our study felt they did not receive sufficient undergraduate exposure to pharmacoeconomics and its principles. This may have adversely affected their understanding of pharmacoeconomic concepts, competence to perform basic analyses, and preparedness to conduct pharmacoeconomic analyses in practice, which urgently needs to be addressed [61]. Encouragingly, few students in our study had an overall poor understanding of pharmacoeconomic concepts or principles. This compares with other studies, where most respondents found these concepts unclear or difficult to understand or indicated that they were “not very familiar” or “slightly familiar” with these concepts [16,46,49,61]. At the same time, the number of students in our study with an overall good understanding of pharmacoeconomic concepts and their application in medicines management was similar to other studies [44,45,47,49]. When questioned about the scope of pharmacoeconomic analyses, more students in our study knew the correct answer to the question, compared with $39.0\%$ of students in the study by Catić and Skrbo [16]. However, more students in our study incorrectly indicated that “pharmacoeconomics examines and calculates the costs of medicines and treatments only” compared with only $13.0\%$ of the respondents recorded in the study by Catić and Skrbo [16]. Of concern, many students in our study incorrectly indicated that budget impact analysis formed part of the scope of pharmacoeconomics, while not being considered as a pharmacoeconomic analysis sub-type in reality. This is an issue to address as budget impact analyses are increasingly important in LMICs when assessing the possible role and value of new medicines [62,63,64,65]. This is balanced against the finding that the number of students correctly indicating that pharmacoeconomic studies compare different therapeutic interventions was appreciably higher than only $7.0\%$ of students reported by Catić and Skrbo [16]. *In* general, more students in our study had a good knowledge of the scope of pharmacoeconomics compared with only $38.9\%$ of students reported by Catić and Skrbo [16]. Nearly a quarter of students in our survey were also able to provide the correct measure to each of the basic pharmacoeconomic analyses concurrent with similar studies [45,46,49]. The level of pharmacoeconomics understanding among the students in our study, especially regarding the scope of pharmacoeconomics, is an important factor to consider going forward. Any healthcare professional tasked with even the most basic pharmacoeconomic analyses would have to know what inputs and outcomes they are measuring when performing these evaluations in practice. Evidently, the level of students’ understanding of pharmacoeconomics is a fundamental cornerstone of their ability to perform pharmacoeconomic analyses successfully. Consequently, education offers a golden opportunity to improve their knowledge and understanding of pharmacoeconomics. The importance of addressing this gap in BPharm students’ knowledge and understanding is supported by earlier studies highlighting that the education of healthcare professionals regarding pharmacoeconomics contributes to the financial sustainability of healthcare systems [3,43,46,48]. Moreover, encouragingly, an appreciable portion of students in our study thought that applying pharmacoeconomics was an essential skill that South African pharmacists should possess. This concurs with the findings of studies conducted amongst different student cohorts in India, Japan, and South Africa [3,43,45,49]. Pharmacists, with their unique knowledge of medicines and key aspects of it, including their acquisition costs, can effectively contribute to any conservation regarding health budgets, which is important in enhancing equal access to pharmaceutical care, especially in developing countries such as South Africa [3,15,66,67]. Consequently, it was encouraging that most students in our study agreed that the application of pharmacoeconomics would benefit the South African healthcare system. This finding strongly correlates with Tahashildar et al. in India [45]. Moreover, how students thought of instances in which pharmacoeconomic evaluations could be used in South Africa was congruent with the findings by Catić and Skrbo from Bosnia and those of Modiba from South Africa [16,43]. Furthermore, encouragingly, most of the surveyed students intended to complete their internship in South African public sector institutions, challenged with significant workforce shortages as well as medicine shortages, requiring pro-active ways to deal with this without seriously impeding on patient care [9,26,68,69]. This is important for the future as South Africa implements UHC with ever-increasing demands on scarce resources, starkly contrasting the findings of the study by Armstrong et al. [ 3,26,49,70]. Only over a third of the students in our current study felt that they needed to receive more pharmacoeconomics exposure in their training. In addition, the majority of those surveyed would have wanted more pharmacoeconomics training at an undergraduate level. This is similar to a 2002 European Pharmaceutical Student Association survey involving 22 European countries, where only over half of the students surveyed ($56\%$) indicated that the level of pharmacoeconomics education they received during their education was poor [3]. This needs addressing in the future, especially as only $19.0\%$ of pharmacy students enrolled at the Lebanese American University School of Pharmacy believed that the number of hours spent preparing them to analyse pharmacoeconomic research was inadequate [71]. Most healthcare professionals and postgraduate medical students participating in the study by Tahashildar et al. did not feel comfortable in conducting pharmacoeconomic analyses. This was despite having undergone a formal assessment of their pharmacoeconomics knowledge [45]. There were similar findings in the study by Umair Khan et al. [ 48]. Both studies concur with our study, where only a limited number of students surveyed felt prepared and competent to perform basic pharmacoeconomic analyses in practice [45,48]. This study finding is emphasised by Kolassa, who suggested that pharmacy curricula did not adequately prepare students [72]. Our study also showed that half of the South African students wanted to obtain additional pharmacoeconomics knowledge through CPD programmes post qualification. This compares to approximately three-quarters of postgraduate students surveyed in the study by Jayasree et al. [ 44]. More students in our study wanted to acquire further pharmacoeconomics knowledge through self-study, which contrasts with $11.0\%$ in the study by Jayasree et al. [ 44]. However, the small number of students in our study wanting to acquire pharmacoeconomics knowledge through postgraduate studies corresponded with findings by Catić and Skrbo, but was in contrast with the $55.0\%$ of respondents in the study by Jayasree et al., who believed that pharmacoeconomics should be included in postgraduate studies [16,44]. We are aware of the limitations with this study. Firstly, four of the nine universities in the country offering the BPharm programme did not participate in the study during the study period as a number of them were unable to provide ethical approval in time. Secondly, two of the five participating universities had a lower response rate compared to the other universities. However, the overall rates are similar to other published studies in this area [44,45,47,55,56]. In this regard, a strength of this study was that it was conducted in the final semester of the BPharm study, assuming that students would have undergone the necessary education to participate. In addition, the study’ the future requirements for pharmacoeconomic teaching for student pharmacists across South Africa to be able to equip them for the future. Consequently, we feel that the overall response rate from 189 pertinent students is extremely helpful, with the findings seen as robust in providing direction for the future. Despite the above, we recognise that the variation in response rates among the universities could have introduced a non-response bias, however, based on the sensitivity analysis results (see Table 7 and Table 8) it is unlikely that variation in the response rate could explain the variation in the study outcomes among the universities. The observed variation could possibly be explained by variations in the academic performance amongst the universities, such as staff to student ratio and emphasis on the teaching of pharmacoeconomics, emphasising standardisation of pharmacoeconomic curricula amongst universities in South Africa. ## 5. Conclusions There is a recognised need to develop a pool of South African personnel who are able to conduct and evaluate pharmacoeconomic analyses as South Africa moves toward UHC. Consequently, it was encouraging to see that most BPharm students surveyed perceived pharmacoeconomics in South African medicines management as relevant to their future needs. In addition, a demand for further pharmacoeconomics education exists among the next generation of pharmacists. South African undergraduate pharmacy students appear to correlate with their international counterparts regarding the gap in their understanding and knowledge of pharmacoeconomic concepts and their preparedness for practical application, which needs addressing going forward. Consequently, pharmacoeconomics education should remain in the South African BPharm programme curriculum; however, the current content requires expansion. Addressing this gap during South African undergraduate pharmacy education should increase students’ understanding of pharmacoeconomic concepts and their preparedness for applying these analyses in practice post qualification to benefit the South African healthcare system. The BPharm curriculum is currently under review and we will continue to monitor it in future research projects. ## References 1. Bootman J., Townsend R., McGhan W.. **Introduction to Pharmacoeconomics**. *Principles of Pharmacoeconomics* (2005) 1-18 2. Drummond M.. **Pharmacoeconomics: Friend or foe?**. *Ann. Rheum. Dis.* (2006) **65** iii44-iii47. DOI: 10.1136/ard.2006.058602 3. Nwokeji E., Rascati K.. **Pharmacoeconomic Education in Colleges of Pharmacy Outside of the United States**. *Am. J. Pharm. Educ.* (2005) **69** 348-355. DOI: 10.5688/aj690352 4. Ofori-Asenso R., Brhlikova P., Pollock A.M.. **Prescribing indicators at primary health care centers within the WHO African region: A systematic analysis (1995–2015)**. *BMC Public Health* (2016) **16**. DOI: 10.1186/s12889-016-3428-8 5. Cameron A., Ewen M., Ross-Degnan D., Ball D., Laing R.. **Medicine prices, availability, and affordability in 36 developing and middle-income countries: A secondary analysis**. *Lancet* (2009) **373** 240-249. DOI: 10.1016/S0140-6736(08)61762-6 6. Godman B., Hill A., Simoens S., Selke G., Selke Krulichova I., Zampirolli Dias C., Martin A.P., Oortwijn W., Timoney A., Gustafsson L.L.. **Potential approaches for the pricing of cancer medicines across Europe to enhance the sustainability of healthcare systems and the implications**. *Expert. Rev. Pharm. Outcomes Res.* (2021) **21** 527-540. DOI: 10.1080/14737167.2021.1884546 7. Luzzatto L., Hyry H.I., Schieppati A., Costa E., Simoens S., Schaefer F., Roos J.C., Merlini G., Kääriäinen H., Garattini S.. **Outrageous prices of orphan drugs: A call for collaboration**. *Lancet* (2018) **392** 791-794. DOI: 10.1016/S0140-6736(18)31069-9 8. Marselis D., Hordijk L.. **From blockbuster to “nichebuster”: How a flawed legislation helped create a new profit model for the drug industry**. *BMJ* (2020) **370** m2983. DOI: 10.1136/bmj.m2983 9. Miot J., Thiede M.. **Adapting Pharmacoeconomics to Shape Efficient Health Systems en route to UHC—Lessons from Two Continents**. *Front. Pharmacol.* (2017) **8** 715. DOI: 10.3389/fphar.2017.00715 10. Gavaza P., Rascati K., Oladapo A., Khoza S.. **The State of Health Economic Research in South Africa**. *Pharmacoeconomics* (2012) **30** 925-940. DOI: 10.2165/11589450-000000000-00000 11. Al-Ziftawi N.H., Shafie A.A., Mohamed Ibrahim M.I.. **Cost-effectiveness analyses of breast cancer medications use in developing countries: A systematic review**. *Expert. Rev. Pharm. Outcomes Res.* (2021) **21** 655-666. DOI: 10.1080/14737167.2020.1794826 12. Erfani P., Bhangdia K., Stauber C., Mugunga J.C., Pace L.E., Fadelu T.. **Economic Evaluations of Breast Cancer Care in Low- and Middle-Income Countries: A Scoping Review**. *Oncologist* (2021) **26** e1406-e1417. DOI: 10.1002/onco.13841 13. Saokaew S., Rayanakorn A., Wu D.B., Chaiyakunapruk N.. **Cost Effectiveness of Pneumococcal Vaccination in Children in Low- and Middle-Income Countries: A Systematic Review**. *Pharmacoeconomics* (2016) **34** 1211-1225. DOI: 10.1007/s40273-016-0439-3 14. Portnoy A., Jit M., Lauer J., Blommaert A., Ozawa S., Stack M., Murray J., Hutubessy R.. **Estimating costs of care for meningitis infections in low- and middle-income countries**. *Vaccine* (2015) **33** A240-A247. DOI: 10.1016/j.vaccine.2014.11.061 15. Purkiss R.. **Pharmacoeconomics-the importance for pharmacists**. *Hosp. Pharm.* (2006) **13** 34 16. Catić T., Skrbo S.. **Pharmacoeconomic Education for Pharmacy Students in Bosnia and Herzegovina**. *Mater. Socio Med.* (2013) **25** 282-285. DOI: 10.5455/msm.2013.25.282-285 17. Soliman A., Hussein M., Abdulhalim A.. **Pharmacoeconomic Education in Egyptian Schools of Pharmacy**. *Am. J. Pharm. Educ.* (2013) **77** 57. DOI: 10.5688/ajpe77357 18. Rattinger G., Jain R., Ju J., Mullins C.. **Principles of Economics Crucial to Pharmacy Students’ Understanding of the Prescription Drug Market**. *Am. J. Pharm. Educ.* (2008) **72** 61-65. DOI: 10.5688/aj720361 19. Reddy M., Rascati K., Wahawisan J., Rascati M.. **Pharmacoeconomic Education in US Colleges and Schools of Pharmacy: An Update**. *Am. J. Pharm. Educ.* (2008) **72** 51. DOI: 10.5688/aj720351 20. Thomas D., Sam Sundararaj K., Shirwaikar A., Tarn Y.. **Inclusion of Pharmacoeconomics course in the Undergraduate Pharmacy Education: A Global Trend Review**. *Indian J. Pharm. Pract.* (2016) **9** 147-151. DOI: 10.5530/ijopp.9.3.3 21. Bosmans S., Kairuz T.. **Undergraduate pharmacy education in two countries in the southern hemisphere**. *Pharm. Educ.* (2009) **9** 44-49 22. Carapinha J.L.. **A comparative review of the pharmacoeconomic guidelines in South Africa**. *J. Med. Econ.* (2017) **20** 37-44. DOI: 10.1080/13696998.2016.1223679 23. **Medicines and Related Substances Act (101 of 1965): Regulations Relating to a Transparent Pricing System for Medicines and Scheduled Substances: Publication of the Guidelines for Pharmacoeconomic Submissions**. (2013) 24. **National Drug Policy for South Africa**. (1996) 25. 25. South African National Department of Health (NDoH) National Health Act, 2003: National Health Insurance Policy: Towards Universal Health CoverageGovernment Gazette, The Government PrinterPretoria, South Africa2017Volume 624. *National Health Act, 2003: National Health Insurance Policy: Towards Universal Health Coverage* (2017) **Volume 624** 26. Meyer J.C., Schellack N., Stokes J., Lancaster R., Zeeman H., Defty D., Godman B., Steel G.. **Ongoing Initiatives to Improve the Quality and Efficiency of Medicine Use within the Public Healthcare System in South Africa; A Preliminary Study**. *Front. Pharmacol.* (2017) **8** 751. DOI: 10.3389/fphar.2017.00751 27. Human A.. **A Tale of Two Tiers: Inequality in South Africa’s Health Care System**. *UBCMJ* (2009) **1** 33 28. Godman B., Leong T., Abubakar A.R., Kurdi A., Kalemeera F., Rwegerera G.M., Patrick O., Lum Niba L., Ibrahim K., Amu A.A.. **Availability and Use of Long-Acting Insulin Analogues Including Their Biosimilars across Africa: Findings and Implications**. *Intern. Med.* (2021) **11** 1-17 29. Edoka I., Kohli-Lynch C., Fraser H., Hofman K., Tempia S., McMorrow M., Ramkrishna W., Lambach P., Hutubessy R., Cohen C.. **A cost-effectiveness analysis of South Africa’s seasonal influenza vaccination programme**. *Vaccine* (2021) **39** 412-422. DOI: 10.1016/j.vaccine.2020.11.028 30. Fraser H., Tombe-Mdewa W., Kohli-Lynch C., Hofman K., Tempia S., McMorrow M., Lambach P., Ramkrishna W., Cohen C., Hutubessy R.. **Costs of seasonal influenza vaccination in South Africa**. *Influ. Other Respir. Viruses* (2022) **16** 873-880. DOI: 10.1111/irv.12987 31. Jamieson L., Johnson L.F., Matsimela K., Sande L.A., d’Elbée M., Majam M., Johnson C., Chidarikire T., Hatzold K., Terris-Prestholt F.. **The cost effectiveness and optimal configuration of HIV self-test distribution in South Africa: A model analysis**. *BMJ Glob. Health* (2021) **6** e005598. DOI: 10.1136/bmjgh-2021-005598 32. Campos N.G., Lince-Deroche N., Chibwesha C.J., Firnhaber C., Smith J.S., Michelow P., Meyer-Rath G., Jamieson L., Jordaan S., Sharma M.. **Cost-Effectiveness of Cervical Cancer Screening in Women Living With HIV in South Africa: A Mathematical Modeling Study**. *J. Acquir. Immune. Defic. Syndr.* (2018) **79** 195-205. DOI: 10.1097/QAI.0000000000001778 33. Pumpalova Y., Rogers A.M., Tan S.X., Herbst C.L., Ruff P., Neugut A.I., Hur C.. **Modeling the Cost-Effectiveness of Adjuvant Chemotherapy for Stage III Colon Cancer in South African Public Hospitals**. *JCO Glob. Oncol.* (2021) **7** 1730-1741. DOI: 10.1200/GO.21.00279 34. Alshreef A., MacQuilkan K., Dawkins B., Riddin J., Ward S., Meads D., Taylor M., Dixon S., Culyer A.J., Ruiz F.. **Cost-Effectiveness of Docetaxel and Paclitaxel for Adjuvant Treatment of Early Breast Cancer: Adaptation of a Model-Based Economic Evaluation From the United Kingdom to South Africa**. *Value Health Reg Issues* (2019) **19** 65-74. DOI: 10.1016/j.vhri.2019.03.001 35. Michaeli D.T., Stoycheva S., Marcus S.M., Zhang W., Michaeli J.C., Michaeli T.. **Cost-Effectiveness of Bivalent, Quadrivalent, and Nonavalent HPV Vaccination in South Africa**. *Clin. Drug Investig.* (2022) **42** 333-343. DOI: 10.1007/s40261-022-01138-6 36. Saxena A., Stacey N., Puech P.D.R., Mudara C., Hofman K., Verguet S.. **The distributional impact of taxing sugar-sweetened beverages: Findings from an extended cost-effectiveness analysis in South Africa**. *BMJ Glob. Health* (2019) **4** e001317. DOI: 10.1136/bmjgh-2018-001317 37. Erzse A., Stacey N., Chola L., Tugendhaft A., Freeman M., Hofman K.. **The direct medical cost of type 2 diabetes mellitus in South Africa: A cost of illness study**. *Glob. Health Action* (2019) **12** 1636611. DOI: 10.1080/16549716.2019.1636611 38. Marsh S.E., Truter I.. **The South African Guidelines for Pharmacoeconomic Submissions’ Evidence Requirements Compared with Other African Countries and The National Institute for Health and Care Excellence in England**. *Expert Rev. Pharm. Outcomes Res.* (2020) **20** 155-168. DOI: 10.1080/14737167.2019.1615451 39. Mfizi E., Niragire F., Bizimana T., Mukanyangezi M.F.. **Analysis of pharmaceutical inventory management based on ABC-VEN analysis in Rwanda: A case study of Nyamagabe district**. *J. Pharm. Policy Pract.* (2023) **16** 30. DOI: 10.1186/s40545-023-00540-5 40. Kivoto P.M., Mulaku M., Ouma C., Ferrario A., Kurdi A., Godman B., Oluka M.. **Clinical and Financial Implications of Medicine Consumption Patterns at a Leading Referral Hospital in Kenya to Guide Future Planning of Care**. *Front. Pharmacol.* (2018) **9** 1348. DOI: 10.3389/fphar.2018.01348 41. Gebretekle G.B., Mariam D.H., Mac S., Abebe W., Alemayehu T., Degu W.A., Libman M., Yansouni C.P., Fenta T.G., Semret M.. **Cost-utility analysis of antimicrobial stewardship programme at a tertiary teaching hospital in Ethiopia**. *BMJ Open* (2021) **11** e047515. DOI: 10.1136/bmjopen-2020-047515 42. **Registered Qualification: Bachelor of Pharmacy (Qualification ID: 72784)**. (2018) 43. Modiba W.. *Pharmacoeconomics in the Healthcare System in Gauteng* (2005) 44. Jayasree D., Uppu B., Devi B.. **A study on awareness of pharmacoeconomics among postgraduates in a tertiary care teaching hospital**. *Int. J. Res. Med. Sci.* (2016) 1597-1603. DOI: 10.18203/2320-6012.ijrms20161233 45. Tahashildar J., Kota K., Kumar K.S., Tahashildar J., Mohanty L., Ahmed S.Y.. **Assessment of the Knowledge and Perceptions about Pharmacoeconomics among Medical postgraduate students and Healthcare professionals at a Tertiary care teaching hospital, Udaipur, India**. *Int. J. Curr. Res. Acad. Rev.* (2015) **3** 135-142 46. Tabassum R., Hussain S., Banday M.. **Evaluation of pharmacoeconomics awareness and its application among postgraduates of a tertiary care hospital: A cross-sectional observational study**. *Asian J. Pharm. Clin. Res.* (2016) **9** 145-147 47. Savkar M., Bhat N., Deepika G.. **Evaluation of pharmacoeconomics awareness among postgraduates: A questionnaire-based study**. *Indian J. Basic Appl. Med. Res.* (2014) **3** 135-141 48. Umair Khan M., Ahmad A., Hussain K., Salam A., Hasnain Z., Patel I.. **The need for redesigned pharmacy practice courses in Pakistan: The perspectives of senior pharmacy students**. *J. Educ. Eval. Health Prof.* (2015) **12** 27. DOI: 10.3352/jeehp.2015.12.27 49. Armstrong E., Akaho E., Fujii M., Hosono K.. **Evaluation of Teaching Pharmacoeconomics to Japanese Students**. *J. Pharm. Teach.* (1996) **5** 67-85. DOI: 10.1300/J060v05n03_06 50. Marinova T., Rascati K.. **Pharmacoeconomics Education in US Colleges and Schools of Pharmacy**. *Am. J. Pharm. Educ.* (2013) **77** 145. DOI: 10.5688/ajpe777145 51. Rascati K., Conner T., Draugalis J.. **Pharmacoeconomic Education in US Colleges and Schools of Pharmacy**. *Am. J. Pharm. Educ.* (1998) **62** 167-169 52. Garcia M.M., Barbosa M.M., Silva R.M., Reis E.A., Alvares J., Acurcio F.D.A., Godman B., Junior A.A.G.. **Indicator of access to medicines in relation to the multiple dimensions of access**. *J. Comp. Eff. Res.* (2019) **8** 1027-1041. DOI: 10.2217/cer-2019-0031 53. Al Zahrani E.M., Al Naam Y.A., AlRabeeah S.M., Aldossary D.N., Al-Jamea L.H., Woodman A., Shawaheen M., Altiti O., Quiambao J.V., Arulanantham Z.J.. **E-Learning experience of the medical profession’s college students during COVID-19 pandemic in Saudi Arabia**. *BMC Med. Educ.* (2021) **21**. DOI: 10.1186/s12909-021-02860-z 54. Gupta M.K., Vohra C., Raghav P.. **Assessment of knowledge, attitudes, and practices about antibiotic resistance among medical students in India**. *J. Fam. Med. Prim. Care.* (2019) **8** 2864-2869. DOI: 10.4103/jfmpc.jfmpc_504_19 55. Burger M., Fourie J., Loots D., Mnisi T., Schellack N., Bezuidenhout S., Meyer J.. **Knowledge and perceptions of antimicrobial stewardship concepts among final year pharmacy students in pharmacy schools across South Africa**. *S. Afr. J. Infect. Dis.* (2016) **31** 84-90 56. Fosnacht K., Sarraf S., Howe E., Peck L.. **How Important are High Response Rates for College Surveys?**. *Rev. High. Educ.* (2017) **40** 245-265. DOI: 10.1353/rhe.2017.0003 57. Akazawa M., Yanagi N.. **Pharmacoeconomics Education in Schools of Pharmacy in Japan: A Questionnaire Survey**. *Value Health* (2018) **21** S55. DOI: 10.1016/j.jval.2018.07.418 58. Freitas G., Balbinotto G.. **Pharmacoeconomic Education in Brazilian Schools of Pharmacy**. *Value Health* (2014) **17** A436. DOI: 10.1016/j.jval.2014.08.1126 59. Jana S., Mondal P.. **Pharmacoeconomics: The need to sensitize undergraduate medical students**. *Indian J. Pharmacol.* (2005) **37** 277. DOI: 10.4103/0253-7613.16849 60. Makhinova T., Makhinova E., Rascati K.. **PHP146 Pharmacoeconomic Education for Pharmacy Students in the Russian Federation**. *Value Health* (2012) **15** A314-A315. DOI: 10.1016/j.jval.2012.08.688 61. Scott D., Pedersen C.. **An Outcomes Research (Pharmacoeconomics/Pharmacoepidemiology) Course for PharmD Students**. *Am. J. Pharm. Educ.* (2000) **64** 114-120 62. Faleiros D.R., Nunes da Silva E., Santos A.C., Godman B.B., Goncalves Pereira R., Guerra Junior A.A.. **Adoption of new therapies in the treatment of Hepatitis: A verification of the accuracy of budget impact analysis to guide investment decisions**. *Expert Rev. Pharm. Outcomes Res.* (2022) **22** 927-939. DOI: 10.1080/14737167.2022.2057950 63. Faleiros D.R., Álvares J., Almeida A.M., De Araújo V.E., Andrade E.I., Godman B.B., Acurcio F.A., Guerra Junior A.A.. **Budget impact analysis of medicines: Updated systematic review and implications**. *Expert. Rev. Pharm. Outcomes Res.* (2016) **16** 257-266. DOI: 10.1586/14737167.2016.1159958 64. Catharina de Beer J., Snyman J., Ker J., Miller-Janson H., Stander M.. **Budget Impact Analysis of Empagliflozin in the Treatment of Patients With Type 2 Diabetes With Established Cardiovascular Disease in South Africa**. *Value Health Reg. Issues* (2022) **33** 91-98. DOI: 10.1016/j.vhri.2022.08.012 65. Elsisi G.H., Anwar M.M., Khattab M., Elebrashy I., Wafa A., Elhadad H., Awad M., Carapinha J.L.. **Budget impact analysis for dapagliflozin in type 2 diabetes in Egypt**. *J. Med. Econ.* (2020) **23** 908-914. DOI: 10.1080/13696998.2020.1764571 66. Dalton K., Byrne S.. **Role of the pharmacist in reducing healthcare costs: Current insights**. *Integr. Pharm. Res. Pract.* (2017) **6** 37-46. DOI: 10.2147/IPRP.S108047 67. Mori A., Gavaza P., Robberstad B.. **Role of pharmacoeconomics in developing countries**. *Farmeconomia* (2013) **14** 3-5. DOI: 10.7175/fe.v14i1.638 68. Modisakeng C., Matlala M., Godman B., Meyer J.C.. **Medicine shortages and challenges with the procurement process among public sector hospitals in South Africa; findings and implications**. *BMC Health Serv. Res.* (2020) **20**. DOI: 10.1186/s12913-020-05080-1 69. Chigome A.K., Matlala M., Godman B., Meyer J.C.. **Availability and Use of Therapeutic Interchange Policies in Managing Antimicrobial Shortages among South African Public Sector Hospitals; Findings and Implications**. *Antibiotics* (2019) **9**. DOI: 10.3390/antibiotics9010004 70. Groenewald Y.. **Fin24**. (2017) 71. Nasser S., Saad A., Karaoui L.. **Mapping of the biomedical literature evaluation competencies based on pharmacy students’ feedback**. *Bio. Med. Cent. Med. Educ.* (2016) **16**. DOI: 10.1186/s12909-016-0583-7 72. Kolassa E.. **A Basic Course in Pharmaceutical Economics**. *J. Pharm. Teach.* (1994) **4** 15-27. DOI: 10.1300/J060v04n03_03
--- title: A Conceptual Model to Strengthen Integrated Management of HIV and NCDs among NIMART-Trained Nurses in Limpopo Province, South Africa authors: - Nthuseni Sharon Murudi-Manganye - Lufuno Makhado - Leepile Alfred Sehularo journal: Clinics and Practice year: 2023 pmcid: PMC10037621 doi: 10.3390/clinpract13020037 license: CC BY 4.0 --- # A Conceptual Model to Strengthen Integrated Management of HIV and NCDs among NIMART-Trained Nurses in Limpopo Province, South Africa ## Abstract Integrated management of human immune deficiency virus (HIV) and non-communicable diseases (NCDs) in primary health care facilities remains a challenge. Despite research that has been conducted in South Africa, it is evident that in Limpopo Province there are slits in the implementation thereof. There is a need to develop a conceptual model to guide in strengthening the clinical competence of nurse-initiated management of antiretroviral therapy (NIMART)-trained nurses to implement the integrated management of HIV and NCDs to improve clinical outcomes of patients with the dual burden of diseases in Limpopo Province, South Africa. This study aimed to develop a conceptual model to strengthen the implementation of integrated management of HIV and NCDs amongst NIMART nurses to improve clinical outcomes of patients with the dual burden of communicable and non-communicable diseases in Limpopo Province, South Africa. An explanatory, sequential, mixed-methods research design was followed. Data were collected from patient records and the skills audit of 25 Primary Health Care (PHC) facilities and from 28 NIMART trained nurses. Donabedian’s structure process outcome model and Miller’s pyramid of clinical competence provided a foundation in the development of the conceptual model. The study revealed a need to develop a conceptual model to strengthen the implementation of integrated HIV and NCDs implementation in PHC, as evidenced by differences in the management of HIV and NCDs. Conclusion: The study findings were conceptualised to describe and develop a model needed to strengthen the implementation of integrated management of HIV and NCDs amongst NIMART nurses working in PHC facilities. The study was limited to Limpopo Province; the model must be implemented in conjunction with the available frameworks to achieve better clinical outcomes. ## 1. Background Many low- and middle-income countries are still faced with a dual burden of HIV and NCDs despite the great achievement of control of HIV and NCDs over the last decade. Amongst others, the human immunodeficiency virus (HIV) and other nutritional diseases such as hypertension and diabetes continue to account for high mortality rates in low- and middle-income countries. The World Health Organisation (WHO) estimates that more than $75\%$ of people living with HIV on antiretroviral therapy (ART) are susceptible to NCDs due to stimulation of inflammatory markers, adverse event of some ART medicine, tobacco use, and alcohol use. Furthermore, $71\%$ of all deaths globally are due to NCDs [1,2,3]. The prevalence of HIV and NCDs varies from country to country, although we have seen a decline in HIV prevalence in the last decade and an increase in the number of patients accessing antiretroviral therapy due to the large funding that is directed towards HIV programmes as compared to the funding allocated towards NCD programmes [4,5,6]. South Africa has the largest number of PLWH with more than $80\%$ of the people on ART [2]. Despite the progress made in the management of HIV, there is a huge number of HIV patients who develop NCDs or were diagnosed with dual conditions at the start of ART or the start of NCDs treatment [7,8]. Furthermore, most developing African countries have many HIV patients who are affected with NCDs [9]. In South Africa, we have witnessed the successful implementation of the HIV programme through different policy strategies such as 90–90–90. In addition, the mortality rate was reduced by one third due to the implementation of NIMART in PHC facilities. The training of NIMART was led by the South African Department of Health supported by CDC and USAID [4,10,11]. Most professional nurses in South Africa were trained in NIMART following the support (financial and technical support) received from PEPFAR [4]. Despite the NIMART training, the quality of patient care related to NCDs was not a priority until the introduction of the Adult Primary Care (APC) guidelines, which included NCDs treatment guidelines [4,12,13]. The concept of integrated management of HIV and NCDs was realised during the implementation of the ideal clinic strategy, which was piloted in three South African provinces—namely Mpumalanga, Western Cape, and KwaZulu-Natal—in the year 2014. In addition, the ideal clinic strategy developed components and subcomponents that guided the rollout, including integrated management of HIV and NCDs [14,15]. Limpopo province is one of the rural provinces in South Africa, with Vhembe District implementing the integrated management of HIV and NCDs without the support of donor-funded organisations after PEPFAR shifted its support to the 27 districts of the 54 districts in South Africa [16,17]. The health care system in Limpopo Province continues to experience challenges that have serious effects on the implementation of integrated management of HIV and NCDs, including patient clinical outcomes. The literature reviewed established the weaknesses and threats which impact on poor implementation of integrated management of HIV and NCDs. In addition, the study verified the prospects to sustain the quality of the implementation thereof [18,19,20,21]. The implementation of integrated management of HIV and NCDs is the second component of the ideal clinic realisation and maintenance framework. Furthermore, it comprises subcomponents such as clinical supportive management and strengthening of support system, amongst others [14,15,16]. Even though the majority of the professional nurses are trained in NIMART, the APC training is still lagging behind, hence the poor implementation of APC guidelines. A literature review found no model that could be used to strengthen the clinical competence of NIMART nurses to implement integrated management of HIV and NCDs with confidence. Therefore, in this study, the researcher strove to develop and describe a conceptual model which may strengthen the implementation of integrated management of HIV and NCDs amongst NIMART-trained nurses to improve clinical outcomes of patients faced with the dual burden of diseases in Limpopo Province, South Africa. The model is designed to guide clinical competence and implementation of the programme thereof. ## 2. Method A mixed-methods approach using an explanatory, sequential, mixed-methods design was chosen to conduct a comprehensive literature review, to obtain a descriptive SWOT analysis, and to obtain an in-depth understanding of the views and experiences of NIMART-trained nurses regarding the implementation of integrated management of HIV and NCDs to generate valid conclusions. The study sought to assess the impact of and barriers influencing NIMART training and implementation to produce more complete and well-validated conclusions [22]. The study was conducted in four phases up until the development of the conceptual model. This included a comprehensive literature review, A SWOT analysis, and an exploratory qualitative study. The conceptual model was developed based on the findings of the first three phases. Donabedian’s [1966] SPO model [22] and Miller’s pyramid of Clinical competence [23] were used concurrently to categorise the characteristics and activities of NIMART nurses’ competencies and the implementation of the integrated management of HIV and NCDs within the PHC context. There is an interdependent relationship between Donabedian’s SPO model and Miller’s pyramid of clinical competence. We used the two frameworks as the starting point for the development of the conceptual model. Furthermore, a selection of the most appropriate information best describing the phenomenon and the activities necessary for the implementation of integrated management of HIV and NCDs was conducted [24,25]. Table 1 depicts the refinement of the methodology to eliminate overlapping activities. ## 3. Results NIMART training enables the professional nurses to initiate ART for all patients diagnosed with HIV, while APC training enables the professional nurses to diagnose and manage the patients with NCDs or with both HIV and NCDs. The two training components apply the same principles for learning and acquiring clinical competencies. The findings confirmed that all nurses were trained in NIMART, and only a few of them were trained in APC guidelines. Furthermore, as indicated, only a few were trained in a post-graduate diploma in Primary Health care, with a very small number trained in PALSA plus guidelines. PHC diploma and Palsa Plus entails training in NCDs management [25]. Furthermore, the qualitative results confirmed that professional nurses had confidence in managing HIV as compared to managing NCDs [25]. Such feelings of non-confidence in managing NCDs in a PHC facility contribute to under-diagnosis of NCDs in patients who are HIV-positive, resulting in serious complications including deaths. The study further verified that, as required by the guidelines, patients were not screened for diabetes and at least half of the patients had their blood pressure measured each time they visited a PHC facility, despite the availability of clinical guidelines in PHC facilities. The APC guidelines $\frac{2016}{2017}$ regulates the management of adults with communicable diseases (HIV) and non-communicable diseases (hypertension and diabetes), amongst others [25]. The latter was found to relate to the qualitative study results, where it clearly showed that there is an imbalance in HIV and NCDs training because NIMART nurses were not confident or rather competent in managing NCDs. In addition, some nurses felt that NCD management is a medical doctor’s responsibility. Unavailability of the updated APC guidelines in PHC facilities and shortage of medical equipment in some facilities may have also impacted the implementation of integrated management of HIV and NCDs. Both the quantitative and qualitative studies revealed that the PHC facilities do not have all the required clinical stationery for recording the patient care rendered to track the patients’ clinical outcomes over time. Furthermore, nurses become frustrated when there are no registers or patient files. Unclear role clarification compromises the clinical competence of the nurses as the nurses sometimes act as clerks as there is a shortage of administrative staff in PHC facilities. Besides the imbalance in the HIV and NCDs training that have been discussed, the study further verified the challenges which can impact the successful implementation of integrated management of HIV and NCDs. There is a vast difference in the implementation of integrated management of HIV and NCDs. Newly trained nurses come into practice without proper training in APC, hence, some patients are not managed in the spirit envisaged in the set guidelines. Usually, the patients are just given HIV treatment without being properly screened for NCDs, or the nurses feel that diagnosing NCDs is a medical doctor’s responsibility. The poor or little support from the NCDs programme managers also contributes to NIMART’s hindrance to providing comprehensive integrated management of HIV and NCDs. The study further identified the use of outdated APC guidelines, thus resulting in poor clinical patient outcomes. In addition, poor infrastructure and poor medical supplies were identified as challenges in this study. The study further confirmed that APC is not considered a requirement for a professional nurse to practice in a PHC setting, and this impacts the quality of patient care rendered [26]. Some opportunities can be strengthened to sustain the implementation of HIV and NCDs. The qualitative study revealed that patient satisfaction is a key to programme implementation, as the care provided becomes patient-centred and stigma is reduced. In addition, support from all stakeholders can change the current picture of integrated management of HIV and NCDs [25]. The study further verified that even if there are challenges, there are suggestions to improve the implementation of HIV and NCDs. This is evident from the qualitative study, which revealed that continuous professional development where professional nurses need to be trained and study further is necessary to improve their skills in providing integrated management of HIV and NCDs. In addition, comprehensive literature review and the quantitative study shows that professional nurses who are already trained in APC can be utilised to teach their peers in PHC facilities [27]. ## 3.1. A Conceptual Model to Strengthen the Implementation of Integrated Management of HIV and NCDs The conceptual model was developed based on the Millers pyramid of clinical competence [1990] and Donabedian SPO model [1966]. Miller’s pyramid enabled the researchers to identify the key elements attached to building clinical competence, whereas the Donabedian SPO model allowed the researchers to incorporate the results of all the study phases to inform the conceptual model. The following segment offers a description of the conceptual model. ## 3.2. Miller’s Pyramid Miller’s pyramid describes four levels that rank the clinical competence of the learners in the workplace, as illustrated in Figure 1 Furthermore, it describes the different levels the learners should go through to achieve and be assessed in the programme being delivered [22]. Miller further argued that at the end of any learning programme, the interest is in observing what learners can do to achieve higher levels of professional authenticity. In addition, the type of assessments should be valid to assist the learner to contribute to the improvement of patients’ clinical outcomes. In this study, the workplace is referred to as a district or the PHC facility—while the learners are referred to as NIMART nurses. ## 3.3.1. A Person to Have Knowledge on the Implementation of Integrated Management of HIV and NCDs (Knows) Knowledge According to Miller [1990] [23], a novice must grow into an expert. For the conceptual model, the NIMART nurse must be trained and assessed on the implementation of integrated management of HIV and NCDs to transit from a novice state to an expert state. A person must have undergone the assessment questions prescribed in the course curriculum. Furthermore, the knowledge that a person has may bring the knowledge attained elsewhere through hearing about it or knowing about it; for instance, a professional nurse who has gathered the facts or has knowledge in the implementation of a certain programme through informal learning. This concurs with Benner’s Novice to Expert Nursing Theory [24], which states that educating nurses is a foundation contributing to the development of specific nursing skills related to clinical guidelines. In this study, the professional nurses are the implementers of integrated management of HIV and NCDs at the PHC level. Therefore, they must know about relevant topics, the stepwise guidelines for adult primary care, and should be assessed on this acquired new knowledge. Missing this first level of competence invariably leads to the poor implementation of integrated management of HIV and NCDs. ## 3.3.2. Application of Knowledge by NIMART Nurses (Knows How/Understand/Competence) According to Miller [1990] [23], the professional nurse must be competent enough to apply what they have learnt through case presentation or answering the set questions to demonstrate how much they know of the integrated management of HIV and NCDs and how this is rendered to improve clinical patient outcomes. In other words, NIMART nurses should be able to utilise the knowledge acquired. The knowledge acquired is referred to here as NIMART and APC guidelines. Another study suggests that this level in Miller’s pyramid limits the nurses critical thinking, as nurses merely focus on the set guidelines rather than bringing their own thoughts to the patient care field [25]. However, this suggestion may be challenged, as nurses can only bring their thinking around patient management through thorough and evidence-based documented research. ## 3.3.3. Demonstration of Learning by NIMART Nurses (Shows How/Performance) Through the learnt skills and topics, NIMART nurses should be allowed to demonstrate these new skills through simulations and objective structured clinical examinations (OSCEs). This level calls upon NIMART nurses to demonstrate the appropriate skills needed for integrated management of HIV and NCDs. For instance, NIMART nurses should be able to develop and implement a treatment plan for patients diagnosed with HIV and NCDs. Furthermore, they should be able to offer health education appropriate to the diagnosis of the patient. The latter has been documented in some studies where it was confirmed that the value of demonstration in the preparation of nurses has a positive impact on the performance of nurses in the clinical environment [28,29]. ## 3.3.4. Actions by NIMART Nurses (Does/Action) According to Miller, this level of the pyramid requires the NIMART nurses to perform patient care through all-learnt patient care approaches. This includes the provision of routine patient care to track the clinical outcomes of the patients. Moreover, recording in the patient clinical stationery adds to the actions of the NIMART nurse as such recording provides a baseline for evaluating the care rendered to the patient. During the quantitative study, we found that patients were not screened for NCDs. Over and above the four levels in the pyramid, Miller affirmed that cognition and behaviour should be displayed by clinicians as an indicator of the nurse being clinically competent. Cognition refers to professional nurses who have never been trained in APC. Anyone who has never been exposed to APC training is certainly more likely to perform poorly compared to those who have received such training. The second term is behaviour, where the NIMART nurses must be tested to see if they can apply what they have learnt into practice. Miller further argued that knowing how does not mean that they will do it on a daily basis. This essentially means that the district is obliged to encourage and support the NIMART nurses who have undergone the APC training to implement integrated management of HIV and NCDs to continuously improve clinical outcomes of the patients. In addition to Miller’s thinking about behaviour and attitude, another researcher reiterates that medical professions, including nursing, depend on individual skills and expertise and rely more on good behaviour and appropriate attitude [28]. Miller further indicates that attitude is the key to clinical competence. In this case, NIMART nurses should demonstrate a willing attitude to transit from a novice state to an expert state in the provision of quality integrated management of HIV and NCDs at the PHC level. ## 3.4. Donabedian SPO Model According to Donabedian, three elements must be taken into consideration if the quality of patient care must be improved. Donabedian was convinced of three elements, as illustrated in Figure 2. Application of Donabedian SPO Model to the Findings. ## 3.4.1. Structure As described above, the structure in this study refers to the district, the PHC facilities, and the first level of Miller’s pyramid. Following Donabedian’s structure description, the PHC facilities’ hours of operation and type, NIMART/APC-trained nurses, clinical mentors, programme managers, availability of policies and guidelines, availability of clinical stationery, availability of APC guidelines, ART consolidated guidelines, and integrated reporting tools are essential for the implementation of integrated management of HIV and NCDs. In addition to the structure, training, communication, and internal and external programme support may strengthen the implementation of integrated management of HIV and NCDs. This study confirmed that the above-mentioned characteristics are essential in the implementation of integrated management of HIV and NCDs. In addition, NIMART nurses must acquire knowledge to enable them to offer quality care to patients with the dual burden of HIV and NCDs, as indicated in Figure 2. ## 3.4.2. Process According to the Donabedian SPO model, all aspects which are considered as the basis of qualified health care should be made available to make sure that patients’ needs are met. In this case, the process is achieved by ensuring that NIMART nurses are assessed on clinical competence. Furthermore, there is an assessment that NIMART nurses should undergo to enable them to diagnose and treat patients with HIV and NCDs. In this study, it was verified that training, programme support, clinical stationery, medical equipment, laboratory equipment, and reporting tools facilitate the process in terms of the implementation of integrated management of HIV and NCDs in PHC facilities. In addition, there should be a concerted use of external or internal motivators to encourage NIMART nurses to adhere to the newly learnt skill. ## 3.4.3. Outcomes Donabedian developed a model aimed to assess the quality of care in clinical practice [22]. In this study, the outcomes were achieved by evaluating the screening of diabetes and HPT, for instance if the blood pressure, blood glucose, and urine tests were conducted at first visit [25]. It is evident in this study that there is no adherence to APC guidelines, even though some patients were treated according to guidelines. Furthermore, adherence to APC guidelines may improve mortality rates amongst HIV-infected and NCDs-diagnosed patients. ## 3.5. Application of Miller’s Pyramid of Clinical Competence and Donabedian’s SPO Model to the Conceptual Model Although Donabedian’s SPO model was designed to evaluate the healthcare system and Miller’s pyramid was developed to assess the doctor’s clinical competence, these two frameworks fit perfectly in the effort to develop a conceptual model to strengthen programme implementation such as integrated management of HIV and NCDs. The two frameworks were merged to forge the conceptual model developed in this study. Figure 3 illustrates the conceptual model to strengthen the implementation of integrated management of HIV and NCDs among NIMART nurses in PHC. There is a link between clinical competence, structure, process, and outcomes. Training, assessment, and completion of a portfolio of evidence should not be overlooked or bypassed. Strengthening the implementation of integrated management of HIV and NCDs among NIMART nurses working at PHC facilities purely relies on enhancing clinical competence. According to the study results, the levels of clinical competence are strengthened through the structure and the process within a district. Nurses should be placed in facilities that are flexible for in-service training or continuous professional development. Resources such as policies, guidelines, clinical records, medical equipment, and laboratory equipment must always be available for NIMART nurses to achieve all the levels required for clinical competence. The process, on the other hand, allows NIMART nurses to clearly define their roles, which subsequently encourages them to develop an interest in enhancing their skills as they will feel a sense of being recognised. In addition, during the process, a supportive environment from the programme managers and external support motivates NIMART nurses to perform much better to improve patient clinical outcomes. The model recognises the NIMART nurse who is not yet trained on APC is placed in a facility where they are expected to implement integrated management of HIV. Often, the NIMART nurses implement the APC without being trained, yet the model emphasizes that novice or untrained nurses should be placed in PHC facilities with necessary skills, or they should be placed in facilities where training can be arranged while they are in service. Structural factors such as availability of medication and clinical records should be a basis for nurses to implement integrated management of HIV and NCDs. For enhanced clinical competence, there should be a process that includes support from various stakeholders, including external motivators. If the structure, process, and clinical competence are addressed, ultimately there will be outcomes such as improved patients’ outcomes. ## 4. Discussion The purpose of this study was to synthesise the study findings so that we could describe and develop a conceptual model that could guide to strengthen the implementation of integrated management of HIV and NCDs among NIMART-trained professional nurses in Limpopo Province, South Africa. The conceptual model was developed based on Miller’s [1990] pyramid of clinical competence and Donabedian’s SPO model [1966]. The study found that the majority of the professional nurses are trained in NIMART, and HIV treatment guidelines are followed by NIMART trained nurses; however, patients who are on ART are not screened for NCDs, as prescribed in the APC guidelines. In addition, the study found that most of the PHC facilities did not have the latest APC guidelines (APC $\frac{2016}{2017}$). Furthermore, there is an imbalance in the management of patients with HIV and those with NCDs, and these may be due to the lack of training of NIMART nurses on APC guidelines. The study findings established that essential diagnostic equipment such as a Blood Pressure (BP) machine and a glucometer were available, however, not all patients were screened for BP or glucose. According to Mboweni et al. [ 27], professional nurses are frustrated when they do not know what to do with patients, especially when they are not trained. Furthermore, another study confirmed that there is little training targeted for professional nurses who are working at the PHC level as compared to the nurses allocated in a hospital setting [28]. The use of developed guidelines to manage both HIV and NCDs enables quality care to patients with comorbidity, therefore, training and mentoring of nurses providing integrated management of HIV and NCDs in rural areas is essential [29,30]. Kane et al. [ 31] indicated that the availability of diagnostic tools and standardised protocols for disease management in PHC facilities are the key to improving patients’ clinical outcomes. Poor implementation of integrated management of HIV and NCDs exposes patients to develop complications related to ART side effects or even death. Furthermore, the longer the patient is on ART medication, the higher the risk of the patient developing NCDs [32,33,34]. The estimated number of NCDs-related deaths by 2025 would be prevented if NCDs management in PHC facilities is considered equally important as HIV management; otherwise, the NCDs may compromise the success of the HIV programme [35,36,37,38,39,40,41,42]. It is evident from this study that the clinical competence of NIMART nurses to manage NCDs is low because most of them did not receive the required APC training. There are also other operational factors which are hindering the implementation of integrated management of HIV and NCDs [43,44,45,46]. NIMART-trained nurses did not comply with the APC guidelines, which compromised integrated management of HIV and NCDs including the quality of patient care. The latter deemed it necessary to develop a conceptual model to strengthen the implementation of integrated management of HIV and NCDs in Limpopo Province. ## 5. Conclusions The implementation of integrated management of HIV and NCDs has proven to increase patient clinical outcomes since its adoption in the last decade. Therefore, training of NIMART nurses on the updated APC guidelines is essential. Enhancing clinical competence among NIMART nurses in Limpopo through training, support from programme managers, availability of equipment and medication, and supporting continued NIMART professional development can assist in the improvement of integrated management of HIV and NCDs at the PHC level in Limpopo Province. However, the implementation of integrated management of HIV and NCDs by NIMART-trained nurses is still a challenge. It is evident from the study findings that many factors such as clinical competence and health care systems influence how NIMART nurses implement the APC guidelines. The developed conceptual model, therefore, has the aptitude to strengthen the implementation of integrated management of HIV and HIV, thus improving patient clinical outcomes. ## 6. Limitations of the Study This study was limited to one district in Limpopo Province. Moreover, the focus was on PHC facilities. However, the findings are significant to other rural provinces in South Africa. ## 7. Practical Implications of the Study The developed conceptual model could assist in strengthening the clinical competence of NIMART-trained nurses, including the proper implementation of integrated management of HIV and NCDs in PHC facilities. Moreover, the conceptual model may be used as a reference during the development of clinical guidelines to ensure that clinical competence is not overlooked. ## References 1. 1. World Health Organization The Top 10 Causes of Death. Fact Sheet No. 310World Health OrganizationGeneva, Switzerland2013. *The Top 10 Causes of Death. Fact Sheet No. 310* (2013.0) 2. 2. World Health Organization Global Action Plan for the Prevention and Control of Noncommunicable Diseases 2013–2020World Health OrganizationGeneva, Switzerland2013. *Global Action Plan for the Prevention and Control of Noncommunicable Diseases 2013–2020* (2013.0) 3. 3. HIV/AIDS JUNPo Global Report: UNAIDS Report on the Global AIDS Epidemic 2010UNAIDSGeneva, Switzerland2010. *Global Report: UNAIDS Report on the Global AIDS Epidemic 2010* (2010.0) 4. 4. World Health Organization Integrated Health Services: What and Why. Technical BriefWorld Health OrganizationGeneva, Switzerland2008. *Integrated Health Services: What and Why. Technical Brief* (2008.0) 5. 5. Institute for Health Metrics and Evaluation Financing Global Health 2013: Transition in an Age of AusterityInstitute for Health Metrics and EvaluationSeattle, WA, USA2014. *Financing Global Health 2013: Transition in an Age of Austerity* (2014.0) 6. Njuguna B., Vorkoper S., Patel P., Reid M., Vedanthan R., Pfaff C., Park P.H., Fischer L., Laktabai J., Pastakia S.D.. **Models of integration of HIV and non-communicable disease care in sub-Saharan Africa: Lessons learned and evidence gaps**. *AIDS* (2018.0) **32** S33-S42. DOI: 10.1097/QAD.0000000000001887 7. Achwoka D., Waruru A., Chen T.H., Masamaro K., Ngugi E., Kimani M., Mukui I., Oyugi J.O., Mutave R., Achia T.. **Noncommunicable disease burden among HIV patients in care: A national retrospective longitudinal analysis of HIV-treatment outcomes in Kenya, 2003–2013**. *BMC Public Health* (2019.0) **19**. DOI: 10.1186/s12889-019-6716-2 8. Duffy M., Ojikutu B., Andrian S., Sohng E., Minior T., Hirschhorn L.R.. **Non-communicable diseases and HIV care and treatment: Models of integrated service delivery**. *Trop. Med. Int. Health* (2017.0) **22** 926-937. DOI: 10.1111/tmi.12901 9. Haregu T.N., Oldenburg B., Sestwe G., Elliott J., Nanayakkara V.. **Epidemiology of comorbidity of HIV/AIDS and non-communicable diseases in developing countries: A systematic review**. *J. Glob. Health Care Syst.* (2012.0) **2** 1-12 10. Visser C.A., Wolvaardt J.E., Cameron D., Marincowitz G.J.. **Clinical mentoring to improve quality of care provided at three NIM-ART facilities: A mixed methods study**. *Afr. J. Prim. Health Care Fam. Med.* (2018.0) **10** 1-7. DOI: 10.4102/phcfm.v10i1.1579 11. Mngqibisa R., Muzigaba M., Ncama B.P., Pillay S., Nadesan-Reddy N.. **Upskilling nursing students and nurse practitioners to initiate and manage patients on ART: An outcome evaluation of the UKZN NIMART course**. *Afr. J. Health Prof. Educ.* (2017.0) **9** 153-158. DOI: 10.7196/AJHPE.2017.v9i3.879 12. Fauci A.S., Eisinger R.W.. **PEPFAR—15 years and counting the lives saved**. *N. Engl. J. Med.* (2018.0) **378** 314-316. DOI: 10.1056/NEJMp1714773 13. Kim Y.. **The effectiveness of PEPFAR’s funding for women and children with HIV/AIDS**. *Int. J. Health Plan. Manag.* (2019.0) **34** e896-e916. DOI: 10.1002/hpm.2706 14. Hunter J.R., Chandran T.M., Asmall S., Tucker J.M., Ravhengani N.M., Mokgalagadi Y.. **The Ideal Clinic in South Africa: Progress and challenges in implementation**. *S. Afr. Health Rev.* (2017.0) **2017** 111-123 15. Lebina L., Alaba O., Ringane A., Hlongwane K., Pule P., Oni T., Kawonga M.. **Process evaluation of implementation fidelity of the integrated chronic disease management model in two districts, South Africa**. *BMC Health Serv. Res.* (2019.0) **19**. DOI: 10.1186/s12913-019-4785-7 16. Katz I.T., Bassett I.V., Wright A.A.. **PEPFAR in transition—implications for HIV care in South Africa**. *N. Engl. J. Med.* (2013.0) **369** 1385-1387. DOI: 10.1056/NEJMp1310982 17. **South Africa. National Department of Health. Integrated Chronic Disease Manual. Pretoria**. (2014.0) 18. Mahomed Ozayr H., Asmall S., Taylor M., Naidoo S.. **Improving the Quality of Nurse Clinical Documentation for Chronic Patients at Primary Care Clinics: A Multifaceted Intervention: Original Research. Curationis [Serial on the Internet]. 2015. SA ePublications Service** 19. Fairall L.R., Folb N., Timmerman V., Lombard C., Steyn K., Bachmann M.O., Bateman E.D., Lund C., Cornick R., Faris G.. **Educational Outreach with an Integrated Clinical Tool for Nurse-Led Non-communicable Chronic Disease Management in Primary Care in South Africa: A Pragmatic Cluster Randomised Controlled Trial**. *PLoS Med.* (2016.0) **13**. DOI: 10.1371/journal.pmed.1002178 20. Davies N., Homfray M., Venables E.. **Nurse and manager perceptions of nurse initiated and managed antiretroviral therapy (NIMART) implementation in South Africa: A qualitative study**. *BMJ Open* (2013.0) **3** e003840. DOI: 10.1136/bmjopen-2013-003840 21. Naidoo S., Mahomed O., Asmall S., Taylor M.. **Nurses’ knowledge of chronic disease management**. *Health SA Gesondheid* (2014.0) **19** 1-8. DOI: 10.4102/hsag.v19i1.809 22. Miller G.E.. **The assessment of clinical skills/competence/performance**. *Acad. Med.* (1990.0) S63-7. DOI: 10.1097/00001888-199009000-00045 23. Donabedian A.. **Evaluating the quality of medical care**. *Milbank Meml. Fund Q.* (1966.0) **44** 166-206. DOI: 10.2307/3348969 24. Murudi-Manganye N., Makhado L., Sehularo L.A.. **Integrated Management of HIV and NCDs in Vhembe District, Limpopo Province: A SWOT Analysis for Primary Health Care Facilities**. *HIV Nurs.* (2021.0) **21** 20-29 25. Murudi-Manganye N.S., Makhado L., Sehularo L.A.. **Exploration of NIMART nurses’ views of the implementation of integrated management of HIV and NCDs in Limpopo Province, South Africa. HIV & AIDS Review**. *Int. J. HIV Relat. Probl.* (2023.0) 26. Murudi-Manganye N.S., Makhado L., Sehularo L.A.. **Integrated Management of HIV and NCDs within the Primary Health Care in the South African context: A comprehensive literature review**. *Open Public Health J.* (2020.0) **13** 447-454. DOI: 10.2174/1874944502013010447 27. Benner P.. **From novice to expert**. *Am. J. Nurs.* (1982.0) **82** 402-407. PMID: 6917683 28. Sullivan N., Swoboda S.M., Breymier T., Lucas L., Sarasnick J., Rutherford-Hemming T., Budhathoki C., Kardong-Edgren S.S.. **Emerging evidence toward a 2: 1 clinical to simulation ratio: A study comparing the traditional clinical and simulation settings**. *Clin. Simul. Nurs.* (2019.0) **30** 34-41. DOI: 10.1016/j.ecns.2019.03.003 29. Goh H.S., Zhang H., Lee C.N., Wu X.V., Wang W.. **Value of nursing objective structured clinical examinations: A scoping review**. *Nurse Educ.* (2019.0) **44** E1-E6. DOI: 10.1097/NNE.0000000000000620 30. Kisembo G.. **Objective Structured Clinical Examination (OSCE) and Academic Performance of Nursing Students in Uganda**. *Ph.D. Thesis* (2016.0) 31. Mitra J., Saha I.. **Attitude and communication module in medical curriculum: Rationality and challenges**. *Indian J. Public Health* (2016.0) **60** 95. PMID: 27350701 32. Creswell J.W.. *Research Design. A Quantitative, Qualitative and Mixed Method Approaches* (2009.0) 33. Mouton J., Marais H.C.. **Basic Concepts in the Methodology of Social Sciences, Human Science Research Council, Pretoria, South Africa**. (1996.0) 34. Friese S.. *Qualitative Data Analysis with ATLAS.ti.* (2019.0) 35. Mboweni S.H., Makhado L.. **Conceptual framework for strengthening nurse-initiated management of antiretroviral therapy training and implementation in North West province**. *Health SA Gesondheid* (2020.0) **25** 1285. DOI: 10.4102/hsag.v25i0.1174 36. Pfaff C., Scott V., Hoffman R., Mwagomba B.. **You can treat my HIV-But can you treat my blood pressure? Availability of integrated HIV and non-communicable disease care in northern Malawi**. *Afr. J. Prim. Health Care Fam. Med.* (2017.0) **9** 1-8. DOI: 10.4102/phcfm.v9i1.1151 37. Venables E., Edwards J.K., Baert S., Etienne W., Khabala K., Bygrave H.. **“They just come, pick and go.” The acceptability of integrated medication adherence clubs for HIV and non-communicable disease (NCD) patients in Kibera, Kenya**. *PLoS ONE* (2016.0) **11**. DOI: 10.1371/journal.pone.0164634 38. Ndayisaba A., Harerimana E., Borg R., Miller A.C., Kirk C., Hann K., Hirschhorn L.R., Manzi A., Ngoga G., Dusabeyezu S.. **A clinical mentorship and quality improvement program to support health centre nurses manage type 2 diabetes in rural Rwanda**. *J. Diabetes Res.* (2017.0) **2017** 2657820. DOI: 10.1155/2017/2657820 39. Kane J., Landes M., Carroll C., Nolen A., Sodhi S.. **A systematic review of primary care models for non-communicable disease interventions in sub-Saharan Africa**. *BMC Fam. Pract.* (2017.0) **18**. DOI: 10.1186/s12875-017-0613-5 40. Casper C., Crane H., Menon M., Money D.. *HIV/AIDS Comorbidities: Impact on Cancer, Non-Communicable Diseases, and Reproductive Health* (2017.0) 41. Feinstein M.J., Bogorodskaya M., Bloomfield G.S., Vedanthan R., Siedner M.J., Kwan G.F., Longenecker C.T.. **Cardiovascular complications of HIV in endemic countries**. *Curr. Cardiol. Rep.* (2016.0) **18** 113. DOI: 10.1007/s11886-016-0794-x 42. Mathebula R.L., Maimela E., Ntuli N.S.. **The prevalence of selected non-communicable disease risk factors among HIV patients on anti-retroviral therapy in Bushbuckridge sub-district, Mpumalanga province**. *BMC Public Health* (2020.0) **20**. DOI: 10.1186/s12889-019-8134-x 43. Patel P., Rose C.E., Collins P.Y., Nuche-Berenguer B., Sahasrabuddhe V.V., Peprah E., Vorkoper S., Pastakia S.D., Rausch D., Levitt N.S.. **HIV. Noncommunicable diseases among HIV-infected persons in low-income and middle-income countries: A systematic review and meta-analysis**. *AIDS* (2018.0) **32** S5. DOI: 10.1097/QAD.0000000000001888 44. Frieden M., Zamba B., Mukumbi N., Mafaune P.T., Makumbe B., Irungu E., Moneti V., Isaakidis P., Garone D., Prasai M.. **Setting up a nurse-led model of care for management of hypertension and diabetes mellitus in a high HIV prevalence context in rural Zimbabwe: A descriptive study**. *BMC Health Serv. Res.* (2020.0) **20**. DOI: 10.1186/s12913-020-05351-x 45. Mboweni S.H., Makhado L.. **Strategies to improve the implementation of nurse-initiated management of antiretroviral therapy (NIMART) training: A systematic review**. *Res. Sq.* (2020.0). DOI: 10.21203/rs.3.rs-16220/v1 46. Ramezanzade Tabriz E., Orooji A., Bikverdi M.. **Investigation clinical competence and its relationship with professional Ethics and spiritual health in nurses**. *Health Spiritual. Med. Ethics* (2017.0) **4** 2-9
--- title: 'Immunoregulatory Biomarkers of the Remission Phase in Type 1 Diabetes: miR-30d-5p Modulates PD-1 Expression and Regulatory T Cell Expansion' authors: - Laia Gomez-Muñoz - David Perna-Barrull - Marta Murillo - Maria Pilar Armengol - Marta Alcalde - Marti Catala - Silvia Rodriguez-Fernandez - Sergi Sunye - Aina Valls - Jacobo Perez - Raquel Corripio - Marta Vives-Pi journal: Non-Coding RNA year: 2023 pmcid: PMC10037622 doi: 10.3390/ncrna9020017 license: CC BY 4.0 --- # Immunoregulatory Biomarkers of the Remission Phase in Type 1 Diabetes: miR-30d-5p Modulates PD-1 Expression and Regulatory T Cell Expansion ## Abstract The partial remission (PR) phase of type 1 diabetes (T1D) is an underexplored period characterized by endogenous insulin production and downmodulated autoimmunity. To comprehend the mechanisms behind this transitory phase and develop precision medicine strategies, biomarker discovery and patient stratification are unmet needs. MicroRNAs (miRNAs) are small RNA molecules that negatively regulate gene expression and modulate several biological processes, functioning as biomarkers for many diseases. Here, we identify and validate a unique miRNA signature during PR in pediatric patients with T1D by employing small RNA sequencing and RT-qPCR. These miRNAs were mainly related to the immune system, metabolism, stress, and apoptosis pathways. The implication in autoimmunity of the most dysregulated miRNA, miR-30d-5p, was evaluated in vivo in the non-obese diabetic mouse. MiR-30d-5p inhibition resulted in increased regulatory T cell percentages in the pancreatic lymph nodes together with a higher expression of CD200. In the spleen, a decrease in PD-1+ T lymphocytes and reduced PDCD1 expression were observed. Moreover, miR-30d-5p inhibition led to an increased islet leukocytic infiltrate and changes in both effector and memory T lymphocytes. In conclusion, the miRNA signature found during PR shows new putative biomarkers and highlights the immunomodulatory role of miR-30d-5p, elucidating the processes driving this phase. ## 1. Introduction Type 1 diabetes (T1D) results from the immune-mediated destruction of the insulin-producing β-cells of the pancreas, which inevitably leads to the appearance of hyperglycemia. This metabolic and autoimmune disease is one of the most prevalent chronic pediatric diseases [1], and patients rely on exogenous insulin therapy for life. The β-cell attack is silent and gradual, reflected by the appearance of islet autoantibodies long before the disease is diagnosed, at which point 60–$80\%$ of these cells have already been destroyed. However, the decline in the β-cell mass is not necessarily linear; instead, patients can course with a relapsing/remitting pattern in which they retain some ability to regenerate β-cells and produce enough endogenous insulin to reduce the requirement of the exogenous one [2,3]. Clinically, this is detected only during the partial remission (PR) phase, called the “honeymoon phase”, a stage experienced by up to $80\%$ of pediatric patients with T1D after the initiation of insulin therapy and that is characterized by low requirements of exogenous insulin and diminished glycated hemoglobin (HbA1c) levels [4]. Although the mechanisms behind this intriguing transient period have been poorly explored, it has been associated with β-cell rest, recovery, and regeneration as well as with immunological changes that could reflect an attempt to restore self-tolerance [5,6]. In fact, peripheral levels of different immune system cells and cytokines may influence the appearance and length of the PR phase [7,8,9,10]. Although metabolic biomarkers and autoantibodies have been useful to monitor and correctly diagnose patients, they do not reflect the variations of the autoimmune response or β-cell regeneration and do not predict the development of the disease. Therefore, there is an unmet need for reliable and cost-effective biomarkers for this heterogeneous disease. Many studies are currently underway to find new immunological biomarkers that can both predict disease progression and response to therapies [11] and stratify patients into different endotypes [12,13]. Furthermore, new biomarkers of remission are arising to understand the immunological mechanisms of this phase, to monitor the early course of T1D, and to stratify patients with better disease prognosis [14,15,16]. MicroRNAs (miRNAs) are a family of endogenous ∼18–25-nucleotide non-coding small RNAs that negatively regulate gene expression at the post-transcriptional level and in a sequence-specific manner. In that sense, miRNAs act on gene expression through different mechanisms, including mRNA degradation, destabilization by deadenylation, and/or translational repression [17], thus modulating many biological processes, such as immune regulation or cellular differentiation, proliferation, metabolism, and apoptosis. Since miRNAs are very stable molecules that can be found both in many mammalian cell types and in cell-free body fluids, they were recently proposed as potentially available blood biomarkers for several disorders, including autoimmunity [18,19,20]. More specifically, in T1D, some differentially expressed miRNAs (DEMs) have been linked to an aberrant T lymphocyte activation, differentiation, and function as well as β-cell apoptosis [21,22,23,24], which suggests a direct role of these molecules in regulating the onset of islet autoimmunity. Indeed, islet autoimmunity induces the expression of miR-142-3p on CD4+ T cells, which impairs regulatory T cell (Treg) homeostasis and functions by targeting TET2, an enzyme that maintains regions within the Foxp3 locus in a hypomethylated state, thus ensuring proper Foxp3 expression in Tregs [25]. Moreover, newly diagnosed T1D patients present a different peripheral miRNA signature when compared to control subjects or patients with long-standing disease [26,27,28,29,30,31,32], indicating that these miRNAs may reflect different stages of the disease or serve both as diagnostic and prognostic biomarkers. Because miRNAs regulate the expression of genes involved in β-cell function, inflammation, and autoimmunity [33], we hypothesize that a specific plasma miRNA profile can be identified in T1D patients during the PR phase that may reflect attempts at immunoregulation or β-cell functional improvement. Here, we identified for the first time a unique peripheral miRNA signature in pediatric patients with T1D during the PR phase and revealed novel immunomodulatory roles for the upregulated miR-30d-5p in the context of autoimmunity in the experimental model of the disease: non-obese diabetic (NOD) mice. ## 2.1. Distinct miRNA Signature during the Partial Remission Phase of Pediatric Patients with Type 1 Diabetes To identify miRNA expression signatures during PR, small RNA was isolated and sequenced from the plasma of the discovery cohort (Table 1; clinical results from discovery cohort were previously and partially published in [14]) composed of 17 newly diagnosed children with T1D (age: 8.7 ± 3.6 years, mean ± SD), of whom 11 were remitters (age: 9.1 ± 4.3 years), and 6 were non-remitters (at eight months after diagnosis; age: 9 ± 2.8). The control group was chosen from 17 age- and sex-matched non-diabetic children (age: 8.8 ± 3.4) who visited the Germans Trias i Pujol University Hospital. However, due to the limited amount of small RNAs of four samples at T1D diagnosis, only 13 samples out of the 17 at this time-point could be profiled on the Illumina NextSeq 1000 System. First, we focused on the comparison between remitters and non-remitters to find a specific miRNA signature that defines the PR. The hierarchical clustering heatmap revealed 16 DEMs (|log(fold change (FC))| > 1 and p-value ≤ 0.05); 12 of them were upregulated, and 4 were downregulated during the remission phase (Figure 1A). To identify the miRNAs with the greatest FC and the lowest p-value, a volcano plot was constructed, where blue dots represent downregulated miRNAs, and red dots represent the upregulated ones. Within this last group, we observed that the miR-30d-5p had the greatest FC with the highest statistical significance, followed by miR-106a-5p and miR-20b-5p (Figure 1B). Additionally, it is worth mentioning that five miRNAs (miR-106a-5p, miR-106b-5p, miR-18b-5p, miR-17-5p, and miR-20b-5p) belong to the same family of miRNAs, the MIR-17 family. Next, we performed a principal component analysis based on each sample’s miRNA expression, which could effectively differentiate the two clusters of patients; the orange are the remitters (PR), and the violet are the non-remitters (non-PR) (Figure 1C). Nevertheless, two remitter patients were clustered within the non-remitter group, which coincides with the two of the heatmap that showed a distinct profile in comparison to the other PR patients (marked with an asterisk); one of them also presented celiac disease. The biplot showing the principal component scores and the miRNAs allowed us to identify which miRNA influences one component or another. In this case, miR-25-5p, let-7b-5p, let-7c-5p, and miR-10393-3p had the most influence in determining the non-PR group (Figure S1). To examine which of these DEMs during PR were present in other comparisons, we made Venn diagrams. We found that the miR-30d-5p was also differentially expressed between the PR phase and both the control group (FC = 1.777, p-value = 0.019; Table 2) and T1D diagnosis (FC = 1.780, p-value = 0.053), which gives it more potency as a specific biomarker of remission. Other miRNAs were more related to the general disease status, such as miR-17-5p, which was further differentially expressed between the control group and the diagnosis of T1D in those future remitter patients (Figure 1D). Table 2 shows the DEMs (|log(FC)| > 1 and p-value ≤ 0.05) with validated target genes between PR vs. non-PR (#14) and PR vs. their diagnosis time-point (#20) as well as the ones that were also differentially expressed between PR vs. controls (#12). The complete list of DEMs between PR vs. controls (#62) and T1D diagnosis vs. controls (#16) can be found in Table S1. ## 2.2. Validation of Differentially Expressed miRNAs during the Partial Remission Phase Then, we selected some of the miRNAs (miR-142-3p, miR-17-5p, miR-106b-5p, miR-20b-5p, let-7b-5p, and let-7c-5p) with the highest number of normalized counts per million and that were present in at least $70\%$ of the forty-seven profiled samples from the discovery cohort in order to validate their expression using single-assay RT-qPCR in another cohort of patients, the validation cohort (Table 1). This cohort was composed of 15 age- and sex-matched non-diabetic controls (age: 9.7 ± 3.6), 8 newly diagnosed children with T1D (age: 11.6 ± 2.7), 10 remitter patients (age: 11.8 ± 2.9), and 9 non-remitter patients (age: 7.3 ± 3.7). Five out of six miRNAs were validated as differentially expressed; miR-142-3p was slightly upregulated, and both miR-17-5p and miR-106b-5p were significantly upregulated during PR in comparison to the non-PR group (Figure 2A–C). Both let-7b-5p and let-7c-5p were also slightly downregulated at this stage (Figure 2E,F). No differences were found regarding the levels of miR-20b-5p between the PR and non-PR groups (Figure 2D). ## 2.3. Potential Enriched Biological Functions and Pathways under the Regulation of Differentially Expressed miRNAs during the Partial Remission Phase To identify potential biological functions and pathways affected by the miRNA signature during PR, we performed in silico miRNA functional analyses based on the inferred miRNA target genes using the DIANA-miRPath v3.0 web server. In this case, we only used miRNAs with experimentally validated target genes, thus ruling miR-10393-3p and miR-10395-3p out of the 16 DEMs from the analysis. Biological processes and pathways were investigated using the DIANA-TarBase v7.0 and a Fisher’s exact test, and the false discovery rate (FDR) method (q-values) was used to calculate the enriched targeted ones. The interactive graph of similar non-redundant gene ontology (GO) terms was retrieved from the web server REVIGO, and it clearly showed two different clusters of enriched biological processes: one that comprises different metabolic processes (e.g., mRNA metabolic process, cellular lipid metabolic process, or cellular protein metabolic process) (upper cluster) and one comprising different signaling pathways related to the immune response (e.g., toll-like receptor TLR6:TLR2 signaling pathway or Fc-gamma receptor signaling pathway involved in phagocytosis), apoptosis (e.g., intrinsic apoptotic signaling pathway), or stress (e.g., stress-activated MAPK cascade) (lower cluster). These are relevant biological functions under the pathogenesis of T1D (Figure 3). The pathway analysis was performed using Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations. Clustering the miRNAs based on their influence on molecular pathways (Figure 4, top) indicated that miR-17-5p, miR-106b-5p, miR-20b-5p, and miR-106a-5p—all from the MIR-17 family—appeared to target similar pathways and were the ones with a broad effect among all the enriched miRNAs. Of those, other interesting pathways were Wnt, mTOR, FoxO, and MAPK signaling pathways. The 10 most enriched KEGG pathways are listed in the table (Figure 4, bottom), with fatty acid biosynthesis being the most statistically significant. Interestingly, the TGF-β signaling pathway was among the top 10, which is known for its crucial role in immune homeostasis and tolerance. ## 2.4. In Vivo Inhibition of miR-30d-5p Potentiates Treg Expansion by Increasing CD200 Levels MiR-30d-5p was the most differentially expressed miRNA during the PR phase in comparison to the non-remission, and its role in the immune system or T1D has been poorly explored. This miRNA was enriched for biological processes such as the immune system, response to stress, cell death, or insulin receptor signaling. In addition, miR-30d-5p was among the miRNAs that presented the highest number of immune-related target genes, including some that have a direct role in the activation of T lymphocytes, such as CTLA-4 or CD226 (Table S2). Thus, to address the immunological relevance of miR-30d-5p in vivo, we next analyzed the effect of miR-30d-5p inhibition in nine-week-old prediabetic NOD mice. We employed a well-known Locked Nucleic Acid (LNA)-miRNA-inhibitor that, when administered systematically, has been shown to accumulate in different tissues, including lymphoid tissues and the pancreas, thus promoting in vivo miRNA silencing [25]. The miR-30d-5p inhibitor or control inhibitor (negative control A probe; miRNA control group) were applied four times by intraperitoneal (i.p.) injection at 9 mg/kg every three days, whereas in the sham group, mice were treated with 200 µL PBS following the same pattern. Along with the treatment with the miRNA inhibitor, mice exhibited normal glycosuria and glycemia at the end of the study and maintained weight within a normal range, which led us to think that the inhibition of this miRNA did not have a toxic effect on mice, nor did it significantly advance the onset of diabetes. Moreover, the viability of splenocytes and pancreatic lymph node (PLN) cells was optimal in the three groups, despite small significant differences (Figure S2). Figure 5A shows representative fluorescence-activated cell sorting (FACS) plots indicating the percentage of CD4+CD25+FoxP3+ Tregs in PLN of the sham (left, grey), control inhibitor (middle, blue), and miRNA inhibitor (right, red) treatment groups. At the end of the study, the blockade of miR-30d-5p resulted in an increased percentage of Treg cells in the PLN in comparison to both the sham group and the miRNA control group (Figure 5B). We then classified the different predicted and validated target genes for miR-30d-5p (approx. 2.000 targets) into three groups, [1] regeneration, [2] metabolism, and [3] immune system (Table S2). Regarding the immune system group, we found interesting target genes such as CD200, which is associated with Treg expansion [34]. To demonstrate the efficacy of the delivered miR-30d-5p inhibitor in the pancreatic tissue and local immune cells, we analyzed the expression of CD200 in the remaining immune cells from PLN. We observed higher levels of CD200 mRNA in the absence of miR-30d-5p in comparison to the miRNA control group, validating the in vivo effect of miR-30d-5p inhibition (Figure 5C). Interestingly, we found a significant positive correlation between CD200 expression levels and the percentage of Treg cells, further confirming the role of this CD200 inhibitory immune checkpoint in immune tolerance and regulation (Figure 5D). ## 2.5. In Vivo Inhibition of miR-30d-5p Reduces PD-1 Expression on Splenocytes Figure 6A,C shows representative FACS plots indicating the percentage of CD4+PD-1+ T cells and PD-1+ Tregs, respectively, in the spleen of the sham (left, grey), control inhibitor (middle, blue), and miRNA inhibitor (right, red) treatment groups. At the end of the study, the blockade of miR-30d-5p resulted in a significant reduction in the percentage and also total cell numbers of CD4+ T cells positive for PD-1 in comparison to both the sham group and the miRNA control group (Figure 6B). Furthermore, this PD-1 expression decline was very pronounced in Treg cells, both in percentage and total cell number (Figure 6D). PDCD1 (PD-1 gene) is not a direct target gene of miR-30d-5p, but we found PRDM1, a repressor of PD-1, which is [35] (Table S2). Thus, in order to link the low expression of PD-1 with the upregulation of one of its repressors in the absence of miR-30d-5p, we analyzed the expression levels of both PRDM1 and PDCD1 on the splenocytes. Although we could not find differences regarding the expression of PRDM1 between miRNA inhibitor and control groups (Figure 6E), we did find decreased PDCD1 mRNA levels, confirming the results obtained by flow cytometry (Figure 6F). We further checked the expression levels of TGFBR2 and PRDM1 in the remaining PLN and of TGFBR2 and CD200 in the splenocytes, but no differences were found between the miRNA control and miRNA inhibitor groups (Figure S3A,B). ## 2.6. In Vivo Inhibition of miR-30d-5p Displays Changes in Additional T Cell Differentiation Subsets To examine the effect of the miR-30d-5p blockade on the maturation of the immune system and different immune cell populations, we analyzed CD4+ and CD8+ T lymphocytes (differentiating between naïve, central memory (CM), effector memory (EM), and pre-effector-like T cells), B lymphocytes, and conventional dendritic cells (DCs). First, no differences in the percentages of total CD4+ and CD8+ T cells, DCs, or B lymphocytes were found between the three treatment groups in either PLN (Figure S4A–D) or the spleen (Figure S4F–I). Representative FACS plots for the analysis of B lymphocytes and DCs in PLN and the spleen are respectively shown in Figure S4E,J. Figure 7A,C show representative FACS plots indicating the percentage of naïve, CM, EM, and pre-effector-like CD4+ and CD8+ T cells, respectively, in the PLN of miRNA control and miRNA inhibitor groups. Although not significant, we found that the miR-30d-5p blockade led to an increase in the percentage of CD4+ EM T lymphocytes (Figure 7B) together with a decrease in the percentage of CD8+ pre-effector-like T lymphocytes (Figure 7D) and an increase in the CD8+ CM ones (Figure 7E). Regarding the spleen, representative FACS plots indicating the percentage of naïve, CM, EM, and pre-effector-like CD4+ T lymphocytes can be found in Figure 8A. In this organ, the miR-30d-5p blockade tended to decrease the percentages and total cell numbers of CD4+ pre-effector-like T cells (Figure 8B) and significantly reduced the percentages of the CD4+ CM T cells, which could also be observed regarding their numbers (Figure 8C). No differences were found within splenic CD8+ T cell subsets. ## 2.7. In Vivo Inhibition of miR-30d-5p Tends to Increase Leukocyte Islet Infiltration The insulitis score was determined at the end of the short treatment. As expected, mice in the sham group showed a similar insulitis degree (0.96 ± 0.39, mean ± SD) as the miRNA control group (1.182 ± 0.25) (Figure 9A). Mice treated with the miR-30d-5p inhibitor showed a biological trend toward an increased insulitis score (1.33 ± 0.51) in comparison to both control groups (Figure 9A); $37.3\%$ of their islets were infiltrated or destroyed (scores from two to four), whereas only $27.3\%$ and $33.5\%$ of the islets were destroyed in the sham and the miRNA control groups, respectively (Figure 9B). ## 3. Discussion In recent years, much progress has been made in the study of the different factors that affect the progression of T1D and its immunopathology, revealing that this disease is more heterogeneous than initially thought [36]. In fact, there is great inter-subject variability in terms of age at diagnosis, autoantibodies, genetics, metabolic control, rate of progression, and immune activity, which has led to the hypothesis that there are different T1D endotypes [37]. This makes the search for new biomarkers essential to understand the different courses of the disease and stratify patients. In this regard, although up to almost $80\%$ of patients experience the PR phase after the initiation of insulin therapy, it is still an under-explored period, although it is of great interest both metabolically and immunologically. Circulating microRNAs have huge potential as a novel class of non-invasive biomarkers that reflect disease activity and attempts at β-cell regeneration or immune regulation [22,38]. Here, we identified for the first time a circulating miRNA signature for PR in a longitudinal pediatric cohort of patients and described an immunoregulatory role for miR-30d-5p in the NOD mouse model by modulating miRNA expression. The onset of islet autoimmunity has been associated with several dysregulated miRNAs both in circulation [28,32,39,40,41,42] and in peripheral blood mononuclear cells or T lymphocytes [25,31,43,44]; however, only a few studies have addressed changes in miRNA levels over time upon diagnosis [26,27]. Specifically in children, one of the first studies found that miR-25 levels in sera are associated with improved glycemic control and stimulated C-peptide three months after diagnosis [29]. Similarly, levels of the miR-23~24~27 cluster in newly diagnosed children predict C-peptide loss over time and are upmodulated upon disease progression [30]. A study conducted on the Danish Remission Phase Cohort found that the miR-197-3p at three months—when PR usually occurs—was the strongest predictor of residual β-cell function one year after diagnosis in children with T1D [45]. Nevertheless, none of them evaluated specific changes in miRNA levels during the PR phase, so that was our first aim. On the one hand, we identified 12 upregulated miRNAs in pediatric remitter patients versus non-remitters. Of those, five belonged to the MIR-17 family (miR-20b-5p, miR-17-5p, miR-106a-5p, miR-106b, and miR-18b-5p). Some of these miRNAs have been previously associated with β-cell apoptosis/regeneration processes and immunomodulation. For instance, the inflammatory microenvironment led β-cells to downregulate the expression of miR-17, a miRNA that negatively affects ERAP1 mRNA, impairing the processing of preproinsulin signal peptide antigen and limiting its recognition by autoreactive CD8+ T cells [46]. Thus, the upregulation of miR-17 during the PR phase could be related to the decreased inflammatory environment and might be involved in decreasing β-cell visibility to the immune system. Moreover, whereas miR-17-92 and miR-106b-25 clusters positively regulate β-cell proliferation and insulin secretion in mice and are important for normal endocrine function [47,48], miR-20b can inhibit T cell proliferation and activation by targeting NFAT [49]. Hence, both facts would contribute to the hypothesis that behind PR, there are processes of immunoregulation and β-cell regeneration that are controlled by epigenetics. On the other hand, we identified four downregulated miRNAs during PR in comparison to non-PR, including let-7b-5p and let-7c-5p. Recently, it was reported that let-7b-5p overexpression impairs insulin production and secretion and inhibits β-cell proliferation in mice [50,51], which could be related to the diminished residual β-function in patients without remission. Notably, of the 16 DEMs, miR-10395-3p, miR-10393-3p, and miR-1277-3p are described for the first time in relation to T1D. Furthermore, the GO analysis revealed several enriched metabolic, immunological, apoptosis, and stress processes, which are pathways closely related to T1D immunopathogenesis. The functional annotation of genes regulated by these miRNAs also implies that TGF-β and FoxO signaling pathways (among others) may be involved in the occurrence of PR, which can control the development and function of Foxp3+ Tregs [52]. MiR-30d-5p was the most upregulated miRNA during PR, where insulin synthesis and secretion are improved. This is a glucose-regulated miRNA that has been associated with both the induction of insulin production by activating MafA in pancreatic β-cells and the protection of β-cell function from impairment caused by proinflammatory cytokines [53,54]. Even though we are uncertain about which cells are producing this miRNA and thus contributing to its circulating pool, it is known to be highly expressed by pancreatic β-cells, suggesting that they are likely one of its primary sources [54]. In cancer cells, miR-30d-5p induces IL-10 expression (an immunosuppressive cytokine), at least in part by repressing the GALNT7 gene, resulting in pro-metastatic effects in vivo [55]. Regarding autoimmunity, miR-30d-5p is capable of regulating the microbiome in the experimental autoimmune encephalomyelitis model, which in turn results in enhanced immunosuppression and the amelioration of the multiple-sclerosis-like symptoms [56]. In the context of T1D, miR-30d is found both up- [57] and downregulated [32] in the plasma of people with T1D versus control subjects and is also increased in exosomes from lactating mothers with T1D [58]. Nevertheless, little is known about its immunological functions. In this work, multiple lines of evidence point to a link between miR-30d-5p and immunoregulatory processes. First, by analyzing its potential targets, we found that miR-30d-5p is mainly enriched for fatty acid biosynthesis pathways, which have key roles in T cell development and immune responses [59]. Second, we managed to classify some of its predicted and validated targets and discovered that this miRNA affects genes such as CD200, CCL5, CTLA4, NFAT5, PRDM1, and TGFBR2. In addition, the direct inhibition of miR-30d-5p in the NOD mouse model led to changes in different immune cell subsets in secondary lymphoid organs and immune cell infiltration in the pancreatic islets. Upon miR-30d-5p inhibition, the levels of Tregs were significantly increased in PLN. We found that this expansion could be explained by the increase in the expression levels of CD200 in PLN cells, which is a direct target of miR-30d-5p [55]. In fact, CD200-CD200R-mediated immunosuppression can occur through the induction of FoxP3+ Tregs [60,61]. Since miR-30d-5p is upregulated in PR vs. non-PR, we should expect lower levels of Tregs along this phase. In a longitudinal study that included the same pediatric patients, we previously found decreased levels of peripheral Tregs at the PR phase in comparison to non-remission, even after 12 months from T1D diagnosis [14], which is accordant with other studies showing diminished Treg levels during the honeymoon phase or after one year [7,62]. Furthermore, although we could not confirm a miR-30d-5p/Blimp-1(PRDM1)/PD-1(PDCD1) axis acting on splenocytes of NOD mice, miR-30d-5p is probably influencing PD-1 expression through other mechanisms that need further research. Interestingly, a recent longitudinal study found a relationship between the PR phase and the restoration of the PD-1/PD-L1 axis on peripheral T cells, suggesting an immunoregulatory mechanism that is absent in patients without remission [15]. These results follow what we saw in the spleen since a higher expression of miR-30d-5p (as happens during PR) was associated with an increased expression of PD-1. Moreover, the reduced levels of pre-effector-like T lymphocytes together with the increased levels of EM and CM T lymphocytes upon miR-30d-5p inhibition could reflect the wave of differentiation into the effector and memory phenotypes, which have a key role in amplifying inflammation. In fact, the expansion of CM T cells might boost the pathogenic potential of the peripheral T cell pool and favor autoimmunity. Finally, the slight increase in the immune cell infiltration into the pancreatic islets could be related to the enhanced effector function of these T lymphocytes. Therefore, the upregulation of miR-30d-5p during the PR might be related to a decreased immune cell infiltration in the islets of Langerhans and the amelioration of the inflammatory microenvironment, with the consequent prevention of β-cell apoptosis. Nevertheless, it would be interesting to identify the nature of the infiltrating T lymphocytes. We must take into consideration that the observed insulitis might be enriched on Tregs since there is an increase in these regulatory cells in the draining lymph nodes. Taking all these data together, we hypothesize that in the absence of miR-30d-5p (non-remission scenario), T lymphocytes do not receive the inhibitory signal through PD-1 because of its low expression levels, which could potentiate their effector functions and their contribution to the inflammatory immune cell infiltrate into the pancreatic islets. At the same time, increased levels of Tregs are generated—in part due to the CD200 upregulation—to try to tackle this enhanced immune response. This study has limitations that must be taken into account when interpreting the results. First of all, the sample size is relatively small, yet we were able to validate some of the miRNAs by RT-qPCR, a highly recommended step [63]. Even though we verified the absence of batch effects and the suitability of the normalized data for the differential expression analysis, another limitation in the RNA sequencing experiment was the absence of technical replicates between runs. Furthermore, although the NOD model spontaneously recapitulates autoimmunity in pancreatic islets, there are key differences in disease development between NOD mice and humans. Another important point to consider is that despite having been able to indirectly test the effect of the miR-30d-5p inhibitor by analyzing the up- or downregulation of some miRNA target genes, it could be interesting to test its specific delivery to T lymphocytes. In this sense, numerous studies have used inhibitory/mimic miRNAs coupled to fluorescent molecules to see their accumulation in vivo in different tissues, including lymph nodes [64,65,66,67]. ## 4.1. Human Sample Collection and T1D Remission Follow-Up The longitudinal discovery cohort for the small RNA sequencing was composed of 17 pediatric patients with new-onset T1D (11 remitters and 6 non-remitters) and 17 age- and sex-matched non-diabetic control subjects. The validation cohort was composed of non-paired samples that included 8 pediatric patients with new-onset T1D, 10 remitters, 9 non-remitters (5 of them also included in the discovery cohort), and 15 age- and sex-matched control subjects. All patients fulfilled the American Diabetes Association classification criteria for T1D [68], with at least one positive anti-islet autoantibody at disease onset (to insulinoma-antigen 2 or glutamic acid decarboxylase). Inclusion criteria were 4–18 years of age and normal body mass index according to the Spanish Body Mass Index pediatric cohort growth chart [69]. Exclusion criteria were being under immunosuppressive or anti-inflammatory treatment, type 2 diabetes, pregnancy, compromised kidney function, or liver diseases. The same inclusion/exclusion criteria were used for non-diabetic controls. T1D data collection occurred for over a year in two University Hospitals of our geographical area, Germans Trias i Pujol (Badalona) and Parc Taulí (Sabadell). Blood samples of 6 mL were obtained at disease onset and PR or 8 months for non-remitter patients in EDTA tubes (BD Biosciences, San Jose, CA, USA). Blood samples from control subjects without T1D were also acquired following the same protocol. Plasma samples were always obtained within the first hour after venipuncture. The tube containing the blood sample was centrifuged at 1900× g at 4 °C for 10 min. The plasma was then aspirated to a 1.5 mL Eppendorf (without disturbing the intermediate layer containing white blood cells and platelets) and centrifuged at 16,000× g at 4 °C for another 10 min to remove additional cellular nucleic acids bound to cellular debris. Finally, 250 μL of clear plasma was pipetted into a 1.5 mL Eppendorf and stored at −80 °C. At disease onset, all samples were collected between 1 and 14 days after diagnosis. To measure PR, we calculated the insulin dose-adjusted HbA1c (IDAA1c) using both the HbA1c value and the insulin requirement as HbA1c (%) + [4 × insulin dose (U/kg/day)]. An IDAA1c equal to or lower than nine indicated the PR phase [70]. Given that the highest percentage of patients in PR is detected within the first 2–6 months after diagnosis, those who did not meet the criteria of PR after 8 months were defined as non-remitters. ## 4.2. Clinical and Laboratory Testing Clinical descriptors on each patient and control subject were collected, including age, sex, and body mass index. Blood samples from patients with T1D were obtained for centralized measurement of HbA1c, basal non-fasting C-peptide (which reflects residual insulin storage), genetics, and immunology; insulin requirements were also recorded. HbA1c was determined by high-performance liquid chromatography (ADAMS A1c HA-8180V, Arkray, MN, USA) in all patients at each time point. Basal non-fasting C-peptide was determined by ELISA (Architect i2000, Abbott, IL, USA) in both controls and patients at each time point. ## 4.3. Total RNA Isolation Total cell-free and exosomal RNA, including miRNA, was isolated from 200 µL of plasma from the discovery cohort using the miRNeasy Serum/Plasma Advanced Kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions. RNA purity, integrity and concentration were determined using TapeStation 2200 (Agilent High Sensitivity Screen Tape, Agilent Technologies Inc., Santa Clara, CA, USA). RNA was stored at −80 °C until use. ## 4.4. RNA Library Preparation, Sequencing, and Data Analysis After the isolation of RNA, 1 µg was used to prepare RNA libraries by D-Plex Small RNA-seq Kit (Diagenode, Liege, Belgium), following the manufacturer’s instructions. After PCR amplification, size selection of fragments and adapter dimer removal were conducted in a $6\%$ polyacrylamide gel (Invitrogen, Carlsbad, CA, USA), and library quality controls were assessed with a TapeStation 2200 using a High Sensitivity D1000 Screen Tape (Agilent). Then, small RNA libraries were sequenced on the Illumina NextSeq 1000 System (SBS-based sequencing technology, Illumina, San Diego, CA, USA) in a run of 92 and 2 × 71 cycles and a high output sequencing mode. Data were retrieved from the sequencer in the form of fastq files. The fastq files of the same sample corresponded to different runs of the same library. In this study, up to six runs were performed for each sample to achieve the desired sequencing depth (ranging from 1.6 to 35.1 million reads depending on the sample). Samples were randomly distributed among the six sequencing runs regardless of the group to which they belong. Trimming steps were further conducted using the Cutadapt tool. This trimming included the removal of the first 16 bp of each read (corresponding to unique molecular identifiers), the polyA tail, and Illumina adapter sequences. Additionally, trimmed sequences of less than 18 bp in length were discarded. After trimming, the quality of the reads (Fastq files) was assessed with FastQC. All the reads were treated as single-end reads, a fact that allowed merging the reads from different runs according to their sample of origin with the multiQC tool. Next, the Subread/Rsubread package was used to map the sequencing reads to the genome of reference and quantify the aligned reads. For the read summarization/quantification step, annotations for precursor and mature miRNAs were obtained from the miRBase v22 database. First, mapped reads were quantified using the mature miRNA annotations contained in miRBase only. Then, unassigned reads were further quantified using the remaining small RNA annotations from miRBase (for pre-miRNA) and Gencode (for other small ncRNAs) databases. Filtering and normalization steps were performed using edgeR 3.34.1 and Limma v.3.48.3 packages from Bioconductor in R. Here, a minimal rule was applied to keep only transcripts that had at least one count per million in at least five samples, and the trimmed mean of M-values normalization method was performed to eliminate composition biases between libraries. Different types of quality controls were also performed (multi-dimensional scaling plot analysis, Euclidean distances between samples) to check that the normalized data were appropriate for the differential expression analysis; by doing so, no outliers or batch effects were detected. For differential expression testing, the Limma’s package v.3.48.3 was used, specifically the limma-voom pipeline. Normalized data were transformed to log2, and DEMs were selected by adjusting a linear model with empirical Bayes moderation of the variance and, in the case of paired samples, a paired design was considered. Data were adjusted for multiple testing to obtain strong control over the FDR using the Benjamini and Hochberg method; however, since these criteria yield too few small RNAs, unadjusted p-values of ≤0.05 were considered for the significance criteria. Therefore, miRNAs with a p-value ≤ 0.05 and log2(FC) >1 were considered upregulated, whereas those with log2 < 1 were considered downregulated. The data for this study were deposited in the European Nucleotide Archive at EMBL-EBI under accession number PRJEB58187 (https://www.ebi.ac.uk/ena/browser/view/PRJEB58187, accessed on 22 December 2022). ## 4.5. Gene Targets for miRNAs In this study, the multiMiR Bioconductor’s package was used to identify miRNA target sites in different databases (miRecords, miRTarBase, and TarBase for validated targets; DIANA-microT, ElMMo, MicroCosm, miRanda, miRDB, PicTar, PITA, and TargetScan for predicted targets). In order to classify the vast number of validated gene targets for each miRNA, a list of keywords was generated and distributed in three different groups: [1] metabolism, [2] regeneration, and [3] immune system. In this way, the metabolism group was composed of keywords like “mTOR signaling” or “insulin signaling pathway”, while the immune system group was composed of words such as “T cell activation” or “dendritic cell”. Then, using a computer logarithm, the Entrez summary of each gene was searched for those keywords, and the genes were consequently classified into one group or another. ## 4.6. Gene Ontology and Pathway Analysis miRNA GO and pathway analysis were performed using the open-access web server DIANA-miRPath v3.0 (http://www.microrna.gr/miRPathv3, accessed on 13 September 2022) [71] using the 14 DEMs with validated target genes between PR and non-PR groups to search for potential biological pathways under their regulation. Biological processes and enriched pathways were investigated using the DIANA-TarBase v7.0 [72], a database that provides high-quality, manually curated and experimentally validated miRNA–target interactions. Significance levels were calculated by using Fisher’s exact test meta-analysis method with Benjamini–Hochberg’s FDR correction (q-value < 0.05). The statistically significant biological processes and their corresponding q-values were introduced in the web server REVIGO (http://revigo.irb.hr/) [73], which takes long lists of GO terms and summarizes them by removing the redundant ones. Interactive graphs showing the relationship between biological processes link highly similar GO terms by edges (using the SimRel semantic similarity measure), where the line width indicates the degree of similarity, and the color of the bubbles is the user-provided p-value. Functional enrichment analysis of miRNA target genes was performed using pathway annotation from the KEGG database and a posteriori of the statistical analysis. In this mode, the server identifies all the significantly targeted pathways by the selected miRNAs. The enrichment analysis is first carried out by the server, and the significance levels (p-values) between each miRNA and each pathway are computed. Subsequently, for each pathway a merged p-value is extracted by combining the previously calculated significance levels using the Fisher’s exact test meta-analysis method and Benjamini–Hochberg’s FDR (q-value) to compensate for multiple testing. Since comparable miRNAs are clustered together, the heatmap makes it possible to identify miRNA subclasses or pathways that define them. ## 4.7. Quantitative RT-qPCR To validate the small RNA-seq results, RT-qPCR was performed on human plasma samples from the validation cohort. Because plasma samples hemolyzed during acquisition can be contaminated by erythrocyte miRNAs [74], the degree of hemolysis was determined based on the optical density at 414 nm (absorbance peak of free hemoglobin) by spectrophotometry (Nanodrop 1000 Spectrophotometer, ThermoFisher Scientific, Waltham, Massachusetts, USA), and the severely hemolyzed samples (OD414 > 0.3) were discarded (Figure S5). Then, RNA isolation was conducted as described above. RNA was reverse transcribed to cDNA with the TaqMan™ Advanced miRNA cDNA Synthesis Kit (ThermoFisher Scientific) following the manufacturer’s instructions and by using the Veriti® Thermal Cycler (ThermoFisher Scientific). To monitor retrotranscription reproducibility, we spiked in 5′-phosphorylated *Arabidopsis thaliana* miR-159a (ath-miR-159a, uuuggauugaagggagcucua), a synthetic oligonucleotide for exogenous control (ThermoFisher Scientific), to cDNA synthesis. Briefly, poly(A) polymerase was used to add a 3′-adenosine tail to the miRNA, which underwent adaptor ligation at the 5′ end. Then, a Universal RT primer (which also incorporates an adaptor) bound to the 3′ poly(A) tail and the miRNA was reverse transcribed. To improve the detection of low-expressing miRNA targets, the cDNA was next pre-amplified using the Universal forward and reverse miR-Amp Primers (which bind to the adaptors) and miR-Amp Master Mix (ThermoFisher Scientific). The 50 µL miR-Amp reaction product was stored at −20 °C until use. Amplified cDNA was 1:10 diluted, and miRNAs were profiled by RT-qPCR using the TaqMan™ Fast Advanced Master Mix (Applied Biosystems, Waltham, MA, USA) with TaqMan Advanced miRNA Assays (ThermoFisher Scientific) in 15 µL PCR reactions in triplicate. Table 3 shows the list of the TaqMan Assays used. MiRNAs to validate were chosen based on [1] the number of reads obtained in the small RNA-seq and their wide expression in most of the samples (>$70\%$), [2] miRNAs most differentially expressed between PR and non-PR patients, and [3] target genes involved in immune system functions. Plates were run on a LightCycler®480 RT-PCR machine (Roche, Mannheim, Germany). All analyzed miRNAs showed a Ct < 30. Relative values were calculated with the 2−∆Ct method [75], and results are given as arbitrary units. Currently, a universally accepted normalization strategy based on endogenous miRNAs is still lacking. NormFinder, an algorithm for identifying the optimal normalization gene among a set of candidates [76], was used to identify the most stable miRNA within our normalized small-RNA seq data to be used as an endogenous control. In our case, the miRNA with the best stability value was miR-16-1-3p. Raw Ct of miR-16-1-3p in samples from the validation cohort can be found in Figure S6. ## 4.8. Mice Wild-type NOD mice were originally obtained from the Jackson Laboratory (Bar Harbor, ME, USA) and then kept in the Animal Facility of the Centre de Medicina Comparativa i Bioimatge de Catalunya (CMCiB) under specific pathogen-free conditions. The colony was subjected to a 12 h dark/12 h light cycle and controlled temperatures between 19–23 °C with 40–$60\%$ humidity and fed with ad libitum access to acidic water at pH 5 and irradiated Teklad Global $18\%$ Protein Rodent Diet (Harlan, Indianapolis, IN, USA). In this study, only prediabetic NOD females of 9 weeks of age were used. In order to detect and exclude mice with T1D, glycosuria levels were monitored weekly using urine test strips (Combiscreen, Analyticon Biotechnologies AG, Lichtenfels, Germany), and T1D was confirmed when glycemia rose above 300 mg/dL in one glucotest control (OneTouch Verio Reflect®, LifeScan IP Holdings, LLC., Zug, Switzerland). ## 4.9. In Vivo miR-30d-5p Inhibitor Administration The mature nucleotide sequence of mmu-miR-30d-5p (5′-UGUAAACAUCCCCGACUGGAAG-3′) was obtained from www.mirbase.org (accessed on 3 May 2022), which is homologous between mice and man. Here, we used an antisense oligonucleotide, called miRCURY LNATM miRNA inhibitor, for miR-30d-5p (LNA-anti-miR-30d-5p) and a miRNA inhibitor negative control A (scrambled LNA) (Exiqon Co., Copenhagen, Denmark). Thus, three groups composed of six prediabetic NOD mice of 9 weeks of age each were respectively treated with [1] miRNA-inhibitor (inhibitor probe mmu-miR-30d-5p), [2] miRNA-inhibitor control (negative control A probe), or [3] saline (PBS, sham group). Mice received four i.p. doses every three days at 9 mg/kg (miRNA-inhibitor or miRNA-inhibitor control) in 200 µL saline solution (PBS, RNase free). Twenty-four hours after the last injection (or 10 days after the first one), mice were euthanized by i.p. ketamine (75 mg/kg)–xylazine (10 mg/kg) injection. Blood was collected via cardiac puncture. The spleen and PLN were harvested and processed. Pancreases were harvested, snap-frozen in an isopentane/cold acetone bath and stored at −80 °C until use. ## 4.10. Insulitis Score The degree of islet infiltration by leukocytes (insulitis) was determined in the pancreas of six mice per group at the end of the study. Briefly, non-overlapping cryosections of 6 µm were obtained, placed on a slide, and stained with hematoxylin and eosin. To score insulitis, a minimum of 40 islets per animal were analyzed under a light microscope, as previously described [77]: 0, intact islets/no insulitis; 1, peri-islet infiltrates; 2, <$25\%$ islet infiltration; 3, 25–$75\%$ islet infiltration; and 4, >$75\%$ islet infiltration or complete islet destruction. A double-blind analysis was performed by independent observers. ## 4.11. Flow Cytometry To determine changes in the percentage of immune cell subpopulations induced by miRNA blockade, the spleen and PLN of all mice were immunophenotyped by flow cytometry. ## 4.11.1. Leukocyte Isolation from Spleen and PLN Splenocyte and leukocyte suspensions were obtained by the mechanical disruption of the spleen and PLN, respectively, and washed twice with RPMI-1640 (Biowest, Nuaille, France) + $10\%$ fetal bovine serum (Gibco, Invitrogen, Carlsbad, CA, USA) (R-10) in order to collect all the cells. The tissue remains were allowed to precipitate for 2 min, and the supernatant was obtained, which was then centrifuged at 400× g for 5 min at room temperature. Afterward, in the case of the splenocyte suspension, erythrocytes were lysed with 5 mL of hemolysis solution [500 mL deionized H2O (Milli-Q Direct, Merck Millipore, Burlington, MA, USA) plus 1.0297 g Trizma Hydrochloride (Sigma-Aldrich, Saint Louis, MO, USA) and 3.735 g NH4Cl (Probus, Badalona, Spain)] for 5 min at room temperature, which was next blocked by adding 5 mL of R-10. Cells were centrifuged again at 400× g for 5 min at room temperature and further washed with another 5 mL of R-10. Finally, leukocyte suspensions from both spleen and PLN were resuspended in 1–3 mL and 200 µL of PBS + $2\%$ fetal bovine serum, respectively. ## 4.11.2. Viability and Cell Counting To assess cell viability and counting, 10 µL of cells were incubated with 2 µL of 7-aminoactinomicina D (7-AAD, BD Biosciences) in 48 µL of PBS for 15 min at room temperature and protected from light. Then, 10 µL of Perfect Count Microspheres (Cytognos SL, Salamanca, Spain) were added to perform cell counting. Cells were acquired by FACSCanto II flow cytometer (BD Biosciences) using the FACSDiva software (BD Biosciences). ## 4.11.3. Immunophenotype Next, the percentages of T and B lymphocytes and DCs were assessed by flow cytometry. For phenotype labeling, 0.5 × 106 cells per panel were used; one of them was designed for the study of different T lymphocyte subpopulations (panel 1), and the other was designed for the detection of conventional DCs (panel 2). For the T lymphocyte panel, surface staining was carried out with CD3 PE, CD4 APC, CD8 V500, CD44 BV786, CD62L APC-Cy7, PD-1 PE-Cy7, and CD25 PerCP-Cy5.5, and intracellular staining was carried out with FoxP3 FITC. For the DC panel, surface staining was carried out with CD3 PE, CD19 V450, CD11c PE-Cy7, and MHC-II APC. Further information on antibodies used can be found in Table 4. First, surface staining was performed. Cells were incubated with the appropriate monoclonal antibodies for 20 min at 4 °C in the dark. If no intracellular staining was needed, cells were then washed with FACSFlow™ Sheath Fluid (ThermoFisher Scientific) and resuspended in the same solution for their acquisition or, for the DC panel, incubated with 4 µL of 7-AAD in 100 µL of PBS for 15 min at room temperature and protected from light. In the case of the intracellular staining with FoxP3, cells were washed with cold FACSFlow™ Sheath Fluid (ThermoFisher Scientific) at 400× g for 5 min and fixed with 1 mL of Fixation/Permeabilization solution (1 Fix/Perm concentrate: 3 diluent solution, ThermoFisher Scientific) for 45 min at 4 °C in the dark. After the incubation, cells were washed twice with 2 mL of 1X Permeabilization buffer (400× g, 5 min, at room temperature), and the supernatant was discarded. Then, fixed and permeabilized cells were stained with FoxP3 FITC for 40 min at 4 °C in the dark. After that, cells were washed twice with 2 mL 1X Permeabilization buffer at 400× g for 5 min and suspended in 200 µL FACSFlow™ Sheath Fluid (ThermoFisher Scientific). At least 100.000 leukocyte events per sample were acquired using FACSCanto II and LSR Fortessa flow cytometers (BD Biosciences). Necrotic and apoptotic cells were excluded from the analysis based on their forward scatter-A/side scatter-A properties and doublets were excluded by forward scatter-A/forward scatter-H. Fluorescence minus one controls were used to assess PD-1, CD25, and FoxP3 staining positivity. Data were analyzed using FlowJo software (Tree Star Inc., Ashland, OR, USA). ## 4.12. Statistical Analysis Data are presented as mean ± SD or SEM or as percentages, where appropriate. The distribution of continuous variables was tested for normality by the Kolmogorov–Smirnov test. For non-normally distributed variables, comparisons between two groups were performed by a non-parametric Mann–Whitney test, and comparisons between three or more groups were performed by Kruskal–Wallis with Dunn’s post hoc test. Correlations between variables were tested by using a two-tailed Spearman’s test. Weight levels of mice throughout the study were compared using two-way ANOVA with Tukey’s multiple comparisons test. Multivariate statistical analysis was performed using principal component analysis. A Fisher’s exact test meta-analysis method with Benjamini–Hochberg’s FDR correction was used to calculate the significant targeted biological processes and KEGG pathways. The complete linkage clustering method was used for the hierarchical clustering of pathways and miRNAs. Squared Euclidian distances were determined as distance measures, absolute p-values were used in all calculations, and the significance levels of the interaction were taken into consideration. For all tests, a two-tailed p-value of ≤0.05 was considered statistically significant. Levels of significance are indicated as: *, p ≤ 0.05; **, p ≤ 0.01; ***, and p ≤ 0.001. Analyses were performed using the programs GraphPad Prism 9 (GraphPad Software Inc, San Diego, CA, USA) and R v4.1.0. ## 4.13.1. Human Samples All the experiments were carried out in strict accordance with the principles outlined in the Declaration of Helsinki for human research and after the approval of the Committee on the Ethics of Research of the Germans Trias i Pujol University Hospital and Parc Taulí University Hospital. ## 4.13.2. Mice This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Generalitat de Catalunya. The procedures carried out with animal models were authorized by the Animal Experimentation Ethics Committee of the CMCiB and IGTP and by the Generalitat de Catalunya, and they followed the principles outlined in the Declaration of Helsinki for animal experimental investigation. All the conducted protocols followed the principles of the 3R, prioritizing the welfare of animals used in research. ## 5. Conclusions Circulating microRNAs have been found to be altered in people with recent-onset T1D, although there are no studies concerning PR (honeymoon). Here, we identify a unique plasma microRNA signature during this phase, providing new microRNA candidate biomarkers for the monitoring of PR in pediatric patients with T1D, which may be used for patient stratification and applied in clinical research. Furthermore, this study demonstrates for the first time that miR-30d-5p—an upregulated miRNA during PR—can modulate various immunoregulatory parameters such as Treg levels or PD-1 expression in the NOD mouse model, shedding light on the mechanisms underlying this phase. ## References 1. Patterson C.C., Harjutsalo V., Rosenbauer J., Neu A., Cinek O., Skrivarhaug T., Rami-Merhar B., Soltesz G., Svensson J., Parslow R.C.. **Trends and cyclical variation in the incidence of childhood type 1 diabetes in 26 European centres in the 25 year period 1989–2013: A multicentre prospective registration study**. *Diabetologia* (2018) **62** 408-417. DOI: 10.1007/s00125-018-4763-3 2. van Megen K.M., Spindler M.P., Keij F.M., Bosch I., Sprangers F., van Royen-Kerkhof A., Nikolic T., Roep B.O.. **Relapsing/remitting type 1 diabetes**. *Diabetologia* (2017) **60** 2252-2255. DOI: 10.1007/s00125-017-4403-3 3. Von Herrath M., Sanda S., Herold K.C.. **Type 1 diabetes as a relapsing–remitting disease?**. *Nat. Rev. Immunol.* (2007) **7** 988-994. DOI: 10.1038/nri2192 4. Fonolleda M., Murillo M., Vázquez F., Bel J., Vives-Pi M.. **Remission Phase in Paediatric Type 1 Diabetes: New Understanding and Emerging Biomarkers**. *Horm. Res. Paediatr.* (2017) **88** 307-315. DOI: 10.1159/000479030 5. Tang R., Zhong T., Wu C., Zhou Z., Li X.. **The Remission Phase in Type 1 Diabetes: Role of Hyperglycemia Rectification in Immune Modulation**. *Front. Endocrinol.* (2019) **10** 824. DOI: 10.3389/fendo.2019.00824 6. Karges B., Durinovic-Belló I., Heinze E., Debatin K.-M., Boehm B., Karges W.. **Immunological mechanisms associated with long-term remission of human type 1 diabetes**. *Diabetes/Metabolism Res. Rev.* (2005) **22** 184-189. DOI: 10.1002/dmrr.600 7. Fitas A.L., Martins C., Borrego L.M., Lopes L., Jörns A., Lenzen S., Limbert C.. **Immune cell and cytokine patterns in children with type 1 diabetes mellitus undergoing a remission phase: A longitudinal study**. *Pediatr. Diabetes* (2018) **19** 963-971. DOI: 10.1111/pedi.12671 8. Gomez-Muñoz L., Perna-Barrull D., Villalba A., Rodriguez-Fernandez S., Ampudia R.-M., Teniente-Serra A., Vazquez F., Murillo M., Perez J., Corripio R.. **NK Cell Subsets Changes in Partial Remission and Early Stages of Pediatric Type 1 Diabetes**. *Front. Immunol.* (2021) **11** 3551. DOI: 10.3389/fimmu.2020.611522 9. Narsale A., Lam B., Moya R., Lu T., Mandelli A., Gotuzzo I., Pessina B., Giamporcaro G., Geoffrey R., Buchanan K.. **CD4+CD25+CD127hi cell frequency predicts disease progression in type 1 diabetes**. *J. Clin. Investig.* (2021) **6** e136114. DOI: 10.1172/jci.insight.136114 10. Schloot N.C., Hanifi-Moghaddam P., Aabenhus-Andersen N., Alizadeh B.Z., Saha M.T., Knip M., Devendra D., Wilkin T., Bonifacio E., Roep B.O.. **Association of immune mediators at diagnosis of Type 1 diabetes with later clinical remission**. *Diabet. Med.* (2007) **24** 512-520. DOI: 10.1111/j.1464-5491.2007.02096.x 11. Long S.A., Thorpe J., DeBerg H.A., Gersuk V., Eddy J.A., Harris K.M., Ehlers M., Herold K.C., Nepom G.T., Linsley P.S.. **Partial exhaustion of CD8 T cells and clinical response to teplizumab in new-onset type 1 diabetes**. *Sci. Immunol.* (2016) **1** eaai7793. DOI: 10.1126/sciimmunol.aai7793 12. Arif S., Leete P., Nguyen V., Marks K., Nor N.M., Estorninho M., Kronenberg-Versteeg D., Bingley P.J., Todd J.A., Guy C.. **Blood and Islet Phenotypes Indicate Immunological Heterogeneity in Type 1 Diabetes**. *Diabetes* (2014) **63** 3835-3845. DOI: 10.2337/db14-0365 13. Leete P., Willcox A., Krogvold L., Dahl-Jørgensen K., Foulis A.K., Richardson S.J., Morgan N.G.. **Differential Insulitic Profiles Determine the Extent of β-Cell Destruction and the Age at Onset of Type 1 Diabetes**. *Diabetes* (2016) **65** 1362-1369. DOI: 10.2337/db15-1615 14. Gomez-Muñoz L., Perna-Barrull D., Caroz-Armayones J.M., Murillo M., Rodriguez-Fernandez S., Valls A., Vazquez F., Perez J., Corripio R., Castaño L.. **Candidate Biomarkers for the Prediction and Monitoring of Partial Remission in Pediatric Type 1 Diabetes**. *Front. Immunol.* (2022) **13** 555. DOI: 10.3389/fimmu.2022.825426 15. Li X., Zhong T., Tang R., Wu C., Xie Y., Liu F., Zhou Z.. **PD-1 and PD-L1 Expression in Peripheral CD4/CD8+ T Cells Is Restored in the Partial Remission Phase in Type 1 Diabetes**. *J. Clin. Endocrinol. Metab.* (2020) **105** 1947-1956. DOI: 10.1210/clinem/dgaa130 16. Villalba A., Fonolleda M., Murillo M., Rodriguez-Fernandez S., Ampudia R.-M., Perna-Barrull D., Raina M.B., Quirant-Sanchez B., Planas R., Teniente-Serra A.. **Partial remission and early stages of pediatric type 1 diabetes display immunoregulatory changes. A pilot study**. *Transl. Res.* (2019) **210** 8-25. DOI: 10.1016/j.trsl.2019.03.002 17. Ha M., Kim V.N.. **Regulation of microRNA biogenesis**. *Nat. Rev. Mol. Cell Biol.* (2014) **15** 509-524. DOI: 10.1038/nrm3838 18. Paul P., Chakraborty A., Sarkar D., Langthasa M., Rahman M., Bari M., Singha R.S., Malakar A.K., Chakraborty S.. **Interplay between miRNAs and human diseases**. *J. Cell. Physiol.* (2017) **233** 2007-2018. DOI: 10.1002/jcp.25854 19. Jin F., Hu H., Xu M., Zhan S., Wang Y., Zhang H., Chen X.. **Serum microRNA Profiles Serve as Novel Biomarkers for Autoimmune Diseases**. *Front. Immunol.* (2018) **9** 2381. DOI: 10.3389/fimmu.2018.02381 20. Simpson L.J., Ansel K.M.. **MicroRNA regulation of lymphocyte tolerance and autoimmunity**. *J. Clin. Investig.* (2015) **125** 2242-2249. DOI: 10.1172/JCI78090 21. Xu G., Thielen L.A., Chen J., Grayson T.B., Grimes T., Bridges S.L., Tse H.M., Smith B., Patel R., Li P.. **Serum miR-204 is an early biomarker of type 1 diabetes-associated pancreatic beta-cell loss**. *Am. J. Physiol. Metab.* (2019) **317** E723-E730. DOI: 10.1152/ajpendo.00122.2019 22. Scherm M.G., Daniel C.. **miRNA-Mediated Immune Regulation in Islet Autoimmunity and Type 1 Diabetes**. *Front. Endocrinol.* (2020) **11** 606322. DOI: 10.3389/fendo.2020.606322 23. Grieco F.A., Sebastiani G., Juan-Mateu J., Villate O., Marroqui L., Ladrière L., Tugay K., Regazzi R., Bugliani M., Marchetti P.. **MicroRNAs miR-23a-3p, miR-23b-3p, and miR-149-5p Regulate the Expression of Proapoptotic BH3-Only Proteins DP5 and PUMA in Human Pancreatic β-Cells**. *Diabetes* (2016) **66** 100-112. DOI: 10.2337/db16-0592 24. Guay C., Kruit J.K., Rome S., Menoud V., Mulder N.L., Jurdzinski A., Mancarella F., Sebastiani G., Donda A., Gonzalez B.J.. **Lymphocyte-Derived Exosomal MicroRNAs Promote Pancreatic β Cell Death and May Contribute to Type 1 Diabetes Development**. *Cell Metab.* (2019) **29** 348-361.e6. DOI: 10.1016/j.cmet.2018.09.011 25. Scherm M.G., Serr I., Zahm A.M., Schug J., Bellusci S., Manfredini R., Salb V.K., Gerlach K., Weigmann B., Ziegler A.-G.. **miRNA142-3p targets Tet2 and impairs Treg differentiation and stability in models of type 1 diabetes**. *Nat. Commun.* (2019) **10** 1-15. DOI: 10.1038/s41467-019-13587-3 26. Erener S., Marwaha A., Tan R., Panagiotopoulos C., Kieffer T.J.. **Profiling of circulating microRNAs in children with recent onset of type 1 diabetes**. *J. Clin. Investig.* (2017) **2** e89656. DOI: 10.1172/jci.insight.89656 27. Samandari N., Mirza A.H., Kaur S., Hougaard P., Nielsen L.B., Fredheim S., Mortensen H.B., Pociot F.. **Influence of Disease Duration on Circulating Levels of miRNAs in Children and Adolescents with New Onset Type 1 Diabetes**. *Non-Coding RNA* (2018) **4**. DOI: 10.3390/ncrna4040035 28. Assmann T.S., Recamonde-Mendoza M., Puñales M., Tschiedel B., Canani L.H., Crispim D.. **MicroRNA expression profile in plasma from type 1 diabetic patients: Case-control study and bioinformatic analysis**. *Diabetes Res. Clin. Pract.* (2018) **141** 35-46. DOI: 10.1016/j.diabres.2018.03.044 29. Nielsen L.B., Wang C., Sørensen K., Bang-Berthelsen C.H., Hansen L., Andersen M.-L.M., Hougaard P., Juul A., Zhang C.-Y., Pociot F.. **Circulating Levels of MicroRNA from Children with Newly Diagnosed Type 1 Diabetes and Healthy Controls: Evidence That miR-25 Associates to Residual Beta-Cell Function and Glycaemic Control during Disease Progression**. *Exp. Diabetes Res.* (2012) **2012** 896362. DOI: 10.1155/2012/896362 30. Garavelli S., Bruzzaniti S., Tagliabue E., Di Silvestre D., Prattichizzo F., Mozzillo E., Fattorusso V., La Sala L., Ceriello A., Puca A.A.. **Plasma circulating miR-23~27~24 clusters correlate with the immunometabolic derangement and predict C-peptide loss in children with type 1 diabetes**. *Diabetologia* (2020) **63** 2699-2712. DOI: 10.1007/s00125-020-05237-x 31. Salas-Pérez F., Codner E., Valencia E., Pizarro C., Carrasco E., Pérez-Bravo F.. **MicroRNAs miR-21a and miR-93 are down regulated in peripheral blood mononuclear cells (PBMCs) from patients with type 1 diabetes**. *Immunobiology* (2013) **218** 733-737. DOI: 10.1016/j.imbio.2012.08.276 32. Garavelli S., Bruzzaniti S., Tagliabue E., Prattichizzo F., Di Silvestre D., Perna F., La Sala L., Ceriello A., Mozzillo E., Fattorusso V.. **Blood Co-Circulating Extracellular microRNAs and Immune Cell Subsets Associate with Type 1 Diabetes Severity**. *Int. J. Mol. Sci.* (2020) **21**. DOI: 10.3390/ijms21020477 33. Margaritis K., Margioula-Siarkou G., Giza S., Kotanidou E.P., Tsinopoulou V.R., Christoforidis A., Galli-Tsinopoulou A.. **Micro-RNA Implications in Type-1 Diabetes Mellitus: A Review of Literature**. *Int. J. Mol. Sci.* (2021) **22**. DOI: 10.3390/ijms222212165 34. Aref S., Azmy E., El-Gilany A.H.. **Upregulation of CD200 is associated with regulatory T cell expansion and disease progression in multiple myeloma**. *Hematol. Oncol.* (2015) **35** 51-57. DOI: 10.1002/hon.2206 35. Zhang J., Jima D.D., Jacobs C., Fischer R., Gottwein E., Huang G., Lugar P.L., Lagoo A.S., Rizzieri D.A., Friedman D.R.. **Patterns of microRNA expression characterize stages of human B-cell differentiation**. *Blood* (2009) **113** 4586-4594. DOI: 10.1182/blood-2008-09-178186 36. Infante M., Alejandro R., Fabbri A., Ricordi C.. *The Heterogeneity of Type 1 Diabetes: From Immunopathology to Immune Intervention* (2022) 83-104. DOI: 10.1016/b978-0-12-824466-1.00001-7 37. Battaglia M., Ahmed S., Anderson M.S., Atkinson M.A., Becker D., Bingley P.J., Bosi E., Brusko T.M., DiMeglio L.A., Evans-Molina C.. **Introducing the Endotype Concept to Address the Challenge of Disease Heterogeneity in Type 1 Diabetes**. *Diabetes Care* (2019) **43** 5-12. DOI: 10.2337/dc19-0880 38. Zheng Y., Wang Z., Zhou Z.. **miRNAs: Novel regulators of autoimmunity-mediated pancreatic β-cell destruction in type 1 diabetes**. *Cell. Mol. Immunol.* (2017) **14** 488-496. DOI: 10.1038/cmi.2017.7 39. Santos A.S., Cunha Neto E., Fukui R.T., Ferreira L.R.P., Silva M.E.R.. **Increased Expression of Circulating microRNA 101-3p in Type 1 Diabetes Patients: New Insights Into miRNA-Regulated Pathophysiological Pathways for Type 1 Diabetes**. *Front. Immunol.* (2019) **10** 1637. DOI: 10.3389/fimmu.2019.01637 40. Snowhite I., Pastori R., Sosenko J., Cayetano S.M., Pugliese A.. **Baseline Assessment of Circulating MicroRNAs Near Diagnosis of Type 1 Diabetes Predicts Future Stimulated Insulin Secretion**. *Diabetes* (2020) **70** 638-651. DOI: 10.2337/db20-0817 41. Marchand L., Jalabert A., Meugnier E., Hende K.V.D., Fabien N., Nicolino M., Madec A.-M., Thivolet C., Rome S.. **miRNA-375 a Sensor of Glucotoxicity Is Altered in the Serum of Children with Newly Diagnosed Type 1 Diabetes**. *J. Diabetes Res.* (2016) **2016** 1869082. DOI: 10.1155/2016/1869082 42. Ventriglia G., Mancarella F., Sebastiani G., Cook D.P., Mallone R., Mathieu C., Gysemans C., Dotta F.. **miR-409-3p is reduced in plasma and islet immune infiltrates of NOD diabetic mice and is differentially expressed in people with type 1 diabetes**. *Diabetologia* (2019) **63** 124-136. DOI: 10.1007/s00125-019-05026-1 43. Zhang X., Gu H., Wang L., Huang F., Cai J.. **MiR-885-3p is down-regulated in peripheral blood mononuclear cells from T1D patients and regulates the inflammatory response via targeting TLR4/NF-κB signaling**. *J. Gene Med.* (2019) **22** e3145. DOI: 10.1002/jgm.3145 44. Zurawek M., Dzikiewicz-Krawczyk A., Izykowska K., Ziolkowska-Suchanek I., Skowronska B., Czainska M., Podralska M., Fichna P., Przybylski G., Fichna M.. **miR-487a-3p upregulated in type 1 diabetes targets CTLA4 and FOXO3**. *Diabetes Res. Clin. Pract.* (2018) **142** 146-153. DOI: 10.1016/j.diabres.2018.05.044 45. Samandari N., Mirza A.H., Nielsen L.B., Kaur S., Hougaard P., Fredheim S., Mortensen H.B., Pociot F.. **Circulating microRNA levels predict residual beta cell function and glycaemic control in children with type 1 diabetes mellitus**. *Diabetologia* (2016) **60** 354-363. DOI: 10.1007/s00125-016-4156-4 46. Thomaidou S., Kracht M.J., van der Slik A., Laban S., de Koning E.J., Carlotti F., Hoeben R.C., Roep B.O., Zaldumbide A.. **β-Cell Stress Shapes CTL Immune Recognition of Preproinsulin Signal Peptide by Posttranscriptional Regulation of Endoplasmic Reticulum Aminopeptidase 1**. *Diabetes* (2020) **69** 670-680. DOI: 10.2337/db19-0984 47. Tsukita S., Yamada T., Takahashi K., Munakata Y., Hosaka S., Takahashi H., Gao J., Shirai Y., Kodama S., Asai Y.. **MicroRNAs 106b and 222 Improve Hyperglycemia in a Mouse Model of Insulin-Deficient Diabetes via Pancreatic β-Cell Proliferation**. *Ebiomedicine* (2016) **15** 163-172. DOI: 10.1016/j.ebiom.2016.12.002 48. Mandelbaum A.D., Kredo-Russo S., Aronowitz D., Myers N., Yanowski E., Klochendler A., Swisa A., Dor Y., Hornstein E.. **miR-17-92 and miR-106b-25 clusters regulate beta cell mitotic checkpoint and insulin secretion in mice**. *Diabetologia* (2019) **62** 1653-1666. DOI: 10.1007/s00125-019-4916-z 49. Xin Y., Cai H., Lu T., Zhang Y., Yang Y., Cui Y.. **miR-20b Inhibits T Cell Proliferation and Activation via NFAT Signaling Pathway in Thymoma-Associated Myasthenia Gravis**. *BioMed Res. Int.* (2016) **2016** 9595718. DOI: 10.1155/2016/9595718 50. Ji H., Fan L., Shan A., Wang W., Ning G., Cao Y., Jiang X.. **Let7b-5p inhibits insulin secretion and decreases pancreatic β-cell mass in mice**. *Mol. Cell. Endocrinol.* (2022) **540**. DOI: 10.1016/j.mce.2021.111506 51. Sung Y., Jeong J., Kang R.J., Choi M., Park S., Kwon W., Lee J., Jang S., Park S.J., Kim S.. **Lin28a expression protects against streptozotocin-induced β-cell destruction and prevents diabetes in mice**. *Cell Biochem. Funct.* (2019) **37** 139-147. DOI: 10.1002/cbf.3376 52. Zurawek M., Fichna M., Fichna P., Czainska M., Rozwadowska N.. **Upregulation of**. *J. Immunol. Res.* (2020) **2020** 9484015. DOI: 10.1155/2020/9484015 53. Tang X., Muniappan L., Tang G., Özcan S.. **Identification of glucose-regulated miRNAs from pancreatic β cells reveals a role for miR-30d in insulin transcription**. *Rna* (2008) **15** 287-293. DOI: 10.1261/rna.1211209 54. Zhao X., Mohan R., Özcan S., Tang X.. **MicroRNA-30d Induces Insulin Transcription Factor MafA and Insulin Production by Targeting Mitogen-activated Protein 4 Kinase 4 (MAP4K4) in Pancreatic β-Cells**. *J. Biol. Chem.* (2012) **287** 31155-31164. DOI: 10.1074/jbc.M112.362632 55. Gaziel-Sovran A., Segura M.F., Di Micco R., Collins M.K., Hanniford D., de Miera E.V.-S., Rakus J.F., Dankert J.F., Shang S., Kerbel R.S.. **miR-30b/30d Regulation of GalNAc Transferases Enhances Invasion and Immunosuppression during Metastasis**. *Cancer Cell* (2011) **20** 104-118. DOI: 10.1016/j.ccr.2011.05.027 56. Liu S., Rezende R.M., Moreira T.G., Tankou S.K., Cox L.M., Wu M., Song A., Dhang F.H., Wei Z., Costamagna G.. **Oral Administration of miR-30d from Feces of MS Patients Suppresses MS-like Symptoms in Mice by Expanding Akkermansia muciniphila**. *Cell Host Microbe* (2019) **26** 779-794.e8. DOI: 10.1016/j.chom.2019.10.008 57. Seyhan A.A., Lopez Y.O.N., Xie H., Yi F., Mathews C., Pasarica M., Pratley R.E.. **Pancreas-enriched miRNAs are altered in the circulation of subjects with diabetes: A pilot cross-sectional study**. *Sci. Rep.* (2016) **6** 31479. DOI: 10.1038/srep31479 58. Frørup C., Mirza A.H., Yarani R., Nielsen L.B., Mathiesen E.R., Damm P., Svare J., Engelbrekt C., Størling J., Johannesen J.. **Plasma Exosome-Enriched Extracellular Vesicles From Lactating Mothers With Type 1 Diabetes Contain Aberrant Levels of miRNAs During the Postpartum Period**. *Front. Immunol.* (2021) **12** 744509. DOI: 10.3389/fimmu.2021.744509 59. Qian X., Yang Z., Mao E., Chen E.. **Regulation of fatty acid synthesis in immune cells**. *Scand. J. Immunol.* (2018) **88** e12713. DOI: 10.1111/sji.12713 60. Coles S.J., Hills R.K., Wang E.C.Y., Burnett A.K., Man S., Darley R.L., Tonks A.. **Increased CD200 expression in acute myeloid leukemia is linked with an increased frequency of FoxP3+ regulatory T cells**. *Leukemia* (2012) **26** 2146-2148. DOI: 10.1038/leu.2012.75 61. Gorczynski R., Chen Z., Khatri I., Yu K.. **sCD200 Present in Mice Receiving Cardiac and Skin Allografts Causes Immunosuppression In Vitro and Induces Tregs**. *Transplantation* (2013) **95** 442-447. DOI: 10.1097/TP.0b013e3182754c30 62. Sanda S., Roep B.O., von Herrath M.. **Islet antigen specific IL-10+ immune responses but not CD4+CD25+FoxP3+ cells at diagnosis predict glycemic control in type 1 diabetes**. *Clin. Immunol.* (2008) **127** 138-143. DOI: 10.1016/j.clim.2007.12.003 63. Fauth M., Hegewald A.B., Schmitz L., Krone D.J., Saul M.J.. **Validation of extracellular miRNA quantification in blood samples using RT-qPCR**. *FASEB BioAdvances* (2019) **1** 481-492. DOI: 10.1096/fba.2019-00018 64. Elmén J., Lindow M., Schütz S., Lawrence M., Petri A., Obad S., Lindholm M., Hedtjärn M., Hansen H.F., Berger U.. **LNA-mediated microRNA silencing in non-human primates**. *Nature* (2008) **452** 896-899. DOI: 10.1038/nature06783 65. Serr I., Scherm M.G., Zahm A.M., Schug J., Flynn V.K., Hippich M., Kälin S., Becker M., Achenbach P., Nikolaev A.. **A miRNA181a/NFAT5 axis links impaired T cell tolerance induction with autoimmune type 1 diabetes**. *Sci. Transl. Med.* (2018) **10** eaag1782. DOI: 10.1126/scitranslmed.aag1782 66. Krützfeldt J., Rajewsky N., Braich R., Rajeev K.G., Tuschl T., Manoharan M., Stoffel M.. **Silencing of microRNAs in vivo with ‘antagomirs’**. *Nature* (2005) **438** 685-689. DOI: 10.1038/nature04303 67. Elmén J., Lindow M., Silahtaroglu A., Bak M., Christensen M., Lind-Thomsen A., Hedtjärn M., Hansen J.B., Hansen H.F., Straarup E.M.. **Antagonism of microRNA-122 in mice by systemically administered LNA-antimiR leads to up-regulation of a large set of predicted target mRNAs in the liver**. *Nucleic Acids Res.* (2007) **36** 1153-1162. DOI: 10.1093/nar/gkm1113 68. **Classification and Diagnosis of Diabetes: Standards of Medical Care in Diabetes—2021**. *Diabetes Care* (2021) **44** S15-S33. DOI: 10.2337/dc21-S002 69. Lezcano A.C., García J.F., Ramos C.F., Longás A.F., López-Siguero J., González E.S., Ruiz B.S., Fernández D.Y.. **Estudio transversal español de crecimiento 2008. Parte II: Valores de talla, peso e &iacute;ndice de masa corporal desde el nacimiento a la talla adulta**. *Anales de Pediatría* (2008) **Volume 68** 552-569. DOI: 10.1157/13123287 70. Mortensen H.B., Hougaard P., Swift P., Hansen L., Holl R.W., Hoey H., Bjoerndalen H., de Beaufort C., Chiarelli F., Danne T.. **New Definition for the Partial Remission Period in Children and Adolescents With Type 1 Diabetes**. *Diabetes Care* (2009) **32** 1384-1390. DOI: 10.2337/dc08-1987 71. Vlachos I.S., Zagganas K., Paraskevopoulou M.D., Georgakilas G., Karagkouni D., Vergoulis T., Dalamagas T., Hatzigeorgiou A.G.. **DIANA-miRPath v3.0: Deciphering microRNA function with experimental support**. *Nucleic Acids Res.* (2015) **43** W460-W466. DOI: 10.1093/nar/gkv403 72. Vlachos I.S., Paraskevopoulou M.D., Karagkouni D., Georgakilas G., Vergoulis T., Kanellos I., Anastasopoulos I.-L., Maniou S., Karathanou K., Kalfakakou D.. **DIANA-TarBase v7.0: Indexing more than half a million experimentally supported miRNA:mRNA interactions**. *Nucleic Acids Res.* (2014) **43** D153-D159. DOI: 10.1093/nar/gku1215 73. Supek F., Bošnjak M., Škunca N., Smuc T.. **REVIGO Summarizes and Visualizes Long Lists of Gene Ontology Terms**. *PLoS ONE* (2011) **6**. DOI: 10.1371/journal.pone.0021800 74. Smith M.D., Leemaqz S.Y., Jankovic-Karasoulos T., McAninch D., McCullough D., Breen J., Roberts C.T., Pillman K.A.. **Haemolysis Detection in MicroRNA-Seq from Clinical Plasma Samples**. *Genes* (2022) **13**. DOI: 10.3390/genes13071288 75. Livak K.J., Schmittgen T.D.. **Analysis of relative gene expression data using real-time quantitative PCR and the 2**. *Methods* (2001) **25** 402-408. DOI: 10.1006/meth.2001.1262 76. Andersen C.L., Jensen J.L., Ørntoft T.F.. **Normalization of Real-Time Quantitative Reverse Transcription-PCR Data: A Model-Based Variance Estimation Approach to Identify Genes Suited for Normalization, Applied to Bladder and Colon Cancer Data Sets**. *Cancer Res.* (2004) **64** 5245-5250. DOI: 10.1158/0008-5472.CAN-04-0496 77. Alba A., Puertas M.C., Carrillo J., Planas R., Ampudia R., Pastor X., Bosch F., Pujol-Borrell R., Verdaguer J., Vives-Pi M.. **IFNβ Accelerates Autoimmune Type 1 Diabetes in Nonobese Diabetic Mice and Breaks the Tolerance to β Cells in Nondiabetes-Prone Mice**. *J. Immunol.* (2004) **173** 6667-6675. DOI: 10.4049/jimmunol.173.11.6667
--- title: 'Student Health and Social Care Professionals’ Health Literacy Knowledge: An Exploratory Study' authors: - Helen Wood - Gabrielle Brand - Rhonda Clifford - Sinead Kado - Kenneth Lee - Liza Seubert journal: Pharmacy year: 2023 pmcid: PMC10037638 doi: 10.3390/pharmacy11020040 license: CC BY 4.0 --- # Student Health and Social Care Professionals’ Health Literacy Knowledge: An Exploratory Study ## Abstract Health literacy is essential for shared decision-making and improved health outcomes, and patients with inadequate health literacy often need additional support from health and social care professionals. Despite global calls for developing tertiary-level health literacy education, the extent of this in Australian health and social care professional degrees is unknown. This research explored students’ health literacy knowledge across five health and social care professional disciplines. A web-based questionnaire was disseminated to student health and social care professionals enrolled in one of two Australian universities. Questions explored students’ factual and conceptual health literacy knowledge, and responses were inductively themed and reported descriptively. Of the 90 students who participated, the depth of health literacy knowledge was low. Students frequently identified understanding as components of health literacy; however, most students did not identify health information access, appraisal and use. Additionally, students’ knowledge of helping patients with inadequate health literacy was limited. Adjusting patient education to their health literacy level and evaluating patient understanding was poorly understood. Without a solid understanding of fundamental health literacy principles, newly-graduated health and social care professionals will be poorly equipped to facilitate patients’ health literacy-related challenges in the community. Further exploration of health literacy education is urgently recommended to identify areas for improvement. ## 1. Introduction Health literacy is a broad concept that describes an individual’s capacity to access, understand, appraise and use health information in a beneficial way [1]. It plays a critical role in achieving positive health outcomes. For example, a person with adequate health literacy is better equipped to take actions that support their own health, such as navigating health and social care systems, understanding health messages and engaging in shared decision-making [1,2,3]. The reverse is also true; a person with inadequate health literacy may struggle to communicate effectively with a health or social care professional, understand health communication or correctly adhere to treatment [1,4,5]. This has been associated with negative health outcomes, such as longer recovery times, development of chronic health conditions or earlier death [4,5,6,7]. While individual factors, such as age or cultural background, influence health literacy [6,8,9,10], the role of health and social care professionals cannot be understated. As first (and often, only) point of patient contact, their professional responsibility extends to ensuring equitable and accessible healthcare for all patients, across the spectrum of individual health literacy capabilities. There is a plethora of evidence, arising primarily from the United States, demonstrating the positive impact that health or social care professionals can have on a patient’s ability to manage their health; initiatives such as ‘Ask Me 3’, ‘Universal Precautions Approach’ or teach-back methods have concentrated on health literacy strategies which help health and social care professionals support patient engagement and understanding [11,12,13,14,15,16]. By implementing specific health literacy-related initiatives, health and social care professionals can improve patients’ treatment adherence [14,17], health-related knowledge [14,16,18,19], patient satisfaction [15,19,20], and health outcomes [16,21,22]. In 2011, an international panel of health literacy advocates published the Calgary Charter on Health Literacy, which proposed the need for developed and evaluated health literacy curricula for health profession students [23]. Similarly, Australia’s National Statement on Health Literacy identified that health literacy education must be integrated into health professionals’ curricula as part of a national action plan—a position recently endorsed by the Australian Medical Association and the World Health Organization [1,24,25]. However, the extent of health literacy education in Australian health and social care professional degrees is largely still unknown. To date, one Australian study evaluated an implemented health literacy curriculum for student pharmacists, using a novel scaffolded education activity where students created and delivered an education session to small patient groups [26]. The success of the education activity was measured using teacher, student and patient evaluations. While promising, this highlights the lack of Australian research that evaluates the health literacy curricula in a range of health and social care professional degrees. When proposing a health literacy training framework, Saunders et al. identified health literacy knowledge as the first of a series of learning outcomes essential to a curriculum, aiming to improve student attitude, knowledge and skills, social health care quality, and patient capacity and satisfaction [27]. This aligns with the position of knowledge as the first step in Bloom’s revised taxonomy, a hierarchical model used to classify components of educational learning [28]. The model proposes that knowledge is the foundation of all learning which students must master in order to attain competence in more complex components [28]. The purpose of this exploratory research was to explore university students’ health literacy knowledge across five health and social care professional disciplines. This will provide a unique insight into students’ interpretation of the current curricula, which is the first step towards the identification of key strategies to help better align health literacy curricula with the National Statement on Health Literacy. ## 2.1. Overview A cross-sectional, web-based questionnaire was chosen for data collection, as it allowed participation from students who may wish to retain anonymity or avoid scrutiny from the researchers [29]. It was disseminated to final-year student health and social care professionals, to gain insight into their understanding of the term ‘health literacy’ and knowledge of its importance to both their patients and themselves as future health and social care professionals. The Human Research Ethics Office at The University of Western Australia granted approval for the study (RA/$\frac{4}{20}$/5960). All data collected were managed and stored securely as per ethics requirements. ## 2.2. Questionnaire Development The questionnaire was created by health professional academics with expertise in health literacy (HW, GB, KL), giving consideration to previously-identified health literacy competencies for health professionals [27,30,31]. To determine students’ level of health literacy knowledge, questions were mapped to the first two dimensions of Bloom’s revised taxonomy of educational learning objectives: factual knowledge and conceptual knowledge [28]. Factual knowledge relates to the basic definitions of a concept which familiarize students with the topic. Conceptual knowledge extends further to the principles and categories which underpin that topic [28]. The questionnaire, provided in Supplementary material SA, was created in Qualtrics (Qualtrics, Provo, UT, USA), an internet-based questionnaire platform, and comprised two sections:Demographics and their understanding of the term ‘health literacy’ encompassing the first dimension, or ‘factual knowledge’. It asked for words or phrases that students think of when they hear the term ‘health literacy’. Conceptual health literacy knowledge encompassing the second dimension, or ‘conceptual knowledge’. This section asked for three signs of inadequate patient health literacy, three potential consequences of inadequate patient health literacy, and three actions that health and social care professionals can take to help a patient with inadequate health literacy. Additional space was provided at the conclusion of the questionnaire for students to write any comments on the health literacy education embedded within their curriculum. For all questions, responses were collected in open text boxes. The consent statement, with a single option to select ‘yes’, required an affirmative response to continue. All other questions prompted for a response, but students were able to leave questions unanswered. The first page sought the students’ understanding of the term ‘health literacy’; each subsequent page provided the following definition of health literacy to ensure all conceptual knowledge questions were answered with the same definition in mind: ‘Health literacy is the degree to which individuals have the capacity to obtain, process and understand basic health information and services needed to make appropriate health decisions’ [32]. The ‘back’ button was disabled on the questionnaire, so students could not amend their original understanding of health literacy based on the provided definition. Participants were prevented from submitting multiple responses by utilizing a function of the survey tool which blocks subsequent access attempts. The questionnaire was piloted with five student health professionals to ensure the questions were appropriate and easily understood, and that online formatting worked as intended on different devices. No changes were recommended during piloting, and data generated from piloting were excluded from analysis. ## 2.3. Participants and Recruitment The target population included students enrolled in the following health disciplines from one of two Western Australian universities: Dental Medicine, Nursing, Pharmacy, Podiatric Medicine, and Social Work and Social Policy. Prior to recruitment, a member of the research team (HW) contacted academic staff from each health or social care discipline to identify potential student cohorts who had completed all health literacy-related content. Students were eligible to participate if they consented and were:Aged 18 years or older;Currently enrolled in one of the aforementioned health or social care disciplines;Able to read and write in English;Advanced enough through their degree that they had completed all health literacy-related content at the time of participation. Age, enrolment status and English language capabilities were assured due to recruiting students in the final year of their degree in an English-speaking university. Completion of health literacy-related content was verified by targeted recruitment to students in their final year at a time-point recommended by the relevant academic staff; a question in the demographics section relating to their progression through the degree was used to double-check eligibility. Recruitment was conducted through broadcasted announcements on Learning Management System, which could be viewed through the software or via email. A reminder announcement was disseminated one week prior to the survey closing. Potential stress or perceived obligation to participate could have existed due to an unequal relationship between some research members and participants, as some researchers were known to potential participants; this was mitigated using an anonymous, web-based questionnaire and voluntary participation. Additionally, research members did not directly approach any student to encourage or request participation. Students had the opportunity to win an AUD25 e-gift card following participation, but survey responses were not linked to contact information. ## 2.4. Sample Size As this was an exploratory study where a wide range of answers was anticipated, we did not aim for data saturation to determine sample size. Rather, we aimed for a mid-range sample size of between 60 and 99 participants [33]. ## 2.5. Analysis and Reporting Anonymous data collected via Qualtrics were automatically generated into a report after data collection was completed. Free-text responses in the questionnaire were inductively themed following a manifest content analysis process. Two researchers (HW and SK, a pharmacy educator and health professions educator, respectively) independently followed a modified version of the process, described by Kleinheksel et al. [ 34]; after data immersion, a codebook was developed by HW for each of the four core knowledge questions. The codebook was developed iteratively and each response was re-coded when an amendment was made to the codebook. Each codebook contained a:Category: each response was linked to at least one relevant category. The category ‘Vague’ was used when a response was ambiguous and the research team could not definitively derive the intent with context (e.g., ‘Impacts other areas’ as a potential consequence of inadequate health literacy could have been interpreted in multiple ways, with no context to derive intended meaning). The category ‘Other’ was used when a response did not fit within an existing category and was not reported frequently enough to warrant its own category (e.g., ‘Stress in health services’ as a sign of inadequate patient health literacy). One-word answers were categorized verbatim (e.g., the response ‘Understanding’ was categorized as ‘understanding’); when students provided more than one word the response was linked to a separate, more descriptive category (e.g., the response ‘Understanding colds and flu’ was categorized as ‘understanding health generally’).Code: within each category, similar responses were grouped together into codes. A single response could be attached to multiple codes (e.g., ‘Feel unable to navigate health system and end up avoiding health appointments’ was linked to ‘Not knowing where to go’ and ‘Low engagement with health professional’). The code ‘Other’ was used when a response fit within a category, but was not mentioned frequently enough to warrant its own code (e.g., ‘Impatience’ as a code within ‘Patient demeanor’).Description: a description of each category or code to ensure consistent theming between researchers. Key example: an example of a direct quote to enhance the description of the category or code. For the question, ‘List three actions that you as a health or social care professional can take to assist a patient with inadequate health literacy’, some codes had so many responses that themes emerged within a code; when this occurred, sub-codes were created. After coding a random $20\%$ of responses using each draft codebook, HW and SK met to compare analyses and discuss inconsistencies. Three codebooks contained no inconsistences and one codebook contained two inconsistencies. The two researchers were able to reach agreement on the appropriate coding without need for moderation; the description of the code was updated to provide clarity. Previously coded data were double-checked by both researchers to ensure the updated description did not introduce further inconsistencies. After coding the full data set, HW and SK met again to compare analyses. There was disagreement with four responses, where one researcher applied a known code and the other used the category “vague”. To remove potential that researcher bias, experience or knowledge had influenced the data analysis of the four responses, the decision was made to defer to the judgment of the researcher who felt it was too vague to accurately code. ## 3.1. Demographics A total of 137 responses were recorded; of those, 90 were submitted as complete and eligible for inclusion. Participating students were all in their final year of study, were more likely to be female ($$n = 61$$; $68\%$), 23 years (interquartile range 2) and enrolled in a pharmacy program ($$n = 31$$; $34\%$) (Table 1). ## 3.2. Understanding of the Term ‘Health Literacy’ A total of 352 health literacy-related words or phrases were assigned to a category; the ten most frequently mentioned categories are shown in Figure 1 (the complete list is available in Supplementary material SB). Each student provided an average of four words or phrases, demonstrating they believe that health literacy relates to understanding health generally ($$n = 59$$; $17\%$), understanding ($$n = 26$$; $7\%$), understanding medical jargon ($$n = 22$$; $6\%$), and education ($$n = 21$$; $6\%$). ## 3.3. Three Signs Which Suggest Inadequate Patient Health Literacy When asked to name three signs which suggest inadequate patient health literacy, students provided 276 responses across 11 categories. The five most frequently provided categories are published in Figure 2 (the complete list is available in Supplementary material SC). The three most mentioned coded signs that could suggest inadequate patient health literacy, as described by students, were low comprehension ($$n = 33$$; $12\%$), poor adherence ($$n = 27$$; $10\%$); and low engagement with health professionals ($$n = 21$$; $8\%$). ## 3.4. Three Potential Consequences of Inadequate Patient Health Literacy Students were asked to describe three potential consequences of inadequate patient health literacy, resulting in 262 responses, across 11 categories. The five most frequently provided categories are shown in Figure 3 (the complete list is available in Supplementary material SD). Students described poor general health ($$n = 51$$; $19\%$), misunderstanding health advice ($$n = 21$$; $8\%$) and treatment failure ($$n = 19$$; $7\%$) to be consequences of inadequate patient health literacy. ## 3.5. Three Actions That Health and Social Care Professionals Can Take to Help a Patient with Inadequate Health Literacy When asked to name three actions that a health and social care professional can take to help a patient with inadequate health literacy, students provided 277 responses across ten categories. The three most frequently provided categories are published in Figure 4 (the complete list is available in Supplementary material SE). Students identified the need to make health information easily understood ($$n = 42$$; $15\%$), have health professional-provided support ($$n = 36$$; $13\%$) and provide basic patient education ($$n = 33$$; $12\%$). Of the 49 suggestions to provide written, visual or verbal education material, nine ($18\%$) mentioned ensuring the provided material was understandable to the patient. ## 3.6. Additional Comments Thirty students wrote additional comments about health literacy within their curriculum ($33\%$). Of these, two students ($7\%$) were satisfied with the current level of health literacy education (‘I think the classes on health literacy in my current degree have been very informative and helpful for us students to have a real understanding of it’). Twenty-eight comments ($93\%$) saw students identify gaps in their own understanding (‘now that I reflect on it I feel like we have under-prepared for this’). Fifteen comments ($50\%$) identified specific areas that students would appreciate more learning on (‘Most of our simulated patients have moderate health literacy, it would be good to have a wider range of health literacy levels. This will allow us to practice adapting our information gathering and counselling methods’). Four comments ($13\%$) linked students’ desire for increased health literacy content to a real-world context (‘*This is* a very important part of being a health practitioner especially in the context of multicultural Australia’). ## 4. Discussion To our knowledge, this is the first study worldwide that has explored student health and social care professionals’ factual and conceptual knowledge of health literacy, starting with their self-described definition of the term. It provides significant and interesting insights into students’ health literacy knowledge following the completion of all relevant curriculum components. The results suggest students had limited understanding of knowledge after completing the health literacy curriculum. As Bloom’s taxonomy identifies, the first fundamental component of a set of learning objectives is knowledge of topic-related definitions [28]. The majority of students did not describe three of the four key components of health literacy—how patients are able to access, appraise and use health information [1]. The fourth component—understanding—did not appear to be well understood, with many students believing it related to understanding general health conditions rather than understanding relevant health information. As understanding of the term ‘health literacy’ was poorly understood in our study cohort, it suggests that health literacy needs to be more explicitly taught in curricula. If students do not understand the definition of health literacy, they do not have the foundational knowledge and understanding necessary for developing their health literacy capabilities as health and social care professionals. Furthermore, we know that health and social care professionals are in a unique position to facilitate patients’ ability to access, understand, appraise and use health information [1,35,36,37]. If students are unaware that the scope of health literacy extends as broadly as it does, as practicing health and social care professionals, they will inadvertently miss many opportunities to have a positive impact on facilitating the full spectrum of patients’ health literacy that impact their long-term health outcomes. Once prompted with a definition of health literacy, students demonstrated some understanding of signs and behaviors of patients with, and potential consequences of, inadequate health literacy. Particularly, many students understood that patients with inadequate health literacy could display low comprehension, reduced engagement with a health professional and poor adherence, all of which are well-known signs of inadequate health literacy [1,4,5]. Similarly, students were aware of commonly-reported consequences of inadequate patient health literacy, such as treatment failure, patient misunderstanding and poorer health outcomes [4,5,6]. However, it was evident that they were unable to describe effective strategies to assist their identification of, and help patients with, inadequate health literacy. Students continued to focus their responses on the ‘understanding’ aspect of health literacy and were more likely to describe a ‘one-size-fits-all’ approach to patient education, where they would supply health information without considering other relevant patient factors, such as their ability to understand the message. Of the responses which addressed patient understanding, it was often suggested to measure understanding through closed questioning or creating a perception that they were open to answering patients’ questions if asked. This strategy is likely to be less effective as patients tend to avoid asking questions or saying they do not understand [1,38,39], a phenomenon more likely to be observed in groups who are at higher risk of having inadequate health literacy (such as older adults) [39]. This gap in students’ knowledge could be addressed by teaching core health literacy principles, such as the teach-back method, an effective way to gauge patient understanding, or the Ask Me 3 initiative, to empower patients to ask important questions designed to further their understanding [11,12,13,14]. From the additional comments, it is apparent that students recognize gaps in health literacy knowledge and, by extension, their own capabilities. A third of students revealed that they would welcome additional opportunities to learn not only the factual and conceptual knowledge around health literacy, but higher dimensions of Bloom’s taxonomy where they learn to apply new knowledge and skills with simulated and/or real patients. When considering the ‘knowledge’ component of published health literacy competencies for health professionals [27,30,31], the student health and social care professional participating in this exploratory research may not yet meet the requisite health literacy knowledge recommended by these publications. If gaps in knowledge do indeed exist, this could have implications for students’ ability to develop their health literacy capabilities in more complex learning components. ## 4.1. Strengths and Limitations The use of an anonymous, web-based questionnaire helped create an environment where students felt safe to provide authentic responses in their own words, rather than choosing or guessing from a list of pre-defined statements. The use of two independent researchers for data analysis provided confirmability to the results. As this was an exploratory study, data were collected from two universities in the same large metropolitan area of Western Australia. There was also unequal representation from each discipline within the study cohort. These two factors limit the transferability of the results; however, the inclusion of five separate health and social care disciplines lessens this impact. We were also unable to access the number of prospective participants enrolled in each degree, so we cannot provide insight into the response rate to the survey. We were unable to access details of the complete health literacy curriculum for all disciplines, so participant eligibility was partly guided by course directors who knew the health literacy curriculum throughout the degree. As we did not have access to specific education activities, we are unable to comment on the completeness of the current health literacy curricula. Nevertheless, whether gaps exist in the curriculum content or in students’ interpretation—or a mix of both—this study has highlighted that further investigation in this area is necessary. Finally, questionnaire responses were in the form of a word, phrase or short statement. The brevity of responses meant that data was not as rich as is typical for qualitative research, and contextual information or participant thoughts around the responses were missing. There is potential that participants provided a short-form response which did not reflect their deeper understanding, and future research should utilize a different methodology to overcome this potential limitation. ## 4.2. Future Research This research is the first step towards a deeper understanding of university students’ health literacy knowledge. Further research is recommended to gain insight into students’ health literacy capabilities within more advanced dimensions of Blooms’ revised taxonomy. Additionally, findings from this research can be used to inform both quantitative research, using a larger participant cohort from other universities and disciplines, and further qualitative research, exploring students’ reasoning behind each response. Insight from the recommended research will identify opportunities to improve upon the current health literacy-related education. ## 5. Conclusions Health and social care professionals play an important role in supporting patients’ health literacy development, but their usefulness is hindered without comprehensive health literacy capabilities. Major gaps were uncovered when exploring student health and social care professionals’ factual and conceptual health literacy knowledge; generally, students were unable to articulate key components of the definition of health literacy and were largely unable to suggest appropriate recommendations to assist patients with inadequate health literacy. This highlights an important extension to the National Statement on Health Literacy—although including health literacy education in health and social care professions curricula is essential, so too is a robust evaluation of the education’s effectiveness. Creating evidence-based and evaluated health literacy education for university will solidly pave the way for future health and social care professional graduates to have the increased capability to achieve better health outcomes for patients across the health literacy spectrum. ## References 1. **Health Literacy: Taking Action to Improve Safety and Quality** 2. Muscat D.M., Shepherd H.L., Nutbeam D., Trevena L., McCaffery K.J.. **Health literacy and shared decision-making: Exploring the relationship to enable meaningful patient engagement in healthcare**. *J. Gen. Intern. Med.* (2021.0) **36** 521-524. DOI: 10.1007/s11606-020-05912-0 3. Liu C., Wang D., Liu C., Jiang J., Wang X., Chen H., Ju X., Zhang X.. **What is the meaning of health literacy? A systematic review and qualitative synthesis**. *Fam. Med. Community Health* (2020.0) **8** e000351. DOI: 10.1136/fmch-2020-000351 4. Pirisi A.. **Low health literacy prevents equal access to care**. *Lancet* (2000.0) **356** 1828. DOI: 10.1016/S0140-6736(05)73297-9 5. Kindig D.A., Panzer A.M., Nielsen-Bohlman L.. *Health Literacy: A Prescription to End Confusion* (2004.0). DOI: 10.17226/10883 6. Baker D.W., Wolf M.S., Feinglass J., Thompson J.A., Gazmararian J.A., Huang J.. **Health literacy and mortality among elderly persons**. *Arch. Intern. Med.* (2007.0) **167** 1503-1509. DOI: 10.1001/archinte.167.14.1503 7. Sentell T., Braun K.L.. **Low health literacy, limited English proficiency, and health status in Asians, Latinos, and other racial/ethnic groups in California**. *J. Health Commun.* (2012.0) **17** 82-99. DOI: 10.1080/10810730.2012.712621 8. Baker D.W., Gazmararian J.A., Williams M.V., Scott T., Parker R.M., Green D., Ren J., Peel J.. **Functional health literacy and the risk of hospital admission among Medicare managed care enrollees**. *Am. J. Public Health* (2002.0) **92** 1278-1283. DOI: 10.2105/AJPH.92.8.1278 9. Schillinger D., Barton L.R., Karter A.J., Wang F., Adler N.. **Does literacy mediate the relationship between education and health outcomes? A study of a low-income population with diabetes**. *Public Health Rep.* (2006.0) **121** 245-254. DOI: 10.1177/003335490612100305 10. Stormacq C., Van den Broucke S., Wosinski J.. **Does health literacy mediate the relationship between socioeconomic status and health disparities? Integrative review**. *Health Promot. Int.* (2019.0) **34** e1-e17. DOI: 10.1093/heapro/day062 11. Centrella-Nigro A.M., Alexander C.. **Using the teach-back method in patient education to improve patient satisfaction**. *J. Contin. Educ. Nurs.* (2017.0) **48** 47-52. DOI: 10.3928/00220124-20170110-10 12. Yen P.H., Leasure A.R.. **Use and effectiveness of the teach-back method in patient education and health outcomes**. *Fed. Pract.* (2019.0) **36** 284-289. PMID: 31258322 13. Lord J., Borkowski N., Weech-Maldonado R.. **Patient Engagement in Home Health: The Role of Health Literacy and “Ask Me Three**. *Home Health Care Manag. Pract.* (2021.0) **33** 202-209. DOI: 10.1177/1084822321996623 14. Milidonis M., Keehan J., Montgomery K., Deuley R., Formoso S., Kopera-Frye K.. **Experience of Physical Therapy Students Mentoring Older Adults with Health Literacy Tools**. *Innov. Aging* (2021.0) **5** 821. DOI: 10.1093/geroni/igab046.3013 15. Pajaro A., Keldo A., Viray T.. **Effect of Ask3Teach3 on Patient Satisfaction on Medication Communication**. *J. Nurs. Pract. Appl. Rev. Res.* (2020.0) **12** 51-57. DOI: 10.13178/jnparr.2022.12.01.1208 16. Shersher V., Haines T.P., Sturgiss L., Weller C., Williams C.. **Definitions and use of the teach-back method in healthcare consultations with patients: A systematic review and thematic synthesis**. *Patient Educ. Couns.* (2021.0) **104** 118-129. DOI: 10.1016/j.pec.2020.07.026 17. Yeung D.L., Alvarez K.S., Quinones M.E., Clark C.A., Oliver G.H., Alvarez C.A., Jaiyeola A.O.. **Low–health literacy flashcards & mobile video reinforcement to improve medication adherence in patients on oral diabetes, heart failure, and hypertension medications**. *J. Am. Pharm. Assoc.* (2017.0) **57** 30-37. DOI: 10.1016/j.japh.2016.08.012 18. Goeman D., Conway S., Norman R., Morley J., Weerasuriya R., Osborne R.H., Beauchamp A.. **Optimising health literacy and access of service provision to community dwelling older people with diabetes receiving home nursing support**. *J. Diabetes Res.* (2016.0) **2016** 2483263. DOI: 10.1155/2016/2483263 19. Nickles D., Dolansky M., Marek J., Burke K.. **Nursing students use of teach-back to improve patients’ knowledge and satisfaction: A quality improvement project**. *J. Prof. Nurs.* (2020.0) **36** 70-76. DOI: 10.1016/j.profnurs.2019.08.005 20. Tothy A.S., Limper H.M., Driscoll J., Bittick N., Howell M.D.. **The ask me to explain campaign: A 90-day intervention to promote patient and family involvement in care in a pediatric emergency department**. *Jt. Comm. J. Qual. Patient Saf.* (2016.0) **42** 281. DOI: 10.1016/S1553-7250(16)42037-4 21. Peter D., Robinson P., Jordan M., Lawrence S., Casey K., Salas-Lopez D.. **Reducing readmissions using teach-back**. *J. Nurs. Adm.* (2015.0) **45** 35-42. DOI: 10.1097/NNA.0000000000000155 22. Daliri S., Hugtenburg J.G., Ter Riet G., van den Bemt B.J., Buurman B.M., Scholte op Reimer W.J., van Buul-Gast M.-C., Karapinar-Çarkit F.. **The effect of a pharmacy-led transitional care program on medication-related problems post-discharge: A before—After prospective study**. *PLoS ONE* (2019.0) **14**. DOI: 10.1371/journal.pone.0213593 23. Coleman C., Kurtz-Rossi S., McKinney J., Pleasant A., Rootman I., Shohet L.. *The Calgary Charter on Health Literacy* (2011.0) 24. **Australian Medical Association Position Statement: Health Literacy** 25. 25. World Health Organization Health Literacy Development for the Prevention and Control of Noncommunicable Diseases: Volume 3. Recommended ActionsWHOGeneva, Switzerland2022. *Health Literacy Development for the Prevention and Control of Noncommunicable Diseases: Volume 3. Recommended Actions* (2022.0) 26. Pearce K.L., Birbeck D., May E.. **The use of animations and the ‘teach-back’technique to facilitate an understanding of health literacy levels within the general community**. *Ergo* (2013.0) **3** 39-45 27. Saunders C., Palesy D., Lewis J.. **Systematic review and conceptual framework for health literacy training in health professions education**. *Health Prof. Educ.* (2019.0) **5** 13-29. DOI: 10.1016/j.hpe.2018.03.003 28. Krathwohl D.R.. **A revision of Bloom’s taxonomy: An overview**. *Physiother. Theory Pract.* (2002.0) **41** 212-218. DOI: 10.1207/s15430421tip4104_2 29. Braun V., Clarke V., Boulton E., Davey L., McEvoy C.. **The online survey as a qualitative research tool**. *Int. J. Soc. Res. Methodol.* (2021.0) **24** 641-654. DOI: 10.1080/13645579.2020.1805550 30. Coleman C.A., Hudson S., Maine L.L.. **Health literacy practices and educational competencies for health professionals: A consensus study**. *J. Health Commun.* (2013.0) **18** 82-102. DOI: 10.1080/10810730.2013.829538 31. Toronto C.. **Health Literacy Competencies for Registered Nurses: An e-Delphi Study**. *Doctoral Thesis* (2016.0) 32. Ratzan S., Parker R., Selden C., Zorn M.. *Health Literacy* (2000.0) 33. Braun V., Clarke V.. *Successful Qualitative Research: A Practical Guide for Beginners* (2013.0) 34. Kleinheksel A., Rockich-Winston N., Tawfik H., Wyatt T.R.. **Demystifying content analysis**. *Am. J. Pharm. Educ.* (2020.0) **84** 7113. DOI: 10.5688/ajpe7113 35. Skinner H., Biscope S., Poland B., Goldberg E.. **How adolescents use technology for health information: Implications for health professionals from focus group studies**. *J. Med. Internet Res.* (2003.0) **5** e20. DOI: 10.2196/jmir.5.4.e32 36. Scheier D.B.. **Barriers to health care for people with hearing loss: A review of the literature**. *J. N. Y. State Nurses Assoc.* (2009.0) **40** 4-11 37. Benjamin R.M.. **Medication adherence: Helping patients take their medicines as directed**. *Public Health Rep.* (2012.0) **127** 2-3. DOI: 10.1177/003335491212700102 38. Murtagh G., Furber L., Thomas A.L.. **Patient-initiated questions: How can doctors encourage them and improve the consultation process? A qualitative study**. *BMJ Open* (2013.0) **3** e003112. DOI: 10.1136/bmjopen-2013-003112 39. Meijers M.C., Potappel A., Kloek C., Hartman T.O., Spreeuwenberg P., van Dulmen S., Noordman J.. **Shifts in patients’ question-asking behaviour between 2007 and 2016: An observational study of video-recorded general practice consultations**. *Patient Educ. Couns.* (2020.0) **103** 1168-1175. DOI: 10.1016/j.pec.2020.01.016
--- title: 'Hand Dexterity Is Associated with the Ability to Resolve Perceptual and Cognitive Interference in Older Adults: Pilot Study' authors: - Marie Schwalbe - Skye Satz - Rachel Miceli - Hang Hu - Anna Manelis journal: Geriatrics year: 2023 pmcid: PMC10037645 doi: 10.3390/geriatrics8020031 license: CC BY 4.0 --- # Hand Dexterity Is Associated with the Ability to Resolve Perceptual and Cognitive Interference in Older Adults: Pilot Study ## Abstract The relationship between hand dexterity and inhibitory control across the lifespan is underexplored. In this pilot study, we examined inhibitory control using a modified Simon task. During the task, participants were presented with right- and left-pointing arrows located either on the right or the left parts of the screen. In the congruent trials, the arrow location and direction matched. In the incongruent trials, they mismatched, thus creating cognitive interference. In $50\%$ of trials, the arrow presentation was accompanied by a task-irrelevant but environmentally meaningful sound that created perceptual interference. Hand dexterity was measured with the 9-hole peg test. Significantly faster reaction time (RT) on the modified Simon task ($p \leq 0.001$) was observed in younger adults, trials with concurrent sound stimuli, and congruent trials. Older adults who reported recent falls had greater difficulty resolving cognitive interference than older adults without recent falls. Hand dexterity significantly moderated the effect of sound on RT, but only in the group of older individuals. Interestingly, older individuals with reduced hand dexterity benefited from concurrent sounds more than those with better hand dexterity. Our findings suggest that task-irrelevant but environmentally meaningful sounds may increase alertness and enhance stimulus perception and recognition, thus improving motor performance in older individuals. ## 1. Introduction Many daily activities, including driving, shopping, and even social interaction, rely on people’s ability to ignore distracting and misleading stimuli to respond effectively to a situation. Inhibitory control is an executive function that helps allocate attention to goal-relevant stimuli while inhibiting extraneous (goal-irrelevant) information coming in the form of perceptual or cognitive interference. Both perceptual interference, such as irrelevant noises or images [1], and cognitive interference, such as misleading information [2], can distract people’s attention from their intended task and worsen task performance. The ability to direct attention to the relevant information and ignore this interference is termed interference resolution, and it is necessary for inhibitory control [3]. Normal aging is associated with worsening inhibitory control, which is demonstrated by the declining performance with age on tasks requiring interference resolution [4]. In addition to playing an important role for cognitive task performance, inhibitory control is critical for effective motor function, such as hand dexterity and strength, where it is necessary to ignore extraneous information [5]. A deeper understanding of the relationship between interference resolution and motor task performance could guide future diagnostics and therapeutic interventions for neurological and movement disorders. Precision of hand movements is determined by hand dexterity. Reduced hand dexterity is often observed in older individuals and is associated with difficulty performing activities of daily living, decreased quality of life, and even depression [6]. Hand dexterity can be compromised in certain neurological conditions, such as multiple sclerosis [6], stroke [7,8], diabetic peripheral neuropathy [9], Parkinson’s disease [10], and Alzheimer’s disease [11]. A small number of studies have shown that hand dexterity is directly related to cognitive function [12,13,14]. For example, better executive function and attentional control were associated with better hand dexterity in healthy community-dwelling older adults [13,14]. Furthermore, better attentional control was associated with better motor dexterity among stroke survivors and patients discharged from the ICU [15]. While these findings are consistent with the idea that decreased motor function in older individuals is associated with underlying reduced cognitive function, the age-related changes in cognitive and perceptual interference resolution, as well as the relationship between interference resolution and hand dexterity, remain underexplored. The current pilot study aims to fill these gaps and explore interference resolution in older and younger individuals using the modified Simon task [16]. During the modified Simon task, participants were presented with right- and left-pointing arrows either in the right or left part of the screen (cognitive interference). A total of $50\%$ of the trials were accompanied by an everyday sound stimulus that was irrelevant to the task but environmentally meaningful, such as birds chirping or thunder (perceptual interference). The traditional version of the Simon task presents participants with congruent and incongruent either visual or auditory stimuli. Our modified Simon task revisits the effects of visual cognitive interference and explores how interference resolution is affected by task-unrelated but environmentally plausible auditory stimuli. This pilot project will allow us to examine the study protocol and the validity of the modified task by replicating the congruency effects observed in previous studies [16,17,18]. We hypothesized that older individuals would be slower and less accurate than younger individuals and that participants would be slower and less accurate on incongruent trials, especially those with concurrent perceptual interference. In addition, we hypothesized that the individuals with reduced hand dexterity would show greater worsening of reaction time (RT) and accuracy on the interference resolution vs. the no interference resolution trials. This effect would be especially pronounced in older, compared to younger, individuals on the trials that required resolution of both cognitive and perceptual interference. Our alternative hypothesis was that concurrent sound presentation would improve task performance independently of hand dexterity. This latter hypothesis was based on the “pip and pop” effect, which suggests that concurrent perceptual stimuli presentation, such as visual or auditory stimuli, may improve task performance [19,20]. ## 2.1. Participants The study was approved by the University of Pittsburgh Institutional Review Board (IRB number STUDY20120048). Written informed consent was obtained from all participants. Eighty-nine participants between 18–85 years of age were recruited from the community, online Pitt + Me and Pepper (IRB number STUDY19090270) registries, as well as from the ongoing NIH-funded study (R01MH114870). Potential participants either called or emailed the research team, expressing their interest in the study, or they were approached by the study team after the person was identified as potentially eligible through the registry. Participants were fluent in English and had premorbid IQ > 85 per the National Adult Reading Test [21]. Exclusion criteria included a history of head injury, neurodevelopmental and neurological disorders, learning disability, and psychiatric disorders other than depressive and anxiety disorders. We excluded from the analyses the individuals who were diagnosed with autism spectrum disorder, bipolar disorder, and schizophrenia ($$n = 3$$), and those who were missing the hand dexterity data ($$n = 4$$), were left-handed or ambidextrous ($$n = 7$$), or did not understand the task instructions ($$n = 2$$). This left 73 individuals in the analyses. Our sample included only right-handed participants to avoid the potential bias in right hand vs. left hand responses when the right hand is dominant for some individuals, but non-dominant for others [22,23,24]. ## 2.2. Study Procedures Participants completed intake interviews and self-report questionnaires, cognitive and neurological assessments, a hand dexterity test, and a computerized modified Simon task during an in-person office visit. Participants were paid for participation. ## 2.2.1. Demographics, Cognitive and Neurological Assessments Intake interviews and self-reports involved the collection of information about general demographics, health, and current medications. A basic neurological examination of cranial nerves, gait, posture, balance, and sensation was administered by a trained team member to screen for possible neurological deficits. After that, the Montreal Cognitive Assessment (MoCA) [25] was used to assess general cognitive functioning across the core domains of cognition. Visual acuity and age-related macular degeneration (AMD) were assessed with the Snellen test and the Amsler grid accordingly. Participants also reported fall history for the past year. ## 2.2.2. Hand Dexterity Assessment The gold standard measurement of hand dexterity is the 9-hole peg test (NHPT) [26,27], which can be used across a wide range of ages and medical conditions [28,29,30]. In this task, the plastic peg board was placed directly in front of the participant, with the dish holding the pegs placed closest to the hand being tested. The participant was instructed to use one hand to take the pegs from the dish, one at a time, and place them into each of the 9 holes on the board as quickly as possible. As soon as all holes were filled, participants removed the pegs one at a time and placed them back into the dish. Participants could fill and empty the holes in any order they choose. The hand that was not being evaluated could be used to steady the board. Performance on the NHPTwas measured using a stopwatch to record the amount of time taken to complete the task in seconds. Recordings were started when the participant touched the first peg and were stopped when the last peg entered the dish. The participants were given an opportunity to practice this task once before the timed trial. ## 2.3. Modified Simon Task The Simon task [16] was modified to measure participants’ ability to resolve cognitive, perceptual, and combined interference (Figure 1). During this task, participants were shown an image of an arrow pointing either to the left or to the right. They were instructed to indicate which direction the arrow was pointing by pressing the “Z” key on a standard QWERTY keyboard with their left index finger for left-pointing arrows and the “M” key with their right index finger for right-pointing arrows as quickly and accurately as possible. The task consisted of 2 runs of 32 trials, whose duration was randomly determined and varied from 6500–10,000 ms in 500 ms increments. Each trial consisted of the screen with a fixation star, the stimulus screen, and an inter-trial interval (Figure 1). The fixation star duration was randomly determined for each trial and was between 500–2000 ms. A stimulus presentation was equal to the participant’s reaction time (RT), but was not longer than 2500 ms. The trials were separated with inter-trial intervals (it is), during which participants were shown a “Please Rest” screen. The ITI duration varied from trial to trial and was equal to the time necessary to complete the duration of the trial. The stimuli were white arrows presented on a black background to the right or to the left of a fixation cross. Of 64 trials, $50\%$ were congruent, and $50\%$ were incongruent. Stimulus congruency refers to the relationship between the arrow location and its direction. In the congruent trials, the arrow location and direction matched (left-pointing arrow presented on the left side of the screen), while in the incongruent trials, the arrow location and direction mismatched (left-pointing arrow on the right side of the screen), thus creating a source of cognitive interference. A concurrent auditory stimulus that could be a natural (birds chirping or a thunderstorm) or a man-made (construction or sirens) sound accompanied $50\%$ of congruent and $50\%$ of incongruent trials. We believe that presenting unrelated auditory stimuli during the task created perceptual interference similar to that which people experience in their everyday lives. The sounds were found on the Internet and modified to have equal loudness across all sound clips. Each sound clip started simultaneously with the onset of the visual stimulus and ended when a participant responded in the trial. Prior to task completion, sounds were tested with each participant to ensure audibility. Volume was adjusted accordingly based on individual needs. Each run had an equal number of congruent/incongruent and sound/no sound trials. The order of the congruent/incongruent trials with and without sound, as well as the trial and fixation durations, was randomized for each participant to avoid systematic bias. ## 2.4. Data Analyses All statistical analyses were conducted using R (https://www.r-project.org (accessed on 24 January 2023)). ## 2.4.1. Demographic and Hand Dexterity Data Analyses We calculated the means and standard deviations for the demographic and hand dexterity data across all participants. Considering the wide age range in the study, we median-split the sample (median age = 58.9 years) into two groups. The individuals whose age was below the median comprised a group of younger adults ($$n = 36$$), while those who were at or above the median age comprised the group of older adults ($$n = 37$$). Demographic, clinical, and hand dexterity variables were then compared between the younger and older adult groups using chi-square and t-tests. ## 2.4.2. Modified Simon Task Data Analyses To understand the effects of age and hand dexterity, in addition to the effects of congruency and sound, on RT and accuracy, we utilized mixed-effects models using the lme4 package in R [31]. Two models were examined: a congruency-by-sound-by-age group interaction model and a congruency-by-sound-by-age group-by hand dexterity interaction model. A significance level $p \leq 0.05$ was utilized for all statistical analyses, except for the analyses that included hand dexterity. Those analyses were conducted for both hands, so the significance level was adjusted to $p \leq 0.025$ to reflect Bonferroni correction ($\frac{0.05}{2}$ = 0.025). In all models, participants were treated as a random factor, and participants’ IQ was used as a covariate. The contrasts and means were estimated from the mixed-effects models using the ‘modebased’ package in R [32]. The p-values, ANOVA, and summary tables were produced using the lmerTest package [33]. RT analysis used only trials with accurate responses. Before entering RT values in the analysis, they were examined for outliers. The values that were outside the 3 IQRs (interquartile ranges) from the first or third quartile were considered as outliers and were excluded from the analyses. RT was fitted using linear mixed-effects models (the lmer function). Accuracy was analyzed as a binomial variable (1 for correct responses, 0 for incorrect responses). It was analyzed using generalized linear mixed-effects models (the glmer function for binomial data), and a Wald’s chi-square and p-values were reported. ## 3.1. Demographic and Clinical The demographics and clinical characteristics of the participants are reported in Table 1. The groups of younger and older participants did not differ in terms of sex, general cognitive function per MoCA, or neurological health. As expected, older adults were significantly older than younger adults ($p \leq 0.001$). Older adults had higher IQ ($p \leq 0.05$), but slower performance on the hand dexterity task in both the dominant and non-dominant hand ($p \leq 0.01$) compared to younger adults. Across all participants, the dominant hand dexterity responses were faster than those in the non-dominant hand (t[72] = −6.9, $p \leq 0.001$). In the group of older, but not younger, participants, 16 individuals ($43\%$) were diagnosed with AMD, and 7 ($19\%$) reported a history of falls during the past 12 months. ## 3.2.1. The Effect of Task Condition and Age Group on RT and Accuracy in the Modified Simon Task Based on the IQR analysis of RT, $1.2\%$ of responses were outliers and were excluded from the analyses. The mixed-effects analysis of the congruency-by-sound-by-age group interaction effect revealed a significant congruency-by-age group interaction [F(1, 4370.1) = 12.7, $p \leq 0.05$], as well as the main effects of congruency [F(1, 4376.6) = 273.9, $p \leq 0.001$], sound [F(1, 4376.1) = 24.9, $p \leq 0.001$] and age group [F[1, 70] = 33.5, $p \leq 0.001$] on RT (Figure 2). Older adults were slower than younger ones ($z = 5.8$, $p \leq 0.001$). Participants were slower on trials without sound than those with sound ($z = 4.99$, $p \leq 0.001$), and they were slower on incongruent trials than congruent trials ($z = 16.55$, $p \leq 0.001$). The analysis of accuracy revealed significant main effects of age group (chi2 = 4.3, $p \leq 0.05$) and congruency (chi2 = 28.9, $p \leq 0.001$). Older participants had lower accuracy than younger participants (z = −2.0, $p \leq 0.05$). The accuracy for congruent trials was higher than that for the incongruent trials ($z = 5.4$, $p \leq 0.001$). Table 2 reports the mean RT and accuracy values that were estimated from the mixed-effects models. ## 3.2.2. The Effect of Task Condition, Hand Dexterity, and Age Group on RT in the Modified Simon Task The mixed-effects analysis for the dominant (right) hand dexterity revealed a significant sound-by-right hand and dexterity-by-age group interaction effect [F(1, 4370.1) = 12.7, $p \leq 0.001$] that survived the Bonferroni correction for 2 tests. In addition, there were significant interaction effects between sound and right hand dexterity [F(1, 4370.1) = 4.5, $$p \leq 0.033$$], sound and age group [F(1, 4370.1) = 14.4, $p \leq 0.001$], and sound and congruency [F(1, 4370.2) = 4.4, $$p \leq 0.036$$], and a main effect of congruency [F(1, 4370.6) = 4.9, $$p \leq 0.027$$]. Of these interaction effects, only sound-by age group interaction survived Bonferroni correction. Further exploration of these effects in each age group revealed a main effect of congruency [F(1, 2197.1) = 6.9, $p \leq 0.01$], with faster responses for congruent vs. incongruent trials [model estimated t(2197.09) = −11.74, $p \leq 0.001$], but no effects of sound or hand dexterity in the younger individuals. In contrast, in older adults, there was a significant sound-by-right hand dexterity interaction effect [F(1, 2173.04) = 15.5, $p \leq 0.001$], suggesting that individuals with reduced hand dexterity benefited from concurrent sound presentation more than those with faster hand dexterity (Figure 3). There was also a main effect of sound [F(1, 2173.05) = 12.0, $p \leq 0.001$], with faster responses on the trials with sound compared to the trials without sound [model estimated t(2173.12) = 2.71, $p \leq 0.01$]. The mixed-effects analysis for the non-dominant (left) hand dexterity revealed a significant sound and age group interaction [F(1, 4370.1) = 4.26, $$p \leq 0.04$$] that, however, did not survive Bonferroni correction. Further exploration of these effects in each age group revealed no significant interactions or main effects in younger individuals. In older adults, there was a significant sound-by-left hand dexterity interaction effect [F(1, 2173.03) = 6.9, $p \leq 0.01$], suggesting that individuals with reduced hand dexterity benefited from concurrent sound presentation more than those with faster hand dexterity (Figure 3). There was also a main effect of sound [F(1, 2173.05) = 4.5, $$p \leq 0.033$$], with faster responses on the trials with sound compared to the trials without sound [model estimated t(2173.12) = 2.72, $p \leq 0.01$]. ## 3.2.3. The Effect of Task Condition, Hand Dexterity, and Age Group on Accuracy in the Modified Simon Task Performance accuracy was above $95\%$ in all conditions and groups (Table 2). The mixed-effects analysis for the dominant (right) hand dexterity revealed a significant congruency-by-right hand dexterity interaction effect [chi2[1] = 4.1, $$p \leq 0.043$$] on the accuracy, with greater differences between accuracy for congruent and incongruent trials observed in those with reduced dominant hand dexterity. This effect did not survive Bonferroni correction, however. The mixed-effects analysis for the non-dominant (left) hand dexterity revealed no significant main or interaction effects of left hand dexterity with the other variables. ## 3.2.4. Exploratory Analyses in the Group of Older Individuals We examined the effects of AMD status (diagnosed vs. not diagnosed with AMD) and recent falls (reported falling vs. not falling during the past 12 months) on RT in the group of older adults. The mixed-effects model that examined the sound-by-congruency-by-AMD status interaction revealed no significant main effects of AMD or any interaction with AMD status on RT in the modified Simon task. The mixed-effects model that examined the sound-by-congruency-by-recent falls status interaction revealed a significant congruency-by-falls interaction effect [F[1, 4370] = 4.26, $p \leq 0.05$]. The difference between the congruent and incongruent conditions was greater for those older adults who reported recent falls vs. those who did not report falling during the past 12 months (no falls: incongruent-congruent difference = 65 msec (SE = 7.5 msec); falls: incongruent-congruent difference = 150.2 (SE = 15.7 msec); Figure 4). ## 4. Discussion In this study, we used a modified Simon task to investigate the effects of cognitive and perceptual interference on the task’s RT and accuracy in neurologically and cognitively normal older versus younger adults. Consistent with previous studies [16,34,35,36], participants’ responses were slower and less accurate on the incongruent, compared to congruent, trials, thus supporting the validity of the modified Simon task. Older adults’ responses were slower and less accurate compared to those of younger adults, reflecting an age-related decline in task performance [4]. Considering that the accuracy in the modified Simon test was very high (almost at ceiling), we will not further discuss the accuracy findings due to their low clinical relevance. Inconsistent with our predictions that perceptual interference would worsen RT, we found that irrelevant but environmentally meaningful sounds presented concurrently with visual stimuli improved RT in both congruent and incongruent trials across all participants. These findings are consistent with the “pip and pop” effect, suggesting that nonspatial, task-irrelevant auditory stimuli may improve performance on visual tasks, including those requiring visual search [19,20]. Although previous studies suggested that the effect of meaningless background noise on cognitive performance depends on the intensity, duration, timing, and type of noise, along with the subjects’ age and the task load [37,38,39], we found benefits of sound in both older and younger adult age groups, as well as in both congruent (easier) and incongruent (more difficult) trials. Our findings may be explained by the nature of the sounds in our study. Even though the sounds were irrelevant to the task at hand, they were environmentally meaningful in that they carried important information about the surrounding environment. For example, if a person hears a screeching sound of a car brake, they should quickly orient their attention toward this sound (even if they are engaged in some other activity) to evaluate the level of danger and produce a quick response if needed. The other goal of this study was to understand how hand dexterity is related to participants’ ability to resolve interference. Consistent with previous studies, reduced hand dexterity was observed in older, compared to younger, individuals [13,40] and for the non-dominant, compared to the dominant, hand [41]. Despite the latter differences, we found that hand dexterity in both dominant and non-dominant hands moderated the effect of sound on RT in the modified Simon task, but only in the group of older individuals. Interestingly, older individuals with reduced hand dexterity benefited from the sound administration more than those with faster dexterity, suggesting a possible implication of these results to improve function in people with movement disorders, such as Parkinson’s disease or multiple sclerosis. Reduced hand dexterity in older adults may be explained by the age-related reduction in strength and muscle mass [42], changes in connective tissues, decrease in number of motor neurons, conditions such as osteoarthritis or rheumatoid arthritis [43,44,45], and a decline in attentional control [5]. Administering the task-irrelevant but environmentally meaningful sounds concurrently with the visual stimuli may increase alertness in older individuals and thus improve motor performance. Given the link between attentional control and motor control [5], it is possible that changes in attention, such as those induced by sound stimuli, are more prominent in individuals with reduced hand dexterity and in individuals with reduced attention (e.g., older adults). Notably, a similar sound effect was observed in children with and without Attention Deficit Hyperactivity Disorder, whose attentional performance was facilitated by administration of task-irrelevant sounds [46]. Lastly, considering that sounds improve visual perception of stimuli [47,48], the sound-related RT improvement in older adults with reduced dexterity could be due to enhanced stimulus perception and faster recognition of whether the arrow was pointing to the right or to the left. The exploratory analyses conducted in the group of older adults showed that those individuals who reported falls within the past 12 months had greater difficulty resolving cognitive interference (i.e., had a greater difference in RT between congruent and incongruent trials) than those individuals who did not report recent falls. With more than 3 million older adults receiving treatment for fall injuries annually [49], understanding cognitive processes that might be associated with increased risk of fall is critically important. Traditionally, it is thought that falls in older adults are associated with the use of various medications, including psychotropics [50], and worsening of balance and gait patterns [51], as well as visual and cognitive decline [52]. One clinical implication of out pilot study (albeit in the small sample) is the improved understanding that a reduced ability to resolve cognitive interference may be another risk factor for falls, even in neurologically and cognitively normal older adults. These results warrant further prospective examination of the relationship between falls in older adults and their ability to resolve cognitive and perceptual interference (e.g., the person fell because they talked to a friend and had not noticed an obstacle in the hallway). With regard to visual disturbance, AMD is one of the most prevalent causes of visual impairment and affects approximately 19.8 million Americans aged 40 and older [53]. In our pilot study, we compared a small sample of older adults with AMD vs. those without. We found no significant differences in RT or accuracy between these groups in either task condition. While the small sample size did not allow us to derive a strong conclusion regarding the effects of AMD on task performance, the findings suggest that performance on cognitive tasks is not necessarily affected by an AMD diagnosis if it is not debilitating and if older adults are neurologically and cognitively healthy. This pilot study’s limitations include the small sample size and cross-sectional approach. Further, all our participants were neurologically healthy, which did not allow us to extrapolate whether the sound administration would be beneficial for older individuals with neurological and movement disorders. Future research should examine the effect of sound on motor response and cognitive/perceptual interference resolution in patients with various neurological disorders (e.g., Alzheimer’s disease, Parkinson’s disease, stroke, and multiple sclerosis), as well as cognitive impairments. For example, approximately $60\%$ of multiple sclerosis patients report impaired hand function in the first year after diagnosis, which makes performance of activities of daily living more difficult [54]. Future studies should examine whether administering task-irrelevant but environmentally meaningful sounds may improve hand function in affected individuals by altering the level of alertness. Neuroimaging studies of the sound effect in older adults may help uncover the neural mechanisms underlying the observed behavior change. ## 5. Conclusions In conclusion, our findings demonstrated that sound presentation improved performance on the modified Simon task independently of cognitive interference across all participants. The effect of sound on RT was moderated by participants’ hand dexterity. However, this effect was only observed in older individuals, of whom participants with reduced hand dexterity benefited from concurrent sound presentation more than their peers with faster hand dexterity. Taken together, these findings warrant further exploration of the role of environmental sounds on cognitive tasks performance as a therapeutic intervention that can be administered along with other treatments for improving hand function, gait, and fall prevention in older adults [55,56]. ## References 1. Wais P.E., Gazzaley A.. **The Impact of Auditory Distraction on Retrieval of Visual Memories**. *Psychon. Bull. Rev.* (2011.0) **18** 1090-1097. DOI: 10.3758/s13423-011-0169-7 2. MacLeod C.M.. **Half a Century of Reseach on the Stroop Effect: An Integrative Review**. *Psychol. Bull.* (1991.0) **109** 163-203. DOI: 10.1037/0033-2909.109.2.163 3. Bessette K.L., Karstens A.J., Crane N.A., Peters A.T., Stange J.P., Elverman K.H., Morimoto S.S., Weisenbach S.L., Langenecker S.A.. **A Lifespan Model of Interference Resolution and Inhibitory Control: Risk for Depression and Changes with Illness Progression**. *Neuropsychol. Rev.* (2020.0) **30** 477-498. DOI: 10.1007/s11065-019-09424-5 4. Collette F., Schmidt C., Scherrer C., Adam S., Salmon E.. **Specificity of Inhibitory Deficits in Normal Aging and Alzheimer’s Disease**. *Neurobiol. Aging* (2009.0) **30** 875-889. DOI: 10.1016/j.neurobiolaging.2007.09.007 5. Rinne P., Hassan M., Fernandes C., Han E., Hennessy E., Waldman A., Sharma P., Soto D., Leech R., Malhotra P.A.. **Motor Dexterity and Strength Depend upon Integrity of the Attention-Control System**. *Proc. Natl. Acad. Sci. USA* (2018.0) **115** E536-E545. DOI: 10.1073/pnas.1715617115 6. Alonso R.N., Eizaguirre M.B., Cohen L., Quarracino C., Silva B., Pita M.C., Yastremiz C., Vanotti S., Garcea O.. **Upper Limb Dexterity in Patients with Multiple Sclerosis**. *Int. J. MS Care* (2021.0) **23** 79-84. DOI: 10.7224/1537-2073.2019-083 7. Nowak D.A., Grefkes C., Dafotakis M., Küst J., Karbe H., Fink G.R.. **Dexterity Is Impaired at Both Hands Following Unilateral Subcortical Middle Cerebral Artery Stroke**. *Eur. J. Neurosci.* (2007.0) **25** 3173-3184. DOI: 10.1111/j.1460-9568.2007.05551.x 8. Sunderland A., Bowers M.P., Sluman S.M., Wilcock D.J., Ardron M.E.. **Impaired Dexterity of the Ipsilateral Hand after Stroke and the Relationship to Cognitive Deficit**. *Stroke* (1999.0) **30** 949-955. DOI: 10.1161/01.STR.30.5.949 9. de Almeida Lima K.C., Borges L.. **da S.; Hatanaka, E.; Rolim, L.C.; de Freitas, P.B. Grip Force Control and Hand Dexterity Are Impaired in Individuals with Diabetic Peripheral Neuropathy**. *Neurosci. Lett.* (2017.0) **659** 54-59. DOI: 10.1016/j.neulet.2017.08.071 10. Proud E.L., Morris M.E.. **Skilled Hand Dexterity in Parkinson’s Disease: Effects of Adding a Concurrent Task**. *Arch. Phys. Med. Rehabil.* (2010.0) **91** 794-799. DOI: 10.1016/j.apmr.2010.01.008 11. de Paula J.J., Albuquerque M.R., Lage G.M., Bicalho M.A., Romano-Silva M.A., Malloy-Diniz L.F.. **Impairment of Fine Motor Dexterity in Mild Cognitive Impairment and Alzheimer’s Disease Dementia: Association with Activities of Daily Living**. *Rev. Bras. Psiquiatr.* (2016.0) **38** 235-238. DOI: 10.1590/1516-4446-2015-1874 12. Abe T., Soma Y., Kitano N., Jindo T., Sato A., Tsunoda K., Tsuji T., Okura T.. **Change in Hand Dexterity and Habitual Gait Speed Reflects Cognitive Decline over Time in Healthy Older Adults: A Longitudinal Study**. *J. Phys. Ther. Sci.* (2017.0) **29** 1737-1741. DOI: 10.1589/jpts.29.1737 13. Walters B.H., Huddleston W.E., O’Connor K., Wang J., Bement M.H., Keenan K.G.. **The Role of Eye Movements, Attention, and Hand Movements on Age-Related Differences in Pegboard Tests**. *J. Neurophysiol.* (2021.0) **126** 1710-1722. DOI: 10.1152/jn.00629.2020 14. Kobayashi-Cuya K.E., Sakurai R., Suzuki H., Ogawa S., Takebayashi T., Fujiwara Y.. **Observational Evidence of the Association between Handgrip Strength, Hand Dexterity, and Cognitive Performance in Community-Dwelling Older Adults: A Systematic Review**. *J. Epidemiol.* (2018.0) **28** 373-381. DOI: 10.2188/jea.JE20170041 15. Elías M.N., Munro C.L., Liang Z.. **Executive Function, Dexterity, and Discharge Disposition in Older Intensive Care Unit Survivors**. *Am. J. Crit. Care* (2020.0) **29** 484-488. DOI: 10.4037/ajcc2020132 16. Simon J.R., Rudell A.P.. **Auditory S-R Compatibility: The Effect of an Irrelevant Cue on Information Processing**. *J. Appl. Psychol.* (1967.0) **51** 300-304. DOI: 10.1037/h0020586 17. Kawai N., Kubo-Kawai N., Kubo K., Terazawa T., Masataka N.. **Distinct Aging Effects for Two Types of Inhibition in Older Adults: A near-Infrared Spectroscopy Study on the Simon Task and the Flanker Task**. *Neuroreport* (2012.0) **23** 819-824. DOI: 10.1097/WNR.0b013e3283578032 18. Proctor R.W., Pick D.F., Vu K.P.L., Anderson R.E.. **The Enhanced Simon Effect for Older Adults Is Reduced When the Irrelevant Location Information Is Conveyed by an Accessory Stimulus**. *Acta Psychol.* (2005.0) **119** 21-40. DOI: 10.1016/j.actpsy.2004.10.014 19. Van der Burg E., Olivers C.N.L., Bronkhorst A.W., Theeuwes J.. **Pip and Pop: Nonspatial Auditory Signals Improve Spatial Visual Search**. *J. Exp. Psychol. Hum. Percept. Perform.* (2008.0) **34** 1053-1065. DOI: 10.1037/0096-1523.34.5.1053 20. Chamberland C., Hodgetts H.M., Vallières B.R., Vachon F., Tremblay S.. **Pip and Pop: When Auditory Alarms Facilitate Visual Change Detection in Dynamic Settings**. *Proceedings of the Human Factors and Ergonomics Society* 284-288 21. Nelson H.E.. **National Adult Reading Test (NART): Test Manual**. *Wind. UK NFER-Nelson* (1982.0) **124** 1-25 22. Andersen K.W., Siebner H.R.. **Mapping Dexterity and Handedness: Recent Insights and Future Challenges**. *Curr. Opin. Behav. Sci.* (2018.0) **20** 123-129. DOI: 10.1016/j.cobeha.2017.12.020 23. Amunts K., Schlaug G., Schleicher A., Steinmetz H., Dabringhaus A., Roland P.E., Zilles K.. **Asymmetry in the Human Motor Cortex and Handedness**. *Neuroimage* (1996.0) **4** 216-222. DOI: 10.1006/nimg.1996.0073 24. Dassonville P., Zhu X.H., Uǧurbil K., Kim S.G., Ashe J.. **Functional Activation in Motor Cortex Reflects the Direction and the Degree of Handedness**. *Proc. Natl. Acad. Sci. USA* (1997.0) **94** 14015-14018. DOI: 10.1073/pnas.94.25.14015 25. Nasreddine Z.S., Phillips N.A., Bédirian V., Charbonneau S., Whitehead V., Collin I., Cummings J.L., Chertkow H.. **The Montreal Cognitive Assessment, MoCA: A Brief Screening Tool for Mild Cognitive Impairment**. *J. Am. Geriatr. Soc.* (2005.0) **53** 695-699. DOI: 10.1111/j.1532-5415.2005.53221.x 26. Kellor M., Frost J., Silberberg N., Iversen I., Cummings R.. **Hand Strength and Dexterity**. *Am. J. Occup. Ther.* (1971.0) **25** 77-83. PMID: 5551515 27. Mathiowetz V., Weber K., Kashman N., Volland G.. **Adult Norms for the Nine Hole Peg Test of Finger Dexterity**. *Occup. Ther. J. Res.* (1985.0) **5** 24-38. DOI: 10.1177/153944928500500102 28. Wang Y.C., Magasi S.R., Bohannon R.W., Reuben D.B., McCreath H.E., Bubela D.J., Gershon R.C., Rymer W.Z.. **Assessing Dexterity Function: A Comparison of Two Alternatives for the NIH Toolbox**. *J. Hand Ther.* (2011.0) **24** 313-321. DOI: 10.1016/j.jht.2011.05.001 29. Mendoza-Sánchez S., Molina-Rueda F., Florencio L.L., Carratalá-Tejada M., Cuesta-Gómez A.. **Reliability and Agreement of the Nine Hole Peg Test in Patients with Unilateral Spastic Cerebral Palsy**. *Eur. J. Pediatr.* (2022.0) **181** 2283-2290. DOI: 10.1007/s00431-022-04423-w 30. Johansson G.M., Häger C.K.. **A Modified Standardized Nine Hole Peg Test for Valid and Reliable Kinematic Assessment of Dexterity Post-Stroke**. *J. Neuroeng. Rehabil.* (2019.0) **16** 8. DOI: 10.1186/s12984-019-0479-y 31. Bates D., Mächler M., Bolker B.M., Walker S.C.. **Fitting Linear Mixed-Effects Models Using Lme4**. *J. Stat. Softw.* (2015.0) **67** 48. DOI: 10.18637/jss.v067.i01 32. Makowski D., Ben-Shachar M.S., Patil I., Lüdecke D.. **Estimation of Model-Based Predictions, Contrasts and Means** 33. Kuznetsova A., Brockhoff P.B., Christensen R.H.B.. **LmerTest Package: Tests in Linear Mixed Effects Models**. *J. Stat. Softw.* (2017.0) **82** 1-26. DOI: 10.18637/jss.v082.i13 34. Mosso M., Freudenberg A., McCracken K., McGivern R.F.. **Sex Differences in Implicit Processing of Allocentric Relationships between Objects and Location in a Simon Task**. *PLoS ONE* (2020.0) **15**. DOI: 10.1371/journal.pone.0235964 35. Proctor R.W., Vu K.P.L.. **Mixing Location-Irrelevant and Location-Relevant Trials: Influence of Stimulus Mode on Spatial Compatibility Effects**. *Mem. Cogn.* (2002.0) **30** 281-293. DOI: 10.3758/BF03195289 36. Aisenberg D., Henik A.. **Stop Being Neutral: Simon Takes Control!**. *Q. J. Exp. Psychol.* (2012.0) **65** 295-304. DOI: 10.1080/17470218.2010.507819 37. Herweg N.A., Bunzeck N.. **Differential Effects of White Noise in Cognitive and Perceptual Tasks**. *Front. Psychol.* (2015.0) **6** 1639. DOI: 10.3389/fpsyg.2015.01639 38. Smith A.. **A Review of the Effects of Noise on Human Performance**. *Scand. J. Psychol.* (1989.0) **30** 185-206. DOI: 10.1111/j.1467-9450.1989.tb01082.x 39. Awada M., Becerik-Gerber B., Lucas G., Roll S.. **Cognitive Performance, Creativity and Stress Levels of Neurotypical Young Adults under Different White Noise Levels**. *Sci. Rep.* (2022.0) **12** 14566. DOI: 10.1038/s41598-022-18862-w 40. Vasylenko O., Gorecka M.M., Rodríguez-Aranda C.. **Manual Dexterity in Young and Healthy Older Adults. 1. Age- and Gender-Related Differences in Unimanual and Bimanual Performance**. *Dev. Psychobiol.* (2018.0) **60** 407-427. DOI: 10.1002/dev.21619 41. Özcan A., Tulum Z., Pinar L., Başkurt F.. **Comparison of Pressure Pain Threshold, Grip Strength, Dexterity and Touch Pressure of Dominant and Non-Dominant Hands within and between Right- and Left-Handed Subjects**. *J. Korean Med. Sci.* (2004.0) **19** 874-878. DOI: 10.3346/jkms.2004.19.6.874 42. Martin J.A., Ramsay J., Hughes C., Peters D.M., Edwards M.G.. **Age and Grip Strength Predict Hand Dexterity in Adults**. *PLoS ONE* (2015.0) **10**. DOI: 10.1371/journal.pone.0117598 43. Kornatz K.W., Christou E.A., Enoka R.M.. **Practice Reduces Motor Unit Discharge Variability in a Hand Muscle and Improves Manual Dexterity in Old Adults**. *J. Appl. Physiol.* (2005.0) **98** 2072-2080. DOI: 10.1152/japplphysiol.01149.2004 44. Carmeli E., Patish H., Coleman R.. **The Aging Hand**. *J. Gerontol.-Ser. Biol. Sci. Med. Sci.* (2003.0) **58** 146-152. DOI: 10.1093/gerona/58.2.M146 45. Clark B.C., Taylor J.L.. **Age-Related Changes in Motor Cortical Properties and Voluntary Activation of Skeletal Muscle**. *Curr. Aging Sci.* (2012.0) **4** 192-199. DOI: 10.2174/1874609811104030192 46. Tegelbeckers J., Schares L., Lederer A., Bonath B., Flechtner H.H., Krauel K.. **Task-Irrelevant Novel Sounds Improve Attentional Performance in Children with and without ADHD**. *Front. Psychol.* (2016.0) **6** 1970. DOI: 10.3389/fpsyg.2015.01970 47. McDonald J.J., Teder-Saälejärvi W.A., Hillyard S.A.. **Involuntary Orienting to Sound Improves Visual Perception**. *Nature* (2000.0) **407** 906-908. DOI: 10.1038/35038085 48. Störmer V.S.. **Orienting Spatial Attention to Sounds Enhances Visual Processing**. *Curr. Opin. Psychol.* (2019.0) **29** 193-198. DOI: 10.1016/j.copsyc.2019.03.010 49. **Facts about Falls|Fall Prevention|Injury Center|CDC** 50. Woolcott J.C., Richardson K.J., Wiens M.O., Patel B., Marin J., Khan K.M., Marra C.A.. **Meta-Analysis of the Impact of 9 Medication Classes on Falls in Elderly Persons**. *Arch. Intern. Med.* (2009.0) **169** 1952-1960. DOI: 10.1001/archinternmed.2009.357 51. Martínez-López E.J., Hita-Contreras F., Jiménez-Lara P.M., Latorre-Román P., Martínez-Amat A.. **The Association of Flexibility, Balance, and Lumbar Strength with Balance Ability: Risk of Falls in Older Adults**. *J. Sport. Sci. Med.* (2014.0) **13** 349-357 52. Ambrose A.F., Paul G., Hausdorff J.M.. **Risk Factors for Falls among Older Adults: A Review of the Literature**. *Maturitas* (2013.0) **75** 51-61. DOI: 10.1016/j.maturitas.2013.02.009 53. **Prevalence Estimates|Vision and Eye Health Surveillance System|Vision Health Initiative (VHI)** 54. Feys P., Lamers I., Francis G., Benedict R., Phillips G., Larocca N., Hudson L.D., Rudick R.. **The Nine-Hole Peg Test as a Manual Dexterity Performance Measure for Multiple Sclerosis**. *Mult. Scler.* (2017.0) **23** 711-720. DOI: 10.1177/1352458517690824 55. Conklyn D., Stough D., Novak E., Paczak S., Chemali K., Bethoux F.. **A Home-Based Walking Program Using Rhythmic Auditory Stimulation Improves Gait Performance in Patients with Multiple Sclerosis: A Pilot Study**. *Neurorehabil. Neural Repair* (2010.0) **24** 835-842. DOI: 10.1177/1545968310372139 56. Shahraki M., Sohrabi M., Taheri Torbati H.R., Nikkhah K., NaeimiKia M.. **Effect of Rhythmic Auditory Stimulation on Gait Kinematic Parameters of Patients with Multiple Sclerosis**. *J. Med. Life* (2017.0) **10** 33-37. PMID: 28255373
--- title: Drug Information Sources for Patients with Chronic Conditions in the Qassim Region, Saudi Arabia authors: - Saeed Alfadly - Mohammed Anaam - Mohammed Alshammari - Saud Alsahali - Ejaz Ahmed - Abdulkareem Bin Mubarak - Abdullah Aldahouk - Muhanna Aljameeli journal: Pharmacy year: 2023 pmcid: PMC10037653 doi: 10.3390/pharmacy11020057 license: CC BY 4.0 --- # Drug Information Sources for Patients with Chronic Conditions in the Qassim Region, Saudi Arabia ## Abstract Appropriate drug information is vital for the correct use of drugs in pharmaceutical practice. Providing patients with educational advice on prescribed medication and on proper medication administration has become an essential part of the pharmaceutical care process. The objectives of this study were to identify patients’ knowledge of prescribed medications, their desire for more information, and the sources of medication information in a population from Qassim, Saudi Arabia, using a cross-sectional descriptive study. Our target population consisted of adult patients with chronic illnesses receiving drugs at outpatient pharmacies. Nineteen pharmacies were selected based on convenience. After collecting their prescriptions, patients were asked to take part in the study by interviewers as they were leaving the pharmacies. The questionnaire used was pretested on 18 patients and then modified accordingly. questions investigated participants’ knowledge of drug information, their wish for more information, and their sources of drug information, other than clinicians. Descriptive analysis was used to describe patients’ physical details. The effect of sex, education, diagnosis, number of drugs, and age on knowledge of the purpose of drugs and the need for additional information were tested using Chi-square test. A total of 270 patients were interviewed, of whom $29.7\%$ reported not knowing the purpose of at least one of their medications, and only reading a portion of the PILs. Of the patients sampled, $56.7\%$ said they read the side effects section of the PIL, $43.3\%$ reported reading the uses, while $27\%$ read the contraindications. The drug -interactions section was the least read, with only $18.9\%$ reporting reading it. A total of $57\%$ of the patients reported that they needed more information about their medications. Highly educated patients reported using the PIL, social media, family and friends, TV, and newspapers as sources of drug information at significantly higher rates than patients with lower levels of education. Healthcare professionals should assess patient comprehension and the need for additional drug information, especially among patients with low levels of education. Additionally, healthcare professionals should consider other information sources used by their patients. ## 1. Introduction The health care in kingdom of Saudi Arabia (KSA) were one of many fields that ex-panded and improved under national transformative plan recently. The focus of pharma-ceutical services within the health service today is on providing effective pharmaceutical care. The community pharmacies were one of the frequently visited places by Saudi healthcare consumers for many reasons including dispensing their prescription, seek ad-vice related to their diseases and medications [1]. Because it’s convenient and easily ac-cessible by the consumers, the community pharmacies play crucial role in pharmaceutical care services offered to the patients in the country. A study conducted in Saudi Arabia reported that patient-centered care is a new concept in community pharmacies in the country and perceived that the knowledge and skills of community pharmacists are insufficient to provide effective patient-centered care [1]. However, the study acknowledged that chain community pharmacy groups in Saudi Arabia do provide diabetes education and other patient-centered care programs in some of their community pharmacies [2]. Few studies in the KSA have demonstrated that Saudi healthcare consumers appreciate community pharmacies providing patient-centered care, and they instead prefer to ask medication- and disease-related questions at community pharmacies over primary healthcare centers and hospitals because of their easy accessibility and convenient opening times [3,4]. Appropriate drug information is critical for proper drug use in pharmacy care practice. providing patients with education about their prescriptions and on proper medication administration has become an essential part of the pharmaceutical care process [5,6,7]. Inadequate patient education on drug therapy can result in therapeutic failure, disease recurrence, drug-induced side effects, and increased costs [7,8]. Patients’ emotional understanding and access to appropriate information should be prioritized when providing healthcare [9]. Drug information is especially important for equipping and preparing patients to take their medications correctly. According to studies, patients who receive accurate drug information better adhere to their drug treatment regimens [8]. Patient understanding of the purpose of their medications and fear of adverse effects can have a significant effect on medication adherence [10,11]. Previous research from various countries has shown that physicians and pharmacists are the most frequently used sources of drug information by patients [12,13,14]. The majority of patients appear to receive insufficient drug-related instructions from their doctors or pharmacists [15]. Furthermore, some patients may not understand all the information given by their healthcare professionals. As a result, these patients may look for alternative sources of Patient Information Leaflets (PILs) [12,13,16]. Written PILs are useful to a large proportion of the public [17]. Most PILs that come with medications contain several instructions: information on drug preparation, mechanism of action, pharmacokinetics, adult and pediatric dosages for various illnesses that can be treated, adverse effects, drug interactions, contraindications, cautions and warnings, a package of drug prescription, and ideal storage conditions [18]. Previous research found that patients also obtain drug information from other sources, including the media (TV, newspapers), their peers, and the Internet [12,13]. However, media coverage of medications may be insufficient in terms of the drugs’ benefits, risks, and costs [19]. This study’s objectives were to identify patients’ knowledge of prescribed medications, understand their desire for more information, and determine their sources of medication information in Saudi Arabia’s Qassim region. ## 2.1. Study Area and Period The target population was adult and elderly patients with chronic illnesses receiving drugs from outpatient pharmacies in the Qassim Region, Saudi Arabia. The study was conducted between March 2018 and June 2018. ## 2.2. Study Design A cross-sectional descriptive study was used to collect data from patients with chronic diseases visiting outpatient pharmacies in the Qassim region. ## 2.3. Study Population Our target population was adult and elderly patients with chronic illnesses receiving drugs from outpatient pharmacies during the study. Patients who were 18 years of age or older were invited to take part in the study. Patients unable to complete the interview due to cognitive issues were excluded. The included age intervals were: below 25 years, 25–39 years, 40–50 years, 51–59 years, and 60 years or older. ## 2.4. Sample Size Determination and Sampling Method Sample size was calculated using the following formula developed by Cochran [20]:[1]N=Z2(PQ)D2=1.962 (0.5∗0.5)(0.06)2=0.96040.0036=266.6 ≈ 270 patients where N is the sample size, Z is the standard error associated with the selected level of confidence, P is the population proportion, Q = (1 − P), and D is the error of estimation (precision). The level of confidence used in this study was $95\%$. The Z value associated with $95\%$ confidence was 1.96. The actual value of P was not known before the study; if P is unknown, researchers typically use 0.5 as an estimation, which made $Q = 1$ − 0.5 = 0.5. the error of estimation was $6\%$. The sample size was 270 patients. A convenience sampling technique was used, and only those who agreed to participate were included in the study. ## 2.5. Study Questionnaire Data were collected using a questionnaire derived from the literature (permission was granted by the main author of Amin et al., 2011 [21]. The questionnaire was pretested on 18 patients ($6\%$ of the sample size) and then modified accordingly. The questionnaire investigated patients’ information, willingness for more information, and drug information sources beyond their healthcare providers. The first questions focused on patient’s knowledge of the purpose of their drug(s) and whether they needed more information about the drugs (Table 1). Subsequent questions were related to reading specific risk-related topics in package inserts, including side effects, contraindications, and drug interactions. Other questions included demographic data and drug information sources. ## 2.6. Data Collection Data collection was carried out at outpatient pharmacies by fifth-year pharmacy students. A total of 270 patients with prescriptions for medications for chronic conditions were interviewed. The interviewers called on patients to take part in the study as they left the pharmacies after collecting their medications. ## 2.7. Data Analysis SPSS version 20.0 software was used to analyze the data. Descriptive statistics (i.e., frequencies and percentages) were used to summarize participants’ responses to the questionnaire. Chi-square test was used to examine any associations that exist between categorical variables. A p-value of <0.05 was considered statistically significant. ## 2.8. Ethical Approval The ethical approval for this study was obtained from the regional research ethics committee, Qassim Region (Reference no. 2018-02-10). The participants were informed with the aims of study and the data protection of the participants. Then, verbal consent was obtained from the patients to participate in the study. All participation was voluntary. ## 3. Results A total of 270 outpatients from 19 pharmacies (16 community pharmacies and three outpatient pharmacies in three major hospitals in the area) were interviewed. The study population comprised $46.1\%$ females and $53.9\%$ males (Table 2). The study population also included a range of education backgrounds, from elementary school to university. Table 2 shows that $43\%$ of the study population had university education, $27\%$ were high -school educated, and $30\%$ had elementary education. Table 3 shows that more than half of our study sample ($52.6\%$) reported regularly receiving 1–3 drugs, $27.8\%$ received up to five drugs, $10.4\%$ received up to seven drugs, and $9.2\%$ were prescribed more than seven drugs. Perhaps unsurprisingly—given that our study focused on patients with chronic conditions—patients below 25 years of age constituted the smallest percentage of participants in our study ($13\%$). Most of our study population consisted of patients aged 51–59 years ($23\%$) and patients older than 60 years ($22.2\%$). Patients between the ages of 25 and 39 years comprised $21.1\%$ of our study population, while patients between the ages of 40 and 50 years comprised $20.7\%$ of the study population, as shown in Table 2. Diagnoses and types of chronic diseases are also reported. diabetes and/or hypertension were the major illnesses in our sample population ($40.3\%$). patients with GIT (gastrointestinal) disorders comprised only $10\%$ of the population. other diagnosed diseases included heart diseases ($4.4\%$), kidney diseases ($3.3\%$), and allergies ($5.6\%$). Patients with diseases not mentioned in the questionnaire had the option of choosing ‘other’. they were subsequently asked to state the type(s) of chronic diseases they were suffering from. These patients comprised $36.4\%$ of the study population (Table 3). The interviewers assessed the patients’ perceived understanding of the purpose of the drugs they were taking. A high number of patients ($70\%$) reported knowing the purpose of the drugs; however, $14.1\%$ of the study population did not know the purpose of the drugs they were consuming, and $15.6\%$ had an idea of the purpose of some, but not all, of the drugs, as indicated in Table 4. The results in Table 4 also show how sex, education, diagnosis, number of drugs, and age affect knowledge of the purpose of medications and the need for additional information. The table shows that both male and female patients had a relatively high level of knowledge of the purpose of their drugs—$67.6\%$ and $72.8\%$, respectively. It was clear that more highly educated patients had a higher likelihood of knowing the purpose of the drugs they were taking ($p \leq 0.05$). For instance, $87\%$ of patients who reported having university education knew the purpose of their prescribed drugs; $64.4\%$ of high -school -educated patients reported knowledge of their drugs, while primary -school -educated patients were the least likely to know the purpose of their drugs ($49.4\%$). Regarding diagnosis, the results show that $67.4\%$ of patients with diabetes identified the purpose of their prescribed medications, $68.2\%$ of hypertensive patients identified the purpose of their respective medications, and $83.3\%$ of patients with heart diseases identified the purpose of their medications. Other notably high proportions of patients with self-reported drug knowledge included patients with GIT disorders and allergies—$85.2\%$ and $86.7\%$, respectively. With respect to the number of drugs, $76.1\%$ [108] of the patients taking 1–3 medications identified the purpose of all their medications, but when patients were taking more than seven medications, this proportion reduced ($44\%$). Table 4 clearly shows that the younger the patient, the higher the likelihood of them knowing the purpose of their drugs: $94.2\%$ of patients under the age of 25 years reported that they knew the purpose of the drugs they were taking; however, this decreased as the patients grew older. The 25–39 and the 40–50 age groups reported knowing the purpose of their medications $82.6\%$ and $73.2\%$ of the time, respectively. In the 51–59 age group, this number was $59.7\%$. Finally, the oldest age group (the 60 years or older age group) reported this $51.7\%$ of the time. Another aspect of the study was to gauge patients’ need for more information about their medications. Table 4 shows that there was little difference between the sexes, as most of both male and female patients reported a need for more drug information—$71.7\%$ and $75.2\%$, respectively. Among university educated patients, $82.6\%$ wanted more drug information, whereas $74\%$ of high-school-educated patients and $55.5\%$ of primary-school-educated patients expressed this desire ($p \leq 0.05$). diabetic and hypertensive patients had similar views regarding the desire for more information, at $69.8\%$ and $68.2\%$, respectively. Patients who had both diabetes and hypertension concurrently reported a need for more information about their prescribed medications $70.4\%$ of the time. Notably, patients suffering from GIT disorders reported the highest need for more information ($88.9\%$), followed by patients with heart disease ($83.3\%$). Of the Other diagnoses, $66.7\%$ of patients with kidney disease and $66.7\%$ of patients with allergies reported a need for more information. The results in Table 5 show that almost half of the patients ($49.6\%$) reported that they read the Package Information Leaflets (PILs) sometimes, while $25.2\%$ reported always reading them, and $25.2\%$ reported never reading the PILs. The results in Table 5 indicate that $25.56\%$ of the patients did not read any of the adverse effects, uses, drug interactions, or contraindication sections of the package inserts. As Table 6 clearly shows, $56.5\%$ of the patients said they read the side effects portion of the PIL, $43.7\%$ reported reading the topic on uses, $27\%$ read about the contraindications, while drug interactions were read the least number of times, by only $18.8\%$ of the patients. The results in Table 7 show that the most frequently used source of drug information was the Internet ($58.75\%$), followed by family and friends ($47\%$) ($p \leq 0.05$). TV as a source of information was used by only $14.2\%$ of the patients, while newspapers were the least used source of information, with only $4.4\%$ of patients relying on them for drug information. Table 7 shows, as expected, that patients with university education mostly used the internet, with $79.3\%$ reporting this as a source of drug information. More than half of patients with a high school education used the Internet ($57.5\%$), while $29.6\%$ of patients with elementary -level education reported using the internet as a source of drug information. There was a significant proportional relationship between using the Internet (which generally requires at least an adequate knowledge of English) and education level: patients with a higher level of education were more likely to use the internet to obtain additional drug information. On the relationship between age and use of the Internet, a high percentage of patients aged 25–39 years reported using the internet ($72\%$). This was also the case in the 40–50 age group ($71.4\%$). More than half ($54.8\%$) of patients aged 51–59 used the Internet. Among patients below 25 years of age, $68.6\%$ reported using the Internet. However, the percentage of older patients (aged 60 years or over) using the Internet was lower than the other age groups, with just $31.7\%$ reporting using the Internet to obtain drug information. ## 4. Discussion Prescriptions are the most common therapeutic interventions in medical practice. As a result, reliable drug information is essential for pharmaceutical care. Barber et al. [ 22] found that a relatively high percentage of drug information needs in their study had not been met. In that study, $52\%$ of elderly patients aged 75 years or older who were on new medications for chronic conditions desired more information on their medications after four weeks. This was slightly lower than the percentage of patients who were 60 years or older ($63.3\%$) who said they needed more drug information in our study. Furthermore, Davis et al. [ 23] indicated that patients with chronic diseases who had a lower level of education were more likely to misunderstand the information written on medication labels. The reason patients with primary school education reported poor understanding and a lower need for more drug information, as shown in Table 4, was that they were less likely to read the Patient Information Leaflet (PIL) topics compared with university educated patients. This finding is in agreement with the findings of other studies [17,24]. Another possible cause is that patients with a low level of education who read the PILs may have difficulty understanding them. With respect to other sources of drug information, patients with a lower level of education were less likely to use TV, newspapers, and the internet as sources of medication information compared with patients with secondary school and university education. These findings are consistent with those of Brodie et al. [ 24]. Based on our findings, patients with a lower level of education depend mainly on periodic pharmacy visits to evaluate and discuss issues relating to their understanding of the purpose, side effects, and drug interactions of their medications. Previous research indicates that a substantial number of patients think that their healthcare providers fail to adequately inform them about drug treatments [15]. Furthermore, many serious medication-related questions arise the moment the patient has left the clinic or pharmacy. Approximately $47\%$ of our study population relied primarily on unreliable sources of information, such as friends and family. This finding was relatively consistent with the $40\%$ found in an Iranian study [25]. Previous research has also found that, in addition to family and friends, TV is a significant source of drug information [25]. This lack of consistent access to information could have serious health consequences, and therefore needs to be addressed and resolved by health policymakers and other relevant stakeholders. In some nations, many patients search for medication information on the Internet [24]. In our study, $58.5\%$ of patients obtained drug information from the Internet. However, in another study [25], the Internet played an insignificant role. In a study undertaken in Egypt by Amin et al. [ 2011] [21], the Internet was the least-used source of information, with only $12\%$ reporting its use, whereas reliance on family and friends, at $48.6\%$, was consistent with our finding. The gap in Internet usage in the two studies is interesting and somewhat logical because it was estimated that $35.6\%$ of the Egyptian population had access to and used the Internet at the time of the study by Amin et al. at 2011 [21], whereas, at the time of our study at 2018, a much larger proportion of the Saudi population had access to the Internet. The instructions attached to the drug container and PILs are the most reliable sources of drug information. Several studies have found that written information effectively improves patient compliance [26]. Our study found that $25.2\%$ of patients reported always reading the package insert, which was in agreement with another study where $27\%$ of the patients read the medication information leaflets [24]. However, in the study conducted in Iran [25], only $14\%$ of the patients frequently used PILs as sources of medication information. Primary-school-educated patients were less likely to read PILs. A lack of communication between clinicians or pharmacists and their patients, as well as only occasionally reading PILs, may make patients turn to untrustworthy sources of information such as family and friends. These patients are more susceptible to medication errors that result in severe health issues [9,27]. Therefore, carefully designed, impartial, scientifically reliable, and updated PILs written in a legible format that is understandable to the average patient are of utmost importance [27]. The number of drugs taken by a patient is an important parameter for physicians and pharmacists to consider when patients require more drug information. According to the study’s findings, the more medications patients took, the less likely they were to understand the reasons for taking them and the less likely they were to read about the adverse effects, uses, drug interactions, and contraindications contained in the PILs. Patients who had more medications were also more likely to need more drug information when they left the pharmacy. Davis et al. [ 2006] reported that a greater number of prescription medications used by patients was significantly associated with misunderstanding the instructions on medications label [23]. It is important for patients taking five or more medications to be aware that there is a higher risk of drug interactions. Our results indicate that patients were more likely to read the side effects than the uses, drug interactions, or contraindications. Side effects have been reported to be the most frequently read topic in written drug information [27,28]. Vander Stichele et al. [ 27] found that patients focused more on side effects and less on contraindications, which is consistent with our results. According to Raynor et al. [ 28], $12\%$ of those who received a PIL did not read it because they did not see it. Van Haecht et al. [ 29] reported no sex differences in PI reading; however, several other studies have found that women read more drug information leaflets than men [27]. There was a small difference in the frequency of PIL reading between the two sexes in our study. It is essential to take the study’s country into account when interpreting these findings. The patient’s age also influenced the perception of the necessity for medication information and the use of other drug information sources. Duggan and Bates [30] reported an inverse relationship between age and the desire for information on medication. This contradicted our finding of unmet drug information needs, which were highest among those under the age of 25 years, and then gradually declined in patients aged 25–39, 40–50, and 51–59 years. Furthermore, younger people used the Internet to obtain drug information more frequently. Elderly patients were found to be less familiar with modern drug information sources such as the Internet [6]. While receiving the same quantity of information, one patient might feel that it is inadequate and another patient might feel that it is adequate. This distinction will have an effect on both satisfaction with current information and the active search for additional information. The need and desire for extra information may be influenced by the patient’s state and disease, such as the type of diabetes or the severity of hypertension. The main limitation of this study was the use of convenience sampling to recruit pa-tients. “ Do you think you need to know more about the drugs you’re taking?” is a question that is positively framed. This question may be prompted patients to answer positively. Also, our study only included patients from the *Qassim area* and conducted in a specific time so, the findings obtained from the population reflect specific time. More re-search should be conducted to study medication practices and behaviors among patients with multiple chronic diseases in different areas of Saudi Arabia to compare the various trends. ## 5. Conclusions After leaving the pharmacy, many patients with chronic diseases stated that they did not grasp the intended use of at least one of their drugs, and many of them said they wanted more information about the drugs. Based on this information, it is clear that many patients in our study did not understand or may not have received adequate information from clinicians or pharmacists; thus, their understanding and desire for information were not adequately assessed. Since pharmacists are the last point of contact for patients, they should assess patient understanding and the need for additional drug information. Patients with a university or secondary school education often rely on other sources of information, such as PILs, to meet their medication information needs, compared with patients with primary school education. These findings imply that healthcare providers should focus on patient education and understanding, particularly among patients with primary school education and those taking multiple medications. Healthcare decision-makers should consider education level when developing accurately designed, impartial, scientifically reliable, and updated PILs written in a legible format that is understandable to the average patient. Healthcare professionals who deliver services should suggest recognized resources suitable for a patient’s disease and level of education, while also encouraging patients to discuss any related issues and concerns raised by other resources. ## References 1. Rasheed M.K., Alqasoumi A., Hasan S.S., Babar Z.-U.. **The community pharmacy practice change towards patient-centered care in Saudi Arabia: A qualitative perspective**. *J. Pharm. Policy Pr.* (2020) **13** 59. DOI: 10.1186/s40545-020-00267-7 2. **Al-Nahdi Diabetes Education**. (2019) 3. Alhaddad M.S.. **Youth Experience With Community Pharmacy Services and Their Perceptions Toward Implementation of Medication Therapy Management Services by Community Pharmacists in the Western Region of Saudi Arabia**. *Ther. Innov. Regul. Sci.* (2019) **53** 95-99. DOI: 10.1177/2168479018769299 4. Dahal R., Bista S.. *Strategies to Reduce Polypharmacy in the Elderly* (2021) 5. Al-Tannir M., Alharbi A.I., Alfawaz A.S., Zahran R.I., AlTannir M.. **Saudi adults satisfaction with community pharmacy services**. *Springerplus* (2016) **5** 774. DOI: 10.1186/s40064-016-2442-8 6. Blom L., Krass I.. **Introduction: The role of pharmacy in patient education and counseling, Patient Education and Counseling**. *Patient Educ. Couns.* (2011) **83** 285-287. DOI: 10.1016/j.pec.2011.05.021 7. Wolf M.S., Davis T.C., Shrank W.H., Neuberger M., Parker R.M.. **A critical review of FDA-approved medication guides**. *Patient Educ. Couns.* (2006) **62** 316-322. DOI: 10.1016/j.pec.2006.06.010 8. Ulfvarson J., Bardage C., Wredling R.A., Von Bahr C., Adami J.. **Adherence to drug treatment in association with how the patient perceives care and information on drugs**. *J. Clin. Nurs.* (2007) **16** 141-148. DOI: 10.1111/j.1365-2702.2005.01477.x 9. Åström K., Carlsson J., Bates I., Webb D.G., Duggan D., Sanghani P., McRobbie D.. **Desire for information about drugs A multi-method study in general medical inpatients**. *Pharm. World Sci.* (2000) **22** 159-164. DOI: 10.1023/A:1008783115420 10. McHorney C.A.. **The Adherence Estimator: A brief, proximal screener for patient propensity to adhere to prescription medications for chronic disease**. *Curr. Med. Res. Opin.* (2009) **25** 215-238. DOI: 10.1185/03007990802619425 11. Svarstad B.L., Chewning B.A., Sleath B.L., Claesson C.. **The Brief Medication Questionnaire: A tool for screening patient adherence and barriers to adherence**. *Patient Educ. Couns.* (1999) **37** 113-124. DOI: 10.1016/S0738-3991(98)00107-4 12. Närhi U., Helakorpi S.. **Sources of medicine information in Finland**. *Health Policy* (2007) **84** 51-57. DOI: 10.1016/j.healthpol.2007.05.005 13. Ho C.H., Ko Y., Tan M.L.. **Patient needs and sources of drug information in Singapore: Is the Internet replacing former sources?**. *Ann. Pharmacother.* (2009) **43** 732-739. DOI: 10.1345/aph.1L580 14. Sleath B., Wurst K., Lowery T.. **Drug information sources and antidepressant adherence**. *Community Ment. Health J.* (2003) **39** 359-368. DOI: 10.1023/A:1024080410284 15. Maywald U., Schindler C., Krappweis J., Kirch W.. **First patient-centered drug information service in Germany—A descriptive study**. *Ann. Pharmacother.* (2004) **38** 2154-2159. DOI: 10.1345/aph.1E210 16. Bandesha G., Raynor D.K., Teale C.. **Preliminary investigation of patient information leaflets as package inserts**. *Int. J. Pharm. Pract.* (1996) **4** 246-248. DOI: 10.1111/j.2042-7174.1996.tb00875.x 17. Nathan J.P., Zerilli T., Cicero L.A., Rosenberg J.M.. **Patients’ use and perception of medication information leaflets**. *Ann. Pharmacother.* (2007) **41** 777-782. DOI: 10.1345/aph.1H686 18. Zehnder S., Beutler M., Bruppacher R., Ehrenhöfer T., Hersberger K.E.. **Needs and use of drug information sources in community pharmacies: A questionnaire based survey in German-speaking Switzerland**. *Pharm. World Sci.* (2004) **26** 197-202. DOI: 10.1023/B:PHAR.0000035881.17853.e0 19. Moynihan R., Bero L., Ross-Degnan D., Henry D., Lee K., Watkins J., Mah C., Soumerai S.B.. **Coverage by the news media of the benefits and risks of medications**. *N. Engl. J. Med.* (2000) **342** 1645-1650. DOI: 10.1056/NEJM200006013422206 20. Cochran W.G.. *Sampling Techniques* (1977) 21. Amin M.E., Chewning B.A., Wahdan A.M.. **Sources of drug information for patients with chronic conditions in Alexandria, Egypt**. *Int. J. Pharm. Pract.* (2011) **19** 13-20. DOI: 10.1111/j.2042-7174.2010.00069.x 22. Barber N., Parsons J., Clifford S., Darracott R., Horne R.. **Patients’ problems with new medication for chronic conditions**. *Qual. Saf. Health Care* (2004) **13** 172-175. DOI: 10.1136/qshc.2003.005926 23. Davis T.C., Wolf M.S., Bass P.F., Thompson J.A., Tilson H.H., Neuberger M., Parker R.M.. **Literacy and misunderstanding prescription drug labels**. *Ann. Intern. Med.* (2006) **145** 887-894. DOI: 10.7326/0003-4819-145-12-200612190-00144 24. Brodie M., Flournoy R.E., Altman D.E., Blendon R.J., Benson J.M., Rosenbaum M.D.. **Health information, the Internet, and the digital divide**. *Health Aff.* (2000) **19** 255-265. DOI: 10.1377/hlthaff.19.6.255 25. Nader F., Mousavizadeh K., Ghafourifar P.. **Patient sources for drug information in Iran: A questionnaire-based survey**. *Pharm. World Sci.* (2008) **30** 764-767. DOI: 10.1007/s11096-008-9254-z 26. Svarstad B.L., Mount J.K., Tabak E.R.. **Expert and consumer evaluation of patient medication leaflets provided in U.S. pharmacies**. *J. Am. Pharm. Assoc.* (2005) **45** 443-451. DOI: 10.1331/1544345054475586 27. Vander Stichele R.H., Van Haecht C.H., Braem M.D., Bogaert M.G.. **Attitude of the public toward technical package inserts for medication information in Belgium**. *DICP* (1991) **25** 1002-1006. DOI: 10.1177/106002809102500916 28. Raynor D.K., Knapp P., Moody A., Young R.. **Patient information leaflets—Impact of. European regulations on safe and effective. use of medicines**. *Pharm. J.* (2005) **275** 609-611 29. Van Haecht C.H., Vander Stichele R., Bogaert M.G.. **Package inserts for antihypertensive drugs: Use by the patients and impact on adverse drug reactions**. *Eur. J. Clin. Pharmacol.* (1990) **39** 551-554. DOI: 10.1007/BF00316093 30. Duggan C., Bates I.. **Medicine information needs of patients: The relationships between information needs, diagnosis and disease**. *Qual. Saf. Health Care* (2007) **17** 85-89. DOI: 10.1136/qshc.2005.017590
--- title: 'Development of core outcome sets of Food for Special Medical Purposes designed for type 2 diabetes mellitus: a study protocol' authors: - Dongyu Mu - Jie Gong - Yaoyao Wei - Muxi Chen - Jiajie Yu - Liang Du - Wen Hu journal: Trials year: 2023 pmcid: PMC10037765 doi: 10.1186/s13063-023-07214-2 license: CC BY 4.0 --- # Development of core outcome sets of Food for Special Medical Purposes designed for type 2 diabetes mellitus: a study protocol ## Abstract ### Background The Chinese government stipulates all food for special medical purposes (FSMP) designed for specific diseases to be tested in clinical trials before approving it for registration. The process of developing core outcome sets (COSs), the minimum sets of outcomes supposed to be measured and reported, provides an economical and practical option for stakeholders to communicate and cooperate in conducting clinical trials as well as in reporting FSMP outcomes. This study uses type 2 diabetes mellitus (T2DM) as an example to develop COS for clinical trials of FSMP. ### Methods The COS for FSMP-T2DM will be divided into 3 phases and developed following COS-STAP and COS-STAD: [1] *Generate a* list of relevant outcomes identified from a systematic review, in which information sources will mainly include published studies, regulatory documentation, and qualitative interviews of stakeholders. The identified outcomes will be categorized using a conceptual framework and formatted into the first round of the Delphi survey questionnaire items. [2] At least 2 rounds of Delphi surveys will be performed among stakeholders to create the COS for FSMP-T2DM. Patients, clinical dietitians, physicians, COS researchers, journal editors, FSMP manufacturers, and regulatory representatives will be invited to score each outcome from aspects of importance. [3] *Hold a* face-to-face or online consensus meeting to refine the content of the COS for FSMP-T2DM. Key stakeholders will be invited to attend the meeting to discuss and agree on the final COS. ### Discussion We have prepared an alternative solution of the Likert scale selection, Delphi survey rounds, scoring group, and consensus definitions in case of an unexpected situation. ### Trial registration COMET [1547]. Registered on March 23, 2020. ### Supplementary Information The online version contains supplementary material available at 10.1186/s13063-023-07214-2. ## Background This study protocol follows the Core Outcome Set-STAndardised Protocol Items (COS-STAP), a checklist of 13 items considered essential documentation in a COS protocol [1]. Food for Special Medical Purposes (FSMP), also known as medical food in the USA, has been widely used in clinical practice since the 1970s [2]. FSMP has made significant contributions in improving nutritional status, promoting rehabilitation, shortening hospital stays, and saving medical expenses [3–7]. The definition of FSMP and different categories of FSMP are shown in Fig. 1.Fig. 1FSMP classification system in China. Superscript lowercase letter a (a) indicates the following: FSMP: formula foods that are specially processed and formulated to meet the special needs for nutrients or diets of people with limited eating, digestion, and absorption disorders, metabolic disorders, or specific disease states. Such products should be consumed alone or in combination with other foods under the guidance of a doctor or clinical dietitian. Superscript lowercase letter b (b) indicates the following: whole-nutrient FSMP (standardized): can be used as a single source of nutrition to meet the nutritional needs of the target population. Superscript lowercase letter c (c) indicates the following: whole-nutrient FSMP (disease specified): can be used as a single nutritional source to meet the nutritional needs of the target population diagnosed with specific diseases or medical conditions. Superscript lowercase letter d (d) indicates the following: nonwhole-nutrient FSMP: can meet part of the nutritional needs of the target population but is not suitable for use as a single source of nutrition However, China entered this field relatively late but passed stricter acts. It has made clinical trials for FSMP compulsory to regulate the domestic FSMP market and achieve the goal of the “Healthy China 2030” initiative [8, 9]. Since 2013, the State Administration for Market Regulation (SAMR), previously known as the China Food and Drug Administration (CFDA), and other government bureaus have gradually issued documentation related to the registration, administration, and supervision of the FSMP. These documentations stipulate that FSMP designed for certain diseases, i.e., whole-nutrient FSMP (disease specified), cannot be registered successfully without clinical trial reports [9–12]. Unbalances have appeared between the supplies and requirements of FSMP. Search results from the Special Food Information Query Platform indicate that 92 FSMP products have been registered; however, only one product has been registered under the category of whole-nutrient FSMP (disease specified) as of November 3, 2022 [13]. Another 39, 23, and 29 FSMP products were registered in the infant FSMP, whole-nutrient FSMP (standardized), and nonwhole-nutrient FSMP segments, respectively. Additionally, the features of existing clinical trials for FSMP, including long conducting time and diverse outcome selections, may exacerbate the imbalances by slowing down the process of FSMP registration and supplies. Our systematic review (SR) enrolled 46 studies of FSMP designed for T2DM from 14,198 items searched from 6 medical databases. The total intervention duration included in the enrolled studies ranged from 6–90 days, of which the more commonly used durations were 7, 14, 28, 84, and 90 days. Those trials are usually divided into 2, 3, or 4 cycles, with an interval of 2–14 days, but the washout period of 7 days is the most common. After integrating another 9 indices obtained from CFDA files to 161 outcomes extracted from the method sections of the enrolled studies, the outcome pool increased to 170, but only 54 ($31.8\%$) outcomes appeared ≥ 3 times [14, 15]. The frequencies of only 3 outcomes, i.e., fasting blood glucose, postprandial blood glucose, and triglyceride levels, touched half of the enrolled studies (≥ 23 times). Outcomes are so diverse in FSMP fields that research waste occurs, as data deviating from clinical needs cannot be combined or compared [16, 17]. Although diverse outcomes give clinical trials of FSMP freedom to design, they increase the difficulty of registration approval for the government. Therefore, a compact and lightweight solution may better suit governmental needs. Core outcome set (COS) is one alternative solution to compressed outcomes of FSMP to the minimum without losing what is essential or required for each stakeholder group. COS is defined as the minimum sets of outcomes that should be measured and reported [18]. It can help standardize the selection, measurement, and reporting of outcomes, improve the practicality and comparability of research results, and optimize the use of individual research data [17, 19–21]. Therefore, COS will be an optimized option to assess and control risks before and after hitting the market, as applying secondary analysis of data will be feasible. This study aims to use T2DM as an example to develop COS for clinical trials of FSMP and provide a multistakeholder agreed tool for regulators to use for the pre- and postmarket management of FSMP. ## Methods The entire study design adopted the approach recommended in COS-STAP and Core Outcome Set-STAndards for Development (COS-STAD), the minimum standard recommendations to improve the methodological approach toward planning a COS study [17, 22, 23]. The results will be reported in the format of the Core Outcome Set–STAndards for Reporting (COS-STAR), a reporting guideline for studies developing COSs [24]. ## Scope The health condition(s) and population(s) covered by the COS This COS is developed for adult T2DM patients who are malnourished or are at nutritional risk; need to receive EN (Enteral Nutrition), without food allergy history such as having symptoms of pruritus or mild cutaneous eruption, etc., after taking milk, eggs, or FSMP; and do not have acute or severe complications, including endocrine diseases such as hyperthyroidism, or other diseases that can significantly influence the estimated energy requirement (EER). They should be capable of taking oral antidiabetic drugs or injecting exogenous insulin. There are no limits to the course of T2DM. Both new and old patients can be included. Women during the gestation or lactation periods will be excluded.[2]The intervention(s) covered by the COS The intervention covers FSMP designed for T2DM used in medical nutrition therapy (MNT), which has previously also been known as the elemental diet, monomeric formula, nonelemental diet, and polymeric formula [25]. T2DM patients can fully or partially obtain the required nutrients from these types of FSMP with or without feeding tubes. Blenderized diets are included but not nutrient modules such as peptides, vitamins, minerals, or dietary fibers.[3]The setting(s) in which the COS is to be appliedResearch and routine clinical practice. Institutions qualified to conduct clinical trials of FSMP are recommended to use this COS. Detailed information on these qualified institutions can be found on the official website of SAMR [26]. ## Stakeholders A diverse representative sample of stakeholders from a broad geographical area within China will participate anonymously. They should either have COS development experience or have been involved in the research and development, registration, production, supervision, and circulation chain of FSMP. Key stakeholders will consist of (a) COCONUT steering committee members and (b) other stakeholders who will have consecutively participated in all rounds of Delphi surveys. They will be given priority to attend the consensus meeting. ## Stakeholder groups Stakeholders will be divided into 4 groups and will be involved in the COS development process at all stages to ensure that the outcomes relevant to all groups are included for the COS to be widely adopted [27]. Some of them are mixed stakeholder representatives. These 4 groups will be as follows:COS developers: They will be recruited from among the corresponding authors of published COS projects in the ChiCOS database and requested to forward the invitation to other authors of those COS projects [28].Journal editors: They will be recruited from the editorial board of the Chinese Journal of Evidence-Based Medicine, etc., in which COS studies have been published. COS users: They will be recruited from (a) researchers of clinical trials enrolled in our SR; (b) nutritional clinical guideline developers; (c) healthcare professionals of clinical nutrition and medicine; (d) regulators from the Special Food Safety Supervision and Administration Department of SAMR; and (e) representative manufacturers in the FSMP industry. Patient representatives: They will be recruited via snowball sampling processes from the Hospital-to-Home Nutrition Management Center (H2H) [29]. Patient inclusion criteria have been described in the scope section. ## COCONUT working group The working group of Core Outcome sets and Core Outcome measurement sets in NUTriology (COCONUT) refers to the coordinator of this study, and it has been initiating a series of nutriological COS studies since 2019 [30]. ## The steering committee of COCONUT COCONUT will bring together approximately 60 committee members to consider the COS with the assistance of the Clinical Nutrition Specialty Alliance of West China Hospital, Sichuan University/Western Medical Nutrition Alliance (WMNA) [31], Chinese Gerontological Society of Nutrition and Food Safety Association (CGSN) [32], the Chinese Evidence-Based Medicine Center (Cochrane China Center) [33], and the Chinese Clinical Trials Core Outcome Set Research Center (ChiCOS) [28]. The steering committee will be convened for the following functions:Monitor and review the results of each round of the Delphi surveyAttend and help facilitate the consensus meetingReview, finalize, and contribute to the publication and dissemination of the COS, reporting guidance of COS for FSMP, and relevant explanatory documents ## The expert panel of COCONUT The panel will include the leader of the steering committee and 3–5 other authoritative committee members. They will review and guide the steering committee at each phase of this project in addition to chairing the consensus meeting. ## The secretariat of COCONUT The secretariat comprising COS developers of this study will be convened to perform the following functions:Maintain contact with the stakeholdersConduct SR and Delphi surveysOrganize the consensus meetingAcquire, analyze, and interpret the data and draft articles ## Overview This COS study will have 3 phases, as shown in Fig. 2. Briefly, these would be as follows:*Generate a* list of relevant outcomes identified from an SR. The information sources of the SR are mainly from (a) published studies, (b) regulatory documentation, and (c) qualitative interviews of stakeholders. Identified outcomes will be categorized via a conceptual framework and formatted into the first round of Delphi survey questionnaire items. Perform at least two consecutive rounds of Delphi surveys among stakeholders invited by the working group of Core Outcome sets and Core Outcome measurement sets in NUTriology (COCONUT) to create the COS for future clinical trials of FSMP. COCONUT will invite patients, clinical dietitians, physicians, COS researchers, journal editors, FSMP manufacturers, and regulatory representatives to score each outcome from the aspects of importance, operability, independence, and cost. Hold a face-to-face or online consensus meeting (if necessary) to refine the COS content. Key stakeholders will be invited to attend the consensus meeting to discuss and reach an agreement regarding the final content of the COS [34].Fig. 2Flowchart of developing the COS of FSMP designed for T2DM ## Phase 1: Generation of a list of outcomes (information sources) SR of outcome selection We will conduct an SR to first identify the preliminary list of outcomes according to the principle of PICOS.PICOS principlesPatient/population/problem, PAs described in the scope section.(b)Intervention, IAs described in the scope section.(iii)© Comparator, CComparators can include blanks, standard enteral nutrition (EN) preparations, therapeutic diets, or nutrition education. The use of FSMP designed for T2DM is the only difference between the experimental and control groups.(iv)Outcome, OOwing to SAMR regulations, outcomes identified from SR will be conceptualized into 4 domains, i.e., safety, nutritional adequacy, special medical effects, and others.(e)Study design, S All original studies related to FSMP designed for T2DM will be enrolled, although RCTs will be preferred. If RCTs do not meet the PICOS criteria, the outcome information will be supplemented by SR, meta-analysis, nonrandomized concurrent controlled studies, and observational studies.[2]Information sources The information sources of the SR comprise (a) published studies, (b) regulatory documentation, and (c) qualitative interviews of stakeholders. Articles related to the subject will be retrieved from 6 medical databases (CNKI, Wanfang, VIP, PubMed, Ovid-Medline, and Cochrane Library). Regulatory documentation related to FSMP will be retrieved from the official SAMR websites and other international regulatory bodies. Online surveys using questionnaires will be performed to collect information from different stakeholders, such as patients, health professionals, FSMP manufacturers, and SAMR officers. Interview surveys will be used as an alternative if online questionnaires are not suitable for every interviewee. Pilot studies to check the readability of questionnaires will be undertaken to ensure that simple language is used and that nonprofessionals can understand the questions.[3]Search strategies and data selection Languages will be limited to Chinese and English. No date limitations. The combination of P and I in the PICOS requirements will be selected for the search strategy to avoid omissions. Keywords will contain MeSH terms and synonyms of FSMP, medical foods, EN preparations, and T2DM. Two trained postgraduate students will independently check bibliographies, extract information, evaluate qualities, and record processes. In case of differences between the two students’ decisions, another senior researcher who has relevant experience will evaluate the concerns and determine its inclusion.[4]Conceptualization of outcomes The China Food and Drug Administration (CFDA), later replaced in 2018 by SAMR after the First Session of the 13th National People’s Congress, published the Clinical Trial Quality Management Practices (for trial implementation) [12] in 2016, in which the observation outcomes were limited to the domains of (a) safety, (b) nutritional adequacy, and (c) special medical effects. The practice also asks qualified institutions to draw conclusions after thoroughly analyzing and explaining the 3 domains. Therefore, we intend to categorize relevant outcomes into the same 3 domains and another domain, (d) others, in case an outcome identified from the SR cannot be categorized under the first 3 domains. ## Phase 2: Prioritization of identified outcomes using multistakeholder Delphi surveys (consensus process 1) We will use a consensus process involving a sequential, multiround Delphi survey followed by a face-to-face or online consensus meeting to reach an agreement among multistakeholder groups on the final COS. A diverse representative sample of key stakeholders from the eastern, central, western, and northeastern regions of China will participate anonymously to avoid the effects of dominant individuals. The geographical regions, also called economic regions, were divided by the National Bureau of Statistics of China based on their social and economic development [35].Round 1(R1) In the first round, stakeholders will be approached by sending out a personalized e-mail with a link to an online questionnaire survey (WJX) [36] with the aid of DelphiManager (Chinese edition), a free software provided by ChiCOS and designed for the Delphi process. Stakeholders will receive background information on the rationale of the development of the COS and the preliminary list of outcomes. Then, they will be asked to provide their basic information, including their name, phone number, fields of expertise, institutions, and stakeholder group categories. Next, they will be suggested to use 9-point Likert scales to score the importance of every outcome in the preliminary list. A score of 1–3 points means not important, 4–6 means important, and 7–9 means critical. Additionally, they will be asked to indicate what outcomes they have newly added. In addition to the Likert scale, an option of “unable to score” is also available in the questionnaire.[2]Round 2 (R2) Stakeholders who participate in R1 will then be invited to undertake R2. The R2 questionnaire will contain what is retained from R1 (see the “Analysis” section) and an anonymized feedback report from R1 in the form of summary scores. There will be the option of transforming the distribution of the summary scores into histograms and inserting them into the questionnaire to avoid unnecessary modifications and help establish consensus [37]. If the score is inconsistent with the previous round, the reason for the inconsistency will be needed. ## Phase 3: Consensus meeting with key stakeholders to discuss and agree on the final COS (consensus process 2) Key stakeholders will be given priority to attend face-to-face or online consensus meetings. One of the COS developers in this study, the leader of the COCONUT steering committee, will chair the meeting, remind the attendees to discuss in simple language, and guide them to poll on every single outcome. The poll results will be simultaneously displayed on the screen in the conference room using a web-based poll program (WJX) [35]. COCONUT secretaries will record the audio of the meeting and transcribe it verbatim. Following the first round of the Delphi survey, outcomes will be categorized as “consensus in,” “consensus out,” or “no consensus” using the definitions in Table 1. Further discussions or meetings will be considered if an agreement is not reached. The same criteria to define consensus and retain/discard outcomes as in Table 1 will be used. Table 1Definitions of consensus in, consensus out, and no consensusCategoryConsensus inConsensus outNo consensusRound 1Any stakeholder group score as critical (7–9 points) ≥ $70\%$ and not important (1–3 points) < $15\%$All stakeholder group score as not important (1–3 points) by ≥ $70\%$ and critical (7–9 points) by < $15\%$Neither criterion of consensus in nor out are metAction 1Outcome retained for round 2Outcome discarded before round 2 (to be ratified at consensus meeting)Outcomes retained for round 2Round 2All stakeholder group score as critical (7–9 points) ≥ $70\%$ and not important (1–3 points) < $15\%$Any stakeholder group score not important (1–3 points) by ≥ $70\%$ and critical (7–9 points) by < $15\%$Neither criterion of consensus in nor out are metAction 2Outcome retained for consensus meetingOutcome discarded before consensus meeting (to be ratified at consensus)Outcomes retained for consensus meetingConsensus meetingAll stakeholder group score as critical (7–9 points) ≥ $70\%$ and not important (1–3 points) < $15\%$Any stakeholder group score not important (1–3 points) by ≥ $70\%$ and critical (7–9 points) by < $15\%$Neither criterion of consensus in nor out are metAction 3Outcome retainedOutcome discardedOutcomes retained for the next consensus meeting ## Sample size The expert panel was born from the steering committee. The stakeholders comprise the steering committee (professionals, stakeholder groups 1–3) and patients (nonprofessionals, stakeholder group 4). A 3:1 ratio of professionals to patient participants is considered appropriate, as the involvement of multiple professional stakeholder subgroups is warranted in the development of COS [34, 37]. Approximately 13–15 more stakeholders are added to account for the possibility of $10\%$ of stakeholders dropping out during the Delphi processes. Therefore, we aim to use purposive and snowball sampling to include approximately 100 professional participants and 30 patient participants, of whom 15–25 key stakeholders will be invited to attend the consensus meeting. Key stakeholders will comprise (a) steering committee members and (b) stakeholders who will have consecutively participated in all rounds of Delphi surveys. ## Outcome scoring/feedback Detailed information is described in phase 2. ## Missing data Maximizing completion To keep the dropout rate as low as possible (preferably less than $20\%$), the language of the questionnaires should be revised to a degree that all stakeholder groups can easily understand. Moreover, the reliability and validity of questionnaires will be tested in a small group of people before use (presurvey, approximately 15 professionals and 5 patients). Text boxes will be inserted at the end of questionnaires, where interviewers can express their thoughts freely, which will help gain more related information. To increase the response rate, questionnaires will be sent with a brief background introduction containing the aim of developing the minimum core set and an official reference letter at the beginning of the email. WeChat messages will be sent as notices at the same time an email is sent. The feedback report of the previous round will consist of a summary of what has been done, including retain outcomes, discard outcomes, and any other adjustments. We will then personalize the feedback report to avoid biases in which stakeholders can only receive information concerning themselves and their stakeholder group.[2]Data quality control Each round of questionnaires will be issued and retrieved by email using the DelphiManager (Chinese edition). Phone numbers and email addresses of the main researchers will be attached to the first page of the questionnaire to answer any questions that stakeholders may have while filling out the questionnaire. If some fields are missed or incorrectly filled, the corresponding stakeholders will be asked to refill them and ensure that each field is complete. If the score is inconsistent with the previous round, reasons should be attached to the changed score. The last edition of the email sent by the stakeholders in every round will be downloaded, printed, and coded, and both paper and electronic forms will be filed.[3]*Data analysis* DelphiManager (Chinese edition) and IBM SPSS 22.0 will be used for data analysis. Questionnaires with omitted values will not be included in the final analysis. ## 9-point or 3-point Likert scale Many COS studies choose a 9-point Likert scale as their score method, but there have been COS studies involving traditional Chinese medicine indicating that the 3-point Likert scale may be better suited to the Chinese language environment than the 9-point Likert scale [38]. If the 9-point Likert scale does not work well in the presurvey, i.e., scores of 2, 5, and 8 account for more than $80\%$ of all scores, we will consider replacing the 9-point Likert scale with a 3-point scale. ## Further round of Delphi surveys Another round (Rn, n ≥ 3) of the Delphi survey will be conducted if significant numbers, i.e., more than $50\%$ of outcomes, remain in the outcome pool after Rn-1. The methods in Rn will be identical to those in Rn-1. The outcomes remaining after Rn-1 and the feedback report from Rn-1 will be included in the Rn questionnaire. Stakeholders who have consecutively taken part in n-1 rounds of Delphi surveys will rescore again. The outcomes remaining after the final Delphi survey round will be taken forward to the consensus meeting. ## Streamlined scoring group COS developers, journal editors, COS users, regulators, and manufacturer representatives will be regarded as one scoring group when conducting statistics if we fail to invite more than 30 persons per subgroup. Patient representatives will be regarded as another scoring group. ## Alternative plan of consensus definition If the consensus definition in Table 1 failed to discard $10\%$ of outcomes in R1. We will consider starting an alternative plan (Table 2) or combining the consensus definition in Table 1 and the alternative plan. A detailed alternative plan description can be found in Additional file 1.Table 2Alternative consensus definitionsCategoryAdded/retainedDeleted directlyDeleted indirectlyRound 1≥ 2 stakeholders agree to add≥ 2 unqualified scales among M, FR, CVPOR ≥ $10\%$≥ 2 unqualified scales among M, FR, CV, POR, PIMR, PIDRM ≤ $60\%$ of the full mark or FR ≤ 0.2 in round 1< 2 unqualified scales among M, FR, CV, POR, PIMR, and PIDR< 2 stakeholders advise to deleteAction 1Outcome added for R2Outcome discarded before R2Outcome discarded before R2 (require expert panel’s approval), otherwise retained for R2Round 2Not allowed to newly add outcomes≥ 2 unqualified scales among M, FR, CVPOR ≥ $10\%$≥ 2 unqualified scales among M, FR, CV, POR, PIMR, PIDRM ≤ $80\%$ of the full mark or FR ≤ 0.3 in R2< 2 unqualified scales among M, FR, CV, POR, PIMR, PIDR< 2 stakeholders advise to deleteAction 2NullOutcome discarded before consensus meetingOutcome discarded before consensus meeting (require expert panel’s approval), otherwise retained for consensus meetingConsensus meetingAll stakeholder group score as critical (7–9 points) ≥ $70\%$ and not important (1–3 points) < $15\%$Any stakeholder group score not important (1–3 points) by ≥ $70\%$ and critical (7–9 points) by < $15\%$Neither criterion of consensus in nor out are metAction 3Outcome retainedOutcome deletedOutcomes retained for the next consensus meetingM arithmetic mean, FR full ratio, CV coefficient of variation, POR poor operability ratio, PIMR poor importance ratio, PIDR poor independence ratio ## Key applications This COS for clinical trials of FSMP will provide a multistakeholder agreed tool for regulators to use for the pre- and postmarket management of FSMP. Premarket, COS will act as a referring guide for manufacturers to produce FSMP in laboratories, allowing them to weed out formulations that do not show effect. The same COS will be used as an assessment criterion in institutions conducting clinical trials of FSMP. Regulators will not approve a product to be registered as FSMP unless its clinical trial reports show acceptable results of COS. Postmarket, real-world data and evidence of COS will help clinical dietitians and doctors choose the best and most suitable FSMP for their patients, providing them with the satisfaction as COS is also what they are concerned about. In summary, COS has the potential to act as a unified metric and communication tool from top to bottom. If this example of COS proves successful, we could potentially apply the experience to FSMPs designed for other diseases. When COS of FSMPs gains momentum, leading to a series of COSs, it may be time to develop technical specifications, guidelines, or standards. We believe that both the clinical nutrition expertise field and the FSMP industry will benefit from COS processes and become stronger in China. ## Strengths and limitations of this study To develop a COS for the premarket evaluation and postmarket monitoring of FSMP designed for specific diseases, we will combine and use typical methods—SR, Delphi, and consensus—in series, as this guarantees the most possibility of success. According to the latest SR conducted by COMET, $53\%$ of COS research used a mixed method, with Delphi combined with other methods being the most common ($54.2\%$), followed by consensus conferences or SR combined with other methods ($14.1\%$) [39]. Furthermore, our information sources are beyond SR. We will also use stakeholder qualitative interviews to generate a preliminary list of outcomes, providing a broader view of FSMP trial outcomes at the start of phase 2 and helping to avoid missing anything. We will develop COS with multistakeholder groups, including patients, clinical dietitians, physicians, FSMP manufacturers, and policymakers. These five groups are responsible for decisions related to the research and development, registration, production, supervision, and circulation chain of FSMP. Achieving consensus among them is challenging, but any agreement reached will refresh the cognition in the entire FSMP industry and its application in clinical nutrition. This COS may have limited applicability outside of China, as our study is not international. Further work is required to implement COS to FSMP designed for other medical states/diseases. ## Trial status This study is ongoing at phase 1 (SR has been finished) as planned by the version 1.0 protocol and is expected to be completed by the end of 2024. ## Supplementary Information Additional file 1: Detailed methods description of the alternative consensus definition plan. ## Patient and public involvement Patients or the public were not involved in the design, conduct, reporting, or dissemination plans of our research. Trained investigators will conduct interviews with patients. Patients who consecutively complete all rounds of Delphi surveys will have priority to be invited to the consensus meeting. ## References 1. Kirkham JJ, Gorst S, Altman DG, Blazeby JM, Clarke M, Tunis S. **Core Outcome Set-STAndardised Protocol Items: the COS-STAP statement**. *Trials* (2019.0) **20** 116. DOI: 10.1186/s13063-019-3230-x 2. Brown B, Roehl K, Betz M. **Enteral nutrition formula selection: current evidence and implications for practice**. *Nutr Clin Pract* (2015.0) **30** 72-85. DOI: 10.1177/0884533614561791 3. Elia M, Ceriello A, Laube H, Sinclair AJ, Engfer M, Stratton RJ. **Enteral nutritional support and use of diabetes-specific formulas for patients with diabetes: a systematic review and meta-analysis**. *Diabetes Care* (2005.0) **28** 2267-2279. DOI: 10.2337/diacare.28.9.2267 4. Jianqin S, Yanfang W, Xiafei C, Yanqiu C, Ying F, Xinyi Z. **An integrated intervention program to control diabetes in overweight Chinese women and men with type 2 diabetes**. *Asia Pac J Clin Nutr* (2008.0) **17** 514-524. PMID: 18818173 5. Stephen R, Vikkie AM, James L, Jianqin S. **Economic analysis of a diabetes-specific nutritional meal replacement for patients with type 2 diabetes**. *Asia Pac J Clin Nutr* (2010.0) **19** 1-7. PMID: 20199981 6. Ojo O, Brooke J. **Evaluation of the role of enteral nutrition in managing patients with diabetes: a systematic review**. *Nutrients* (2014.0) **6** 5142-5152. DOI: 10.3390/nu6115142 7. Sanz-Paris A, Álvarez Hernández J, Ballesteros-Pomar MD, Botella-Romero F, León-Sanz M, Martín-Palmero Á. **Evidence-based recommendations and expert consensus on enteral nutrition in the adult patient with diabetes mellitus or hyperglycemia**. *Nutrition* (2017.0) **41** 58-67. DOI: 10.1016/j.nut.2017.02.014 8. 8.SC. “Health China 2030” planning outline; 2016. Available from: http://www.gov.cn/xinwen/2016-10/25/content_5124174.htm. 9. 9.SAMRRegistration management methods of Food for Special Medical Purposes2016. *Registration management methods of Food for Special Medical Purposes* (2016.0) 10. 10.SAMRTechnical guidelines for clinical trials of specific whole nutrient food for special medical purposes2019. *Technical guidelines for clinical trials of specific whole nutrient food for special medical purposes* (2019.0) 11. 11.SAMRQuality management standard for clinical trials of Food for Special Medical Purposes (Trial)2016. *Quality management standard for clinical trials of Food for Special Medical Purposes (Trial)* (2016.0) 12. 12.CFDAClinical trial quality management practices (for trial implementation)2016. *Clinical trial quality management practices (for trial implementation)* (2016.0) 13. 13.SAMRSpecial food information query platform2016. *Special food information query platform* (2016.0) 14. Lei S, Dongyu M, Jie G, Zhiyong R, Yuan L, Jiajie Y. **Using Delphi method to preliminarily perfect the comprehensive evaluation index system of food for special medical purposes designed for type 2 diabetes mellitus patients**. *Chin J Evid Based Med* (2019.0) **20** 1012-1019 15. Dongyu M. *Preliminarily constructing an evaluation index system of the food for special medical purposes exampled by adult patients with type 2 diabetes mellitus based on evidence based medicine* (2019.0) 16. Macleod MR, Michie S, Roberts I, Dirnagl U, Chalmers I, Ioannidis JPA. **Biomedical research: increasing value, reducing waste**. *Lancet* (2014.0) **383** 101-104. DOI: 10.1016/S0140-6736(13)62329-6 17. Mingyan Z, Junhua Z, Liang D, Yonggang Z, Hongjie Z, Rui F. **Introduction of COS-STAD recommendations and its application of TCM**. *Chin J Evid Based Med* (2018.0) **18** 392-396 18. Ruijin Q, Min L, Songjie H, Tianmai H, Ya H, Jing C. **Interpretation of the COMET handbook (version 1.0) and its insight for developing core outcome sets in clinical trials of traditional Chinese medicine**. *Chin J Evid Based Med* (2017.0) **17** 1482-1488 19. Mingyan Z, Fengwen Y, Yue L, Junhua Z. **Core Outcome Set-STAndards for Reporting: the COS-STAR statement (Translated)**. *Chin J Evid Based Med* (2017.0) **17** 470-474 20. Mingyan Z. **Introduction of Core Outcome Set-STAndards for Reporting (COS-STAR)**. *Chin J Evid Based Med* (2017.0) **17** 857-861 21. Dongmei X, Junhua Z, Boli Z. **Approach to developing core outcome sets for clinical trials of traditional Chinese medicine**. *CJTCMP* (2014.0) **29** 1352-1355 22. Kirkham JJ, Davis K, Altman DG, Blazeby JM, Williamson PR. **Core Outcome Set-STAndards for Development: the COS-STAD recommendations**. *PLoS Med* (2017.0) **14** e1002447. DOI: 10.1371/journal.pmed.1002447 23. Mingyan Z, Junhua Z, Liang D, Yonggang Z, Fengwen Y, Rui F. **Development standard of core outcome sets: recommended by COS-STAD (translated)**. *Chin J Evid Based Med* (2018.0) **18** 753-757 24. Kirkham JJ, Gorst S, Altman DG, Blazeby JM, Williamson PR. **Core Outcome Set–STAndards for Reporting: the COS-STAR statement**. *PLoS Med* (2016.0) **13** e1002148. DOI: 10.1371/journal.pmed.1002148 25. Chengyu H. *Medical Nutriology* (2003.0) 26. 26.SAMRWork specifications for special food verification and evaluation of technical institutions2017. *Work specifications for special food verification and evaluation of technical institutions* (2017.0) 27. Williamson PR, Altman DG, Bagley H, Barnes KL, Blazeby JM, Brookes ST. **The COMET Handbook: version 1.0**. *Trials* (2017.0) **18** 280. DOI: 10.1186/s13063-017-1978-4 28. 28.ChiCOSChinese clinical trials core outcome set research center2020. *Chinese clinical trials core outcome set research center* (2020.0) 29. Wen H, Dongyu M, Jie G, Ziqi Z, Fangfang P, Xuemei L. **Evidence-based nutrition and national health promotion**. *Chin J Evid Based Med* (2020.0) **20** 851-861 30. Jie G, Lei S, Dongyu M, Xiaofan J, Yu X, Yuan L. **Initiative evidence-based establishment of an indicator pool for clinical evaluation of nutritional and functional food for the elderly with sarcopenia**. *Chin J Mult Organ Dis Elderly* (2021.0) **20** 731-737 31. 31.WCHSCUClinical nutrition specialty alliance of West China Hospital, Sichuan University/Western Medical Nutrition Alliance2018. *Clinical nutrition specialty alliance of West China Hospital, Sichuan University/Western Medical Nutrition Alliance* (2018.0) 32. 32.CGSNChinese gerontological society of nutrition and food safety association2015. *Chinese gerontological society of nutrition and food safety association* (2015.0) 33. 33.Cochrane. Cochrane China Center 1999. Available from: https://china.cochrane.org/. 34. Avery K, Blazeby J, Wilson N, Macefield R, Cousins S, Main B. **Development of reporting guidance and core outcome sets for seamless, standardised evaluation of innovative surgical procedures and devices: a study protocol for content generation and a Delphi consensus process (COHESIVE study)**. *BMJ Open* (2019.0) **9** e029574. DOI: 10.1136/bmjopen-2019-029574 35. 35.NBSDivision method of East-West Central and Northeast China2011. *Division method of East-West Central and Northeast China* (2011.0) 36. 36.WJXQuestionnaire Star2022. *Questionnaire Star* (2022.0) 37. Kirkham JJ, Gorst S, Altman DG, Blazeby J, Clarke M, Devane D. **COS-STAR: a reporting guideline for studies developing core outcome sets (protocol)**. *Trials* (2015.0) **16** 373. DOI: 10.1186/s13063-015-0913-9 38. Ruijin Q. *Construction of core outcome set in clinical research of traditional Chinese Medicine: taking non-valvular Atrial Fibrillation as an example* (2018.0) 39. Gargon E, Gorst SL, Williamson PR. **Choosing important health outcomes for comparative effectiveness research: 5th annual update to a systematic review of core outcome sets for research**. *PLoS One* (2019.0) **14** e0225980. DOI: 10.1371/journal.pone.0225980
--- title: Differences in child and adolescent exposure to unhealthy food and beverage advertising on television in a self-regulatory environment authors: - Monique Potvin Kent - Julia Soares Guimaraes - Meghan Pritchard - Lauren Remedios - Elise Pauzé - Mary L’Abbé - Christine Mulligan - Laura Vergeer - Madyson Weippert journal: BMC Public Health year: 2023 pmcid: PMC10037770 doi: 10.1186/s12889-023-15027-w license: CC BY 4.0 --- # Differences in child and adolescent exposure to unhealthy food and beverage advertising on television in a self-regulatory environment ## Abstract ### Background Food and beverage promotion is a contributor to children’s dietary behaviours, and ultimately, downstream health consequences. Broadcast television remains an important source of such advertising. The objective of this study was to examine and compare children and adolescent’s exposure to food advertising on television in Canada over an entire year in a self-regulatory environment. ### Methods Television advertising data for 57 selected food and beverage categories were licensed from Numerator for 36 stations in Toronto, for 2019. The estimated average number of advertisements viewed by children aged 2–11 and adolescents aged 12–17 was determined overall, by food category, and by marketing technique. The healthfulness of advertisements was also assessed using Health Canada’s Nutrient Profile Model. ### Results Overall in 2019, children viewed 2234.4 food ads/person/yr while adolescents viewed 1631.7 ads, exposure for both groups stemmed primarily from stations with general appeal, and both age groups were exposed to a range of powerful marketing techniques. Exposure to advertising for restaurants, snacks, breakfast food and candy and chocolate was high among both age groups and the healthfulness of most advertised products was considered poor. Adolescents were exposed to $36.4\%$ more food products classified as unhealthy, had higher exposure to all marketing techniques examined, and were exposed to substantially more child-related marketing techniques compared to children. ### Conclusion Children and adolescents were heavily exposed to food advertisements on television in 2019. Despite current self-regulatory policies, children’s exposure to unhealthy food and beverages remains high. Differences in exposure to food advertisements by food category and healthfulness may suggest that adolescents are being disproportionately targeted by food companies as a result of self-regulatory marketing restrictions. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12889-023-15027-w. ## Introduction Obesity among youth is an issue worldwide [1–3]. In Canada, obesity in children and adolescents has increased significantly over time, and data from 2015 indicate that $10.4\%$ of children aged 5 to 11 years and $13.8\%$ of adolescents aged 12–17 years have obesity [4]. Poor dietary intake contributes to obesity and youth diets are characterised by high levels of ultra-processed foods and beverages that are elevated in added fats, sugar, and salt [5]. Evidence from systematic reviews has demonstrated that exposure to the marketing of unhealthy food and beverage products can contribute to obesity and other chronic diseases associated with poor diet by influencing food preferences and food intake [6, 7]. Although children and adolescents are spending increasing amounts of their leisure time on digital devices, broadcast television remains an important source of food marketing exposure among young people [8, 9]. According to $\frac{2017}{18}$ data, children in Canada spend an average of 17.3. hours/week watching television, while adolescents spend 13.9 hours/week viewing this media [10]. Much of the literature to date on food advertising on television has focused on school-aged children. One study involving 22 countries demonstrated that advertisements containing nutrient-poor foods were more common during popular children’s viewing times, and that advertisements for these foods frequently contained persuasive and child-directed messaging [11, 12]. This is especially problematic since children lack the cognitive ability to comprehend persuasive marketing techniques [13–15]. Indeed, research shows that the persuasive techniques used in food and beverage advertising to children can alter attitudes, preferences, and food consumption which can ultimately result in poorer health outcomes [7, 16]. Adolescents are also a key target market for the food and beverage industry as they are more independent and have greater purchasing power than children [17]. Neurocognitively, they are also vulnerable due to their overactive reward pathways, poor impulse control as well as heightened peer influence [18]. A recent study found that Canadian television stations aired 3.3 food advertisements per hour on programs designated by broadcasters as targeted at adolescents (ages 12–17) compared to 1.5 food advertisements per hour on programs targeted at children (ages 2–11) [8]. Another recent study examining Canadian adolescent exposure to food and beverage advertising found that despite the decreases in exposure between May 2011 and 2016, adolescents aged 12–17 living in Toronto were nonetheless exposed to over 150 food and beverage advertisements on television in May 2016 [19]. In light of the evidence base on the harmful effects of food marketing, countries such as the UK, Mexico, and Chile have introduced government policies to protect children from unhealthy television food advertising [20]. With the exception of the UK, few countries have extended these protections to adolescents despite their vulnerability to unhealthy food advertising. Other countries such as the United States, Australia, and Canada (with the exception of the province of Quebec) have taken a self-regulatory approach to restricting food advertising and here again, adolescents have been excluded from protections [21, 22]. In Canada, advertising to children under age 12 is self-regulated and 15 food companies have committed to restricting unhealthy food advertising to children in a variety of media and settings through the Children’s Food and Beverage Advertising Initiative [18]. Though this initiative has been shown to be ineffective at protecting children, much of the research evaluating this initiative has been conducted using television data collected over a one-month period [12, 23]. Research on adolescent exposure to food advertising is also very limited and based on monthly data [23]. Little is known about whether the current self-regulatory policy confers any protection to adolescents or whether, in the alternative, adolescents are exposed to higher levels of food advertising given that the food and beverage industry has not committed to protecting this age group. Given this critical gap in data, comparing child and adolescent exposure to food advertising is essential to inform food marketing restrictions. A new bill to restrict food marketing to children was introduced in the Canadian House of Commons in February 2022, Bill C-252, and once again, adolescents have been overlooked [24]. The purpose of the current study was to determine whether there are differences in child and adolescent exposure to food and beverage advertising on television and to marketing techniques used in food advertising. It was hypothesized that children would be exposed to fewer food and beverage advertisements compared to adolescents. ## Data sample Data on television viewership and advertisements airing from January to December 2019 were licensed from Numerator, a marketing information and advertising intelligence company. The advertising data included 57 selected food categories broadcast on 36 television stations in Toronto. The 57 food categories were selected from 112 possible categories because they were either known to be advertised heavily to children and adolescents [11] and/or because of their contribution to children’s food intake and diet quality, including both less healthy and healthier product categories. These food categories were then aggregated into 13 food categories which included bread; sweet baked goods/desserts; candy and chocolate; breakfast food; dairy; condiments; entrees and meat (including fish, poultry, and meat products); fruit and vegetables; beverages (excluding milk and water); miscellaneous; snacks; water; and restaurants. See Additional file 1 for a detailed breakdown of all food categories and their definitions. Toronto was selected as it is the largest media market in Canada and has the largest panel size of children aged 2–11 years ($$n = 175$$) and adolescents aged 12–17 years ($$n = 106$$). Television audience viewership data are provided by Numerator though these data are collected by Numeris, an organization that maintains a panel and collects audience viewership data from a stratified random sample of households proportional to the population. Each person on Numeris’ panel wears a portable device which captures the stations to which the television set is tuned when each panel member is near the television. Data are then weighted according to demographic variables including age, sex, and household size. This enables the examination of advertising viewership by age group (e.g., children aged 2–11 years and adolescents aged 12–17 years). All stations captured for the Toronto market ($$n = 36$$) were examined and all 24 hours of television per day were analyzed. ## Frequency of food and beverage advertisements The frequency of food and beverage advertisements was extracted from AdQuest, an advertising platform licensed from Numerator. These data included the number of unique advertisements and the weighted frequency of advertisements. An advertisement was considered unique if it differed from others in terms of content, language, or duration. To calculate the weighted frequency, the number of products in an advertisement was multiplied by the number of times an advertisement was broadcasted. For example, if there were 2 products in an advertisement broadcasted 500 times, the weighted frequency would be 1000. If the number of unique products exceeded 3, then it was only counted as having 3 products (i.e., 4 products in an advertisement broadcasted 500 times would have a weighted frequency of 1500). The limit was capped at 3 products to remain consistent with Numerator methodology. ## Exposure to food and beverage advertisements Numerator expresses exposure as “ratings”, which is the approximate percentage of a population that has viewed an advertisement. Ratings summed across a defined period of time (in this case 24-hour programming broadcast from January to December 2019) are known as gross rating points (GRPs). GRPs are calculated by dividing the total impressions by the total population of the media market and multiplying by 100 (Impressions / Population × 100). GRPs were determined by age group and were divided by 100 to calculate the estimated average exposure to food and beverage advertising, overall or by food category. GRPs will hereinafter be referred to as “exposure”, however note that this measure is approximate and not an absolute measure of child and adolescent exposure. ## Content analysis of food marketing techniques A content analysis of all advertisements aired on the captured stations in 2019 in Toronto was conducted. Excluding duplicates and data that was missing due to technical problems extracting advertisements from the AdQuest platform ($$n = 22$$), a total of 1365 unique ads were coded (total frequency = 1,670,912, $97.1\%$ of total food advertising frequency). Marketing techniques as described in Table 1 were recorded as present or absent and were counted once per ad, regardless of how many products were featured. The advertisements were analyzed by three trained research assistants using a previously developed coding manual [25]. During training, interrater reliability of 0.93 was calculated based on practice samples. Advertisements were then coded by randomly distributing unique advertisements among the three coders, and a researcher oversaw the data looking for any inconsistencies, which were settled via consensus. Table 1Marketing techniques examinedMarketing techniqueDescriptionChild actorsMain characters in the advertisement are children (0–12 years), or have childlike voicesChild productsA product that appeals to children due to the type/nature of product (e.g., candy, compartment snacks), its shape, colour and/or designChild-appealing charactersCartoon characters, animals, or imaginary, fantasy, or virtual creaturesChild languageThe level of language is that commonly used by children or language is directed at children (e.g., “Hey Kids”)Child-appealing special effectsLettering, colours, special effects, animation, music, songs or jingles that appeal to childrenChild themesChild-appealing themes linked to fantasy, magic, mystery, suspense, adventure, or virtual worlds are featuredUse of spokes-charactersE.g., brand-owned characters such as Tony the TigerParent-child situationsSituations that play on the parent-child relationship or other authority-based relationship (i.e., coach-child or teacher-child)Use of licensed charactersE.g., Dora the Explorer or SpidermanCross-promotionsCross-promotions to movies or television shows watched by childrenChild incentivesFree gifts including toys, books, collectibles aimed at childrenTeen actorsYouth (12–17 years) were prominently featuredTeen languageE.g., “ hey dude”Teen musicE.g., rapTeen themesThemes based on adolescent activities or interests (e.g., socializing, school-related activities like dances, sports or extreme sports/risk-taking behavior, adolescent-directed humor, freedom, popular music/culture, video games)Teen incentivesE.g., gift card to movie theatreTeen humourE.g., boy wiping out on a skateboardContest/sweepstakesPrizes are given away at no charge to the participants and there is a competition (not every consumer will win).Celebrity endorsementsE.g., musical groups, film stars, athletes, etc. Health claimsHealth or nutrition claimsPrice promotionsPrice related premiums or rebates (e.g., bonus offers, calls to action to encourage purchase)Call to action - onlineSending viewers online to access brand website, app, etc. ## Nutritional analysis The nutritional information for each product featured in an advertisement was primarily collected using the 2017 Food Label Information Program (FLIP), a large database containing food label information for over 17,000 Canadian products from three grocery retailers (Metro, Sobeys, and Loblaws) [26]. For restaurant and fast-food items, the 2016 Menu-FLIP with over 12,000 restaurant food items was used [27]. If products were not in FLIP or Menu-FLIP, the nutrition information was obtained from 1) the company’s Canadian website, 2) the product’s Nutrition Facts table found online, or 3) the company’s American website, or 4) a similar product from the Canadian Nutrient File was substituted if the original product could not be found. Each advertisement was further classified as “unhealthy” or “healthy” based on the Nutrient Profile Model (NPM) by Health Canada designed to identify products that should not be marketing to children [28]. This classification is based on nutrient thresholds for added fat, sugar and salt. If any product with added fat, sodium and/or sugar within advertisements exceeded the thresholds, then the entire advertisement was deemed “unhealthy”. See Additional file 2 for the detailed criteria outlined by Health Canada. Note that only products were included in the analysis; data were not collected for brand advertisements (e.g., advertisement featuring no identifiable food products). ## Data analysis The frequency of advertisements and exposure to advertisements were determined by media market, age group, and by food category. The frequencies of marketing techniques and proportion of advertisements that were “unhealthy” were calculated. The relative and absolute differences between children and adolescent’s exposure to food and beverage advertising were also calculated, using adolescents as the comparator group. ## Results Overall, a total of 1,720,763 food advertisements were broadcast across 36 stations in Toronto in 2019 as shown in Table 2. Each child viewed 2234.4 food advertisements on these stations in 2019, while each adolescent in Toronto viewed 1631.7 food advertisements over the entire year. In relative and absolute terms, adolescents viewed $27.0\%$ less or 602.7 fewer advertisements in 2019 compared to children. Children’s greatest exposure was on Citytv (257.4 ads/person/year), YTV (198.8 ads/person/year), CTV (198.3 ads/person/year), SportsNet Ontario (176.8 ads/person/year), and Global (140.8 ads/person/year.) Adolescents’ greatest exposure was from CTV (142.1 ads/person/year), YTV (133.0 ads/person/year), Citytv (125.0 ads/person/year), Global (112.7 ads/person/year), and TSN4 (94.3 ads/person/year).Table 2Exposure to food products advertised on 36 stations in Toronto in 2019StationsFreq (%)Ads/person/yearChildrenAds/person/yrAdolescentsAds/person/yrAbsolute differenceChildren vs adolescentsRelative difference%CBC42,672 (2.5)64.346.8−17.5−$27.2\%$CBC News Network24,120 (1.4)$7.611.33.748.7\%$CHCH18,756 (1.1)14.412.9−1.5−$10.4\%$Citytv56,887 (3.3)257.4125.0− 132.4− $51.4\%$CMT71,022 (4.1)$19.620.00.42.0\%$CP245229 (0.3)29.326.7−2.6−$8.9\%$CTV45,295 (2.6)198.3142.1−56.2−$28.3\%$CTV 247,099 (2.7)69.646.1−23.5−$33.8\%$CTV Comedy Channel56,703 (3.3)138.580.6−57.9− $41.8\%$CTV Drama Channel52,341 (3.0)$20.932.811.956.9\%$CTV Life Channel76,656 (4.5)16.910.0−6.9−$40.8\%$CTV Sci Fi Channel61,568 (3.6)84.848.5−36.3−$42.8\%$Discovery Channel45,966 (2.7)43.830.1−13.7−$31.3\%$Disney Channel26,168 (1.5)52.342.6−9.7−$18.5\%$DTour59,942 (3.5)$6.318.512.2193.7\%$Food Network71,387 (4.1)77.170.3−6.8−$8.8\%$Global53,986 (3.1)140.8112.7−28.1−$20.0\%$HGTV37,133 (2.2)69.240.4−28.8−$41.6\%$History50,800 (3.0)$30.135.65.518.3\%$Investigation Discovery14,677 (0.9)$2.510.27.7308.0\%$MTV72,599 (4.2)35.225.7−9.5−$27.0\%$Much84,728 (4.9)$47.063.116.134.3\%$National Geographic13,592 (0.8)$4.35.81.534.9\%$OLN70,146 (4.1)$12.821.78.969.5\%$Omni50,466 (2.9)2.00.5−1.5−$75.0\%$Showcase61,333 (3.6)69.047.2−21.8−$31.6\%$Slice59,398 (3.5)$12.021.69.680.0\%$Sportsnet 36058,678 (3.4)19.217.9−1.3−$6.8\%$SportsNet Ontario49,654 (2.9)176.889.3−87.5− $49.5\%$Teletoon67,984 (4.0)84.843.0−41.8−$49.3\%$The Weather Network25,117 (1.5)$3.78.44.7127.0\%$TSN211,405 (0.7)4.93.2−1.7−$34.7\%$TSN445,981 (2.7)157.194.3− 62.8−$40.0\%$VisionTV11,061 (0.6)$1.83.21.477.8\%$W Network52,594 (3.1)$61.690.428.846.8\%$YTV67,530 (3.9)198.8133.0−65.8−$33.1\%$Total1,720,673 (100.0)2234.41631.7−602.7−$27.0\%$Average by station47,79662.145.3−16.7−$27.0\%$Source: Numerator, 2019. Analysis based on the 57 selected food categories ## Differences in children and adolescent’s exposure to food advertising by food category The most frequently advertised food categories in 2019 were restaurants ($49.1\%$ of advertisements), snacks ($9\%$), candy and chocolate ($9\%$), and dairy ($8.4\%$) (Table 3). Both children and adolescents were most exposed to advertisements for restaurants (1145.5 and 813.6 ads/person/year, respectively) and snacks (204.2 and 149.6 ads/person/year respectively). Additionally, children were also highly exposed to advertising for breakfast food (188.3 ads/person/year), candy and chocolate (161.6 ads/person/year) and beverages 117.5 ads/person/year) while adolescents were highly exposed to dairy (136.4 ads/person/year), breakfast food (132.4 ads/person/year) and candy and chocolate (122.6 ads/person/year). Child and adolescent exposure to other categories, such as water, fruits and vegetables, and bread were markedly lower. Adolescents were exposed to less advertisements compared to children across all food categories; in relative terms this was most notable for restaurants (− $29\%$), fruits and vegetables (− $31.8\%$), and breakfast food (− $29.7\%$). In absolute terms, the greatest negative differences between children and adolescent’s exposure were for restaurants (− 331.9 ads/child), snacks (− 54.6 ads/child), and breakfast food (− 55.9 ads/child).Table 3Exposure to food products advertised on 36 stations in Toronto by food category in 2019Food categoryFreq (%)Ads/person/yearChildrenAds/ person /yrAdolescentsAds/ person /yrAbsolute differenceChildren vs adolescentsRelative difference%Bread12,485 (0.7)17.413.7−3.7−$21.3\%$Sweet baked goods/desserts61,723 (3.6)59.844.6−15.2−$25.4\%$Candy and chocolate154,970 (9.0)161.6122.6−39.0−$24.1\%$Breakfast food105,882 (6.2)188.3132.4−55.9−$29.7\%$Dairy145,342 (8.4)170.1136.4−33.7−$19.8\%$Condiments21,944 (1.3)39.330.9−8.4−$21.4\%$Entrees59,815 (3.5)66.553.0−13.5−$20.3\%$Fruits/vegetables26,866 (1.6)37.725.7−12.0−$31.8\%$Beverages89,143 (5.2)117.587.1−30.4−$25.9\%$Miscellaneous70,608 (4.1)69.352.6−16.7−$24.1\%$Snacks155,323 (9.0)204.2149.6−54.6−$26.7\%$Water14,346 (0.8)13.310.8−2.5−$18.8\%$Restaurants844,224 (49.1)1145.5813.6−331.9−$29.0\%$Total1,720,673 (100.0)2234.41631.7−602.7−$27.0\%$Source: Numerator, 2019. Analysis based on the 57 selected food categories ## Differences in children and adolescent’s exposure to marketing techniques Children and adolescents’ exposure to examined marketing techniques is presented in Table 4. Overall, calls to action ($34.7\%$ of all advertisements), health appeals ($32.3\%$), and child-appealing products ($30.3\%$) were the most frequently featured marketing techniques over the entirety of 2019 in Toronto. Children and adolescents were heavily exposed to similar marketing techniques, and exposure was highest for both age groups to calls to action (547.1 and 764 ads/person/year for children and adolescents respectively), health appeals (526.7 and 720.4 ads/person/year for children and adolescents respectively), child-appealing products (486.4 and 664.3 ads/person/year for children and adolescents respectively), and child-appealing special effects (432.1 and 586.6 ads/person/year for children and adolescents respectively). Adolescents had higher exposure to all marketing techniques examined compared to their younger counterparts. The greatest relative differences in exposure were for adolescent humour (+ $44.7\%$), adolescent music (+ $42\%$), and adolescent language (+ $44.4\%$) while the greatest absolute differences were for call to action (+ 216.9 ads/child), health appeal (+ 193.7 ads/child), and child-appealing product (+ 177.9 ads/child).Table 4Exposure to food products advertised on 36 stations in Toronto by marketing techniques in 2019Marketing techniquesFreq (%)Ads/person/yearChildrenAds/ person /yrAdolescentsAds/ person /yrAbsolute differenceChildren vs. adolescentsRelative difference%Child actor339,343 (20.3)$330.2442.8112.634.1\%$Child-appealing product506,107 (30.3)$486.4664.3177.936.6\%$Child-appealing characters344,919 (20.6)$367.5500.0132.536.1\%$Child language93,567 (5.6)$107.6150.342.739.7\%$Child-appealing special effects426,094 (25.5)$432.1586.6154.535.8\%$Child themes230,821 (13.8)$247.9335.387.435.3\%$Use of spokes-characters323,490 (19.4)$339.9462.8122.936.2\%$Use of licensed characters10,138 (0.6)$9.912.93.030.3\%$Cross-promotions53,760 (3.2)$52.672.119.537.1\%$Child incentives18,615 (1.1)$18.624.15.529.6\%$Adolescent actor229,355 (13.7)$223.7302.078.335.0\%$Adolescent language37,575 (2.2)$49.371.221.944.4\%$Adolescent music30,280 (1.8)$43.862.218.442.0\%$Adolescent themes306,385 (18.3)$304.4415.0110.636.3\%$Adolescent incentives5289 (0.3)$6.98.11.217.4\%$Adolescent humour33,527 (2.0)$43.462.819.444.7\%$Contest/sweepstakes59,618 (3.6)$63.681.017.427.4\%$Celebrity endorsement77,925 (4.7)$83.5110.527.032.3\%$Parent-child situations316,086 (18.9)$281.5371.590.032.0\%$Health appeal538,910 (32.3)$526.7720.4193.736.8\%$Price promotion353,754 (21.2)$331.6461.9130.339.3\%$Call to action579,213 (34.7)$547.1764.0216.939.6\%$Source: Numerator, 2019. Analysis based on the 57 selected food categories ## Differences in children and adolescent’s exposure to food advertising by nutritional content A greater proportion ($91.3\%$) of the advertisements broadcast in 2019 in Toronto were unhealthy compared to those considered healthy ($8.7\%$) (Table 5). Children’s exposure to food advertisements that were classified as unhealthy was 760 ads/person/year while adolescent exposure was to such advertisements was markedly higher at 1036.7 ads/person/year. Adolescents viewed $36.4\%$ or 276.7 more advertisements per person that were unhealthy compared to children. Table 5Exposure to food products advertised on 36 stations in Toronto by healthfulness in 2019NPM classificationAds/person/yearFreq (%)ChildrenAds/person/yrAdolescentsAds/ person /yrAbsolute differenceChildren vs. adolescentsRelative difference%Healthy74,617 (8.7)$64.390.426.140.6\%$Unhealthy783,855 (91.3)$760.01036.7276.736.4\%$Source: Numerator, 2019. Analysis based on the 57 selected food categories on products where nutritional information was available ## Discussion Overall, our results indicate that while children in Toronto were exposed to higher levels of food advertising than adolescents and exposed to a myriad of marketing techniques in 2019, adolescents’ exposure to unhealthy food advertising and marketing techniques was more than $30\%$ greater than children. ## Differences in child and adolescent exposure This study revealed that both children and adolescents were heavily exposed to food advertising on television in Toronto in 2019. Though it was hypothesized that children would be exposed to fewer food and beverage advertisements compared to adolescents, our results showed the opposite. Children aged 2–11 years were exposed to 2234 food advertisements on the examined stations or 6.1 ads per day while adolescents aged 12–17 years viewed $27\%$ fewer ads and were exposed to 1631.7 food advertisements or 4.5 ads per day. This is consistent with past research in Canada that showed higher rates of food advertising in programming targeted at children [8, 29]. Similar trends have been noted in the United States, where a study found that in 2017, adolescents viewed an average of 9.4 ads per day, while children viewed an average of 10 ads per day [30]. The noted exposure differences between age groups may in part, be due to changes in media consumption among adolescents, with less time being spent watching broadcast television among adolescents aged 12 to 17 years compared to children aged 2 to 11 years [10]. Other research in Canada has shown that there was a decline in adolescent exposure to food advertising on television between May 2011 and May 2016 [19]. Given decreased adolescent broadcast television viewership, the food and beverage industry may be re-directing their advertising dollars from television to digital media where teens are omnipresent in order to reach this market [31, 32]. Both children and adolescents were exposed to similar food categories. Restaurants, snack foods, breakfast foods, and candy and chocolate were figured prominently, while exposure to other categories, such as fruits and vegetables or water, was markedly lower. Restaurants (fast food and non-fast food) accounted for the largest source of advertising exposure relative to other food categories for both children and adolescents, with children exposed to 1145.5 advertisements (more than 3 ads/day) while adolescents were exposed to 813.6 advertisements (2.2 ads/day) in 2019. This high level of restaurant advertising is concerning from a public health perspective as experimental research has shown that exposure to fast food advertising induces reward-related neural pathways in adolescents and contributes to an overall greater intake of fast food [33]. Consuming fast food has also been linked with poor health outcomes in children and adolescents including weight gain/obesity, increased risk of cardiovascular disease, difficulties in insulin functioning, and diabetes [34]. Furthermore, research conducted in the UK has indicated that eating at full-service restaurants contributes to excessive caloric intake, with many meals from these types of restaurants exceeding calorie counts of meals from fast food restaurants [35]. High levels of restaurant advertising on television could be stymied by government policy. To date, only one restaurant manufacturer participates in the self-regulatory Children’s Food and Beverage Advertising Initiative (CAI) [21]. Government legislation that restricts unhealthy food marketing to children and adolescents would level the playing field here by applying to all food and beverage companies including all fast food and non-fast food restaurants. Though children were exposed to more food advertising overall, adolescents were exposed to $36.4\%$ more unhealthy food advertisements in 2019. Given that unhealthy food advertising has been causally linked to diet, this is worrisome [36]. Adolescent diets are characterized by high consumption of sugar, fats, and salts and low consumption of fruits and vegetables [37, 38]. Such results beg the question of whether focusing marketing restrictions exclusively on children is adequate when adolescents are clearly being bombarded with unhealthy food advertising on television. Here, the UK has taken the lead as their television advertising restrictions apply to children under the age of 16 [39]. Our results suggest that Canada should seriously consider following this example. ## Child and adolescent exposure to marketing techniques Advertising impact is a function of both exposure and power (i.e. the design and execution of the advertising message) [40]. Children and adolescents in Toronto were both highly exposed to a broad range of powerful marketing techniques in 2019 though overall, adolescents had higher exposure (> $30\%$) to all marketing techniques examined compared to their younger counterparts. Most notably, the highest exposure across both age groups was for calls to action, health appeals, the use of child-appealing products, and child-appealing special effects. Advertisements viewed by children and adolescents also commonly featured child and teen actors and themes and child-appealing characters such as cartoons. Children and adolescents viewed 340 and 463 exposures respectively to spokes-characters in television food advertising and 83 and 111 exposures respectively to celebrity endorsements in 2019. Both of these techniques have been shown to be particularly appealing to both children and adolescents [41]. The frequent inclusion of such marketing techniques in television advertisements to children suggests the Children’s Food and Beverage Advertising Initiative (CAI) is not effectively preventing the use of most marketing techniques intended to appeal to children and that much remains to be done to reduce children’s exposure to these techniques. The fact that adolescents were exposed to these marketing techniques at a higher level than children suggest they are being heavily targeted. The food marketing literature is evolving to recognize that adolescents’ ability to understand and recognize persuasive appeals may not provide an adequate defense against the harmful impact of unhealthy food advertisements [42]. Again, governments need to consider extending protections to this age group. ## Policy implications Overall, our results further reinforce previous evidence which indicates that self-regulatory policies are not efficiently reducing children’s exposure to unhealthy food advertising on television or the power of such marketing [40, 43, 44]. The CAI has been previously criticized for its low participation rates, inadequate definition of advertising, high audience thresholds to trigger pledges, and weak nutritional criteria [8, 44]. In the summer of 2021, a new industry-wide voluntary food code, the Food and Beverage Advertising Code was announced and is to be implemented by the summer of 2023 [45]. This new code applies to children under age 13 and is a poor imitation of the Quebec Consumer Protection Act which restricts all commercial advertising directed to children under the age of 13. In particular, this new industry code excludes many media forms, and does not explicitly outline how child targeting will be determined, provide nutritional criteria for in-store and out of home meals nor include financial penalties for infractions. With the growing body of literature pointing to the ineffectiveness of self-regulation of food marketing to children, Bill C-252 which was recently introduced in the Canadian federal House of Commons, is a more viable option given the breadth of its media coverage in terms of protecting children from unhealthy food marketing across media and settings. This bill would restrict unhealthy food marketing to children under age 13 in a variety of media and child settings as well as restrict the use of marketing techniques commonly used to target children [24]. The results of this study can be used as baseline data to evaluate the effectiveness of the new potential Canadian law and industry code. The results of this study also suggest a need to include adolescents in food marketing policy efforts. While children are widely recognized as a priority age group for food marketing restrictions, our results point to the insufficiency of regulation solely targeting children [46, 47]. The levels of exposure observed among adolescents in this study represent a significant loophole in current self-regulatory policies and in the new proposed Canadian Bill C-252. As previously described, adolescents face a unique set of conditions given their neurodevelopmental stage, increasing independence and financial resources, and greater susceptibility to peer influence [18, 33]. To compensate for profits lost to food marketing restrictions to children, companies may pivot to target adolescents by designing food advertisements that capitalize on these unique developmental vulnerabilities [48]. The new Bill C-252 proposes a review 5 years post-implementation of the impact of the new legislation on unhealthy food advertising directed at adolescents aged 13–16 years [24]. Continued monitoring of both children and adolescents’ exposure to food advertising on television is warranted to inform future policy efforts in Canada. ## Limitations Although this is the first study to compare child and adolescent exposure to food marketing and its associated marketing techniques using licensed commercial full-year data across a large number of stations, some limitations should be noted. First, the exposure data from this study can only be considered as a proxy measure since the portable devices used to measure exposure only captured the television stations being broadcasted. As a result, it cannot be stated with certainty that the advertisements were viewed by participants, only that the participants were near their televisions at the time of broadcast. Another limitation is that this study only presents data from one market (Toronto) and though it is the largest media market in Canada, it may not be representative of the entire country. Furthermore, although we were able to compare children and adolescents, other comparisons by demographic characteristics (socioeconomic status, sex, race) were not included, either due to not having sufficient sample sizes (i.e., sex) or simply lack of data availability (i.e., race, socioeconomic status). There were also several key limitations linked with the nutritional data collected in this study. We were unable to collect nutritional information for $51.4\%$ of the advertisements. Particularly products from brand advertisements, sit-down restaurants and new seasonal products that were not included in the FLIP databases and could not be derived from other sources. Sit-down restaurants in particular did not provide complete data (serving sizes or complete nutrient information) for available products on their websites which precluded a nutritional analysis. Nutritional information could also not be collected on brand advertisements as these ads do not feature any food products. Since FLIP and Menu FLIP data were collected in 2017 and 2016, respectively, the nutritional data collected did not reflect any product reformulation that may have taken place in the last few years. As a result, some products may have been misclassified according to healthfulness. ## Conclusion This study has presented a unique perspective on the differences in exposure to food and beverage advertising among children and adolescents. Despite lower exposure among adolescents compared to children overall, both children and adolescents were exposed to unhealthy food categories and powerful marketing techniques. These data highlight the failure of current self-regulatory policies to protect children and the lack of consideration given to the protection of adolescents. Given the unique developmental vulnerabilities among both children and adolescents, government policies that protect all young people from the dangers of food and beverage advertising must be developed. ## Supplementary Information Additional file 1. Numerator definition of food categories. Additional file 2. Health Canada Nutrient Profile Model Thresholds. ## References 1. 1.Public Health Agency of CanadaChildhood obesity2016. *Childhood obesity* (2016.0) 2. 2.OECD. Obesity Update 2017. [cited 2021 Jul 10]. Available from https://www.oecd.org/els/health-systems/Obesity-Update-2017.pdf. 3. Janssen I. **The public health burden of obesity in Canada**. *Can J Diabetes* (2013.0) **37** 90-96. DOI: 10.1016/j.jcjd.2013.02.059 4. 4.Statistics Canada. Measured children and youth body mass index (BMI) (World Health Organization classification), by age group and sex, Canada and provinces, Canadian Community Health Survey - Nutrition. [cited 2022 Nov 24]. Available from: https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=1310079501 5. Hack S, Jessri M, L’Abbé MR. **Nutritional quality of the food choices of Canadian children**. *BMC Nutrition* (2021.0) **7** 16. DOI: 10.1186/s40795-021-00422-6 6. Qutteina Y, De Backer C, Smits T. **Media food marketing and eating outcomes among pre-adolescents and adolescents: a systematic review and meta-analysis**. *Obes Rev* (2019.0) **20** 1708-1719. DOI: 10.1111/obr.12929 7. Cairns G, Angus K, Hastings G, Caraher M. **Systematic reviews of the evidence on the nature, extent and effects of food marketing to children. A retrospective summary**. *Appetite* (2013.0) **62** 209-215. DOI: 10.1016/j.appet.2012.04.017 8. Pinto A, Pauzé E, Mutata R, Roy-Gagnon MH, Potvin KM. **Food and beverage advertising to children and adolescents on television: a baseline study**. *IJERPH.* (2020.0) **17** 1999. DOI: 10.3390/ijerph17061999 9. Potvin Kent M, Pauzé E, Roy EA, de Billy N, Czoli C. **Children and adolescents’ exposure to food and beverage marketing in social media apps**. *Pediatr Obes* (2019.0) **14** e12508. DOI: 10.1111/ijpo.12508 10. 10.CRTC. Communications Monitoring Report 2019;2019;338. [cited 2022 March 23]. Available from: https://crtc.gc.ca/pubs/cmr2019-en.pdf. 11. Kelly B, Halford JCG, Boyland EJ, Chapman K, Bautista-Castaño I, Berg C. **Television food advertising to children: a global perspective**. *Am J Public Health* (2010.0) **100** 1730-1736. DOI: 10.2105/AJPH.2009.179267 12. Kelly B, Vandevijvere S, Ng S, Adams J, Allemandi L, Bahena-Espina L. **Global benchmarking of children’s exposure to television advertising of unhealthy foods and beverages across 22 countries**. *Obes Rev* (2019.0) **20** 116-128. DOI: 10.1111/obr.12840 13. Mcginnis JM, Gootman JA, Kraak VI. *Food marketing to children and youth: threat or opportunity?* (2006.0) 14. Sadeghirad B, Duhaney T, Motaghipisheh S, Campbell NRC, Johnston BC. **Influence of unhealthy food and beverage marketing on children’s dietary intake and preference: a systematic review and meta-analysis of randomized trials**. *Obes Rev* (2016.0) **17** 945-959. DOI: 10.1111/obr.12445 15. Story M, French S. **Food advertising and marketing directed at children and adolescents in the US**. *Int J Behav Nutr Phys Act* (2004.0) **1** 3. DOI: 10.1186/1479-5868-1-3 16. Smith R, Kelly B, Yeatman H, Boyland E. **Food marketing influences Children’s attitudes, preferences and consumption: a systematic critical review**. *Nutrients.* (2019.0) **11** 875. DOI: 10.3390/nu11040875 17. Powell LM, Szczypka G, Chaloupka FJ. **Adolescent exposure to food advertising on television**. *Am J Prev Med* (2007.0) **33** S251-S256. DOI: 10.1016/j.amepre.2007.07.009 18. Pechmann C, Levine L, Loughlin S, Leslie F. **Impulsive and self-conscious: adolescents’ vulnerability to advertising and promotion**. *J Public Policy Mark* (2005.0) **24** 202-221. DOI: 10.1509/jppm.2005.24.2.202 19. Czoli CD, Pauzé E, Potvin KM. **Exposure to food and beverage advertising on television among Canadian adolescents, 2011 to 2016**. *Nutrients.* (2020.0) **12** 428. DOI: 10.3390/nu12020428 20. 20.Global Food Research Program, UNC Chapel Hill. National policies regulating food marketing to children. Available from: https://www.globalfoodresearchprogram.org/wp-content/uploads/2023/03/GFRP-UNC_Marketing_map_2023_03.pdf. 21. 21.Ad StandardsThe Canadian Children’s food and beverage advertising initiative [internet]2018. *The Canadian Children’s food and beverage advertising initiative [internet]* (2018.0) 22. 22.Government of QuebecAdvertising Directed at Children under 13 Years of Age2012. *Advertising Directed at Children under 13 Years of Age* (2012.0) 23. Pauzé E, Potvin KM. **Children’s measured exposure to food and beverage advertising on television in Toronto (Canada), may 2011–may 2019**. *Can J Public Health* (2021.0) **112** 1008-1019. DOI: 10.17269/s41997-021-00528-1 24. 24.Private Member’s Bill C-252 (44–1) - First Reading - Child Health Protection Act - Parliament of Canada [Internet]. Bill C-252. Available from: https://www.parl.ca/DocumentViewer/en/44-1/bill/C-252/first-reading 25. Potvin Kent M, Soares Guimaraes J, Remedios L, Bagnato M, Pauzé E, Pinto A. *Child and Youth Exposure to Unhealthy Food and Beverage Marketing on Television in Canada in 2019: A report to Health Canada* (2021.0) 26. Franco-Arellano B, Arcand J, Kim MA, Schermel A, L’Abbé MR. **Progress towards eliminating industrially produced trans-fatty acids in the Canadian marketplace, 2013-2017**. *Public Health Nutr* (2020.0) **23** 2257-2267. DOI: 10.1017/S1368980019004816 27. Murphy SA, Weippert MV, Dickinson KM, Scourboutakos MJ, L’Abbé MR. **Cross-sectional analysis of calories and nutrients of concern in Canadian chain restaurant menu items in 2016**. *Am J Prev Med* (2020.0) **59** e149-e159. DOI: 10.1016/j.amepre.2020.05.005 28. 28.Health CanadaHealth Canada’s proposed nutrient profile model for restricting marketing to children2019. *Health Canada’s proposed nutrient profile model for restricting marketing to children* (2019.0) 29. Pinto A, Pauzé E, Roy-Gagnon MH, Dubois L, Potvin KM. **The targeting of preschoolers, children, adolescents and adults by the Canadian food and beverage industry on television: a cross-sectional study**. *Appl Physiol Nutr Metab* (2021.0) **46** 651-660. DOI: 10.1139/apnm-2020-0621 30. Frazier WC, Harris JL. *Trends in television food advertising to young people: 2017 update* (2018.0) 31. Gruzd A, Mai P. *The state of social Media in Canada 2020* (2020.0) 32. 32.Media Technology Monitor Jr. Sneak Peek [Internet]. Available from: https://www.mtm-otm.ca/Download.ashx?file=emailers/infographics/mtmjr-launch/infographics-en.pdf 33. Gearhardt AN, Yokum S, Harris JL, Epstein LH, Lumeng JC. **Neural response to fast food commercials in adolescents predicts intake**. *Am J Clin Nutr* (2020.0) **111** 493-502. DOI: 10.1093/ajcn/nqz305 34. Bahadoran Z, Mirmiran P, Azizi F. **Fast food pattern and Cardiometabolic disorders: a review of current studies**. *Health Promot Perspect* (2016.0) **5** 231-240. DOI: 10.15171/hpp.2015.028 35. Robinson E, Jones A, Whitelock V, Mead BR, Haynes A. **(over)eating out at major UK restaurant chains: observational study of energy content of main meals**. *BMJ.* (2018.0) **363** k4982. DOI: 10.1136/bmj.k4982 36. Norman J, Kelly B, Boyland E, McMahon A. **The impact of marketing and advertising on food behaviours: evaluating the evidence for a causal relationship**. *Curr Nutr Rep* (2016.0) **5** 139-149. DOI: 10.1007/s13668-016-0166-6 37. Garriguet D. **Canadians’ eating habits**. *Health Rep* (2007.0) **18** 17-32. PMID: 17578013 38. Moubarac JC. *Ultra-processed foods in Canada: consumption, impact on diet quality and policy implications* (2017.0) 39. Conway L. *Advertising to children - commons library research briefing* (2021.0) 40. 40.World Health OrganizationA Framework for implementing the set of recommendations on the marketing of foods and non-alcoholic beverages to children2012. *A Framework for implementing the set of recommendations on the marketing of foods and non-alcoholic beverages to children* (2012.0) 41. Boyland EJ, Harrold JA, Kirkham TC, Halford JCG. **Persuasive techniques used in television advertisements to market foods to UK children**. *Appetite.* (2012.0) **58** 658-664. DOI: 10.1016/j.appet.2011.11.017 42. Folkvord F. *The Psychology of Food Marketing and (Over)eating* (2019.0) 43. Potvin Kent M, Wanless A. **The influence of the Children’s food and beverage advertising initiative: change in children’s exposure to food advertising on television in Canada between 2006-2009**. *Int J Obes* (2014.0) **38** 558-562. DOI: 10.1038/ijo.2014.4 44. Potvin Kent M, Pauzé E. **The effectiveness of self-regulation in limiting the advertising of unhealthy foods and beverages on children’s preferred websites in Canada**. *Public Health Nutr* (2018.0) **21** 1608-1617. DOI: 10.1017/S1368980017004177 45. 45.Association of Canadian Advertisers. Code and Guide for the Responsible Advertising of Food and Beverage Products to Children | Association of Canadian Advertisers. [cited 2021 Nov 17]. Available from: https://acaweb.ca/en/resource/code-guide-responsible-advertising-of-food-beverage-to-children/ 46. Reeve B, Magnusson R. **Regulation of food advertising to children in six jurisdictions: a framework for analyzing and improving the performance of regulatory instruments**. *Ariz J Int’l Comp L* (2018.0) **35** 60 47. Rincón-Gallardo Patiño S, Rajamohan S, Meaney K, Coupey E, Serrano E, Hedrick VE. **Development of a responsible policy index to improve statutory and self-regulatory policies that protect Children’s diet and health in the America’s region**. *Int J Environ Res Public Health* (2020.0) **17** E495. DOI: 10.3390/ijerph17020495 48. Powell LM, Harris JL, Fox T. **Food marketing expenditures aimed at youth putting the numbers in context**. *Am J Prev Med* (2013.0) **45** 453-461. DOI: 10.1016/j.amepre.2013.06.003
--- title: Intense solar activity reduces urinary 6-sulfatoxymelatonin in patients with COPD authors: - Carolina L. Zilli Vieira - Petros Koutrakis - Man Liu - Daniel J. Gottlieb - Eric Garshick journal: Respiratory Research year: 2023 pmcid: PMC10037776 doi: 10.1186/s12931-023-02390-w license: CC BY 4.0 --- # Intense solar activity reduces urinary 6-sulfatoxymelatonin in patients with COPD ## Abstract ### Background Little is known about the link between solar activity and variations in melatonin. In this study, we investigated if melatonin's major urinary metabolite, urinary 6-sulfatoxymelatonin (aMT6s), is lowest under periods of intense solar activity. ### Methods We investigated associations between high-energy solar particle events [Coronal Mass Ejection (CME) mass, speed and energy] on creatinine-adjusted aMT6s (aMT6sr) concentrations in 140 patients with chronic obstructive pulmonary disease (COPD) using up to four seasonal urine samples ($$n = 440$$). Mixed effect models with a random intercept for each subject were used to estimate associations, including effect modification attributable to diabetes, obesity, and reduced pulmonary function. ### Results Higher values of CME were associated with reduced aMT6sr concentrations, with stronger associations in patients with diabetes. An interquartile range (IQR) increase in natural log CMEspeed averaged through two days before urine collection was associated with a reduction of $9.3\%$ aMT6sr ($95\%$CI: − $17.1\%$, − $0.8\%$) in aMT6sr. There was a greater reduction in aMT6sr in patients with diabetes (− $24.5\%$; $95\%$CI: − $35.9\%$, − $11.6\%$). In patients without diabetes there was no meaningful association (− $2.2\%$; $95\%$CI: − $12\%$, $8.4\%$). There were similar associations with CMEenergy and CMEmass. There was no effect modification attributable to reduced pulmonary function or obesity. ### Conclusions This is the first study in patients with COPD to demonstrate strong detrimental impact of high-energy solar particle events on aMT6sr, with greater associations in patients with diabetes. Since melatonin is an anti-oxidant, it is possible that adverse effects of intense solar activity may be attributable to a reduction in circulating melatonin and that patients with both COPD and diabetes may be more susceptible. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12931-023-02390-w. ## Background Melatonin (N-acetyl-5-methoxytryptamine) is a potent nocturnal antioxidant hormone and efficient immuno-inflammatory regulator, which is primarily synthetized in the pineal gland [1]. Melatonin is also produced by other organs such as the retina and the gastrointestinal tract [1]. It is well known that pineal gland function and melatonin secretion are modulated by the environmental light/dark cycle via the suprachiasmatic nucleus (SCN) in the hypothalamus [2]. SCN is the “master clock” that regulates the 24 h-circadian rhythm, including melatonin/cortisol rhythms, gene expression, and autonomic nervous system (ANS) function [2–8]. The disruption of 24-h circadian rhythm with concomitant reduction of melatonin synthesis and serum levels seem also to be modulated by solar and geomagnetic activity (SGA)-related visible and non-visible electromagnetic radiation [3]. This may be relevant clinically, as we have previously described reductions in pulmonary function attributable to SGA [9]. We have also observed additional adverse effects of SGA that include increases in blood pressure, atrial and ventricular arrhythmias, and increases in circulating biomarkers of systemic inflammation [10–13]. In this study, we hypothesize that intense solar activity reduce levels of urinary 6-sulfatoxymelatonin (aMT6s), thereby providing a biologic pathway with the potential to contribute to the adverse effects of SGA. Circulating melatonin has a short half-life (~ up to 45 min) and is rapidly metabolized in the liver with the major enzymatic metabolite of melatonin (aMT6s), which highly correlates with plasma melatonin levels [13, 14]. Studies in animals have suggested a link between seasonal rhythmicity of aMT6s levels and ~ 11-year solar cycles [8]. Melatonin levels can also be affected by genetic factors, sex, pineal gland size, and modifiable factors such as seasonal changes, latitudinal locations, medications (e.g., β-blockers and non-steroidal anti-inflammatory drugs), sleep quality, depression, smoking status, and comorbidities including diabetes and cancer [2, 13–20]. Levels of melatonin vary throughout life, reaching highest levels in children younger than 4 years, and declining within age [2, 19]. In the lung, melatonin may mitigate respiratory disease severity by modulating pro-inflammatory cytokines such as interleukin1β and 6, and TNF-α, reducing oxidative stress [1, 13, 16]. A reduction in melatonin also can increase the risk of the development of type 2 diabetes mellitus by impairing insulin sensitivity and glucose tolerance [6, 7]. Understanding the impact of solar activity on melatonin levels is critical to understanding natural exposures that may affect health outcomes, particularly in association with oscillations of 11-solar cycles, even though the relationship between variation in solar activity and melatonin levels in humans has not been well established. To investigate this, we took advantage of previously collected urine samples in a cohort of patients with COPD [21, 22] to assess the link between short-term variation in solar activity intensity and melatonin excretion, using aMT6s concentrations to estimate melatonin levels. We also assessed effect modification of comorbidities (diabetes mellitus, obesity, and reduced pre-bronchodilator pulmonary function) on aMT6s. ## Study subjects Between 2013 and 2017, 143 patients with chronic obstructive pulmonary disease (COPD) were recruited at Veterans Affairs (VA) Boston Healthcare System from Eastern Massachusetts and vicinity to investigate the impact of particulate air pollution exposures [20–22]. Participants had up to 4 seasonal visits scheduled at least 2 weeks after therapy for a COPD exacerbation. Participants were former smokers with 10 pack-years or more of lifetime smoking, and had a ratio of post-bronchodilator forced expiratory volume in one second to forced vital capacity (FEV1/FVC) of < 0.70 at a screening visit or emphysema on chest computed tomography. Individuals with malignancies other than local skin or stable prostate cancer, a systemic inflammatory disease such as rheumatoid arthritis; or with unstable heart disease, were excluded. By study design, in order to substantially reduce exposure to sources of indoor combustion, we excluded patients who were current smokers or lived with a current smoker, or who had a major source of indoor air pollution (e.g., wood stove or fireplace, frequent burning of incense or candles). At each study visit spirometry pre- and post-bronchodilator was conducted [23], medication use was reviewed, and height and weight were measured. At study entry, participants were asked if they had ever being told by doctor that they had sleep apnea or diabetes. The study protocol was approved by Institutional Review Boards at VA Boston and Harvard Medical School. We obtained informed consent from all participants prior to study procedures. ## Urinary 6-sulfatoxymelatonin (aMT6s) assessment Study visits occurred during daylight hours where the time of urine sample collection was noted [median 11:22 AM (interquartile range (IQR): 2 h 34 min), mean 11:35 AM (standard deviation: 1 h 51 min);]. Samples were put on ice and transported to the VA Boston core laboratory and frozen at − 80 °C. For this analysis aMT6s (ng/ml) was measured in these stored samples in duplicate by an ELISA assay (Alpco, Salem, NH) in the Department of Laboratory Medicine, Boston Children’s Hospital, Boston, MA, USA. The day-to-day variabilities of the assay at concentrations of 6.8, 95 and 248 ng/mL are 11.0, 6.3 and $5.2\%$, respectively. The assay is sensitive to 1.0 ng/mL. To account for differences in urinary dilution, we measured urinary creatinine (mg/ml), using the ratio of aMT6s and creatinine (aMT6sr) as the study outcome (in ng/mg creatinine). ## Solar activity parameters Parameters of solar activity events [corona mass ejection (CME)] were obtained from the NASA SOHO/LASCO CME (https://cdaw.gsfc.nasa.gov/CME_list/). CME is high-energy plasma ejected from the outer surface of the Sun that interacts with the earth’s magnetic field, producing geomagnetic disturbances and increased electromagnetic phenomena in the earth’s systems (e.g. atmosphere, geosphere). CME can take hours to days to reach earth depending on its energy, mass and speed. We converted hourly CME data in UTC time to local Boston time, and created daily averages. Daily CME data included the mean of CMEmass (grams of solar mass), CMEenergy [in erg, unit of energy equal to 10−7 J (100 nJ)], and CMEspeed (km/s). CME can occur many times per day or none, depending on the intensity of solar activity (higher intensity of solar activity results in an increased number of CME events). Because the CME values can range from negligible to trillions of units of CMEmass, energy and speed, we used the natural log of CME variables. ## Statistical analysis We used mixed effect models with random intercept for each subject to analyze the association of solar activity on the natural log transformed aMT6sr to normalize its distribution and meet model assumptions. As CME can take hours to days to reach Earth, we analyzed five windows of moving averages of exposure from the day of urine collection (day 0) to up to 4-days (day 4) prior to the urine sample collection. The primary models included a single exposure variable (corona mass ejection energy, speed or mass) and covariates. Melatonin can be affected by sex, seasonal changes, medications (including β-blockers and non-steroidal anti-inflammatory drugs), sleep apnea, and comorbidities including diabetes [2, 13–20]. Therefore, model covariates included a priori were race (white vs. other), sex (male/female), age, body mass index (BMI), beta blocker use and non-steroidal anti-inflammatory medication use within 1 day of urine collection, diabetes, time of urine collection, history of sleep apnea, and season (winter, spring, summer, fall), and BMI.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${aMT6sratio}_{ID} \sim {X}_{1 lCME\left[X\right]}+{X}_{2}{\mathrm{urine}}_{\mathrm{season}}+{X}_{3}\mathrm{sex}+{X}_{4}\mathrm{bmi}+{X}_{5}\mathrm{age}+{X}_{6}\mathrm{race} +{X}_{6}\mathrm{med}1 \left(\mathrm{NSAID}\right)+{X}_{7}\mathrm{med}2 \left(\mathrm{beta blockers}\right)+{X}_{8}{\mathrm{baseline}}_{\mathrm{diabetes}}+{X}_{9}{\mathrm{baseline}}_{\mathrm{apnea}}+{X}_{10}{\mathrm{urine}}_{\mathrm{time}},$$\end{document}aMT6sratioID∼X1lCMEX+X2urineseason+X3sex+X4bmi+X5age+X6race+X6med1NSAID+X7med2betablockers+X8baselinediabetes+X9baselineapnea+X10urinetime,where: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${aMT6sratio}_{ID}$$\end{document}aMT6sratioID is the natural log of aMT6s/creatinine per visit; X1 is the natural log CME mass, energy or speed. Visits where persons reported melatonin use the night before urine collection were excluded in addition to when the aMT6sr was above the upper 95th percentile (aMT6sr > 77.8), suggesting unreported melatonin use (from 143 patients, 3 were excluded). That is consistent with the upper limit of physiologic levels reported in the literature [12]. We assessed the effect modification by obesity (BMI > 30), diabetes, and %-predicted pre-bronchodilator FEV1 < $50\%$ and ≥ $50\%$ predicted using multiplicative interaction terms and stratified effect estimates. Effects on aMT6sr were calculated by multiplying each beta and corresponding $95\%$ confidence interval values by log IQR of CME mass, speed and energy and exponentiating. After subtracting 1 from each estimate and multiplying by 100, the results are expressed as percent increase in aMT6sr per log IQR of each exposure. We assessed the correlation among the exposure variables (Additional file 1) by calculating Pearson and Spearman correlation coefficients (Additional file 1). We also examined each model residuals to assess model fit. We performed all analyses using SAS 9.4 software. ## Characteristics of the study population After excluding visits with aMT6sr above the upper 95th percentile, there were 140 participants with 440 visits (75 patients with 4 visits, 25 patients with 3 visits, 25 patients with two visits, and 15 patients with only one visit) and approximately $95\%$ of the visits were completed over one year. The clinical characteristics of the patients are showed in the Table 1. Most of them were elderly white men, having a mean age of 72.7 (SD = 8.2) years. Thirty-six percent ($25.7\%$) reported diabetes, and 66($47.2\%$) were obese at study entry. There were 73 ($52.1\%$) patients using beta blockers, and 21($15\%$) using non-steroidal taking anti-inflammatory medication (Table 1). aMT6sr levels varied widely among study participants [mean = 14.8 (SD = 11.4); median = 11.2 (IQR = 11.7)]. There was a strong correlation among CME parameters (R > 0; p-value > 0.05) (details in the Additional file 1).Table 1Characteristics of patients with COPD between 2013 to 2017First visitOveralln = 140nobs = 440Age (years)72.7 (8.2)73.1 (8.3)BMI (kg/m2)30 (5.7)30.3 (6.1)Past lifetime smoking (packyears)59.9 (5.7)58.3 (37.1)Pre-bronchodilator % predicted FEV164.5 (21.9)*64.7 [22]**Pre-bronchodilator % predicted FVC85.5 (20.2)*85 (20.5)**FEV1/FVC ratio0.54 (0.13)*0.55 (0.12)**aMT6s (ng/ml)15.6 (13.9)17.4 (18.2)Log-aMT6sr2.5 (0.7)2.5 (0.7)Creatinine (mg/dl)112.1 (57.4)120 (68.9)N (%)N (%)Race White125 (89.3)390 (88.6) Non-white15 (10.7)50 (11.4)Sex Female4 (2.8)11 (2.5) Male136 (97.2)429 (97.5)Season Winter21 [15]101 [23] Spring35 [25]103 (23.4) Summer48 (34.3)121 (27.5) Fall36 (25.7)115 (26.2)ComorbiditiesN (%)N (%)Reduced lung function (pre-bronchodilator FEV1 < $50\%$ predicted)***35 (25.8)112 (25.9)Obesity (BMI > 30)66 (47.2)209 (47.5)Diabetes36 (25.7)113 (25.7)Sleep apnea42 [30]139 (31.6)Medications (within 1 day of urine collection)N (%)N (%)Beta blocker73 (52.1)232 (52.7)Non-steroidal anti-inflammatory medication21 [15]61 (13.8)Solar activity assessment [Coronal Mass Ejection (CME)]Mean (SD)Mean (SD)Log-CMEspeed7.3 (0.4)7.2 (0.5)Log-CMEmass35.8 (0.8)35.6 (0.9)Log-CMEenergy70.2 (1.2)69.9 (1.3)CMEspeed1519.4 (627.2)1503.6 (694.9)CMEmass4.2 × 1016 (2.5 × 1016)3.8 × 1015 (2.6 × 1015)CMEenergy5.1 × 1030 (5.3 × 1030)4.5 × 1030 (5.4 × 1030)*$$n = 136$$, **$$n = 433$$, ***based on Hankinson et al. [ 24] ## Primary analysis There was inverse relationship between each CME parameter moving average (from day 0 to 4 days prior to the urine collection) and creatinine-adjusted melatonin, consistent with an association between intense solar particle events and reduction in aMT6sr (Fig. 1). The point estimates for CMEspeed were more negative than those of the other CME parameters, indicating a slightly greater reduction (Fig. 1). For an increase of an IQR of 1.1 (2.1 km/s) in natural log CMEspeed two days prior to the day of examination, there was a reduction of $9.3\%$ ($95\%$CI: − $17.1\%$, − $0.8\%$; p-value: 0.02) or − 0.1 units ($95\%$CI: − 0.2; − 0.01) of natural log-transformed aMT6sr levels (Fig. 1; Additional file 1).Fig. 1Associations of natural log CME (energy, mass, and speed) with aMT6s/Creatinine ratio. The primary models included a single exposure variable (log corona mass ejection energy, speed or mass) and model covariates ## Effect modification There were greater CME-related effects in patients with diabetes mellitus (p-interaction = 0.006) (Fig. 2; Additional file 1). For example, an IQR of 1.1 (2.1 km/s) natural log CMEspeed two days prior to the sample collection was associated with a reduction of $24.5\%$ ($95\%$CI: − $35.9\%$, − $11.6\%$; p-value: 0.0006) or − 0.3 units ($95\%$CI: − 0.5, − 0.1) in natural log aMT6sr levels in patients with diabetes. In patients without diabetes there was no association [aMT6sr: − $2.2\%$ ($95\%$CI: − $12\%$, $8.4\%$; p-value: 0.66) or − 0.02 natural log units ($95\%$CI: − 0.1, 0.1] (Fig. 2; Additional file 1). There was no evidence of effect modification attributable to obesity and reduced pulmonary function (Additional file 1).Fig. 2Associations of natural log CME with aMT6s/creatinine ratio modified by diabetes. Effect modification of diabetes. The models included a single exposure variable (log corona mass ejection energy, speed or mass) and model covariates ## Discussion To the best of our knowledge, this is the first study to demonstrate a reduction of aMT6sr during periods of intense high-energy solar particle events. The associations were stronger and more robust in patients with diabetes mellitus. Although all CME parameters are strongly correlated and represent the ejection of high energy solar material, the association between CME speed and MT6sr was greater than other parameters. As CMEspeed parameter seems to play a critical role on the atmospheric photoionization processes and energetic particle precipitation, and on Earth’s magnetic field disturbances, these factors may be linked for the observed impact on melatonin levels [23]. Intense solar activity can impact human health directly, by the modulation and disruption of the 24 h-circadian rhythm, and indirectly by inducing physicochemical properties and toxicity of atmospheric aerosols [9, 12]. Solar activity continuously emits a broad-spectrum range of electromagnetic radiation that modulates changes in the solar–terrestrial environment over time scales ranging from minutes to millennia. Sun-Earth interaction plays major roles in the radioactive, physiochemical processes and dynamics of Earth systems, which affect the human health possibly by the reduction of melatonin synthesis and ANS dysregulation related to the disruption of the 24 h circadian rhythm located in the SCN [15]. While it is unclear why patients with diabetes were more susceptible to the detrimental impact of intense solar activity on aMT6sr levels, the literature describes a link between aMT6s and diabetes risk [4, 7, 17]. Reduced levels of melatonin can impair diabetes management and disrupt blood sugar control [7], which suggests that diabetic patients may experience periods of poor health during intense solar activity periods. A bidirectional relationship between melatonin levels and insulin secretion may explain our findings in patients with diabetes. The disruption of the 24 h circadian rhythm with subsequent lower nocturnal melatonin secretion is associated with insulin secretion and resistance [4, 17], which may influence the development and progression of type 2 diabetes [4]. The circadian rhythm regulates the sleep/wake and feeding/fasting cycles through melatonin, impacting glucose homeostasis by influencing the timing of insulin secretion from pancreatic β cells, glucose production by the liver, insulin-dependent glucose GLUT 4 expression in skeletal muscles, among other effects [4]. There is a synchronization between insulin levels and melatonin during the night and day, in which melatonin action is inhibited by insulin release (and reduce glucose tolerance) through its membrane receptors MT1 and MT2, and the secondary messengers 3′,5′-cyclic adenosine monophosphate, 3’,5’-cyclic guanosine monophosphate, and inositol 1,4,5-trisphosphate [6]. High levels of insulin are observed when melatonin levels are low, and during the night low levels of insulin are observed with high levels of melatonin [6]. Reduction of the melatonin receptor MT2 may play a role on the development of type 2 diabetes and metabolic diseases [6]. Concomitantly, age-related reduction levels of melatonin accompany an increase of age-related insulin resistance and type 2 diabetes risk [6]. Patients with both diabetes and COPD appear to be more susceptible to solar activity reducing circulating melatonin, and potentially antioxidant effects of melatonin. This finding suggests that differences in the response to solar activity could in part explain a greater susceptibility to effects of illnesses and environmental exposures that promote oxidative stress in patients with diabetes [25]. This study has limitations and strengths. As the original COPD study was not designed to assess effects linked to melatonin, the time of urine collection was not standardized and there is no information about melatonin determinants such as sleep habits and nocturnal light exposures, and the cohort included mainly white males. However, we improved our estimation by adjusting our models for covariates linked to impaired melatonin secretion, including urine collection time, sleep apnea history and treatment, beta blocker use, and diabetes history as covariates. As melatonin excretion is greatest at night, urine from a first morning void might be more informative, while 24-h urine collection would be ideal to assess overall impact on melatonin secretion. This limitation should bias towards a null result; therefore, the effect that we observed is likely underestimated. As melatonin was assessed at only one time of day, we are unable to comment on the effects of solar activity on circadian rhythm per se, which would require repeated measures across the day. Moreover, our findings may not be generalized to other populations. Our study strengths are the availability of this cohort with stored urine samples for every visit, extensive information regarding personal and clinical characteristics, which created a unique opportunity to test our hypotheses. ## Conclusions Our findings contribute to understanding relationships between solar activity and susceptibility to disruption to 24 h-circadian rhythm that result in lower levels of melatonin, a circulating anti-oxidant. These results provide evidence for a biologic pathway related to intense solar activity that may be responsible for adverse health effects. Hence, our study findings have critical relevance to understand the impact of the periodicity of solar activity intensity on aMT6sr levels in high risk patients with COPD, which impacts the progression and prognosis of the disease and comorbidities. ## Supplementary Information Additional file 1: Table S1 Supplementary descriptive analysis. Table S2 Pearson correlation analysis [is this for $$n = 440$$? Show Pearson for natural log as this is what you use in the models. Table S3 Associations of log CME with aMT6s/Creatinine modified by pre bronchodilator %-predicted FEV1, obesity, and diabetes in patients with COPD. Results expressed as % -change per overall IQR of log CMEspeed, CMEenergy, and CMEmass for each moving average starting with the day 0, the day of urine collection through 4 days before collection. Table S4 Associations of log CME with aMT6s/Creatinine modified in patients with COPD. Results expressed as % -change per overall IQR of log CMEspeed, CMEenergy, and CMEmass for each moving average starting with the day 0, the day of urine collection through 4 days before collection. Fig. S1. Associations of CMEspeed, CMEenergy, CMEmass with MT6s/creatinine modified by obesity (BMI>30). Fig. S2. Associations of CMEspeed, CMEenergy, CMEmass with MT6s/creatinine modified by post-bronchodilator %-predicted FEV1 <$50\%$. ## References 1. Opie LH, Lecour S. **Melatonin has multiorgan effects**. *Eur Heart J Cardiovasc Pharmacother* (2016.0) **2** 258-265. DOI: 10.1093/ehjcvp/pvv037 2. Nogueira LM, Sampson JN, Chu LW. **Individual variations in serum melatonin levels through time: implications for epidemiologic studies**. *PLoS ONE* (2013.0) **8** e83208. DOI: 10.1371/journal.pone.0083208 3. Burch JB, Reif JS, Yost MG. **Geomagnetic disturbances are associated with reduced nocturnal excretion of a melatonin metabolite in humans**. *Neurosci Lett* (1999.0) **266** 209-212. DOI: 10.1016/s0304-3940(99)00308-0 4. Oosterman JE, Wopereis S, Kalsbeek A. **The circadian clock, shift work, and tissue-specific insulin resistance**. *Endocrinology* (2020.0) **161** bqaa180. DOI: 10.1210/endocr/bqaa180 5. Karamitri A, Jockers R. **Melatonin in type 2 diabetes mellitus and obesity**. *Nat Rev Endocrinol* (2019.0) **15** 105-125. DOI: 10.1038/s41574-018-0130-1 6. Espino J, Pariente JA, Rodríguez AB. **Role of melatonin on diabetes-related metabolic disorders**. *World J Diabetes* (2011.0) **2** 82-91. DOI: 10.4239/wjd.v2.i6.82 7. Reutrakul S, Siwasaranond N, Nimitphong H. **Associations between nocturnal urinary 6-sulfatoxymelatonin, obstructive sleep apnea severity and glycemic control in type 2 diabetes**. *Chronobiol Int* (2017.0) **34** 382-392. DOI: 10.1080/07420528.2016.1278382 8. Bartsch H, Mecke D, Probst H. **Search for seasonal rhythmicity of pineal melatonin production in rats under constant laboratory conditions: spectral chronobiological analysis, and relation to solar and geomagnetic variables**. *Chronobiol Int* (2012.0) **29** 1048-1061. DOI: 10.3109/07420528.2012.719958 9. 9.Anand K, Vieira CLZ, Garshick E, et al. Solar and geomagnetic activity reduces pulmonary function and enhances particulate pollution effects. Sci Total Environ. 10. Schiff JE, Vieira CL, Garshick E. **The role of solar and geomagnetic activity in endothelial activation and inflammation in the NAS cohort**. *PLoS ONE* (2022.0) **17** e0268700. DOI: 10.1371/journal.pone.0268700 11. Wang VA, Zilli Vieira CL, Garshick E. **Solar activity is associated with diastolic and systolic blood pressure in elderly adults**. *J Am Heart Assoc* (2021.0) **10** e021006. DOI: 10.1161/JAHA.120.021006 12. Zilli Vieira CL, Link MS, Garshick E. **Solar and geomagnetic activity enhance the effects of air pollutants on atrial fibrillation**. *EP Europace* (2022.0) **24** 713-720. DOI: 10.1093/europace/euab269 13. Rzepka-Migut B, Paprocka J. **Melatonin-measurement methods and the factors modifying the results a systematic review of the literature**. *Int J Environ Res Public Health* (2020.0) **17** 1916. DOI: 10.3390/ijerph17061916 14. Braam W, Spruyt K. **Reference intervals for 6-Sulfatoxymelatonin in urine: a meta-analysis**. *Sleep Med Rev* (2022.0) **3** 101614. DOI: 10.1016/j.smrv.2022.101614 15. Reiter RJ. **Static and extremely low frequency electromagnetic field exposure: reported effects on the circadian production of melatonin**. *J Cell Biochem* (1993.0) **51** 394-403. DOI: 10.1002/jcb.2400510403 16. Wang W, Gao J. **Effects of melatonin on protecting against lung injury**. *Exp Thera Med.* (2021.0) **21** 1 17. McMullan CJ, Schernhammer ES, Rimm EB. **Melatonin secretion and the incidence of type 2 diabetes**. *JAMA* (2013.0) **309** 1388-1396. DOI: 10.1001/jama.2013.2710 18. Gläser S, Krüger S, Merkel M. **Chronic obstructive pulmonary disease and diabetes mellitus: a systematic review of the literature**. *Respiration* (2015.0) **89** 253-264. DOI: 10.1159/000369863 19. Touitou Y. **Human aging and melatonin**. *Clin Relev Exp Gerontol* (2001.0) **36** 1083-1100. DOI: 10.1016/S0531-5565(01)00120-6 20. Garshick E, Grady ST, Hart JE. **Indoor black carbon and biomarkers of systemic inflammation and endothelial activation in COPD patients**. *Environ Res* (2018.0) **165** 358-364. DOI: 10.1016/j.envres.2018.05.010 21. Grady ST, Koutrakis P, Hart JE. **Indoor black carbon of outdoor origin and oxidative stress biomarkers in patients with chronic obstructive pulmonary disease**. *Environ Int* (2018.0) **115** 188-195. DOI: 10.1016/j.envint.2018.02.040 22. Busenkell E, Collins CM, Moy ML. **Modification of associations between indoor particulate matter and systemic inflammation in individuals with COPD**. *Environ Res* (2022.0) **209** 112802. DOI: 10.1016/j.envres.2022.112802 23. Gopalswamy N, Tsurutani B, Yan Y. **Short-term variability of the Sun-Earth system: an overview of progress made during the CAWSES-II period**. *Prog Earth Planet Sci* (2015.0) **2** 1-41. DOI: 10.1186/s40645-015-0043-8 24. Hankinson JL, Odencrantz JR, Fedan KB. **Spirometric reference values from a sample of the general US population**. *Am J Respir Crit Care Med* (1999.0) **159** 179-187. DOI: 10.1164/ajrccm.159.1.9712108 25. Bonini MG, Sargis RM. **Environmental toxicant exposures and type 2 diabetes mellitus: two interrelated public health problems on the rise**. *Curr Opin Toxicol.* (2018.0) **7** 52-59. DOI: 10.1016/j.cotox.2017.09.003
--- title: Prevalence of preterm birth and associated factors among mothers who gave birth in public hospitals of east Gojjam zone, Ethiopia authors: - Tafere Birlie Ayele - Yikeber Abebaw Moyehodie journal: BMC Pregnancy and Childbirth year: 2023 pmcid: PMC10037778 doi: 10.1186/s12884-023-05517-5 license: CC BY 4.0 --- # Prevalence of preterm birth and associated factors among mothers who gave birth in public hospitals of east Gojjam zone, Ethiopia ## Abstract ### Backgrounds Preterm birth is defined as babies born alive before 37 weeks of pregnancy or fewer than 259 days since the first day of a woman’s last menstrual period. Globally, 14.84 million babies were preterm births. Preterm infants are at risk for specific diseases related to the immaturity of various organ systems. This study aimed to assess the prevalence of preterm birth and associated factors among mothers who gave birth in public hospitals of east Gojjam zone, Ethiopia. ### Methods An institutional-based cross-sectional study was conducted from April 1 up to June 30, 2021, in public hospitals in the east Gojjam zone. Systematic random sampling was used. Data were collected through structured questionnaires, patient interviews and patient card reviews. We used binary logistic regression analysis with $95\%$ CI and P-value < 0.05 to identify the significant factors with preterm birth. ### Results Out of 615 mothers, $13.2\%$ gave a preterm birth. Antenatal care (AOR = 2.87; $95\%$ CI = (1.67, 5.09)), educational status of mother (AOR = 2.79; $95\%$ CI = (1.27, 6.67)), husband educational status(AOR = 2.11; $95\%$ CI = (1.10, 4.18)), Average monthly family income(AOR = 1.95; $95\%$ CI = (1.05, 3.75)),family size(AOR = 0.15; $95\%$ CI = (0.03, 0.67)), multifetal gestation (AOR = 3.30; $95\%$ CI = (1.29, 8.69), having Premature Rupture Of Membrane (AOR = 6.46; $95\%$ CI= (2.52, 18.24)), history of chronic illness (AOR = 3.94; $95\%$ CI = (1.67, 9.45)), being HIV positive(AOR = 6.99; $95\%$ CI= (1.13, 44.65)), Ante-Partum Hemorrhage (AOR = 3.62; $95\%$ CI= (1.12, 12.59)), pregnancy Induced Hypertension (AOR = 3.61; $95\%$ CI= (1.19, 11.84)), mode of delivery (AOR = 7.16; $95\%$ CI = (2.09, 29.29)), and onset of labor (AOR = 0.10; $95\%$ CI = (0.03, 0.29)) were found to be significantly associated with preterm birth. ### Conclusions antenatal care, educational status of the mother, husband’s educational status, family income, family size, multifetal gestation, Premature Rupture of the membrane, history of chronic illness, being HIV positive, Ante-Partum Hemorrhage, pregnancy Induced Hypertension, mode of delivery, and the onset of labor were found to be significantly associated with preterm birth. To minimize the proportion of preterm birth focusing on this important variables, timely identification of obstetric complications, strengthening early screening of HIV and high-risk pregnancies like multiple gestations, PIH and APH were important. ## Background Preterm birth (PTB) is defined as babies born alive before 37 weeks of pregnancy or fewer than 259 days since the first day of a woman’s last menstrual period [1]. Globally, 14.84 million babies were preterm births. The majority of these births occurred in Asia and sub-Saharan Africa [2]. Preterm birth is a global problem, with $60\%$ of preterm births occurring in Africa and South Asia. On average, $12\%$ of babies born in the poorest countries are premature, compared with $9\%$ in higher-income countries [3]. Direct complications of preterm birth account for one million deaths yearly, and preterm birth is a risk factor in over $50\%$ of all neonatal deaths [4]. Preterm infants have a higher risk of developing specific diseases due to the immaturity of various organ systems and the causes of preterm birth. As a result of their prematurity, preterm babies are subjected to serious illnesses or deaths during their neonatal period. In the absence of appropriate treatment, survivors are more likely to suffer a lifelong disability and a compromised quality of life. Prematurity complications are the leading cause of neonatal death and the second leading cause of death among children under 5 years old. [ 5]. Preterm birth complications were the leading cause of death in children younger than 5 years of age globally, accounting for approximately $16\%$ of all deaths and responsible for $35\%$ of deaths among newborn babies [6]. Prematurity now takes the first place for neonatal intensive care unit (NICU) admission, longer hospital stay, the second leading cause of death in children under 5 years, and the single most important direct cause of death in the critical first month of life of infants [7, 8]. In low- and middle-income (LMIC) countries, preterm births account for more than $60\%$ of all births, and the rate has steadily increased. Despite this high preterm birth rate, it is challenging to determine the trend of preterm birth in the majority of low-income economies due to a lack of accurate data [9]. Sub-Saharan African countries have a high preterm birth rate: $23.7\%$ in Nigeria [10], $18.3\%$ in Kenya [11], and $16.3\%$ in Malawi [12]. The overall prevalence of preterm birth in Ethiopia was $10.48\%$ [13]. There is a high rate of infant mortality (48 deaths per 1,000 live births) and neonatal mortality (29 deaths per 1,000 live births) and complication of preterm birth is a major risk factor of this mortality [14]. In the Amhara region, the prevalence of preterm birth was $11.41\%$ [15]. The complication of infants born at preterm gestational age results in a trivial cost to the health sector, parents, and society. The prediction and prevention of preterm birth is a major health care priority [16]. Global efforts to further reduce child mortality demand urgent action to address preterm birth. However, it is a complex multifactorial process associated with diverse pathogenic mechanisms and the prevalence of preterm delivery is one of the strongest predictors of neonatal mortality in our country. During the neonatal period, preterm babies are at a higher risk of serious illness or death. Those who survive preterm delivery without proper care are at great risk of chronic impairment and poor quality of life [17]. Previous studies conducted in different regions indicated that several risk factors were identified for preterm birth. This includes having a previous preterm birth, having a short cervix, short intervals between pregnancies, and certain pregnancy complications (including multiple pregnancy, pregnancy-induced hypertension, premature rupture of the membrane, and vaginal bleeding), chronic illness, educational status, Multiple pregnancies, maternal age, residing in rural areas, antenatal care visits, being HIV positive, family number, and income [15, 18–26]. PTB is a major public health problem. However, in most low-income countries, including Ethiopia, little emphasis is given to PTB intervention as a means of reducing infant mortality. Health care providers or other stakeholders who worked in this public health problem need data related to common factors associated with preterm birth. However, in our country, the studies conducted about this problem were minimal. Although few studies have been conducted in some areas of Ethiopia, the magnitude and possible risk factors of PTB vary by area. A number of methodological issues are addressed in our study (sample size calculation, sampling technique, and multicenter study area) that were not considered in previous studies. Studying a large region is essential to designing effective interventions and programs in public health. Most of the previous study focused mainly on the prevalence of PTB, rather than associated factors of PTB. In addition, different cultures and socioeconomic statuses within a society have varied factors of preterm birth. Therefore studying in different and multicenter settings was important. Determining preterm birth prevalence and associated factors greatly guides health professionals and health policymakers to identify indicators for monitoring preterm birth strategy and applying necessary preventive and appropriate measures to decrease preterm birth. Moreover, the study area’s magnitude of preterm births and associated factors were unknown. Therefore, this study was conducted to assess the prevalence of preterm birth and to identify factors associated with preterm birth in public hospitals in the east Gojjam zone. ## Study design The hospital based cross-sectional study design was conducted using interviewer administered questionnaire from April 1 up to June 30, 2021. Additional information was obtained from medical records of the mothers and babies. ## Study area The study was conducted in public hospitals in the east Gojjam zone, Amhara National Regional State, Ethiopia. Debre Markos city is a zonal administrative city. It covers 14,010 square kilometers and is divided into 18 administrative districts, further subdivided into 49 urban and 392 rural kebeles, the smallest administrative units [27]. The East Gojam zone is located in northwest Ethiopia, which is 265 Km far from Bahirdar, the capital city of Amhara region, 299 km from Addis Ababa, the country’s capital. The Oromia region borders the zone on the south, West Gojjam on the west, South Gondar on the north, and South Wollo on the east. With a population of 3.8 million, the East Gojjam Zone has 21 Woreda, 480 Kebeles, 10 government hospitals (1 referral hospital, 9 primary hospitals), 102 health centers, and 423 health posts [28–30]. This study was conducted in five randomly selected public hospitals. The randomly selected public hospitals were Shegaw *Motta* general hospital, Bichena primary hospital, Dejen primary hospital, Lumame primary hospital, and Debre Markos referral hospital. All these health institutions are currently providing maternal and child health care services. ## Participants All mothers who gave birth in randomly selected public hospitals in the east Gojjam zone during the study period were our study population. Mothers who gave birth and had known either LNMP or had early ultrasound (before 24 weeks) diagnosis were included for this study. Mothers who were deaf and comatose, had unknown last normal menstrual period (LNMP) and had no early ultrasound (before 24 weeks) were excluded from this study. ## Sampling technique and sample size determination A systematic sampling technique was used. The sample was arranged based on the three-month patient flow before the data collection period by referring to the hospital’s delivery registration book/ record. To calculate K, the summation of the three months delivery report for the hospital was 1698.Then K = N/n, $\frac{1698}{615}$ = 2.8 ≈ 3. Where k = interval, N = total population, n = sample size. Every third mother’s was interviewed, gestational age of the newborn was calculated based on the mothers LNMP or first-trimester ultrasound result, in estimating gestational week, when there are extra days it was counted to the near lowest gestational age. The next participant was taken when the selected study participant was not eligible for the study. The sample size was determined by using a single population proportion formula by considering the prevalence (p) of preterm birth = $15.5\%$, which was obtained from the previous study in Ethiopia [17], $95\%$ confidence interval (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z\frac{\alpha }{2}$$\end{document}=1.96), and level of precision (d) = 0.03. 1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n=\frac{\left({{Z}_{\frac{\alpha }{2}}}^{2}\right)\left(p\right)(1-p)}{{d}^{2}}$$\end{document} 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n=\frac{{ 1.96}^{2}\left(0.155\right)\left(0.845\right)}{\left({0.03}^{2}\right)}=559$$\end{document} Finally, after taking a $10\%$ non-response rate the total sample size (n) was 615. An average delivery report for a month before the actual data collecting period was computed for each hospital by analyzing the client’s registration book to distribute the sample size proportionally to each hospital. The sample size was then proportionally allocated to each hospital. Finally, the immediate postnatal mother with her baby had been selected every three intervals using a systematic sampling technique. ## Data collection Data were collected and extracted by reviewing medical records of mother’s. The data collectors were four fourth year undergraduate midwives students from Debre Markos University. One staff midwife from each hospital was assigned to supervise the data collection process. The data collection process was supervised by both the principal investigator and the supervisor. Information collected from the mother included Residency, Age, Religion, Education status, Occupation, Marital status, Average monthly Average monthly family income, Antenatal Care (ANC), inter pregnancy interval parity, Educational Status of husband, Ante-Partum Hemorrhage (APH), Premature Rupture Of Membrane (PROM), Pregnancy Induced Hypertension (PIH), multiple pregnancies, polyhydramnios, anemia, cardiac disease, hypertension, HIV status, Modern contraceptive use before current pregnancy, Urinary Tract Infection (UTI) and malaria were the predictor variable for this study. Anemia was defined as an HGB level below 11gm/dl (HCT < $33\%$). Obstetric complications was defined as challenges or problems that happen during labor or delivery. PIH was defined clinically as a blood pressure of > $\frac{140}{90}$ mmHg after 20 weeks of gestation with or without proteinuria and/or edema as diagnosed and documented by the attending clinician. APH was defined as any vaginal bleeding in the mother after 24 weeks of gestation as documented in the records by the attending clinician. UTI was defined as a documented clinical/laboratory diagnosis of UTI any time during the pregnancy and/or a positive history of treatment of burning sensation with micturition as reported by the mother. Birth to pregnancy interval was defined as the time between the start of the index pregnancy and the preceding live birth. Last normal menstrual period was defined as the date of the starting of last normal menstruation the women had to index pregnancy. Preterm birth was defined as a newborn with a gestational age of 28 weeks to less than 37 weeks. To assure the quality of data, the questionnaire was pre-testing on 31 mothers in public hospitals in the east Gojjam zone, and the questionnaire’s fitness was confirmed from pre-testing. One-day practical training on how to collect data was given to the data collectors and the supervisor before data collection. During the data collection period, the collected data were reviewed, checked for completeness, and signed by the supervisor at the end of each day. ## Data processing and analysis All questionnaires were checked, coded, and entered into the SPSS version 25 software packages, which were then analyzed using R 4.1.3. The data were presented using frequency tables and graphs. The relevant determinants of preterm birth were identified using binary logistic regression. The researchers used both bivariable and multivariable analyses. In the bivariable analysis, independent variables with a p-value less than 0.25 were chosen for the multivariable analysis. An adjusted odds ratio with a $95\%$ confidence level was used to examine the degree of relationship between independent and dependent variables, and variables with a p-value of 0.05 were considered statistically significant. ## Socio-demographic characteristics the of the respondents All participants completed the interview ($100\%$ response rate). The majority of the respondents 358 ($58.2\%$) were rural residents. The majority of the study participants 360 ($58.5\%$) were between 25 and 34 years old. The entire respondent belongs to Amhara by ethnicity, and 516 ($96.1\%$) were Orthodox Christians in religion. Regarding the marital status of the respondents, the majority of them 596($96.9\%$) were married. 253 ($41.1\%$) were have no formal education, 217($35.3\%$) of the respondents were taken education in primary school, and 253($41.1\%$) of them can only read and write, and the remaining 145 ($23.6\%$) were secondary and above education level. The occupational status of most of the respondents 228($37.1\%$) were farmers, 144 ($23.4\%$) of the respondents were housewives, 150 ($24.4\%$) of the respondents were employers, and the remaining respondents had other occupations (labor work, no work, students, work-seekers). 264 ($42.9\%$) of the husband were have no formal education. 264($42.9\%$) of the respondent’s family average monthly income is less than or equal to 3000 Birr and 212($34.5\%$) of the respondent’s family average monthly income is average greater than 5000 Birr. The majority of the respondents, 415 ($67.5\%$) had between 3 and 5 family members (Table 1). Table 1Socio-Demographic Characteristics of the study participants in public hospitals of east Gojjam zone, Ethiopia, 2021 from April to JuneVariablesCategoriesFrequenciesPercent (%) ($$n = 615$$)ResidenceUrban25741.8Rural35858.2Age15–2418229.625–3436058.5> 357311.9Marital statusMarried59696.9Others(Divorced, Widowed, and Single)193.1ReligionOrthodox59196.1Others(protestant, and Muslim)243.9Educational status of Mothersno formal education25341.1primary education21735.3secondary and above14523.6Occupational status of the respondentsHousewife14423.4Employer15024.4Farmer22837.1Other9315.1Educational Status of the husbandno formal education26442.9primary education13922.6secondary and above21234.5Average monthly family income<=3000 Birr26442.93001–5000 Birr13922.6> 5000 Birr21234.5Family size< 317428.33–541567.5> 5264.2 ## Obstetric and medical-related characteristics Most of the respondents ($96.3\%$) had ANC follow up and $56.1\%$ of the respondent had at least 4 visits. The majority of the respondents used modern contraceptives ($94.8\%$) before their pregnancy, and the majority of mothers had their pregnancy wanted and planned ($99.3\%$). More than two-thirds of the respondents had birth-to-pregnancy intervals greater than or equal to 36 months ($83.5\%$). Labor spontaneously started in $87.8\%$ of the respondents. Three-quarters of the respondents ($74.3\%$) gave birth by SVD. Out of the total respondents, 595($96.7\%$) were tested for HIV status, and 36 ($5.7\%$) respondents were HIV positive. [ 133] $21.6\%$ of women had APH, [142] $23.1\%$, of women had PIH, [70] $11.4\%$, of mothers had multiple pregnancies, and women had Polyhydramnios [10] $1.6\%$. 105($17.1\%$) of the respondents had a history of chronic illness. 70($11.4\%$) of mothers had urinary tract infection (Table 2). Table 2Obstetric and medical related characteristics of the study participants in public hospitals of east Gojjam zone, Ethiopia, 2021 from April to JuneVariablesCategoriesFrequencyPercent (%) ($$n = 615$$)ANC visitYes59296.3No233.7ANC follow up< 4 follow up34556.1>=4 follow up27043.9Pregnancy statusPlanned61199.3Unplanned40.7Modern contraceptive use before current pregnancyYes58394.8No325.2Birth to pregnancy interval in months< 366516.5≥ 3632983.5Mode of deliverySVD45774.3CS or Instrumental delivery15825.7Onset of laborSpontaneous54087.8Induced7512.2History of HIV testingYes59596.7No203.3Status of HIV ($$n = 595$$)Positive365.7Negative55991.1Un known203.3Blood RH factorPositive55590.2Negative609.8PROMYes21935.6No39664.4HGB (mg/dl)< 11294.7≥ 1158695.3APHYes13321.6No48278.4PIHYes14223.1No47376.9Multiple pregnanciesYes7011.4No54588.6PolyhydramniosYes101.6No60598.4History of Chronic illnessYes10517.1No51082.9Urinary Tract Infection. Yes7011.4No54588.6 ## The proportion of preterm birth The proportion of preterm birth in this study was found to be $13.2\%$ (CI: 0.11, 0.16) (Fig. 1). Fig. 1proportion of preterm birth in the east Gojjam zone ## Factors associated with preterm birth All independent variables were analyzed using binary logistic regression with the dependent variable preterm birth and those, which were significant at a p-value of < 0.25 were transferred to multivariable logistic regression analysis. The variable with a p-value < 0.05 was significant. In bivariable analysis ANC follow-up, educational status of mother, husband educational status, family income, marital status, occupation, family size, PROM, being HIV positive, obstetric complication, APH, PIH, history of chronic illness, multifetal gestation, RH factor, pregnancy status, mode of delivery, onset of labor, and UTI were found to be significantly associated with pre term birth. In multivariable binary logistic regression analysis ANC follow up (AOR = 2.87; $95\%$ CI = (1.67, 5.09)), educational status of mother (AOR = 2.79; $95\%$ CI = (1.27, 6.67)), husband educational status(AOR = 2.11; $95\%$ CI = (1.10, 4.18)), Average monthly family income(AOR = 1.95; $95\%$ CI = (1.05, 3.75)),family size(AOR = 0.15; $95\%$ CI = (0.03, 0.67)), multifetal gestation (AOR = 3.30; $95\%$ CI = (1.29, 8.69), having PROM(AOR = 6.46; $95\%$ CI= (2.52, 18.24)), history of chronic illness (AOR = 3.94; $95\%$ CI = (1.67, 9.45)), being HIV positive(AOR = 6.99; $95\%$ CI= (1.13, 44.65)), APH(AOR = 3.62; $95\%$ CI= (1.12, 12.59)), PIH (AOR = 3.61; $95\%$ CI= (1.19, 11.84)), mode of delivery (AOR = 7.16; $95\%$ CI = (2.09, 29.29)), and onset of labor (AOR = 0.10; $95\%$ CI = (0.03, 0.29)) were found to be statistically significant at p-value of < 0.05 (Table 3). Table 3Factors associated with preterm birth among study participants in public hospitals of east Gojjam zone, Ethiopia, 2021 from April to JuneVariablesN = 615 PretermCOR ($95\%$ CI)AOR($95\%$ CI)YesNo ANC follow up < 4602852.50(1.50, 4.31)2.87(1.67, 5.09)≥ 42124911 *Marital status* Married745220.24(0.09, 0.67)0.44(0.06, 3.92)Others(Divorced, Widowed, and Single)71211 *Education status* of mothers no formal education522013.91(1.95, 8.73)2.79 (1.27, 6.67)primary education201971.53(0.70, 3.64)1.53(0.65, 3.86)secondary and above913611 Husbands education status no formal education451533.35(1.88, 6.24)2.11(1.10, 4.18)primary education191861.18(0.59, 2.36)0.88(0.42, 1.85)secondary and above1719511 Average monthly family income <=3000 Birr522122.81(1.60, 5.16)1.95(1.05, 3.75)3001–5000 Birr121271.08(0.49, 2.33)0.86(0.38, 1.94)> 5000 Birr1719511 Family size < 3301440.47(0.19,1.23)0.51(0.09, 2.68)3–5433720.26(0.11, 0.66)0.15(0.03, 0.67)> 581811Occupational status of the respondentsHousewife261181.64(0.79, 3.64)1.42(0.63, 3.35)employer81420.42(0.157, 1.08)0.61(0.22, 1.64)Farmer361921.39(0.69, 3.00)0.60(0.27, 1.41)Other118211 Mode of delivery SVD693882.16(1.18, 4.30)7.16(2.09, 29.29)CS or Instrumental delivery1214611 Multifetal gestation Yes23474.12(2.30, 7.20)3.30(1.29, 8.69)No5848711PROMYes481713.09(1.92, 5.02)6.46(2.52,18.24)No3336311APHYes44895.94(3.64,9.78)3.62(1.12, 12.59)No3744511PIHYes52908.85(5.36, 14.85)3.61(1.19, 11.84)No2944411 History of Chronic illness Yes35705.04(3.03, 8.37)3.94(1.67, 9.45)No4646411 UTI Yes26445.26(2.99, 9.18)2.41(0.37, 16.42)No5549011 Status of HIV($$n = 595$$) Positive15215.83(2.8, 11.86)6.99(1.13, 44.65)Negative6149811Pregnancy statusPlanned795320.15(0.02, 1.25)0.19(0.01, 35.87)Unplanned2211RH Factorpositive794764.81(1.46, 29.73)1.82(0.31,16.75)negative25811 onset of labor Spontaneous504900.14(0.08, 0.25)0.10(0.03, 0.29)Induced314411Abbreviations: 1 = Reference, ANC Antenatal Care, AOR Adjusted Odds Ratio, APH Ante-Partum Hemorrhage, CI Confidence Interval, COR Crud Odds Ratio, CS Cesarean Section, HIV Human Immune Deficiency Virus, PIH pregnancy Induced Hypertension, PROM Premature Rupture of Membrane, SVD Spontaneous Vaginal Delivery, UTI Urinary Tract Infection ## Discussion This study was conducted to assess the magnitude of preterm birth and its associated factors in public hospitals in the east Gojjam zone. During the study period, the overall proportion of preterm birth was found to be $13.2\%$. The finding was greater than the preterm birth rate for the world ($9.8\%$) and North America ($9\%$) [2, 31]. It was also more than the prevalence of preterm birth in Ethiopia, $10.1\%$, which was reported by the Global Action Report on Preterm Birth [32]. Compared with cross-sectional studies conducted in our country the finding was found to be in line with the finding conducted in Debre Tabor town health institutions, which was $12.8\%$, and with the finding conducted in Axum and Adwa town public hospitals in which the prevalence of preterm birth was $13.3\%$ [33, 34]. The proportion of preterm birth in this study however was higher than another study conducted in our country at Gondar town health institutions, which reported a prevalence rate of $4.4\%$ [18]. This discrepancy may be due to differences in exclusion criteria for multiple pregnancies. In our study, mothers with multiple pregnancies were included, whereas these mothers were excluded from the mentioned study. Therefore, a lower rate was expected in their study, as over distention of the uterus as in multiple pregnancies and polyhydramnios is one of the scientifically explained causative factors for preterm labor. The prevalence of preterm was also higher than the prevalence of most developed nations. The preterm birth rate from the study conducted in Sweden which was estimated to be $5.03\%$ is good evidence [35]. The low preterm birth rate in developed nations like Sweden may be due to high socio-demographic status of the population and improved preconception and ANC services, which are important in early identifying and preventing risk factors. The proportion of preterm birth in this study was found to be lower than in some studies conducted in low and middle-income countries. The study conducted in Malawi shows that the prevalence of preterm birth was $16.3\%$ [12]. This higher prevalence of preterm birth in Malawi may be due to the country’s higher HIV infection rate, where one in four women are HIV positive. In Brazil, the prevalence of preterm birth among young women attending public hospitals was $21.7\%$ which was higher compared to this study [36]. The discrepancy may be due to variation in the study population. Only parturient mothers aged 15–24 were included in their study. A similar study conducted in Nigeria reported the prevalence of preterm birth of $16.9\%$ [37], which was also higher than this study. This variation is maybe because of the difference in the study area where their study was at a referral hospital with referrals of more complicated cases from other general hospitals. The current study’s finding was lower than those conducted in Kenya National Hospital and Jemma University Specialized Hospital, which reported the prevalence of preterm birth was $20.2\%$ and $25.9\%$, respectively [23, 38]. This variation might be due to the difference in the study time, reflecting that FMOH has currently improved maternal health care service. Another possible reason for this variation might be due to differences in the study area, the study done in Kenya and Jimma indicates that the high prevalence of alcohol consumption and substance intake during pregnancy may be the contributing factors to the increased magnitudes of preterm birth. Mothers with less than 3000 birr Average monthly family income were 1.95 times more likely to develop preterm than those with greater than 5000 birr family income. This study was in line with the study in SSA [26, 39],which shows Average monthly family income positively affected preterm. This might be due to financial insecurity, psychosocial stress, and low health care utilization. This study was not in line with the studies in Ethiopia, which show Average monthly family income was not associated with preterm mothers [18]. Mothers from three up to five family members were 0.15 times less likely to develop/give preterm as compared to mothers from greater than five family members. This study was in line with the study in Ethiopia [17], which shows family members had a significant positive effect on preterm birth. The education status of mothers were significant factors of preterm. Mothers who had no formal education were 2.79 times more likely to develop/give preterm than mothers who had secondary and above education levels. This study was in line with the studies in Sub-Saharan African and European countries [39–41], which show education had been associated with preterm birth. This study was not in line with the studies in Ethiopia and Kenya [38, 42], which show education statutes had no significant effect on preterm. Mothers who had husbands who had no formal education were 2.11 times more likely to develop/give preterm than mothers who had husbands with secondary and above education levels. This study is in line with the previous study in Iran [43], which shows a significant association between husbands’ education level and preterm birth. Mothers with APH were 3.62 times more likely to have a preterm birth than mothers without APH. This finding is in line with the study conducted in Kenya, Ethiopia, and East Africa [38, 44, 45]. This suggests that obstetric problems caused by APH may significantly impact the occurrence of PTB. This could be due to decreased placental blood flow, impacting the mother-fetus exchange of nutrients and oxygen. As a result, fetal growth would be slowed, and the chance of PTB would be increased. Mothers with multifetal gestation were 3.30 times more likely to have a preterm birth than mothers without multifetal gestation. This finding is in line with the study conducted in Ethiopia, Korea, Greek [46–48]. It might be since multiple pregnancies are more likely to be associated with a variety of problems, including preeclampsia, PROM, and polyhydramnios, all of which could contribute to iatrogenic PTB. Furthermore, this could be related to uterine overstretching and deciding to terminate the pregnancy before it reaches term. Mothers who delivered with SVD were 7.16 times more likely to develop preterm as compared to mothers delivered with CS or Instrumental delivery. This finding is in line with the study conducted in Ethiopia [49], which shows delivered with SVD had a positive significant effect on preterm. This study is not in line with [50], which shows women delivering by previous cesarean section had a significantly higher risk of preterm birth when compared to women with vaginal delivery. Mothers with spontaneous labor were 0.10 times less likely to have preterm birth as compared to mothers with induced labor. This finding is in line with the study conducted in Ethiopia [49], which shows spontaneous labor had a negative impact on preterm mothers. In addition, this study is in line with the study conducted in Ethiopia [51], which shows *Labor status* was associated with preterm birth. History of chronic illness was significantly associated with the outcome variable, mothers who had a history of chronic illness were 3.94 times more likely to give preterm birth than mothers who had no chronic illness history. This finding is in line with a study conducted in Ethiopia [18, 44, 49], which shows chronic illness positively associated with preterm. This might be due to maternal illnesses that impede or impair the placental transfer of oxygen and nutrients to the developing fetus in the uterus can raise the chance of preterm birth. This study revealed a significant association between pregnancy-induced hypertension and preterm birth. Mothers who had complications PIH were 3.61 times increased risk of having a preterm birth than those mothers without this problem during the index pregnancy. This finding is in line with the study conducted in East Africa, Ethiopia, Nigeria, Iran, Ghana, and Kenya [18, 38, 45, 46, 51–55], which shows PIH had significant effect on preterm birth. This might be due to the vascular damage of the placenta caused by PIH, which results in preterm labor and delivery. Another significant association was found between mothers who had premature rupture of membranes (PROM) and preterm birth. Mothers with PROM were 6.46 times more likely to have a preterm birth than their counterparts. This finding was in line with the previous findings in Tehran, Iran, Ethiopia, East Africa, Nigeria [7, 26, 46, 53, 56], showing a significant association between PROM and preterm. This might be because in the absence of any clinical intervention labor will spontaneously initiate within hours after term PROM and within a week after preterm PROM in the majority of the cases. In this study, the ANC follow up were significantly associated with the outcome variable preterm birth. Mothers with < 4 times ANC follow-up in the index pregnancy were 2.87 times more likely to have a preterm birth than mothers who had the ANC visits ≥ 4 times. This finding is in line with a study conducted in Ghana, Nigeria teaching hospitals, and western Ethiopia [20, 44, 52, 54, 57], which shows a significant association between ANC visits and preterm. WHO recommends at least four ANC visits, this will assure the quality of care and early detection of high-risk pregnancies which result in the prevention and proper management of obstetric complications. This may happen regularly. The benefits of an ANC visit include health promotion, early detection, and treatment of obstetric problems. However, this study is not in line with the study conducted in Ethiopia [18], which shows ANC follow-up had no significant effect on preterm. Being HIV positive was significantly associated with preterm birth. HIV-positive mothers were 6.99 times more likely to give preterm compared to HIV-negative mothers. This finding is in line with a study in Malawi and Ethiopia [12, 18, 41, 58, 59], which shows a significant association between prevalence of HIV and preterm birth. This might be due to the drug effect and immunity of the mother as risk factors for preterm birth. This study was not in line with studies conducted in Botswana, and Malawi [60, 61]. This might be due to the drug’s impact and the mother’s immunity to premature birth. In contradiction to the previous studies in our study, age, marital status, RH factor, Pregnancy status, and UTI [42, 62, 63]. This difference can be due to differences in the study area, design, period, population, and culture differences. ## Limitation of the study Despite efforts to reduce recollection biases by informing local events, there may be a recall bias. When ultrasound facilities are not accessible, the menstrual history and clinical examination are used to confirm gestational age and which may subject to considerable error. Another drawback of our study is using secondary data for some variables. ## Conclusion The proportion of preterm birth in public hospitals in the east Gojjam zone is $13.7\%$. number of ANC, educational status of mother, husband educational status, family income, family size, occupation, multifetal gestation, having PROM, history of chronic illness, being HIV positive, APH, PIH, mode of delivery, and onset of labor were found to be significantly associated with preterm birth. As a result, focusing on these important variables would reduce the number of premature births. Furthermore, it was suggested that educating the community about the importance of ANC service utilization of mothers and preventing preterm birth be strengthened. Moreover, strengthening early screening of HIV and high-risk pregnancies like multiple gestations, PIH and APH were important to prevent preterm. ## References 1. 1.Organization WH. World health statistics 2013: a wealth of information on global public health. World Health Organization; 2013. 2. Chawanpaiboon S. **Global, regional, and national estimates of levels of preterm birth in 2014: a systematic review and modelling analysis**. *The Lancet Global Health* (2019.0) **7** e37-e46. DOI: 10.1016/S2214-109X(18)30451-0 3. Dey AC. **Magnitude of problems of prematurity-national and global perspective: a review**. *Bangladesh J Child Health* (2012.0) **36** 146-52. DOI: 10.3329/bjch.v36i3.14288 4. Blencowe H. **Born too soon: the global epidemiology of 15 million preterm births**. *Reproductive health* (2013.0) **10** 1-14. PMID: 23279882 5. 5.Organization WH. WHO recommendations on interventions to improve preterm birth outcomes 2015. 6. 6.Hug L, Sharrow D, You D. Levels and trends in child mortality: report 2017. The World Bank; 2017. 7. Blencowe H. **National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: a systematic analysis and implications**. *The lancet* (2012.0) **379** 2162-72. DOI: 10.1016/S0140-6736(12)60820-4 8. Brown HK. **Neonatal morbidity associated with late preterm and early term birth: the roles of gestational age and biological determinants of preterm birth**. *Int J Epidemiol* (2014.0) **43** 802-14. DOI: 10.1093/ije/dyt251 9. Lawn JE. **Global report on preterm birth and stillbirth (1 of 7): definitions, description of the burden and opportunities to improve data**. *BMC Pregnancy Childbirth* (2010.0) **10** 1-22. DOI: 10.1186/1471-2393-10-S1-S1 10. Onankpa B, Isezuo K. **Pattern of preterm delivery and their outcome in a tertiary hospital**. *Int J Health Sci Res* (2014.0) **4** 59-65 11. 11.Wagura P. factors associated with preterm birth at kenyatta national Hospital BMC Pregnancy Childbirth, (18). 12. van den Broek NR, Jean-Baptiste R, Neilson JP. **Factors associated with preterm, early preterm and late preterm birth in Malawi**. *PLoS ONE* (2014.0) **9** e90128. DOI: 10.1371/journal.pone.0090128 13. Muchie KF. **Epidemiology of preterm birth in Ethiopia: systematic review and meta-analysis**. *BMC Pregnancy Childbirth* (2020.0) **20** 1-12. DOI: 10.1186/s12884-020-03271-6 14. 14.Zewudie AT, Gelagay AA, Enyew EF. Determinants of Under-Five Child Mortality in Ethiopia: Analysis Using Ethiopian Demographic Health Survey, 2016 International Journal of Pediatrics, 2020. 2020. 15. Adugna DG. **Prevalence and associated risk factors of preterm birth among neonates in referral hospitals of Amhara Region, Ethiopia**. *PLoS ONE* (2022.0) **17** e0276793. DOI: 10.1371/journal.pone.0276793 16. 16.Waitzman NJ, Jalali A, Grosse SD. Preterm birth lifetime costs in the United States in 2016: an update. In seminars in Perinatology. Elsevier; 2021. 17. Woldeyohannes D. **Factors associated with preterm birth among mothers who gave birth in Dodola town hospitals, Southeast Ethiopia: institutional based cross sectional study**. *Clin Mother Child Health* (2019.0) **16** 2 18. 18.Gebreslasie K. Preterm birth and associated factors among mothers who gave birth in Gondar town health institutions Advances in Nursing, 2016. 2016. 19. Belaynew W. **Effects of inter pregnancy interval on preterm birth and associated factors among postpartum mothers who gave birth at Felege Hiwot referral hospital**. *World J Pharm Pharm Sci* (2015.0) **4** 12-25 20. Woday A, Muluneh MD, Sherif S. **Determinants of preterm birth among mothers who gave birth at public hospitals in the Amhara region, Ethiopia: a case-control study**. *PLoS ONE* (2019.0) **14** e0225060. DOI: 10.1371/journal.pone.0225060 21. Aynalem YA. **Determinants of neonatal mortality among preterm births in Black Lion Specialized Hospital, Addis Ababa, Ethiopia: a case–cohort study**. *BMJ open* (2022.0) **12** e043509. DOI: 10.1136/bmjopen-2020-043509 22. 22.Kelkay B et al. Factors associated with singleton preterm birth in Shire Suhul general hospital, northern Ethiopia, 2018 Journal of Pregnancy, 2019. 23. Bekele I, Demeke T, Dugna K. **Prevalence of preterm birth and its associated factors among mothers delivered in Jimma university specialized teaching and referral hospital, Jimma Zone, Oromia Regional State, South West Ethiopia**. *J Women’s Health Care* (2017.0) **6** 1-10 24. Hosny AE-DM. **Risk factors associated with preterm labor, with special emphasis on preterm premature rupture of membranes and severe preterm labor**. *J Chin Med Association* (2020.0) **83** 280-7. DOI: 10.1097/JCMA.0000000000000243 25. Vogel JP, Lee AC, Souza JP. **Maternal morbidity and preterm birth in 22 low-and middle-income countries: a secondary analysis of the WHO Global Survey dataset**. *BMC Pregnancy Childbirth* (2014.0) **14** 1-14. DOI: 10.1186/1471-2393-14-56 26. Bekele T, Amanon A, Gebreslasie K. **Preterm birth and associated factors among mothers who gave birth in Debremarkos-town health institutions 2013 institutional basedcrosssectional study**. *Gynecol Obstet (Sunnyvale)* (2015.0) **5** 1000292 27. Asemahagn MA, Alene GD, Yimer SA. **Geographic accessibility, readiness, and barriers of health facilities to offer tuberculosis services in East Gojjam Zone, Ethiopia: a convergent parallel design**. *Res Rep Trop Med* (2020.0) **11** 3. PMID: 32099509 28. 28.Agency CS. 2007 population and housing census of Ethiopia. Central Statistical Agency Addis Ababa, Ethiopia; 2007. 29. Bizuayew H. **Post-cesarean section surgical site infection and associated factors in East Gojjam zone primary hospitals, Amhara region, North West Ethiopia, 2020**. *PLoS ONE* (2021.0) **16** e0261951. DOI: 10.1371/journal.pone.0261951 30. Aynalem BY, Melesse MF. **Health extension service utilization and associated factors in East Gojjam zone, Northwest Ethiopia: a community-based cross-sectional study**. *PLoS ONE* (2021.0) **16** e0256418. DOI: 10.1371/journal.pone.0256418 31. 31.Beck S et al. The worldwide incidence of preterm birth: a systematic review of maternal mortality and morbidity Bulletin of the world health organization, 2010. 88: p. 31–38. 32. Blencowe H, Cousens S, Oestergaard D. **Lawn data from national, regional and worldwide estimates of preterm birth rate**. *Glob Action Rep Preterm Birth* (2010.0) **4** 1-2 33. Fritz T. **Outcome of extremely preterm infants after iatrogenic or spontaneous birth**. *Acta Obstet Gynecol Scand* (2018.0) **97** 1388-95. DOI: 10.1111/aogs.13387 34. Bakhteyar K. **Factors associated with preterm delivery in women admitted to hospitals in Khorramabad: a case control study**. *Int J Health Allied Sci* (2012.0) **1** 147. DOI: 10.4103/2278-344X.105064 35. Cnattingius S. **Maternal obesity and risk of preterm delivery**. *JAMA* (2013.0) **309** 2362-70. DOI: 10.1001/jama.2013.6295 36. Miranda AE. **Prevalence and correlates of preterm labor among young parturient women attending public hospitals in Brazil**. *Revista Panam de Salud Pública* (2012.0) **32** 330-4. DOI: 10.1590/S1020-49892012001100002 37. Iyoke CA. **Prevalence and perinatal mortality associated with preterm births in a tertiary medical center in South East Nigeria**. *Int J women’s health* (2014.0) **6** 881. DOI: 10.2147/IJWH.S72229 38. Wagura P. **Prevalence and factors associated with preterm birth at kenyatta national hospital**. *BMC Pregnancy Childbirth* (2018.0) **18** 1-8. DOI: 10.1186/s12884-018-1740-2 39. Alamneh TS. **Preterm birth and its associated factors among reproductive aged women in sub-saharan Africa: evidence from the recent demographic and health surveys of sub-sharan african countries**. *BMC Pregnancy Childbirth* (2021.0) **21** 1-11. DOI: 10.1186/s12884-021-04233-2 40. Ruiz M. **Mother’s education and the risk of preterm and small for gestational age birth: a DRIVERS meta-analysis of 12 european cohorts**. *J Epidemiol Community Health* (2015.0) **69** 826-33. DOI: 10.1136/jech-2014-205387 41. 41.Wakeyo D, Addisu Y, Mareg M. Determinants of Preterm Birth among Mothers Who Gave Birth in Dilla University Referral Hospital, Southern Ethiopia: A Case-Control Study BioMed Research International, 2020. 42. Berhe T, Gebreyesus H, Desta H. **Determinants of preterm birth among mothers delivered in Central Zone Hospitals, Tigray, Northern Ethiopia**. *BMC Res Notes* (2019.0) **12** 1-6. DOI: 10.1186/s13104-019-4307-z 43. Khezri R. **RELATIONSHIP BETWEEN BODY MASS INDEX BEFORE PREGNANCY AND WEIGHT GAIN DURINGPREGNANCY WITH PRETERM BIRTH**. *Stud Med Sci* (2016.0) **26** 890-9 44. Zewde GT. **Preterm birth and associated factors among mother who gave birth in public health hospitals in harar town eastern Ethiopia 2019**. *OSP J Health Care Med* (2020.0) **1** 1-3 45. Laelago T, Yohannes T, Tsige G. **Determinants of preterm birth among mothers who gave birth in East Africa: systematic review and meta-analysis**. *Ital J Pediatr* (2020.0) **46** 1-14. DOI: 10.1186/s13052-020-0772-1 46. Mulualem G, Wondim A, Woretaw A. **The effect of pregnancy induced hypertension and multiple pregnancies on preterm birth in Ethiopia: a systematic review and meta-analysis**. *BMC Res Notes* (2019.0) **12** 1-7. DOI: 10.1186/s13104-019-4128-0 47. Lee KJ. **The clinical usefulness of predictive models for preterm birth with potential benefits: a KOrean Preterm collaboratE Network (KOPEN) registry-linked data-based cohort study**. *Int J Med Sci* (2020.0) **17** 1. DOI: 10.7150/ijms.37626 48. Bouras G. **Preterm birth and maternal psychological health**. *J Health Psychol* (2015.0) **20** 1388-96. DOI: 10.1177/1359105313512353 49. 49.Aregawi G et al. Preterm births and associated factors among mothers who gave birth in Axum and Adwa Town public hospitals, Northern Ethiopia, 2018. BMC research notes, 2019. 12(1): p. 1–6. 50. Zhang Y. **Mode of delivery and preterm birth in subsequent births: a systematic review and meta-analysis**. *PLoS ONE* (2019.0) **14** e0213784. DOI: 10.1371/journal.pone.0213784 51. 51.Abaraya M, Seid SS, Ibro SA. Determinants of preterm birth at Jimma university medical center, Southwest Ethiopia Pediatric health, medicine and therapeutics, 2018. 9: p. 101. 52. Iyoke C. **Maternal risk factors for singleton preterm births and survival at the University of Nigeria Teaching Hospital, Enugu, Nigeria**. *Niger J Clin Pract* (2015.0) **18** 744-50. DOI: 10.4103/1119-3077.158141 53. Tehranian N, Ranjbar M, Shobeiri F. **The prevalence and risk factors for preterm delivery in Tehran, Iran**. *J Midwifery Reproductive Health* (2016.0) **4** 600-4 54. Adu-Bonsaffoh K. **Determinants and outcomes of preterm births at a tertiary hospital in Ghana**. *Placenta* (2019.0) **79** 62-7. DOI: 10.1016/j.placenta.2019.01.007 55. Gejo NG. **Factors associated with preterm birth at Wachemo University Nigist Eleni Mohammed memorial hospital, southern Ethiopia: case-control study**. *BMC Pregnancy Childbirth* (2021.0) **21** 1-9. DOI: 10.1186/s12884-020-03503-9 56. 56.Iyoke CA et al. Prevalence and perinatal mortality associated with preterm births in a tertiary medical center in South East Nigeria.International Journal of Women’s Health, 2014: p.881–888. 57. Abadiga M. **Determinants of preterm birth among women delivered in public hospitals of western Ethiopia, 2020: unmatched case-control study**. *PLoS ONE* (2021.0) **16** e0245825. DOI: 10.1371/journal.pone.0245825 58. Deressa AT. **Factors associated with spontaneous preterm birth in Addis Ababa public hospitals, Ethiopia: cross sectional study**. *BMC Pregnancy Childbirth* (2018.0) **18** 1-5. DOI: 10.1186/s12884-018-1957-0 59. Mekonen DG. **Proportion of Preterm birth and associated factors among mothers who gave birth in Debretabor town health institutions, northwest, Ethiopia**. *BMC Res Notes* (2019.0) **12** 1-6. DOI: 10.1186/s13104-018-4037-7 60. 60.Kourtis AP, Fowler MG. Antiretroviral use during pregnancy and risk of preterm delivery: more questions than answers. Oxford University Press; 2011. pp. 493–4. 61. van den Akker T. **HIV care need not hamper maternity care: a descriptive analysis of integration of services in rural Malawi**. *BJOG: An International Journal of Obstetrics & Gynaecology* (2012.0) **119** 431-8. DOI: 10.1111/j.1471-0528.2011.03229.x 62. El-Sayed AM, Tracy M, Galea S. **Life course variation in the relation between maternal marital status and preterm birth**. *Ann Epidemiol* (2012.0) **22** 168-74. DOI: 10.1016/j.annepidem.2012.01.002 63. Shah R. **Incidence and risk factors of preterm birth in a rural bangladeshi cohort**. *BMC Pediatr* (2014.0) **14** 1-11. DOI: 10.1186/1471-2431-14-112
--- title: Explaining biological differences between men and women by gendered mechanisms authors: - Hélène Colineaux - Lola Neufcourt - Cyrille Delpierre - Michelle Kelly-Irving - Benoit Lepage journal: Emerging Themes in Epidemiology year: 2023 pmcid: PMC10037796 doi: 10.1186/s12982-023-00121-6 license: CC BY 4.0 --- # Explaining biological differences between men and women by gendered mechanisms ## Abstract ### Background The principal aim of this study was to explore if biological differences between men and women can be explained by gendered mechanisms. ### Methods We used data from the 1958 National Child Development Study, including all the living subjects of the cohort at the outcome collection wave (44–45 years). We explored several biomarkers as outcomes: systolic blood pressure, triglycerides, LDL cholesterol, HbA1c, CRP, and cortisol. Three conceptualizations of gender have been used to define methodological strategies: (a) Gender as an individual characteristic; (b) Gender as an effect of sex on socio-behavioural characteristics; (c) Gender as an interaction between sex and the social environment, here the early-life social environment. We estimated the total effect of sex and the proportion of total effect of sex at birth eliminated by gender, measured by 3 different ways according to these 3 concepts, using g-computation. ### Results The average level of each biomarker was significantly different according to sex at birth, higher in men for cardiometabolic biomarkers and higher in women for inflammatory and neuroendocrine biomarkers. The sizes of the differences were always smaller than one standard deviation but were larger than differences due to early-life deprivation, except for CRP. We observed gender mechanisms underlying these differences between men and women, even if the mediation effects were rarely statistically significant. These mechanisms were of three kinds: [1] mediation by socio-behavioural characteristics; [2] attenuation by gendered mechanisms; [3] interaction with early social environment. Indeed, we observed that being born into a deprived rather than non-deprived family increased metabolic and inflammatory biomarkers levels more strongly in females than in males. ### Conclusions The biological differences between men and women seem to not be purely explained by biological mechanisms. The exploration of gender mechanisms opens new perspectives, in terms of methodology, understanding and potential applications. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12982-023-00121-6. ## Background Observed differences between men and women* (see Table 1 for definitions of used terms*) in variables of a biological kind seem to be frequently attributed to mechanisms of biological kind (we will refer to them as sexed* mechanisms). These assumptions are sometimes explicitly expressed, for example in a course about “Sex and Gender in the Analysis of Data”, where haemoglobin, kidney function, height, lean body mass are described as “sex-related variables”, defined as “measurable biological or physiological parameters that systematically differ between men and women due to genetic or hormonal factors” [1]. Assuming sexed mechanisms can also be implicit, for example when sex-specific thresholds are set to control for presumed sexual dimorphism [2–4]. We even find this idea in the classic and often cited definition of sex and gender* in the biomedical literature, where “gender” refers to the social differences between men and women and “sex” refers to the biological differences between men and women [5], which can be understood as all the biological differences. While this simple definition may facilitate a broad understanding of ‘what is sex’ versus ‘what is gender’, it also formalizes a false dichotomy where sex/ gender is understood as biological/ social. Table 1Definitions given to the used termsTermsDefinitionsMen and womenHere, we use the term “men” or “women” in the sense of “male-born” and “female-born” individuals, without referring to gender identity or performanceSex or sexedWe use the term "sexed" instead of “sex” to qualify mechanisms related to biological sexual dimorphism. For example, "sex differences" designates differences observed according to the sex assigned at birth, without hypotheses on their underlying biological or social mechanisms, whereas "sexed differences" would designate differences explained by biological mechanisms linked to sexGender“Gender” is a complex concept and can refer to several phenomena. Especially in social sciences, it may refer to the process which polarises humanity and the characteristics of humanity into two categories "masculine" or "feminine", i.e., the binary and hierarchical categorisation system, including the phenomenon of domination relationship between the two categories [6]. The term of “gender” can also refer to an experience of self (gender personality or identity) [7, 8]In epidemiology, the term often seems to refer [1] either to the level at which the social characteristics and behaviours of an individual fits the stereotypes/ norms of masculinity or femininity; [2] or to the fact that, or the process by which, social characteristics and behaviours are differently distributed according to the binary sex at birth [9]. The gender concepts we used here are described in the "methods" section However, social and biological phenomena are not so compartmentalized. Human life courses, environments, experiences, and behaviours shape our biology. Social life is biologically “embodied” [10]. Men and women live different life courses in our gendered binary world. Indeed, individuals, according to their sex assigned at birth, will not be subjected during their lives to the same physical, social, economic, cultural, and emotional exposures, they will not experience the same stressors, nor perceive and react to them in the same way, they will not adopt the same behaviours, etc. All these exposures that are distributed differently according to the sex assigned at birth will have different biological consequences in men and women. The gendered social structuring leads to the gendered construction of biology and health. Therefore, the biological and health differences observed between men and women could, at least in part, be explained by gender-based social mechanisms. The concept of allostatic load emerged in the field of neurobiology, proposed by Mc Ewen to designate the cumulative multi-system physiological consequences of the repeated activation of adaptation (allostasis) processes by the organism in the face of the challenges it encounters during its existence [11]. This concept has been adopted by social epidemiology [12], since Teresa Seeman’s work [13], to explore how various social experiences—as adversity, discrimination, education, behaviours, etc. –, through stressful experiences and regulation, are biologically embodied [12]. In 1997, Seeman proposed the first epidemiological measurement of allostatic load [14], based on various biomarkers, which has since been widely used in the field of social epidemiology. These biomarkers measured the effect of experiences and regulations of stress in several physiological systems: the primary system (neuroendocrine), and, more often, secondary systems (cardiovascular, metabolic, inflammatory) [15]. Many of the biomarkers used to measure allostatic load have different distributions between men and women. These differences have been attributed to sexual dimorphisms [2–4], due to difference in sex hormones (e.g. oestrogen and testosterone) [16–18]. Yet socio-cultural factors and behaviours could also explain some of the differences in distribution [16, 19]. In this study, we aimed to explore whether differences in biomarkers observed between men and women are explained, at least partly, by three different gender mechanisms analysed using three different analytical strategies. ## Data and population We used data from the 1958 National Child Development Study (NCDS), one of the national British birth cohorts, which includes all people born during one week in 1958 ($$n = 18$$,555). Data on life conditions and experiences, about family, education, work, and health, were collected in twelve waves from birth to age 62 by the Centre for Longitudinal Studies. The NCDS has been described in detail elsewhere [20]. Detailed review of the ethical practices throughout NCDS is available at [https://cls.ucl.ac.uk/wp-content/uploads/$\frac{2017}{07}$/NCDS-Ethical-review-an-Consent-2014.pdf]. For this study, we used data collected during the first (1958, birth, $$n = 17$$,638), fourth (1981, 23 years, $$n = 12$$,357), fifth (1991, 33 years, $$n = 16$$,174) and biomedical waves (2002–2004, 44–45 years, $$n = 9$$,377). See the flow chart in Fig. 1. To reduce selection bias, we included all the living subjects at the time of the biomedical waves [21], when outcome variables had been collected ($$n = 17$$,272). Indeed, the total cohort is assumed to be representative of the generation, but the subjects are not missing at random at each wave. As a consequence, including only non-missing participants can leads to collider bias [22]. We therefore chose to include all living participants, to preserve the population structure, and imputed missing data. We however also performed a sensitivity analysis on participants who participated at the four used collection waves, involving more selection bias but fewer missing data ($$n = 7$$,021).Fig. 1Flow chart ## Gender concepts In this study, we explored gender as [1] the level at which the social characteristics and behaviours of an individual fits the stereotypes/ norms of masculinity or femininity (gender performance); and [2] the fact that, or the process by which, social characteristics and behaviours are differently distributed according to the binary sex at birth (gender pressure) [9]. These concepts can be operationalized in three ways within the epistemological and methodological framework of epidemiology, as detailed elsewhere [9] (see Fig. 2):Gender as an individual characteristic: gender refers to how an individual performs their gender, according to the norms of gender in the population in which they are socially active. This corresponds to the concept of gender performance. E.g., an individual can be said to have a “feminine” gender if they have mainly social characteristics considered as feminine, like having more care activities (childcare, looking after older people, nursing). This conceptualization implies understanding gender as an individually defined variable. Gender as an effect of sex on socio-behavioural characteristics: gender refers to the fact that socio-behavioural characteristics are differently distributed according to the sex at birth. This corresponds to gender as a gender pressure [9]. E.g., gender refers to the systemic process by which women are more likely to engage in caregiving activities than men in a given population. This conceptualization involves understanding gender as an effect of sex on one or more social-behavioural characteristics. Gender as an interaction between sex and the early-life social environment: if sex differences are not stable between social groups, we can explain these sex differences by gender mechanisms [9]. This third way of thinking about gender also refers to gender as a gender pressure as in conceptualization (b), but it takes into account the fact that the systemic process of gender varies, in its form or intensity, across social groups. E.g., if care activities are more often found in women in population A but in men in population B, we can conclude that the fact that care attitudes are associated with a sex is not “natural” but linked to systemic gender mechanisms. This is symmetrically equivalent to the fact that a given social environment does not have the same effect, through socialisation, on an individual, depending on their sex attributed at birth [9]. This conceptualisation involves understanding gender as a difference in effects, i.e. an interaction. In theory, this effect can concern the whole social environment, at any age, but to simplify the approach, we here considered only the early-life social environment, which is a priori independent of the sex at birth. This third conceptualisation can be seen as part of an intersectional approach to gender [23].Fig. 2Conceptual graphs for three conceptualizations of gender Here, we did not address gender as an experience of self (gender identity) or as a given kind of psyche (gender personality). The conceptualizations of gender imply specific analytical strategies to meet the objective of identifying gender mechanisms to explain sex differences in biomarkers (See “Analyses” section). We refer to the corresponding strategy by the letter for each corresponding conceptualization (a, b, c). ## Outcomes: biomarkers When individuals were about 45 years old, biomedical data were collected through a survey and a home-based clinical assessment (blood, saliva samples and anthropometric measurements) [24]. We explored several of these biomarkers, representing the four most frequent systems used to construct the score of allostatic load [25]: systolic blood pressure (SBP) for the cardiovascular system; triglycerides, low density lipoprotein (LDL) cholesterol and haemoglobin A1c (HbA1c) for the metabolic system; C-reactive protein (CRP) for the inflammatory system; and cortisol for the neuroendocrine system. When the distribution was too asymmetric, the variables were log transformed (triglycerides, CRP and cortisol). ## Exposures: sex and early social environment As our main exposure measure, we used sex attributed at birth. Relative to the effect of sex on the outcome, the early-life social environment was a competitive exposure and a confounder of the mediator-outcome relationship in strategies (a) and (b). In strategy (c), the early-life social environment was a modifier of the effect of sex on the outcome (see Fig. 2). We used two variables to characterise the early-life social environment at the time of the cohort member’s birth: educational level of the cohort member’s mother (school leaving age of 15 versus stayed at school beyond age 15, i.e., “O level”) and their other parent’s (or mother’s partner) social class (manual or non-manual social class). We used these variables to define two groups: the deprived group if the mother had a short education and the other parent a manual social class, and the non-deprived group in all other cases. ## Mediators: gender scores and socio-behavioural characteristics In strategies (a) and (b), we explore gender processes through mediator(s) (see Fig. 2). In strategy (a) gender was conceptualised as an individual characteristic which was measured by a gender score based on socio-behavioural variables. The mediator was this score. In strategy (b), gender was conceptualised as a sex effect on socio-behavioural characteristics. In this approach, mediators were the same socio-behavioural variables than those used to compute the gender score but kept separated. ## Choice of socio-behavioural characteristics In each strategy, we used the same set of socio-behavioural variables, either to compute the score (a) or separately (b). We consider that gender processes do not simply impact aspects of life classically described as gendered (like domestic load, type of occupation, etc.) but diffuses into all socio-behavioural dimensions. We had therefore chosen to use a larger set of socio-behavioural individual characteristics for which we assumed a priori to be distributed differently according to sex, because of the gender processes, and which may a priori have an impact on biomarkers and health. Gender processes are multi-level, multidimensional and highly diffuse [7], so much so that we could say that every aspect of a human's life is impacted by the gender norms of the society in which they live. It impacts their identity (“how an individual sees themselves”), their roles (behaviours, experiences, expectations), their relations (“how individuals interact with and are treated by others”) and their relative power in different institutions (“political, educational, religious, media, medical, cultural and social institutions”) [7]. It seems impossible to capture all the aspects of life impacted by gender phenomenon [9]. We therefore sought to characterize, as broadly as possible, various dimensions of social life from the data available in the cohort. Individual social characteristics that have an impact on health are equally multi-level, multi-dimensional and diffuse. They can be classified in two types: behaviours and social advantages/ disadvantages, i.e., resources which give the individual a varying degree of control and resilience over their environment, their experiences, and their life course. According to Bourdieu, these advantages/ disadvantages can be categorised into three dimensions: cultural capital, economic capital and social capital [26]. We therefore used several variables to characterise the three dimensions of capital and the behaviours. We made the a priori hypothesis that the distributions of these variables may vary according to sex, due to the gender processes:Cultural capital refers to knowledge, skills and integrated attitudes that will influence the way an individual sees the world, thinks, behaves, lives and acts [26]. In this study, we have represented these resources through five measures: “educational level at 23” (more or less than O level), “literacy at 23” (declared difficulties or not), “numeracy at 23” (declared difficulties or not), “often reads books at 23” (at least once a month) and “driver’s license at 33”. The driver's license is not a classic criterion of cultural capital, but we considered that it corresponded to the definition of a skill giving an increased control on the environment, an increased capacity to act in social life. Economic capital refers to the material and financial resources of individuals and the means to produce them [26]. In this study, we measured economic reserves with “personal savings at 23” and qualified the resources to produce them through “paid work at 33” and “social class at 33” (manual or non-manual).Social capital refers to an individual's social network, its size, value, and the degree of usefulness of these relationships [26]. In this study, we used the frequency of friends’ visits at 23 (more or less than once a week) as a marker of social support and “being religious at 23” as a marker of belonging to a community. The three variables “child(ren) at 23”, “married at 23”, and “doing laundry at 33” are markers of an affective and family support, but also of the domestic burden, counterpart of this resource. To characterize behaviours, we chose behavioural variables which can be considered as protective or a risk for health: “smoking (≥ 1 cigarette/day)” at 23 and 33, “everyday alcohol drinking” at 23 and 33, “frequent fried food” at 33 (more than once a week), “sport” at 23 and 33 (at least once a month). Risk taking being "a value and reality associated with masculinity" and which penalizes men by "causing them to perish" [27], we also used a proxy of risk-taking behaviours with the variable “have attended hospital or casualty department for any kind of accident or assault between 23 and 33”. ## Impact of the choice of variables The choice of variables to explore the gender phenomenon is largely based on their availability in the database and would have varied widely if another cohort had been used. To explore the impact of the variables availability on results, we constituted three different sets of variables and performed the analyses, for the strategies (a) and (b), with these different sets, "as if", in each case, we had only these variables to characterize the same phenomenon of interest. The sets were:Complete set: all the listed-above variables had been used. It was the main analysis. Behavioural set: only the behavioural variables had been used, as if only these variables were available. Small set: only 4 social characteristics (educational level at 23, social class at 33, frequency of friends’ visits at 23, and marital status at 23) and 4 variables for behaviours (sport, diet, smoking and alcohol at 33), as if only these variables were available. ## Scores computation In strategy (a), we wanted to capture gender as an individual characteristic, corresponding to the level at which the social characteristics and behaviours of an individual meets the standards of masculinity or femininity. To measure the gender corresponding to this definition, we used the "gender diagnosis" method [28–30]. The gender score produced by this method corresponds to a measure of the level at which an individual complies with a set of elements constituting femininity or masculinity in a given population, place and time, i.e., as the probability of being "predicted male or female" from social dimensions [9]. To construct the score, we modelled sex at birth by socio-behavioural characteristics using logistic regression, for each set of variables defined above. The gender score corresponded to the predicted probability by the model of sex at birth, a continuum from 0 “predict female by their socio-behavioural characteristics”, proxy of “gendered in a feminine way”, to 1 “predict male by their socio-behavioural characteristics”, proxy of “gendered in a masculine way”. ## Analyses All the analyses were performed with R release 4.1.3, with the Tidyverse packages. To deal with missing data, we performed a single stochastic imputation using the MICE package in R [31] on each of 1,000 bootstrapped databases, also used to computed $95\%$ confidence intervals [32, 33]. ## Descriptive analyses We first described the exposures, mediators, and outcomes in excluded participants (dead at the biomedical waves) and included subjects (living at the biomedical wave), with number of missing data, and number and percentage of the variable categories, or mean and standard for quantitative variables. We also described these variables in the imputed bootstrapped databases, with mean of percentages or means and confidence intervals (2.5 and 97.5 percentiles), computed on 1,000 bootstrapped imputed datasets. The results of this description are given in Additional file 1. We then described socio-behavioural characteristics and gender scores by sex, with mean value and confidence intervals (2.5 and 97.5 percentiles) of percentages or means, computed on 1,000 bootstrapped imputed datasets. ## Causal analyses The mean value of each biomarker at 44–45 years of age was estimated under several counterfactual scenarios that differed by exposure and/ or mediator assignment. The notation \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathbb{E}}\left({Y}_{A=a}\right)$$\end{document}EYA=a represents the expected potential outcome (mean of Y) in the counterfactual scenario in which the exposure is set to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A = a$$\end{document}A=a. Under the randomization assumption (no residual confounding) and the consistency assumption (effect of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A$$\end{document}A is the same whether observed or given by intervention), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathbb{E}}({Y}_{A=a}$$\end{document}E(YA=a) was estimated using g-computation. Linear regressions were used to estimate conditional expectations of the outcome, denoted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(A,L\right)= {\mathbb{E}}(Y|A,L)$$\end{document}Q¯A,L=E(Y|A,L), with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L$$\end{document}L, the confounders. From the estimated \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(A,L\right)$$\end{document}Q¯A,L functions, we predicted the value of biomarker \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y$$\end{document}Y for each member \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document}i under the counterfactual scenarios. Target causal parameters (estimands), described below, were defined in an additive scale as the difference between the mean of potential outcomes in two scenarios. ## Total effect of sex We first aimed to measure the size of sex-differences in biomarker. The estimands were the total effect \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TE$$\end{document}TE of sex at birth \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S$$\end{document}S on each biomarker \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Y$$\end{document}Y, defined as the difference between the mean outcome had all the population been born male and the mean outcome had all the population been born female, denoted:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TE= {\mathbb{E}}[{Y}_{S=male }-{Y}_{S=female}]$$\end{document}TE=E[YS=male-YS=female] In the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(S,E\right)$$\end{document}Q¯S,E functions used to estimate the potential outcomes, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E contained the early-life social environment variable. We included an interaction term between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S$$\end{document}S and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E. The outcomes were first used in their original scale, the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TE$$\end{document}TE is therefore expressed in these units of measure, e.g., in mmHg for systolic blood pressure. They were then standardized as: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$z=\frac{y- \mu }{\sigma }$$\end{document}z=y-μσ where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mu$$\end{document}μ and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sigma$$\end{document}σ are the mean and standard deviation of the outcome in each imputed bootstrapped data set. The \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$TE$$\end{document}TE is therefore expressed in standard deviation, e.g., a total effect of 1 corresponds to a mean difference of 1 standard deviation between men and women. ## Strategy (a): mediation by a gender score The principal objective of the study was to identify gender mechanisms that explain sex-differences in biomarkers. With gender conceptualized as an individual characteristic (a), the estimand corresponded to the proportion of the total effect of sex on biomarkers which disappear when all the individuals are gendered in the same way, i.e., the eliminated proportion by gender score \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G$$\end{document}G, denoted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{G}$$\end{document}EPG. The eliminated proportion \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{G}$$\end{document}EPG was measured as the difference between the total effect of sex and the remaining effect of sex when all the population is gendered in the same way (gender score fixed at 0.5), divided by the total effect of sex [34]. The remaining effect of sex when all the population is gendered in the same way corresponds to the controlled direct effect \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${CDE}^{G}$$\end{document}CDEG, which was defined here as the difference between the mean outcome had all the population been born male and the mediator (gender score) set at a given value (here 0.5) and the mean outcome had all the population been born female and the mediator (gender score) set at the same value (0.5) [35]. In the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(S,E\right)$$\end{document}Q¯S,E and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(S,G,E\right)$$\end{document}Q¯S,G,E functions used to estimate the potential outcomes, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E contained the early-life social environment variable (mediator-outcome confounder). We included an interaction term between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S$$\end{document}S and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E, but not with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G$$\end{document}G. Finally, we had: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${CDE}^{G}= {\mathbb{E}}[{Y}_{S=male, $G = 0.5$}-{Y}_{S=female,$G = 0.5$}]$$\end{document}CDEG=E[YS=male,$G = 0.5$-YS=female,$G = 0.5$] and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{G}=\frac{TE-{CDE}^{G}}{TE}$$\end{document}EPG=TE-CDEGTE ## Strategy (b): mediation by social characteristics With gender conceptualized as an effect of sex on socio-behavioural characteristics (b), the estimand corresponded to the proportion of the total effect of sex on biomarkers which disappears when all the individuals have the same socio-behavioural characteristics, i.e., the eliminated proportion by the set of socio-behavioural characteristics \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Sigma$$\end{document}Σ, denoted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{\Sigma }$$\end{document}EPΣ. The eliminated proportion \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{\Sigma }$$\end{document}EPΣ was measured as the difference between the total effect of sex and the remaining effect of sex when all the population has the same socio-behavioural characteristics (see Additional file 1 for detailed fixed values \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }^{*}$$\end{document}ε∗ for each variables), divided by the total effect of sex [34]. The remaining effect when all the population has the same socio-behavioural characteristics corresponds to the controlled direct effect \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${CDE}^{\Sigma }$$\end{document}CDEΣ, which was defined here as the difference between the mean outcome had all the population been born male and all the socio-behavioural characteristics set at a given value (see Additional file 1) and the mean outcome had all the population been born female and all the socio-behavioural characteristics set at the same value [35]. In the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(S,E\right)$$\end{document}Q¯S,E and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(S,\Sigma,E\right)$$\end{document}Q¯S,Σ,E functions used to estimate the potential outcomes, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E contained the early-life social environment variable. We included an interaction term between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S$$\end{document}S and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E, but not with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Sigma$$\end{document}Σ. We therefore had: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${CDE}^{\Sigma }= {\mathbb{E}}[{Y}_{S=male,\Sigma ={\varepsilon }^{*}}-{Y}_{S=female,\Sigma ={\varepsilon }^{*}}]$$\end{document}CDEΣ=E[YS=male,Σ=ε∗-YS=female,Σ=ε∗] and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{\Sigma }=\frac{TE-{CDE}^{\Sigma }}{TE}$$\end{document}EPΣ=TE-CDEΣTE ## Strategy (c): considering an interaction between sex and the early-life social environment With gender conceptualized as an interaction between the sex at birth and the social environment (c), the estimand corresponded to the proportion of the total effect of sex on biomarkers which disappears when all the individuals have a non-gendered social environment. We considered that an observed non-gendered social environment was not realistic, so we rather considered a “less-gendered environment”. Here, to define the social environment, we considered only the early-life social environment \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E. We made the a priori hypothesis that the non-deprived group was less gendered than the deprived group. Therefore, the eliminated proportion \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{\mathrm{E}}$$\end{document}EPE was measured as the difference between the total effect of sex and the remaining effect of sex when all the population is exposed to a non-deprived early-life social environment \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$$E = 0$$$\end{document}$E = 0$, divided by the total effect of sex. The remaining effect when all the population is exposed to a non-deprived early-life social environment corresponds to the total effect of sex when \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$$E = 0$$$\end{document}$E = 0$, denoted \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${TE}^{0}$$\end{document}TE0 and defined as the difference between the mean outcome had all the population been born male and the early-life social environment been non-deprived, and the mean outcome had all the population been born female and the early-life social environment been non-deprived. The model \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{Q }\left(S,E\right)$$\end{document}Q¯S,E used to estimate potential outcomes under these scenarios considered an interaction term between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S$$\end{document}S and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E$$\end{document}E. We finally had: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${TE}^{0} = {\mathbb{E}}[{Y}_{S=male,$E = 0$}-{Y}_{S=female,$E = 0$}]$$\end{document}TE0=E[YS=male,$E = 0$-YS=female,$E = 0$] and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${EP}^{\mathrm{L}}=\frac{TE-{TE}^{0} }{TE}$$\end{document}EPL=TE-TE0TE ## Complementary analyses regarding interactions We also present in the results section a more detailed description of the interaction effects, with mean value and confidence intervals (2.5 and 97.5 percentiles) of means, computed on 1,000 bootstrapped imputed datasets, for each biomarker in each category of sex and early-life social environment. We also computed the total effect of sex in each stratum of early social environment and total effect of early social environment in each stratum of sex. We finally estimated the interaction effect \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$IE$$\end{document}IE of sex and early social environment, defined as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$IE = \,{\mathbb{E}}[{Y}_{S=male $E = 1$}-{Y}_{S=male,$E = 0$}-{Y}_{S=female,$E = 1$}+{Y}_{S=female,$E = 0$}]$$\end{document}IE=E[YS=maleE=1-YS=male,$E = 0$-YS=female,$E = 1$+YS=female,$E = 0$] ## Description of population We included the 17,272 participants alive at the biomedical waves, $51\%$ of whom were born male (see Additional file 1 for detailed description, Table a). Within our sample, $75\%$ had a short-educated mother at their birth, and for $73\%$, the other parent was from manual social class. The 1,286 participants who died before the biomedical survey at 44 years (see Additional file 1: Table a) were more often men ($59\%$) and more often socially deprived ($80\%$ had a short-educated mother at their birth and for $78\%$, the other parent was from manual social class). The sensitivity analysis population—those with complete data for all the waves—(see Additional file 1: Table d) is globally more advantaged and with fewer men ($48\%$) than in the main analysis population. Most of the analysed variables of social resources, experiences, and behaviours at 23 and 33 were differently distributed according to sex at birth (see Table 2). In this specific population, on average, male-born and female-born individuals did not have the same type of cultural capital: a higher percentage of women than men had a level of education above O-level, declared fewer literacy difficulties and read more frequently than men, while a higher percentage of men had a driving licence than women. On average, economic capital also differed: a higher percentage of women had a non-manual social class at 33 compared to men, whereas men had a higher amount of savings at 23 and had more frequently a paid job at 23. Social capital was higher among women: a higher proportion of them were married and with children at age 23, reported being part of a religious community and saw their friends more frequently. As a counterpart, they were also much more likely to carry the domestic load, measured here by the laundry load. Regarding behaviours, fatty diets, risk-taking behaviours, regular alcohol consumption and smoking were more common in individuals born male, as the fact of being physically active at age 23.Table 2Distribution of social characteristics at 23 and 33 by sex, NCDS-58 cohort ($$n = 17$$,272)Male-bornFemale-bornM—F%$95\%$CI%$95\%$CI%Cultural capital Less than O level at 2342.6[41.2 to 43.9]35.9[34.6 to 37.2] + 6.7 No numeracy problems at 2394.8[94.1 to 95.4]94.5[93.8 to 95.1] + 0.3 Literacy problems at 2312.6[11.6 to 13.5]7.0[6.3 to 7.8] + 5.6 Does not often read at 2345.6[44.2 to 46.9]32.6[31.2 to 34.0] + 13.0 Driver’s license at 3392.5[91.7 to 93.3]80.9[79.6 to 82.1] + 11.6Economic capital Personal savings > median at 2354.5[53.0 to 56.0]45.7[44.4 to 47.2] + 8.8 Paid work at 2390.1[89.1 to 91.0]68.5[67.2 to 69.9] + 21.6 Manual social class at 3350.8[49.3 to 52.4]33.4[32.0 to 34.9] + 17.4Social capital Does not often see friend at 2335.6[34.1 to 37.0]27.6[26.4 to 28.8] + 8.0 Not married at 2365.2[63.9 to 66.5]45.7[44.4 to 47.1] + 19.5 No child at 2382.0[80.9 to 83.1]66.6[65.3 to 67.8] + 15.4 Does not do laundry at 3386.0[84.7 to 87.3]4.2[3.5 to 4.9] + 81.8 *Not religious* at 2349.6[48.2 to 51.1]32.5[31.2 to 33.8] + 17.1Behaviours Smoking at 2339.9[38.6 to 41.2]38.3[36.9 to 39.5] + 1.6 Smoking at 3332.6[31.3 to 34.0]32.0[30.5 to 33.4] + 0.6 Alcohol every day at 2331.4[30.1 to 32.7]9.8[9.0 to 10.6] + 21.6 Alcohol every day at 3317.6[16.4 to 18.7]7.0[6.4 to 7.8] + 10.6 Often eats fried food at 3356.6[55.0 to 58.2]35.0[33.7 to 36.5] + 21.6 Often practices sport at 2359.3[57.8 to 60.7]35.8[34.5 to 37.1] + 23.5 Does not practice sport at 3377.8[76.4 to 79.0]77.4[76.1 to 78.7] + 0.4 Accident between 23 and 3358.3[56.8 to 59.8]24.8[23.5 to 26.0] + 33.5Gender scores, total population Complete set*0.87[0.85 to 0.88]0.13[0.12 to 0.14] + 0.74 Behavioural set*0.61[0.60 to 0.62]0.37[0.36 to 0.38] + 0.24 Small set*0.56[0.55 to 0.57]0.42[0.41 to 0.43] + 0.14Gender scores, in deprived-born group Complete set*0.87[0.86 to 0.88]0.12[0.11 to 0.13] + 0.75 Behavioural set*0.62[0.61 to 0.64]0.37[0.36 to 0.38] + 0.25 Small set*0.55[0.55 to 0.57]0.41[0.40 to 0.42] + 0.14Gender scores, in advantaged-born group Complete set*0.85[0.84 to 0.87]0.14[0.13 to 0.16] + 0.71 Behavioural set*0.59[0.58 to 0.60]0.37[0.36 to 0.38] + 0.22 Small set*0.57[0.56 to 0.58]0.43[0.42 to 0.45] + 0.14All categorical variables were binary. Showing variable categories that were the most frequent among male-born participants“$95\%$CI” corresponds to the confidence intervals computed on 1,000 bootstrapped imputed datasets“M-F” corresponds to the male to female differences of observed probabilities (%)* = quantitative variables, mean and $95\%$CI of mean are given Regarding the gender scores, the mean was 0.13 (sd = 0.19) in female-born individuals and 0.87 (sd = 0.24) in male-born participants (complete set). The higher the number of variables used to estimate the score, the more discriminating the score. Groups defined by the early-life social environment were slightly differently gendered. Results confirmed the a priori hypothesis that a non-deprived early-life social environment was in a certain way less gendered than deprived early-life social environment, as the sex-gap was (slightly) smaller in the non-deprived group. The sensitivity analysis performed only on participants who attended all 4 waves of data collection ($$n = 7021$$) showed similar distributions (see Additional file 1 Table e). ## Total effect of sex on biomarkers The distributions of all the analysed biomarkers were significantly different according to sex at birth (see Table 3). Cardiometabolic biomarkers were all on average higher for male-born individuals, whereas inflammatory and endocrine biomarkers were on average higher for female-born individuals. For example, the systolic blood pressure at 44–45 years old was on average 12.45 mmHg (95CI = [11.77 to 13.18]) higher in individuals born male than in those born female. However, the size of differences varied and was always smaller than 1 standard deviation (sd), from − 0.07 [− 0.14 to − 0.01] standard deviation for logarithm of cortisol, to + 0.75 [0.71 to 0.79] standard deviation for systolic blood pressure. Table 3Total effect (TE) of being born male on biomarkers at 44–45, NCDS-58 cohort ($$n = 17$$,272)FemaleOriginal scaleZ-scoresMean (sd)TE$95\%$CITE$95\%$CISystolic Blood Pressure (mmHg)120.5 (15.5) + 12.45[11.77 to 13.18] + 0.75[0.71 to 0.79]Log (Triglycerides (g/L))0.32 (0.5) + 0.42[0.39 to 0.45] + 0.70[0.65 to 0.75]LDL Cholesterol (mmol/L)3.30 (0.9) + 0.31[0.25 to 0.37] + 0.33[0.27 to 0.39]HbA1c (%)5.21 (0.6) + 0.13[0.09 to 0.16] + 0.17[0.13 to 0.22]Log (CRP (mg/L))0.13 (1.3)− 0.13[− 0.2 to − 0.05]− 0.11[− 0.17 to − 0.04]Log (Cortisol (µg))2.94 (0.5)− 0.04[− 0.08 to − 0.01]− 0.07[− 0.14 to − 0.01]TE = total effect of being born male rather than female; $95\%$CI = bootstrapped confidence intervals ($$n = 1$$,000) In comparison, being born into a deprived family was associated with increased levels of all biomarkers from + 0.04 sd [− 0.02 to 0.10] for cortisol to + 0.20 sd [0.13 to 0.26] for CRP (see Additional file 1: Table c). The sensitivity analysis performed only on participants who attended the 4 waves of data collection ($$n = 7021$$) yielded similar conclusions (see Additional file 1: Tables f and g) ## Explained proportion of sex effect Strategies (a) and (b) provided very similar results (see Table 4). According to the set of variables used to estimate the eliminated proportion, results varied. We describe results of strategies (a) and (b) in this section and results of strategy (c) in the following section. Table 4Eliminated proportion (EP) of sex effect, NCDS-58 cohort ($$n = 17$$,272)(a) By gender-score mediator(b) By socio-behavioural mediators(c) By early social environmentEP (%)$95\%$CIEP (%)$95\%$CIEP (%)$95\%$CISystolic Blood Pressure− 1.86[− 6.7 to 3.0] Complete set9.31[− 2.6 to 21.4]9.15[− 3.2 to 21.1] Behavioural set2.12[− 1.7 to 5.9]2.04[− 1.9 to 5.9] Small set4.38[1.5 to 6.9]*4.32[1.4 to 6.9]*Log (Triglycerides)− 7.67[− 13.2 to − 2.1]* Complete set− 2.51[− 18.7 to 12.4]− 5.52[− 22.0 to 10.2] Behavioural set− 0.22[− 5.0 to 4.6]− 0.35[− 5.3 to 4.4] Small set1.32[− 2.1 to 4.9]1.43[− 2.0 to 5.1]LDL Cholesterol− 14.28[− 29.0 to − 1.8]* Complete set2.75[− 28.6 to 33.5]− 8.87[− 42.5 to 23.4] Behavioural set− 0.03[− 10.1 to 10.0]0.01[− 9.9 to 10.1] Small set− 1.90[− 9.4 to 5.7]− 1.69[− 9.3 to 5.8]HbA1c− 4.95[− 28.1 to 17.0] Complete set− 7.88[− 71.6 to 50.0]− 32.73[− 98.2 to 29.4] Behavioural set− 13.20[− 31.6 to 4.9]− 14.60[− 32.5 to 3.7] Small set6.27[− 7.6 to 21.9]6.24[− 7.4 to 21.7]Log (CRP)63.34[23.2 to 170.0]* Complete set− 20.31[− 158.4 to 103.2]13.93[− 121.5 to 160.8] Behavioural set0.53[− 43.4 to 38.5]1.24[− 42.5 to 38.8] Small set− 16.45[− 63.3 to 11.9]− 16.33[− 64.0 to 13.1]Log (Cortisol)− 8.46[− 121.3 to 71.1] Complete set108.41[− 143.5 to 562.0]112.98[− 136.4 to 545.9] Behavioural set5.90[− 72.8 to 99.3]6.44[− 70.2 to 97.7] Small set0.64[− 70.3 to 69.1]1.24[− 69.8 to 68.7]EP = Eliminated proportion; $95\%$CI = bootstrapped confidence intervals ($$n = 1000$$); * = $95\%$CI does not include zero Cardiovascular biomarker: Systolic blood pressure at 44 was on average higher in male-born individuals of this population. Setting the gender score to 0.5 (a) or the socio-behavioural characteristics at their values of reference (b), we eliminated up to $9.3\%$ [− 2.6 to 21.4] of the total effect of sex. Metabolic biomarkers: Triglycerides, LDL Cholesterol and HbA1c at 44 were on average higher in male-born individuals. Setting the gender score to 0.5 (a) or the socio-behavioural characteristics at their values of reference (b), the eliminated proportions of sex effect on lipids varied from − $9\%$ to + $3\%$ according to the considered set of variables and were never statistically significant. Regarding HbA1c, setting the gender score to 0.5 (a) or the socio-behavioural characteristics at their values of reference (b), the eliminated proportions of sex effect varied from − $33\%$ to + $6\%$ according to the considered set of variables and were never statistically significant. Inflammatory biomarker: CRP at 44 was on average lower in male-born individuals. Setting the gender score to 0.5 (a) or the socio-behavioural characteristics at their values of reference (b), the eliminated proportions of sex effect varied from − $20\%$ to + $14\%$ according to the considered set of variables and were never statistically significant. Neuroendocrine biomarker: Cortisol at 44 was on average lower in male-born individuals. Setting the gender score to 0.5 (a) or the socio-behavioural characteristics at their values of reference (b), the eliminated proportions of sex effect varied from + $1\%$ to + $112\%$ according to the considered set of variables and were never statistically significant. The sensitivity analysis performed only on participants who attended the 4 waves of data collection ($$n = 7021$$) yielded the same conclusions (see Additional file 1: Table h). ## Results regarding interaction Results of strategy (c) was difficult to interpret only with the eliminated proportion. Moreover, the results seemed contradictory with the strategies (a) and (b) for several biomarkers. A more detailed description of interaction effects is given Table 5, with, for each biomarkers: mean values in each category of sex and early-life social environment (in original scale and as z-scores); total effect of sex in each stratum of early-life social environment; total effect of early-life social environment in each stratum of sex, additive interaction between sex and early-life social environment, as defined in the Methods section. Table 5Effects of sex and early-life social environment on biomarkers, NCDS-58 cohort ($$n = 17$$,272)Advantaged- bornDeprived-bornTE of early deprivationSBPOriginal scaleMale born (mean)132.0[131.3 to 132.7]133.6[133.0 to 134.3] + 1.6[0.7 to 2.5]Female born (mean)119.3[118.6 to 120.1]121.3[120.6 to 121.9] + 1.9[1.0 to 2.9]TE of being born male + 12.7[11.8 to 13.6] + 12.3[11.5 to 13.2](− 0.38)[− 1.38 to 0.60]Z-scoreMale born (mean)0.31[0.27 to 0.35]0.40[0.38 to 0.43] + 0.09[0.04 to 0.15]Female born (mean)− 0.46[− 0.5 to − 0.42]− 0.34[− 0.37 to − 0.31] + 0.12[0.06 to 0.18]TE of being born male + 0.77[0.72 to 0.82] + 0.74[0.70 to 0.79](− 0.03)[− 0.08 to 0.04]Log(triglycerides)Original scaleMale born (mean)0.70[0.67 to 0.73]0.77[0.74 to 0.80] + 0.07[0.03 to 0.11]Female born (mean)0.25[0.22 to 0.28]0.37[0.34 to 0.40] + 0.12[0.09 to 0.16]TE of being born male + 0.45[0.41 to 0.49] + 0.40[0.36 to 0.44](− 0.05)[− 0.09 to − 0.01]Z-scoreMale born (mean)0.27[0.22 to 0.32]0.39[0.35 to 0.42] + 0.12[0.06 to 0.18]Female born (mean)− 0.48[− 0.53 to − 0.43]− 0.28[− 0.32 to − 0.24] + 0.20[0.14 to 0.26]TE of being born male + 0.75[0.69 to 0.82] + 0.67[0.61 to 0.72](− 0.08)[− 0.15 to − 0.02]LDL cholesterolOriginal scaleMale born (mean)3.58[3.52 to 3.63]3.63[3.57 to 3.70] + 0.05[− 0.01 to 0.12]Female born (mean)3.23[3.18 to 3.28]3.35[3.30 to 3.40] + 0.12[0.06 to 0.18]TE of being born male + 0.35[0.29 to 0.42] + 0.28[0.21 to 0.35](− 0.07)[− 0.13 to − 0.01]Z-scoreMale born (mean)0.13[0.08 to 0.17]0.18[0.14 to 0.22] + 0.05[− 0.01 to 0.12]Female born (mean)− 0.25[− 0.29 to − 0.20]− 0.12[− 0.16 to − 0.07] + 0.13[0.07 to 0.19]TE of being born male + 0.37[0.31 to 0.44] + 0.30[0.23 to 0.37](− 0.07)[− 0.14 to − 0.01]HbA1cOriginal scaleMale born (mean)5.29[5.26 to 5.33]5.36[5.33 to 5.40] + 0.07[0.03 to 0.12]Female born (mean)5.16[5.14 to 5.19]5.24[5.21 to 5.28] + 0.08[0.05 to 0.12]TE of being born male + 0.13[0.09 to 0.17] + 0.12[0.08 to 0.16](− 0.01)[− 0.05 to 0.04]Z-scoreMale born (mean)0.02[− 0.02 to 0.07]0.12[0.09 to 0.16] + 0.10[0.04 to 0.16]Female born (mean)− 0.16[− 0.19 to − 0.12]− 0.04[− 0.08 to − 0.01] + 0.11[0.07 to 0.15]TE of being born male + 0.18[0.12 to 0.23] + 0.17[0.11 to 0.22](− 0.01)[− 0.07 to 0.05]Log (CRP)Original scaleMale born (mean)− 0.11[− 0.18 to − 0.03]0.07[0.01 to 0.13] + 0.17[0.09 to 0.26]Female born (mean)− 0.06[− 0.13 to 0.02]0.25[0.18 to 0.32] + 0.31[0.22 to 0.4]TE of being born male− 0.05[− 0.13 to 0.04]− 0.18[− 0.27 to − 0.09](− 0.13)[− 0.21 to − 0.05]Z-scoreMale born (mean)− 0.14[− 0.19 to − 0.08]0.00[− 0.04 to 0.05] + 0.14[0.07 to 0.22]Female born (mean)− 0.1[− 0.16 to − 0.05]0.15[0.11 to 0.20] + 0.25[0.18 to 0.33]TE of being born male− 0.04[− 0.11 to 0.03]− 0.15[− 0.22 to − 0.08](− 0.11)[− 0.17 to − 0.04]Log (Cortisol)Original scaleMale born (mean)2.89[2.86 to 2.92]2.91[2.89 to 2.94] + 0.03[− 0.01 to 0.06]Female born (mean)2.93[2.9 to 2.96]2.95[2.92 to 2.98] + 0.02[− 0.02 to 0.06]TE of being born male− 0.04[− 0.08 to 0.00]− 0.04[− 0.07 to 0.00](+ 0.00)[− 0.03 to 0.04]Z-scoreMale born (mean)− 0.06[− 0.12 to − 0.01]− 0.02[− 0.06 to 0.02] + 0.05[− 0.02 to 0.11]Female born (mean)0.01[− 0.03 to 0.07]0.05[0.01 to 0.09] + 0.04[− 0.03 to 0.10]TE of being born male− 0.08[− 0.15 to − 0.01]− 0.07[− 0.14 to 0.00](+ 0.01)[− 0.05 to 0.08]SBP = Systolic Blood Pressure; TE = Total effect; $95\%$CI = bootstrapped confidence intervals ($$n = 1$$,000); results in bold and in brackets are the measures of additive interaction \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${(Y}_{11}-{Y}_{10})-\left({Y}_{01}-{Y}_{00}\right)$$\end{document}(Y11-Y10)-Y01-Y00 Regarding cardiometabolic biomarkers, we observed that, contrary to what was expected, the sex gaps were smaller (negative eliminated proportions) or tended to be observed in people deprived versus non-deprived in early life. The detailed results (see Table 5) showed that a deprived early-life social environment increased the level of the four cardiometabolic biomarkers at 44 years old and these effects on lipids were stronger in women. E.g., the total effect of being born deprived on LDL cholesterol was + 0.05 standard deviation [− 0.01 to 0.12] in men and + 0,13 standard deviation [0.07 to 0.19] in women; so, the sex gap was + 0.37 standard deviation [0.31 to 0.44] in non-deprived group against + 0.30 standard deviation [0.23 to 0.37] in deprived group. In other words, being born male and being born deprived increased, on average, the level of triglycerides and LDL cholesterol, but the effect of being born deprived was stronger in women. So, symmetrically, the effect of being born male was smaller in the deprived group, which explained the negative eliminated proportions of the sex effect by being born into a non-deprived early social environment (see Table 4). Regarding the inflammatory biomarker, the sex gap was reduced by almost $64\%$ in born non-deprived group (see Table 4). Indeed, the total effect of sex was − 0.04 standard deviation [− 0.11 to 0.03] in the non-deprived group and − 0.15 standard deviation [− 0.22 to − 0.08] in the deprived group (see Table 5). For this biomarker too, the effect of early deprivation was stronger in women than in men: + 0.25 [0.18 to 0.33] versus + 0.14 [0.07 to 0.22] standard deviation. Regarding the neuroendocrine biomarker, the sex effect was similar in each group of early social environment and, symmetrically, the early-deprivation effect was similar between men and women. The sensitivity analysis performed only on participants who attended the 4 waves of data collection ($$n = 7021$$) yielded the same conclusions (see Additional file 1: Table i). ## Main results The distribution of each biomarker was significantly different according to sex at birth, higher in men for cardiometabolic biomarkers and higher in women for inflammatory and neuroendocrine biomarkers. The sizes of the differences were always smaller than one standard deviation but were larger than differences due to early-life deprivation, except for CRP. We observed gender mechanisms underlying these differences between men and women, even if the mediation effect was rarely statistically significant. These mechanisms were of three kinds: [1] mediation by socio-behavioural characteristics, e.g., for SBP, $4.3\%$ of the observed sex differences significantly disappeared when socio-behavioural characteristics measured at 23 and 33 years or gender score were set; [2] attenuation by gendered mechanisms: e.g., for HbA1c, sex differences tended to be larger when the socio-behavioural characteristics at 23 and 33 years were set to a common level; [3] interaction with early social environment: for metabolic and inflammatory biomarkers; e.g., we observed that being born into a deprived rather than non-deprived family increased triglycerides, LDL cholesterol and CRP levels more strongly in females than in males. ## Comparison with literature In our study, cardiometabolic biomarkers (systolic blood pressure (SBP), triglycerides, LDL cholesterol and HbA1c) were on average higher in male-born individuals. The differences in SBP were partly mediated by the studied socio-behavioural characteristics. For triglycerides and LDL cholesterol, the sex difference varied with social environment at birth, suggesting that the effect of deprivation is different according to sex at birth and therefore that levels of serum lipids are influenced by gender mechanisms. These results are consistent with prior knowledge: if innate biologic factors explained a part of the biomarkers levels variability [36] and sex hormone levels [37], lifestyle also plays an important role on the whole cardiovascular and metabolic system. Indeed, medications, smoking, alcohol consumption, diet, sedentary lifestyle, obesity, etc. are identified risk factors of dyslipidaemia [38–40], high blood pressure [41, 42] and diabetes [43, 44]. These behaviours varying by sex at birth, through gender processes, it seems expected that part of the sex differences can be explained by socio-behavioural mechanisms. Pelletier et al. also showed that risks of hypertension and diabetes varied by sex and that these effects were partially mediated by a gender score in a different population [30]. This also confirms that sex disparities are partly explained by gender mechanisms. In our study, CRP levels were on average slightly higher in female-born individuals. In the mediation approaches, the sex difference tended to be larger when socio-behavioural characteristics or gender scores were set at the same value. But these effects were not stable nor statistically significant. However, the sex difference varied with early social environment: female-born individuals had a greater increase in CRP with deprivation than male-born individuals. Genetic factors would account for approximately 40 to $50\%$ of the variance of CRP [45]. Besides, genes related to immunity have been identified on the X chromosome, and would explain some immune differences between men and women, including the higher immune reactivity and the higher risk of autoimmune pathologies in women [46]. Oestrogen levels, particularly during pregnancy or menopause, are also known to interfere with CRP [45]. However, independently or in interaction with genetic factors, the levels of inflammatory biomarkers are higher in cases of high body fat, obesity, diabetes, smoking and hypertension, lower in case of certain deficiencies and with alcohol consumption and vary with diet, physical activity, year of schooling and with the use of certain medications [45, 47–49]. These socio-behavioural determinants varying by sex at birth, through gender processes, it seems expected that part of the sex differences can be explained by socio-behavioural mechanisms. In our study, cortisol levels were on average slightly higher in female-born individuals. In the mediation approaches, the sex-difference tended to disappear when socio-behavioural characteristics or gender scores (complete set) were set at the same value. But these effects were not stable nor statistically significant. The sex difference did not vary either with early social environment. Differences between men and women in cortisol responsiveness to stress have been extensively described [18]. In particular, cortisol levels vary when oestrogen levels vary (menstrual cycle, pregnancy contraceptive use) [50]. However, there are socio-behavioural factors involved in the variability of cortisol of cortisol levels: the number and intensity of stressful situations [2], the perception of stress [51, 52], the type of stressors and their context [13, 50, 53–55]. In our study, we were not able to explain the differences between men and women in cortisol levels by gender mechanisms, but the observed differences were very small. We assume that, although differences have been observed in experimental situations of acute stress, salivary cortisol seems to be rather insensitive to identify differences in the daily life of the general population. While cortisol is a central biomarker of the allostatic load theory, it is the only biomarker that does not increase significantly with early social disadvantage, contrary to what one might expect [56, 57]. These findings provide a better understanding of the gendered biological incorporation, as a pathway to explain the links described in the literature between sex/gender and mental and physical health [30, 58–60] ## Strengths and limits In this study, we used a large and well-known cohort, with a prospective collection of social and health data. Studying this cohort however involves several limitations. First, a large part of participants was lost to follow-up, and those are more disadvantaged than others, which can lead to selection bias. We chose therefore to include all the living subject (no missing data for death status) at the time of outcome collection to limit these selection bias. Secondly, and as a consequence of including not-attending participants, some data are missing, especially concerning biomarkers. To deal with this, we have made single stochastic imputations on 1,000 bootstrapped databases. We also performed a sensitivity analysis only on participants who attended all the four waves of data collection that we used, which produced similar results. Our approach remained focused on an individual-centred definition of gender. We have not addressed the issues of gender structural impact, sexism, discrimination affecting queer people, etc., which are of course part of the gender issue, and which certainly also impact biology and health. Within our specified scope, we had defined three methodological strategies to explore how gender could explain health differences between men and women. The use of several strategies, with several sets of variables, allowed us to capture different phenomena, explore the robustness of our analyses and reveal the advantages and disadvantages of these different approaches, as summarized in Table 6.Table 6How to capture Gender effect: advantages and disadvantages of three approachesApproachesMain advantagesMain disadvantages(a) Gender as an individual characteristic∙ *Through a* single score, allows the measurement of a total effect of gender∙ Sensitive to the type and number of variables to compute the score∙ Difficult to interpret and conceptually questionable(b) Gender as a sex-differentiated distribution of socio-behavioural characteristics∙ Less information lost∙ Less difficult to interpret∙ Different paths could be analysed in a second step∙ Sensitive to the type and number of variables∙ Gender total effect not directly measurable(c) Gender as a differential effect of the social environment depending on the sex at birth∙ More consistent with the systemic concept of gender∙ Depends on the heterogeneity of gender processes between the social groups in the studied population∙ Complexity of the choice of scale, the presentation of results and the interpretation∙ Only early-life variables can be used with these method (cannot be mediators of sex or other methods would be needed) We consider that the strategy (c) is more consistent with the systemic concept of gender, whose variation across social categories is inherent. Intersectionality is a term proposed in the 1980s by the American law professor Kimberlé Crenshaw to refer to the fact that the relations of domination experienced by racialised women are not the same as those experienced by white women (Black Feminism movement) [61]. The term now refers more generally to the fact that the experiences of any individual are situated at the intersections of multiple categories, including gender, but also class, race, age, sexual orientation, ability, etc., which influence each other to create distinct experiences in different combinations [23]. In other words, the effect of these different categories do not add up but "intersect in dynamic, complex, and surprising ways depending on context" [23]. From an analytical point of view, the intersectional approach allows to question which variables are relevant to understanding a health phenomenon, but also how they interact with each other and how their effects vary over time or between populations [23]. This analytical framework has been used primarily in the social sciences, but many quantitative methods also allow for an intersectional analysis of a phenomenon in an intersectional way, such as interaction analyses [62, 63], which we applied here. However, in our study, the population seemed to be relatively homogeneous in term of gender processes, which reduces the relevance of the interaction approach with the early-life social deprivation here. The study of the intersection with ethnicity/ race/ migration might have been more sensitive, as it has been shown that the intersectional effect of these categories on health can be strong [58, 64], but this was not possible due to a small number of minoritized individuals in the study data. It will be interesting to replicate this analysis in other populations, as the dynamics of gender phenomena can vary extensively according to the generation, age, culture, context, etc., and the consequences on observed sex differences are probably also very diverse from one population to another. In a previous work [9], we had proposed two strategies: one based on an individual approach of gender, measured by a gender score and the other based on an interaction analysis. Here we propose an additional approach (b), close to the strategy (a) but without a score. These two strategies lead to similar results. However, they do not have the same conceptual and methodological implications. First, the production of a score is, in our opinion, more at risk of over-interpretation of the results by forgetting the loss of information, simplification and non-exhaustiveness of an attempt to capture such a diffuse, complex, multilevel, intersectional phenomenon through one variable. It could also lead to an essentialization, and immobilization of what gender would be. In this sense, considering gender as being an effect of sex on socio-behavioural characteristics rather than directly composed by these characteristics, seems to us less risky. Secondly, the score complexifies the interpretation: it aims to capture the latent, diffuse phenomenon of gender performance that would have an impact on biology through various mechanisms which are not necessarily the variables used to calculate the score. Using these variables separately (b) implies, on the contrary, more causal, mechanistic hypotheses, i.e., that the effect may actually pass through each of the identified variables. This strategy renders assumptions and interpretations less obscure and gives more guidance for interventions [60]. Finally, the level of a gender score is not necessarily due to gender pressure alone, but also to other factors such as social class, age, generation, etc. [ 58]. The effect of this type of variable is therefore difficult to interpret as a strict gender effect, which would not include part of an effect of class, age, culture, etc. Again, the use of separate variables avoids over-interpretation of the results. One of the central points of our work concerning the consideration of gender in epidemiology is to decentralise the problem from a question of gender measure, to propose a more dynamic and structural approach, focused on the strategy. We demonstrated that we can analyse the impact of gender without a gender variable, by conceptualising it as a structural phenomenon and operationalise it as an effect or a difference in effect. ## Perspectives This study has identified gender mechanisms but has also opened many questions, which themselves raise complex methodological issues. First, it would seem interesting to explore more finely the paths of mediation involved. For example, we explained almost $10\%$ of the sex-differences in SBP, but what exactly was this due to? To answer this, it would require making more precise hypotheses on the causal sequences between the different mediators in order to consider the intermediate confounding. Secondly, it would be potentially important to explore the interaction between sex and mediators. Indeed, the social characteristics, not only at birth but also across the life-course, probably does not have the same effect depending on the sex at birth. For example, having a paid job at 23 probably does not have the same impact on a woman's life than on a man's, as it might not have the same “value” socially. It would therefore be interesting to continue this work in order to consider not only other early-life social categories (material condition, social support, etc.), but also other social categories throughout the life course. The 4-way decomposition proposed by VanderWeele could allow us to explore these phenomena [65], but this must be analysed mediator by mediator. Finally, it would be interesting to combine the approaches of mediation and interaction in order to explore if we “captured” or not the same mechanisms with the mediation approach (strategy b) and the interaction approach (strategy c). These could be explored with a combination of the 4-way decomposition and a measure of interaction between two independent exposures [66]. An important implication of these results concerns the methods for calculating allostatic load. The methodological questions around the most appropriate way to measure allostatic load from biomarkers are diverse and still debated [4, 67]. The original and most widely used method is based on counting the number of biomarkers where the individual is "at risk", i.e., at one extreme of the distribution (often the highest or lowest quartile). When such data-driven thresholds are used to defined the “at-risk” groups, the question of whether these thresholds should be set according to sex or not is important, and unresolved [4]. Some authors justify a sex-specific threshold approach because they attribute differences in the average level of biomarkers to sexual dimorphisms which they want to control [2–4]. Many of the biomarkers used to measure allostatic load may indeed have physiologically different distributions between the sexes due to difference in sex hormones (e.g. oestrogen and testosterone) [16–18]. But socio-cultural factors and behaviours also explain some of the differences in distribution [16, 19], as the results of this study showed. Only few studies have looked at this issue and, if the method based on sex-specific thresholds seems to be more successful in terms of prediction [2, 4], the question of the methodological and conceptual relevance of this approach has not been asked. *In* general, dichotomisation should be avoid, mainly as it lead to an information loss [68]. But, beyond this point, a differentiated dichotomization is equivalent to controlling not only for the effect of sex but also to the effect of gender (as gender might explain some of the sex difference), on these variables. This can lead to gender bias, especially if the studied phenomena are social and gendered. We would, therefore, recommend using non-dichotomizing methods, such as the sum of z-scores, to calculate the allostatic load; and, at least, not to calculate a sex-specific score, but rather to adjust for sex in the model, including an interaction term between sex and allostatic load, to control for the effect of sexual dimorphism on biomarkers. On the other hand, some sex differences are large, larger than a strong social determinant such as early-life deprivation. It was the case here for systolic blood pressure for example. However, clinical norms for this biomarker are not sex-differentiated, which may lead to underdiagnosis and therefore undertreatment in women, in this example. ## Conclusion The biological differences between men and women seem not to be purely explained by sexed mechanisms. The exploration of gender mechanisms opens new perspectives, in terms of methodology, understanding and potential applications. ## Supplementary Information Additional file 1. Additional tables. ## References 1. 1.Canadian Institutes of Health Research. Can Inst Health Res n.d. https://www.cihr-irsc-igh-isfh.ca/. Accessed March 1, 2022. 2. Juster R-P, de Torre MB, Kerr P, Kheloui S, Rossi M, Bourdon O. **Sex Differences and gender diversity in stress responses and allostatic load among workers and LGBT people**. *Curr Psychiatry Rep* (2019.0) **21** 110. DOI: 10.1007/s11920-019-1104-2 3. D’Amico D, Amestoy ME, Fiocco AJ. **The association between allostatic load and cognitive function: a systematic and meta-analytic review**. *Psychoneuroendocrinology* (2020.0) **121** 104849. DOI: 10.1016/j.psyneuen.2020.104849 4. McLoughlin S, Kenny RA, McCrory C. **Does the choice of Allostatic Load scoring algorithm matter for predicting age-related health outcomes?**. *Psychoneuroendocrinology* (2020.0) **120** 104789. DOI: 10.1016/j.psyneuen.2020.104789 5. Hammarström A, Johansson K, Annandale E, Ahlgren C, Aléx L, Christianson M. **Central gender theoretical concepts in health research: the state of the art**. *J Epidemiol Community Health* (2014.0) **68** 185-190. DOI: 10.1136/jech-2013-202572 6. 6.Delphy C. L’ Ennemi principal. Editions Syllepse; 1998. 7. Johnson JL, Greaves L, Repta R. **Better science with sex and gender: Facilitating the use of a sex and gender-based analysis in health research**. *Int J Equity Health* (2009.0) **8** 14. DOI: 10.1186/1475-9276-8-14 8. Bem SL. **The measurement of psychological androgyny**. *J Consult Clin Psychol* (1974.0) **42** 155-162. DOI: 10.1037/h0036215 9. Colineaux H, Soulier A, Lepage B, Kelly-Irving M. **Considering sex and gender in Epidemiology: a challenge beyond terminology From conceptual analysis to methodological strategies**. *Biol Sex Differ* (2022.0) **13** 23. DOI: 10.1186/s13293-022-00430-6 10. Kelly-Irving M, Delpierre C. **Framework for understanding health inequalities over the life course: the embodiment dynamic and biological mechanisms of exogenous and endogenous origin**. *J Epidemiol Community Health* (2021.0) **75** 1181-1186. DOI: 10.1136/jech-2021-216430 11. McEwen BS, Stellar E. **Stress and the individual. Mechanisms leading to disease**. *Arch Intern Med* (1993.0) **153** 2093-2101. DOI: 10.1001/archinte.1993.00410180039004 12. 12.Kelly-Irving K. Allostatic load: how stress in childhood affects life-course health outcomes. Health Found Work Pap 2019. 13. Seeman TE, McEwen BS. **Impact of social environment characteristics on neuroendocrine regulation**. *Psychosom Med* (1996.0) **58** 459-471. DOI: 10.1097/00006842-199609000-00008 14. Seeman TE, McEwen BS, Singer BH, Albert MS, Rowe JW. **Increase in urinary cortisol excretion and memory declines: MacArthur studies of successful aging**. *J Clin Endocrinol Metab* (1997.0) **82** 2458-2465. DOI: 10.1210/jcem.82.8.4173 15. 15.Mcwen BS, Seeman TE. Protective and damaging effects of mediators of stress. Elaborating and testing the concepts of allostasis and allostatic load. Ann N Y Acad Sci 1999;896:30–47. 10.1111/j.1749-6632.1999.tb08103.x. 16. Juster R-P, Bizik G, Picard M, Arsenault-Lapierre G, Sindi S, Trepanier L. **A transdisciplinary perspective of chronic stress in relation to psychopathology throughout life span development**. *Dev Psychopathol* (2011.0) **23** 725-776. DOI: 10.1017/S0954579411000289 17. Juster R-P, Almeida D, Cardoso C, Raymond C, Johnson P, Pfaus J. **Gonads and strife: sex hormones vary according to sexual orientation for women and stress indices for both sexes**. *Psychoneuroendocrinology* (2016.0). DOI: 10.1016/j.psyneuen.2016.06.011 18. Juster R-P, Raymond C, Desrochers AB, Bourdon O, Durand N, Wan N. **Sex hormones adjust “sex-specific” reactive and diurnal cortisol profiles**. *Psychoneuroendocrinology* (2016.0) **63** 282-290. DOI: 10.1016/j.psyneuen.2015.10.012 19. Lundberg U. **Stress hormones in health and illness: The roles of work and gender**. *Psychoneuroendocrinology* (2005.0) **30** 1017-1021. DOI: 10.1016/j.psyneuen.2005.03.014 20. Power C, Elliott J. **Cohort profile: 1958 British birth cohort (National Child Development Study)**. *Int J Epidemiol* (2006.0) **35** 34-41. DOI: 10.1093/ije/dyi183 21. Atherton K, Fuller E, Shepherd P, Strachan DP, Power C. **Loss and representativeness in a biomedical survey at age 45 years: 1958 British birth cohort**. *J Epidemiol Community Health* (2008.0) **62** 216-223. DOI: 10.1136/jech.2006.058966 22. Munafò MR, Tilling K, Taylor AE, Evans DM, Davey SG. **Collider scope: when selection bias can substantially influence observed associations**. *Int J Epidemiol* (2018.0) **47** 226-235. DOI: 10.1093/ije/dyx206 23. Shattuck-Heidorn H, Richardson SS. **Sex/Gender and the Biosocial Turn**. *Sch Fem Online* (2019.0) **15** 1 24. Fuller E, Power C, Shepherd P, Strachan D. **Technical report on the National Child Development Study biomedical survey 2002–2004**. *Natl Cent Soc Res* (2006.0) **1** 1 25. Johnson SC, Cavallaro FL, Leon DA. **A systematic review of allostatic load in relation to socioeconomic position: poor fidelity and major inconsistencies in biomarkers employed**. *Soc Sci Med* (2017.0) **192** 66-73. DOI: 10.1016/j.socscimed.2017.09.025 26. La BP. *Distinction: Critique sociale du jugement* (1979.0) 27. 27.Fassin D. Inégalité, genre et santé, entre l’universel et le culturel. In: Burnier M-J, Preiswerk Y, editors. Tant Qu’on Santé Déterm. Socio-Économiques Cult. Santé Dans Relat. Soc. Entre Femmes Hommes, Genève: Graduate Institute Publications; 2016, p. 119–30. 28. Lippa RA, Martin LR, Friedman HS. **Gender-Related Individual Differences and Mortality in the Terman Longitudinal Study: Is Masculinity Hazardous to Your Health?**. *Pers Soc Psychol Bull* (2000.0) **26** 1560-1570. DOI: 10.1177/01461672002612010 29. Smith PM, Koehoorn M. **Measuring gender when you don’t have a gender measure: constructing a gender index using survey data**. *Int J Equity Health* (2016.0) **15** 82. DOI: 10.1186/s12939-016-0370-4 30. Pelletier R, Ditto B, Pilote L. **A composite measure of gender and its association with risk factors in patients with premature acute coronary syndrome**. *Psychosom Med* (2015.0) **77** 517-526. DOI: 10.1097/PSY.0000000000000186 31. van Buuren S. *Flexible Imputation of Missing Data* (2018.0) 32. Daniel RM, De Stavola BL, Cousens SN. **Gformula: estimating causal effects in the presence of time-varying confounding or mediation using the G-computation formula**. *Stata J* (2011.0) **11** 479-517. DOI: 10.1177/1536867X1201100401 33. Zhang YT, Laraia BA, Mujahid MS, Tamayo A, Blanchard SD, Warton EM. **Does food vendor density mediate the association between neighborhood deprivation and BMI? A G-computation mediation analysis**. *Epidemiol Camb Mass* (2015.0) **26** 344-352. DOI: 10.1097/EDE.0000000000000271 34. VanderWeele TJ. **Policy-relevant proportions for direct effects**. *Epidemiol Camb Mass* (2013.0) **24** 175-176. DOI: 10.1097/EDE.0b013e3182781410 35. 35.Pearl J. Direct and Indirect Effects. Probabilistic Causal Inference Works Jud. Pearl, vol. 36. 1st ed., New York, NY, USA: Association for Computing Machinery; 2022, p. 373–92. 36. 36.Chang M, Yesupriya A, Ned RM, Mueller PW, Dowling NF. Genetic variants associated with fasting blood lipids in the U.S. population: Third National Health and Nutrition Examination Survey. BMC Med Genet 2010;11:62. 10.1186/1471-2350-11-62. 37. Xu Y, Cao W, Shen Y, Tang J, Wang Y, Ma X. **The relationship between sex hormones and glycated hemoglobin in a non-diabetic middle-aged and elderly population**. *BMC Endocr Disord* (2022.0) **22** 91. DOI: 10.1186/s12902-022-01002-w 38. Eliasson M, Janlert U, Jansson J-H, Stegmayr B. **Time trends in population cholesterol levels 1986–2004: influence of lipid-lowering drugs, obesity, smoking and educational level. The northern Sweden MONICA study**. *J Intern Med* (2006.0) **260** 551-9. DOI: 10.1111/j.1365-2796.2006.01730.x 39. Wilson PW. **The epidemiology of hypercholesterolemia A global perspective**. *Am J Med* (1989.0) **87** 5S-13S. DOI: 10.1016/s0002-9343(89)80593-5 40. Hegsted DM, McGandy RB, Myers ML, Stare FJ. **Quantitative effects of dietary fat on serum cholesterol in man**. *Am J Clin Nutr* (1965.0) **17** 281-295. DOI: 10.1093/ajcn/17.5.281 41. Pazoki R, Dehghan A, Evangelou E, Warren H, Gao H, Caulfield M. **Genetic predisposition to high blood pressure and lifestyle factors**. *Circulation* (2018.0) **137** 653-661. DOI: 10.1161/CIRCULATIONAHA.117.030898 42. Desai AN. **High blood pressure**. *JAMA* (2020.0) **324** 1254-1255. DOI: 10.1001/jama.2020.11289 43. Forouhi NG, Wareham NJ. **The EPIC-interact study: a study of the interplay between genetic and lifestyle behavioral factors on the risk of type 2 diabetes in european populations**. *Curr Nutr Rep* (2014.0) **3** 355-363. DOI: 10.1007/s13668-014-0098-y 44. Jannasch F, Kröger J, Schulze MB. **Dietary patterns and type 2 diabetes: a systematic literature review and meta-analysis of prospective studies**. *J Nutr* (2017.0) **147** 1174-1182. DOI: 10.3945/jn.116.242552 45. Greenfield JR, Samaras K, Jenkins AB, Kelly PJ, Spector TD, Gallimore JR. **Obesity is an important determinant of baseline serum c-reactive protein concentration in monozygotic twins Independent of Genetic Influences**. *Circulation* (2004.0) **109** 3022-3028. DOI: 10.1161/01.CIR.0000130640.77501.79 46. Libert C, Dejager L, Pinheiro I. **The X chromosome in immune functions: when a chromosome makes the difference**. *Nat Rev Immunol* (2010.0) **10** 594-604. DOI: 10.1038/nri2815 47. 47.Colman RW. Hemostasis and Thrombosis: Basic Principles and Clinical Practice. Lippincott Williams & Wilkins; 2006. 48. Kao PC, Shiesh S-C, Wu T-J. **Serum C-reactive protein as a marker for wellness assessment**. *Ann Clin Lab Sci* (2006.0) **36** 163-169. PMID: 16682512 49. Ridker PM, Cannon CP, Morrow D, Rifai N, Rose LM, McCabe CH. **C-reactive protein levels and outcomes after statin therapy**. *N Engl J Med* (2005.0) **352** 20-28. DOI: 10.1056/NEJMoa042378 50. Kirschbaum C, Klauer T, Filipp S-H, Hellhammer DH. **Sex-specific effects of social support on cortisol and subjective responses to acute psychological stress**. *Psychosom Med* (1995.0) **57** 23-31. DOI: 10.1097/00006842-199501000-00004 51. Dickerson SS, Kemeny ME. **Acute stressors and cortisol responses: a theoretical integration and synthesis of laboratory research**. *Psychol Bull* (2004.0) **130** 355. DOI: 10.1037/0033-2909.130.3.355 52. Stroud LR, Salovey P, Epel ES. **Sex differences in stress responses: social rejection versus achievement stress**. *Biol Psychiatry* (2002.0) **52** 318-327. DOI: 10.1016/S0006-3223(02)01333-1 53. House JS, Robbins C, Metzner HL. **The association of social relationships and activities with mortality: prospective evidence from the Tecumseh Community Health Study**. *Am J Epidemiol* (1982.0) **116** 123-140. DOI: 10.1093/oxfordjournals.aje.a113387 54. Kaplan GA, Salonen JT, Cohen RD, Brand RJ, Leonard SS, Puska P. **Social connections and mortality from all causes and from cardiovascular disease: prospective evidence from eastern finland**. *Am J Epidemiol* (1988.0) **128** 370-380. DOI: 10.1093/oxfordjournals.aje.a114977 55. Seeman TE, Berkman LF, Blazer D, Rowe JW. **Social ties and support and neuroendocrine function: the macarthur studies of successful aging2**. *Ann Behav Med* (1994.0) **16** 95-106. DOI: 10.1093/abm/16.2.95 56. Bozovic D, Racic M, Ivkovic N. **Salivary cortisol levels as a biological marker of stress reaction**. *Med Arch Sarajevo Bosnia Herzeg* (2013.0) **67** 374-377. DOI: 10.5455/medarh.2013.67.374-377 57. Dowd JB, Simanek AM, Aiello AE. **Socio-economic status, cortisol and allostatic load: a review of the literature**. *Int J Epidemiol* (2009.0) **38** 1297-1309. DOI: 10.1093/ije/dyp277 58. Wandschneider L, Sauzet O, Razum O, Miani C. **Exploring gendered practices by social position in epidemiology: the gender score applied to Germany**. *Eur J Public Health* (2021.0) **31** 164. DOI: 10.1093/eurpub/ckab164.204 59. Ballering AV, Bonvanie IJ, Olde Hartman TC, Monden R, Rosmalen JGM. **Gender and sex independently associate with common somatic symptoms and lifetime prevalence of chronic disease**. *Soc Sci Med* (2020.0) **253** 112968. DOI: 10.1016/j.socscimed.2020.112968 60. Nielsen MW, Stefanick ML, Peragine D, Neilands TB, Ioannidis JPA, Pilote L. **Gender-related variables for health research**. *Biol Sex Differ* (2021.0) **12** 23. DOI: 10.1186/s13293-021-00366-3 61. 61.Crenshaw K. Demarginalizing the intersection of race and sex: A Black feminist critique of antidiscrimination doctrine, feminist theory and antiracist politics.(pp. 383–395) In D. Kelly Weisbert (Ed.) Feminist legal theory: Foundations 1989. 62. Else-Quest NM, Hyde JS. **Intersectionality in quantitative psychological research: I. Theoretical and epistemological issues**. *Psychol Women Q* (2016.0) **40** 155-70. DOI: 10.1177/0361684316629797 63. Mahendran M, Lizotte D, Bauer GR. **Describing intersectional health outcomes: an evaluation of data analysis methods**. *Epidemiol Camb Mass* (2022.0) **33** 395-405. DOI: 10.1097/EDE.0000000000001466 64. Gkiouleka A, Huijts T. **Intersectional migration-related health inequalities in Europe: exploring the role of migrant generation, occupational status & gender**. *Soc Sci Med* (2020.0) **267** 113218. DOI: 10.1016/j.socscimed.2020.113218 65. VanderWeele TJ. **A unification of mediation and interaction: a 4-way decomposition**. *Epidemiol Camb Mass* (2014.0) **25** 749-761. DOI: 10.1097/EDE.0000000000000121 66. Colineaux H, Lamy S, Kelly-Irving M, Lepage B. **Décomposition d’un effet d’interaction entre deux expositions indépendantes**. *Rev DÉpidémiologie Santé Publique* (2020.0) **68** S99. DOI: 10.1016/j.respe.2020.03.004 67. Lia Y, Rosemberg M-AS, Dalton VK, Lee SJ, Seng JS. **Exploring the optimal allostatic load scoring method in women of reproductive age**. *J Adv Nurs* (2019.0) **75** 2548-58. DOI: 10.1111/jan.14014 68. Altman DG, Royston P. **The cost of dichotomising continuous variables**. *BMJ* (2006.0) **332** 1080. DOI: 10.1136/bmj.332.7549.1080
--- title: 'Eubiotic effect of rifaximin is associated with decreasing abdominal pain in symptomatic uncomplicated diverticular disease: results from an observational cohort study' authors: - Vladimir Ivashkin - Oleg Shifrin - Roman Maslennikov - Elena Poluektova - Alexander Korolev - Anna Kudryavtseva - George Krasnov - Nona Benuni - Giovanni Barbara journal: BMC Gastroenterology year: 2023 pmcid: PMC10037807 doi: 10.1186/s12876-023-02690-x license: CC BY 4.0 --- # Eubiotic effect of rifaximin is associated with decreasing abdominal pain in symptomatic uncomplicated diverticular disease: results from an observational cohort study ## Abstract ### Background Rifaximin effectively treats symptomatic uncomplicated diverticular disease (SUDD) and has shown eubiotic potential (i.e., an increase in resident microbial elements with potential beneficial effects) in other diseases. This study investigated changes in the fecal microbiome of patients with SUDD after repeated monthly treatment with rifaximin and the association of these changes with the severity of abdominal pain. ### Methods This was a single-center, prospective, observational, uncontrolled cohort study. Patients received rifaximin 400 mg twice a day for 7 days per month for 6 months. Abdominal pain (assessed on a 4-point scale from 0 [no pain] to 3 [severe pain]) and fecal microbiome (assessed using 16 S rRNA gene sequencing) were assessed at inclusion (baseline) and 3 and 6 months. The Spearman’s rank test analyzed the relationship between changes in the gut microbiome and the severity of abdominal pain. A p-value ≤ 0.05 was considered statistically significant. ### Results Of the 23 patients enrolled, 12 patients completed the study and were included in the analysis. Baseline abdominal pain levels decreased significantly after 3 ($$p \leq 0.036$$) and 6 ($$p \leq 0.008$$) months of treatment with rifaximin. The abundance of Akkermansia in the fecal microbiome was significantly higher at 3 ($$p \leq 0.017$$) and 6 ($$p \leq 0.015$$) months versus baseline. The abundance of Ruminococcaceae ($$p \leq 0.034$$), Veillonellaceae ($$p \leq 0.028$$), and Dialister ($$p \leq 0.036$$) were significantly increased at 6 months versus baseline, whereas Anaerostipes ($$p \leq 0.049$$) was significantly decreased. The severity of abdominal pain was negatively correlated with the abundance of Akkermansia (r=-0.482; $$p \leq 0.003$$) and Ruminococcaceae (r=-0.371; $$p \leq 0.026$$) but not with Veillonellaceae, Dialister, or Anaerostipes. After 3 months of rifaximin, abdominal pain was significantly less in patients with Akkermansia in their fecal microbiome than in patients without Akkermansia ($$p \leq 0.022$$). ### Conclusion The eubiotic effect of rifaximin was associated with decreased abdominal pain in patients with SUDD. ## Introduction Diverticula are small sac-like protrusions that form in the large intestine wall and represent the most frequent anatomical alteration of the colon. The presence of diverticula in the intestine, defined as diverticulosis, may be asymptomatic or proceed as symptomatic uncomplicated or complicated diverticular disease [1]. The main symptoms of symptomatic uncomplicated diverticular disease (SUDD) are episodes of abdominal pain without evidence of inflammation of diverticula (i.e., without diverticulitis) [2]. The pathogenesis of abdominal pain in SUDD is poorly understood [1, 2]. However, studies suggest that the bacteria that inhabit the colon (gut microbiota) may play a role in its pathogenesis [3, 4]. For example, compared with patients with asymptomatic diverticulosis, patients with SUDD had a decreased abundance of *Clostridium cluster* IX, Fusobacterium, and Lactobacillaceae [5]. Compared with healthy controls, patients with SUDD had a decreased abundance of Porphyromonadaceae and *Bacteroides fragilis* in their fecal microbiome [6], whereas an increased abundance of *Akkermansia muciniphila* was identified in the fecal samples of SUDD patients in a separate study [7]. Similarly, the abundance of Enterobacteriaceae was increased in colonic mucosa biopsies of patients with SUDD compared with patients without diverticular disease [8]. Several treatment options for SUDD have been proposed, including the use of the non-absorbable antibiotic rifaximin, which decreases both the severity of SUDD symptoms and the incidence of complications of diverticular disease [9–12]. The eubiotic (i.e., improving the composition of the gut microbiota) [13] effect of rifaximin has been reported in experimental studies in rats [14, 15] and in patients with Crohn’s disease [16], cirrhosis [17], and non-constipated irritable bowel syndrome [18]. Very few studies have investigated changes in the gut microbiome after treatment of diverticular disease with rifaximin. However, two of these studies assessed only 4–7 patients with SUDD alongside patients with other intestinal diseases (i.e., ulcerative colitis, Crohn’s disease, or irritable bowel syndrome) [19, 20], while a third study treated SUDD patients with other therapeutic approaches (i.e., fiber supplementation, mesalazine, probiotic mixture VivoMixx(R)) [21]. More recently, a larger study of 43 patients with SUDD identified significant variation in the composition of the gut microbiota in stool samples taken before versus after treatment with rifaximin [22]. However, these patients received only 7 days of treatment with rifaximin. Consequently, studies that evaluate the long-term effect of rifaximin on the gut microbiota in patients with diverticular disease are lacking. Our study aimed to investigate changes in the fecal microbiome composition in patients with SUDD after repeated monthly treatment with rifaximin and the association of these changes with the severity of abdominal pain. ## Ethics approval and consent to participate This single-center, prospective, observational, uncontrolled cohort study was conducted according to the Declaration of Helsinki and approved by the Independent Interdisciplinary Ethics Committee (Resolution No. 13 dated 21.07.2017). All participants gave written informed consent. ## Patients This study enrolled consecutive patients with exacerbation of SUDD who were aged > 18 years and attended the Clinic for Internal Medicine, Gastroenterology, and Hepatology of Sechenov University. The exacerbation of SUDD was defined as the presence of abdominal pain recorded in the lower left quadrant for > 24 h in patients with diverticulosis and absence of any complications (stenosis, abscesses, fistulas) [7]. Enrolled patients also had to have received dietary fiber for at least 6 months prior to study entry to prevent constipation as a risk factor for the development of diverticulitis and other complications of diverticular disease. The exclusion criteria were as follows: contraindications to the use of rifaximin (history of drug allergy to rifaximin), the use of rifaximin during the previous 6 months, cancer, acute complications of diverticular disease (development of acute diverticulitis and/or intestinal bleeding) during the previous 6 months, planned surgery, participation in other clinical trials, pregnancy, and breastfeeding. In addition, patients were excluded from the study analysis if they refused to continue, violated the rifaximin intake regimen, required additional treatment for SUDD, or used other antibacterial drugs. ## Intervention All patients received rifaximin (Alpha Normix®) at a dose of 400 mg twice a day for 7 days per month for 6 months. ## Outcomes Abdominal pain and the fecal microbiome were assessed at study inclusion (baseline) and after 3 and 6 months. Abdominal pain was assessed on a 4-point scale: 0 = no pain; 1 = mild pain (easily tolerated); 2 = moderate pain affecting daily activities; 3 = severe pain that interferes with daily activities. The maximum score for the 2 weeks before the assessment was also considered. The fecal microbiome was analyzed using 16S rRNA gene sequencing according to the method described by Maslennikov and colleagues [23]. Briefly, stool samples were collected by each patient in a sterile disposable container on the morning of admission and immediately frozen at -80°C [24]. DNA was isolated from the stool sample, and two rounds of PCR amplification were used to prepare libraries for sequencing. In the first round, specific primers (16S-F, TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG; 16S-R, GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC) were used to amplify the v3-v4 region of the 16S ribosomal RNA gene. During the second round of PCR, specific adapters were attached to the PCR product to enable multiplex sequencing. After measuring their concentration and quality, the prepared libraries were mixed in equal proportions, and pair-end readings of 300 + 300 nucleotides were obtained on a MiSeq (Illumina) device. Reads were trimmed from the 3’-tail with Trimmomatic (Illumina) and merged into a single amplicon with the MeFiT tool [25, 26]. Amplicon sequences were then classified using the Ribosomal Database Project (RDP) classifier and RDP database [27]. ## Statistical analysis Data are reported as median [interquartile range (IQR)]. The Mann-Whitney test was used to assess differences between continuous variables. For differences between categorical variables, the Fisher’s exact test was used. Variations in the abundance of gut microbiome taxa were analyzed using the Wilcoxon test. The Spearman’s rank test was used to assess correlations between the variables computed. A p-value ≤ 0.05 was considered statistically significant. Statistical analysis was performed using STATISTICA 10 (StatSoft Inc., USA). ## Results Thirty patients were assessed for eligibility. Twenty-three patients were included in the study, and 12 patients completed the study (Fig. 1). Eleven patients were lost to follow-up; 6 ($55\%$) patients stopped taking rifaximin due to persistent improvement, 2 patients refused to participate further in the study or took systemic antibiotics, and 1 patient required additional drugs due to the persistence of abdominal pain. Fig. 1CONSORT 2010 Flow Diagram The median [IQR] age of patients who completed the study was 68 [55–71] years, body mass index was 26.4 [24.8–27.6] kg/m2, and $50\%$ of patients were male. All patients were Caucasian. Complete blood counts and main biochemical blood biomarkers were normal in all patients. None of the patients had taken probiotics or antibiotics in the 6 weeks prior to study entry. Four patients received therapy for arterial hypertension. We found no evidence in the literature that these drugs have a significant effect on abdominal pain or gut microbiota. The remaining patients reported no concomitant medication. Compared with baseline levels, significant improvement in abdominal pain was identified after 3 months of rifaximin ($$p \leq 0.036$$), with improvement further pronounced at 6 months ($$p \leq 0.008$$) (Fig. 2). None of the patients developed complications of diverticular disease or side effects from rifaximin. Fig. 2Distribution of patients according to the severity of abdominal pain (3 - severe pain; 2 – moderate pain; 1- mild pain; 0 - no pain) at inclusion, after 3 and 6 courses of rifaximin Analyses of the fecal microbiome (Table 1) identified significant increases in the abundance of the phylum Verrucomicrobia and the genus Akkermansia (phylum Verrucomicrobia) at both 3 months ($$p \leq 0.018$$ and $$p \leq 0.017$$, respectively) and 6 months (both $$p \leq 0.015$$) compared with baseline levels. Significant increases were also observed at 6 months in the abundance of Ruminococcaceae ($$p \leq 0.034$$), Veillonellaceae ($$p \leq 0.028$$), and Dialister ($$p \leq 0.036$$) (family Veillonellaceae). In contrast, the abundance of Anaerostipes ($$p \leq 0.049$$) (family Lachnospiraceae) decreased significantly after 6 months of rifaximin compared with baseline levels. No significant differences in abundance were identified for the remaining taxa investigated. Table 1Changes in the fecal microbiome in patients with symptomatic uncomplicated diverticular disease after 3 and 6 months of rifaximinTaxonAt inclusionAfter 3 coursesAfter 6 coursesp*p**p***Main phylaFirmicutes68.6 [59.4–74.3]70.4 [60.6–76.7]72.7 [64.8–81.8]0.8140.6380.814Bacteroidetes28.3 [23.9–38.2]27.3 [16.2–38.1]25.2 [15.6–33.1]0.8140.5830.694Proteobacteria1.6 [0.8–3.9]1.3 [0.7–2.5]1.4 [1.2–2.4]0.2720.4801.000Actinobacteria0.2 [0.1–0.8]0.1 [0.0–0.7]0.1 [0.0–1.0]0.3460.6950.433Verrucomicrobia0.0 [0.0–0.0]0.1 [0.0–1.0]0.3 [0.0–0.7] 0.018 0.799 0.015 Main classes Clostridia 65.9 [55.1–72.5]66.9 [55.3–74.4]69.2 [59.8–76.7]0.8750.4330.875 Bacilli 0.2 [0.1–0.7]0.5 [0.1–1.6]0.1 [0.0–1.1]0.1580.7210.374 Negativicutes 1.2 [0.5–1.4]1.5 [1.1–1.7]1.5 [0.9–2.0]0.1360.9370.272 Main families Lachnospiraceae 26.3 [18.6–32.4]21.0 [14.1–30.3]21.0 [13.2–30.0]0.1170.6380.158 Ruminococcaceae 33.5 [27.3–38.0]37.0 [30.1–41.1]41.7 [36.5–47.4]0.2390.427 0.034 Bacteroidaceae 15.6 [10.8–27.4]13.6 [6.2–29.5]20.1 [7.6–26.8]0.6951.0000.937 Prevotellaceae 3.81 [0.37–12.44]2.73 [0.91–8.05]2.66 [1.07–4.03]0.7220.5940.477 Porphyromonadaceae 1.64 [0.82–2.66]1.35 [0.96–1.73]1.78 [1.33–3.44]0.5300.2090.308 Rikenellaceae 0.97 [0.22–2.69]0.49 [0.07–1.61]0.61 [0.16–1.78]0.1820.8140.308 Peptostreptococcaceae 0.63 [0.19–1.32]0.06 [0.00–0.55]0.19 [0.13–0.62]0.2410.4240.308 Sutterellaceae 0.59 [0.05–0.66]0.42 [0.21–0.64]0.20 [0.09–0.56]0.4800.5830.209 Streptococcaceae 0.24 [0.06–0.66]0.23 [0.05–1.10]0.04 [0.00–0.97]0.4800.7210.514 Veillonellaceae 0.17 [0.00–0.45]0.08 [0.00–1.09]0.84 [0.00–1.83]0.1390.799 0.028 Coriobacteriaceae 0.09 [0.03–0.18]0.03 [0.00–0.10]0.09 [0.01–0.16]0.2130.2860.582 Bifidobacteriaceae 0.08 [0.00–0.51]0.00 [0.00–0.65]0.03 [0.00–0.84]0.5080.6780.721 Enterobacteriaceae 0.05 [0.00–0.72]0.25 [0.12–0.69]0.39 [0.00–1.46]0.7990.3880.575 Genera with a significant change in abundance Dialister 0.00 [0.00–0.25]0.00[0.00–1.51]0.49 [0.00–1.46]0.0930.779 0.036 Akkermansia 0.00 [0.00–0.00]0.11[0.00–1.02]0.31 [0.01–0.70] 0.017 0.721 0.015 Anaerostipes 0.42 [0.18–1.06]0.28[0.09–1.03]0.17 [0.11–0.37]0.9290.272 0.049 Data are presented as median percentage [interquartile range]. * p-value between 3 months and baseline; **p-value between 6 months and 3 months; ***p-value between 6 months and baseline. A p-value ≤ 0.05 was considered statistically significant (highlighted in bold) Akkermansia was detected in the fecal microbiome in 2 of 12 ($16.7\%$) patients at inclusion. This increased to 7 of 12 ($58.3\%$; $$p \leq 0.045$$) patients at 3 months, and 9 of 12 ($75.0\%$; $$p \leq 0.006$$) patients after 6 months of rifaximin. The patients were divided into subgroups based on the presence or absence of Akkermansia in the fecal microbiome at 3 months as follows; patients with Akkermansia (Akkermansia[+]; $$n = 7$$) versus patients without Akkermansia (Akkermansia[-]; $$n = 5$$). The severity of abdominal pain was significantly less in the Akkermansia[+] group than in the Akkermansia[-] group after 3 months (median 1.0 [0.0–1.0] versus 2.0 [1.0–2.0] points, respectively; $$p \leq 0.022$$) and 6 months (median 0.0 [0.0–1.0] versus 1.0 [1.0–1.0] points, respectively; $$p \leq 0.023$$) of rifaximin, whereas no significant between-group differences were identified at baseline (median 2.0 [1.0–2.0] versus 2.0 [1.0–2.0] points, respectively; $$p \leq 0.876$$). Significant decreases in abdominal pain were observed in the Akkermansia[+] group at 3 and 6 months compared with baseline (both $$p \leq 0.028$$) (Fig. 3a). The severity of abdominal pain also decreased, albeit without significance, in the Akkermansia[-] group after 6 months of rifaximin compared with 3 months (Fig. 3b), however, the decrease in abdominal pain was only observed in patients with detectable Akkermansia in their fecal microbiome. The severity of abdominal pain was negatively correlated with the abundance of Akkermansia (r=-0.482; $$p \leq 0.003$$), Verrucomicrobia (r=-0.440; $$p \leq 0.007$$), and Ruminococcaceae (r=-0.371; $$p \leq 0.026$$) in the fecal microbiome. No significant correlation was identified between the severity of abdominal pain and the abundance of Veillonellaceae ($$p \leq 0.486$$), Dialister ($$p \leq 0.101$$), or Anaerostipes ($$p \leq 0.867$$). Fig. 3Distribution of patients according to the severity of abdominal pain (3 - severe pain; 2 – moderate pain; 1- mild pain; 0 - no pain) at inclusion, after 3 and 6 courses of rifaximin in the Akkermansia[+] (a) and Akkermansia[-] (b) groups ## Discussion To the best of our knowledge, this study is the first to evaluate the long-term (i.e., 6 months) effect of rifaximin on the gut microbiota in patients with SUDD. We show here that rifaximin significantly reduced the severity of abdominal pain, which is consistent with previous studies [9–12]. Whereas treatment with rifaximin over the 6 months was not accompanied by significant changes in the abundance of most major taxa of the fecal microbiome, increases in the abundance of Akkermansia, Verrucomicrobia, and Ruminococcaceae were observed and inversely correlated with the severity of abdominal pain. Changes in the abundance of Veillonellaceae, Dialister, and Anaerostipes were also observed after 6 months of rifaximin, however, they were not correlated with abdominal pain severity. Thus, it is likely that these bacteria are not involved in the development of abdominal pain in SUDD. In our study, Akkermansia were identified in the fecal microbiome of 2 patients ($16.7\%$) at inclusion and in 9 patients ($75.0\%$) after 6 months of rifaximin, whereas these bacteria were detected in $90.5\%$ of healthy individuals in a separate study ($p \leq 0.001$) (unpublished data from [23]). However, a significantly higher abundance of A. muciniphila has previously been reported in fecal samples of patients with SUDD compared with healthy controls [7]. It should be noted, however, that both our study and the study by Tursi and colleagues [7] excluded patients with a recent history of acute diverticulitis. It is possible that, in patients with diverticulosis, the number of Akkermansia may increase as a compensatory reaction. Therefore, patients with abundant Akkermansia develop asymptomatic diverticulosis or SUDD, while patients who do not have high enough numbers of Akkermansia for this compensatory reaction develop acute diverticulitis. It should also be noted that, in the study by Tursi and colleagues [7], patients with SUDD had a lower abundance of A. muciniphila in the gut microbiota than patients with asymptomatic diverticulosis, however, this difference did not reach the limits of significance, which may have been due to the small patient population (15 and 13 patients, respectively). New larger studies should be performed to resolve this problem. Akkermansia is the main representative of the *Verrucomicrobia phylum* in the gut microbiome. This bacterium has several beneficial properties, including an anti-inflammatory effect [28–32]. Specifically, the presence of Akkermansia increases the thickness of the mucin layer and improves the intestinal epithelial barrier, preventing the translocation of harmful bacteria and their components into the intestinal wall [31]. This bacterial translocation results in low-level inflammation, which is believed to play an important role in the pathogenesis of abdominal pain in SUDD [1]. Moreover, the intensity of infiltration of the mucous membrane of the diverticula by inflammatory cells inversely correlates with the abundance of Akkermansia in the mucosal microbiome [5]. The positive effect of Akkermansia on the epithelial barrier and mucous layer is believed to be because these bacteria degrade mucin to molecules that stimulate its formation by feedback and are used by bacteria from the family Ruminococcaceae that form butyrate [29, 31], which is known to strengthen the intestinal barrier [33, 34]. In our study, the abundance of Akkermansia and Ruminococcaceae increased significantly after treatment with rifaximin. However, while the increase in the abundance of Akkermansia was significant at both 3 and 6 months, the increased abundance of Ruminococcaceae was only significant after 6 months of treatment with rifaximin. This result supports the hypothesis of the synergistic effect of these two groups of bacteria on decreasing intestinal permeability, bacterial translocation, low-level inflammation, and abdominal pain associated with patients with SUDD. Although rifaximin has been reported to increase the abundance of bacteria under the Ruminococcaceae family, there have been no published data to show that its use increases the abundance of Akkermansia [13]. In a previous study [22], rifaximin significantly altered the relative abundance of specific bacteria in patients with SUDD, with a significantly greater abundance of Bacteroidaceae, Citrobacter, and Coprococcus, and a deficiency of Mogibacteriaceae, Christensenellaceae, Dehalobacteriaceae, Pasteurellaceae, Anaerotruncus, Blautia, Eggerthella lenta, Dehalobacterium, SMB53, and *Haemophilus parainfluenzae* (p-adj < 0.05) reported. However, as patients received only 7 days of treatment with rifaximin, these results must be viewed with caution as they may not represent the long-term effect of rifaximin on the gut microbiota in patients with SUDD. Two small studies investigated the difference in the gut microbiome between patients with asymptomatic diverticulosis and SUDD [5, 7]. Although neither study reported a significant between-group difference in the abundance of Akkermansia in the gut microbiome, counts of A. muciniphila species were numerically lower in patients with SUDD than in those with asymptomatic diverticulosis (-3.56 ± 1.27 versus − 3.41 ± 1.13, respectively) [7]. However, larger studies are required to confirm the hypothesis that the decreased abundance of Akkermansia in patients with diverticulosis is associated with their transition from asymptomatic to symptomatic. Furthermore, a cohort study of patients with asymptomatic diverticulosis and periodic analysis of their gut microbiota may identify predictors of SUDD. All patients in our study consumed dietary fiber to prevent constipation, a risk factor for complications of diverticular disease. However, since we selected patients who had consumed dietary fiber for a minimum of 6 months before enrollment, this is unlikely to have influenced our results. In addition, we did not evaluate the severity of stool disturbances or bloating in our patients, since these may depend on dietary fiber intake. Several limitations of the present study must be acknowledged. Firstly, the number of participants was low, and a substantial proportion of patients were lost at follow-up. Nonetheless, our preliminary results are promising and may support the design of larger controlled studies. The small number of participants can also be explained by self-termination of rifaximin due to persistent improvement or other divergences that led to the exclusion of these patients from the study analysis. Another significant limitation of our study is the lack of a placebo control arm essential to demonstrate an unambiguous symptomatic benefit of rifaximin. Large randomized controlled trials are therefore required to support our findings. ## Conclusion In conclusion, our study is the first to describe changes in the gut microbiome after multiple courses of rifaximin to treat SUDD and the association of these changes with the severity of abdominal pain. We show significant increases in the abundance of the beneficial bacteria, Akkermansia and Ruminococcaceae, which was associated with a decrease in the severity of abdominal pain. Our study also highlights the eubiotic effect of rifaximin from as early as 3 months after starting treatment. ## References 1. Tursi A, Scarpignato C, Strate LL, Lanas A, Kruis W, Lahat A. **Colonic diverticular disease**. *Nat Rev Dis Primers* (2020.0) **6** 20. DOI: 10.1038/s41572-020-0153-5 2. 2.Feuerstein JD, Falchuk KR. Diverticulosis and Diverticulitis. Mayo Clin Proc. 2016;91(8):1094 – 104. 3. Ticinesi A, Nouvenne A, Corrente V, Tana C, Di Mario F, Meschi T. **Diverticular Disease: a gut microbiota perspective**. *J Gastrointestin Liver Dis* (2019.0) **28** 327-37. DOI: 10.15403/jgld-277 4. Tursi A. **Current and evolving concepts on the pathogenesis of Diverticular Disease**. *J Gastrointestin Liver Dis* (2019.0) **28** 225-35. DOI: 10.15403/jgld-184 5. Barbara G, Scaioli E, Barbaro MR, Biagi E, Laghi L, Cremon C. **Gut microbiota, metabolome and immune signatures in patients with uncomplicated diverticular disease**. *Gut* (2017.0) **66** 1252-61. DOI: 10.1136/gutjnl-2016-312377 6. Lopetuso LR, Petito V, Graziani C, Schiavoni E, Paroni Sterbini F, Poscia A. **Gut microbiota in Health, Diverticular Disease, irritable bowel syndrome, and Inflammatory Bowel Diseases: time for microbial marker of gastrointestinal Disorders**. *Dig Dis* (2018.0) **36** 56-65. DOI: 10.1159/000477205 7. Tursi A, Mastromarino P, Capobianco D, Elisei W, Miccheli A, Capuani G. **Assessment of Fecal Microbiota and Fecal Metabolome in Symptomatic Uncomplicated Diverticular Disease of the Colon**. *J Clin Gastroenterol* (2016.0) **50** 9-S12. DOI: 10.1097/MCG.0000000000000626 8. Linninge C, Roth B, Erlanson-Albertsson C, Molin G, Toth E, Ohlsson B. **Abundance of Enterobacteriaceae in the colon mucosa in diverticular disease**. *World J Gastrointest Pathophysiol* (2018.0) **9** 18-27. DOI: 10.4291/wjgp.v9.i1.18 9. Brandimarte G, Bafutto M, Kruis W, Scarpignato C, Mearin F, Barbara G. **Hot topics in Medical Treatment of Diverticular Disease: evidence pro and cons**. *J Gastrointestin Liver Dis* (2019.0) **28** 23-9. PMID: 31930227 10. De Bastiani R, Sanna G, Bertolusso L, Casella G, De Polo M, Zamparella M. **General practitioners’ management of symptomatic uncomplicated diverticular disease of the colon by using rifaximin, a non-adsorbable antibiotic**. *Eur Rev Med Pharmacol Sci* (2021.0) **25** 423-30. PMID: 33506932 11. Di Mario F, Miraglia C, Cambie G, Violi A, Nouvenne A, Franceschi M. **Long-term efficacy of rifaximin to manage the symptomatic uncomplicated diverticular disease of the colon**. *J Investig Med* (2019.0) **67** 767-70. DOI: 10.1136/jim-2018-000901 12. Pietrzak AM, Dziki A, Banasiewicz T, Regula J. **Cyclic rifaximin therapy effectively prevents the recurrence of symptoms after exacerbation of symptomatic uncomplicated diverticular disease: a retrospective study**. *Prz Gastroenterol* (2019.0) **14** 69-78. PMID: 30944680 13. Ponziani FR, Zocco MA, D’Aversa F, Pompili M, Gasbarrini A. **Eubiotic properties of rifaximin: disruption of the traditional concepts in gut microbiota modulation**. *World J Gastroenterol* (2017.0) **23** 4491-9. DOI: 10.3748/wjg.v23.i25.4491 14. Colucci R, Pellegrini C, Fornai M, Tirotta E, Antonioli L, Renzulli C. **Pathophysiology of NSAID-Associated Intestinal Lesions in the rat: luminal Bacteria and mucosal inflammation as targets for Prevention**. *Front Pharmacol* (2018.0) **9** 1340. DOI: 10.3389/fphar.2018.01340 15. Xu D, Gao J, Gillilland M, Wu X, Song I, Kao JY. **Rifaximin alters intestinal bacteria and prevents stress-induced gut inflammation and visceral hyperalgesia in rats**. *Gastroenterology* (2014.0) **146** 484-96. DOI: 10.1053/j.gastro.2013.10.026 16. Maccaferri S, Vitali B, Klinder A, Kolida S, Ndagijimana M, Laghi L. **Rifaximin modulates the colonic microbiota of patients with Crohn’s disease: an in vitro approach using a continuous culture colonic model system**. *J Antimicrob Chemother* (2010.0) **65** 2556-65. DOI: 10.1093/jac/dkq345 17. Bajaj JS, Heuman DM, Sanyal AJ, Hylemon PB, Sterling RK, Stravitz RT. **Modulation of the metabiome by rifaximin in patients with cirrhosis and minimal hepatic encephalopathy**. *PLoS ONE* (2013.0) **8** e60042. DOI: 10.1371/journal.pone.0060042 18. Soldi S, Vasileiadis S, Uggeri F, Campanale M, Morelli L, Fogli MV. **Modulation of the gut microbiota composition by rifaximin in non-constipated irritable bowel syndrome patients: a molecular approach**. *Clin Exp Gastroenterol* (2015.0) **8** 309-25. DOI: 10.2147/CEG.S89999 19. Ponziani FR, Scaldaferri F, De Siena M, Mangiola F, Matteo MV, Pecere S. **Increased Faecalibacterium abundance is associated with clinical improvement in patients receiving rifaximin treatment**. *Benef Microbes* (2020.0) **11** 519-25. DOI: 10.3920/BM2019.0171 20. Ponziani FR, Scaldaferri F, Petito V, Paroni Sterbini F, Pecere S, Lopetuso LR. **The role of antibiotics in gut microbiota modulation: the Eubiotic Effects of Rifaximin**. *Dig Dis* (2016.0) **34** 269-78. DOI: 10.1159/000443361 21. Laghi L, Mastromarino P, Elisei W, Capobianco D, Zhu CL, Picchio M. **Impact of treatments on fecal microbiota and fecal metabolome in symptomatic uncomplicated diverticular disease of the colon: a pilot study**. *J Biol Regul Homeost Agents* (2018.0) **32** 1421-32. PMID: 30574746 22. De Vincentis A, Santonico M, Del Chierico F, Altomare A, Marigliano B, Laudisio A. **Gut microbiota and related electronic Multisensorial System changes in subjects with symptomatic uncomplicated Diverticular Disease Undergoing Rifaximin Therapy**. *Front Med* (2021.0) **8** 655474. DOI: 10.3389/fmed.2021.655474 23. Maslennikov R, Ivashkin V, Efremova I, Alieva A, Kashuh E, Tsvetaeva E. **Gut dysbiosis is associated with poorer long-term prognosis in cirrhosis**. *World J Hepatol* (2021.0) **13** 557-70. DOI: 10.4254/wjh.v13.i5.557 24. Fouhy F, Deane J, Rea MC, O’Sullivan O, Ross RP, O’Callaghan G. **The effects of freezing on faecal microbiota as determined using MiSeq sequencing and culture-based investigations**. *PLoS ONE* (2015.0) **10** e0119355. DOI: 10.1371/journal.pone.0119355 25. Bolger AM, Lohse M, Usadel B. **Trimmomatic: a flexible trimmer for Illumina sequence data**. *Bioinformatics* (2014.0) **30** 2114-20. DOI: 10.1093/bioinformatics/btu170 26. Parikh HI, Koparde VN, Bradley SP, Buck GA, Sheth NU. **MeFiT: merging and filtering tool for illumina paired-end reads for 16S rRNA amplicon sequencing**. *BMC Bioinformatics* (2016.0) **17** 491. DOI: 10.1186/s12859-016-1358-1 27. Wang Q, Garrity GM, Tiedje JM, Cole JR. **Naive bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy**. *Appl Environ Microbiol* (2007.0) **73** 5261-7. DOI: 10.1128/AEM.00062-07 28. de Vos WM. **Microbe Profile: Akkermansia muciniphila: a conserved intestinal symbiont that acts as the gatekeeper of our mucosa**. *Microbiol (Reading)* (2017.0) **163** 646-8. DOI: 10.1099/mic.0.000444 29. Derrien M, Belzer C, de Vos WM. **Akkermansia muciniphila and its role in regulating host functions**. *Microb Pathog* (2017.0) **106** 171-81. DOI: 10.1016/j.micpath.2016.02.005 30. Jayachandran M, Chung SSM, Xu B. **A critical review of the relationship between dietary components, the gut microbe Akkermansia muciniphila, and human health**. *Crit Rev Food Sci Nutr* (2020.0) **60** 2265-76. DOI: 10.1080/10408398.2019.1632789 31. Macchione IG, Lopetuso LR, Ianiro G, Napoli M, Gibiino G, Rizzatti G. **Akkermansia muciniphila: key player in metabolic and gastrointestinal disorders**. *Eur Rev Med Pharmacol Sci* (2019.0) **23** 8075-83. PMID: 31599433 32. Zhang T, Li Q, Cheng L, Buch H, Zhang F. **Akkermansia muciniphila is a promising probiotic**. *Microb Biotechnol* (2019.0) **12** 1109-25. DOI: 10.1111/1751-7915.13410 33. 33.Deleu S, Machiels K, Raes J, Verbeke K, Vermeire S. Short chain fatty acids and its producing organisms: An overlooked therapy for IBD? EBioMedicine. 2021;66:103293. 34. Snelson M, de Pasquale C, Ekinci EI, Coughlan MT. **Gut microbiome, prebiotics, intestinal permeability and diabetes complications**. *Best Pract Res Clin Endocrinol Metab* (2021.0) **35** 101507. DOI: 10.1016/j.beem.2021.101507
--- title: 'Self-reported insomnia symptoms are associated with urinary incontinence among older Indian adults: evidence from the Longitudinal Ageing Study in India (LASI)' authors: - Siqi Leng - Yuming Jin - Michael V. Vitiello - Ye Zhang - Rong Ren - Lin Lu - Jie Shi - Xiangdong Tang journal: BMC Public Health year: 2023 pmcid: PMC10037814 doi: 10.1186/s12889-023-15472-7 license: CC BY 4.0 --- # Self-reported insomnia symptoms are associated with urinary incontinence among older Indian adults: evidence from the Longitudinal Ageing Study in India (LASI) ## Abstract ### Background Insomnia and urinary incontinence (UI) are both diseases burdening older adults. However, the association between them has not been well elucidated. The purpose of this study is to assess the correlation between insomnia symptoms and UI in a large community‐dwelling sample of older Indian adults. ### Methods Data were from Wave 1 (2017–2018) of the Longitudinal Ageing Study of India (LASI). Male and female participants aged ≥ 60 years who provided complete information on insomnia symptoms, UI, stress UI (SUI), and covariates were included. Insomnia symptoms were identified by a report of: trouble falling asleep, waking up at night, or waking too early, ≥ 5 times/week. UI was defined by self-reported diagnosis. SUI was identified by self-report of involuntary urine leakage when sneezing, coughing, laughing, or lifting weights. Multivariable logistic regression analyses evaluated the associations between insomnia symptoms and UI and SUI. Stratified linear regression evaluated for interactions in prespecified subgroups. ### Results Twenty-six thousand eight hundred twenty-one LASI participants met entry criteria. 2979 ($11.11\%$) reported insomnia symptoms, 976 ($3.64\%$) UI, and 2726 ($10.16\%$) SUI. After full adjustment, insomnia symptoms were associated with both UI and SUI among males (OR 1.53; $95\%$CI 1.20–1.96 and OR 1.51; $95\%$CI 1.25–1.83) and females (OR 1.53; $95\%$ CI 1.21–1.92 and OR 1.50; $95\%$ CI 1.31–1.73). A significant interaction effect by age was observed between insomnia symptoms and SUI among both males ($$p \leq 0.048$$) and females ($$p \leq 0.042$$). ### Conclusions Insomnia symptoms were associated with UI and with SUI in both male and female older Indian adults. Further prospective study is called for to better characterize these associations and to explore underlying mechanisms. ## Introduction Increasing life expectancy and falling fertility rates have led to older adults becoming a rapidly expanding portion of the Indian population. More than 316 million adults aged 60 years or older are projected in India by 2050, representing $19.1\%$ of the total population [1]. This increasing proportion of older adults is a global phenomenon and with it comes rises in age-related physiological and pathological changes, and age-related diseases [2]. Insomnia, a patient-reported complaint of difficulties in falling asleep, maintaining sleep or early morning awakening, accompanied by adverse daytime consequences [3], is a very common health concern, afflicting $15\%$ of older adults in India [4, 5]. Growing evidence has documented that insomnia is associated with and a risk factor for diseases such as metabolic syndrome, nocturia, sarcopenia, depression, and neurodegenerative disorders including Alzheimer's disease [6–8]. Urinary incontinence (UI) also increases in prevalence with age [9, 10]. UI is common among older adults and results in physical, psychological, and social adverse consequences, contributing to functional limitations and decreased quality of life. Patients living with UI experience restriction of normal activities of daily living, resulting in physical discomfort, emotional burdens of shame and embarrassment, and social isolation [11]. Moreover, older adults with UI are likely to be functionally dependent, leading to heavy caregiver burden and unmet healthcare need [12]. Given that UI patients experience persistent symptoms for more than 10 years and that the condition can worsen over time, the economic burden of this disease is substantial [13, 14]. The prevalence of UI among Indian women is about $12\%$, while the prevalence of UI among men is less often measured in Indian population [9, 15, 16]. The prevalence of UI in community-dwelling older men in a systematic review is $11\%$-$34\%$, and the prevalence in older women was 1.3–2.0 times that of older men [17]. UI commonly presents as stress UI (SUI), urgency UI (UUI), overflow UI (OUI), and mixed UI (MUI) [18]. SUI, defined as involuntary leakage of urine on effort or physical exertion, or on sneezing or coughing, is reported to be the most common subtype by most previous UI studies in India [19]. Studies have reported on the association of insomnia and various urologic symptoms, such as nocturia, the commonly observed phenomenon of individuals awakening and realizing that they have to void, which can be a cause of insomnia [20]. However, nocturia is only one of a number of urologic dysfunctions that may be associated with insomnia [21]. Insomnia contributes to neurodegeneration and endocrine dysfunction in ways similar to those seen in ageing, suggesting that insomnia may contribute to the frequency and the severity of age-related chronic disorders, such as UI [22]. However, there is limited evidence demonstrating the link between insomnia symptoms and UI, and even fewer studies have reported the potential relationship between insomnia symptoms and SUI. Considering the sex differences in anatomical structures, risk factors, causes as well as pathophysiological mechanism of UI/SUI, we evaluated the relationships between insomnia symptoms and UI and SUI, separately for males and females, employing the nationally representative data from Wave 1 of the Longitudinal Ageing Study in India (LASI) [10, 23]. We hypothesized that compared to participants without insomnia symptoms those with insomnia symptoms would have higher prevalence rates of UI and SUI. ## Data Data for our study were drawn from wave 1 of LASI, collected from 2017 to 2018. LASI is a representative national study that includes 72,250 individuals aged 45 years and over and their spouses irrespective of age, across all 35 states (except Sikkim) and union territories of India [24]. The LASI database focuses on health and socioeconomic determinants and consequences of the aging process. The survey utilized a multi-stage clustering sampling design. Detailed study methods and microdata for LASI can be accessed at https://lasi-india.org and https://g2aging.org [25]. Ethical approvals and necessary guidelines were approved by The Indian Council of Medical Research (ICMR, Delhi) and the International Institute for Population Sciences (IIPS, Mumbai) with participant informed consent obtained prior to wave 1 survey data collection. Our study was restricted to the data of older adults aged ≥ 60 years, which included 26,821 eligible participants (12,717 males and 14,104 females). We excluded individuals with incomplete information on insomnia symptoms ($$n = 22$$), on diagnosed UI and/or SUI ($$n = 686$$), and missing data on covariates ($$n = 3948$$) (Fig. 1).Fig. 1Sample selection for the study. LASI, Longitudinal Ageing Study of India; UI, urinary incontinence; SUI, stress urinary incontinence ## Insomnia symptoms Detailed data about insomnia symptoms were collected, including three symptoms: 1) trouble falling asleep; 2) waking up at night and having trouble getting back to sleep; 3) waking too early in the morning and not being able to fall asleep. Those three symptoms were selected on the basis of previous studies and were in line with the clinically diagnosed insomnia [26–28], correspondingly, difficulty in initiating sleep, difficulty in maintaining sleep, and early morning awakening were analyzed. The frequency of insomnia symptoms was also recorded, as rarely or never (0–2 times per week), occasionally (3–4 times per week), and frequently (5 or more times per week). We defined a participant as having insomnia symptoms based on their report of at least one of the three insomnia symptoms occurring five or more times per week. ## Diagnosed UI and SUI UI was defined based on self-report of whether respondents had ever been diagnosed with UI. SUI was defined based on self-report of respondents ever having passed urine while sneezing, coughing, laughing, or lifting heavy objects. ## Covariates To ensure that our results are representative and be applied to a wide range of individuals, we adjusted for several potential covariates including sociodemographic and biological factors. Sociodemographic variables included; age in years, sex, level of education (no schooling/less than 5 years complete/5–9 years complete/10 or more years complete), working status (currently unemployed/currently employed), marital status (married or partnered/widowed/others), living arrangement (co-residential living/separate living, “co-residential living” for “living with spouse, children or other household member”, and “separate living” for “living alone”), place of residence (urban/rural), economic status (low/middle/high, trichotomized by annual per capita household consumption. Annual per capita household consumption was used as a proxy for economic status based on prior studies [29, 30]. Household consumption included self-reported expenditure on food, household utilities, fees, durable goods, education, transit, remittances, and discretionary spending and outpatient and inpatient health care in the previous year. Annual per capita household consumption was calculated by taking household consumption divided by the total number of household members), religion (Hindu/Muslim/Christian/others) and caste (scheduled caste/scheduled tribal/other backward class/none of the above). Insomnia symptoms were associated with underweight, chronic diseases and physically inactivity based on prior study using LASI data [26].Thus, we included biological factors, i.e., body mass index (BMI), waist-to-hip ratio, frequency of vigorous physical activity, number of chronic diseases, medication/treatment status, self-rated health (SRH), drinking status, smoking status, depression, and pain. Medication/*Treatment status* was categorized into “no” for “never having taken medication or used other treatments to help sleep” and “yes” for “having taken medication or used other treatments to help sleep”. BMI was categorized as < 18.5, 18.5–25, 25–29.9, ≥ 30 kg/m2. Waist-to-hip ratio was dichotomized into low risk (< 0.90 for male, while < 0.85 for female) and high risk (≥ 0.90 for male, while ≥ 0.85 for female). Vigorous physical activity was about respondents’ involvement in running or jogging, swimming, going to a health center/gym, cycling, digging with a spade or shovel, heavy lifting, chopping, farm work, fast bicycling, and cycling with loads and was classified by frequency as “every day”, “more than once a week”, “once a week”, “1–3 times a month”, “hardly ever or never”. Number of chronic diseases included self-reported hypertension, diabetes, tumor, lung disease, chronic heart disease, stroke, arthritis, mental disease, Alzheimer's disease, hypercholesterolemia, asthma, congestive heart failure, heart attack, abnormal heart rate, osteoporosis, abnormal thyroid function, digestive disease, skin disease, kidney stone, presbyopia, cataract, glaucoma, myopia, hyperopia, tooth decay, and periodontal disease. The variable was recorded as “0” if the respondent did not have any chronic disease, “1”, if the respondent had only one chronic disease and “2 + ” if the respondent had more than two chronic diseases. SRH was sorted by a five‐point Likert scale as “excellent,” “very good,” “good,” “fair” and “poor”, which was a proxy indicator for health status. Drinking status, defined as consumption of any form of alcohol in one’s lifetime, was categorized into “no” for “never having had an alcoholic drink” and “yes” for “currently or ever having consumed any alcoholic beverages.” Smoking status, defined as consumption any form of tobacco in one’s lifetime, was categorized into “no” for “never having smoked” and “yes” for “current or ever having smoked.” Depression was dichotomized into “no” for “not diagnosed with depression” and “yes” for “diagnosed.” The presence of depression was evaluated using internationally validated 10-item Center for Epidemiologic Studies Depression Scale (CESD-10), in which score ≥ 4 out of overall 10 score was defined as depression) [31]. Pain was divided into “no” for “not troubled with pain” and “yes” for “often troubled with pain.” ## Statistical analysis We expressed continuous variables as mean and standard deviation and categorical variables as proportions. Kruskal Wallis H test (continuous variables) and chi-square tests (categorical variables) were used to calculate statistical differences in baseline characteristics among groups dichotomized by whether reporting insomnia symptoms. Besides, we conducted multivariate logistic regression dividing the individuals by sex to evaluate the associations between insomnia symptoms and diagnosed UI or SUI. The incremental models were constructed adjusting for covariates: no covariates in the unadjusted model; sociodemographic covariates in model 1 i.e., age, level of education, work status, marital status, living arrangement, place of residence, economic status, religion, and caste. And for the fully adjusted model 2, we adjusted for the sociodemographic mentioned above and for the biological covariates: medication/treatment status, BMI, vigorous physical activity, waist-to-hip ratio, number of chronic diseases, SRH, drinking status, smoking status, depression, and pain. We performed interaction analyses to evaluate the heterogeneity of association between insomnia symptoms and UI or SUI stratified by covariates (including age, BMI, drinking status, smoking status and medication/treatment status). Given that average life expectancy in *India is* 69.4 years in 2014–18, age was categorized as < 70 and ≥ 70 years [32]. BMI levels were categorized into underweight (< 18.5 kg/m2), normal (18.5 to 24.9 kg/m2) and overweight/obese (≥ 25 kg/m2) subgroups, since the sample size of obese subjects (BMI ≥ 30 kg/m2) was limited. Drinking and smoking status were dichotomized into “yes” and “no” as mentioned above. As a control for potential sex-based differences, our study investigated the interaction separately for males and females. The subgroup analyses were performed using stratified linear regression models, while the p for interaction was calculated using the log-likelihood ratio test to compare the differences between models with and without the interaction of covariates. All statistical analyses were conducted using the statistical software packages R (http://www.R-project.org, The R Foundation) and Empower (http://www.empowerstats.com). Two-tailed P-values were performed with a significance level of < 0.05. ## Baseline characteristics The characteristics and related covariates of participants are summarized in Table 1. The prevalence of insomnia symptoms was $11.11\%$, Patients with insomnia symptoms were more likely to be older, female, less educated, unemployed, unmarried, living separately, living in rural areas, Hindu, lower proportion of high-risk waist-to-hip ratio, physically inactivity, BMI ≥ 30 or < 18.5 kg/m2, with two or more chronic diseases, and with poor SRH.Table 1Baseline characteristics of participantsCharacteristicsTotalInsomnia symptomsNo ($$n = 23$$,842)Yes ($$n = 2979$$)p-valueAge, year (mean ± SD)68.61 ± 7.2768.50 ± 7.2069.49 ± 7.72 < 0.001Sex, n (%) < 0.001 Male12,717 ($47.41\%$)11,470 ($48.11\%$)1247 ($41.86\%$) Female14,104 ($52.59\%$)12,372 ($51.89\%$)1732 ($58.14\%$)Level of education, n (%) < 0.001 No schooling14,384 ($53.63\%$)12,662 ($53.11\%$)1722 ($57.80\%$) Less than 5 years complete8463 ($31.55\%$)7540 ($31.62\%$)923 ($30.98\%$) 5–9 years complete2796 ($10.42\%$)2551 ($10.70\%$)245 ($8.22\%$) 10 or more years complete1178 ($4.39\%$)1089 ($4.57\%$)89 ($2.99\%$)Working status, n (%) < 0.001 Currently unemployed17,538 ($65.39\%$)15,460 ($64.84\%$)2078 ($69.75\%$) Currently employed9283 ($34.61\%$)8382 ($35.16\%$)901 ($30.25\%$)UI, n (%) < 0.001 No25,845 ($96.36\%$)23,083 ($96.82\%$)2762 ($92.72\%$) Yes976 ($3.64\%$)759 ($3.18\%$)217 ($7.28\%$)SUI, n (%) < 0.001 No24,095 ($89.84\%$)21,628 ($90.71\%$)2467 ($82.81\%$) Yes2726 ($10.16\%$)2214 ($9.29\%$)512 ($17.19\%$)Medications/Treatments to help sleep, n (%) < 0.001 No26,014 ($96.99\%$)23,342 ($97.90\%$)2672 ($89.69\%$) Yes807 ($3.01\%$)500 ($2.10\%$)307 ($10.31\%$)BMI, kg/m2, n (%) < 0.001 < 18.56174 ($23.02\%$)5367 ($22.51\%$)807 ($27.09\%$) 18.5–2514,094 ($52.55\%$)12,647 ($53.05\%$)1447 ($48.57\%$) 25–304996 ($18.63\%$)4458 ($18.70\%$)538 ($18.06\%$) ≥ 301557 ($5.81\%$)1370 ($5.75\%$)187 ($6.28\%$)Waist-to-hip ratio, n (%) < 0.001 Low risk2851 ($10.63\%$)2444 ($10.25\%$)407 ($13.66\%$) High risk23,970 ($89.37\%$)21,398 ($89.75\%$)2572 ($86.34\%$)Vigorous physical activity, n (%) < 0.001 Everyday4874 ($18.17\%$)4432 ($18.59\%$)442 ($14.84\%$) More than once a week1502 ($5.60\%$)1375 ($5.77\%$)127 ($4.26\%$) *Once a* week855 ($3.19\%$)766 ($3.21\%$)89 ($2.99\%$) One to three times a month1200 ($4.47\%$)1078 ($4.52\%$)122 ($4.10\%$) Hardly ever or never18,390 ($68.57\%$)16,191 ($67.91\%$)2199 ($73.82\%$)Number of chronic diseases, n (%) < 0.001 04410 ($16.44\%$)4127 ($17.31\%$)283 ($9.50\%$) 15771 ($21.52\%$)5293 ($22.20\%$)478 ($16.05\%$) 2 + 16,640 ($62.04\%$)14,422 ($60.49\%$)2218 ($74.45\%$)SRH, n (%) < 0.001 Excellent750 ($2.80\%$)708 ($2.97\%$)42 ($1.41\%$) Very good3889 ($14.50\%$)3619 ($15.18\%$)270 ($9.06\%$) Good9767 ($36.42\%$)8989 ($37.70\%$)778 ($26.12\%$) Fair8827 ($32.91\%$)7751 ($32.51\%$)1076 ($36.12\%$) Poor3588 ($13.38\%$)2775 ($11.64\%$)813 ($27.29\%$)Drinking status, n (%)0.91 Never22,222 ($82.85\%$)19,756 ($82.86\%$)2466 ($82.78\%$) Current/ever4599 ($17.15\%$)4086 ($17.14\%$)513 ($17.22\%$)Smoking status, n (%)0.058 Never21,394 ($79.77\%$)19,057 ($79.93\%$)2337 ($78.45\%$) Current/ever5427 ($20.23\%$)4785 ($20.07\%$)642 ($21.55\%$)Depression, n (%) < 0.001 No19,486 ($72.65\%$)17,846 ($74.85\%$)1640 ($55.05\%$) Yes7335 ($27.35\%$)5996 ($25.15\%$)1339 ($44.95\%$)Pain, n (%) < 0.001 No15,920 ($59.36\%$)14,618 ($61.31\%$)1302 ($43.71\%$) Yes10,901 ($40.64\%$)9224 ($38.69\%$)1677 ($56.29\%$)Marital status, n (%) < 0.001 Married or partnered17,212 ($64.17\%$)15,479 ($64.92\%$)1733 ($58.17\%$) Widowed9068 ($33.81\%$)7880 ($33.05\%$)1188 ($39.88\%$) Others541 ($2.02\%$)483 ($2.03\%$)58 ($1.95\%$)Living arrangement, n (%)0.001 Co-residential living25,429 ($94.81\%$)22,641 ($94.96\%$)2788 ($93.59\%$) Separate living1392 ($5.19\%$)1201 ($5.04\%$)191 ($6.41\%$)Place of residence, n (%) < 0.001 Urban8911 ($33.22\%$)8010 ($33.60\%$)901 ($30.25\%$) Rural17,910 ($66.78\%$)15,832 ($66.40\%$)2078 ($69.75\%$)Economic status, n (%)0.357 Low9570 ($35.68\%$)8511 ($35.70\%$)1059 ($35.55\%$) Middle9141 ($34.08\%$)8153 ($34.20\%$)988 ($33.17\%$) High8110 ($30.24\%$)7178 ($30.11\%$)932 ($31.29\%$)Religion, n (%) < 0.001 Hindu19,700 ($73.45\%$)17,366 ($72.84\%$)2334 ($78.35\%$) Muslim3158 ($11.77\%$)2808 ($11.78\%$)350 ($11.75\%$) Christian2632 ($9.81\%$)2475 ($10.38\%$)157 ($5.27\%$) Others1331 ($4.96\%$)1193 ($5.00\%$)138 ($4.63\%$)Caste, n (%) < 0.001 Scheduled caste4410 ($16.44\%$)3846 ($16.13\%$)564 ($18.93\%$) Scheduled trible4465 ($16.65\%$)4146 ($17.39\%$)319 ($10.71\%$) Other backward class10,287 ($38.35\%$)9028 ($37.87\%$)1259 ($42.26\%$) Other castes7659 ($28.56\%$)6822 ($28.61\%$)837 ($28.10\%$)SD standard deviation, UI urinary incontinence, SUI stress urinary incontinence, BMI body mass index, SRH self-rated healthMean ± SD for continuous variables: P value was calculated by Kruskal Wallis H testNumber (%) for categorical variables: P value was calculated by chi-square test $3.64\%$ of respondents reported UI and $10.16\%$ reported SUI. UI was reported by $7.28\%$ of participants with insomnia symptoms compared to $3.18\%$ by participants without insomnia symptoms. SUI was reported by $17.19\%$ of participants with insomnia symptoms and by $9.29\%$ of participants without insomnia symptoms. ## Insomnia symptoms and associated UI and SUI Results from the multivariable linear regression analysis of insomnia symptoms and UI, and insomnia symptoms and SUI are shown in Table 2. Having separated the respondents by sex, we found that, after adjusting for only sociodemographic covariates (model 1), insomnia symptoms were associated with UI for both male respondents (OR 2.24; $95\%$ CI 1.78–2.83) and female respondents (OR 2.23; $95\%$ CI 1.80–2.77). Further adjustment for health characteristics (model 2) moderately attenuated the association for both sexes (OR 1.53; $95\%$ CI 1.20–1.96 for males, and OR 1.53; $95\%$ CI 1.21–1.92 for females). Similarly, insomnia symptoms were significantly related to SUI symptoms in model 1 for both males (OR 2.13; $95\%$ CI 1.78–2.54) and females (OR 1.90; $95\%$ CI 1.66–2.17). Full adjustment (model 2) attenuated the association for both sexes (OR 1.51; $95\%$ CI 1.25–1.83 for males, and OR 1.50; $95\%$ CI 1.31–1.73 for females).Table 2Relationship between insomnia symptoms and associated UI and SUIInsomnia symptomsOR ($95\%$ CI)Male ($$n = 12$$,717)Female ($$n = 14$$,104)UI Unadjusted model2.50 (1.99, 3.15)2.33 (1.88, 2.89) Model 12.24 (1.78, 2.83)2.23 (1.80, 2.77) Model 21.53 (1.20, 1.96)1.53 (1.21, 1.92)SUI Unadjusted model2.22 (1.86, 2.64)1.53 (1.21, 1.92) Model 12.13 (1.78, 2.54)1.90 (1.66, 2.17) Model 21.51 (1.25, 1.83)1.50 (1.31, 1.73)OR odds ratio, $95\%$ CI $95\%$ Confidence interval, UI urinary incontinence, SUI stress urinary incontinenceUnadjusted model: no covariates were adjustedModel 1 adjusted for: age, level of education, work status, marital status, living arrangement, place of residence, economic status, religion, casteModel 2 adjusted for: age, level of education, work status, marital status, religion, place of residence, living arrangement, economic status, caste, medication/treatment status, body mass index (BMI), vigorous physical activity, waist-to-hip ratio, number of chronic diseases, self-rated health (SRH), drinking status, smoking status, depression, pain ## Subgroup analysis We conducted interaction tests to further assess the relationship between insomnia symptoms and UI and SUI the results of which are presented in Fig. 2. And we recategorized BMI into three groups: < 18.5, 18.5–25, ≥ 25 kg/m2 in subgroup analysis, due to the limited sample size of BMI ≥ 30 kg/m2. No significant interactions were found in stratified analyses by age, BMI, drinking status, smoking status and medication/treatment status for the associations between insomnia symptoms and UI for males ($$p \leq 0.560$$, 0.556, 0.732, 0.204, 0.146, respectively) or for females ($$p \leq 0.183$$, 0.745, 0.934, 0.610, 0.317, respectively). Comparable lack of significant insomnia/SUI interactions were found in stratified analyses for BMI, drinking, smoking and medications/treatments taking for both males ($$p \leq 0.357$$, 0.281, 0.066, 0.265, respectively) and females ($$p \leq 0.381$$, 0.252, 0.836, 0.381, respectively). However, in the subgroup analysis stratified by age, there was a significant interaction of insomnia symptoms and SUI for both males ($$p \leq 0.048$$) and females ($$p \leq 0.042$$), with positive associations observed among participants aged both < 70 years (OR 1.68; $95\%$ CI 1.26–2.23 for males, while OR 1.55; $95\%$ CI 1.28–1.87 for females) and ≥ 70 years (OR 1.40; $95\%$ CI 1.08–1.81 for males, and OR 1.44; $95\%$ CI 1.17–1.77 for females).Fig. 2Subgroup analysis of relationship between insomnia and associated UI and SUI. We recoded the BMI and recategorized it into three groups: < 18.5, 18.5–25, ≥ 25 kg/m2, due to the limited sample size of BMI ≥ 30 kg/m2. OR, odds ratio; $95\%$ CI, $95\%$ Confidence interval; UI, urinary incontinence; SUI, stress urinary incontinence; BMI, body mass index. Model 2 adjusted for: age, level of education, work status, marital status, religion, place of residence, living arrangement, economic status, caste, medication/treatment status, BMI, vigorous physical activity, waist-to-hip ratio, number of chronic diseases, self-rated health (SRH), drinking status, smoking status, depression, pain except the subgroup variable ## Discussion Insomnia symptoms were present in $11.11\%$ of the study sample and was significantly associated with both diagnosed UI and with SUI in both males and females. These associations remained statistically significant after adjustment for multiple potential covariates. Notably, we did not detect a significant sex difference on the relationship between insomnia symptoms and UI and SUI. moreover, only a single a significant interaction was observed, that of age in the association of insomnia symptoms and SUI in both males and females, with the insomnia/SUI association positive among participants aged < 70 years and ≥ 70 years, which did not differ by sex. The prevalence of diagnosed UI was $3.64\%$, and $10.16\%$ respondents reported symptoms of SUI. Diagnosed UI prevalence was lower than that of SUI symptoms. Compared with the relatively strict diagnosis of UI, SUI status was based on reporting ever having experienced any SUI event, a very lax definition which likely resulted in the higher reported prevalence. Moreover, low treatment seeking for UI may also contribute to this discrepancy, which is in line with prior studies, suggesting that UI remains underrecognized and underestimated, with fewer than $40\%$ of affected females seeking treatments to diagnose this disease [33]. Nevertheless, the overall prevalence of UI was relatively lower that reported in several small-sample studies focusing on the general Indian population [15, 16]. It can be interpreted by the fact that there is wide variability in UI prevalence estimates ($5\%$ to $70\%$) from various countries, depending on the definitions of UI used, the study population, and the assessment tools and availability of health care, and with higher prevalence rates reported in western countries [9, 34, 35]. Inconsistent relationships between sleep quality and UI have been reported in earlier studies. Siddiqui et al. carried out cross-sectional analysis on a sample of 510 treatment seeking females with UI, finding out that there was no difference in sleep quality based on the presence and severity of urinary incontinence after adjustment [36]. Dasdemir Ilkhan et al. reported that sleep status was not associated with UI and incontinence-related life quality among 1150 older adults residing in nursing homes in Istanbul ($P \leq 0.05$) [37]. In contrast, Araujo et al. conducted a prospective cohort study of 4145 individuals, reported a bi-directional link between sleep-related problems and UI, with BMI possibly mediating the relationship [21]. Ge et al. in a study of fifty-one overactive bladder (OAB) patients, reported a positive correlation between sleep quality and UI status [38]. Yilmaz Bulut et al. found a similar association between UI symptoms and sleep quality among 140 older females lived in Turkey [39]. One possible explanation of our finding is that mental factors could mediate the association between insomnia symptoms and UI. Several longitudinal studies and meta-analyses have identified insomnia as a risk factor for depression, anxiety, and other mental disorders among adults [40, 41]. It is conceivable that insomnia’s adverse impact on mental health may affect symptoms of UI. This is supported by studies reporting that depression is related to UI symptom severity, functional impairment, and incontinence-related life quality [42]. In addition, the negative emotional impacts caused by UI can in turn contribute to sleep disturbance, given that the association between insomnia and psychological distress is bidirectional [43, 44]. Finally, insomnia is also accompanied by daytime function difficulties and cognitive decline, which may further contribute to the burden of UI, making it more difficult for UI patients to lose weight, reduce caffeine and nicotine intake, and conduct pelvic floor muscle training [45, 46]. Another possible link between insomnia symptoms and UI is metabolic disturbance. Acute and chronic sleep deprivation is associated with metabolic disorders, which influence hormonal secretion patterns, autonomic nervous system balance and vasopressin secretion [22, 47]. These pathways each impact the regulation of smooth muscle tone, which is fundamental to relaxation/contraction of the detrusor and bladder musculature, and thus may be linked with urination function [21]. Additionally, obesity and related metabolic disorders are also related to systemic inflammation, pro-inflammatory cytokine release and oxidative stress, thereby altering collagen metabolism, accompanied with increased intra-abdominal pressure, leading to the progression of SUI [48, 49]. Accumulating evidence also suggests a possible neuro-molecular mechanism underlying the association between insomnia symptoms and UI and SUI. Disturbed sleep, which has long been considered a symptom of neurodegenerative conditions, may in fact be a risk factor for and trigger the onset of these diseases in the early stages via processes such as endoplasmic reticulum stress and neuronal damage [50–52]. Studies on the molecular mechanism of SUI have shown that SUI is related to the differential expression of neuronal cell-specific proteins and neurodegeneration-related proteins, which indicates the potential involvement of a neurodegeneration process in SUI [53]. We speculate that potential neuropathological effects of insomnia symptoms may be related to the development of SUI. Our study has clear limitations. Given that it is cross-sectional, only associational and not causal relationships can be inferred. Future waves of data collection in LASI will allow for identification of potential causal relationships. Monthly recall of insomnia symptoms can only provide short-term information and may not reflect participants’ usual sleep status, while the UI and SUI status can provide long-term information based on ever having been diagnosed UI or experienced any SUI event. Thus, the mismatch of time timeframe limits our ability to identity the direction of causality for the relationships we studied. Self-report of health conditions, including insomnia symptoms, UI, and SUI, are subject to recall bias, idiosyncratic interpretation of the question and other reporting errors. An additional limitation the lack of severity and duration measures for insomnia symptoms, UI and SUI, such that the contribution of severity and duration to associations of insomnia symptoms and UI and SUI could not be assessed. An analytic limitation is that data analyses were limited to variables collected in the parent study, which limited our ability to define better insomnia symptoms and SUI, as well as prohibited the examination of other forms of UI. Finally, the LASI database, although extensive, limited the number of potential co-variates available for analysis. Nevertheless, our study also has several strengths. Data in our study were collected from a large nationally representative sample, using standardized processes and protocols to ensure high quality data. To the best of our knowledge, this is the first attempt to explore the relationships between insomnia symptoms and UI and SUI among older Indian males and females. To ensure that our results were nationally representative our data analyses were adjusted for multiple potential covariates. After full adjustment, the observed associations between insomnia symptoms and UI and SUI remained unchanged, although their magnitudes were diminished, which suggests that the study’s findings are robust with results that are stable and reliable. Additionally, our results indicate the need for future longitudinal study of the association between insomnia symptoms and UI and SUI to determine its directionality and explore potential underlying mechanisms. ## Conclusions Insomnia symptoms were associated with greater prevalence both of diagnosed UI and of SUI among older Indian males and females, independent of covariates. It can be recommended that older adults with insomnia symptoms need early screening and appropriate treatment of UI and SUI, which improve their quality of life and gain in public health benefits in low- and middle-income countries, as an unprecedented pace of population aging continues. And further research is called for to better characterize this association, determine underlying mechanisms and to explore the potential therapeutic implications. ## References 1. Ravindranath V, Sundarakumar JS. **Changing demography and the challenge of dementia in India**. *Nat Rev Neurol* (2021.0) **17** 747-758. DOI: 10.1038/s41582-021-00565-x 2. Campisi J, Kapahi P, Lithgow GJ, Melov S, Newman JC, Verdin E. **From discoveries in ageing research to therapeutics for healthy ageing**. *Nature* (2019.0) **571** 183-192. DOI: 10.1038/s41586-019-1365-2 3. Riemann D, Nissen C, Palagini L, Otte A, Perlis ML, Spiegelhalder K. **The neurobiology, investigation, and treatment of chronic insomnia**. *Lancet Neurol* (2015.0) **14** 547-558. DOI: 10.1016/S1474-4422(15)00021-6 4. Buysse DJ. **Insomnia**. *JAMA* (2013.0) **309** 706-716. DOI: 10.1001/jama.2013.193 5. Koyanagi A, Garin N, Olaya B, Ayuso-Mateos JL, Chatterji S, Leonardi M, Koskinen S, Tobiasz-Adamczyk B, Haro JM. **Chronic conditions and sleep problems among adults aged 50 years or over in nine countries: a multi-country study**. *PLoS ONE* (2014.0) **9** e114742. DOI: 10.1371/journal.pone.0114742 6. Patel D, Steinberg J, Patel P. **Insomnia in the elderly: a review**. *J Clin Sleep Med* (2018.0) **14** 1017-1024. DOI: 10.5664/jcsm.7172 7. Nakakubo S, Doi T, Tsutsumimoto K, Kurita S, Ishii H, Shimada H. **Sleep duration and progression to sarcopenia in Japanese community-dwelling older adults: a 4 year longitudinal study**. *J Cachexia Sarcopenia Muscle* (2021.0) **12** 1034-1041. DOI: 10.1002/jcsm.12735 8. Irwin MR, Vitiello MV. **Implications of sleep disturbance and inflammation for Alzheimer's disease dementia**. *The Lancet Neurology* (2019.0) **18** 296-306. DOI: 10.1016/S1474-4422(18)30450-2 9. Milsom I, Gyhagen M. **The prevalence of urinary incontinence**. *Climacteric* (2019.0) **22** 217-222. DOI: 10.1080/13697137.2018.1543263 10. Vaughan CP, Markland AD. **Urinary incontinence in women**. *Ann Intern Med* (2020.0) **172** ITC17-ITC32. DOI: 10.7326/AITC202002040 11. Landefeld CS, Bowers BJ, Feld AD, Hartmann KE, Hoffman E, Ingber MJ. **National Institutes of Health state-of-the-science conference statement: prevention of fecal and urinary incontinence in adults**. *Ann Intern Med* (2008.0) **148** 449-58. DOI: 10.7326/0003-4819-148-6-200803180-00210 12. Yang E, Lisha NE, Walter L, Obedin-Maliver J, Huang AJ. **Urinary incontinence in a national cohort of older women: implications for caregiving and care dependence**. *J Womens Health (Larchmt)* (2018.0) **27** 1097-1103. DOI: 10.1089/jwh.2017.6891 13. Hagan KA, Erekson E, Austin A, Minassian VA, Townsend MK, Bynum JPW, Grodstein F. **A prospective study of the natural history of urinary incontinence in women**. *Am J Obstet Gynecol* (2018.0) **218** 502.e1-502.e8. DOI: 10.1016/j.ajog.2018.01.045 14. Aoki Y, Brown HW, Brubaker L, Cornu JN, Daly JO, Cartwright R. **Urinary incontinence in women**. *Nat Rev Dis Primers* (2017.0) **3** 17042. DOI: 10.1038/nrdp.2017.42 15. Agarwal BK, Agarwal N. **Urinary incontinence: prevalence, risk factors, impact on quality of life and treatment seeking behaviour among middle aged women**. *Int Surg J* (2017.0) **4** 1953. DOI: 10.18203/2349-2902.isj20172131 16. Kumari S, Jain V, Mandal AK, Singh A. **Behavioral therapy for urinary incontinence in India**. *Int J Gynaecol Obstet* (2008.0) **103** 125-130. DOI: 10.1016/j.ijgo.2008.06.019 17. Thom D. **Variation in estimates of urinary incontinence prevalence in the community: effects of differences in definition, population characteristics, and study type**. *J Am Geriatr Soc* (1998.0) **46** 473-80. DOI: 10.1111/j.1532-5415.1998.tb02469.x 18. Lukacz ES, Santiago-Lastra Y, Albo ME, Brubaker L. **Urinary incontinence in women**. *Jama* (2017.0) **318** 1592. DOI: 10.1001/jama.2017.12137 19. Haylen BT, de Ridder D, Freeman RM, Swift SE, Berghmans B, Lee J, Monga A, Petri E, Rizk DE, Sand PK. **An International Urogynecological Association (IUGA)/International Continence Society (ICS) joint report on the terminology for female pelvic floor dysfunction**. *Neurourol Urodyn* (2010.0) **29** 4-20. DOI: 10.1002/nau.20798 20. Ancoli-Israel S, Bliwise DL, Norgaard JP. **The effect of nocturia on sleep**. *Sleep Med Rev* (2011.0) **15** 91-97. DOI: 10.1016/j.smrv.2010.03.002 21. Araujo AB, Yaggi HK, Yang M, McVary KT, Fang SC, Bliwise DL. **Sleep related problems and urological symptoms: testing the hypothesis of bidirectionality in a longitudinal, population based study**. *J Urol* (2014.0) **191** 100-106. DOI: 10.1016/j.juro.2013.07.011 22. Spiegel K, Leproult R, Van Cauter E. **Impact of sleep debt on metabolic and endocrine function**. *The Lancet* (1999.0) **354** 1435-1439. DOI: 10.1016/S0140-6736(99)01376-8 23. Gacci M, Sakalis VI, Karavitakis M, Cornu JN, Gratzke C, Herrmann TRW, Kyriazis I, Malde S, Mamoulakis C, Rieken M. **European association of urology guidelines on male urinary incontinence**. *Eur Urol* (2022.0) **82** 387-398. DOI: 10.1016/j.eururo.2022.05.012 24. Bloom DE, Sekher TV, Lee J. **Longitudinal Aging Study in India (LASI): new data resources for addressing aging in India**. *Nature Aging* (2021.0) **1** 1070-1072. DOI: 10.1038/s43587-021-00155-y 25. Perianayagam A, Bloom D, Lee J, Parasuraman S, Sekher TV, Mohanty SK, Chattopadhyay A, Govil D, Pedgaonkar S, Gupta S. **Cohort profile: the Longitudinal Ageing Study in India (LASI)**. *Int J Epidemiol* (2022.0) **51** e167-e176. DOI: 10.1093/ije/dyab266 26. 26.Muhammad T, Gharge S, Meher T. The associations of BMI, chronic conditions and lifestyle factors with insomnia symptoms among older adults in India. PLoS One. 2022;17(9):e0274684. 10.1371/journal.pone.0274684. 27. Chiu HF, Xiang YT, Dai J, Chan SS, Leung T, Yu X, Hou ZJ, Ungvari GS, Caine ED. **The prevalence of sleep problems and their socio-demographic and clinical correlates in young Chinese rural residents**. *Psychiatry Res* (2012.0) **200** 789-794. DOI: 10.1016/j.psychres.2012.03.050 28. Sutton EL. **Insomnia**. *Ann Intern Med* (2021.0) **174** ITC33-ITC48. DOI: 10.7326/AITC202103160 29. Zhao Y, Atun R, Oldenburg B, McPake B, Tang S, Mercer SW, Cowling TE, Sum G, Qin VM, Lee JT. **Physical multimorbidity, health service use, and catastrophic health expenditure by socioeconomic groups in China: an analysis of population-based panel data**. *Lancet Glob Health* (2020.0) **8** e840-e849. DOI: 10.1016/S2214-109X(20)30127-3 30. Farron MR, Kabeto MU, Dey AB, Banerjee J, Levine DA, Langa KM. **Hypertension and cognitive health among older adults in India**. *J Am Geriatr Soc* (2020.0) **68** S29-S35. PMID: 32815597 31. 31.IIrwin M, Artin KH, Oxman MN. Screening for depression in the older adult: criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D). Arch Intern Med. 1999;159(15):1701–4. 10.1001/archinte.159.15.1701. 32. 32.Office of the Registrar General & Census Commissioner IO. Sample Registration System (SRS)-Abridged Life Tables 2014-2018. In. India; 2020. Available from: https://censusindia.gov.in/nada/index.php/catalog/42517. 33. Waetjen LE, Xing G, Johnson WO, Melnikow J, Gold EB. **Study of women's health across the N: factors associated with reasons incontinent midlife women report for not seeking urinary incontinence treatment over 9 years across the menopausal transition**. *Menopause* (2018.0) **25** 29-37. DOI: 10.1097/GME.0000000000000943 34. Cao C, Zhang C, Sriskandarajah C, Xu T, Gotto G, Sutcliffe S, Yang L. **Trends and racial disparities in the prevalence of urinary incontinence among men in the USA, 2001–2020**. *Eur Urol Focus* (2022.0) **8** 1758. DOI: 10.1016/j.euf.2022.04.015 35. Abufaraj M, Xu T, Cao C, Siyam A, Isleem U, Massad A, Soria F, Shariat SF, Sutcliffe S, Yang L. **Prevalence and trends in urinary incontinence among women in the United States, 2005–2018**. *Am J Obstet Gynecol* (2021.0) **225** 166 e161-166 e112. DOI: 10.1016/j.ajog.2021.03.016 36. Siddiqui NY, Wiseman JB, Cella D, Bradley CS, Lai HH, Helmuth ME, Smith AR, Griffith JW, Amundsen CL, Kenton KS. **Mental health, sleep and physical function in treatment seeking women with urinary incontinence**. *J Urol* (2018.0) **200** 848-855. DOI: 10.1016/j.juro.2018.04.076 37. DasdemirIlkhan G, Celikhisar H. **The effect of incontinence on sleep quality in the elderly**. *Int J Clin Pract* (2021.0) **75** e13965. PMID: 33368941 38. Ge TJ, Vetter J, Lai HH. **Sleep disturbance and fatigue are associated with more severe urinary incontinence and overactive bladder symptoms**. *Urology* (2017.0) **109** 67-73. DOI: 10.1016/j.urology.2017.07.039 39. Yilmaz Bulut T, Altay B. **Sleep quality and quality of life in older women with urinary incontinence residing in Turkey: a cross-sectional survey**. *J Wound Ostomy Continence Nurs* (2020.0) **47** 166-171. DOI: 10.1097/WON.0000000000000615 40. Hertenstein E, Feige B, Gmeiner T, Kienzler C, Spiegelhalder K, Johann A, Jansson-Frojmark M, Palagini L, Rucker G, Riemann D. **Insomnia as a predictor of mental disorders: a systematic review and meta-analysis**. *Sleep Med Rev* (2019.0) **43** 96-105. DOI: 10.1016/j.smrv.2018.10.006 41. Baglioni C, Battagliese G, Feige B, Spiegelhalder K, Nissen C, Voderholzer U, Lombardo C, Riemann D. **Insomnia as a predictor of depression: a meta-analytic evaluation of longitudinal epidemiological studies**. *J Affect Disord* (2011.0) **135** 10-19. DOI: 10.1016/j.jad.2011.01.011 42. Melville JL, Walker E, Katon W, Lentz G, Miller J, Fenner D. **Prevalence of comorbid psychiatric illness and its impact on symptom perception, quality of life, and functional status in women with urinary incontinence**. *Am J Obstet Gynecol* (2002.0) **187** 80-87. DOI: 10.1067/mob.2002.124839 43. Fang H, Tu S, Sheng J, Shao A. **Depression in sleep disturbance: A review on a bidirectional relationship, mechanisms and treatment**. *J Cell Mol Med* (2019.0) **23** 2324-2332. DOI: 10.1111/jcmm.14170 44. Konjarski M, Murray G, Lee VV, Jackson ML. **Reciprocal relationships between daily sleep and mood: A systematic review of naturalistic prospective studies**. *Sleep Med Rev* (2018.0) **42** 47-58. DOI: 10.1016/j.smrv.2018.05.005 45. Ma Y, Liang L, Zheng F, Shi L, Zhong B, Xie W. **Association between sleep duration and cognitive decline**. *JAMA Netw Open* (2020.0) **3** e2013573. DOI: 10.1001/jamanetworkopen.2020.13573 46. Morin CM, Drake CL, Harvey AG, Krystal AD, Manber R, Riemann D, Spiegelhalder K. **Insomnia disorder**. *Nat Rev Dis Primers* (2015.0) **1** 15026. DOI: 10.1038/nrdp.2015.26 47. Yaggi HK, Araujo AB, McKinlay JB. **Sleep duration as a risk factor for the development of type 2 diabetes**. *Diabetes Care* (2006.0) **29** 657-61. DOI: 10.2337/diacare.29.03.06.dc05-0879 48. Doumouchtsis SK, Loganathan J, Pergialiotis V. **The role of obesity on urinary incontinence and anal incontinence in women: a review**. *BJOG* (2022.0) **129** 162-170. DOI: 10.1111/1471-0528.16848 49. Lin G, Shindel AW, Banie L, Deng D, Wang G, Hayashi N, Lin CS, Lue TF. **Molecular mechanisms related to parturition-induced stress urinary incontinence**. *Eur Urol* (2009.0) **55** 1213-1222. DOI: 10.1016/j.eururo.2008.02.027 50. Musiek ES, Holtzman DM. **Mechanisms linking circadian clocks, sleep, and neurodegeneration**. *Science* (2016.0) **354** 1004-1008. DOI: 10.1126/science.aah4968 51. Naidoo N, Ferber M, Master M, Zhu Y, Pack AI. **Aging impairs the unfolded protein response to sleep deprivation and leads to proapoptotic signaling**. *J Neurosci* (2008.0) **28** 6539-6548. DOI: 10.1523/JNEUROSCI.5685-07.2008 52. Benedict C, Cedernaes J, Giedraitis V, Nilsson EK, Hogenkamp PS, Vagesjo E, Massena S, Pettersson U, Christoffersson G, Phillipson M. **Acute sleep deprivation increases serum levels of neuron-specific enolase (NSE) and S100 calcium binding protein B (S-100B) in healthy young men**. *Sleep* (2014.0) **37** 195-198. DOI: 10.5665/sleep.3336 53. Post WM, Widomska J, Grens H, Coenen MJH, Martens FMJ, Janssen DAW, IntHout J, Poelmans G, Oosterwijk E, Kluivers KB. **Molecular processes in stress urinary incontinence: a systematic review of human and animal studies**. *Int J Mol Sci* (2022.0) **23** 3401. DOI: 10.3390/ijms23063401
--- title: 'Polycystic ovary syndrome is associated with a higher risk of premalignant and malignant endometrial polyps in premenopausal women: a retrospective study in a tertiary teaching hospital' authors: - Ling Lu - Jianbo Luo - Jie Deng - Chaolin Huang - Chanyu Li journal: BMC Women's Health year: 2023 pmcid: PMC10037815 doi: 10.1186/s12905-023-02269-4 license: CC BY 4.0 --- # Polycystic ovary syndrome is associated with a higher risk of premalignant and malignant endometrial polyps in premenopausal women: a retrospective study in a tertiary teaching hospital ## Abstract ### Background Polycystic ovary syndrome (PCOS) is characterized by anovulation, insufficient progesterone, hyperandrogenism, and insulin resistance. These factors can disrupt the endometrium of PCOS patients and can lead to chronic low-grade inflammation in the endometrium, endometrial hyperplasia, or even endometrial cancer. ### Objective The aim of this study was to investigate the prevalence of premalignant and malignant endometrial polyps in premenopausal women and to further explore whether PCOS is associated with premalignant and malignant changes in endometrial polyps. ### Methods This study was conducted by retrieving the medical data of 4236 premenopausal women who underwent hysteroscopic polypectomies between January 2015 and December 2021. Demographic and clinical data regarding age, height, weight, parity, hormone replacement therapy, oral contraceptives, abnormal uterine bleeding, hypertension, diabetes mellitus, PCOS, number of polyps, and size of polyps were collected, and their associations with premalignant and malignant changes in endometrial polyps were analysed. ### Result Among the endometrial polyps removed by hysteroscopic polypectomy in premenopausal women, the prevalence of premalignant and malignant polyps was $2.15\%$, which comprised hyperplasia with atypia at $1.13\%$ and endometrial carcinoma at $1.02\%$. PCOS was associated with a higher risk of premalignant and malignant endometrial polyps in premenopausal women after adjustment for potential confounding factors. ### Conclusion PCOS is associated with a higher risk of premalignant and malignant endometrial polyps in premenopausal women. Therefore, it is important to evaluate the endometrium in PCOS patients with ultrasonography or hysteroscopy, and active management involving hysteroscopic polypectomy should be offered to PCOS patients diagnosed with endometrial polyps regardless of symptoms. ## Introduction Endometrial polyps are localized overgrowths of endometrial tissue that form finger-like projections from the surface of the endometrium. They consist of stroma, glands, and blood vessels and can be single or multiple [1–3]. According to previous studies, the prevalence of endometrial polyps in women of all age groups ranges from 10 to $40\%$ [4, 5]. Endometrial polyps may be asymptomatic, and the most common symptoms include excessive leukorrhea, abnormal uterine bleeding, and infertility [2, 6]. Most endometrial polyps are benign lesions, but approximately 3–$5\%$ of endometrial polyps have been reported as premalignant or malignant [7–9]. However, the factors associated with premalignant and malignant changes in endometrial polyps are not completely understood. Polycystic ovary syndrome (PCOS) is a common reproductive endocrine disease in women of reproductive age. It has been reported that $5\%$ to $10\%$ of women of reproductive age are diagnosed with PCOS worldwide, and the symptoms of PCOS include amenorrhea, oligomenorrhea, hirsutism, obesity, infertility, and acne [10–13]. PCOS is characterized by anovulation, insufficient progesterone, hyperandrogenism, and insulin resistance. These factors can disrupt the endometrium of PCOS patients and can lead to chronic low-grade inflammation in the endometrium, infertility, endometrial hyperplasia, or even endometrial cancer [13–16]. The risk of endometrial cancer has been shown to be between 2–6 times higher in women with PCOS than in women without PCOS [13]. The pathogenesis of endometrial cancer in PCOS patients is thought to be related to the prolonged stimulation of the endometrium by unopposed oestrogen in the setting of anovulation and prevention of endometrial exfoliation [17]. Thus, it is imperative to determine whether PCOS patients diagnosed with endometrial polyps are at greater risk of developing cancer so that a more judicious indication regarding hysteroscopic polypectomy can be established. The aim of the current study was to investigate the prevalence of premalignant and malignant endometrial polyps that were removed by hysteroscopy in premenopausal women and to further explore whether PCOS is associated with premalignant and malignant changes in endometrial polyps. ## Study design and data collection This retrospective study was conducted by searching the medical record database for subjects diagnosed with endometrial polyps that were removed by hysteroscopic polypectomy at the First Affiliated Hospital of Chengdu Medical College between January 2015 and December 2021. This hospital is one of the largest tertiary hospitals situated in the megacity of Chengdu in Southwest China. The study complied with the Declaration of Helsinki and was approved by the Ethics Committee of The First Affiliated Hospital of Chengdu Medical College (No. CYFY17143031). Informed consent was obtained from all participants before the study when they visited the outpatient department. The study population was selected based on the following inclusion and exclusion criteria. The inclusion criteria were as follows: [1] women who were diagnosed with endometrial polyps either by transvaginal ultrasonography or by hysteroscopy; and [2] premenopausal women. The exclusion criteria were as follows: [1] histopathology results consistent with submucosal uterine leiomyomas; [2] macroscopic malignancy; [3] incomplete information; [4] presence of intrauterine contraceptive device use; and [5] cases where polypectomy was not performed. A total of 4236 women who met the inclusion and exclusion criteria were included in the study. A flowchart of the study design is shown in Fig. 1.Fig. 1Flowchart of the study design The demographic and clinical characteristics were collected by retrieving information from the medical record database. These included age, height, weight, parity, history of hormone replacement therapy, oral contraceptives, abnormal uterine bleeding, hypertension, diabetes mellitus, PCOS, number of polyps, and size of polyps. The diagnosis of PCOS was based on the *Rotterdam criteria* that two of the following three criteria should be met: [1] clinical or biochemical hyperandrogenism; [2] ovulatory dysfunction; and [3] polycystic ovaries [18]. The size of the polyps was evaluated by measuring the maximum diameter for a single polyp or the largest polyp in the presence of multiple polyps. The indications for hysteroscopy were satisfied by women with abnormal uterine bleeding and asymptomatic women in whom the intrauterine lesions were incidental findings during other imaging scans. All the women included in the study had their endometrial polyps removed by hysteroscopic polypectomy. Endometrial assessment was also performed by endometrial aspiration or dilatation and curettage after polypectomy in the same setting. The endometrial polyp specimens and endometrium samples were placed in separate sample bags for histologic examination. In this study, the histologic findings were from the endometrial polyp specimens. The histologic findings of endometrial polyp specimens were defined as endometrial polyp, hyperplasic (simple or complex hyperplasia without atypia), premalignant (hyperplasia with atypia), and malignant (endometrial carcinoma) according to the World Health Organization (WHO) Classification Systems for Endometrial Hyperplasia [19]. Based on the Chinese National Health Commission, obesity was defined as BMI ≥ 28.0 kg/m2 in adults [20]. ## Statistical analysis SPSS software (version 16.0, Chicago, IL, USA) was used to analyse statistical significance. Data with normal distributions are presented as the mean ± SD, and nonnormally distributed data are presented as frequencies. The independent-sample t test or Mann‒Whitney U test was used for continuous variables, and the chi-square test or Fisher’s exact test was used for categorical variables. The association between the premalignant and malignant changes in endometrial polyps and PCOS was determined by multivariate logistic regression analysis after adjustment for potential confounding factors. The results are presented as the odds ratio (OR) and $95\%$ confidence interval (CI). A $p \leq 0.05$ was considered indicative of statistical significance. The sample size was calculated using the following formula as described in previous studies: n = Z2 × P × (1-P)/e2, where n = the required sample size, $Z = 1.96$ at a $95\%$ CI, P = the prevalence of premalignant and malignant polyps ($3\%$–$5\%$) and e = the margin of error ($5\%$) [21, 22]. PASS (Power Analysis and Sample Size) software was used to evaluate the statistical power and effect size. ## Histologic findings regarding the endometrial polyps removed from the study population A total of 4236 women who met the inclusion and exclusion criteria were included in this study. The demographic and clinical characteristics of the study population are shown in Table 1. Histological examination of the removed endometrial polyps showed that the prevalence of the premalignant and malignant polyps was $2.15\%$. Based on the histologic findings, the specimens were classified into a benign polyp group and a premalignant and malignant polyp group (Table 2). In the benign polyp group, 3057 ($72.17\%$) cases were diagnosed as endometrial polyps, 732 ($17.28\%$) cases were diagnosed as simple hyperplasia without atypia, and 356 ($8.4\%$) cases were diagnosed as complex hyperplasia without atypia. In the premalignant and malignant polyp group, 48 ($1.13\%$) cases were diagnosed as hyperplasia with atypia, and 43 ($1.02\%$) cases were diagnosed as endometrial carcinoma (Table 2).Table 1The demographic and clinical characteristics of the study population ($$n = 4236$$)ParametersValueAge (year)42 ± 7.5BMI (kg/m2)26.53 ± 4.3Menopausal status premenopausal4236 postmenopausal0Gravidity 0534 ($12.61\%$) ≥ 13702 ($87.39\%$)Parity 0704 ($16.62\%$) ≥ 13532 ($83.38\%$)Abnormal uterine bleeding Yes1045 ($24.67\%$) No3191 ($75.33\%$)Hypertension Yes516 ($12.18\%$) No3720 ($87.82\%$)*Diabetes mellitus* Yes749 ($17.68\%$) No3487 ($82.32\%$)PCOS Yes337 ($7.96\%$) No3899 (92.04)Hormone replacement therapy Yes168 ($3.97\%$) No4068 ($96.03\%$)Oral contraceptives Yes501 ($11.83\%$) No3735 ($88.17\%$)Polyp number 13012 ($71.1\%$) ≥ 21224 ($28.9\%$)Polyp size (cm) < 22994 ($70.68\%$) ≥ 21242 ($29.32\%$)Table 2Histologic findings regarding the endometrial polyp removed from the study population ($$n = 4236$$)Histology categoryFrequency (%)Benign polyp4145 (97.85) Endometrial polyps3057 (72.17) *Simple hyperplasia* without atypia732 (17.28) *Complex hyperplasia* without atypia356 (8.4)(Pre)malignant polyp91 (2.15) Hyperplasia with atypia48 (1.13) Endometrial carcinoma43 (1.02) ## Univariate analysis of risk factors for premalignant and malignant endometrial polyps Univariate analysis was performed to identify risk factors for premalignant and malignant endometrial polyps. The results showed that age (≥ 40 years), obesity, nulliparity, diabetes mellitus, PCOS, and polyp number were significantly associated with a higher risk of premalignant and malignant polyps ($p \leq 0.05$) (Table 3). The ORs ($95\%$ CIs) for age (≥ 40 years), obesity, nulliparity, diabetes mellitus, PCOS, and polyp number were 1.74 (1.12–2.71), 1.64 (1.07–2.53), 2.8 (1.81–4.35), 3.3 (2.16–5.06), 2.96 (1.74–5.02), and 1.71 (1.12–2.61), respectively. However, no statistically significant associations were observed between the premalignant and malignant polyps and other variables, including abnormal uterine bleeding, gravidity, hypertension, history of hormone replacement therapy, oral contraceptives, and the size of polyps ($p \leq 0.05$) (Table 3).Table 3Univariate analysis of risk factors for premalignant and malignant endometrial polyps ($$n = 4236$$)Variable(Pre)Malignant polypsn = 91Benign polypsn = 4145OR$95\%$ CIp-valueAge1.741.12–2.710.017 ≥ 40612231 < 40301914Obesity (BMI ≥ 28.0 kg/m2)1.641.07–2.530.03 Yes582143 No332002Gravidity 0175171.610.94–2.750.108 ≥ 1743628Parity2.81.81–4.35<0.0001 032672 ≥ 1593473Abnormal uterine bleeding1.370.87–2.150.172 Yes281017 No633128Hypertension1.3190.74–2.350.434 Yes14502 No773643Diabetes mellitus3.302.16–5.06<0.0001 Yes37712 No543433PCOS2.961.74–5.02<0.0001 Yes18319 No733826Hormone replacement therapy2.060.94–4.530.117 Yes7161 No843984Oral contraceptives1.490.85–2.610.220 Yes15486 No763659Polyp number1.711.12–2.610.012 2 ≥ 371187 1542958Polyp size (cm)0.660.43–1.010.069 2 ≤ 562938 > 2351207 ## Multivariate analysis of risk factors for premalignant and malignant endometrial polyps The risk factors that were identified by univariate analysis were further evaluated by a multivariate logistic regression analysis model. The results showed that only PCOS was significantly associated with a higher risk of premalignant and malignant endometrial polyps after adjustment for confounding factors, including age, obesity, parity, diabetes mellitus, and polyp number; the OR ($95\%$ CI) was 2.75 (1.02–3.45) (Table 4). This finding suggested that PCOS was an independent risk factor for premalignant and malignant endometrial polyps. The PCOS factor was also used to build a risk prediction model for premalignant and malignant polyps. An ROC curve was constructed, and the AUC was 0.746 with a sensitivity of $72.3\%$ and a specificity of $81.4\%$.Table 4Multivariate analysis of risk factors for premalignant and malignant endometrial polypsVariablesOR$95\%$ CIp-valueAge1.320.95—2.070.113Obesity2.451.02—3.410.185Parity3.232.15—4.270.089Diabetes mellitus4.223.02—5.870.221Polyp number1.230.35—2.170.217PCOS2.751.02—3.45 < 0.001Adjustment for age, obesity, parity, diabetes mellitus, and polyp number ## The association between PCOS and different histological types of endometrial polyps The prevalence of PCOS was evaluated in different histological types of endometrial polyps, and the results showed that the prevalence of PCOS was $8.01\%$ in endometrial polyps, $9.02\%$ in simple hyperplasia without atypia, $8.71\%$ in complex hyperplasia without atypia, $25\%$ in hyperplasia with atypia, and $23.26\%$ in endometrial carcinoma (Fig. 2). The association between PCOS and different histological types of endometrial polyps was further evaluated by logistic regression analysis (Table 5). The results showed that PCOS was significantly associated with a higher risk of hyperplasia with atypia and endometrial carcinoma; the ORs ($95\%$ CIs) were 3.96 (2.04–7.69) and 3.58 (1.75–7.34), respectively (Table 5). However, no associations were observed between PCOS and endometrial polyps, simple hyperplasia without atypia, and complex hyperplasia without atypia ($p \leq 0.05$).Fig. 2The prevalence of PCOS in different histological types of endometrial polypsTable 5The association between PCOS and different histological types of endometrial polyps ($$n = 4236$$)Histology categoryPCOS($$n = 337$$)Non-PCOS($$n = 3899$$)OR$95\%$CIp-valueEndometrial polyps ($$n = 3057$$)24528121.030.80–1.320.87Simple hyperplasia without atypia ($$n = 732$$)666661.180.89–1.570.275Complex hyperplasia without atypia ($$n = 356$$)313251.110.757–1.640.656Hyperplasia with atypia ($$n = 48$$)12363.962.04–7.69 < 0.0001Endometrial carcinoma ($$n = 43$$)10333.581.75–7.340.0006 ## Discussion Our study revealed that the prevalence of premalignant and malignant endometrial polyps was $2.15\%$, comprising rates of $1.13\%$ for hyperplasia with atypia and $1.02\%$ for endometrial carcinoma, among the polyps removed by hysteroscopic polypectomy in premenopausal women (Table 2). Moreover, PCOS was found to be associated with a higher risk of premalignant and malignant endometrial polyps in premenopausal women after adjustment for potential confounding factors (Table 4). These findings may provide guidance for clinical practice in the management of endometrial polyps among premenopausal women with PCOS. A meta-analysis study that included 35,345 premenopausal and postmenopausal women showed that the prevalence of premalignant and malignant endometrial polyps was $2.73\%$ [8]. Another meta-analysis study involved recruitment of 21,057 patients and reported that $3.4\%$ of patients presented with premalignant and malignant endometrial polyps [9]. However, our study revealed that the prevalence of premalignant endometrial polyps was $1.13\%$, and the prevalence of malignant endometrial polyps was $1.02\%$ (Table 2). The discrepancies could be attributed to differences in the characteristics of the study populations; these other studies included premenopausal women as well as postmenopausal women. However, only premenopausal women were included in our study, which may explain the lower prevalence in our population compared to that in the populations of these other studies when considering menopause as a risk factor for malignant changes in endometrial polyps [4]. Definition of the standards of obesity may also be an important factor that contributes to the difference in the prevalence of premalignant and malignant polyps between our population and others. We used the *Chinese criteria* BMI ≥ 28.0 kg/m2 to define obesity, while the other studies used the World Health Organization recommended criteria BMI ≥ 30.0 kg/m2 to define obesity [20]. Another explanation for the discrepancies might be attributed to the different methods used for the removal of polyps. Some studies performed the removal of polyps by uterine curettage, which usually fails to extract the whole polyp and obtains only a mixed specimen of polyps and endometrial mucosae. This may result in inconsistency of histological diagnosis. Our study involved performing the removal of polyps by hysteroscopic polypectomy, which is a more reliable technique for removing the entire polyp and allowing a more complete histological examination. In the univariate analysis, we found that age (≥ 40 years), obesity, nulliparity, diabetes mellitus, PCOS, and polyp number were significantly associated with the risk of premalignant and malignant polyps in premenopausal women. However, analysed by a multivariate logistic regression model, only PCOS was found to be significantly associated with the risk of premalignant and malignant polyps when potential confounding factors were controlled (Table 4). The prevalence of hyperplasia with atypia and endometrial carcinoma was 3.96 times and 3.58 times greater in women with PCOS than in those without PCOS (Table 5). This may be explained by the operation of a mechanism that involves endocrinologic and metabolic disorders in PCOS, specifically chronic anovulation, hyperandrogenism, and insulin resistance [17]. Chronic anovulation results in endometrial proliferation by long-term exposure to oestrogen without the opposing action of progesterone [23, 24]. Androgens can convert to oestrogens and indirectly stimulate endometrial proliferation [25]. Insulin resistance is accompanied by hyperinsulinism, which increases the levels of free androgen in the plasma by reducing the production of sex hormone-binding globulin, and high levels of androgen and insulin in the plasma can affect endometrial cell differentiation [26, 27]. In the present study, the prevalence of premalignant and malignant polyps was not associated with the presence of abnormal uterine bleeding. However, other studies have reported that abnormal uterine bleeding is associated with an increased risk of malignant polyps in postmenopausal women [28]. The discrepancy may be attributed to our study including only premenopausal women. It may also be explained by the fact that some small premalignant and malignant lesions existed on the surface or inside the polyps that were detected only by the pathological examinations and did not cause abnormal uterine bleeding. Therefore, when women are diagnosed with endometrial polyps, we propose that active management involving hysteroscopic polypectomy should be offered to premenopausal women with PCOS regardless of symptoms and to postmenopausal women with symptoms. Moreover, although screening for premalignant and malignant polyps with hysteroscopy is not recommended when considering the invasive nature of the procedure and the economic costs, we recommend routine pelvic ultrasonography for premenopausal women as well as for postmenopausal women regardless of symptoms. Evidence-based management of incidental ultrasound findings such as endometrial polyps has gained importance for modern gynaecologists. Therefore, it will be of great value to understand the significance of both symptomatic and asymptomatic endometrial polyps and their proposed management. Our study has many strengths. First, this study included a large sample size, and polyp removal was performed by a standardized procedure. Second, potential confounding factors were controlled in our evaluation of the association between PCOS and the premalignant and malignant endometrial polyps. Third, this study focused on premenopausal women because a considerable proportion of endometrial polyps are asymptomatic and are found incidentally in premenopausal women. Nevertheless, this study also has limitations. First, this study was conducted by a retrospective review of patient data and is subject to potential selection bias. Second, although a variety of potential confounding factors were controlled when we evaluated the association between PCOS and the premalignant and malignant endometrial polyps in premenopausal women, we cannot rule out the effect of any residual confounding factors on the findings. Third, the number of PCOS cases may have been underestimated in this study because we did not routinely screen PCOS in all outpatients. Finally, since this is a single-centre study, further multicentre studies are needed to confirm our findings. ## Conclusions PCOS is associated with a higher risk of premalignant and malignant endometrial polyps in premenopausal women. Therefore, it is necessary to evaluate the endometrium in PCOS patients with ultrasonography or hysteroscopy, and active management involving hysteroscopic polypectomy should be offered to PCOS patients diagnosed with endometrial polyps regardless of symptoms. ## References 1. Vitale SG, Haimovich S, Lagana AS. **Endometrial polyps. An evidence-based diagnosis and management guide**. *Eur J Obstet Gynecol Reprod Biol.* (2021) **260** 70-77. DOI: 10.1016/j.ejogrb.2021.03.017 2. Raz N, Feinmesser L, Moore O. **Endometrial polyps: diagnosis and treatment options - a review of literature**. *Minim Invasive Ther Allied Technol* (2021) **30** 278-287. DOI: 10.1080/13645706.2021.1948867 3. Ludwin A, Lindheim SR, Booth R. **Removal of uterine polyps: clinical management and surgical approach**. *Climacteric* (2020) **23** 388-396. DOI: 10.1080/13697137.2020.1784870 4. Karakas LA, Atilgan AO, Akilli H. **Nulliparity and postmenopausal status are independent factors of malignancy potential of endometrial intraepithelial neoplasia in polyps**. *Int J Gynaecol Obstet* (2021) **152** 433-438. DOI: 10.1002/ijgo.13448 5. Lieng M, Istre O, Qvigstad E. **Treatment of endometrial polyps: a systematic review**. *Acta Obstet Gynecol Scand* (2010) **89** 992-1002. DOI: 10.3109/00016349.2010.493196 6. Clark TJ, Stevenson H. **Endometrial Polyps and Abnormal Uterine Bleeding (AUB-P): What is the relationship, how are they diagnosed and how are they treated?**. *Best Pract Res Clin Obstet Gynaecol* (2017) **40** 89-104. DOI: 10.1016/j.bpobgyn.2016.09.005 7. Wong CLH, So PL. **Prevalence and risk factors for malignancy in hysteroscopy-resected endometrial polyps**. *Int J Gynaecol Obstet* (2021) **155** 433-441. DOI: 10.1002/ijgo.13656 8. Uglietti A, Buggio L, Farella M. **The risk of malignancy in uterine polyps: A systematic review and meta-analysis**. *Eur J Obstet Gynecol Reprod Biol* (2019) **237** 48-56. DOI: 10.1016/j.ejogrb.2019.04.009 9. Sasaki LMP, Andrade KRC, Figueiredo ACMG. **Factors Associated with Malignancy in Hysteroscopically Resected Endometrial Polyps: A Systematic Review and Meta-Analysis**. *J Minim Invasive Gynecol.* (2018) **25** 777-785. DOI: 10.1016/j.jmig.2018.02.004 10. Zhao H, Zhao Y, Ren Y. **Epigenetic regulation of an adverse metabolic phenotype in polycystic ovary syndrome: the impact of the leukocyte methylation of PPARGC1A promoter**. *Fertil Steril.* (2017) **107** 467-474 e5. DOI: 10.1016/j.fertnstert.2016.10.039 11. Goodarzi MO, Dumesic DA, Chazenbalk G. **Polycystic ovary syndrome: etiology, pathogenesis and diagnosis**. *Nat Rev Endocrinol* (2011) **7** 219-231. DOI: 10.1038/nrendo.2010.217 12. Taghavi SA, Bazarganipour F, Hugh-Jones S. **Health-related quality of life in Iranian women with polycystic ovary syndrome: a qualitative study**. *BMC Womens Health* (2015) **30** 111. DOI: 10.1186/s12905-015-0272-4 13. Teede HJ, Misso ML, Costello MF. **Recommendations from the international evidence-based guideline for the assessment and management of polycystic ovary syndrome**. *Fertil Steril* (2018) **110** 364-379. DOI: 10.1016/j.fertnstert.2018.05.004 14. Palomba S, Daolio J, La Sala GB. **Oocyte Competence in Women with Polycystic Ovary Syndrome**. *Trends Endocrinol Metab* (2017) **28** 186-198. DOI: 10.1016/j.tem.2016.11.008 15. Palomba S. **Is fertility reduced in ovulatory women with polycystic ovary syndrome? An opinion paper**. *Hum Reprod* (2021) **36** 2421-2428. DOI: 10.1093/humrep/deab181 16. Palomba S, de Wilde MA, Falbo A. **Pregnancy complications in women with polycystic ovary syndrome**. *Hum Reprod Update.* (2015) **21** 575-92. DOI: 10.1093/humupd/dmv029 17. Tokmak A, Kokanali MK, Guzel AI. **Polycystic ovary syndrome and risk of endometrial cancer: a mini-review**. *Asian Pac J Cancer Prev* (2014) **15** 7011-7014. DOI: 10.7314/APJCP.2014.15.17.7011 18. **Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome (PCOS)**. *Hum Reprod* (2004) **19** 41-47. DOI: 10.1093/humrep/deh098 19. Mills AM, Longacre TA. **Endometrial hyperplasia**. *Semin Diagn Pathol* (2010) **27** 199-214. DOI: 10.1053/j.semdp.2010.09.002 20. He W, Li Q, Yang M. **Lower BMI cutoffs to define overweight and obesity in China**. *Obesity (Silver Spring)* (2015) **23** 684-691. DOI: 10.1002/oby.20995 21. Pourhoseingholi MA, Vahedi M, Rahimzadeh M. **Sample size calculation in medical studies**. *Gastroenterol Hepatol Bed Bench.* (2013) **6** 14-7. PMID: 24834239 22. Serdar CC, Cihan M, Yucel D. **Sample size, power and effect size revisited: simplified and practical approaches in pre-clinical, clinical and laboratory studies**. *Biochem Med (Zagreb)* (2021) **31** 010502. DOI: 10.11613/BM.2021.010502 23. Xue Z, Li J, Feng J. **Research Progress on the Mechanism Between Polycystic Ovary Syndrome and Abnormal Endometrium**. *Front Physiol* (2021) **12** 788772. DOI: 10.3389/fphys.2021.788772 24. Palomba S, Piltonen TT, Giudice LC. **Endometrial function in women with polycystic ovary syndrome: a comprehensive review**. *Hum Reprod Update* (2021) **27** 584-618. DOI: 10.1093/humupd/dmaa051 25. Shafiee MN, Chapman C, Barrett D. **Reviewing the molecular mechanisms which increase endometrial cancer (EC) risk in women with polycystic ovarian syndrome (PCOS): time for paradigm shift?**. *Gynecol Oncol* (2013) **131** 489-492. DOI: 10.1016/j.ygyno.2013.06.032 26. McCartney CR, Marshall JC. **CLINICAL PRACTICE. Polycystic Ovary Syndrome**. *N Engl J Med.* (2016) **375** 54-64. DOI: 10.1056/NEJMcp1514916 27. Shafiee MN, Khan G, Ariffin R. **Preventing endometrial cancer risk in polycystic ovarian syndrome (PCOS) women: could metformin help?**. *Gynecol Oncol* (2014) **132** 248-253. DOI: 10.1016/j.ygyno.2013.10.028 28. Bel S, Billard C, Godet J. **Risk of malignancy on suspicion of polyps in menopausal women**. *Eur J Obstet Gynecol Reprod Biol* (2017) **216** 138-142. DOI: 10.1016/j.ejogrb.2017.07.013
--- title: NMR derived changes of lipoprotein particle concentrations related to impaired fasting glucose, impaired glucose tolerance, or manifest type 2 diabetes mellitus authors: - Tina Kalbitzer - Kristina Lobenhofer - Silke Martin - Markus Beck Erlach - Werner Kremer - Hans Robert Kalbitzer journal: Lipids in Health and Disease year: 2023 pmcid: PMC10037821 doi: 10.1186/s12944-023-01801-7 license: CC BY 4.0 --- # NMR derived changes of lipoprotein particle concentrations related to impaired fasting glucose, impaired glucose tolerance, or manifest type 2 diabetes mellitus ## Abstract ### Background Type 2 diabetes mellitus (T2D) and corresponding borderline states, impaired fasting glucose (IFG) and/or glucose tolerance (IGT), are associated with dyslipoproteinemia. It is important to distinguish between factors that cause T2D and that are the direct result of T2D. ### Methods The lipoprotein subclass patterns of blood donors with IFG, IGT, with IFG combined with IGT, and T2D are analyzed by nuclear magnetic resonance (NMR) spectroscopy. The development of lipoprotein patterns with time is investigated by using samples retained for an average period of 6 years. In total 595 blood donors are classified by oral glucose tolerance test (oGTT) and their glycosylated hemoglobin (HbA1c) concentrations. Concentrations of lipoprotein particles of 15 different subclasses are analyzed in the 10,921 NMR spectra recorded under fasting and non-fasting conditions. The subjects are assumed healthy according to the strict regulations for blood donors before performing the oGTT. ### Results Under fasting conditions manifest T2D exhibits a significant concentration increase of the smallest HDL particles (HDL A) combined with a decrease in all other HDL subclasses. In contrast to other studies reviewed in this paper, a general concentration decrease of all LDL particles is observed that is most prominent for the smallest LDL particles (LDL A). Under normal nutritional conditions a large, significant increase of the concentrations of VLDL and chylomicrons is observed for all groups with IFG and/or IGT and most prominently for manifest T2D. As we show it is possible to obtain an estimate of the concentrations of the apolipoproteins Apo-A1, Apo-B100, and Apo-B48 from the NMR data. In the actual study cohort, under fasting conditions the concentrations of the lipoproteins are not increased significantly in T2D, under non-fasting conditions only Apo-B48 increases significantly. ### Conclusion In contrast to other studies, in our cohort of “healthy” blood donors the T2D associated dyslipoproteinemia does not change the total concentrations of the lipoprotein particles produced in the liver under fasting and non-fasting conditions significantly but only their subclass distributions. Compared to the control group, under non-fasting conditions participants with IGT and IFG or T2D show a substantial increase of plasma concentrations of those lipoproteins that are produced in the intestinal tract. The intestinal insulin resistance becomes strongly observable. ## Background Metabolic syndrome, insulin resistance, and type-2 diabetes are associated with a typical dyslipoproteinemia. Dyslipoproteinemia can be studied elegantly by high resolution 1H-NMR spectroscopy as shown initially by the groups of Otvos [1, 2] and Ala-Korpela [3] in this field. An important feature of NMR spectroscopy is the possibility of classifying subgroups of lipoproteins by their size under high-throughput conditions, giving additional information on the size distribution and corresponding particle numbers. Whereas initially NMR analytics was based solely on a chemical shift analysis of the lipoprotein spectra, later the method was improved by additionally using diffusion effects measured by pulsed magnetic field gradients [4, 5]. Since the NMR visibility of different lipoproteins varies strongly in different lipoprotein classes [6], the data evaluation procedure has to be calibrated carefully by the gold standard method analytical ultracentrifugation. However, the latter method is not suitable for large scale studies, therefore, NMR spectroscopy was mainly used in the past for characterizing changes in lipoprotein particle patterns caused by different forms of prediabetes (defined by NIDKK as impaired fasting glucose (IFG) and/or impaired glucose tolerance (IGT)) and manifest type-2 diabetes (T2D) itself [7–14]. In a prospective study by Festa et al. [ 7] the lipoprotein particle sizes and concentrations were determined by NMR spectroscopy with an average follow-up time of 5.2 years. Increased concentrations of small HDL and large VLDL were positively associated with an increased risk for the development of type-2 diabetes. Another follow-up study of 13 years by Mora et al. [ 9] showed that increased concentrations of small HDL, small LDL, and large VLDL particles were predictive for a higher risk to develop type-2 diabetes. In a multi-ethnic study of atherosclerosis by Mackey et al. [ 11], besides changes in lipid concentrations, increased concentrations of VLDL were found to be associated with the development of diabetes mellitus. Wang et al. [ 10] analyzed the lipoprotein concentrations from male Finnish individuals in native blood serum. It is the only NMR based study where an accurate metabolic classification by oral glucose tolerance test (oGTT) has been performed. The participants were assigned to five classes, non-diabetic participants, participants with impaired fasting glucose (IFG), impaired glucose tolerance (IGT), IFG combined with IGT, and with newly diagnosed type-2 diabetes (T2D). Increased VLDL concentrations were associated with abnormal glucose tolerance as found in IGT and T2D. Decrease of large HDL and increase of small HDL concentrations were consistently observed for individuals with abnormal fasting glucose (IFG and T2D). Sokooti et al. [ 12] compared the HDL-particle distribution of the post transplantation diabetes mellitus with type-2 diabetes mellitus from other sources. They found the risk to develop type-2 diabetes is decreased, when larger HDL particles prevail. The same was found for the risk to develop T2D for non-transplanted subjects [13]. Tranes et al. [ 14] studied a small group of lean Chinese with and without insulin resistance and found no differences of the lipoprotein subclass distributions in the two groups. They concluded that mechanistically there is a dissociation between the insulin resistance at the level of glucose metabolism (impaired glucose tolerance) and the dyslipoproteinemia usually described in type 2 diabetes mellitus. In the present study we focus on a different population, German long time blood donors that are assumed to be healthy according to the rules applying for blood donors. IFG, IGT or manifest T2D were not known for this group before including them in this study. The metabolic state was determined by the “gold standard” oGTT. Based on the oGTT the participants were divided in five metabolic groups, the healthy control group and the four groups with different disorders of the glucose metabolism (impaired fasting glucose, impaired glucose tolerance, umpaired fasting glucose combined with impaired glucose tolerance, manifest type 2 diabetes mellitus). The differences in the lipoprotein particle numbers and concentrations in five different groups were analysed. In addition, we could retrospectively follow the changes in the lipoprotein parameters in retained samples taken at different times in an average period of 6 years before testing. ## Study design The participants of this study were preselected from the pool of blood donors of the Bavarian Red Cross (BRK) by sending the FindRisk questionnaires [15] to 60,000 individuals of this group. 51,021 correctly filled out FindRisk forms were returned. The FindRisk score questionnaire predicts the risk to develop type-2 diabetes within the next 10 years. The concentration of glycosylated hemoglobin HbA1c was determined for 12,773 of these blood donors. On the basis of their FindRisk score and their HbA1c concentration, 4017 persons were invited to the oGTT. Six hundred seventy-one persons accepted the invitation to perform an oral glucose tolerance test (oGTT). Finally, 595 persons fulfilled all criteria for the inclusion in the present diabetes NMR study (Table 1). In addition to the general, quite strict exclusion criteria for blood donors, the following criteria led to an exclusion from the study: [1] previously known perturbations of the glucose metabolism (treated and untreated), [2] medication known to influence the lipid metabolism, [3] failures in the preparation phase of the oGTT. The participants had to adhere to a normal diet rich in carbohydrates (> 150 g/day) at least 3 days prior to testing, a fasting period of 8–11 h prior to testing, and [4] the capillary glucose concentration immediately before oGTT had to be less than 150 mg/mL. The participants were distributed into 5 classes, a control group, groups of individuals with impaired fasting glucose (IFG), with impaired glucose tolerance (IGT), with impaired fasting glucose combined with impaired glucose tolerance (IFG + IGT), and with newly diagnosed type-2 diabetes (T2D). As control group served blood donors with no signs of diabetes. They were selected on the basis of their normal HbA1c concentrations and their normal oGTT values. Fasting EDTA plasma samples before the start of the oGGT test and 120 min after oral administration of 75 g glucose were immediately frozen at -80 C. For most of these candidates, reserve EDTA plasma samples stored at -80 C were available for NMR-spectroscopy from the BIOBANK of the BRK (www.biobank.de). According to the American Diabetes Association (ADA) we define IFG by venous plasma glucose concentrations c($t = 0$) ≥ 100 mg/dL and ≤ 125 mg/dL and c($t = 120$) ≥ 140 mg/dL and < 200 mg/dL, IGT by c($t = 0$) < 100 mg/dL and c($t = 120$) ≥ 140 mg/dL and < 200 mg/dL, T2D by c($t = 0$) > 125 mg/dL or c($t = 120$) ≥ 200 mg/dL, healthy by c($t = 0$) < 100 mg/dL and c($t = 120$) < 140 mg/dL. The majority of Bavarian blood donors have Caucasian ethnicity. Table 1Description of the study cohortaoGTT groupAll ($$n = 595$$)Control ($$n = 251$$)IFG ($$n = 193$$)IGT ($$n = 26$$)IFG + IGT ($$n = 68$$)T2D ($$n = 57$$)Male380140130184547Female2151116382310Age [years]54.3 ± 8.952.8 ± 10.154.6 ± 7.957.2 ± 7.756 ± 8.156.4 ± 7.0BMI [kg/m2]28.8 ± 4.127.9 ± 4.029.0 ± 3.829.4 ± 3.530.3 ± 4.330.4 ± 4.3HbA1c [%]5.9 ± 0.45.8 ± 0.25.9 ± 0.35.9 ± 0.46 ± 0.36.3 ± 0.7FindRisk Score [a.u]12.5 ± 3.611.4 ± 3.812.9 ± 3.312.7 ± 2.814 ± 3.013.7 ± 3.4c ($t = 0$)a [mg/dL]103.4 ± 13.493.8 ± 4.7106.7 ± 5.695.2 ± 3.9109.9 ± 6.6130.3 ± 18.8c ($t = 120$)b [mg/dL]122.1 ± 43.6101.2 ± 20.6106.2 ± 20.3155.3 ± 14.6163.8 ± 17.2210.9 ± 61.4aConcentration of glucose in the oGTT at $t = 0$ minbConcentration of glucose in the oGTT at $t = 120$ min ## NMR-spectroscopy and primary data evaluation After thawing the frozen samples, 400 μL of EDTA-plasma were used for the NMR experiments performed at 600.2 MHz 1H-frequency with a Bruker Avance II NMR spectrometer. The samples were used without any addition of reagents. Three spectra were recorded for every sample at 310 K, one 1D NOESY spectrum (pulse program noesygppr1D) with a mixing time of 10 ms and a repetition time of 5.4 s, and two stimulated spin echo spectra (LED, pulse program ledbpgppr2s1d) using different gradient strengths. The total measuring time per sample was 6 min. The NMR spectra were evaluated using an adapted proprietary software (version 2011) from LipoFit GmbH, Regensburg, Germany, as disclosed in our published patents [4, 5]. Fifteen different subclasses of lipoproteins with varying mean diameters d were defined in the evaluation program: HDL A, 7.75 nm, HDL B, 9.25 nm, HDL C, 11.5 nm, HDL D, 14.5 nm, LDL A, 17.5 nm, LDL B, 20 nm, LDL C, 21.5 nm, LDL D, 23.5 nm, LDL E, 27.5 nm, IDL, 35 nm, VLDL A, 50 nm, VLDL B, 70 nm, chylomicron remnants (CM Re), 90 nm, small chylomicrons (CM A), 125 nm, and large chylomicrons (CM B) 375 nm. For additional information see Table 3. The intensities were corrected according to Baumstark et al. [ 6] for visibility effects at 310 K by multiplying the lipoprotein concentrations obtained from the direct integration by 1.00, 1.30, 1.11, 1.19, 1.27, and 1.36, VLDL (including IDL), LDL, HDL A, HDL B, HDL C, and HDL D, respectively. For chylomicrons correction factors have not been published and thus the correction factor was set to 1.0. The concentrations of Apo-A1, Apo-B100, and Apo-B48 were calculated from the particle concentrations of the corresponding particles determined by NMR. The Apo-A1 concentration was calculated from the total concentrations of the HDL particles, the Apo-B100 concentration from the total concentrations of the LDL and VLDL particles, and the Apo-B48 concentration from the total concentrations of chylomicron particles, assuming a stoichiometry of 2:1, 1:1, and 1:1, respectively. Since the apolipoprotein concentrations are indirectly derived from the NMR particle numbers, the suffix NMR is used for these concentrations in the following. ## Statistical evaluation Date were evaluated with the SPSS-software package, version 25.0 for windows (IBM) and the R-program, version 3.6.1 (R Core Team, 2019). The Kolmogorov–Smirnov test together with the Lilifors correction was used to test the normal distribution of data. The t-test was used to determine the significance of differences between classes when they were sufficiently well normally distributed. Otherwise the non-parametric Kruskal–Wallis H-test and the Mann–Whitney U-test were used. Particle concentrations at different times were fitted to a linear function of time. Time $t = 0$ is the extrapolation to the time of the oGTT test. The result of the oGGT was not included into the fit since it was recorded under fasting conditions. ## Ethical aspects The ethical aspects of the study were positively reviewed by the ethics commission of the Bayerische Landesärztekammer (#08,055) at July 29, 2008. They abide the Declaration of Helsinki. ## Description of the study cohort Table 1 summarizes the features of the study cohort, Fig. 1 represents the distribution of the participants to the different groups. The majority of Bavarian blood donors have Caucasian ethnicity. Overall, 215 females and 380 males were accepted in the study. The ratio of 0.56 approximates also the ratio of the two sexes in the cohort of blood donors. The BMIs in all groups are quite similar with an average of 28.8 kg m−2, compared to the control group of healthy participants (nonIGT + nonIFG + non2TD), the average value of diabetics is only higher by $9\%$ (Table 1).Fig. 1The study cohort. ( Top) Distribution of the different oGTT groups in the study. ( Bottom) Distribution of the frequency n of whole set of donors that gave the first sample –t years before the final oGTT. ( Blue) absolute frequency, (orange) cumulative frequency in % All participants were healthy according to the German strict rules for blood donors, diabetes was not known before performing the oGTT tests of the study. Especially, the subjects did not obtain treatment of possible perturbations of their glucose or lipid metabolism. Only persons of age 18 to 68 are accepted as blood donors and thus to the present study. The age distribution was also quite similar in all groups with an average age of 54.3 years and the mean values between the group of heathy subjects and subjects with T2D differ only by 3.5 years. Only small differences between the groups are observed for HbAc1 concentrations and the FindRisk score. Note that subjects with previously known impairment of the glucose metabolism were excluded from the study. In summary, the five groups of the study cohort are quite well matched with respect to age, BMI, and the general risk to develop T2D as predicted by the FindRisk score. Only for a part of the study participants (all of them were subjected to an oGTT under fasting conditions) suitable NMR samples directly frozen at -80 C at the oGTT were provided for this study (Table 1). That means that for some participants only reserve samples for the lipoprotein analytics by NMR were available. The cohort studied here is part of a larger cohort used to find out the correlation between the FindRisk score and the HbAc1 concentrations [16]. ## Estimation of apolipoprotein concentrations by NMR Generally, it is to be expected that the apolipoproteins directly involved in the recognition of their specific receptor have a fixed stoichiometry for a given type of lipoprotein. There is a good evidence that Apo-A1 is the characteristic apolipoprotein for HDL-particles, Apo-B100 for the non-HDL particles that is LDL and VLDL, and Apo-B48 for the chylomicrons and chylomicron remnants. As consequence, their concentrations should be proportional to the concentrations of their correspondent particles. The proportionality between the NMR derived particle numbers and the apolipoprotein concentrations has already shown experimentally for HDL and non-HDL particles in the serum [17, 18]. However, the exact stoichiometry is still under discussion. We assumed the most likely stoichiometry of 2:1, 1:1, and 1:1 for calculating the apolipoprotein concentrations of Apo-A1, Apo-B100, and Apo-B48 determined by NMR and accordingly call this value Apo-A1NMR, Apo-B100NMR, and Apo-B48NMR (see also Discussion). ## Lipoprotein particle concentrations related to fasting and normal nutritional conditions From all individuals of the study cohort oGGT test results were available (Table 1). However, not for all of them plasma samples for NMR analysis were taken during the oGTT test (time $t = 0$) that would reflect the NMR derived lipoprotein state under fasting conditions. For most of the participants reserve samples were taken before $t = 0$, when they donated blood, and were stored in the BioBank at -80 C. For more than $80\%$ of the participants of this study reserve samples older than 6 years were available, with some samples taken at t = -10 years (Fig. 1). From all samples NMR spectra were recorded, in total 10,921 sets of plasma NMR spectra were analyzed. The time dependence of most of the lipoprotein particle concentrations can be fitted sufficiently well with a linear relation. An example of a blood donor with freshly diagnosed T2D is shown in Fig. 2. The blood donors are asked to eat as every day before blood donation. Of course, their actual NMR lipoprotein profile determined will also depend on the time, when the plasma is taken after food ingestion and the food content itself. Therefore, larger fluctuation of the lipoprotein particles concentrations around the line of best fit are to be expected. The mean changes of particle concentrations (the slopes of the straight lines) are summarized in Table 2. For the control group, for the group with combined IFG and IGT, and for the T2D-group the slope is always positive, meaning that all concentrations increase with time. Only in the groups of participants diagnosed with IFG or IGT, a decrease of concentrations of some lipoproteins with time is observed. Especially, impaired glucose tolerance leads to a decrease of lipoprotein concentrations of almost all particle classes. The relative concentration increases per year are moderate (of the order of $1\%$). The spread between individuals is significantly larger (data not shown). However, always clear trends are observed with time, allowing to calculate the values expected at time $t = 0.$ They represent the average particle concentrations under “normal” nutritional conditions. The values for non-fasting conditions determined by the long-term fit of the data extrapolated to time $t = 0$ are probably more representative of the “normal” non-fasting state of the individuals, since this method decreases variations caused by isolated cases of extensive food consumption. Fig. 2Time dependence of the lipoprotein particle concentrations of an individual with newly diagnosed T2D. Time $t = 0$ is the time of the oGGT. A Concentrations of CM B, CM A, CM Re, B of VLDL B, VLDL A, IDL, C of LDL E, LDL D, LDL C, LDL B, LDL A, D of HDL B, HDL C, HDL D (E) of HDL ATable 2Time course of lipoprotein particle concentrations in the years before testingaControlIFGIGTIFG + IGTT2D<Δc/Δt >[nM/y][%/y][nM/y][%/y][nM/y][%/y][nM/y][%/y][nM/y][%/y]HDL A103.660.61-17.89-0.10178.851.0623.000.1383.220.46HDL B68.261.4888.331.96-98.19-2.19138.342.89111.332.54HDL C35.041.4642.711.88-38.69-1.6868.263.0246.362.22HDL D12.411.0811.681.01-0.37-0.0323.001.9715.701.37Apo-A1NMR109.680.2262.420.1220.810.04126.290.25128.300.25LDL A5.111.105.841.30-4.38-0.959.132.024.751.13LDL B3.650.992.560.70-2.56-0.686.211.673.290.92LDL C4.381.331.830.57-1.83-0.565.111.571.830.59LDL D1.460.650.000.00-1.46-0.641.460.650.370.17LDL E1.831.030.000.00-2.92-1.562.561.331.100.58IDL0.370.43-0.37-0.43-1.46-1.690.370.410.000.00VLDL A0.000.00-0.37-0.79-1.10-2.280.731.480.370.75VLDL B0.000.000.000.00-0.37-3.400.000.000.000.00Apo-B100NMR2.100.121.190.07-2.01-0.123.190.191.460.09CM Re0.000.000.000.000.000.000.000.000.000.00CM A0.000.000.000.000.000.000.000.000.000.00CM B0.000.000.000.000.000.000.000.000.000.00Apo-B48NMR0.000.000.000.000.000.000.000.000.000.00aData are taken from the reserve samples. Note that disorders of the glucose metabolism were newly diagnosed for all participants of this study and that values correspond to normal nutritional conditions. The particle concentrations c were linearly fitted as function of the time t before the oGTT. <Δc/Δt > is the mean slope of the line of best fit, either expressed in nM/year or %/year. The percents are related to the particle concentrations extrapolated to $t = 0.$ Number of participants 595. The Apo-A1NMR concentrations were calculated from the HDL concentrations assuming 2 Apo-A1 molecules /HDL-particle, the Apo-B100NMR concentrations were calculated as the sum of VLDL, IDL and LDL particle concentrations assuming 1 Apo-B100 molecule/particle, the Apo-B48 concentrations were calculated as the sum of all chylomicron particle concentrations assuming 1 Apo-B48 molecule/particle (see Discussion) Table 3 summarizes the particle concentrations in the five groups under fasting and average nutritional conditions. Fasting has only a small influence on the mean lipoprotein concentrations of the main lipoprotein classes for the control group of healthy volunteers. Even the concentrations of particles related directly to food intake (chylomicrons and chylomicron remnants) are not changed much by fasting in this group. This confirms that the food intake happened usually several hours before blood donation. Only the average volume of these particles is increased after food intake in the healthy control group indicating that more lipids are transported from the guts (Table 4). We observed somewhat larger changes in the subclass patterns of the control group by fasting with the largest relative decrease by -$2.2\%$ for the average values of the smallest HDL particles (HDL A) compensated by an increase of larger HDL particles (HDL C and HDL D). Much larger particle concentration changes induced by fasting are clearly observed for subjects with impaired glucose metabolism with the largest effects observed for manifest type 2 diabetes mellitus. It strongly reduces the general concentrations of lipoprotein particles of all classes. The largest changes are observed for the large particles (chylomicrons and VLDL) with a maximum effect found for large chylomicrons (CM B) with an increase by $22.7\%$. The average volumes and particle concentrations of chylomicrons increase by $6.8\%$ and $19.2\%$ under average nutritional conditions, respectively (Table 3).Table 3Comparison of lipoproteins concentrations obtained under fasting and non-fasting conditionsaControlIFGIGTIFG + IGTT2DNf196167212919Nnf251193266857LipoproteinFasting statec [nM]c [nM]c [nM]c [nM]c [nM]HDL AFasting16,628 ± 283916,885 ± 293916,725 ± 265817,107 ± 244517,408 ± 2776(7 – 8.5 nm)Non-fasting16,963 ± 332017,362 ± 277716,799 ± 242017,285 ± 225118,014 ± 2454Δcnf-f [%]22.80.413.5HDL BFasting4688 ± 14514247 ± 13314586 ± 16924724 ± 15304218 ± 1123(8.5 – 10 nm)Non-fasting4623 ± 14424501 ± 12674479 ± 17784779 ± 14934390 ± 1156Δcnf-f [%]-1.46-2.31.24.1HDL CFasting2445 ± 9032106 ± 6952309 ± 1086 2259 ± 6122004 ± 593(10 – 13 nm)Non-fasting2407 ± 8272270 ± 7492300 ± 10492263 ± 6142085 ± 565Δcnf-f [%]-1.67.8-0.40.24.1HDL DFasting1159 ± 2061111 ± 1701184 ± 2281174 ± 1831125 ± 147(13 – 16 nm)Non-fasting1154 ± 1931158 ± 1761193 ± 2351170 ± 1551149 ± 162Δcnf-f [%]-0.54.20.7-0.32.1HDL totalFasting24,921 ± 348824,349 ± 342324,805 ± 327225,265 ± 284024,754 ± 3003(7 – 16 nm)Non-fasting25,147 ± 387425,290 ± 323424,771 ± 310425,498 ± 294725,638 ± 3128Δcnf-f [%]0.93.9-0.10.93.6Apo-A1NMRbFasting49,842 ± 697648,698 ± 684549,611 ± 654350,529 ± 567949,508 ± 6006Non-fasting50,294 ± 774950,581 ± 646849,543 ± 620750,996 ± 589351,276 ± 6255Δcnf-f [%]0.93.9-0.10.93.6LDL AFasting469.2 ± 127.2426.4 ± 97.6469.6 ± 172.6455.1 ± 81.2395.1 ± 97.6(16 – 19 nm)Non-fasting465.5 ± 118.5449.9 ± 105.7461.4 ± 174.8452.0 ± 87.9420.7 ± 86.3Δcnf-f [%]-0.85.5-1.7-0.76.5LDL BFasting370.0 ± 74.3354.3 ± 61.0375.2 ± 85.6374.5 ± 59.8339.1 ± 59.0(19 – 21 nm)Non-fasting369.4 ± 72.2367.3 ± 66.0374.2 ± 89.8371.1 ± 57.4358.8 ± 59.1Δcnf-f [%]-0.23.7-0.3-0.95.8LDL CFasting328.5 ± 66.5314.9 ± 56.4331.2 ± 78.5328.6 ± 54.5296.2 ± 63.4(21 – 22 nm)Non-fasting329.9 ± 70.1323.0 ± 64.7326.7 ± 82.9324.6 ± 53.9311.5 ± 55.9Δcnf-f [%]0.42.6-1.4-1.25.1LDL DFasting221.3 ± 48.7216.9 ± 45.1225.0 ± 54.2229.0 ± 51.1205.1 ± 53.7(22 – 25 nm)Non-fasting223.6 ± 57.1223.1 ± 47.1227.5 ± 59.1224.0 ± 47.1219.2 ± 47.0Δcnf-f [%]12.81.1-2.26.9LDL EFasting175.6 ± 41.9180.3 ± 41.0185.4 ± 35.7196.6 ± 58.0169.8 ± 44.1(25 – 30 nm)Non-fasting177.6 ± 53.2183.2 ± 46.1187.6 ± 38.2191.5 ± 45.3188.8 ± 47.8Δcnf-f [%]1.11.61.2-2.611.2LDL totalFasting1564.7 ± 328.11492.9 ± 273.41586.4 ± 406.01583.7 ± 266.81405.2 ± 286.3(16 – 30 nm)Non-fasting1566.0 ± 336.11546.5 ± 301.01577.2 ± 423.91563.1 ± 262.41498.8 ± 271.7Δcnf-f [%]0.13.6-0.6-1.36.7IDLFasting83.41 ± 20.9584.43 ± 20.2783.41 ± 19.3191.8 ± 28.3778.71 ± 23.38(30 – 40 nm)Non-fasting84.87 ± 27.2485.67 ± 21.4886.58 ± 19.488.13 ± 21.9786.57 ± 22.98Δcnf-f [%]1.81.53.8-410VLDL AFasting42.93 ± 13.7645.08 ± 13.7945.22 ± 12.1851.88 ± 23.0943.17 ± 15.54(40 – 60 nm)Non-fasting43.33 ± 17.7445.98 ± 14.4248.03 ± 11.9349.4 ± 15.748.95 ± 16.87Δcnf-f [%]0.926.2-4.813.4VLDL BFasting9.92 ± 3.2210.32 ± 3.2610.1 ± 3.0312.28 ± 6.029.35 ± 3.23(60 – 80 nm)Non-fasting10.06 ± 4.3410.5 ± 3.4210.74 ± 2.6811.35 ± 3.6111.19 ± 4.11Δcnf-f [%]1.41.76.3-7.619.7IDL and VLDLFasting136.26 ± 37.08139.83 ± 36.48138.72 ± 33.32155.95 ± 56.55131.22 ± 40.77(30 – 80 nm)Non-fasting138.27 ± 48.51142.15 ± 38.65145.35 ± 32.75148.88 ± 40.54146.71 ± 43.34Δcnf-f [%]1.51.74.8-4.511.8Apo-B100NMRbFasting1700.9 ± 352.01632.7 ± 300.91725.1 ± 431.31739.6 ± 310.71536.4 ± 318.3Non-fasting1704.2 ± 372.01688.6 ± 329.41722.6 ± 449.11712.0 ± 294.31645.5 ± 306.3Δcnf-f [%]0.23.4-0.1-1.67.1CM ReFasting0.66 ± 0.280.7 ± 0.280.7 ± 0.250.86 ± 0.520.68 ± 0.33(80 – 100 nm)Non-fasting0.66 ± 0.350.73 ± 0.280.77 ± 0.230.8 ± 0.330.78 ± 0.34Δcnf-f [%]04.310-714.7CM AFasting0.42 ± 0.170.44 ± 0.180.43 ± 0.160.56 ± 0.350.4 ± 0.17(100 – 150 nm)Non-fasting0.42 ± 0.230.45 ± 0.180.47 ± 0.140.51 ± 0.20.5 ± 0.23Δcnf-f [%]02.39.3-8.925CM BFasting0.075 ± 0.0390.079 ± 0.040.076 ± 0.0410.109 ± 0.0830.066 ± 0.041(> 150 nm)Non-fasting0.073 ± 0.0510.082 ± 0.0380.0810 ± 0.0300.095 ± 0.0460.088 ± 0.048Δcnf-f [%]-2.73.86.6-12.833.3CM totalFasting1.15 ± 0.481.22 ± 0.491.21 ± 0.451.53 ± 0.941.14 ± 0.54(80 – 430 nm)Non-fasting1.15 ± 0.621.26 ± 0.491.32 ± 0.381.41 ± 0.571.37 ± 0.61Δcnf-f [%]03.39.1-7.820.2Apo-B48NMRbFasting1.15 ± 0.481.22 ± 0.491.21 ± 0.451.53 ± 0.941.14 ± 0.54Non-fasting1.15 ± 0.621.26 ± 0.491.32 ± 0.381.41 ± 0.571.37 ± 0.61Δcnf-f [%]03.39.1-7.820.2aFor lipoprotein nomenclature see Methods. The values given are the mean ± the standard deviation. The fasting values are from the spectra taken before the oGTT, the non-fasting values were extrapolated to time $t = 0$ from the spectra of reserve samples of the blood donors. Δcnf-f [%], relative particle concentrations c under non-fasting conditions minus those under fasting conditions. Number of participants analysed under fasting conditions (Nf) and non-fasting conditions (Nnf) 432 and 595, respectively. First column, values in bracket represent the diameters assumed for different subclassesbThe NMR derived apolipoprotein concentrations were calculated as described in Table 2Table 4Change of lipoprotein particle diameters and volumes after fastingaFasting stateControlIFGIGTIFG + IGTT2DNf196167212919Nnf251193266857Lipoprotein HDL totalFasting (7 – 16 nm) < d > [nm]8.628.558.618.598.52 < V > [nm3]721701717711693Non-Fasting < d > [nm]8.608.578.608.588.53 < V > [nm3]715707716709694Δdnf-f [%]-0.260.23-0.05-0.060.06ΔVnf-f [%]-0.870.75-0.10-0.250.18Δcnf-f [%]0.93.9-0.10.93.6 LDL totalFasting (16 – 30 nm) < d > [nm]20.020.220.120.220.2 < V > [nm3]9.07E + 39.27E + 39.14E + 39.29E + 39.28E + 3Non-Fasting < d > [nm]20.020.120.120.120.2 < V > [nm3]9.10E + 39.20E + 39.19E + 39.25E + 39.35E + 3Δdnf-f [%]0.16-0.270.20-0.130.20ΔVnf-f [%]0.31-0.720.57-0.390.74Δcnf-f [%]0.13.6-0.6-1.306.7 IDL and VLDLFasting (30 – 80 nm) < d > [nm]42.342.442.442.742.4 < V > [nm3]9.06E + 49.15E + 49.15E + 49.38E + 49.12E + 4Non-Fasting < d > [nm]42.242.442.542.642.7 < V > [nm3]9.05E + 49.16E + 49.22E + 49.30E + 49.32E + 4Δdnf-f [%]-0.060.040.25-0.230.62ΔVnf-f [%]-0.140.110.76-0.872.18Δcnf-f [%]1.51.74.8-4.511.8 CM totalFasting (80 – 430 nm) < d > [nm]121121120123119 < V > [nm3]4.55E + 64.54E + 64.44E + 64.88E + 64.24E + 6Non-Fasting < d > [nm]121121120122121 < V > [nm3]4.47E + 64.54E + 64.35E + 64.69E + 64.52E + 6Δdnf-f [%]-0.36-0.09-0.42-0.941.72ΔVnf-f [%]-1.840.08-2.01-3.996.75Δcnf-f [%]0.03.39.1-7.820.2aThe fasting values are from the spectra taken before the oGTT, the non-fasting vales were extrapolated to time $t = 0$ from the spectra of reserve samples of the blood donors. < d >, average particle diameter and < V > average volume calculated with the diameters given in Methods assuming a spherical shape.. Δdnf-f, ΔVnf-f, and Δcnf-f, relative differences (in %) of particle diameters, volumes, and concentrations under non-fasting conditions minus those under fasting conditions. Nf and Nnf, number of participants analysed under fasting or non-fasting conditions, respectively. Differences of mean values between fasting and non-fasting subjects with an error probability ≤ 0.05 are presented in bold letters Compared to fasting conditions, on average, normal nutritional conditions do not lead to a larger change of plasma apolipoprotein concentrations determined by NMR in the control group, the group without disorders of the glucose metabolism. Here, the concentrations of Apo-A1NMR, Apo-B100 NMR, and Apo. B48 NMR increase by only 0.9, 0.2, and $0\%$, respectively (Table 3). In contrast, in T2D the Apo-A1 Apo-B100, and Apo-B48 concentrations are significantly increased under non-fasting conditions by 3.6, 7.1, and $20.2\%$, respectively (Tables 3 and 5). This probably indicates an increased additional fatty acid synthesis in the liver when the plasma glucose concentration is strongly increased by food intake in T2D. The strongest effect of food intake is again observed in the LDL, VLDL, and chylomicron main classes for components with larger size that can carry larger amounts of lipids. The average volume (related to the absolute concentration of lipids transported) is largely increased (Table 4). The concentration of VLDL of the largest subclass VLDL B is increased by more than $19\%$ during normal food intake compared to the situation observed after fasting (Table 3). Even the number and average size of HDL particles is influenced by fasting in persons with T2D.Table 5Significance of lipoproteins concentration differences between healthy and (pre)diabetic participants under fasting and non-fasting conditionsaLipoproteinFasting stateIFGIGTIFG + IGTT2DHDL AFastingΔcx-H [%] + 1.5 + 0.6 + 2.9 + 4.7(7 – 8.5 nm)Non-fastingΔcx-H [%] + 2.4-1.0 + 1.9 + 6.2FastingPx-H0.2530.8510.2640.008Non-fastingPx-H0.0810.8900.1000.002HDL BFastingΔcx-H [%]-9.4-2.2 + 0.8-10.0(8.5 – 10 nm)Non-fastingΔcx-H [%]-2.7-3.1 + 3.4-5.0FastingPx-H0.0050.6470.7130.146Non-fastingPx-H0.4940.5140.3640.258HDL CFastingΔcx-H [%]-13.9-5.6-7.6-18.0(10 – 13 nm)Non-fastingΔcx-H [%]-5.7-4.5-6.0-13.4FastingPx-H < 0.0050.2310.4320.005Non-fastingPx-H0.0410.3430.3440.005HDL DFastingΔcx-H [%]-4.1 + 2.2 + 1.3-3.0(13 – 16 nm)Non-fastingΔcx-H [%] + 0.3 + 3.4 + 1.5-0.4FastingPx-H0.0220.9720.8760.460Non-fastingPx-H0.6150.2490.2050.827HDL totalFastingΔcx-H [%]-2.3-0.5 + 1.4-0.7(7 – 16 nm)Non-fastingΔcx-H [%] + 0.6-1.5 + 1.42.0Apo-A1NMRbFastingPx-H0.1000.7820.6830.861Non-fastingPx-H0.5280.8130.2790.154LDL AFastingΔcx-H [%]-9.1 + 0.1-3.0-15.8(16 – 19 nm)Non-fastingΔcx-H [%]-3.4-0.9-2.9-9.6FastingPx-H0.0010.4960.7020.008Non-fastingPx-H0.1630.4370.6530.015LDL BFastingΔcx-H [%]-4.2 + 1.4 + 1.2-8.4(19 – 21 nm)Non-fastingΔcx-H [%]-0.6 + 1.3 + 0.4-2.9FastingPx-H0.0510.8460.8200.081Non-fastingPx-H0.9970.7710.5200.437LDL CFastingΔcx-H [%]-4.1 + 0.80.0-9.8(21 – 22 nm)Non-fastingΔcx-H [%]-2.1-1.0-1.6-5.6FastingPx-H0.0510.7290.8050.031Non-fastingPx-H0.4420.6940.8000.081LDL DFastingΔcx-H [%]-2.0 + 1.7 + 3.4-7.4(22 – 25 nm)Non-fastingΔcx-H [%]-0.2 + 1.8 + 0.2-2.0FastingPx-H0.4950.9220.7090.176Non-fastingPx-H0.7520.6550.5550.696LDL EFastingΔcx-H [%] + 2.7 + 5.6 + 11.9-3.3(25 – 30 nm)Non-fastingΔcx-H [%] + 3.1 + 5.6 + 7.8 + 6.3FastingPx-H0.2170.2200.01320.653Non-fastingPx-H0.0740.1560.0120.083LDL totalFastingΔcx-H [%]-4.6 + 1.4 + 1.2-10.2(16 – 30 nm)Non-fastingΔcx-H [%]-1.2 + 0.7-0.2-4.3FastingPx-H0.0410.7760.8540.032Non-fastingPx-H0.8030.9940.6940.230IDLFastingΔcx-H [%] + 1.20.0 + 10.1-5.6(30 – 40 nm)Non-fastingΔcx-H [%] + 0.9 + 2.0 + 3.8 + 2.0FastingPx-H0.5450.9160.2020.309Non-fastingPx-H0.3950.5560.1360.546VLDL AFastingΔcx-H [%] + 5.0 + 5.3 + 20.8 + 0.6(40 – 60 nm)Non-fastingΔcx-H [%] + 6.1 + 10.8 + 14.0 + 13.0FastingPx-H0.0990.3640.0530.852Non-fastingPx-H0.0260.0490.0030.020VLDL BFastingΔcx-H [%] + 4.0 + 1.8 + 23.8-5.7(60 – 80 nm)Non-fastingΔcx-H [%] + 4.4 + 6.8 + 12.8 + 11.2FastingPx-H0.2050.8190.0320.417Non-fastingPx-H0.0960.1560.0060.055IDL and VLDLFastingΔcx-H [%] + 2.6 + 1.8 + 14.5-3.7(30 – 80 nm)Non-fastingΔcx-H [%] + 2.8 + 5.1 + 7.7 + 6.1FastingPx-H0.2480.7340.1030.629Non-fastingPx-H0.1600.2400.0290.180Apo-B100NMRbFastingΔcx-H [%]-4.0 + 1.4 + 2.3-9.7Non-fastingΔcx-H [%]-0.9 + 1.1 + 0.5-3.4FastingPx-H0.1270.8490.6670.071Non-fastingPx-H0.9540.8810.5540.360CM ReFastingΔcx-H [%] + 6.1 + 6.1 + 30.3 + 3.0(80 – 100 nm)Non-fastingΔcx-H [%] + 10.6 + 16.7 + 21.2 + 18.2FastingPx-H0.0620.4430.0340.927Non-fastingPx-H0.0040.016 < 0.0010.011CM AFastingΔcx-H [%] + 4.8 + 2.4 + 33.3-4.8(100 – 150 nm)Non-fastingΔcx-H [%] + 7.1 + 11.9 + 21.4 + 19.0FastingPx-H0.1220.5640.0110.588Non-fastingPx-H0.0160.041 < 0.0010.016CM BFastingΔcx-H [%] + 5.3 + 1.3 + 45.3-12.0(> 150 nm)Non-fastingΔcx-H [%] + 12.3 + 11.0 + 30.1 + 20.5FastingPx-H0.2050.9360.0120.284Non-fastingPx-H0.0110.154 < 0.0010.048CM totalFastingΔcx-H [%] + 6.1 + 5.2 + 33.0-0.9(80 – 430 nm)Non-fastingΔcx-H [%] + 9.6 + 14.8 + 22.6 + 19.1Apo-B48NMRbFastingPx-H0.0830.5460.0200.804Non-fastingPx-H0.0060.022 < 0.0010.012aΔcx-H represents the relative particle concentrations c in the groups of (pre)diabetic participants (x) minus those in the control group of healthy participants (H). The error probabilities P are obtained by using the non-parametric Mann–Whitney U-test. They compare the particle concentrations of the healthy control group with the different (pre)diabetic cohorts. Δc values with an error probability Px-H ≤ 0.05 are presented in bold letters. First column, values in bracket represent the diameter ranges assumed for different subclasses. For more details see Table 3bThe NMR derived apolipoprotein concentrations were calculated as described in Table 2 A pattern of diet dependent lipoprotein changes similar to T2D is observed for all groups with impaired fasting glucose or impaired glucose tolerance. The only difference is that the lipoprotein concentration changes are smaller. Interestingly, impaired fasting glucose is more similar to T2D concerning the concentration changes of lipoproteins that carry lipids synthesized in the liver (HDL, LDL, IDL, VLDL). In contrast, subjects with impaired glucose tolerance mainly show an increase of lipoproteins synthesized in the intestinal system (chylomicrons). This may indicate differences in the pathophysiology of the two forms of prediabetes (IFG, IGT). Surprisingly, the group characterized by IFG + IGT shows a very different response to fasting than the other groups. Here, fasting leads to a statistically significant increase (not decrease!) of LDL/VLDL type as well as chylomicron particles, in contrast to the changes observed in IFG, IGT, and T2D but particles with some similarities to the control group of healthy subjects. ## Differences in lipoprotein particle concentrations in subjects with and without disorder of glucose metabolism The differences of lipoprotein particle concentrations between healthy people and people with isolated IFG, isolated IGT, IFG combined with IGT, or manifest T2D are summarized in Tables 3 and 5. These differences can be a consequence of the perturbation of the glucose metabolism but can also represent the consequence of a risk factor associated with a general dyslipoproteinemia. However, a reasonable hypothesis is that they represent lipoprotein concentration changes that mainly are the consequence of the actual metabolic state. People with impaired fasting glucose (IFG) have an increased number of chylomicrons and VLDL particles compared to the control group under fasting and normal nutritional conditions. The increase of concentrations is about twice as large under non-fasting conditions and becomes statistically significant for all chylomicron subclasses (CM B, CM A, CM Re). The highest increase by more than $12\%$ is observed for the largest chylomicrons. A smaller increase is also observed for VLDL that is with $6.1\%$ significant for VLDL A under non-fasting conditions. IDL concentrations are barely influenced. The total LDL particle number is decreasing relative to the control group with a stronger decrease observable under fasting conditions. The strongest decrease is observed for the smallest LDL particles (LDL A) with -$9.1\%$. However, the decrease of particle numbers is not uniform in all LDL subclasses, in fact, the particle concentration of the largest LDL particles (LDL E) has the tendency to increase (not statistical significant). For HDL, in total a small decrease of particle numbers relatively to the control group is observed with a stronger effect under fasting conditions. Again, the magnitude and sign of these effects vary from subclass to subclass. Statistically significant are the decrease of particle numbers in the HDL B subclass under fasting conditions and the HDL C subclass for fasting and non-fasting conditions (Table 5). Qualitatively, the same lipoprotein subclass patterns are observable for IFG and T2D. Quantitatively, in cases, that are significant in both groups, the increase relative to the control group is also much stronger (approximately twice) in T2D compared to IFG. This would suggest that the insulin resistance and/or concomitant increased blood glucose are the common factor influencing the lipoprotein profile. As in IFG and T2D, impaired glucose tolerance leads to strong concentration increases of large lipoprotein particles (chylomicrons and VLDL) compared with the control groups under fasting and normal nutrition conditions. *In* general, the concentration increase of these particles is larger than in IFG but smaller than in T2D. Under non-fasting conditions the Apo-48NMR concentration is significantly increased by almost $15\%$ (Table 5). Again IGT shows quite small effects on the HDL and LDL subclass concentrations. The only differences concern HDL B and HDL C concentrations under fasting conditions that are strongly reduced in IFG but not in IGT. Under non-fasting conditions the IFG + IGT group with impaired fasting glucose combined with impaired glucose tolerance shows much larger increases of particles concentrations of most of the subclasses larger particles (VLDL including IDL and chylomicrons) than the groups with isolated IFG or IGT. The particle number changes in the different subgroups in IFG and IGT do not simply sum up to the values found in the IFG + IGT group. Only for the largest particles (VLDL and chylomicrons) this seems to be the case under fasting and non-fasting conditions. Under non-fasting conditions, the increase of large particle concentrations in the IFG + IGT group corresponds rather well to that observed in the T2D-group. Under fasting conditions, the concentration changes observed differ clearly from those observed for T2D. For large chylomicrons (CM B), an increase by + $45.3\%$ is calculated for IFG + IGT, whereas for T2D a decrease by -$9.3\%$ is observed (Table 5). Unfortunately, the number of test persons of the IFG + IGT group was too low to reach an error probability $P \leq 0.05$, meaning that this may also be a statistical error. It would also be interesting to see, if people without disorders of the glucose metabolism shows differences in their lipoprotein patterns compared to the whole cohort of subjects with IFG, IGT, or T2D. If this cohort behaves like a homogeneous group, the statistical significance should also increase because of the larger number of participants. Under fasting conditions, the changes found to be significant for T2D alone (decrease of the particle numbers of HDL C and LDL A) remain significant for the whole group. The strong increase of HDL A particle numbers is found to be specific for T2D. In addition, an increase of the particle numbers of chylomicron remnants ($P \leq 0.004$), and a decrease of the particle numbers of HDL B ($$P \leq 0.007$$), HDL C ($$P \leq 0.001$$), LDL A ($$P \leq 0.001$$), LDL B ($$P \leq 0.04$$), and LDL C ($$P \leq 0.03$$) get significant for the whole group. Under non-fasting conditions, the increase of VLDL A, and of all chylomicron particles numbers becomes even more significant ($P \leq 0.003$). The decrease of HDL C and the increase of LDL E and VLDL B observed also for T2D gets now significant with the data of the whole group ($$P \leq 0.027$$, 0.006, 0.006, respectively). ## The study cohort It is a general problem that the results of studies primarily reflect statistical properties of the specific cohort of individuals studied. As already mentioned above, the ratio of 0.58 of the two sexes in our study approximates also the ratio of the two sexes in the complete cohort of blood donors of the BRK. The five groups of the study cohort including the control group were matched with respect to age, BMI, and the general risk to develop T2D as shown by the FindRisk score. Our study cohort consists of long-term blood donors that are healthy according to the strict rules defined for blood donors by the Bavarian Red Cross. A disorder of the glucose metabolism was unknown to the individuals, before they became recruited to the study. It also means that they did not have severe metabolic symptoms leading to a consultation of a medical doctor and thus leading to a specific treatment with antidiabetics. IFG, IGT, or T2D were newly diagnosed by oGTT. In line with this, the mean HbAc1-value of the control group was with 5.8 not much lower than 6.3 of the T2D group. For most subjects with T2D the relative HbAc1-concentration was below the limit of $6.5\%$, traditionally used for diagnosing T2D. This indicates that in these subjects T2D with increased glucose concentrations had probably prevailed for a relatively short period of time, in agreement with the accession criteria to the study. They were chosen with the aim to identify persons that just developed their disorder of glucose metabolism. This apparently short history of impaired glucose metabolism may also influence the actually observed apolipoprotein concentrations in T2D. Therefore, the metabolic changes described in this study may better characterize the effect of “pure” insulin resistance of diabetes on the lipoprotein profile and not other concomitant metabolic changes that are known to be partly causative for the development of T2D as the metabolic syndrome. Such a dissociation of the insulin resistance per se and the usually observed dyslipoproteinemia was also described for a cohort of lean Chinese subjects recently [14]. The lipoprotein concentrations determined from the reserve samples change only moderately with time (Table 2). The particle concentrations in all particle classes increased with time for the healthy control group, the group with combined IFG and IGT, and the T2D group. This is to be expected since the BMI and the coupled plasma lipid concentrations usually increase with time (age). However, the mean increase of particle numbers and of apolipoprotein concentrations derived by NMR increased only by 0.25 and $0.12\%$/year for Apo-A1 and Apo-B100, respectively. Surprisingly, the group with isolated IGT showed a different trend. The particle concentrations of almost all particle groups decreased with time. The only exception were small HDL particles (HDL A) whose concentration increased with time. The reason for this difference is not clear for us but may give information about the pathomechanism of IGT. It would be interesting to study that in more detail. ## Determination of lipoprotein particle and apolipoprotein concentrations by NMR Initially, only the intensities of NMR lines separated by size dependent chemical shifts and sometimes (as in our case) the size dependent diffusion constants could be used to count the number of lipid protons in the different lipoprotein subclasses. Together with the approximate lipid composition from these line intensities, the particle concentrations were derived. Baumstark et al. [ 6] showed that a substantial part of the lipid signals are NMR invisible and thus may introduce large errors in the particle concentration determination. With the data from [6] we corrected our concentrations dependent on the specific subclass under consideration and the experimental temperature during data recording (see Methods). Actually, more and more groups realize that apolipoprotein concentrations determined by alternative methods such as immunoassays can be used as independent check of the lipoprotein particle numbers obtained by NMR spectroscopy. An easily conceivable idea is that vice versa correct particle numbers determined e. g. by NMR can be used to estimate the corresponding apolipoprotein. It is now established in the lipidomics community that the apolipoprotein stoichiometry is fixed in the different main classes of proteins. Chylomicrons and chylomicron remnants contain just one Apo-B48 [19, 20], LDL, IDL, and VLDL on Apo-B100 [17, 21]. Molecular evidence shows that most probably 2 Apo-A1-molecules [22–27] are arranged in an antiparallel manner for stabilizing ordered phospholipid membranes. Based on this fact, recombinant Apo-A1 is used routinely since more than a decade to produce artificial nanodiscs for x-ray and NMR structural studies. These nanodiscs form spontaneously in the presence of lipids and Apo-A1. In some older publications, also more than two ApoA-I are assumed to be bound to large HDL particles [25]. The most likely stoichiometry for Apo-A1 is 2 apolipoproteins per particles. The proportionality of the Apo-A1 and Apo-B100 concentration to the corresponding particle concentrations is experimentally well-established [17, 18]. Accepting this stoichiometry, the concentrations of these apolipoproteins can be approximated by summing up the particle concentrations in the different classes. Since a final verification of the exact stoichiometry is still missing, we annotate the apolipoprotein concentrations determined by NMR with the suffix “NMR” in the tables. Under fasting conditions we obtain for the control group of subjects without disorders of glucose metabolism using the above stoichiometry Apo-A1 and Apo-B100, concentrations of 49.9 μM and 1.74 μM, respectively. These values obtained from our visibility corrected particle numbers for Apo-A1 and Apo-B100 are quite close to the average values of 51.9 μM and 1.70 μM reported for their control group by Monsoni-Centelles [18] by apolipoprotein specific immunoassays. This consistency check strongly supports the quantitative validity of our NMR analysis. The concentrations of Apo-B48 determined here for the control group are with 1.16 nM substantially smaller than 8.4 nM and 18.5 nM determined by immunoassays and reported by Masuda et al. [ 19] and Tian et al. [ 28], respectively. The determination of the chylomicron particle concentrations by NMR is more tricky, since NMR can distinguish particles on the basis of their size only. Because of the overlap in size of very large VLDL particles with small chylomicrons and chylomicron remnants (see e. g. [29, 30]) these groups can only be partly separated by NMR. This means that the particle numbers of large VLDL B and smaller chylomicrons can only be approximated by using a suitable size cutoff. In addition, weighting factors for the visibility of chylomicrons are not published yet. Because of lack of information we set the weighting factor to 1 in the calculations of chylomicron particle concentrations. Another factor possibly reducing the measured fraction of chylomicrons may be the freezing of samples before NMR analysis. *More* generally, freezing and (long-term) storage may influence the outcome of the actual measurements under fasting conditions as well as the spectroscopy of the reserve samples. Storage of frozen biological samples at -80 C is assumed to preserve very well their integrity. This is the basis of all biological data bases that intend to provide long-term samples. The only typical effects are sometimes very slow oxidation processes depending on details of the composition of the samples. *In* general, the critical point is freezing and thawing for complex samples (see e.g. [6]). Figure 3 (top) shows an example for the increase of the chylomicron specific NMR signals (maybe partly superposed by signals of large VLDL) of the corresponding methyl and methylene groups after an intake of a fat rich diet. Figure 3 (bottom) shows the effect of freezing and thawing cycles of this sample. After two cycles the chylomicron NMR signal is only slightly reduced. However, our samples are only frozen once and a signal reduction less than $5\%$ is to be expected (Fig. 3). After several freezing and thawing cycles a stronger reduction of the signal is observed meaning that the chylomicron particle structure is partly destroyed. Note that the NMR signal of LDL and HDL is still unchanged after 6 cycles. There is ample evidence that long-term storage at -80 C or only -20 C does not have an effect on the lipoprotein analysis by NMR (see e. g. [31–34]).Fig. 31H spectra showing effects of food intake and sample freezing on the methyl and methylene signals of chylomicrons. ( Top) Changes of the CH2 (CM- CH2) and CH3 (CM- CH3) signals of chylomicrons recorded at different times after food intake. ( Bottom) Changes of the CH2 (CM- CH2) and CH3 (CM- CH3) signals of chylomicrons recorded after repeated freezing and thawing of the samples (0-times, 2-times, 5-times, 6-times and stored at 253 K). Note that the signals of smaller chylomicrons may overlap with signals of very large VLDLs Taking as comparison the particle numbers from published NMR based diabetes studies presented in Fig. 4, one obtains Apo-AI concentrations in the control groups of 14.6 μM [10], 68.2 μM [9], and 41.4 μM [12]. The variation of Apo-AI concentrations in the control groups of the different studies are quite large. The Apo-B100 concentrations calculated from the particle concentrations for the control groups are 0.66 μM [10] and 1.28 μM [9]. Compared with 51.9 μM (Apo-A1) and 1.70 μM (Apo-B100) [18] mentioned above it suggests that these values have to be considered with care, even when taking into account that the apolipoprotein concentration determination by immunoassays has an error of about $15\%$ and the control groups are not identical. This means, that one has to be very careful when absolute values of particle concentrations determined by NMR by different programs are essential. However, more important in medical diagnosis are the concentration changes relative to a reference value given by the provider of a test. Indeed, when analyzing the effects of T2D on the lipoprotein subclass concentration changes consistent results are obtained in all studies (see below).Fig. 4Changes of lipoprotein particle concentrations by impairment of glucose metabolism as described in literature. Concentration increase (+) or decrease [-] in with IFG, IGT, and T2D, green, significant increase, blue significant decrease. Note that. f, fasting, nf, non-fasting. The nomenclature of the lipoprotein subclasses and the correspondent particle diameters d were similar to those given by Huber et al. [ 4] and Kaess et al. [ 35]. The definitions of subclasses and particle diameters d vary from study to study. The results of these studies were assigned as good as possible by the diameters to the subclasses given here. Data are taken from Kalbitzer (this study); Festa et al., [ 7]: Mora et al. [ 9]; Hodge et al. [ 8]); Mackey et al. [ 11]; Wang et al. [ 10]; Sokooti et al. [ 13] ## Variations of apolipoprotein concentrations in impaired glucose tolerance, impaired fasting glucose. and in manifest type 2 diabetes mellitus Recently, variations of apolipoprotein concentrations in diseases linked to dyslipoproteinemia such as atherosclerosis and coronary heart disease have invoked new interests. The American Societies of Cardiology [36] and the European Society of Cardiology [37] recommend the preferential determination of Apo-B100 concentrations as basic risk assessment for atherosclerosis and coronary heart disease. Mainly the number of apo-B100 particles is predictive for the CHD risk, not the classical cholesterol linked values [38]. The Apo-B100 concentration is also recommended as more suited for therapy control with statins. In our study, we find a decrease of Apo-B100NMR by -9.8 and -$3.2\%$ for fasting and non-fasting conditions when participants with newly detected T2D are compared to the healthy control group. Unfortunately, this decrease is not statistically significant at $P \leq 0.05$ but for fasting it is significant at an error probability $P \leq 0.07.$ In contrast, in the other studies represented in Fig. 4 an increase of Apo-B100NMR is observed when calculated from the particle concentrations. The simplest explanation for these differences is that this mainly is an effect of the cohort studied. Our data set is compared with data sets containing long-term diabetics together with possible other health problems. For Apo-B48 under normal nutritional conditions a statistically significant increase can be observed for IGT, IGT + IFG, and T2D itself. ## Comparison of the present study with results of similar studies of diabetes related lipoprotein changes As mentioned in Background there are a number of studies that relate lipoprotein particle concentrations determined by NMR with type 2 diabetes mellitus [7–13]. They differ from each other and from our study as well in many aspects, the number of lipoprotein subclasses and particle sizes defined, the composition of the study cohorts, the fasting state and the method of diagnosis of type 2 diabetes mellitus. Concerning the diagnosis of T2D only Wang et al. [ 10] used the gold standard for diabetes diagnosis, oGGT, a method that also was used in the present paper. In the other studies the metabolic state of the participants was not unambiguously defined, the diagnosis T2D was not clearly verified and their cohorts may include also subjects with IFG and/or IGT only. The definitions of lipoprotein subclass particle sizes used in our study are given in Table 3. Although the exact definitions differ considerably from study to study, it is possible to find a kind of consensus pattern for the different studies. These definitions are used in Fig. 4 to represent schematically the changes of lipoprotein particle concentrations in T2D described in the present paper and the cited publications. Under standard fasting conditions in all studies the concentration of the smallest HDL particles (HDL A) is significantly increased in T2D, in our case by $5.9\%$. Under non-fasting condition in our study its increase is even higher, but only statistical significant when people with IFG and/or IGT are included. The particle concentrations of the other HDL subclasses are decreased in most studies. Including the present study (decrease by -$18\%$), the HDL C subclass particle concentrations are significantly decreased in all studies. In our study, the total concentration of HDL particles of all subclasses together (and thus of Apo-A1) is only weakly increased in T2D, it is almost not influenced in our cohort of blood donors (Fig. 4, Table 5). Similar results are also obtained in other studies where Apo-A1 is increased by $0.6\%$ [10] or decreased by -3 or -$8\%$ [12, 13]. The concentrations of small LDL particles (LDL A and LDL B) are significantly increased in the five published studies by diabetes (Fig. 4), whereas they are significantly decreased in our study under non-fasting conditions when also the participants with IFG and/or IGT are included. The main difference may be that in our cohort all participants were “healthy” according to the standards defined for blood donors just before diabetes had been detected by oGGT. In our group the average BMI of people with T2D is less than $10\%$ higher than for the control group of healthy volunteers and quite moderate for a group with an average age of 53 and 57 years, respectively (Table 1). This probably means that obesity related dyslipoproteinemia effects are quite small. The concentration of larger LDL particles (LDL C, D, and E) is significantly reduced in three of the five studies and also in our study. VLDL particle concentrations are increased in T2D in all published studies relative to the control group. In our case, only under non-fasting conditions, significance for an increase is reached. The Apo-B100NMR concentration change in subjects with T2D calculated from the particle concentrations varies from + $2.4\%$ [10] to + $26.6\%$ [9]. However, in our case, we find a (not significant) decrease of the Apo-B100 concentration in T2D under fasting and non-fasting conditions (Tables 2 and 4). Except of our present study, only Wang et al. [ 10] studied the changes of the chylomicron concentrations in T2D under fasting conditions. They showed a remarkable increase of chylomicron concentrations in diabetic people by $107\%$. In our case for all three subclasses of chylomicron particles also a significant concentration increase is observed (Fig. 4). Under non-fasting conditions only for large chylomicrons (CM B) a significant concentration increase is observed for people with impaired glucose metabolism. Note that in [10] the nutritional state of the participants was mixed including postprandial data. As discussed above, it is impossible to separate very large VLDLs from small chylomicrons or chylomicron remnants by NMR only, that is the chylomicron fraction may contain a significant contribution of very large VLDL particles in both studies. Concentrations of chylomicrons and very large VLDLs, respectively, under non-fasting conditions may be the most sensitive marker for T2D in the lipoprofile. This has also proposed by Mora et al. [ 9] for T2D. Postprandial VLDL and chylomicron concentrations seem also be the most sensitive marker for the risk for cardiovascular diseases [39, 40]. What exactly is happening in type 2 diabetes mellitus pathophysiologically is an open question, is it the lipid resorption, the fatty acid resynthesis, the chylomicron synthesis and clearance or (most likely) all factors together that cause this increase of chylomicrons in blood serum [41, 42]. In T2D insulin does not downregulate the chylomicron synthesis as it does in healthy individuals according to Nogueira et al. [ 43]. This leads to increased intestinal chylomicron synthesis and secretion in insulin resistance and T2D [44] as also being observed in this study. ## Conclusions As we have shown the absolute, NMR derived, particle concentrations in many publications cited here vary substantially and the quantifications are probably partly incorrect. Using the temperature dependent visibility/invisibility values from Baumstark et al. [ 6] as done in the present study gives quite reliable absolute concentrations. In the long-term, a standardization of the method would be required that is based on well-defined size distributions and apolipoprotein concentrations in the different subclasses. The definition of sizes and apolipoprotein stoichiometry should be a future task for the regulatory bodies. However, for the effects of impaired glucose metabolism (IFG, IGT, T2D) on the lipoprotein profile relative concentration changes are mainly important. Here, all studies observe a concentration increase of small HDL particles and a decrease of large HDL particles in T2D (Fig. 4). In addition, an increase of VLDL particle concentrations, and, where data are available, of chylomicrons and chylomicron remnants is observed. Our study is different to most other studies that the diabetic state of our subjects was not known before, our study subjects were healthy according to the criteria set for blood donors (that is also seen in the quite low HbAc1 values) and the diagnosis was based on the gold standard method oGGT. In addition, we can present data for the same subjects under fasting and “normal” nutritional conditions. Here, the diabetic metabolism is easier to observe. Contrary to other studies, in our “healthy” cohort of blood donors the T2D associated dyslipoproteinemia does not significantly change the total concentrations of the lipoproteins produced in the liver under fasting and non-fasting conditions but selectively their subclass distributions. In contrast, under normal nutritional conditions persons with IFG, IGT or T2D show a substantial increase of plasma concentrations of those lipoproteins that are produced in the intestinal tract. An important effect of the insulin resistance gets visible here. Different to other studies, we observe a slight, significant decrease of the average concentration of small LDL particles (LDL A). This may be again an effect of our cohort of blood donors but one has to be somewhat careful since our cohort itself is not very large. ## Duality of interests This study has funded by an investigator-initiated research grant of the Bavarian Research Foundation. Lipofit GmbH (now numares AG) had no control over details of data evaluation but provided the environment for the primary NMR data recording at their 600 MHz NMR spectrometer and their software for the lipoprotein particle number estimation. All experiments as well as primary and secondary data evaluations had performed by members of the University of Regensburg. The Bavarian Red Cross (BRK) organized the acquisition of volunteers out of their pool of blood donors according to the rules defined by the grant application. They organized the oGTT tests and provided the corresponding retained samples out of their pool. ## Prior presentation Except for the PhD-examination of K. L. the data were not presented elsewhere before. ## References 1. Otvos JD, Jeyarajah EJ, Bennett DW. **Quantification of plasma lipoproteins by protein nuclear magnetic resonance spectroscopy**. *Clin Chem* (1991.0) **37** 377-386. DOI: 10.1093/clinchem/37.3.377 2. Otvos JD. *Method and apparatus for measuring classes and subclasses of lipoproteins," Patent US 5343389 A* (1994.0) 3. Ala-Korpela M, Korhonen A, Keisala J, Horkko S, Korpi P, Ingman LP, Jokisaari J, Savolainen MJ, Kesäniemi YA. **1H NMR-based absolute quantitation of human lipoproteins and their lipid contents directly from plasma**. *J Lipid Res* (1994.0) **35** 2292-2304. DOI: 10.1016/S0022-2275(20)39935-1 4. 4.Huber F, Kalbitzer H R, Kremer W. Verfahren zur Bestimmung von Lipoproteinen in Körperflüssigkeiten und Messanordnung dafür. DE 10 2004 026 903 B4 2006.05.18. 5. Kremer W, Kalbitzer HR, Huber F. *Process for the determination of lipoproteins in body fluids. US 7,927,878 B2* (2011.0) 6. Baumstark D, Kremer W, Boettcher A, Schreier C, Sander P, Schmitz G, Kirchhoefer R, Huber F, Kalbitzer HR. **1H NMR spectroscopy quantifies visibility of lipoproteins, subclasses, and lipids at varied temperatures and pressures**. *J Lipid Res* (2019.0) **60** 1516-1534. DOI: 10.1194/jlr.M092643 7. Festa A, Williams K, Hanley AJG, Otvos JD, Goff DC, Wagenknecht LE, Haffner SM. **Nuclear magnetic resonance lipoprotein abnormalities in prediabetic subjects in the insulin resistance atherosclerosis study**. *Circulation* (2005.0) **111** 3465-3472. DOI: 10.1161/CIRCULATIONAHA.104.512079 8. Hodge AM, Jenkins AJ, English DR, O’Dea K-, Giles GG. **NMR-determined lipoprotein subclass profile predicts type 2 diabetes**. *Diabetes Res Clin Pract* (2009.0) **83** 132-139. DOI: 10.1016/j.diabres.2008.11.007 9. Mora S, Otvos JD, Rosenson RS, Pradhan A, Buring JE, Ridker PM. **Lipoprotein particle size and concentration by nuclear magnetic resonance and incident type 2 diabetes in women**. *Diabetes* (2010.0) **59** 1153-1160. DOI: 10.2337/db09-1114 10. Wang J, Stančáková A, Soininen P, Kangas AJ, Paananen J, Kuusisto J, Laakso M. **Lipoprotein subclass profiles in individuals with varying degrees of glucose tolerance: a population-based study of 9399 Finnish men**. *J Intern Med* (2012.0) **2012** 562-572. DOI: 10.1111/j.1365-2796.2012.02562.x 11. Mackey RH, Mora S, Bertoni AG, Wassel CL, Carnethon MR, Sibley CT, Goff DC. **Lipoprotein particles and incident type 2 diabetes in the multi-ethnic study of atherosclerosis**. *Diabet Care* (2015.0) **38** 628-636. DOI: 10.2337/dc14-0645 12. Sokooti S, Szili-Torok T, Flores-Guerrero JL, Osté MCC, Gomes-Neto AW, Kootstra-Ros JE, Heerspink HJL, Connelly MA, Bakker SJL, Dullaart RPF. **High-density lipoprotein particles and their relationship to posttransplantation diabetes mellitus in renal transplant recipients**. *Biomol* (2020.0) **10** 481 13. Sokooti S, Flores-Guerrero JL, Kieneker LM, Heerspink HJL, Connelly MA, Bakker SJL, Dullaart RPF. **HDL particle subspecies and their association with incident type 2 diabetes: the PREVEND study**. *J Clin Endocrin Metabol* (2021.0) **106** 1761-1772. DOI: 10.1210/clinem/dgab075 14. Tranaes K, Ding C, Chooi YC, Chan Z, Choo J, Leow MK-S, Magkos F. **Dissociation between insulin resistance and abnormalities in lipoprotein particle concentrations and sizes in normal-weight Chinese adults**. *Front Nutr* (2021.0) **8** 651199. DOI: 10.3389/fnut.2021.651199 15. Bergmann A, Li J, Wang L, Schulze J, Bornstein SR, Schwarz PEH. **A simplified finnish diabetes risk score to predict type 2 diabetes risk and disease evolution in a German population**. *Horm Metabol Res* (2007.0) **39** 677-682. DOI: 10.1055/s-2007-985353 16. Martin E, Ruf E, Landgraf R, Hauner F, Weinauer S. **FINDRISK Questionnaire combined with HbA1c testing as a potential screening strategy for undiagnosed diabetes in a healthy population**. *Horm Metab Res* (2011.0) **43** 782-787. DOI: 10.1055/s-0031-1286333 17. Delatour V, Clouet-Foraison N, Gaie-Levrel F, Marcovina SM, Hoofnagle Z, Kuklenyik MP, Caulfield JD, Otvos RM, Krauss KR, Kulkarni JH, Contois AT, Remaley HW, Vesper CM, Cobbaert P. **Comparability of Lipoprotein Particle Number Concentrations Across ES-DMA, NMR, LC-MS/MS, immunonephelometry, and VAP: In Search of a candidate reference measurement procedure for apoB and non-HDL-P standardization**. *Clin Chem* (2018.0) **64** 1485-1495. DOI: 10.1373/clinchem.2018.288746 18. Monsonis-Centelles S, Hoefsloot HCJ, Engelsen SB, Smilde AK, Lind MV. **Repeatability and reproducibility of lipoprotein particle profile measurements in plasma samples by ultracentrifugation**. *Clin Chem Lab Med* (2020.0) **58** 03-115 19. Matsuda D, Nishida M, Arai T, Hanada H, Yoshida H, Yamauchi-Takihara K, Moriyama T, Tad N, Yamashita S. **Reference interval for apolipoprotein B-48 concentration in healthy Japanese Individuals**. *J Atheroscler Thromb* (2014.0) **21** 618-627. DOI: 10.5551/jat.22558 20. 20.Phillips ML, Pullinger C, Kroes I, Kroes J, Hardman DA, Chen G, Curtiss LK, Gutierrez MM, Kane JP, Schumaker VN. A single copy of apolipoprotein B-48 is present on the human chylomicron remnant. J Lipid Res 1997;38:1170–7. 21. Wiklund O, Dyer CA, Tsaoe BP, Curtiss LK. **Stoichiometric binding of apolipoprotein specific monoclonal antibodies to low density lipoproteins**. *J Biol Chem* (1985.0) **20** 10956-10960. DOI: 10.1016/S0021-9258(17)39129-9 22. Matera R, Horvath KV, Nair H, Schaefer EJ, Asztalos BF. **HDL Particle measurement: comparison of 5 methods**. *Clin Chem* (2018.0) **64** 492-50.0. DOI: 10.1373/clinchem.2017.277632 23. He Y, Song HD, Anantharamaiah GM, Palgunachari MN, Bornfeldt KE, Segrest JP, Heinecke JW. **Apolipoprotein A1 Forms 5/5 and 5/4 antiparallel dimers in human high-density lipoprotein**. *Mol Cell Proteomics* (2019.0) **18** 854-864. DOI: 10.1074/mcp.RA118.000878 24. Fang Y, Gursky O, Atkinson D. **Lipid-binding studies of human apolipoprotein A-I and its terminally truncated mutants**. *Biochemistry* (2003.0) **42** 13260-13268. DOI: 10.1021/bi0354031 25. Bibow S, Polyhach Y, Eichmann C, Chi CN, Kowal J, Albiez S, McLeod RA, Stahlberg H, Jeschke G, Günter P, Riek R. **Solution structure of discoidal high-density lipoprotein particles with a shortened apolipoprotein A-I. Mature**. *Struct Mol Biol* (2017.0) **24** 187-193. DOI: 10.1038/nsmb.3345 26. Brouillette CG, Anantharamaiah GM, Engler JA, Borhani DW. **Structural models of human apolipoprotein A-I: a critical analysis and review**. *Biochim Biophys A* (2001.0) **1531** 4-46. DOI: 10.1016/S1388-1981(01)00081-6 27. Hutchins PM, Ronsein GE, Monette JS, Pamir N, Wimberger J, He Y, Anantharamaiah GM, Kim DS, Ranchalis JE, Jarvik GP, Vaisar T, Heinecke JW. **Accurate quantification of high density lipoprotein particle concentration by calibrated ion mobility analysis**. *Clin Chem* (2014.0) **60** 1393-1401. DOI: 10.1373/clinchem.2014.228114 28. Tian J, Chen H, Liu P, Wang C, Chen Y. **Fasting apolipoprotein B48 is associated with large artery atherosclerotic stroke: a case control study**. *Sci Rep* (2019.0) **9** 3729. DOI: 10.1038/s41598-019-40385-0 29. Konishi T, Fujiwara R, Saito T, Satou N, Hayashi Y, Crofts N, Iwasaki I, Abe Y, Kawata S, Shikawa T. **Human lipoproteins comprise at least 12 different classes that are lognormally distributed**. *PLoS ONE* (2022.0) **17** e0275066. DOI: 10.1371/journal.pone.0275066 30. Nakajima K, Nakano T, Tokita Y, Nagamine T, Inazu A, Kobayashi J, Mabuchi H, Kimber L, Stanhope KL, Havel PJ, Okazaki M, Ai M, Tanaka A. **Postprandial lipoprotein metabolism VLDL vs chylomicrons**. *Clin Chem A* (2011.0) **412** 1306-1318. DOI: 10.1016/j.cca.2011.04.018 31. Kronenberg F, Lobentam EV, Konig P, Utermann G, Dieplinger H. **Effect of sample storage on the measurement of lipoprotein[a], apolipoproteins B and A-IV, total and high density lipoprotein cholesterol and triglycerides**. *J Lipid Res* (1994.0) **35** 1318-1328. DOI: 10.1016/S0022-2275(20)39975-2 32. Zivkovic AM, Wiest MM, Nguyen UT, Davis R, Watkins SM, German JB. **Effects of sample handling and storage on quantitative lipid analysis in human serum**. *Metabolomics* (2009.0) **5** 507-516. DOI: 10.1007/s11306-009-0174-2 33. Cuhadar S, Koseoglu M, Atay A, Dirican A. **The effect of storage time and freeze-thaw cycles on the stability of serum samples**. *Biochem Med* (2013.0) **23** 70-77. DOI: 10.11613/BM.2013.009 34. Loo RL, Lodge S, Kimhofer T, Bong SH, Begum S, Whiley L, Gray NC, Lindon JC, Nitschke P, Lawler NG, Schafer H, Spraul M, Richards T, Nicholson JK, Holmes E. **Quantitative in-vitro diagnostic NMR spectroscopy for lipoprotein and metabolite measurements in plasma and serum: recommendations for analytical artifact minimization with special reference to COVID-19/SARS-CoV-2 Samples**. *J Proteome Res* (2020.0) **19** 4428-4441. DOI: 10.1021/acs.jproteome.0c00537 35. Kaess B, Fischer M, Baessler A, Stark K, Huber F, Kremer W, Kalbitzer HR, Schunkert H, Riegger G, Hengstenberg C. **(2008) The lipoprotein subfraction profile: heritability and identification of quantitative trait loci**. *J Lipid Res* (2008.0) **49** 715-723. DOI: 10.1194/jlr.M700338-JLR200 36. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, Braun LT, de Ferranti S, Faiella-Tommasino J, Forman DE, Goldberg R, Eidenreich PA, Hlatky MA, Jones DW, Lloyd-Jones D, Lopez-Pajares N, Ndumele CF, Orringer CE, Peralta CA, Saseen JJ, Smith SC, Sperling L, Virani DS, Yeboah J. **AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/ APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol: executive summary**. *J Am Coll Cardiol* (2019.0) **73** 3168-3209. DOI: 10.1016/j.jacc.2018.11.002 37. 37.Mach F, Baigent C, Catapano AL, Koskinas KC, Casula M, Badimon L, Chapman MJ, De Backer GG, Delgado V, Ference BA, Graham IM (Ireland), Halliday A, Landmesser U, Mihaylova B, Pedersen TR, Riccardi G, Richter DJ, Sabatine MS, Taskinen MR, Tokgozoglu L, Wiklund O. 2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk. Eur Heart J. 2020;41:111-188. 38. Kohli-Lyncha CN, Thanassoulisc G, Morana AE, Sniderman AD. **The clinical utility of apoB versus LDL-C/non-HDL-C**. *Clin Chim A* (2020.0) **508** 103-108. DOI: 10.1016/j.cca.2020.05.001 39. Mora S, Rifai N, Buring JE, Ridker PM. **Fasting compared with nonfasting lipids and apolipoproteins for predicting incident cardiovascular events**. *Circulation* (2008.0) **118** 993-1001. DOI: 10.1161/CIRCULATIONAHA.108.777334 40. Bansal S, Buring JE, Rifai N, Mora S, Sacks FM, Ridker PM. **Fasting compared with nonfasting triglycerides and risk of cardiovascular events in women**. *JAMA* (2007.0) **298** 309-316. DOI: 10.1001/jama.298.3.309 41. Haas ME, Attie AD, Biddinger SB. **The regulation of ApoB metabolism by insulin**. *Trends Endocrinol Metabol* (2013.0) **24** 391-397. DOI: 10.1016/j.tem.2013.04.001 42. Giammanco A, Cefalù AB, Noto D, Averna MR. **The pathophysiology of intestinal lipoprotein production**. *Front Physiol* (2015.0) **6** 61. DOI: 10.3389/fphys.2015.00061 43. Nogueira J-P, Maraninchi M, Béliard S, Padilla N, Duvillard L, Mancini J, Nicolay A, Xiao C, Vialettes B, Lewis GF, Valéro R. **Absence of acute inhibitory effect of insulin on chylomicron production in type 2 diabetes**. *Arterioscler Thromb Vasc Biol* (2012.0) **32** 1039-1044. DOI: 10.1161/ATVBAHA.111.242073 44. Stahel P, Xiao C, Nahmias A, Lewis GF. **(2020) Role of the Gut in Diabetic Dyslipidemia**. *Froniers Endocrinol* (2020.0) **11** 116. DOI: 10.3389/fendo.2020.00116
--- title: Association of clinical, laboratory and imaging biomarkers with the occurrence of acute myocardial infarction in patients without standard modifiable risk factors – rationale and design of the “Beyond-SMuRFs Study” authors: - Dimitrios V. Moysidis - Stylianos Daios - Vasileios Anastasiou - Alexandros C. Liatsos - Andreas S. Papazoglou - Efstratios Karagiannidis - Vasileios Kamperidis - Kali Makedou - Aikaterini Thisiadou - Paraskevi Karalazou - Marios Papadakis - Christos Savopoulos - Antonios Ziakas - George Giannakoulas - Vassilios Vassilikos - Georgios Giannopoulos journal: BMC Cardiovascular Disorders year: 2023 pmcid: PMC10037837 doi: 10.1186/s12872-023-03180-4 license: CC BY 4.0 --- # Association of clinical, laboratory and imaging biomarkers with the occurrence of acute myocardial infarction in patients without standard modifiable risk factors – rationale and design of the “Beyond-SMuRFs Study” ## Abstract ### Background Acute myocardial infarction (AMI) remains the leading cause of mortality worldwide. The majority of patients who suffer an AMI have a history of at least one of the standard modifiable risk factors (SMuRFs): smoking, hypertension, dyslipidemia, and diabetes mellitus. However, emerging scientific evidence recognizes a clinically significant and increasing proportion of patients presenting with AMI without any SMuRF (SMuRF-less patients). To date, there are no adequate data to define specific risk factors or biomarkers associated with the development of AMIs in these patients. ### Methods The ‘‘Beyond-SMuRFs Study’’ is a prospective, non-interventional cohort trial designed to enroll patients with AMI and no previous coronary intervention history, who undergo coronary angiography in two academic hospitals in Thessaloniki, Greece. The rationale of the study is to investigate potential relations between SMuRF-less AMIs and the clinical, laboratory and imaging profile of patients, by comparing parameters between patients with and without SMuRFs. Complete demographic and comprehensive clinical data will be recorded, Venous blood samples will be collected before coronary angiography and the following parameters will be measured: total blood count, standard biochemistry parameters, coagulation tests, hormone levels, glycosylated hemoglobin, N- terminal pro-B-type natriuretic peptide and high-sensitivity troponin T levels- as well as serum levels of novel atherosclerosis indicators and pro-inflammatory biomarkers. Furthermore, all participants will undergo a complete and comprehensive transthoracic echocardiographic assessment according to a pre-specified protocol within 24 h from admission. Among others, 2D-speckle-tracking echocardiographic analysis of cardiac chambers and non-invasive calculation of myocardial work indices for the left ventricle will be performed. Moreover, all patients will be assessed for angiographic parameters and the complexity of coronary artery disease using the SYNTAX score. Multivariable linear and logistic regression models will be used to phenotypically characterize SMuRF-less patients and investigate independent clinical, laboratory, echocardiographic and angiographic biomarkers-predictors of SMuRF-less status in AMI.The first patient was enrolled in March 2022 and completion of enrollment is expected until December 2023. ### Discussion The ‘‘Beyond-SmuRFs’’ study is an ongoing prospective trial aiming to investigate potential clinical, laboratory and imaging biomarkers associated with the occurrence of AMIs in SMuRF-less patients. The configuration of these patients’ profiles could lead to the development of personalized risk-stratification models predicting the occurrence of cardiovascular events in SΜuRF-less individuals. ### Trial Registration ClinicalTrials.gov Identifier: NCT05535582 / September 10, 2022. ## Background Acute myocardial infarction (AMI) remains the leading cause of mortality worldwide [1]. The incidence of coronary artery disease (CAD) and -its most adverse manifestation- AMI, has been proven to rise along with the increasing prevalence of major cardiovascular risk factors, such as obesity, smoking, and hypercholesterolemia [2]. These comorbidities have been well recognized as risk factors of coronary artery disease (CAD) and are often used to evaluate the risk of sustaining an acute coronary event, including AMI. Furthermore, the primary and secondary prevention of AMIs has primarily focused on the modification and treatment of standard modifiable risk factors (SMuRFs), namely smoking, diabetes mellitus, dyslipidaemia, and hypertension [2]. However, recent registries indicate a growing population of patients suffering an acute coronary syndrome (ACS) without any SMuRF (SMuRF-less patients) [3–6]. Over the past decade, the prevalence of such cases among patients presenting with ST-segment elevation myocardial infarction (STEMI) has increased from 13 to $27\%$ in a large, national registry [3]. To date, scientific evidence on the pathogenesis and etiology of SMuRF-less AMI remains limited, although it constitutes an increasingly recognized clinical entity. A popular hypothesis implicates systematic inflammation and high levels of intra-coronary pro-inflammatory cytokines, but data are scarce. Pro-inflammatory and atherosclerosis biomarkers, such as lipoprotein (a) [Lp(a)], C-reactive protein and fibrinogen, have been shown to have higher specificity in predicting worse prognosis of AMI in SMuRF-less patients, as compared to patients with SMuRFs, but no studies have been conducted to investigate their role as predictors of SMuRF-less AMIs [4, 7]. Furthermore, large observational studies have indicated several demographic features potentially linked to SMuRF-less status in AMI, but the results are contradictory, probably due to marked heterogeneity in studied populations [8–10]. Hence, pinpointing sensitive clinical and laboratory parameters, as well as diagnostic approaches, for the prediction of AMI in patients without SMuRFs is crucial, as it concerns an increasing proportion of patients with CAD, who are rather under-represented in registries and clinical trials of cardiovascular risk-assessment [11]. The aim of this study is, therefore, to investigate potential clinical and/or laboratory characteristics associated with SMuRF-less AMIs by comparing the prevalence of clinical parameters and levels of laboratory and imaging indicators among patients with and without SMuRFs. The ultimate goal is the development of a predictive risk stratification model capable of recognizing patients without SMuRFs at high risk for AMI. Secondarily, the study aims at investigating differences in the prognosis of SMuRF-less patients compared to those with SMuRFs. ## Study design and population The ‘‘Beyond-SMuRFs Study’’ (ClinicalTrials.gov Identifier: NCT05535582) is an investigator-initiated, prospective, non-interventional cohort trial involving patients suffering from AMI and undergoing coronary angiography. The study is performed in accordance with the general principles outlined in the Declaration of Helsinki [12] and the rules of good clinical practice (GCP), and has been approved by the Ethics Committee of the Aristotle University of Thessaloniki (reference number: $\frac{136945}{2022}$). A total of 500 consecutive patients presenting with STEMI or Non-STEMI (NSTEMI) at two academic hospitals in Thessaloniki, Greece, undergoing primary or emergency coronary angiography, will be enrolled in the study. All eligible participants will provide informed written consent before enrollment. Patients with a history of previous AMI or previous coronary intervention, either percutaneous or surgical, will be excluded, as the calculation of the SYNTAX score for these patients is not possible. Detailed eligibility criteria are described in ​Table 1. Table 1Inclusion and exclusion criteria of the study populationInclusion CriteriaExclusion Criteria• Age > 18 years• Hospitalization for acute myocardial infarction (AMI) with or without ST elevation (based on the Fourth Universal Definition of Myocardial Infarction) within the previous 4 weeks• Coronary angiography before or after hospitalization for AMI, in which at least one stenosis > $50\%$ in a major epicardial coronary artery (left anterior descending artery, left circumflex artery, right coronary artery) or a branch thereof with a diameter of at least 2 mm was observed.• Inability or refusal to provide informed consent• Age > 80 years• History of hospitalization due to AMI prior to the present AMI• History of coronary revascularization prior to the present AMI• Previous coronary angiography (prior to the present AMI) showing > $50\%$ stenosis in a major epicardial coronary artery Patients will be divided into two groups based on their medical history: (i) Group A: SMuRF-less patients, (ii) Group B: Patients with SMuRFs, defined as those who fulfilled at least one of the following criteria: (i) known history of hypertension and/or antihypertensive treatment prior to AMI, (ii) self-reported use of tobacco products on a systematic basis for up to 12 months before AMI, (iii) history of diabetes mellitus type 1 or 2 and/or treatment with antidiabetic tablets or insulin before AMI or diagnosis of diabetes mellitus based on HbA1c during AMI hospitalization, (iv) known hypercholesterolemia (total cholesterol > 200 mg/dl / LDLc > 150 mg/dl) or treatment with statins or PCSK9is, before AMI. SMuRF-less patients (Group A) are defined as those suffering an AMI in the absence of these comorbidities. ## Data collection and study procedures After obtaining written informed consent, the following clinical characteristics will be recorded for each patient: demographics, socioeconomic parameters, complete medical history and medication, prior diagnostic and therapeutic interventions. A self-reported measurement of patients’ physical activity will be provided using the International Physical Activity Questionnaire (IPAQ), which will be completed by each patient before discharge. IPAQ is a validated questionnaire utilized to objectively measure and stratify physical activity by dividing populations into three levels: low, moderate and high physical activity [13]. Obesity will be diagnosed by categorization of body mass index (BMI; Kg/m2) measured before echocardiographic study. Recorded socioeconomic parameters will include education, marital status, employment, income, migration background and ethnicity. Moreover, the 36-item short form (SF-36) standardized questionnaire will be administered to obtain a self-reported measure of their health-related perception of quality of life before the AMI. In addition, patient laboratory data will be recorded on admission and during hospitalization. These include total blood count, standard biochemistry parameters, coagulation tests, thyroid hormone and thyroid-stimulating hormone levels, HbA1c, NTproBNP and HsTnT levels on admission, peak values of HsTnT, and NTproBNP. Moreover, levels of LP(a), apolipoproteins B and A1 (ApoB and ApoA1) interleukin-6 (IL-6) and soluble urokinase plasminogen activator receptor (suPAR) on admission will be assessed. Additionally, coronary angiographic images of every patient will be evaluated by two experienced independent interventional cardiologists blinded as to the demographic and clinical patient characteristics. Angiographic parameters such as lesion characteristics, coronary dominance and the SYNTAX score will be calculated. A complete and comprehensive transthoracic echocardiographic assessment (TTE) will be performed within 24 hours from admission. All TTE studies will be conducted by certified sonographers/cardiologists using high-end scanners (e.g. Vivid E95, GE Healthcare, Chicago, IL, USA). All analyses will be separately performed by two dedicated expert cardiologists, blinded to the clinical data of all participants. All cardiac chamber sizing quantification, two-dimensional (2-D) and Doppler measurements will be performed in accordance with current recommendations [14, 15]. Simpson’s biplane method will be employed for the calculation of left ventricular ejection fraction (LVEF) for the left ventricle (LV) and abnormal values of conventional LV diastolic parameters will be determined based on recently published criteria [15]. LV diastolic function parameters include mitral inflow and annular velocities and the derived trans-mitral to averaged septal and lateral annular early diastolic velocity ratio (E/e’). All right ventricular (RV) systolic function parameters including tricuspid annular plane systolic excursion (TAPSE), fractional area change (FAC), systolic movement of the RV lateral wall using tissue Doppler imaging (S’), pulmonary artery systolic pressure (PASP) will be evaluated as per current guidelines [16]. Two-dimensional (2D) speckle tracking echocardiography will be employed to calculate strain measurements for the LV, left atrium (LA), and RV. Global longitudinal strain (GLS) will be derived from the calculation of the average of the peak systolic longitudinal strain of all segments for each chamber. To estimate myocardial work indices, LV GLS and peak systolic LV pressure measurements will be integrated to the aforementioned module [17]. Four different indices of myocardial work will be calculated including (i) LV global work index (LVGWI, mmHg %), representing the total work within the LV pressure-strain loops, (ii) LV global constructive work (LVGCW, mmHg %), defined as the work performed during myocardial shortening in systole and the work during myocardial lengthening in isovolumic relaxation, (iii) LV global wasted work (LVGWW, mmHg %), representing the work contributing to the lengthening of the cardiac myocytes during systole and the shortening during isovolumic relaxation, and (iv) LV global work efficiency (LVGWE, %), defining the percentage of effectively spend work by the LV myocytes and obtained by the following formula: (LVGCW/[LVGCW + LVGWW]) × $100\%$. The primary outcome of the study is to compare clinical, laboratory and imaging parameters among SMuRF-less patients and patients with SMuRFs, thereby exploring clinical, laboratory, echocardiographic and angiographic biomarkers potentially associated with SMuRF-less status in AMI. Subsequently, we aim to assess these parameters as potential independent predictors of SMuRF-less AMIs (logistic regression analysis). Secondary goals include a comparison of short- and long-term mortality and major adverse cardiovascular events (MACE), as well as of the complexity and severity of CAD, between AMI patients with and without history of SMuRFs (Fig. 1). Short-term outcomes include all-cause death and MACE during hospitalization and/or 30 days after hospital admission. Patients will also be followed-up for a median period of 24 months after enrollment to evaluate long-term prognosis. All deaths will be ascertained by searching in the Greek web-based national health insurance system. Apart from death, MACE will be documented by independent physicians either through hospital reports or via in-person or telephonic interviews. The first patient was enrolled in April 2022 and completion of enrolment is expected until August 2023. Fig. 1Visual Overview of the ‘‘Beyond-SMuRFs Study’’. Study processes required for the association of SMuRF-less myocardial infarctions with patients’ clinical, laboratory and imaging biomarkers (primary outcome) are depicted. (* This figure is original and, therefore, permission for publication was not needed to be obtained from a third-party) ## Statistical analysis Clinical parametersc laboratory findings and imaging indices of interest will be compared among patients with and without SMuRFs to phenotypically characterize the SMuRF-less group, and identify any particular associations of these biomarkers with the SMuRF-less status. Subsequently, univariate logistic regression analysis will be performed to identify independent predictors of SMuRF-less AMIs. A multivariate logistic regression model will be constructed by forcing univariably significant and clinically relevant variables into the multivariable model. The G*Power software was utilized by a specialist statistician in order to calculate the sample size required to derive statistical significance. It was estimated that to detect an odds ratio > 2 with a prevalence of $15\%$ of SMuRF-less status among AMIs, approximately 75 SMuRF-less patients will be required (with a probability for Type I error: 0.80, and statistical significance level: 0.05). Therefore, 500 patients (with and without SMuRFs) will be included in the present study. An odds ratio > 2 will be interpreted as a prediction of a 2-fold higher probability of having SMuRF-less AMI than an AMI attributed to SMuRFs. Baseline patient characteristics of each group will be examined and compared using the chi-square test for categorical variables and the 2-sided Student’s t-test for continuous variables or non-parametric tests (Wilcoxon, Mann-Whitney U etc.), when assumptions of normality are not met. Categorical variables will be represented by frequencies and percentages (%) and continuous variables will be summarized by mean ± standard deviation (SD) or median (1st -3rd quartile), as appropriate. In terms of statistical tests recruited for outcomes calculation, the two groups of patients will be compared to each other to identify clinical, laboratory and imaging biomarkers associated with a higher or lower probability of developing a SMuRF-less AMI. The chi-square test will be used to compare the prevalence of clinical parameters among patients with and without SMuRFs, and the t-test and Mann-Whitney U test will be utilized to compare the mean levels of laboratory and imaging biomarkers between the two groups. As mentioned before, univariate linear and logistic regression analyses will be applied to calculate univariate odds ratios, and then multivariate regression analyses will be performed. Finally, in order to identify the optimal combination of epidemiological, clinical, laboratory and imaging biomarkers associated with SMuRF-less AMIs, supervised machine learning algorithms will be used. The resulting clinical-laboratory prediction models will be evaluated with Receiver Operating Characteristic (ROC) curves. A time-to-event analysis will be performed to assess whether the absence of SMuRFs is associated with better or worse clinical prognosis of patients (secondary outcome). Event rates will be compared by the long-rank test. Μultivariable Cox proportional hazard models will be utilized to adjust the results for clinically relevant and univariately significant variables. The 2-tailed p value of 0.05 will be considered the significance threshold for all statistical tests. All outcomes will be reported with $95\%$ confidence intervals. Data management and statistical analyses will be conducted using SPSS software, version 26 (IBM SPSS Statistics) and R version 3.4.4 (R Foundation for Statistical Computing, Vienna, Austria). ## Discussion ‘’Beyond-SMuRF’’ is a prospective cohort trial, enrolling recently hospitalized patients with AMI and aspiring to identify clinical, laboratory and imaging biomarkers associated with SMuRF-less status. So far, suggested polygenic clinical risk-score models seem to underestimate the risk in SMuRF-less patients and are usually incapable of identifying cardiovascular risk factors on top of SMuRFs [18]. Therefore, the rationale of this study is -by comparing clinical, laboratory, echocardiographic and angiographic parameters between the SMuRF-less group and the group of patients with SMuRFs- to assess novel biomarkers as predictors of SMuRF-less AMIs and ultimately generate a risk-stratification tool for this increasingly recognized entity. To our knowledge, this is the first study to comprehensively attempt an in-depth, multiparametric evaluation of the profile of SMuRF-less population with AMI. Although several studies have reported clinical outcomes in SMuRF-less patients, there is a lack of evidence regarding clinical, laboratory and imaging findings in these patients. Several studies have reported baseline demographic characteristics among patients with and without SMuRFs [3, 11, 18, 19]. Ηowever, no marked differences have been explored so far between the two groups in terms of basic demographic features such as age and sex. Moreover, differences in baseline clinical characteristics and comorbidities have not been thoroughly evaluated on a systematic basis, which adds to the novelty of this study. For instance, rheumatic and autoimmune diseases, as well as abdominal obesity, alcohol consumption and drug use, have been associated with increased cardiovascular risk but have never been assessed as potential absolute explainers of SMuRF-less AMIs [18, 20]. Furthermore, mental health status and psychosocial risk factors have been linked to CAD, but data regarding the association thereof with SMuRF-less status in AMI are lacking [20–24]. Additionally, the lack of physical activity, as well as specific socioeconomic parameters, have been proven drivers of CAD, but never evaluated in the context of SMuRF-less AMIs [24–26]. Moreover, the laboratory profile of SMuRF-less patients remains understudied. *In* general, there is an absence of available and specific blood work-up biomarkers of CAD beyond markers indicating SMuRFs. In a retrospective analysis of the SWEDEHEART registry, the authors found lower body-mass index, lower triglyceride concentrations, and higher HDL-C concentrations in the SMuRF-less group versus patients with SMuRFs, which probably suggests that these factors might not drive the atherosclerosis [27]. Only few studies have been conducted to indicate blood markers associated with AMI in these patients, but none of them showed remarkable results or studied non-conventional biomarkers [4, 27, 28]. Consistent with the hypothesis that a significant number of mechanisms underlying atherosclerosis and ACS without SMuRFs remains undiscovered, our study will be the first to elaborate metabolomic and inflammatory biomarkers, such as Lp(a), ApoB and ApoA1 interleukin-6 (IL-6) and suPAR. Although an increasing number of registries assess the impact of such biomarkers on patients with AMI, none of them have focused on SMuRF-less patients [29–31]. The logic behind this effort is that the metabolic and inflammatory status of patients have been shown to play a pivotal role in cardiovascular disease and specifically in CAD and ACS [32–40]. On top of that, genetic analyses, Mendelian randomization studies, and the determination of specific polymorphism responsible for atherogenesis in SMuRF-less patients should drive future research to elucidate additional pathogenetic aspects, and develop -potentially with the use of artificial intelligence- clinically relevant polygenic risk scores [41, 42]. Regarding imaging biomarkers, the innovation of the study lies in the fact that it recruits conventional and novel echocardiographic parameters and analytical angiographic assessment to unravel differences among SMuRF-less patients compared to patients with SMuRFs. As novel imaging modalities have proved, plaque vulnerability and rupture electromechanical complications are not always directly associated with systematically assessed biomarkers and risk factors such as dyslipidaemia [43, 44]. Moreover, an observational study by Figtree et al. indicated differences in the prevalence of left main and left anterior descending culprit lesions in SMuRF-less patients compared to those with SMuRFs [27]. As explained by the authors, this may be partly attributed to the family history of premature CAD and is an important finding potentially leading to a more adverse AMI risk profile in SMuRF-less patients. A potential criticism that our study could deal with is that differences in imaging parameters between patients with and without SMuRFs are likely to be the consequence of AMI rather than a predictive parameter. However, the main objective of the study is to identify associations of clinical, laboratory and imaging variables with SMuRF-less status which should be tested in future larger trials for their predictive value. Thus, the elaboration of additional biomarkers, such as imaging parameters, could contribute to the characterization of the clinical profile of SMuRF-less patients hospitalized with AMI, and shed light on the etiology of this emerging clinical entity. Finally, this study aims to provide data on the prognosis of SMuRF-less AMIs and evaluate emerging evidence highlighting the worse clinical course of these patients [3, 5, 6, 45–49]. The reason behind this observation could be correlated with increased time-to-reperfusion time due to reduced suspicion of AMI in these individuals, but also with undiscovered pathogenetic mechanisms and ‘‘hidden’’ comorbidities which might increase their cardiovascular risk. ## Limitations Certain limitations of this study should be properly acknowledged. The main limitation is its observational nature which does not allow to conclude causal associations. Second, this study is not a multi-center one and its population consists of Greek patients exclusively. Future studies should be conducted to include and study other ethnicities and races to account for inherent variability of different patient populations and test the generalizability of our results. Moreover, the authors set an upper age limit as inclusion criterion in the study in an attempt to exclude very old patients with age-relating comorbidities and intricated AMI pathophysiology. As there are no evidence on cutoff values to support this limit, this could lead to selection bias in the study and ambiguous resuls across different age groups. Finally, our study -due to restricted resources- will study only some of the potential predictive laboratory biomarkers. Comprehensive and complete analysis of SMuRF-less patients’ metabolomic and genetic profiling should be taken into consideration for future research. This real-world, prospective, non-interventional cohort trial study of patients hospitalized with AMI has the potential to identify clinical, laboratory and imaging biomarkers associated with the occurrence of AMI in SMuRF-less patients. SMuRF-less individual profiling could ultimately lead to the development of personalized risk-stratification models predicting adverse cardiovascular events. and acronyms: ACS = acute myocardial infarction; CAD = coronary artery disease; NSTEMI = Non-ST-segment elevation myocardial infarction; SmuRF = standard modifiable risk factors; STEMI = ST-segment elevation myocardial infarction. ## References 1. Roth GA, Mensah GA, Johnson CO. **Global Burden of Cardiovascular Diseases and Risk factors, 1990–2019: Update from the GBD 2019 study**. *J Am Coll Cardiol* (2020.0) **76** 2982-3021. DOI: 10.1016/j.jacc.2020.11.010 2. Visseren FLJ, Mach F, Smulders YM. **2021 ESC Guidelines on cardiovascular disease prevention in clinical practice**. *Eur Heart J* (2021.0) **42** 3227-337. DOI: 10.1093/eurheartj/ehab484 3. Vernon ST, Coffey S, Bhindi R. **Increasing proportion of ST elevation myocardial infarction patients with coronary atherosclerosis poorly explained by standard modifiable risk factors**. *Eur J Prev Cardiol* (2017.0) **24** 1824-30. DOI: 10.1177/2047487317720287 4. Vernon ST, Coffey S, D’Souza M. **ST-Segment-Elevation myocardial infarction (STEMI) patients without Standard Modifiable Cardiovascular risk factors-how common are they, and what are their outcomes?**. *J Am Heart Assoc* (2019.0) **8** e013296. DOI: 10.1161/JAHA.119.013296 5. Figtree GA, Redfors B, Kozor R. **Clinical outcomes in patients with ST-Segment Elevation MI and No Standard Modifiable Cardiovascular Risk factors**. *JACC Cardiovasc Interv* (2022.0) **15** 1167-75. DOI: 10.1016/j.jcin.2022.03.036 6. Papazoglou AS, Farmakis IT, Zafeiropoulos S. **Angiographic severity in acute coronary syndrome patients with and without standard modifiable risk factors**. *Front Cardiovasc Med* (2022.0) **9** 934946. DOI: 10.3389/fcvm.2022.934946 7. Wang L, Cong H-L, Zhang J-X. **Prognostic performance of multiple biomarkers in patients with acute coronary syndrome without standard cardiovascular risk factors**. *Front Cardiovasc Med* (2022.0) **9** 916085. DOI: 10.3389/fcvm.2022.916085 8. Moledina SM, Rashid M, Nolan J. **Addressing disparities of care in non-ST-segment elevation myocardial infarction patients without standard modifiable risk factors: insights from a nationwide cohort study**. *Eur J Prev Cardiol* (2022.0) **29** 1084-92. DOI: 10.1093/eurjpc/zwab200 9. Chen X, Liu H-X, Yu X-Q, Yang S-Q, Qi L-Y, Cai L. **Standard Modifiable Cardiovascular Risk factors and prognosis of Acute Coronary syndrome in younger patients**. *J Coll Physicians Surg Pak* (2021.0) **31** 1394-8. DOI: 10.29271/jcpsp.2021.12.1394 10. Vernon ST, Hansen T, Kott KA, Yang JY, O’Sullivan JF, Figtree GA. **Utilizing state-of-the-art “omics” technology and bioinformatics to identify new biological mechanisms and biomarkers for coronary artery disease**. *Microcirculation* (2019.0) **26** e12488. DOI: 10.1111/micc.12488 11. Kong G, Chew NWS, Ng CH. **Prognostic outcomes in Acute myocardial infarction patients without standard modifiable risk factors: a multiethnic study of 8,680 asian patients**. *Front Cardiovasc Med* (2022.0) **9** 869168. DOI: 10.3389/fcvm.2022.869168 12. **Declaration of Helsinki: ethical principles for medical research involving human subjects**. *JAMA* (2013.0) **310** 2191-4. DOI: 10.1001/jama.2013.281053 13. Rzewnicki R, Vanden Auweele Y, De Bourdeaudhuij I. **Addressing overreporting on the International Physical Activity Questionnaire (IPAQ) telephone survey with a population sample**. *Public Health Nutr* (2003.0) **6** 299-305. DOI: 10.1079/PHN2002427 14. Lang RM, Badano LP, Mor-Avi V. **Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging**. *J Am Soc Echocardiogr Off Publ Am Soc Echocardiogr* (2015.0) **28** 1-39e14. DOI: 10.1016/j.echo.2014.10.003 15. Nagueh SF, Smiseth OA, Appleton CP. **Recommendations for the evaluation of left ventricular diastolic function by Echocardiography: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging**. *J Am Soc Echocardiogr Off Publ Am Soc Echocardiogr* (2016.0) **29** 277-314. DOI: 10.1016/j.echo.2016.01.011 16. Rudski LG, Lai WW, Afilalo J. **Guidelines for the echocardiographic assessment of the right heart in adults: a report from the American Society of Echocardiography endorsed by the European Association of Echocardiography, a registered branch of the European Society of Cardiology, and**. *J Am Soc Echocardiogr Off Publ Am Soc Echocardiogr* (2010.0) **23** 685-8. DOI: 10.1016/j.echo.2010.05.010 17. Russell K, Eriksen M, Aaberge L. **A novel clinical method for quantification of regional left ventricular pressure-strain loop area: a non-invasive index of myocardial work**. *Eur Heart J* (2012.0) **33** 724-33. DOI: 10.1093/eurheartj/ehs016 18. Lønnebakken MT. **The risk of no risk in STEMI**. *Lancet (London England)* (2021.0) **397** 1039-40. DOI: 10.1016/S0140-6736(21)00375-5 19. Rasmussen TB, Borregaard B, Palm P. **Patient-reported outcomes, sociodemographic and clinical factors are associated with 1-year mortality in patients with ischemic heart disease-findings from the DenHeart cohort study**. *Qual life Res an Int J Qual life Asp Treat care Rehabil* (2022.0) **31** 389-402. DOI: 10.1007/s11136-021-02956-5 20. Gerdts E, Regitz-Zagrosek V. **Sex differences in cardiometabolic disorders**. *Nat Med* (2019.0) **25** 1657-66. DOI: 10.1038/s41591-019-0643-8 21. 21.Rutledge T, Redwine LS, Linke SE, Mills PJ. A Meta-Analysis of Mental Health Treatments and Cardiac Rehabilitation for Improving Clinical Outcomes and Depression Among Patients With Coronary Heart Disease. Psychosom Med. 2013;75(4). https://journals.lww.com/psychosomaticmedicine/Fulltext/2013/05000/A_Meta_Analysis_of_Mental_Health_Treatments_and.3.aspx. 22. Berg SK, Rasmussen TB, Thrysoee L. **Mental health is a risk factor for poor outcomes in cardiac patients: findings from the national DenHeart survey**. *J Psychosom Res* (2018.0) **112** 66-72. DOI: 10.1016/j.jpsychores.2018.07.002 23. Yusuf S, Hawken S, Ounpuu S. **Effect of potentially modifiable risk factors associated with myocardial infarction in 52 countries (the INTERHEART study): case-control study**. *Lancet (London England)* (2004.0) **364** 937-52. DOI: 10.1016/S0140-6736(04)17018-9 24. Timmis A, Townsend N, Gale CP. **European Society of Cardiology: Cardiovascular Disease Statistics 2019**. *Eur Heart J* (2020.0) **41** 12-85. DOI: 10.1093/eurheartj/ehz859 25. 25.Stomby A, Strömberg S, Theodorsson E, Olsen Faresjö Ã, Jones M, Faresjö T. Standard Modifiable Cardiovascular risk factors mediate the Association between elevated hair cortisol concentrations and coronary artery disease. Front Cardiovasc Med. 2022;8. 10.3389/fcvm.2021.765000. https://www.frontiersin.org/articles/. 26. Khan SU, Nguyen RT, Javed Z. **Socioeconomic status, cardiovascular risk profile, and premature coronary heart disease**. *Am J Prev Cardiol* (2022.0) **11** 100368. DOI: 10.1016/j.ajpc.2022.100368 27. Figtree GA, Vernon ST, Hadziosmanovic N. **Mortality in STEMI patients without standard modifiable risk factors: a sex-disaggregated analysis of SWEDEHEART registry data**. *Lancet* (2021.0) **397** 1085-94. DOI: 10.1016/S0140-6736(21)00272-5 28. Blaum C, Brunner FJ, Kröger F. **Modifiable lifestyle risk factors and C-reactive protein in patients with coronary artery disease: implications for an anti-inflammatory treatment target population**. *Eur J Prev Cardiol* (2021.0) **28** 152-8. DOI: 10.1177/2047487319885458 29. 29.Kohlhauer M, Dawkins S, Costa ASH, et al. Metabolomic profiling in Acute ST-Segment-Elevation myocardial infarction identifies succinate as an early marker of Human Ischemia-Reperfusion Injury. J Am Heart Assoc. 2018;7(8). 10.1161/JAHA.117.007546. 30. Guo M, Fan X, Tuerhongjiang G. **Targeted metabolomic analysis of plasma fatty acids in acute myocardial infarction in young adults**. *Nutr Metab Cardiovasc Dis* (2021.0) **31** 3131-41. DOI: 10.1016/j.numecd.2021.06.024 31. Karagiannidis E, Sofidis G, Papazoglou AS. **Correlation of the severity of coronary artery disease with patients’ metabolic profile- rationale, design and baseline patient characteristics of the CorLipid trial**. *BMC Cardiovasc Disord* (2021.0) **21** 79. DOI: 10.1186/s12872-021-01865-2 32. Blake GJ, Ridker PM. **Inflammatory bio-markers and cardiovascular risk prediction**. *J Intern Med* (2002.0) **252** 283-94. DOI: 10.1046/j.1365-2796.2002.01019.x 33. Papazoglou AS, Kartas A, Moysidis DV. **Glycemic control and atrial fibrillation: an intricate relationship, yet under investigation**. *Cardiovasc Diabetol* (2022.0) **21** 39. DOI: 10.1186/s12933-022-01473-0 34. McGarrah RW, Crown SB, Zhang G-F, Shah SH, Newgard CB. **Cardiovascular Metabolomics**. *Circ Res* (2018.0) **122** 1238-58. DOI: 10.1161/CIRCRESAHA.117.311002 35. Karagiannidis E, Moysidis DV, Papazoglou AS. **Prognostic significance of metabolomic biomarkers in patients with diabetes mellitus and coronary artery disease**. *Cardiovasc Diabetol* (2022.0) **21** 70. DOI: 10.1186/s12933-022-01494-9 36. Sabatine MS, Liu E, Morrow DA. **Metabolomic identification of novel biomarkers of myocardial ischemia**. *Circulation* (2005.0) **112** 3868-75. DOI: 10.1161/CIRCULATIONAHA.105.569137 37. Ali SE, Farag MA, Holvoet P, Hanafi RS, Gad MZ. **A comparative Metabolomics Approach reveals early biomarkers for metabolic response to Acute myocardial infarction**. *Sci Rep* (2016.0) **6** 36359. DOI: 10.1038/srep36359 38. Stalikas N, Papazoglou AS, Karagiannidis E. **Association of stress induced hyperglycemia with angiographic findings and clinical outcomes in patients with ST-elevation myocardial infarction**. *Cardiovasc Diabetol* (2022.0) **21** 140. DOI: 10.1186/s12933-022-01578-6 39. Seropian IM, Sonnino C, Van Tassell BW, Biasucci LM, Abbate A. **Inflammatory markers in ST-elevation acute myocardial infarction**. *Eur Hear Journal Acute Cardiovasc Care* (2016.0) **5** 382-95. DOI: 10.1177/2048872615568965 40. Oprescu N, Micheu MM, Scafa-Udriste A, Popa-Fotea N-M, Dorobantu M. **Inflammatory markers in acute myocardial infarction and the correlation with the severity of coronary heart disease**. *Ann Med* (2021.0) **53** 1041-7. DOI: 10.1080/07853890.2021.1916070 41. Moysidis DV, Papazoglou AS, Karagiannidis E. **The SMuRFs and the SMuRF-Less Gargamel**. *JACC Cardiovasc Interv* (2022.0) **15** 1886. DOI: 10.1016/j.jcin.2022.07.025 42. Mittas N, Chatzopoulou F, Kyritsis KA. **A risk-stratification machine learning Framework for the prediction of coronary artery Disease Severity: insights from the GESS Trial**. *Front Cardiovasc Med* (2021.0) **8** 812182. DOI: 10.3389/fcvm.2021.812182 43. Massimo V, Giovanna G. **Myocardial infarction in patients without Cardiovascular Risk factors**. *JACC Cardiovasc Interv* (2022.0) **15** 1176-8. DOI: 10.1016/j.jcin.2022.04.040 44. Papazoglou AS, Karagiannidis E, Moysidis DV. **Current clinical applications and potential perspective of micro-computed tomography in cardiovascular imaging: a systematic scoping review**. *Hell J Cardiol* (2021.0) **62** 399-407. DOI: 10.1016/j.hjc.2021.04.006 45. Figtree GA, Vernon ST. **Coronary artery disease patients without standard modifiable risk factors (SMuRFs)- a forgotten group calling out for new discoveries**. *Cardiovasc Res* (2021.0) **117** e76-8. DOI: 10.1093/cvr/cvab145 46. Del Gonzalez MI, Peiro Ibanez OM, Vaquez-Nunez K. **The absence of standard modifiable cardiovascular risk factors does not predict better outcomes in patients with acute coronary syndrome**. *Eur Heart J* (2020.0) **41** ehaa9461341. DOI: 10.1093/ehjci/ehaa946.1341 47. 47.Sia C-H, Ko J, Zheng H, et al. Comparison of mortality outcomes in Acute myocardial infarction patients with or without Standard Modifiable Cardiovascular Risk factors. Front Cardiovasc Med. 2022;9. 10.3389/fcvm.2022.876465. https://www.frontiersin.org/articles/. 48. Shrestha B, Mochon A, Poudel B, Poudel D, Donato A. **Trends and Outcomes of ST-Segment-Elevation MI in Hospitalized Patients without Standard Modifiable Cardiovascular Risk factors**. *Curr Probl Cardiol* (2022.0) **47** 101271. DOI: 10.1016/j.cpcardiol.2022.101271 49. Li S, Gao X, Yang J. **Number of standard modifiable risk factors and mortality in patients with first-presentation ST-segment elevation myocardial infarction: insights from China Acute myocardial infarction registry**. *BMC Med* (2022.0) **20** 217. DOI: 10.1186/s12916-022-02418-w
--- title: Ratiometric electrochemical OR gate assay for NSCLC-derived exosomes authors: - Fanyu Meng - Wenjun Yu - Minjia Niu - Xiaoting Tian - Yayou Miao - Xvelian Li - Yan Zhou - Lifang Ma - Xiao Zhang - Kun Qian - Yongchun Yu - Jiayi Wang - Lin Huang journal: Journal of Nanobiotechnology year: 2023 pmcid: PMC10037838 doi: 10.1186/s12951-023-01833-2 license: CC BY 4.0 --- # Ratiometric electrochemical OR gate assay for NSCLC-derived exosomes ## Abstract Non-small cell lung cancer (NSCLC) is the most common pathological type of LC and ranks as the leading cause of cancer deaths. Circulating exosomes have emerged as a valuable biomarker for the diagnosis of NSCLC, while the performance of current electrochemical assays for exosome detection is constrained by unsatisfactory sensitivity and specificity. Here we integrated a ratiometric biosensor with an OR logic gate to form an assay for surface protein profiling of exosomes from clinical serum samples. By using the specific aptamers for recognition of clinically validated biomarkers (EpCAM and CEA), the assay enabled ultrasensitive detection of trace levels of NSCLC-derived exosomes in complex serum samples (15.1 particles μL−1 within a linear range of 102–108 particles μL−1). The assay outperformed the analysis of six serum biomarkers for the accurate diagnosis, staging, and prognosis of NSCLC, displaying a diagnostic sensitivity of $93.3\%$ even at an early stage (Stage I). The assay provides an advanced tool for exosome quantification and facilitates exosome-based liquid biopsies for cancer management in clinics. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12951-023-01833-2. ## Introduction Lung cancer (LC) ranks top among all malignancies, leading to a quarter of cancer-related deaths, according to World Health Organization [1]. In particular, non-small-cell lung carcinoma (NSCLC) constitutes about ~ $85\%$ of LC cases, showing a 5-year survival rate as low as 10–$15\%$ only [2–4]. Early diagnosis of NSCLC reduces disease mortality with an enhanced five-year survival rate of up to $85\%$, and precise staging and prognosis relieve disease burden with lowered treatment cost by $61.5\%$ [5, 6]. However, routine methods using tissue biopsy or low-dose computed tomography are limited, concerning nonquantitative examinations, invasive sampling, and poor diagnosis accuracy [7–9]. Therefore, there is an urgent need to develop an assay for clinical diagnosis, staging, and prognosis of NSCLC populations. Blood test is crucial in liquid biopsy industries and contributes to ~ $66\%$ of clinical diagnosis. In particular, blood contains abundant disease biomarkers, considered information-rich and readily available biospecimens suitable for point of care testing [10–16]. In particular, exosomes are membrane-enclosed nanovesicles (30–150 nm) secreted by diverse cell types into various biological fluids, participating in physiological and pathophysiological processes, e.g., immunomodulation and tumor promotion [17–19]. Exosomes serve as potent mediators of intercellular communication and demonstrate key roles during tumorigenesis. Especially, the NSCLC-derived exosomes from tumor cells (e.g., A549, H460, and H1299) carry different expression levels of multitudinous proteins (e.g., epithelial cell adhesion molecule (EpCAM) and carcinoembryonic antigen (CEA)), resulting in distinct surface phenotypes correlated with cancer occurrence and progression [20–22]. Further considering the high concentration (up to 1011 mL−1) and stability in circulating blood, surface protein phenotypes of exosomes are explored as a promising biomarker in liquid biopsies for NSCLC patients [23]. Nowadays, most studies regarding exosome surface proteins mainly focus on discovering biomarkers through mass spectrometry or western blots [24]. Little progress has been made in their adaption to the clinical diagnosis of NSCLC, due to the lack of feasible and accurate profiling tools. Electrochemical sensing assays have been developed for the simple detection of disease biomarkers, while suffering from unsatisfactory specificity and sensitivity in exosome profiling [25–27]. For specificity, human exosomes have more heterogeneous compositions than cell line-derived exosomes, and thus it is difficult to obtain comprehensive information of multiple surface proteins on exosomes [28–30]. DNA logic gates are capable of modeling complicated networks and leveraging valuable information within observed data for the accurate estimation and prediction of practical samples [31–33]. They have shown superior performance in analyzing electrochemical signals in complex samples, thus becoming trustworthy solutions to this issue [34, 35]. For sensitivity, the concentration of exosome biomarkers is ultralow at the early stage of NSCLC (e.g., 7 ng/mL) [30, 36]. Unlike the previous assay using electrochemical absolute values of a single reporter, ratiometric methods measure relative signals by two redox reporters (i.e., target-responsive and drift-correcting reference reporters) with opposite variation, offering higher sensitivity with enhanced response magnitude [37]. In addition, ratiometric signals can be further regulated by DNA-based amplification strategies (e.g., hybridization chain reaction (HCR)), to achieve better analytical performance toward trace abundance of targets [38, 39]. Inspired by the above findings, we presented a ratiometric electrochemical biosensor with the assistance of DNA OR logic gates to form an assay for the detection of multiple NSCLC-derived exosomes (Fig. 1). Two NSCLC-related protein markers, EpCAM and CEA, were selected as targets of exosomes. The DNA probe for OR gate operation contains two parts: an aptamer targeting EpCAM or CEA and an extension region to partially complementary block strands. Upon the recognition of both target proteins on exosome surfaces by OR gate operation probes, block strands were released and separated by the capture of exosome surface proteins via aptamers (Fig. 1A). Then, we acquired the electrochemical readouts as the exosome quantitative signals by targeting to the surface proteins, based on the ratiometric biosensor assisted with DNA OR logic gates and HCR (Fig. 1B). As a proof-of-concept application in clinical diagnostics, we profiled the surface proteins on serum exosomes derived from NSCLC patients and normal controls, suggesting that the differential exosome levels obtained from protein profiling enabling the precise diagnosis, staging, and prognosis of NSCLC (Fig. 1C). Overall, this work establishes a novel liquid biopsy to profile surface proteins and quantify disease-derived exosome in sera, facilitating precision diagnostics of various diseases including but not limited to cancer. Fig. 1Principle of the ratiometric electrochemical OR gate assay for non-small cell lung cancer (NSCLC)-derived exosomes. A Exosomes captured by specific aptamers. B Coupling ratiometric electrochemical biosensor with OR logic gate for exosome detection. C Workflow of serum exosome detection for the diagnosis, staging, and prognosis of NSCLC cohort. Abbreviations: gold nanoparticle (GNP); Aptamers (Apt1/Apt2); Block strands (B1/B2); DNA tetrahedron (T1/T2); Reference electrode (RE); Counter electrode (CE); Working electrode (WE) ## Electrode preparation and electrochemical measurements Gold electrodes (2 mm diameter) were first pretreated with piranha solution ($98\%$ H2SO4:$30\%$ H2O2 = 3:1) for 5 min (caution: danger, violent reaction). Then, the electrodes were mechanically polished with P4000 silicon carbide paper, and then 1-, 0.3-, and 0.05-mm alumina slurry, respectively. The polished electrode was sonicated in ethanol and ultrapure water for 5 min. Next, the electrodes were electrochemically cleaned with 0.5 M H2SO4 for 20 cycles to remove any remaining impurities. Finally, the electrodes were dried for mirror-like surfaces, ready to use as working electrodes (WEs). All electrochemical measurements were carried out on a CHI 660E electrochemical workstation (Chenhua Instruments Co., Shanghai, China). The conventional three-electrode system consisted of a DNA-bound gold WE, an Ag/AgCl reference electrode (RE), and a platinum counter electrode (CE). Differential pulse voltammetry (DPV) measurements were performed in 10 mM phosphate buffered saline (PBS, pH 7.4). Electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV) measurements were performed in 5 mM [Fe(CN)6]3−/4− and 1 M KCl buffer. The experimental parameters were as follows. DPV: scan rate 50 mV/s, sweep range − 0.6–0.7; EIS: bias potential 0.232 V, amplitude 5 mV, frequency range 0.1–100,000 Hz; and CV: sweep range 0.6 to − 0.2 V, scan rate 50 mV/s. The ZSimpWin software was used to fit the Nyquist plots based on a Randles equivalent circuit. ## Aptamer capture of exosome 20 μL of gold nanoparticles (GNPs) were centrifuged at 10,000 rpm for 15 min and resuspended in 1000 μL of ultrapure water for DNA modification. To inactivate aptamer, 1 μM aptamer strand (Apt1/Apt1) and 2 μM block strand (B1/B2) were incubated at 37 ℃ for 1 h to form the Apt1-B1 and Apt2-B2 duplex. Then, a mixture of the DNA duplex (1 μM) at a 1:10 volume ratio was incubated with GNPs at 37 ℃ for 12 h to form a GNPs-DNA structure. The load of oligonucleotides was increased by aging treatment with 2 M NaCl, which was added 6 times every 20 min to a final concentration of 0.1 M. Subsequently, the mixture was centrifuged (10,000 rpm, 10 min) to remove the uncoupled duplex DNA, and the remaining GNPs-DNA conjugates were dispersed in PBS (pH 7.4). Different concentrations of exosomes were added to the solution for 1 h, captured by aptamer strand (Apt1/Apt2) with block strand (B1/B2) released into the solution. After centrifugation, the released block strand (B1/B2) was collected and used for the following experiments. For control experiments, GNPs were modified with one aptamer strand and control strand (C) to form GNPs-DNA structure with Apt1/C or Apt2/C. After the exosome capture, the corresponding released B1 or B2 was collected for further electrochemical detection. ## Fabrication of DNA tetrahedron and exosome detection DNA tetrahedrons (T1/T2) were assembled by four single-stranded DNAs (D1/D2, D3, D4, and D5) by an annealing process. Oligonucleotide sequences used in the experiments were listed in Supplementary Information (Additional file 1: Table S1). Equimolar amounts of the four DNAs were blended in 20 mM Tris–HCl buffer (50 mM MgCl2, pH 8.0). The mixture was heated to 95℃ for 5 min and then slowly cooled down to room temperature for complete hybridization until forming a stable structure. The electrode was modified with 5 μM tetrahedral DNAs (T1/T2, 30 μL) via Au–S bond at room temperature overnight. It was subsequently incubated in 1 mM MCH for 0.5 h to block the nonspecific binding sites, followed by blocking with 5 μM substrate strands (S1-MB/S2-MB) for 1 h. Then, 150 μL of the released block strand (B1/B2), as described above, at various concentrations was incubated with the modified electrode at 37℃ for 1 h and then washed with PBS (10 mM, pH 7.4). Next, 30 μL of a mixture containing 2 μM hairpin DNAs (H1-Fc/H2-Fc) was dropped onto the obtained electrode. The reaction solution was incubated at 37 ℃ for 1.5 h before the electrochemical measurements. The control experiments were prepared separately and performed under the same conditions. All experiments were repeated three times. Polyacrylamide gel electrophoresis (PAGE) analysis was performed to characterize the DNA tetrahedron fabrication. For DNA tetrahedron characterization, 10 μL of sample 1 (5 μM, D1), sample 2 (5 μM, D2), sample 3 (5 μM, D1 + D3), sample 4 (5 μM, D1 + D3 + D4), sample 5 (5 μM, D1 + D3 + D4 + D5), and sample 6 (5 μM, D2 + D3 + D4 + D5) were placed on a $12\%$ polyacrylamide gel. The electrophoresis was performed in 0.5 × Tris-borate-EDTA (pH 8.0) at 100 V constant voltage for 1.5 h. After that, the gel was scanned using a gel imaging analyzer. ## Study population and serum harvesting A total of 135 subjects were recruited in Shanghai Chest Hospital, affiliated to Shanghai Jiao Tong University School of Medicine, including 105 NSCLC patients and 30 normal controls (healthy donors (HD, $$n = 15$$), pneumonia ($$n = 5$$), bronchitis ($$n = 5$$), fibrosis ($$n = 5$$)). All cancerous subjects were verified with pathological results. The tumour was staged according to the international standards for tumour, node, and metastasis (TNM) staging, including 30 stage I, 25 stage II, 25 stage III, and 25 stage IV [40, 41]. NSCLC patients included three cancer types (adenocarcinoma, ADC; squamous-cell carcinoma, SCC; and large-cell carcinoma, LCC). All samples were anonymized, and relevant pathological diagnoses were recorded (Additional file 1: Tables S3–S5). All blood samples were drawn into BD Vacutainer® SST™ Tubes by venipuncture and clotted at room temperature within 40 min. Serum was collected at 3000 × g for 10 min of centrifugation from the blood and immediately stored at – 80 ℃ for further analysis. All the investigation protocols were approved by the Institutional Ethics Committees of Shanghai Chest Hospital, under the approved protocol No. KS22025. All subjects provided informed consent to participate in the study and approved the use of their biological samples for analysis. ## Exosome isolation for clinical assay For cell model investigation, NSCLC cell lines (A549, H460, H1299, H1975, H2030, and Calu-1 cells) were cultured in Dulbecco's modified Eagle's medium (Gibco, USA) supplemented with $10\%$ fetal bovine serum (Gibco) in $5\%$ CO2 in an incubator at 37 °C. At the exponential growth phase, cells were collected with trypsinization and centrifuged at 1000 rpm for 5 min. Exosomes were extracted from the cancer cells using a miRCURY® Exosome Kits (Qiagen, Hilden, Germany). The samples were counted and characterized by NanoFCM Flow NanoAnalyzer (NanoFCM Inc., Xiamen, China), and then stepwise diluted into solutions to different final concentrations. For clinical serum applications, the exosomes from the serum of HDs, benign diseases, and NSCLC patients were extracted using an ExoQuick™ Kit (System Biosciences, USA). Extracted exosomes were firstly spiked into the provided buffer by 100-fold dilutions to different concentrations, for better assessment of the practical utility of the assay. ## Statistical analysis All statistical analyses in this work (including significance analyses, receiver operating characteristic curve (ROC) construction, and area under curve (AUC) calculation) were performed on IBM SPSS Statistics software (Version 26.0.0). Sensitivity referred to the probability that the assay result indicated "positive" among all cancerous subjects. Specificity was the fraction of those without cancer, which showed a negative assay result. The $95\%$ confidence intervals (CIs) were calculated using a binomial distribution based on receiver operator characteristic analysis. The significant difference was calculated using a two-tailed Student's t-test, with all significance levels set as $5\%$. Three independent experiments were performed with data shown as the mean ± SD ($$n = 3$$). Plots and charts were performed using Origin 2021 software. ## Principle for exosome detection Figure 1 illustrates the assay principle to quantify NSCLC-derived exosomes as a function of surface biomarker expression. CEA and EpCAM, dual biomarkers enriched on the surface of cancer exosomes, were selected as specific targets to achieve this goal [42]. The strategy consists of three steps: [1] construction of DNA OR logic gate for signal input: exosomes competitively bind to CEA and EpCAM aptamers (Apt1/Apt2) and release block strands (B1/B2) from the aptamer-block hybridization complex. [ 2] Triggering of DNA tetrahedron (T1/T2) for signal output and introduction of HCR for signal amplification: on the electrode surface, the released block strands (B1/B2) hybridize with the primer strands (D1/D2) of the DNA tetrahedron (T1/T2) and displace the substrate strand (S1-MB/S2-MB), generating the decrease of MB intensity (IMB). Then, block strands activate HCR with hairpin DNAs (H1-Fc/H2-Fc) to produce long double-stranded DNAs, resulting in a significant signal enhancement of Fc (IFc). Finally, the ratiometric intensity of Fc and MB signal (IFc/IMB) is used as signal output, corresponding to the level of exosome. [ 3] Analysis of serum exosome for diagnostic application: exosomes were detected from serum samples of NSCLC patients and correlated to different disease status, including NSCLC patients with stage I-IV, BDs, and HDs. For signal input, DNA OR logic gate was designed for multiple exosome biomarkers (CEA and EpCAM) in parallel, to obtain comprehensive tumor information toward precision medicine. In detail, gold nanoparticles (GNPs) were functionalized with particular aptamers (Apt1/Apt2) targeting CEA and EpCAM of exosome to form a DNA OR logic gate. Notably, block strands consist of two parts: one is partially complementary with the aptamer sequence and another is complementary to the primer strands of the DNA tetrahedron. The designed aptamers (Apt1/Apt2) had a higher affinity with exosomes compared with the block strands, leading to the release of block strands (B1/B2) from the aptamer-block hybridization complex. Therefore, the exosome content was converted to the level of released B1/B2. For signal output, DNA tetrahedron with controlled shape was functionalized on electrode surface to detect the specific binding of signal strands, avoiding steric hindrance and molecular entanglement. In particular, the working electrode (WE) was incubated with substrate strands (S1-MB/S2-MB), in which MB was introduced as reference signal. Once B1/B2 strands hybridize with the primer strands and replace the substrate strands, the MB signal will decrease. As a result, the decrement extent of the reference signal (IMB) can serve as an indicating readout for exosome contents. During the test process, the existence of either one single biomarker from exosome surface or two biomarkers led to an electrochemical signal (IFc/IMB) output. Only when there is no existence of any biomarker, the DNA OR logic gate provide no electrochemical signal output. For signal amplification, the trace abundance of recognition biomarkers on exosome surface places obstacle for direct detection [43]. Therefore, we designed a cascade amplification via HCR for the enhancement of Fc signal (IFc). It is worth noting that the intensity of *Fc is* accompanied by a signal decline of referenced MB, further amplifying the ratiometric current (IFc/IMB) for accurate detection of trace exosomes. Specifically, the clinical application of the assay is not limited to LC diagnosis. In another diagnostic scenario targeting different exosome biomarkers (e.g., HER2 for breast cancer), the operation is facile by simply changing the aptamer sequences and the corresponding block strands. Therefore, universal disease diagnosis can be realized through the substitution of multiple aptamers in the assay. ## Characterization of exosome, GNP, and DNA tetrahedron By using a commercial extraction kit, exosomes with sizes ranging from 30 to 150 nm were isolated from both NSCLC cell culture media and clinical serum biospecimens of HDs and patients with NSCLC. A flatted round shape was observed in the transmission electron microscopy (TEM) images of exosomes (Fig. 2A). We also examined size parameters of the exosomes extracted from A549 cells by nanoflow cytometry approach. The acquired morphology revealed an average size of ∼100 nm of exosomes (Fig. 2B), consistent with the TEM result as well as previous literature confirming the successful extraction of exosomes [44].Fig. 2Characterization of exosome, GNP, and DNA tetrahedron. A Transmission electron microscopic (TEM) image and B nanoparticle flow cytometry (NanoFCM) characterization of exosomes. C TEM image and D dynamic light scattering analysis of gold nanoparticles (GNPs) dispersed in water at room temperature. E Ultraviolet–visible spectroscopy absorption spectra and F Zata potential results of the GNPs and aptamer-functionalized GNPs (GNPs-DNA). G Assembly process of DNA tetrahedrons (T1/T2). H Polyacrylamide gel electrophoresis analysis of T1/T2 (formed by D1/D2, D3, D4 and D5) GNPs are critical in the aptamer recognition and capture process (see Methods for details). As shown in TEM image, well-distributed spheres were observed possessing an average diameter of 13 ± 3 nm (Fig. 2C). In the dynamic light scattering experiment, GNPs displayed an average hydrodynamic size of 18 ± 3 nm (Fig. 2D) agreed to TEM characterizations, demonstrating a typical particle size in use according to literature [45]. In a typical ultraviolet–visible absorption spectrum, the light absorption band of GNPs occurred at 520 nm. The new absorption peak at 260 nm, a characteristic band of nucleic acid composites, verified the satisfied functionalization of GNPs with thiol-aptamers (Fig. 2E), in regards to the binding affinity between thiol groups and gold [46]. In addition, GNPs were negatively charged with zeta potential of 31 ± 3 mV (Fig. 2F), beneficial for the distribution of aptamers and preventing non-specific bindings. A significant decrease in zeta potential of GNPs to 50 ± 2 mV ($p \leq 0.05$) also confirmed the surface modification via positively charged aptamers. DNA tetrahedrons are the basic structure for the electrochemical reaction, assembled from four single-stranded oligonucleotides (D1/D2, D3, D4, and D5 for T1/T2, respectively) by an annealing process (Fig. 2G). The self-assembly process of DNA tetrahedron was investigated by native polyacrylamide gel electrophoresis (PAGE, Fig. 2H). The lanes from left to right (corresponding to Lane 1 to $\frac{5}{6}$) in PAGE results indicated a stepwise increase in molecular weight during self-assembly process. As shown in Lane 5 and 6, the DNA tetrahedrons were prepared through consecutive hybridization among DNA oligonucleotides. In particular, the consecutive bands were witnessed due to decreased mobility, consistent with the significantly increased hydrodynamic sizes of DNA structures. Therefore, the increase in molecular weight as well as mobility together validated successful self-assembly of DNA tetrahedrons on WE. ## Construction and feasibility verification of the biosensor The stepwise construction process of the biosensor has been validated by electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV). As depicted in Fig. 3A, the bare electrode presented an electron transfer resistance (Ret) of about 72 Ω (curve a). After the successive assembly of T1/T2 and MCH on the electrode, the Ret increased significantly (curve b) due to the hindered electron transfer after the conjugation of DNA tetrahedrons. When S1-MB/S2-MB were attached to the electrode surface, the Ret value increased further (curve c), which could be ascribed to the repulsion of [Fe(CN)6]3−/4− by the DNA backbone. The Ret value continued to increase after the replacement of S1-MB/S2-MB by the released B1/B2 after hybridizing B1/B2 with T1/T2 (curve d). In the presence of H1 and H2, the Ret value increased considerably (curve e), which implied that the HCR amplification response was initiated on the electrode. Sequential fabrication steps for the ratiometric biosensor were also investigated using CV measurements of the current changes. As seen in Fig. 3B, the voltammograms of [Fe(CN)6]3−/4− gradually changed during the modification of the electrodes. All the electrochemical performances indicated the successful fabrication of the biosensor, promising the simultaneous analysis of multiple exosome surface proteins [47, 48].Fig. 3Construction and feasibility verification of the biosensor. A Electrochemical impedance spectroscopy and B cyclic voltammetry curves of the bare electrode (curve a), bare + T1/T2 (curve b), bare + T1/T2 + S1-MB/S2-MB (curve c), bare + T1/T2 + S1-MB/S2-MB + B1/B2 (curve d), and bare + T1/T2 + S1-MB/S2-MB + B1/B2 + H1-Fc/H2-Fc (curve e). C Differential pulse voltammetry (DPV) curves under different experimental conditions. D DPV peak current of IMB and IFc under different signal inputs. E Validation of OR logic gate in the presence/absence of epithelial cell adhesion molecule (EpCAM) and carcinoembryonic antigen (CEA). F Optimization of incubation time for exosome captured by aptamers. The concentrations of exosomes were set as 105, 2 × 106, and 2 × 108 particles µL−1, respectively. G Single MB signal (IMB), single Fc signal (IFc), and H IFc/IMB ratio signal for exosome detection. The concentrations of exosomes were set as 103, 105, and 107 particles µL−1, respectively The feasibility of the electrochemical biosensor for exosome detection was also evaluated. As shown in Fig. 3C, after modified with DNA tetrahedron (T1/T2), barely any signal was observed in the absence of the released B1/B2, S1-MB/S2-MB, and H1-Fc/H2-Fc, indicating that the majority of the signals were not immobilized on the electrode. Once the reference signal from S1-MB/S2-MB was present in the reaction system, we observed an obvious signal at − 0.267 V, indicating the intercalation of MB on electrode surface via hybridization. When the exosome surface protein of CEA or EpCAM was captured by Apt1 or Apt2, the corresponding block strand (B1 or B2) was released into electrolyte and thus served as signal input (Input 1 or Input 2, respectively) for downstream sensing. In the co-existence of CEA and EpCAM, B1 and B2 were released in parallel and served as Input 1/Input 2. In the presence of released block strands (B1, B2, and B1/B2) and H1-Fc/H2-Fc, the electrochemical signal showed a further enhancement of Fc signal at + 0.278 V and a reduced MB signal at − 0.267 V. As a result, the corresponding signal intensities of Fc signal (IFc) and MB signal (IMB) were illustrated under different input conditions (Fig. 3D, E). Notably, the current ratio of IFc/IMB (used as "Output") was significantly increased, owing to two signals varied in an opposite manner induced by target exosomes. Therefore, the constructed biosensor is capable of built-in correction analysis, improving sensing sensitivity with DNA-based HCR amplification [49, 50]. We further set out to identify the optimized experimental conditions, including the aptamer DNA volume ratio and the recognition time for exosome capture, as well as the DNA tetrahedron concentrations and HCR amplification time for electrochemical exosome detection. Different volume ratios were first investigated with constant exosomes, and the DPV intensity reached the highest at a concentration of 1:10 (Additional file 1: Fig. S1A). Next, the recognition time between the exosome analytes and aptamers was optimized to 1 h, yielding the highest IFc/IMB for the specific recognition and capture (Fig. 3F). Both T1/T2 concentration and HCR amplification time decide the signal intensity of the electrochemical exosome detection. The value of IFc/IMB increased with the rising T1/T2 concentration, and reached the platform after 5 μM added. As a result, the optimum concentration and amplification time were determined to be 5 μM and 1.5 h respectively (Additional file 1: Fig. S1B, C), offering the saturation of signal binding and improving the amplification efficiency. Electrochemical sensing is typically designed for single biomarkers, hardly depicting a comprehensive picture under pathological stimuli [51, 52]. To address the analyte limitations, introducing multiple electrochemical active substances or nanostructures are normally required [53–55]. However, lack of sensibility and susceptible to environmental conditions hampers the universal application of the above electrochemical techniques in clinics. Ratiometric biosensors serve as a practical alternative for detection of multiple biomarkers, due to the ability to amplify the signal changes and eliminate the fluctuations by external factors [38, 39, 56, 57]. In this work, single signal biosensors displayed a maximum DPV response within $30\%$ only (current change of $29.4\%$ for IMB and $20.5\%$ for IFc, Fig. 3G). In comparison, our constructed ratiometric biosensor demonstrated a decrease in IMB and an increase in IFc (Fig. 3H), facilitating the sensitive detection with an increased change of $58.9\%$ in signal ratio (IFc/IMB). In addition, the data acquired from ratiometric biosensor has higher reproducibility and accuracy with a lower coefficient of variation (CV) of $3.5\%$ (Fig. 3H), as compared with that from single signal biosensors (CV = $8.9\%$ for IMB, CV = $5.5\%$ for IFc; Fig. 3G). Therefore, our approach addressed the current challenges with enhanced biomarker throughput (CEA and EpCAM as a proof-of-concept demonstration in this work) and robust analyte quantitation (lower CV = $3.5\%$), by introducing a reference signal to construct a ratiometric biosensor. ## Ultrasensitive detection of exosome The proposed ratiometric biosensor enabled the exosome detection in an accurate, specific, sensitive, reproducible, and stable manner. For detection accuracy, we found that quantitation of exosome concentrations using the proposed biosensor were in agreement with those measured via a commonly-used nanoflow cytometry method, with a Pearson correlation coefficient of 0.917 (Fig. 4A). For detection specificity, the assay displayed preference toward exosomes with highest IFc/IMB response (Fig. 4B), against typical interfering substances that co-existed in human blood (e.g., neuron-specific enolase (NSE), squamous cell carcinoma antigen (SCCA), pro-gastrin-releasing peptide (Pro-GRP), carcinoembryonic antigen 125 (CA125), section 19 of CYFRA 21-1 (CYF21-1), and three-base mismatched aptamers (Mis-3)). For detection sensitivity, the subtle changes of different exosomes were determined by the proposed biosensor under optimal experimental conditions (Fig. 4C). The current intensity ratio IFc/IMB was linearly proportional (R2 = 0.986) to the exosome concentrations (Fig. 4D), yielding a low limit of detection of 15.1 particles μL−1 within a wide linear range of 102–108 particles μL−1 (IFc/IMB = 0.504 LgCexosome − 0.487). For detection reproducibility, we recorded current intensities of 2 × 103 particles μL−1 of exosomes at intra-batch level (5 parallel measurements for three independent experiments), affording coefficients of variation within $4.8\%$ (Fig. 4E). For detection stability, the electrochemical signals were retained at $88.1\%$ after storage for one week and $77.8\%$ for two weeks (Fig. 4F), indicating the long-term stability suitable for clinical use. Fig. 4Ultrasensitive detections of exosomes. A Comparison of the biosensor with NanoFCM for exosome detection. B Determination of the selectivity of exosome detection against typical interfering substances, including 1 KU/mL of CA125, 1 mg/mL of NSE, Pro-GRP, SCCA, and CYF21-1, respectively. C DPV analysis of different concentrations of exosomes ranging 108–102 particles µL−1 (from a to g). D Linear calibration curve between the ratiometric intensities and logarithm of exosome concentrations. E DPV responses acquired from 5 working electrodes, under the constant level of exosomes (2 × 103 particles µL−1). F DPV responses acquired over two weeks, under the constant level of exosomes (107 particles µL−1). G Exosome expression were profiled in different types of NSCLC cell lines (A549, H460, H1299, H1975, H2030, and Calu-1 cells). H Linear correlation between ratiometric intensities IFc/IMB and A549 cell numbers ranging 103–107. I Ratiometric intensities of exosomes spiked into three clinical serum samples at final concentrations of $\frac{103}{105}$ particles µL−1. Error bars referred to the standard derivation obtained from three independent experiments Having optimized the biosensor, we validated the feasibility of our approach in exosome sensing in both cell media and real-case biospecimens. We first profiled the exosome expression in different NSCLC cell lines (A549, H460, H1299, H1975, H2030, and Calu-1 cells). A significant increase in ratiometric intensity was detected in NSCLC cell lines as compared with that in phosphate buffered saline (Fig. 4G), suggesting a positive correlation between the expression levels of surface proteins CEA and EpCAM on the exosomes and LC pathogenesis [30, 36, 42]. Next, we selected A549 cells as an example to validate the capability of the proposed biosensor for exosome quantitation. The peak current difference exhibited a linear correlation with the logarithmic number of cells, yielding a wide linear range of 103–107 cells (Fig. 4H). Exosomes were spiked into human serum to concentrations of $\frac{103}{105}$ particles μL−1. The biosensor afforded the average recovery of $90.1\%$-$105.6\%$ with CV within $5.4\%$ (Fig. 4I), comparable to the traditional electrochemical method according to previous reports [20]. The analytical performances of the developed biosensor and previously reported methods for exosome detection are summarized in Supplementary Information (Additional file 1: Table S2). As can be seen, the proposed biosensor exhibited a lower limit-of-detection within a wider linear range, compared with other techniques (e.g., fluorescence, surface plasmon resonance, and electrochemiluminescence) [58–60]. We attributed the superior sensitivity of the proposed biosensor to the ratiometric strategy and aptamer-facilitated HCR for signal amplification [61, 62]. In addition to the sensitivity performance, the proposed biosensor also displayed an improved specificity by introducing a DNA OR logic gate for the simultaneous detection of two surface protein targets, which provides more comprehensive information for cancer diagnosis [63–65]. In particular, single biomarker hardly distinguishes patient groups with satisfied performance, since diseases are accompanied by abnormal regulation of multiple biomarkers. The multiplex and simultaneous analysis of multiple biomarkers depicted the systematic alteration under disease stimuli, playing a key role especially in the era of precision medicine. Notably, the distinct signal intensities from different systems suggested a positive correlation between exosome expression and LC pathogenesis (Fig. 4G). Therefore, we concluded that the ratiometric biosensor achieved accurate, specific, sensitive, reproducible, and stable profiling and quantitation of exosomes in real-case biospecimens toward further diagnostics. ## Assay of clinical sample-derived exosome for NSCLC diagnosis Having shown the heterogeneous nature of protein expression on exosome surface, we next demonstrated the possibility of the assay for NSCLC diagnosis. We recruited 60 NSCLC patients with stage I-IV and 15 HDs. All the patient subtypes had been verified by histological findings, including: adenocarcinoma, (ADC, $$n = 20$$), squamous-cell carcinoma (SCC, $$n = 20$$), and large-cell carcinoma (LCC, $$n = 20$$). The sample demographics are provided (Additional file 1: Tables S3–S5). Surface protein markers of serum exosomes were profiled by the assay (Fig. 5A). In particular, marked differences in ratiometric intensity between NSCLC patients and HD group were observed ($p \leq 0.001$, Fig. 5B). We obtained the optimized diagnostic sensitivity of $98.3\%$ with specificity of $86.7\%$ based on the cut-off value of 1.23 according to the receiver operating characteristic (ROC) curve (area under curve of ROC (AUC) of 0.973 ($95\%$CI: 0.798–1.000)), in discriminating NSCLC patients from HD group. Importantly, the readouts acquired from the assay accounted for a sensitivity of $93.3\%$, specificity of $86.7\%$ with AUC of 0.916 ($95\%$CI: 0.798–1.000) in the early diagnosis of NSCLC staging I (Fig. 5C).Fig. 5Assay of clinical sample-derived exosome for NSCLC diagnosis. A Ratiometric levels and B the corresponding scatter intervals of serum samples using the exosome-based assay, to differentiate NSCLC patients from healthy donors (HDs). The optimized threshold (107 particles µL.−1) according to the receiver operating characteristic (ROC) curve was denoted by the dashed line in A. C ROC curves of the exosome-based assay and six clinically adopted serum biomarkers, in differentiating stage I NSCLC patients from HDs. D Heat map and E scatter intervals of ratiometric intensities for exosome detection in HDs and NSCLC patients at stage I-IV ($$n = 15$$, respectively). F Heat map of ratiometric intensities for exosome detection in patients with benign disease, NSCLC patients without treatment, NSCLC patients with treatment and recurrence/no recurrence ($$n = 15$$, respectively). G Scatter intervals and H ROC curves of the exosome-based assay and six clinically adopted serum biomarkers, in differentiating NSCLC patients (stage I, no treatment) from patients with benign disease. I Scatter interval for NSCLC patients with treatment and recurrence or not. J ROC curves of the exosome-based assay and six clinically adopted serum biomarkers, in differentiating NSCLC patients with recurrence and no recurrence. *** $p \leq 0.001$, ****$p \leq 0.0001$ Exosome overexpression is reported to function as tumor markers, highly correlated with the occurrence and metastasis of NSCLC [66, 67]. We further investigated the applicability of the assay in tumor staging. The ratiometric intensity was strongly predictive of NSCLC progression (Fig. 5D), independent of the cancer subtypes. For instance, early cancer patients (stage I) had higher levels of exosomes as compared to HD, which kept on upregulating in advanced NSCLC (stage IV, $p \leq 0.0001$). In detail, each stage of NSCLC showed a heterogeneous rise in the exosome level, in contrast with the low expression in HD ($p \leq 0.0001$, Fig. 5E). Following cancer treatment, exosomes hold promise as predictive biomarkers for therapy efficacy evaluation and recurrence risk monitoring [68, 69]. We analyzed the exosome protein expression from a new patient cohort (Fig. 5F), including benign disease ($$n = 15$$); cancer without treatment ($$n = 15$$); and cancer with treatment (recurrence: $$n = 15$$; no recurrence: $$n = 15$$). As depicted in Fig. 5G, NSCLC patients had lower levels of exosome compared to patients with benign disease (t-test, $p \leq 0.0001$). As a result, we achieved sensitivity of $100.0\%$ and specificity of $80.0\%$ with an AUC of 0.933 ($95\%$CI: 0.847–1.000), in differentiating early NSCLC (stage I, no treatment) from benign disease (red curve in Fig. 5H). We further explored the capability of the assay in predicting the recurrence risk after the surgical/radical operation. We profiled exosome surface proteins and observed significant difference between recurrence and no recurrence groups (t-test, $p \leq 0.0001$; Fig. 5I). Similarly, we achieved both sensitivity and specificity of $100\%$, with an AUC of 1.000 ($95\%$CI: 1.000–1.000) in the differentiation of these two groups (red curve in Fig. 5J). The current liquid biopsy in clinics mainly relies on serum tumor protein markers (e.g., CEA, CA125, SCCA, NSE, Pro-GRP, and CYF21-1) [70–73]. However, the classification models built on the six serum biomarkers underperformed our proposed exosome assay. Consequently, the combination of protein biomarkers demonstrated an AUC of 0.844, 0.733, and 0.947 for the staging, diagnosis, and prognosis of NSCLC in the current datasets (black curves in Fig. 5C, H, J), as well as the reported values (AUC = 0.82, 0.84, and 0.87, respectively) [42, 74]. Importantly, the serum protein markers are far from satisfactory in the early diagnosis of NSCLC (stage I), displaying AUCs ranging 0.460–0.796 consistent with references [23, 75]. Apart from the clinically-adopted serum biomarkers, it was also difficult to differentiate between benign and malignant lesions at an early stage (e.g., SI NSCLC) by other emerging liquid biopsy approaches [72, 76], such as integrated analysis of circulating proteins and mutations in cell-free DNA (CancerSeek) showing an accuracy of $43\%$ and sequencing analysis of circulating tumour DNA (Lung-CLiP) showing a sensitivity of $63\%$ [77, 78]. Taken together, the surface proteins profiled from exosomes by the assay were capable of identifying malignance, superior to the protein biomarkers recommended by medical association guidelines for liquid biopsy of NSCLC. ## Conclusion As a limitation of our work, larger sample size is needed to further validate the diagnostic performance of the proposed assay. In addition, the integration of the sensing system with microfluidics remains to be a promising path for one-step extraction, separation, and analysis toward the dedicated use of exosome-based assay in clinical settings. In summary, a sensitive and specific diagnosis assay composed of ratiometric biosensor and OR logic gate was developed for the detection of NSCLC-derived exosomes. The assay enabled ultrasensitive sensing of trace exosomes (as low as 15.1 particles μL−1), by combining the HCR amplification with two redox reporter signals. In clinical demonstrations, the assay was superior in detecting stage I NSCLC from HDs with sensitivity of $93.3\%$ than the combination of serum protein biomarkers (AUC of 0.844). In addition, the assay is capable of monitoring the tumor progression, showing an upregulated level of exosome level in advanced NSCLC, in contrast with the low expression in early stage. For patients who underwent biopsies, the readouts acquired from the assay are highly predictive of the cancer recurrence (AUC of 1.000), holding promise in evaluating the treatment efficacy in clinics. Taken together, the expression of dysregulated surface proteins was profiled and correlated to clinical status during cancer progression. As a proof-of-concept validation, the exosome-based assay provided enhanced differentiation outcomes compared to the reported blood biomarker paradigm for the diagnosis, staging, and prognosis of NSCLC patients. We anticipated the assay could be easily translated into the diagnostic workflow not exclusively specific to malignant tumors. ## Supplementary Information Additional file 1: Materials and reagents, oligonucleotide sequences used in the experiments; GNP synthesis and characterization; Optimization of experimental conditions; Comparison of different methods for exosome detection; Clinical demographics of enrolled cohorts. Table S1. Oligonucleotide sequences used in the experiments. Table S2. Comparison of different methods for exosome detection. Table S3. Clinical demographics for NSCLC diagnosis. Table S4. Clinical demographics for differentiation of NSCLC from benign diseases. Table S5. Clinical demographics for NSCLC prognosis. Figure S1. The intensity of the electrochemical response on (A) the volume ratio of the GNPs and DNA (1 μM), (B) the incubation time between exosomes and aptamers, and (C) the DNA tetrahedron concentration. The concentrations of exosomes were set as 2×106, 6×106, 4×106 particles μL−1, respectively. Error bars represented the standard deviation of three repetitive measurements. ## References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A. **Global cancer statistics 2020: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *Ca-Cancer J Clin* (2021.0) **71** 209-249. DOI: 10.3322/caac.21660 2. Liang W, Liu J, He J. **Driving the improvement of lung cancer prognosis**. *Cancer Cell* (2020.0) **38** 449-451. DOI: 10.1016/j.ccell.2020.09.008 3. Crosby D, Bhatia S, Brindle KM, Coussens LM, Dive C, Emberton M. **Early detection of cancer**. *Science* (2022.0) **375** eaay9040. DOI: 10.1126/science.aay9040 4. Rawal S, Patel M. **Bio-nanocarriers for lung cancer management: Befriending the barriers**. *Nano-Micro Lett* (2021.0) **13** 1-54. DOI: 10.1007/s40820-021-00630-6 5. Liu S, Jie G, Wu Y. **Toward a cure for lung cancer: Important advances in operable non-small cell lung cancer**. *Sci Bull* (2022.0) **67** 1402-1405. DOI: 10.1016/j.scib.2022.06.008 6. Huang Y, Du S, Liu J, Huang W, Liu W, Zhang M. **Diagnosis and prognosis of breast cancer by high-performance serum metabolic fingerprints**. *P Natl Acad Sci USA* (2022.0) **119** e2122245119. DOI: 10.1073/pnas.2122245119 7. de Koning HJ, van der Aalst CM, de Jong PA, Scholten ET, Nackaerts K, Heuvelmans MA. **Reduced lung-cancer mortality with volume CT screening in a randomized trial**. *New Engl J Med* (2020.0) **382** 503-513. DOI: 10.1056/NEJMoa1911793 8. Huang L, Wang L, Hu X, Chen S, Tao Y, Su H. **Machine learning of serum metabolic patterns encodes early-stage lung adenocarcinoma**. *Nat Commun* (2020.0) **11** 3556. DOI: 10.1038/s41467-020-17347-6 9. Su H, Li X, Huang L, Cao J, Zhang M, Vedarethinam V. **Plasmonic alloys reveal a distinct metabolic phenotype of early gastric cancer**. *Adv Mater* (2021.0) **33** 2007978. DOI: 10.1002/adma.202007978 10. Kim M, Chen C, Wang P, Mulvey JJ, Yang Y, Wun C. **Detection of ovarian cancer via the spectral fingerprinting of quantum-defect-modified carbon nanotubes in serum by machine learning**. *Nat Biomed Eng* (2022.0) **6** 267-275. DOI: 10.1038/s41551-022-00860-y 11. Meng F, Yu W, Chen C, Guo S, Tian X, Miao Y. **A versatile electrochemical biosensor for the detection of circulating microrna toward non-small cell lung cancer diagnosis**. *Small* (2022.0) **18** e2200784. DOI: 10.1002/smll.202200784 12. Nie M, Yao K, Zhu X, Chen N, Xiao N, Wang Y. **Evolutionary metabolic landscape from preneoplasia to invasive lung adenocarcinoma**. *Nat Commun* (2021.0) **12** 6479. DOI: 10.1038/s41467-021-26685-y 13. Wan J, Massie C, Garcia-Corbacho J, Mouliere F, Brenton JD, Caldas C. **Liquid biopsies come of age: towards implementation of circulating tumour DNA**. *Nat Rev Cancer* (2017.0) **17** 223-238. DOI: 10.1038/nrc.2017.7 14. Huang L, Gurav DD, Wu S, Xu W, Vedarethinam V, Yang J. **A multifunctional platinum nanoreactor for point-of-care metabolic analysis**. *Matter* (2019.0) **1** 1669-1680. DOI: 10.1016/j.matt.2019.08.014 15. Cohen JD, Li L, Wang Y, Thoburn C, Afsari B, Danilova L. **Detection and localization of surgically resectable cancers with a multi-analyte blood test**. *Science* (2018.0) **359** 926-930. DOI: 10.1126/science.aar3247 16. Huang L, Wan J, Wei X, Liu Y, Huang J, Sun X. **Plasmonic silver nanoshells for drug and metabolite detection**. *Nat Commun* (2017.0) **8** 220. DOI: 10.1038/s41467-017-00220-4 17. Theodoraki MN, Laban S, Jackson EK, Lotfi R, Schuler PJ, Brunner C. **Changes in circulating exosome molecular profiles following surgery/(chemo)radiotherapy: early detection of response in head and neck cancer patients**. *Br J Cancer* (2021.0) **125** 1677-1686. DOI: 10.1038/s41416-021-01567-8 18. Zhang J, Ji C, Zhang H, Shi H, Mao F, Qian H. **Engineered neutrophil-derived exosome-like vesicles for targeted cancer therapy**. *Sci Adv* (2022.0) **8** eabj8207. DOI: 10.1126/sciadv.abj8207 19. Li M-Y, Liu L-Z, Dong M. **Progress on pivotal role and application of exosome in lung cancer carcinogenesis, diagnosis, therapy and prognosis**. *Mol Cancer* (2021.0) **20** 1-22. DOI: 10.1186/s12943-021-01312-y 20. Zhu C, Li L, Wang Z, Irfan M, Qu F. **Recent advances of aptasensors for exosomes detection**. *Biosens Bioelectron* (2020.0) **160** 112213. DOI: 10.1016/j.bios.2020.112213 21. Lyu Y, Cui D, Huang J, Fan W, Miao Y, Pu K. **Near-infrared afterglow semiconducting nano-polycomplexes for the multiplex differentiation of cancer exosomes**. *Angew Chem Int Edit* (2019.0) **58** 4983-4987. DOI: 10.1002/anie.201900092 22. Hu L, Zhang T, Ma H, Pan Y, Wang S, Liu X. **Discovering the secret of diseases by incorporated tear exosomes analysis via rapid-isolation system: itears**. *ACS Nano* (2022.0) **16** 11720-11732. DOI: 10.1021/acsnano.2c02531 23. Kalluri R, LeBleu VS. **The biology, function, and biomedical applications of exosomes**. *Science* (2020.0) **367** eaau6977. DOI: 10.1126/science.aau6977 24. Chen Y, Zhu Q, Cheng L, Wang Y, Li M, Yang Q. **Exosome detection via the ultrafast-isolation system: exodus**. *Nat Methods* (2021.0) **18** 212-218. DOI: 10.1038/s41592-020-01034-x 25. Sahu SS, Cavallaro S, Haag P, Nagy A, Karlstrom AE, Lewensohn R. **Exploiting electrostatic interaction for highly sensitive detection of tumor-derived extracellular vesicles by an electrokinetic sensor**. *ACS Appl Mater Inter* (2021.0) **13** 42513-42521. DOI: 10.1021/acsami.1c13192 26. Zhuang J, Tan J, Wu C, Zhang J, Liu T, Fan C. **Extracellular vesicles engineered with valency-controlled DNA nanostructures deliver crispr/cas9 system for gene therapy**. *Nucleic Acids Res* (2020.0) **48** 8870-8882. DOI: 10.1093/nar/gkaa683 27. Lv WY, Li CH, Li YF, Zhen SJ, Huang CZ. **Hierarchical hybridization chain reaction for amplified signal output and cascade DNA logic circuits**. *Anal Chem* (2021.0) **93** 3411-3417. DOI: 10.1021/acs.analchem.0c04483 28. Li X, Yang Y, Zhang B, Lin X, Fu X, An Y. **Lactate metabolism in human health and disease**. *Signal Transduct Target Ther* (2022.0) **7** 305. DOI: 10.1038/s41392-022-01151-3 29. Peng H, Newbigging AM, Reid MS, Uppal JS, Xu J, Zhang H. **Signal amplification in living cells: a review of microrna detection and imaging**. *Anal Chem* (2020.0) **92** 292-308. DOI: 10.1021/acs.analchem.9b04752 30. Park J, Park JS, Huang CH, Jo A, Cook K, Wang R. **An integrated magneto-electrochemical device for the rapid profiling of tumour extracellular vesicles from blood plasma**. *Nat Biomed Eng* (2021.0) **5** 678-689. DOI: 10.1038/s41551-021-00752-7 31. Lv M, Zhou W, Fan D, Guo Y, Zhu X, Ren J. **Illuminating diverse concomitant DNA logic gates and concatenated circuits with hairpin DNA-templated silver nanoclusters as universal dual-output generators**. *Adv Mater* (2020.0) **32** e1908480. DOI: 10.1002/adma.201908480 32. Yin F, Wang F, Fan C, Zuo X, Li Q. **Biosensors based on DNA logic gates**. *View* (2020.0) **2** 2 33. Fan D, Wang J, Wang E, Dong S. **Propelling DNA computing with materials' power: recent advancements in innovative DNA logic computing systems and smart bio-applications**. *Adv Sci* (2020.0) **7** 2001766. DOI: 10.1002/advs.202001766 34. Zhao W, Hu J, Liu J, Li X, Sun S, Luan X. **Si nanowire bio-fet for electrical and label-free detection of cancer cell-derived exosomes**. *Microsyst Nanoeng* (2022.0) **8** 57. DOI: 10.1038/s41378-022-00387-x 35. Yang Y, Wang J, Huang W, Wan G, Xia M, Chen D. **Integrated urinalysis devices based on interface-engineered field-effect transistor biosensors incorporated with electronic circuits**. *Adv Mater* (2022.0) **34** 2203224. DOI: 10.1002/adma.202203224 36. Liu C, Zhao J, Tian F, Cai L, Zhang W, Feng Q. **Low-cost thermophoretic profiling of extracellular-vesicle surface proteins for the early detection and classification of cancers**. *Nat Biomed Eng* (2019.0) **3** 183-193. DOI: 10.1038/s41551-018-0343-6 37. Ling X, Gong D, Shi W, Xu Z, Han W, Lan G. **Nanoscale metal-organic layers detect mitochondrial dysregulation and chemoresistance via ratiometric sensing of glutathione and ph**. *J Am Chem Soc* (2021.0) **143** 1284-1289. DOI: 10.1021/jacs.0c11764 38. Wu N, Zhang XY, Xia J, Li X, Yang T, Wang JH. **Ratiometric 3d DNA machine combined with machine learning algorithm for ultrasensitive and high-precision screening of early urinary diseases**. *ACS Nano* (2021.0) **15** 19522-19534. DOI: 10.1021/acsnano.1c06429 39. Lu Y, Zhang B, Tian Y, Guo Q, Nie G. **Ultrasensitive ratiometric photoelectrochemical immunoassay for prostate specific antigen based on nanoscale heterojunction**. *Sensor Actuat B-Chem* (2021.0) **326** 128994. DOI: 10.1016/j.snb.2020.128994 40. Huang L, Zhou Y, Zhu Y, Su H, Yang S, Feng L. **Dual-modal nanoplatform integrated with smartphone for hierarchical diabetic detection**. *Biosens Bioelectron* (2022.0) **210** 114254. DOI: 10.1016/j.bios.2022.114254 41. Yoon JY, Sigel K, Martin J, Jordan R, Beasley MB, Smith C. **Evaluation of the prognostic significance of TNM staging guidelines in lung carcinoid tumors**. *J Thorac Oncol* (2019.0) **14** 184-192. DOI: 10.1016/j.jtho.2018.10.166 42. Yu D, Li Y, Wang M, Gu J, Xu W, Cai H. **Exosomes as a new frontier of cancer liquid biopsy**. *Mol Cancer* (2022.0) **21** 1-33. DOI: 10.1186/s12943-022-01509-9 43. Sun JD, Sun XL. **Recent advances in the construction of DNA nanostructure with signal amplification and ratiometric response for miRNA sensing and imaging**. *Trac-Trend Anal Chem* (2020.0) **127** 115900. DOI: 10.1016/j.trac.2020.115900 44. Deng J, Zhao S, Li J, Cheng Y, Liu C, Liu Z. **One-step thermophoretic and gate operation on extracellular vesicles improves diagnosis of prostate cancer**. *Angew Chem Int Ed Edit* (2022.0) **61** e202207037 45. Wang Q, Zou L, Yang X, Liu X, Nie W, Zheng Y. **Direct quantification of cancerous exosomes via surface plasmon resonance with dual gold nanoparticle-assisted signal amplification**. *Biosens Bioelectron* (2019.0) **135** 129-136. DOI: 10.1016/j.bios.2019.04.013 46. Chen H, Huang C, Wu Y, Sun N, Deng C. **Exosome metabolic patterns on aptamer-coupled polymorphic carbon for precise detection of early gastric cancer**. *ACS Nano* (2022.0) **16** 12952-12963. DOI: 10.1021/acsnano.2c05355 47. Wu W, Liu X, Li W. **Progress and challenges in functional nanomaterial-based suspension array technology for multiplexed biodetection**. *View* (2022.0) **3** 20200140. DOI: 10.1002/VIW.20200140 48. 48.Song S, Zhu L, Wang C, Yang Y. In vitro diagnostic technologies for the detection of extracellular vesicles: Current status and future directions. View. 2022: 10.1002/VIW.20220011.20220011.. DOI: 10.1002/VIW.20220011.20220011. 49. Chai H, Cheng W, Jin D, Miao P. **Recent progress in DNA hybridization chain reaction strategies for amplified biosensing**. *ACS Appl Mater Inter* (2021.0) **13** 38931-38946. DOI: 10.1021/acsami.1c09000 50. Kim J, Shim JS, Han BH, Kim HJ, Park J, Cho I-J. **Hydrogel-based hybridization chain reaction (HCR) for detection of urinary exosomal mirnas as a diagnostic tool of prostate cancer**. *Biosens Bioelectron* (2021.0) **192** 113504. DOI: 10.1016/j.bios.2021.113504 51. Sadat Mousavi P, Smith SJ, Chen JB, Karlikow M, Tinafar A, Robinson C. **A multiplexed, electrochemical interface for gene-circuit-based sensors**. *Nat Chem* (2020.0) **12** 48-55. DOI: 10.1038/s41557-019-0366-y 52. Zhou J, Wu Z, Hu J, Yang D, Chen X, Wang Q. **High-throughput single-ev liquid biopsy: Rapid, simultaneous, and multiplexed detection of nucleic acids, proteins, and their combinations**. *Sci Adv* (2020.0) **6** eabc1204. DOI: 10.1126/sciadv.abc1204 53. Somasundaram S, Easley CJ. **A nucleic acid nanostructure built through on-electrode ligation for electrochemical detection of a broad range of analytes**. *J Am Chem Soc* (2019.0) **141** 11721-11726. DOI: 10.1021/jacs.9b06229 54. Lu C, Chen X. **Nanostructure engineering of graphitic carbon nitride for electrochemical applications**. *ACS Nano* (2021.0) **15** 18777-18793. DOI: 10.1021/acsnano.1c06454 55. Zhou Y, Mazur F, Fan Q, Chandrawati R. **Synthetic nanoprobes for biological hydrogen sulfide detection and imaging**. *View* (2022.0) **3** 20210008. DOI: 10.1002/VIW.20210008 56. Chen H, Yu J, Men X, Zhang J, Ding Z, Jiang Y. **Reversible ratiometric nadh sensing using semiconducting polymer dots**. *Angew Chem Int Edit* (2021.0) **133** 12114-12119. DOI: 10.1002/ange.202100774 57. Yang R, Gao Y, Ouyang Z, Shi X, Shen M. **Gold nanostar-based complexes applied for cancer theranostics**. *View* (2022.0) **3** 20200171. DOI: 10.1002/VIW.20200171 58. Moura SL, Pallares-Rusinol A, Sappia L, Marti M, Pividori MI. **The activity of alkaline phosphatase in breast cancer exosomes simplifies the biosensing design**. *Biosens Bioelectron* (2022.0) **198** 113826. DOI: 10.1016/j.bios.2021.113826 59. Yang T, Yu R, Yan Y, Zeng H, Luo S, Liu N. **A review of ratiometric electrochemical sensors: from design schemes to future prospects**. *Sensor Actuat B-Chem* (2018.0) **274** 501-516. DOI: 10.1016/j.snb.2018.07.138 60. Cao J, Xiao Y, Zhang M, Huang L, Wang Y, Liu W. **Deep learning of dual plasma fingerprints for high-performance infection classification**. *Small* (2022.0). DOI: 10.1002/smll.202206349 61. Kogikoski S, Paschoalino WJ, Cantelli L, Silva W, Kubota LT. **Electrochemical sensing based on DNA nanotechnology**. *Trac-Trend Anal Chem* (2019.0) **118** 597-605. DOI: 10.1016/j.trac.2019.06.021 62. Phillips JW, Prominski A, Tian B. **Recent advances in materials and applications for bioelectronic and biorobotic systems**. *View* (2022.0) **3** 20200157. DOI: 10.1002/VIW.20200157 63. Song X, Reif J. **Nucleic acid databases and molecular-scale computing**. *ACS Nano* (2019.0) **13** 6256-6268. DOI: 10.1021/acsnano.9b02562 64. Ye F, Zhao Y, Ei-Sayed R, Muhammed M, Hassan M. **Advances in nanotechnology for cancer biomarkers**. *Nano Today* (2018.0) **18** 103-123. DOI: 10.1016/j.nantod.2017.12.008 65. Xu W, Lin J, Gao M, Chen Y, Cao J, Pu J. **Rapid computer-aided diagnosis of stroke by serum metabolic fingerprint based multi-modal recognition**. *Adv Sci* (2020.0) **7** 2002021. DOI: 10.1002/advs.202002021 66. Jiang C, Zhang N, Hu X, Wang H. **Tumor-associated exosomes promote lung cancer metastasis through multiple mechanisms**. *Mol Cancer* (2021.0) **20** 1-17. DOI: 10.1186/s12943-021-01411-w 67. Shin H, Oh S, Hong S, Kang M, Kang D, Ji Y-g. **Early-stage lung cancer diagnosis by deep learning-based spectroscopic analysis of circulating exosomes**. *ACS Nano* (2020.0) **14** 5435-44. DOI: 10.1021/acsnano.9b09119 68. Zhang J, Fu B, Li M, Mi S. **Secretome of activated fibroblasts induced by exosomes for the discovery of biomarkers in non-small cell lung cancer**. *Small* (2021.0) **17** 2004750. DOI: 10.1002/smll.202004750 69. Salunkhe S, Basak M, Chitkara D, Mittal A. **Surface functionalization of exosomes for target-specific delivery and in vivo imaging & tracking: strategies and significance**. *J Control Release* (2020.0) **326** 599-614. DOI: 10.1016/j.jconrel.2020.07.042 70. Yang G, Xiao Z, Tang C, Deng Y, Huang H, He Z. **Recent advances in biosensor for detection of lung cancer biomarkers**. *Biosens Bioelectron* (2019.0) **141** 111416. DOI: 10.1016/j.bios.2019.111416 71. Wang R, Gu Z, Wang Y, Yin X, Liu W, Chen W. **A “one-stop shop” decision tree for diagnosing and phenotyping polycystic ovarian syndrome on serum metabolic fingerprints**. *Adv Funct Mater* (2022.0) **32** 2206670. DOI: 10.1002/adfm.202206670 72. Yin X, Yang J, Zhang M, Wang X, Xu W, Price C-AH. **Serum metabolic fingerprints on bowl-shaped submicroreactor chip for chemotherapy monitoring**. *ACS Nano* (2022.0) **16** 2852-65. DOI: 10.1021/acsnano.1c09864 73. Ongarora BG. **Recent technological advances in the management of chronic wounds: a literature review**. *Health Science Reports* (2022.0) **5** e641. DOI: 10.1002/hsr2.641 74. Seijo LM, Peled N, Ajona D, Boeri M, Field JK, Sozzi G. **Biomarkers in lung cancer screening: achievements, promises, and challenges**. *J Thorac Oncol* (2019.0) **14** 343-357. DOI: 10.1016/j.jtho.2018.11.023 75. Vargas AJ, Harris CC. **Biomarker development in the precision medicine era: Lung cancer as a case study**. *Nat Rev Cancer* (2016.0) **16** 525-537. DOI: 10.1038/nrc.2016.56 76. Moinzad H, Akbarzadeh MH. **How to improve information technology strategic planning effectiveness using balanced scorecard, risk and maturity analysis, case study health information technology? A qualitative study.&nbsp;**. *Health Science Reports* (2022.0) **5** e926. DOI: 10.1002/hsr2.926 77. Chabon JJ, Hamilton EG, Kurtz DM, Esfahani MS, Moding EJ, Stehr H. **Integrating genomic features for non-invasive early lung cancer detection**. *Nature* (2020.0) **580** 245-251. DOI: 10.1038/s41586-020-2140-0 78. Wang Y, Liu Y, Yang S, Yi J, Xu X, Zhang K. **Host-guest self-assembled interfacial nanoarrays for precise metabolic profiling**. *Small* (2023.0). DOI: 10.1002/smll.202207190
--- title: 'Association between obesity and short- and medium-term mortality in critically ill patients with atrial fibrillation: a retrospective cohort study' authors: - Duo Yang - Shujun Ye - Kaihong Zhang - Zhiliang Huang - Longsheng Zhang journal: BMC Cardiovascular Disorders year: 2023 pmcid: PMC10037857 doi: 10.1186/s12872-023-03179-x license: CC BY 4.0 --- # Association between obesity and short- and medium-term mortality in critically ill patients with atrial fibrillation: a retrospective cohort study ## Abstract ### Background There has been controversy about how obesity affects the clinical prognosis for patients with atrial fibrillation (AF), and the relationship between obesity and outcomes in critically ill patients with AF remains unclear. The purpose of this study was to explore the association between obesity and short- and medium-term mortality in critically ill patients with AF. ### Methods The Medical Information Mart for Intensive Care-IV (MIMIC-IV) database was used to conduct a retrospective cohort analysis on 9282 critically ill patients with AF. Patients were categorized into four groups based on their body mass index (BMI) values: underweight, normal-weight, overweight, and obese. The outcomes of this study were 30-day, 90-day, and 1-year all-cause mortality. Cox proportional-hazards models and restricted cubic spline analyses were performed to investigate the association between BMI and mortality. ### Results For 30-day mortality, after adjustment for all confounding factors, the hazard ratio (HR) with $95\%$ confidence interval (CI) for the underweight, overweight, and obese categories were 1.58 (1.21, 2.07), 0.82 (0.72, 0.93), and 0.79 (0.68, 0.91), respectively, compared to the normal-weight category. Using multivariable-adjusted restricted cubic spline analysis, an “L-shaped” correlation was observed between BMI and 30-day mortality. For each 1 kg/m2 increase in BMI when BMI was less than 30 kg/m2, the risk of 30-day mortality decreased by $6.4\%$ (HR, $95\%$ CI: 0.936 [0.918, 0.954]; $P \leq 0.001$); however, this relationship was not present when BMI was greater than or equal to 30 kg/m2. Similar results were observed for 90-day and 1-year mortality. ### Conclusions There was a nonlinear relationship between BMI and all-cause mortality among critically ill patients with AF. All-cause mortality and the BMI were negatively correlated when the BMI was less than 30 kg/m2. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12872-023-03179-x. ## Background As the most prevalent kind of persistent arrhythmia, atrial fibrillation (AF) is associated with severe morbidity and mortality, placing a heavy burden on patients and healthcare systems worldwide [1]. It is estimated that the number of people suffering from AF will reach 72 million by 2050 [2]. AF is characterized by a severe disturbance of atrial electrical activity, resulting in ineffective atrial contractions [3]. It is a significant cause of heart failure, stroke, and all-cause mortality [4]. AF is also the most prevalent arrhythmia in patients admitted to the intensive care unit (ICU) [5]. ICU patients with AF experience worse clinical outcomes than patients without AF [6]. Reducing the adverse prognoses of critically ill patients with AF has always been an important clinical goal [7]. Identifying prognostic factors among critically ill patients with AF is crucial for clinicians to make wiser treatment strategies and improve clinical outcomes. Over the past 40 years, obesity has shown a clear upward trend in both developed and developing countries and has become a global epidemic [8]. Obesity was found to be an independent risk factor for the occurrence, recurrence, and progression of AF in previous studies [9–11]. Interestingly, some studies showed that overweight and obesity were positively associated with better outcomes among patients with AF [12, 13]. The “obesity paradox” has been used to describe this counter-intuitive phenomenon. In fact, the obesity paradox has been reported to exist in various cardiovascular diseases, including heart failure [14], hypertension, and coronary artery disease [15], as well as various non-cardiovascular diseases, including chronic liver disease [16] and chronic obstructive pulmonary disease [17]. Nonetheless, there is conflicting evidence regarding how obesity affects clinical outcomes in AF patients [18]. In a prospective cohort study, overweight and obesity were indicated to be associated with adverse outcomes in AF patients [19]. In addition, most of the previous studies have focused on the association of obesity with long-term outcomes in AF patients. To date, in critically ill patients with AF, the relationship between obesity and outcomes remains unclear. Therefore, we aimed to use real-world data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database to evaluate the relationship between obesity and all-cause mortality for critically ill patients with AF. ## Data source The data for this retrospective cohort study was collected from the MIMIC-IV database (version 2.1), a publicly accessible database containing the clinical information on over 73,000 ICU stays at Beth Israel Deaconess Medical Center (BIDMC) from 2008 to 2019. The use of the database has been approved by the Massachusetts Institute of Technology and the BIDMC institutional review boards. Duo Yang, one of the authors, has obtained access to the database (certification number: 48247201). Information in this database was anonymous, and informed consent from the participants was waived. This study adhered to the Declaration of Helsinki and the guidelines for Strengthening the Reporting of Observational Studies in Epidemiology. ## Study Population The inclusion criteria of our study were adult critically ill patients (age ≥ 18 years) diagnosed with atrial fibrillation according to the ninth or tenth revision of the International Classification of Diseases (ICD) code. ICD-9 code 427.31, ICD-10 codes I48.0, I48.1x, I48.2x, and I48.91 were used to identify the patients [20, 21]. Patients were excluded if key data (weight on the first day admitted to the ICU, height, or follow-up information) were missing. Additionally, patients with a length of ICU stay less than 24 h were also excluded. Finally, for patients who had multiple ICU stays, only the data of the first ICU admission were analyzed in this study. ## Data extraction Navicat Premium 12 software and Structure Query Language were utilized to extract the following information from the MIMIC-IV database: [1] weight on the first day of ICU stay and height; [2] demographics: age, sex, and race; [3] comorbidities: hypertension, diabetes, congestive heart failure (CHF), peripheral vascular disease (PVD), cerebrovascular disease, chronic pulmonary disease (CPD), renal disease, liver disease, malignancy, sepsis, acute kidney injury (AKI), acute heart failure (AHF), and stroke; [4] disease severity scores: sequential organ failure assessment (SOFA), simplified acute physiology score II (SAPS II), and Charlson comorbidity index (CCI); [5] the first laboratory parameter and vital sign value after ICU admission: hemoglobin, white blood cell (WBC), platelet, red cell distribution width (RDW), anion gap, blood urea nitrogen (BUN), creatinine, glucose, heart rate (HR), mean blood pressure (MBP), respiratory rate (RR), and saturation of pulse oximetry (SpO2); [6] treatment information: medication administration during hospitalization (antiplatelet agents, anticoagulant agents, antiarrhythmic agents, and vasopressors), mechanical ventilation (MV), and renal replacement treatment (RRT). ## Exposure variable and study endpoints Body mass index (BMI), the most widely used anthropometric measure of adiposity, was the exposure variable in this study. The formula for calculating BMI is weight (in kilograms)/height (in meters)2. According to the World Health Organization BMI classifications, participants were categorized into four groups: underweight (BMI < 18.5 kg/m2), normal-weight (BMI: 18.5 to < 25 kg/m2), overweight (BMI: 25 to < 30 kg/m2), and obese (BMI ≥ 30 kg/m2). The endpoints of our study were 30-day, 90-day, and 1-year all-cause mortality after ICU admission. 30-day and 90-day mortality were referred to as short-term, whereas 1-year mortality as medium-term mortality. ## Statistical analysis Participant characteristics were analyzed based on the BMI categories. Categorical variables were described as numbers (percentages) and chi-square test was used for comparison between groups. Continuous variables were described as mean ± standard deviation (SD) for normal distributions or median and interquartile range (IQR) for skewed distributions. One-way analysis of variance or Kruskal–Wallis H-test was applied to evaluate the statistical difference of continuous variables. Survival analyses were visualized using Kaplan–Meier curves and compared with a log-rank test. Cox proportional-hazards models were performed to evaluate the hazard ratio (HR) and $95\%$ confidence interval (CI) for the relationship between BMI and mortality. For multivariable Cox analyses, covariates were chosen according to previous findings and clinical constraints. We also adjusted for covariates that had P values less than 0.1 in single-factor *Cox analysis* or that, when added to the model, changed the matched odds ratio by at least $10\%$ [22]. The variance inflation factor (VIF) was utilized to identify the multicollinearity of the covariates in the fully adjusted models, and the VIFs of all covariates cannot be greater than 5. In the minimally adjusted model (Model 1), adjusted variables included age, sex, and race. In the fully adjusted model (Model 2), we further adjusted for hypertension, diabetes, CHF, PVD, cerebrovascular disease, CPD, renal disease, liver disease, malignancy, sepsis, AKI, AHF, stroke, SOFA, SAPS II, CCI, hemoglobin, WBC, platelet, RDW, anion gap, BUN, creatinine, glucose, HR, MBP, RR, SpO2, antiplatelet agents, anticoagulant agents, antiarrhythmic agents, MV, RRT, and vasopressors. Multivariable-adjusted restricted cubic spline analyses were conducted to evaluate the possible nonlinear association between BMI and all-cause mortality. If a nonlinear association was detected, an optimal inflection point for BMI was determined, and a two-segment linear regression model was used to assess the threshold effect of the BMI on all-cause mortality. The log-likelihood ratio test was performed to compare the one-line linear model with a two-segment linear regression model. Subgroup analyses were conducted by age, sex, and comorbidities that may influence the relationship between BMI and mortality. The likelihood ratio tests were used to examine interactions between subgroups. We simply replaced the missing data using the mean and median because no more than $2\%$ of each covariate in the data was missing. Sensitivity analyses were performed to evaluate the robustness of our results. We analyzed whether the association between BMI and mortality would change after excluding patients with missing data. Furthermore, to investigate study outcomes distinctively in different levels of obesity, patients with BMI ≥ 30 kg/m2 were divided into two groups: obese (BMI: 30 to < 40 kg/m2) and morbidly obese (BMI ≥ 40 kg/m2). Data analyses were conducted using the statistical software package R. version 4.0.5 (R Foundation, Vienna, Austria) and Free Statistics software version 1.7.1. Differences with a two-sided $P \leq 0.05$ were considered to be statistically significant. ## Selection of participants Of the 73,141 ICU admissions, 13,330 patients with atrial fibrillation were identified. A flowchart of this study is presented in Fig. 1. The cohort for the final analysis included 9,282 participants. Fig. 1Flowchart of the study cohort. Abbreviations: ICU, intensive care unit; MIMIC-IV, Medical Information Mart for Intensive Care-IV ## Baseline characteristics of participants The baseline characteristics of the participants stratified by BMI values are shown in Table 1. According to the BMI values, 199, 2357, 3114, and 3612 patients belonged to the underweight, normal-weight, overweight, and obese categories, respectively. The mean age of all patients was 74.2 ± 11.5 years, with males accounting for $60.5\%$ of the population. The median SOFA score of all participants was 5.0 (IQR [3.0, 8.0]); SAPS II scores averaged 40.5 ± 13.0; and CCI scores averaged 6.6 ± 2.5. Patients with a higher BMI were younger, and men accounted for a larger percentage of those patients. Compared with normal-weight patients, overweight and obese patients had a higher prevalence of hypertension and diabetes, whereas underweight patients presented more history of cerebrovascular disease, CPD, and malignancy. Overall, patients with a higher BMI had lower 30-day, 90-day, and 1-year all-cause mortality. Interestingly, patients with a higher BMI were more likely to receive relevant treatments, including antiplatelet agents, vasopressors, MV, and RRT. Table 1Baseline characteristics of the study population according to BMI categoryVariablesTotalBMI category (kg/m2)P-ValueUnderweightNormal-weightOverweightObese< 18.5≥ 18.5, < 25≥ 25, < 30≥ 30N9282199235731143612Age (years)74.2 ± 11.577.3 ± 13.278.0 ± 11.575.1 ± 11.070.8 ± 11.0< 0.001Sex (male)5612 (60.5)76 (38.2)1261 (53.5)2068 (66.4)2207 (61.1)< 0.001Race (White)6905 (74.4)146 (73.4)1731 (73.4)2323 (74.6)2705 (74.9)0.622 Comorbidities Hypertension4266 (46.0)72 (36.2)1024 (43.4)1438 (46.2)1732 [48]< 0.001Diabetes2938 (31.7)26 (13.1)482 (20.4)896 (28.8)1534 (42.5)< 0.001CHF4193 (45.2)88 (44.2)1072 (45.5)1347 (43.3)1686 (46.7)0.044PVD1494 (16.1)28 (14.1)402 (17.1)514 (16.5)550 (15.2)0.202Cerebrovascular disease1495 (16.1)41 (20.6)422 (17.9)523 (16.8)509 (14.1)< 0.001CPD2756 (29.7)89 (44.7)695 (29.5)844 (27.1)1128 (31.2)< 0.001Renal disease2497 (26.9)41 (20.6)594 (25.2)813 (26.1)1049 [29]< 0.001Liver disease734 (7.9)15 (7.5)175 (7.4)224 (7.2)320 (8.9)0.058Malignancy1146 (12.3)39 (19.6)338 (14.3)414 (13.3)355 (9.8)< 0.001Sepsis5463 (58.9)103 (51.8)1372 (58.2)1812 (58.2)2176 (60.2)0.046AKI7471 (80.5)134 (67.3)1742 (73.9)2472 (79.4)3123 (86.5)< 0.001AHF2129 (22.9)43 (21.6)530 (22.5)683 (21.9)873 (24.2)0.148Stroke828 (8.9)27 (13.6)241 (10.2)291 (9.3)269 (7.4)< 0.001 Disease severity scores SOFA5.0 (3.0, 8.0)5.0 (3.0, 7.0)5.0 (3.0, 7.0)5.0 (3.0, 8.0)5.0 (3.0, 8.0)< 0.001SAPS II40.5 ± 13.041.7 ± 13.241.5 ± 12.540.5 ± 13.139.8 ± 13.2< 0.001CCI6.6 ± 2.56.9 ± 2.56.7 ± 2.46.6 ± 2.66.5 ± 2.60.013 Laboratory parameters Hemoglobin (g/dL)10.6 ± 2.310.4 ± 2.310.6 ± 2.310.6 ± 2.410.7 ± 2.30.008WBC (×109/L)11.0 (7.9, 15.1)10.6 (7.0, 14.2)10.5 (7.3, 14.5)10.8 (7.8, 15.0)11.6 (8.4, 15.8)< 0.001Platelet (×109/L)179.0 (132.0, 242.0)221.0 (155.0, 318.0)184.0 (129.0, 250.0)173.0 (127.0, 236.0)180.0 (136.8, 237.0)< 0.001RDW (%)15.0 ± 2.215.5 ± 2.515.0 ± 2.314.9 ± 2.215.0 ± 2.1< 0.001Anion gap (mmol/L)14.7 ± 4.615.1 ± 4.514.9 ± 4.614.6 ± 4.714.7 ± 4.50.079BUN (mg/dL)21.0 (15.0, 33.0)22.0 (16.0, 37.0)22.0 (15.0, 32.0)21.0 (15.0, 33.0)21.0 (16.0, 35.0)0.116Creatinine (mg/dL)1.0 (0.8, 1.5)0.9 (0.6, 1.4)1.0 (0.7, 1.4)1.0 (0.8, 1.5)1.1 (0.8, 1.6)< 0.001Glucose (mg/dL)126.0 (106.0, 156.0)116.0 (98.5, 138.0)122.0 (102.0, 150.0)126.0 (106.0, 154.0)131.0 (110.0, 165.0)< 0.001 Vital signs HR (beats/minute)88.1 ± 21.292.0 ± 22.088.0 ± 21.287.8 ± 21.488.1 ± 20.90.056MBP (mmHg)81.6 ± 18.184.2 ± 21.782.0 ± 17.581.8 ± 18.181.1 ± 18.10.034RR (beats/minute)18.6 ± 5.819.6 ± 6.018.9 ± 6.018.6 ± 6.018.4 ± 5.6< 0.001SpO2 (%)97.1 ± 4.196.6 ± 5.197.1 ± 4.497.2 ± 4.097.1 ± 4.00.085 Therapy Antiplatelet agents3288 (35.4)53 (26.6)775 (32.9)1133 (36.4)1327 (36.7)< 0.001Anticoagulant agents4348 (46.8)103 (51.8)1093 (46.4)1415 (45.4)1737 (48.1)0.077Antiarrhythmic agents4421 (47.6)98 (49.2)1094 (46.4)1472 (47.3)1757 (48.6)0.354MV3586 (38.6)55 (27.6)813 (34.5)1154 (37.1)1564 (43.3)< 0.001RRT657 (7.1)11 (5.5)141 [6]198 (6.4)307 (8.5)< 0.001Vasopressors5111 (55.1)93 (46.7)1226 [52]1719 (55.2)2073 (57.4)< 0.001 Outcomes 30-day all-cause mortality1405 (15.1)66 (33.2)451 (19.1)458 (14.7)430 (11.9)< 0.00190-day all-cause mortality1912 (20.6)82 (41.2)642 (27.2)617 (19.8)571 (15.8)< 0.0011-year all-cause mortality2714 (29.2)109 (54.8)891 (37.8)872 [28]842 (23.3)< 0.001Note: Variables are presented as mean ± SD, median (IQR), or N (%)Abbreviations: BMI, body mass index; CHF, congestive heart failure; PVD, peripheral vascular disease; CPD, chronic pulmonary disease; AKI, acute kidney injury; AHF, acute heart failure; SOFA, sequential organ failure assessment; SAPS II, simplified acute physiology score II; CCI, Charlson comorbidity index; WBC, white blood cell; RDW, red cell distribution width; BUN, blood urea nitrogen; HR, heart rate; MBP, mean blood pressure; RR, respiratory rate; SpO2, saturation of pulse oximetry; MV, mechanical ventilation; RRT, renal replacement therapy ## Effects of BMI on Mortality The Kaplan–Meier curves for 30-day, 90-day, and 1-year survival are presented in Fig. 2 (all P values for log-rank were less than 0.001). Both single-factor and multivariable Cox proportional risk models were conducted to evaluate the association between the BMI and all-cause mortality in critically ill patients with AF. The results of the single-factor Cox regression analysis of covariates and all-cause mortality are presented in Supplementary Table 1. In Table 2, we show both the crude and adjusted models. When BMI was considered a categorical variable (the four BMI categories), we used the normal-weight category as a reference group for comparison with other groups. For 30-day mortality, in the fully adjusted model (Model 2), the HR with $95\%$ CI for the underweight, overweight, and obese categories were 1.58 (1.21, 2.07), 0.82 (0.72, 0.93), and 0.79 (0.68, 0.91), respectively, compared to the normal-weight category (all P values < 0.05). For 90-day mortality, in Model 2, the HR with $95\%$ CI for the underweight, overweight, and obese categories were 1.46 (1.15, 1.85), 0.76 (0.68, 0.85), and 0.7 (0.62, 0.79), respectively, compared to the normal-weight category (all P values < 0.05). For 1-year mortality, in Model 2, the HR with $95\%$ CI for the underweight, overweight, and obese categories were 1.45 (1.19, 1.78), 0.76 (0.69, 0.83), and 0.7 (0.63, 0.77), respectively, compared to the normal-weight category (all P values < 0.05). Fig. 2Kaplan–Meier curves of 30-day (A), 90-day (B), and 1-year (C) all-cause mortality by BMI categories. Abbreviations: BMI, body mass index Table 2Relationship between BMI and all-cause mortality among critically ill patients with atrial fibrillationVariablesCrude modelModel 1Model 2 h ($95\%$ CI)P-valueHR ($95\%$ CI)P-valueHR ($95\%$ CI)P-value30-day all-cause mortalityBMI categoryUnderweight1.89 (1.46, 2.44)< 0.0011.92 (1.48, 2.48)< 0.0011.58 (1.21, 2.07)0.001Normal-weight1.00 (Reference)1.00 (Reference)1.00 (Reference)Overweight0.75 (0.66, 0.85)< 0.0010.84 (0.74, 0.96)< 0.0010.82 (0.72, 0.93)0.003Obese0.6 (0.52, 0.68)< 0.0010.78 (0.68, 0.9)< 0.0010.79 (0.68, 0.91)0.001P for trend< 0.001< 0.001< 0.00190-day all-cause mortalityBMI categoryUnderweight1.7 (1.35, 2.14)< 0.0011.73 (1.38, 2.18)< 0.0011.46 (1.15, 1.85)0.002Normal-weight1.00 (Reference)1.00 (Reference)1.00 (Reference)Overweight0.7 (0.63, 0.78)< 0.0010.79 (0.7, 0.88)< 0.0010.76 (0.68, 0.85)< 0.001Obese0.55 (0.49, 0.61)< 0.0010.71 (0.63, 0.8)< 0.0010.7 (0.62, 0.79)< 0.001P for trend< 0.001< 0.001< 0.0011-year all-cause mortalityBMI categoryUnderweight1.69 (1.39, 2.06)< 0.0011.71 (1.4, 2.09)< 0.0011.45 (1.19, 1.78)< 0.001Normal-weight1.00 (Reference)1.00 (Reference)1.00 (Reference)Overweight0.7 (0.63, 0.76)< 0.0010.78 (0.71, 0.85)< 0.0010.76 (0.69, 0.83)< 0.001Obese0.56 (0.51, 0.62)< 0.0010.72 (0.65, 0.79)< 0.0010.7 (0.63, 0.77)< 0.001P for trend< 0.001< 0.001< 0.001Note: Crude model was adjusted for none; Model 1 was adjusted for age, sex and race; Model 2 was further adjusted (from Model 1) for hypertension, diabetes, CHF, PVD, cerebrovascular disease, CPD, renal disease, liver disease, malignancy, sepsis, AKI, AHF, stroke, SOFA, SAPS II, CCI, hemoglobin, WBC, platelet, RDW, anion gap, BUN, creatinine, glucose, HR, MBP, RR, SpO2, antiplatelet agents, anticoagulant agents, antiarrhythmic agents, MV, RRT, and vasopressorsAbbreviations: BMI, body mass index; HR, hazard ratio; CI, confidence interval; CHF, congestive heart failure; PVD, peripheral vascular disease; CPD, chronic pulmonary disease; AKI, acute kidney injury; AHF, acute heart failure; SOFA, sequential organ failure assessment; SAPS II, simplified acute physiology score II; CCI, Charlson comorbidity index; WBC, white blood cell; RDW, red cell distribution width; BUN, blood urea nitrogen; HR, heart rate; MBP, mean blood pressure; RR, respiratory rate; SpO2, saturation of pulse oximetry; MV, mechanical ventilation; RRT, renal replacement therapy We also treated BMI as a continuous variable. Using multivariable-adjusted restricted cubic spline analysis, an “L-shaped” correlation between BMI and 30-day mortality was evident (P for non-linearity < 0.001), and a similar correlation was found for 90-day and 1-year all-cause mortality (Fig. 3). Combining the graphical interpretation with clinical utility, an optimal inflection point for BMI was determined to be 30 kg/m2. For each 1 kg/m2 increase in BMI when BMI was less than 30 kg/m2, the risk of 30-day mortality decreased by $6.4\%$ (HR, $95\%$ CI: 0.936 [0.918, 0.954]; $P \leq 0.001$). However, this relationship was not present when BMI was greater than or equal to 30 kg/m2 (HR, $95\%$ CI: 0.996 [0.974, 1.017]; $$P \leq 0.689$$). Similarly, for each 1 kg/m2 increase in BMI when BMI was less than 30 kg/m2, the risk of 90-day and 1-year mortality decreased by $6.9\%$ (HR, $95\%$ CI: 0.931 [0.916, 0.947]; $P \leq 0.001$) and $6.8\%$ (HR, $95\%$ CI: 0.932 [0.919, 0.946]; $P \leq 0.001$), respectively. This relationship was also not present when BMI was greater than or equal to 30 kg/m2 for 90-day (HR, $95\%$ CI: 0.992 [0.974, 1.011]; $$P \leq 0.412$$) and 1-year mortality (HR, $95\%$ CI: 0.996 [0.981, 1.011]; $$P \leq 0.588$$) (Table 3). Fig. 3Multivariable-adjusted restricted cubic spline analyses of relationship between BMI and 30-day (A), 90-day (B), and 1-year (C) all-cause mortality. The upper limit of the BMI is restricted to 99th. The purple lines represent the estimated risk of all-cause mortality, and the gray bands represent the point-by-point $95\%$ CI adjusted for covariates. HRs were adjusted for age, sex, race, hypertension, diabetes, CHF, PVD, cerebrovascular disease, CPD, renal disease, liver disease, malignancy, sepsis, AKI, AHF, stroke, SOFA, SAPS II, CCI, hemoglobin, WBC, platelet, RDW, anion gap, BUN, creatinine, glucose, HR, MBP, RR, SpO2, antiplatelet agents, anticoagulant agents, antiarrhythmic agents, MV, RRT, and vasopressors. Abbreviations: BMI, body mass index; HR, hazard ratio; CI, confidence interval; CHF, congestive heart failure; PVD, peripheral vascular disease; CPD, chronic pulmonary disease; AKI, acute kidney injury; AHF, acute heart failure; SOFA, sequential organ failure assessment; SAPS II, simplified acute physiology score II; CCI, Charlson comorbidity index; WBC, white blood cell; RDW, red cell distribution width; BUN, blood urea nitrogen; HR, heart rate; MBP, mean blood pressure; RR, respiratory rate; SpO2, saturation of pulse oximetry; MV, mechanical ventilation; RRT, renal replacement therapy. Table 3Threshold effect analysis of the relationship between BMI and all-cause mortalityThreshold of BMI30-day all-cause mortality90-day all-cause mortality1-year all-cause mortalityHR ($95\%$CI)P-valueHR ($95\%$CI)P-valueHR ($95\%$CI)P-value< 30 kg/m20.936 (0.918, 0.954)< 0.0010.931 (0.916, 0.947)< 0.0010.932 (0.919, 0.946)< 0.001≥ 30 kg/m20.996 (0.974, 1.017)0.6890.992 (0.974, 1.011)0.4120.996 (0.981, 1.011)0.588Likelihood ratio test< 0.001< 0.001< 0.001Note: HRs were adjusted for age, sex, race, hypertension, diabetes, CHF, PVD, cerebrovascular disease, CPD, renal disease, liver disease, malignancy, sepsis, AKI, AHF, stroke, SOFA, SAPS II, CCI, hemoglobin, WBC, platelet, RDW, anion gap, BUN, creatinine, glucose, HR, MBP, RR, SpO2, antiplatelet agents, anticoagulant agents, antiarrhythmic agents, MV, RRT, and vasopressorsAbbreviations: BMI, body mass index; HR, hazard ratio; CI, confidence interval; CHF, congestive heart failure; PVD, peripheral vascular disease; CPD, chronic pulmonary disease; AKI, acute kidney injury; AHF, acute heart failure; SOFA, sequential organ failure assessment; SAPS II, simplified acute physiology score II; CCI, Charlson comorbidity index; WBC, white blood cell; RDW, red cell distribution width; BUN, blood urea nitrogen; HR, heart rate; MBP, mean blood pressure; RR, respiratory rate; SpO2, saturation of pulse oximetry; MV, mechanical ventilation; RRT, renal replacement therapy ## Subgroup analyses and sensitivity analyses Further subgroup analyses were performed on 5670 patients with BMI < 30 kg/m2 to verify the consistency of the correlation between mortality and BMI. The interaction tests were not statistically significant for most stratification variables (Supplementary Fig. 1). Hypertension, AHF, and AKI modified the association between 30-day mortality and BMI. Hypertension, sepsis, AHF, and AKI modified the association between 90-day mortality and BMI. Sex, hypertension, and AHF modified the association between 1-year mortality and BMI. However, in all of these interactive stratifications, BMI acted as a protective factor. In the sensitivity analyses, results remained robust after excluding participants with incomplete data of covariates (Supplementary Tables 2, Supplementary Tables 3, and Supplementary Fig. 2). When patients with BMI ≥ 30 kg/m2 were divided into obese and morbidly obese group, the results of multivariable Cox regression analyses indicated that morbidly obese patients still had a survival advantage, compared to normal-weight patients (Supplementary Table 4). For 30-day mortality, in the fully adjusted model, the HR with $95\%$ CI for the morbidly obese category was 0.77 (0.6, 0.98), compared to the normal-weight category ($$P \leq 0.034$$). Similar results were found for 90-day and 1-year mortality. ## Discussion This study focused on the association between BMI and short- and medium-term mortality among critically ill patients with AF. We found that underweight patients had higher 30-day, 90-day, and 1-year all-cause mortality compared with normal-weight patients, even after adjusting for essential confounders, including important disease severity scores (SOFA, SAPS II, and CCI). Conversely, overweight and obese patients had a relatively low risk of death. Furthermore, when BMI was considered a continuous variable, we observed a curvilinear association between BMI and mortality among critically ill patients with AF. The dose-response effect of BMI on all-cause mortality was significantly different when it was below or above the threshold of 30 kg/m2, showing that a threshold effect was present. Stratified analyses and sensitivity analyses suggested the robustness of our results. This information will contribute to an in-depth understanding of the relationship between the BMI and all-cause mortality of critically ill patients with AF, thereby strengthening the physician’s ability to risk-stratify patients. The obesity paradox has recently received a lot of attention in various diseases. In previous studies, overweight and obese patients with AF were reported to have a long-term survival advantage compared to normal-weight patients [12, 13, 23]. Obesity has also been shown to have a protective effect in other critically ill patient populations, including sepsis [24], AKI [25], and coronary care unit patients [26]. Most of the participants in these studies were from Western populations, and our study cohort also included only a very small number of Asians. Differences in obesity standards between Asian and Western populations may lead to ethnic differences in findings on the obesity paradox. Nevertheless, in a Chinese study, overweight AF patients had lower all-cause mortality and cardiovascular mortality compared with normal-weight and underweight patients [27]. In another study conducted in Japan, being overweight was associated with a reduced risk of all-cause death among AF patients [28]. Our findings in critically ill patients with AF were in accordance with these studies. However, a prospective cohort study showed that overweight and obesity are risk factors for adverse clinical outcomes among AF patients [19]. In this research, the study population was not critically ill patients, and the study outcome was a composite endpoint of ischemic stroke, thromboembolism, or death. Therefore, the results do not represent the relationship between obesity and mortality in critically ill patients with AF. Another study conducted by Wang et al. showed that overweight or obesity may be a risk factor for poor prognoses in AF patients [29]. But their study population was also not exactly the same as ours. Although there are a lot of studies on the obesity paradox, the detailed mechanisms remain poorly understood. One reason for this phenomenon may relate to the higher metabolic reserve in patients with obesity. As reported, AF patients with a higher BMI are better able to endure the increased catabolic stress associated with disease development [30]. A catabolic state with increased energy expenditure is an important feature in the early stages of critical illness [31]. Patients with obesity have a better ability to supply more substrate synthesis energy to fulfill the increased demands during such a period because of their large lipid reserves [32]. In addition to increasing available energy, adipose tissue can play a role in critical illness by improving insulin sensitivity, reducing dyslipidemia, and enhancing thermoregulation. Although the underlying mechanism is not fully understood, white adipose tissue browning may be an important pathway for these effects [33]. Other protective effects provided by adipose tissue and adipocytes have been reported in previous studies as well. Adipocytes release adipokines and inflammatory factors, including leptin and interleukin-10, which may attenuate adverse immunological responses and thus help improve survival from critical diseases [34]. Adiponectin, a peptide secreted by adipocytes, was also reported to play an important role in critical illness. It is recognized as an insulin sensitizer, anti-atherosclerotic agent, and anti-inflammatory agent. Lower adiponectin levels on admission, followed by gradual elevation, may be a useful signal for a better prognosis of critical illness [35]. There is also evidence that patients with obesity can supply more lipoproteins, which can bind to endotoxins and lessen their toxic actions [36]. Additionally, tumor necrosis factor-alpha receptor has been observed to be upregulated in adipose tissue, which may aid in the dispersal of cardiomyocyte-activated arrhythmogenic substrates and inflammation [37]. Finally, obese patients have lower levels of natriuretic peptide, which has been shown to predict stroke and death among AF patients [38]. Other potential mechanisms included the fact that obese patients receive more attention from clinical staff and differences in baseline characteristics across BMI categories. Whether the obesity paradox really exists or is the result of selection bias has been hotly debated. Obese patients are often considered to have an increased risk of mortality and complications than normal-weight patients, which may result in earlier ICU admissions and more aggressive treatments [39, 40]. Interestingly, in our study population, overweight and obese patients received more relevant treatments, including MV, RRT, vasopressors, and antiplatelet agents. This is in line with the findings of these previous studies. Use of MV, RRT, and vasopressors are important organ support therapies in critically ill patients. Antiplatelet agents are regularly used for the prevention of stroke and thrombotic events in patients with atrial fibrillation [41, 42]. These aggressive treatments may be one of the reasons for the reduced risk of death in obese patients. Additionally, previous studies have shown that obese and overweight patients in the ICU tend to be younger, and our study has the same results [24, 43]. In our study, we observed a higher prevalence of hypertension and diabetes among overweight and obese patients, which may result in an earlier onset of AF [44]. Age is a significant predictor for stroke and mortality [45]; consequently, obese patients with AF may have better outcomes due to their younger age. In addition, it seems logical that, in our data, obese patients had a lower proportion of comorbid strokes in comparison to normal-weight patients. It has also been shown that less cachexia is related to the obesity paradox [46]. As a result, it is not surprising that underweight patients presented more history of malignancy in our cohort. Notwithstanding adjustments for age and other confounders in the multivariable regression models, this may not fully account for the differences in baseline characteristics across BMI categories. The findings should therefore be interpreted with caution. Lower sympathetic activity in obese patients may also be involved in the mechanisms of the obesity paradox [47]. Studies have demonstrated that sympathetic overdrive plays a significant role in the progression of AF and is associated with poorer outcomes among AF patients [48]. Hence, lower activation of the sympathetic nervous system may have a protective effect in obese patients. Another potential mechanism focused on the genetic factors. Lean patients with cardiovascular disease (CVD) may develop CVD because of their completely different etiologies and genetic predispositions, which may be associated with poorer outcomes [49]. The current study has several limitations. First, data from the MIMIC-IV database did not include long-term mortality over one year or cardiovascular mortality, which limited our further analysis of the association between obesity and outcomes in critically ill patients with AF. Second, due to the complexity of ICD codes, errors may occur when they are used by medical staff. Therefore, identifying AF patients by ICD codes might not be the most accurate method. Unfortunately, we were unable to find a more accurate way to identify AF patients because of the nature of the MIMIC-IV database. Additionally, BMI is unable to quantify body fat percentage or distribution because it represents the sum of fat-free mass index and fat-mass index [50]. As in many previous studies, the definition of obesity in our study was based on BMI, which does not take into account different obesity phenotypes. Recent studies have suggested that various obesity phenotypes with different cardiovascular risk profiles coexist within the same BMI category [51]. When markers of central obesity, such as waist circumference or waist-hip ratio, are used instead of BMI, the opposite results may occur. Hence, to overcome the limitations of the traditional definition of obesity, a new classification of obesity based on different variables needs to be established in future studies. Finally, although our study from the MIMIC-IV database has a large sample size, there were only a small number of participants in the underweight category, which may affect the statistical power and reliability of our data analyses. Despite the above limitations, this study provides valuable information on the association between BMI and prognoses among critically ill patients with AF. ## Conclusions This study revealed a nonlinear relationship between BMI and short- and medium-term all-cause mortality among critically ill patients with AF. All-cause mortality and the BMI were negatively correlated when the BMI was less than 30 kg/m2. These findings suggested that the obesity paradox may also be suitable for critically ill patients with AF. The mechanisms underlying this relationship were worthy of further investigation. ## Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 ## References 1. Benjamin EJ, Muntner P, Alonso A. **Heart Disease and Stroke Statistics-2019 update: a Report from the American Heart Association**. *Circulation* (2019.0) **139** e56-28. DOI: 10.1161/CIR.0000000000000659 2. Chiang CE, Okumura K, Zhang S. **2017 consensus of the Asia Pacific Heart Rhythm Society on stroke prevention in atrial fibrillation**. *J Arrhythm* (2017.0) **33** 345-67. DOI: 10.1016/j.joa.2017.05.004 3. Hindricks G, Potpara T, Dagres N. **2020 ESC Guidelines for the diagnosis and management of atrial fibrillation developed in collaboration with the European Association for Cardio-Thoracic surgery (EACTS): the Task Force for the diagnosis and management of atrial fibrillation of the European Society of Cardiology (ESC) developed with the special contribution of the European Heart Rhythm Association (EHRA) of the ESC**. *Eur Heart J* (2021.0) **42** 373-498. DOI: 10.1093/eurheartj/ehaa612 4. Kalra S, Kishor K, Batra A, Aggarwal S. **Atrial fibillation in diabetes: need for cardiovigilance**. *J Pak Med Assoc* (2019.0) **69** 437-9. PMID: 30890843 5. Moss TJ, Calland JF, Enfield KB. **New-Onset Atrial Fibrillation in the critically ill**. *Crit Care Med* (2017.0) **45** 790-7. DOI: 10.1097/CCM.0000000000002325 6. Bosch NA, Cimini J, Walkey AJ. **Atrial fibrillation in the ICU**. *Chest* (2018.0) **154** 1424-34. DOI: 10.1016/j.chest.2018.03.040 7. Shao Y, Hu J. **U-Shaped Association between blood pressure and mortality risk in ICU patients with Atrial Fibrillation: the MIMIC-III database**. *Front Cardiovasc Med* (2022.0) **9** 866260. DOI: 10.3389/fcvm.2022.866260 8. Chooi YC, Ding C, Magkos F. **The epidemiology of obesity**. *Metabolism* (2019.0) **92** 6-10. DOI: 10.1016/j.metabol.2018.09.005 9. Wang TJ, Parise H, Levy D. **Obesity and the risk of new-onset atrial fibrillation**. *JAMA* (2004.0) **292** 2471-7. DOI: 10.1001/jama.292.20.2471 10. Guglin M, Maradia K, Chen R, Curtis AB. **Relation of obesity to recurrence rate and burden of atrial fibrillation**. *Am J Cardiol* (2011.0) **107** 579-82. DOI: 10.1016/j.amjcard.2010.10.018 11. Tsang TS, Barnes ME, Miyasaka Y. **Obesity as a risk factor for the progression of paroxysmal to permanent atrial fibrillation: a longitudinal cohort study of 21 years**. *Eur Heart J* (2008.0) **29** 2227-33. DOI: 10.1093/eurheartj/ehn324 12. Sandhu RK, Ezekowitz J, Andersson U. **The ‘obesity paradox’ in atrial fibrillation: observations from the ARISTOTLE (Apixaban for reduction in stroke and other thromboembolic events in Atrial Fibrillation) trial**. *Eur Heart J* (2016.0) **37** 2869-78. DOI: 10.1093/eurheartj/ehw124 13. Pandey A, Gersh BJ, McGuire DK. **Association of Body Mass Index with Care and Outcomes in patients with Atrial Fibrillation: results from the ORBIT-AF Registry**. *JACC Clin Electrophysiol* (2016.0) **2** 355-63. DOI: 10.1016/j.jacep.2015.12.001 14. Curtis JP, Selter JG, Wang Y. **The obesity paradox: body mass index and outcomes in patients with heart failure**. *Arch Intern Med* (2005.0) **165** 55-61. DOI: 10.1001/archinte.165.1.55 15. Uretsky S, Messerli FH, Bangalore S. **Obesity paradox in patients with hypertension and coronary artery disease**. *Am J Med* (2007.0) **120** 863-70. DOI: 10.1016/j.amjmed.2007.05.011 16. Curcic IB, Berkovic MC, Kuna L. **Obesity Paradox in Chronic Liver Diseases: product of Bias or a real thing**. *J Clin Transl Hepatol* (2019.0) **7** 275-9. DOI: 10.14218/JCTH.2019.00029 17. Chittal P, Babu AS, Lavie CJ. **Obesity paradox: does fat alter outcomes in chronic obstructive pulmonary disease**. *COPD* (2015.0) **12** 14-8. DOI: 10.3109/15412555.2014.915934 18. 18.Zhu W, Wan R, Liu F et al. Relation of Body Mass Index With Adverse Outcomes Among Patients With Atrial Fibrillation: A Meta-Analysis and Systematic Review.J Am Heart Assoc. 2016. 5(9). 19. Overvad TF, Rasmussen LH, Skjøth F, Overvad K, Lip GY, Larsen TB. **Body mass index and adverse events in patients with incident atrial fibrillation**. *Am J Med* (2013.0) **126** 640e9-17. DOI: 10.1016/j.amjmed.2012.11.024 20. Zubair Khan M, Gupta A, Patel K. **Association of atrial fibrillation and various cancer subtypes**. *J Arrhythm* (2021.0) **37** 1205-14. DOI: 10.1002/joa3.12589 21. Tripathi B, Atti V, Kumar V. **Outcomes and resource utilization Associated with Readmissions after Atrial Fibrillation Hospitalizations**. *J Am Heart Assoc* (2019.0) **8** e013026. DOI: 10.1161/JAHA.119.013026 22. Jaddoe VW, de Jonge LL, Hofman A, Franco OH, Steegers EA, Gaillard R. **First trimester fetal growth restriction and cardiovascular risk factors in school age children: population based cohort study**. *BMJ* (2014.0) **348** g14. DOI: 10.1136/bmj.g14 23. Proietti M, Lane DA, Lip GY. **Relation of Nonvalvular Atrial Fibrillation to Body Mass Index (from the SPORTIF trials)**. *Am J Cardiol* (2016.0) **118** 72-8. DOI: 10.1016/j.amjcard.2016.04.013 24. Li S, Hu X, Xu J. **Increased body mass index linked to greater short- and long-term survival in sepsis patients: a retrospective analysis of a large clinical database**. *Int J Infect Dis* (2019.0) **87** 109-16. DOI: 10.1016/j.ijid.2019.07.018 25. 25.Wang B, Li D, Gong Y, Ying B, Cheng B, Sun L. Body Mass Index Is Associated with the Severity and All-Cause Mortality of Acute Kidney Injury in Critically Ill Patients: An Analysis of a Large Critical Care Database. Biomed Res Int. 2021. 2021: 6616120. 26. Yan J, Li X, Long W, Yuan T, Xian S. **Association between obesity and lower short- and long-term mortality in Coronary Care Unit Patients: a cohort study of the MIMIC-III database**. *Front Endocrinol (Lausanne)* (2022.0) **13** 855650. DOI: 10.3389/fendo.2022.855650 27. Wang J, Yang YM, Zhu J. **Overweight is associated with improved survival and outcomes in patients with atrial fibrillation**. *Clin Res Cardiol* (2014.0) **103** 533-42. DOI: 10.1007/s00392-014-0681-7 28. Inoue H, Kodani E, Atarashi H, Okumura K, Yamashita T, Origasa H. **Impact of body Mass Index on the prognosis of japanese patients with non-valvular atrial fibrillation**. *Am J Cardiol* (2016.0) **118** 215-21. DOI: 10.1016/j.amjcard.2016.04.036 29. Wang HJ, Si QJ, Shan ZL, Guo YT, Lin K, Zhao XN. **Effects of body mass index on risks for ischemic stroke, thromboembolism, and mortality in chinese atrial fibrillation patients: a single-center experience**. *PLoS ONE* (2015.0) **10** e0123516. DOI: 10.1371/journal.pone.0123516 30. Davos CH, Doehner W, Rauchhaus M. **Body mass and survival in patients with chronic heart failure without cachexia: the importance of obesity**. *J Card Fail* (2003.0) **9** 29-35. DOI: 10.1054/jcaf.2003.4 31. Secombe P, Harley S, Chapman M, Aromataris E. **Feeding the critically ill obese patient: a systematic review protocol**. *JBI Database System Rev Implement Rep* (2015.0) **13** 95-109. DOI: 10.11124/jbisrir-2015-2458 32. Karampela I, Chrysanthopoulou E, Christodoulatos GS, Dalamaga M. **Is there an obesity Paradox in critical illness? Epidemiologic and metabolic considerations**. *Curr Obes Rep* (2020.0) **9** 231-44. DOI: 10.1007/s13679-020-00394-x 33. Alipoor E, Hosseinzadeh-Attar MJ, Rezaei M, Jazayeri S, Chapman M. **White adipose tissue browning in critical illness: a review of the evidence, mechanisms and future perspectives**. *Obes Rev* (2020.0) **21** e13085. DOI: 10.1111/obr.13085 34. Hauner H. **Adipose tissue inflammation: are small or large fat cells to blame**. *Diabetologia* (2010.0) **53** 223-5. DOI: 10.1007/s00125-009-1605-3 35. Alipoor E, Mohammad Hosseinzadeh F, Hosseinzadeh-Attar MJ. **Adipokines in critical illness: a review of the evidence and knowledge gaps**. *Biomed Pharmacother* (2018.0) **108** 1739-50. DOI: 10.1016/j.biopha.2018.09.165 36. Umbrello M, Fumagalli J, Pesenti A, Chiumello D. **Pathophysiology and management of Acute respiratory distress syndrome in obese patients**. *Semin Respir Crit Care Med* (2019.0) **40** 40-56. DOI: 10.1055/s-0039-1685179 37. Lee SH, Chen YC, Chen YJ. **Tumor necrosis factor-alpha alters calcium handling and increases arrhythmogenesis of pulmonary vein cardiomyocytes**. *Life Sci* (2007.0) **80** 1806-15. DOI: 10.1016/j.lfs.2007.02.029 38. Hijazi Z, Wallentin L, Siegbahn A. **N-terminal pro-B-type natriuretic peptide for risk assessment in patients with atrial fibrillation: insights from the ARISTOTLE Trial (Apixaban for the Prevention of Stroke in subjects with Atrial Fibrillation)**. *J Am Coll Cardiol* (2013.0) **61** 2274-84. DOI: 10.1016/j.jacc.2012.11.082 39. Ball L, Serpa Neto A, Pelosi P. **Obesity and survival in critically ill patients with acute respiratory distress syndrome: a paradox within the paradox**. *Crit Care* (2017.0) **21** 114. DOI: 10.1186/s13054-017-1682-5 40. O’Brien JM, Philips GS, Ali NA, Aberegg SK, Marsh CB, Lemeshow S. **The association between body mass index, processes of care, and outcomes from mechanical ventilation: a prospective cohort study**. *Crit Care Med* (2012.0) **40** 1456-63. DOI: 10.1097/CCM.0b013e31823e9a80 41. Chang KW, Xian Y, Zhao X, Mi X, Matsouaka R, Schwamm LH. **Antiplatelet patterns and outcomes in patients with atrial fibrillation not prescribed an anticoagulant after stroke**. *Int J Cardiol* (2020.0) **321** 88-94. DOI: 10.1016/j.ijcard.2020.08.011 42. Vandiver JW, Diane Beavers K. **Combining oral anticoagulation and antiplatelet therapies: appropriate patient selection**. *J Thromb Thrombolysis* (2018.0) **45** 423-31. DOI: 10.1007/s11239-018-1635-0 43. Abhyankar S, Leishear K, Callaghan FM, Demner-Fushman D, McDonald CJ. **Lower short- and long-term mortality associated with overweight and obesity in a large cohort study of adult intensive care unit patients**. *Crit Care* (2012.0) **16** R235. DOI: 10.1186/cc11903 44. Aksnes TA, Schmieder RE, Kjeldsen SE, Ghani S, Hua TA, Julius S. **Impact of new-onset diabetes mellitus on development of atrial fibrillation and heart failure in high-risk hypertension (from the VALUE Trial)**. *Am J Cardiol* (2008.0) **101** 634-8. DOI: 10.1016/j.amjcard.2007.10.025 45. Joundi RA, Cipriano LE, Sposato LA, Saposnik G. **Ischemic stroke risk in patients with Atrial Fibrillation and CHA2DS2-VASc score of 1: systematic review and Meta-analysis**. *Stroke* (2016.0) **47** 1364-7. DOI: 10.1161/STROKEAHA.115.012609 46. Lavie CJ, McAuley PA, Church TS, Milani RV, Blair SN. **Obesity and cardiovascular diseases: implications regarding fitness, fatness, and severity in the obesity paradox**. *J Am Coll Cardiol* (2014.0) **63** 1345-54. DOI: 10.1016/j.jacc.2014.01.022 47. Esler M, Lambert G, Schlaich M, Dixon J, Sari CI, Lambert E. **Obesity Paradox in Hypertension: is this because sympathetic activation in obesity-hypertension takes a Benign Form**. *Hypertension* (2018.0) **71** 22-33. DOI: 10.1161/HYPERTENSIONAHA.117.09790 48. Carnagarin R, Kiuchi MG, Ho JK, Matthews VB, Schlaich MP. **Sympathetic nervous system activation and its modulation: role in Atrial Fibrillation**. *Front Neurosci* (2018.0) **12** 1058. DOI: 10.3389/fnins.2018.01058 49. Lavie CJ, De Schutter A, Parto P. **Obesity and prevalence of Cardiovascular Diseases and prognosis-the obesity Paradox updated**. *Prog Cardiovasc Dis* (2016.0) **58** 537-47. DOI: 10.1016/j.pcad.2016.01.008 50. Donini LM, Pinto A, Giusti AM, Lenzi A, Poggiogalle E. **Obesity or BMI Paradox? Beneath the tip of the Iceberg**. *Front Nutr* (2020.0) **7** 53. DOI: 10.3389/fnut.2020.00053 51. Vecchié A, Dallegri F, Carbone F, Bonaventura A, Liberale L, Portincasa P. **Obesity phenotypes and their paradoxical association with cardiovascular diseases**. *Eur J Intern Med* (2018.0) **48** 6-17. DOI: 10.1016/j.ejim.2017.10.020
--- title: 'Impact of hypertension diagnosis on morbidity and mortality: a retrospective cohort study in primary care' authors: - Jesus Martín-Fernández - Tamara Alonso-Safont - Elena Polentinos-Castro - Maria Dolores Esteban-Vasallo - Gloria Ariza-Cardiel - Mª Isabel González-Anglada - Luis Sánchez-Perruca - Gemma Rodríguez-Martínez - Rafael Rotaeche-del-Campo - Amaia Bilbao-González journal: BMC Primary Care year: 2023 pmcid: PMC10037862 doi: 10.1186/s12875-023-02036-2 license: CC BY 4.0 --- # Impact of hypertension diagnosis on morbidity and mortality: a retrospective cohort study in primary care ## Abstract ### Background Hypertension is responsible for a huge burden of disease. The aim of this study was to evaluate the impact of newly diagnosed hypertension on the occurrence of kidney or cardiovascular events (K/CVEs) and on mortality among community dwellers. ### Methods Retrospective cohort study, conducted from January, 2007, to December, 2018. All patients (age > 18) newly diagnosed with hypertension and no previous K/CVEs in 2007 and 2008, in the primary care centers of Madrid (Spain) ($$n = 71$$,770), were enrolled. The control group ($$n = 72$$,946) included patients without hypertension, matched by center, sex and age. The occurrence of kidney or CV events, including mortality from these causes and total mortality were evaluated using Cox regression and multistate models. Data were collected from three sources: personal data from administrative records, clinical data from medical records, and mortality data from regional and national databases. ### Results The median follow-up was 138.61 months (IQR: 124.68–143.97 months). There were 32,896 K/CVEs (including 3,669 deaths from these causes) and 12,999 deaths from other causes. Adjusted for sex, smoking, diabetes and socioeconomic status, K/CVEs HR was 4.36 ($95\%$ CI: 3.80–5.00) for diagnoses before 45 years of age, 2.45($95\%$ CI: 2.28- 2.63) for diagnosis between 45 to 54 years, and HR decreased to 1.86 ($95\%$ CI: 1.64–210) for diagnoses over age 85. Total mortality risk was only higher for hypertension diagnosed before 55 years of age (HR: 2.47, $95\%$ CI: 1.90–3.19 for ages 18 to 44; and HR: 1.14, $95\%$ CI: 1.02–1.28 for ages 45 to 54). ### Conclusion The diagnosis of hypertension in the community environment, in patients without evidence of previous kidney or CV disease, is associated with a large increase in the risk of K/CVEs, but especially in individuals diagnosed before the age of 55. This diagnosis is only associated with an increase in kidney or cardiovascular mortality or overall mortality when it occurs before age 55. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12875-023-02036-2. ## Background Arterial hypertension (HTN) is one of the most prevalent pathological conditions. One in three people over age 30 has been diagnosed with HTN, and although the age-adjusted prevalence has remained stable, the total number of diagnoses has doubled in the past 30 years [1]. HTN is an enormous burden responsible for the loss of 143 million disability-adjusted life years (DALYs) worldwide by 2015, considering the threshold of $\frac{140}{90}$ mm Hg for its definition. These figures represented an increase of more than $30\%$ of the DALYs lost for the same reason in 1990 [2]. The prevalence of HTN increases throughout life; among a cohort of previously healthy patients ages 25, 45, and 65, $0.3\%$, $6.5\%$, and $37\%$ were diagnosed, respectively [3]. The number of people over 65 is growing steadily. Specifically, in Europe, it is expected that their number will double during the next 50 years, reaching 150 million people, and those who reach the average life expectancy without HTN have more than a $90\%$ probability of developing the disease during their remaining life [4]. The excess mortality produced by HTN is mainly mediated by CV disease [2, 5]. Although the control of HTN through pharmacological and lifestyle measures has been shown to decrease mortality from these causes [6–10], it seems that hypertensive patients have an excess risk of CVEs [11] and overall mortality [12–14]. There has been a reduction in mortality in hypertensive patients over time, but there is a differential mortality compared to nonhypertensive patients with the same characteristics [15]. However, some authors have questioned this interpretation [16]. The diagnosis of hypertension occurs more frequently in people with other cardiovascular risk factors (CVRFs) [17]. Further, the presence of certain inflammatory markers, which have been associated with cardiovascular disease, is associated with the risk of being diagnosed with hypertension [18]. In fact, the existence of other CVRFs in patients newly diagnosed with hypertension is more frequent than in the general population of the same age [19]. Although the association between hypertension and cardiovascular disease, and mortality seems solid, the causality is debatable. The impact of hypertension on overall mortality is attenuated as a function of the age of onset, going from mortality risks 2.5 times higher when hypertension is diagnosed before the age of 45 to excess mortality risk lower than $30\%$ when this diagnosis is made over 65 years [19]. It is possible that the increase in blood pressure levels with age may constitute a protective mechanism against the dysfunction of certain organs [20, 21]. Additionally, it has been described that the differences in mortality (by cardiovascular or global cause) between patients with and without hypertension disappear when only the group of patients with treated and good control is considered [22]. In this context, the objective of evaluating the impact of newly diagnosed hypertension on the occurrence of kidney or cardiovascular events, mortality from these causes and total mortality in different age groups in the community under clinical practice conditions was proposed. ## Method A retrospective cohort study was designed. Inclusion criteria for the hypertensive cohort included to be over 18 years of age at recruitment, have been diagnosed of hypertension (code CIAP2 K86) from January 1, 2007, to December 31, 2008, and the absence of kidney or cardiovascular disease prior to such diagnosis. The diagnosis of hypertension implied that the mean of two or more correctly measured systolic blood pressure readings at each of two or more clinic visits was ≥ 140 mmHg or that diastolic blood pressure readings at each of two or more clinic visits was ≥ 90 mmHg. When the record referred to a diagnosis prior to that time, the subject was excluded. The comparison cohort was constructed by pairing each individual with another person from the same Primary Care Center (PCC), without hypertension of the same sex and age range who did not have kidney or cardiovascular (CV) disease. Exclusion criteria were being younger than 18 years, having suffered a kidney disease or CV event or having been diagnosed with hypertension before the start of the study. Subjects in each cohort were selected from all PCC in the Community of Madrid. Additional file 1 details the construction of both cohorts. The follow-up lasted until December 31, 2018, or until the patient was removed from the community health records or died. Sociodemographic and clinical variables were collected. Age in years at diagnosis, sex and deprivation index of the area at the time of inclusion in the study were recorded. This deprivation index was developed for the MEDEA Project using Principal Component Analysis with the Census data. MEDEA index detects small areas of large cities with unfavorable socioeconomic characteristics and is related to general mortality [23]. The index was assigned to each census area using the following five socioeconomic indicators: manual workers, unemployment, temporary wage earners, total insufficient education and in youth. Each patient was assigned the MEDEA Index (in quintiles, the fifth quintile represents the least advantaged group) of their PCC, as an approximation to the place of residence. The presence of the following clinical conditions recorded in the Clinical History of Primary Care, which uses the International Classification of Primary Care (ICPC-2), was collected [24]: Diabetes Mellitus (DM)-ICPC2 T89 and T90-, tobacco use -ICPC2 P17-, or any reference to active tobacco consumption in the Electronic Health Record (HER) at the time of inclusion or in the year prior to inclusion. In the follow-up, three types of dependent variables were collected: Occurrence of kidney or cardiovascular event (K/CVE): ischemic heart disease (acute myocardial infarction (ICPC2 K75), angina–(ICPC2 K74), chronic ischemia (ICPC2 K76), heart failure (ICPC2 K77), cerebrovascular disease (ICPC2 K90, K91), peripheral arterial disease (ICPC2 K92), chronic kidney disease (ICPC2 U99.1), or appearance of maintained urinary microalbuminuria, or proteinuria. Mortality from all causes. The International Classification of Diseases 10th edition (ICD-10) was used to study the causes of mortality [25].Kidney or CV mortality: deaths due to chronic kidney disease (ICD10: N18), cerebrovascular accident (ICD10: G46; I60-I69), ischaemic heart disease (ICD10: I20-I25), heart failure (ICD10: I50) and peripheral arterial disease (ICD10: I70, I71, I72, I74), were classified as kidney or CV mortality. The appearance of hypertension in the cohort that initially did not express this condition was also collected. ## Data sources Potential participants were identified by applying the eligibility criteria to the Center for Basic Strategic Information for Health care Environments (CIBELES). Clinical data were collected using a coding algorithm from the Centralized EHR for Primary Care PC of the Community of Madrid (AP-Madrid®). The EHR electronic source was linked to the mortality database of the Statistics National Institute and copied to a normalized database. ## Analysis Prior to analysis, investigators implemented and verified several data quality processes for error identification and had access to the database population. For the study of the occurrence of events, it should be taken into account that the subjects of the unexposed cohort could be diagnosed with hypertension in the follow-up. To use time-dependent covariates, the observation periods must be broken down into parts, depending on whether there is exposure. Once this procedure is performed, the data can be analyzed using a Cox proportional hazards model [26]. Given that contextual data were used, standard errors were calculated using robust methods and adjusted for 401 clusters (centers) [27]. The first analytical approach was Cox regression with time-dependent covariates. On the other hand, mortality from kidney or CV causes and mortality from other causes can be considered competitive risks. One of the events could increase by decreasing the other. To address this problem, multistate models were constructed [28]. These models assume that the probability of transitioning to another state only depends on the present situation and allows treating some competitive risks as mutually exclusive absorbing states. Each transition between the states can be evaluated by Cox regressions, and the probabilities of transition to the same state from two different intermediate states can be compared [26]. The transitions defined for the multistate models can be seen in Fig. 1a and b. In our case, the comparisons of interest were transitions 2 versus 4 and 3 versus 5 to assess the risk of death from other causes and death from kidney or CV causes, respectively (Fig. 1a). To assess the impact of the association between HTN and total mortality, transitions 2 and 3 of Fig. 1b were studied. The comparison between models was made by assessing the Akaike Information Criteria (AIC) and Bayes Information Criteria (BIC) [29]. The analyses were performed with Stata® 14 using the “multistate” module designed by Crowther and Lambert [30].Fig. 1Transitions evaluated using multistate models. a Mortality from different causes. b Total mortality ## Ethical and legal aspects A positive opinion was obtained from the Ethical Committee of the Alcorcon Foundation University Hospital. This study was funded by the Carlos III Health Institute (ISCIII) through project PI$\frac{18}{00370}$ and co-financed by the European Union. ## Results A total of 71,770 patients with an incidental diagnosis of hypertension and 72,946 matched controls by age group, sex and health center were included in the study (see Additional file 1). Table 1 shows the baseline characteristics of both cohorts. Table 1Characteristics of the selected cohortsHypertensive cohortNot hypertensive at baselineTotal71.77072.946Ages 18 to 44 years$19.06\%$$19.13\%$ 45 to 54 years$24.30\%$$24.29\%$ 55 to 64 years$26.33\%$$26.33\%$ 65 to 74 years$18.70\%$$18.63\%$ 75 to 84 years$9.54\%$$9.53\%$ 85 and older$2.07\%$$2.09\%$Women$51.76\%$$51.78\%$Smokers$16.39\%$$18.05\%$Diabetes mellitus$10.15\%$$3.97\%$ The median follow-up was 138.61 months (interquartile range, IQR 124.68–143.97 months). During the follow-up time, 15,042 patients in the unexposed cohort were diagnosed with hypertension. Of these, 1,327 had suffered a K/CVE prior to the diagnosis of hypertension, so their observation period ended at that time (with the occurrence of K/CVE). The remaining 13,715 were considered subjects with hypertension. ## Study of the occurrence of kidney or CV events A total of 32,896 K/CVEs were recorded, including death due to these causes, 13,008 in the initially unexposed cohort ($17.83\%$; $95\%$ CI: 17.55–$18.11\%$) and 19,888 ($27.71\%$; $95\%$ CI: 27.38–$28.03\%$) in the cohort diagnosed with hypertension. For patients who had a K/CVE, the median follow-up until the event was 70.14 months (IQR: 39.79–100.86 months). The follow-up was performed on 144,716 subjects, with a total of 16,462,184.38 person-months of observation and an event incidence rate of 0.0020 per person-month. The risk of event occurrence was studied for each age range (constructing an age-hypertension interaction) and adjusting it for the following covariates: sex, smoking, diabetes diagnosis and socioeconomic status of the area. Table 2 shows the results of the best Cox model. The risk of event occurrence is up to four times higher when hypertension diagnosis is made between ages 18 and 44 and gradually decreases with increasing age of diagnosis, but the association is relevant in all age ranges. Table 2Cox model for cardiovascular events, including kidney or cardiovascular mortality, hypertensive vs. non-hypertensive subjectsVariableHRHR CI $95\%$ p valueHypertension in each age group a 18 to 44 years4.3583.799- 4.999 < 0.001 45 to 54 years2.4452.275–2.628 < 0.001 55 to 64 years1.9621.870–2.060 < 0.001 65 to 74 years1.9191.836–2.005 < 0.001 75 to 84 years1.7181.629–1.812 < 0.001 85 and older1.8561.644–2.096 < 0.001Female vs. male0.8660.844–0.889 < 0.001Diabetes mellitus1.4941.443–1.547 < 0.001Baseline smoking1.3111.274–1.349 < 0.001Socioeconomic group < 0.001 2nd vs. 1st quintile1.0720.997–1.1530.061 3rd vs. 1st quintile1.1691.083- 1.263 < 0.001 4th vs. 1st quintile1.1781.101–1.261 < 0.001 5th vs. 1st quintile1.2241.128–1.328 < 0.001Characteristics of the modelAkaike Information Criteria (AIC): 739,153.3Bayes Information Criteria (BIC): 739,332.7 a The HR (hazard ratio) expresses the risk of an event calculated from the linear combination of the hypertension coefficients and their interaction with age ## Study of global mortality The follow-up was performed on 144,716 subjects, with a total of 18,137,117.42 person-months of observation and an incidence rate of death of 0.0009 per person-month. At the end of the follow-up, 16,668 subjects had died, $10.74\%$ ($95\%$ CI: 10.51–$10.97\%$) of the cohort initially diagnosed with hypertension and $12.28\%$ ($95\%$ CI: 12.05- $12.52\%$) of the cohort initially undiagnosed. Of the deaths observed, 3,669 were caused by kidney or CV events and 12,999 by other causes. For patients who died, the median follow-up was 84.40 months (IQR: 52.53–112.21 months). The overall mortality risk was studied for each age range (constructing an age-hypertension interaction) and adjusting it for the same variables as in the previous case. Table 3 shows the results of the best model. Adjusted for sex, smoking, diabetes and the socioeconomic level of the area, hypertension increases the risk of mortality up to 54 years but stops doing so when diagnosed after this age. Table 3Cox model for the total mortality event, hypertensive vs. non-hypertensive subjectsVariableHRHR CI $95\%$ p valueHypertension in each age groupa 18 to 44 years2.4651.904–3.192 < 0.001 45 to 54 years1.1411.017–1.2810.025 55 to 64 years0.8640.803- 0.932 < 0.001 65 to 74 years0.7940.742–0.850 < 0.001 75 to 84 years0.7380.696- 0.782 < 0.001 85 and older0.9970.907–1.0950.950Female vs. male0.6180.597–0.640 < 0.001Diabetes mellitus1.2431.179–1.310 < 0.001Baseline smoking1.4581.391–1.528 < 0.001Socioeconomic group0.041 2nd vs. 1st quintile1.0811.007–1.1590.030 3rd vs. 1st quintile1.0180.949–1.0420.618 4th vs. 1st quintile0.9940.933–1.0600.872 5th vs. 1st quintile1.0550.988–1.1280.108Characteristics of the modelAkaike Information Criteria (AIC): 357,599.3Bayes Information Criteria (BIC): 357,778.8 aThe HR (hazard ratio) expresses the risk of an event calculated from the linear combination of the hypertension coefficients and their interaction with age ## Study of mortality from different causes as competitive risks A multistate model was developed in which the probability of mortality due to kidney or cardiovascular causes or other causes in the two cohorts was studied. Table 4 shows the results of the association of hypertension with mortality in the different age groups. The adjustment variables were the same: sex, smoking, diabetes and socioeconomic status of the area. Table 4Proportional risks of death, hypertensive vs. non-hypertensive patients by age groups and causes (use of multistate models, Figs. 1a and b). Models adjusted for age, sex, diabetes mellitus, smoking and socioeconomic status of the area Mortality due to kidney or cardiovascular causesAge groupHRIC $95\%$ HR p value 18 to 44 years9.3093.819–22.694 < 0.001 45 to 54 years1.7871.280–2.4930.001 55 to 64 years1.0450.855–1.2780.665 65 to 74 years0.9590.828–1.1110.578 75 to 84 years0.9000.811–0.9990.049 85 and older1.1320.976–1.3130.102Characteristics of the modelAkaike Information Criteria (AIC): 75,448.71Bayes Information Criteria (BIC): 75,628.22Mortality due to other causesAge groupHRIC $95\%$ HR p value 18 to 44 years2.0441.557–2.683 < 0.001 45 to 54 years1.0820.959–1.2220.200 55 to 64 years0.8700.801–0.945 < 0.001 65 to 74 years0.8140.757–0.875 < 0.001 75 to 84 years0.7500.701–0.802 < 0.001 85 and older0.9870.881–1.1060.818Characteristics of the modelAkaike Information Criteria (AIC): 280,556.7Bayes Information Criteria (BIC): 280,736.2Total mortalityAge groupHRIC $95\%$ HR p value 18 to 44 years2.4661.905–3.193 < 0.001 45 to 54 years1.1541.028–1.2950.016 55 to 64 years0.8910.827–0.9600.003 65 to 74 years0.8370.783–0.895 < 0.001 75 to 84 years0.7850.740–0.833 < 0.001 85 and older1.0310.938–1.1340.522Characteristics of the modelAkaike Information Criteria (AIC): 356,480.7Bayes Information Criteria (BIC): 356,660.2 *Hypertension is* strongly associated with kidney or cardiovascular mortality when diagnosed in individuals less than 45 years of age, and this association remains but with lower intensity until 54 years of age. After this age, there is no association between the two. When other causes of mortality are evaluated, hypertension is associated with an increase in mortality only if it is diagnosed before age 45 and presents an inverse association after the age of 55. ## Discussion The diagnosis of HTN in patients without previous kidney or CV disease is associated with an increase in the occurrence of K/CVEs (including death due to these causes) throughout the entire life course, but especially when HTN is diagnosed before the age of 55. The diagnosis of hypertension is only associated with kidney or cardiovascular mortality or total mortality, when it occurs before 55 years. An inverse association has been observed between HTN diagnosis over 55 years and overall mortality. The described associations were found in patients without previous kidney or CV disease who were followed and treated in a health system with full access to the general population and adjusted for the effect of DM, smoking, and socioeconomic situation. HTN is associated with an increase in CVEs [2, 22, 31] and an association has also been described between HTN and all-cause mortality [14]. In some studies, CVEs’ incidence was twice if HTN was diagnosed under the age of 45 and an the excess of risk was about $60\%$ for patients diagnosed between 45 and 55 years [19]. The results presented indicate higher risks (HR 4.36 and 2.45 for each of these age ranges) but include kidney or CV death as an event. But association with mortality from all causes has only been found when HTN is diagnosed at the earliest ages of life. The differences are more subtle when we compare the results with studies that analyze newly diagnosed HTN by age strata. While the diagnosis of HTN has been associated with an excess risk of mortality from all causes of 2.5 times (HR 2.59) when it occurs before age 45, this excess mortality does not reach $30\%$ (HR 1.29) when it occurs over age 65 [19]. In the age group under 45 years, our results are very similar for this association (HR 2.47). The decrease in risk with the latest diagnosis of HTN is consistent with what has been previously described. An increase in the probability of developing target organ injury has been reported in patients with HTN diagnosed before 35 years, which was not observed when the diagnosis was made over age 45 [32]. Some studies have reported an association between well-controlled HTN and all-cause mortality in patients younger than 70 years [33], but other ones reported no association observed with all-cause mortality in patients with HTN under treatment, older than 75 years [7]. The differences found when assessing the risk of total mortality may be due to several reasons. Some of the primary studies that mentioned cohorts were recruited more than two decades ago and those that have more recent recruitments find more uncertain results for the association of HTN with all-cause mortality [14]. Improved survival in HTN patients has been demonstrated over time, and when blood pressure levels are better controlled with antihypertensive medication [7, 9]. In one of the studies with the longest reported follow-up (median 19.1 years), although a strong association between HTN and the occurrence of CVEs and all-cause mortality for untreated or poorly controlled patients was established, no such association was found in treated and controlled patients [22]. The inverse association found between HTN and total mortality over 55 years should not be explained from a causal perspective, as it is not plausible. It has been reported that a more intense use of PC services was associated with lower mortality in hypertensive patients [34], and certain promotional interventions have been shown to decrease CV risk in elderly hypertensive patients in PC [35]. Additionally a strong association has been reported between higher continuity of care and reduced mortality rate among hypertensive patients [36]. Our health system has practically universal coverage. Chronic disease is the main explanatory factor for the use of family doctor visits [37], and HTN occupies a relevant consultation time in PC [38]. The health care to which hypertensive patients older than 55 years are subjected, as well as the action on other coexisting CVRFs, may contribute to explaining, at least in part, the association. Regarding the confounding role of the variables studied, both the occurrence of kidney or CV events and mortality from all causes increased in diabetic patients and smokers. The role of these CVRFs in mortality is well known, and both factors are used as adjustment variables in most of the studies discussed [14, 31, 33, 39]. In women with HTN, the risk is lower. It has already been described that avoidable mortality is lower in women than in hypertensive men [40], and it has been estimated that the burden of disease for hypertension is lower in women than in men for all ages except over 75 years [2]. The association of a worse socioeconomic situation and the events associated with HTN has also been widely described [41, 42]. This study has limitations inherent to retrospective cohort studies. The strength of the data is determined by the quality of the information collected. Some of the classical CVRFs, such as hypercholesterolemia or obesity, or other clinical circumstances such as the time at which diabetes was diagnosed, were not included as adjustment variables because they had not been validated in the EHR or their collection could have been differential for the groups compared. The diagnoses recorded in the EHR of PC for HTN and DM have been previously validated [43]. Among the strengths, all cases diagnosed in PC in the Autonomous Community during a period of two years were included and the secondary data sources allowed us to identify the final state in a reliable way, with very limited losses to follow-up. Given the characteristics of the health system (in 2020, $86\%$ of assigned people were visited in their PCC) the generalizability of the results is important. The value of the findings presented is based on the fact that they are data in real clinical practice conditions, in a specific environment, in which paradoxically, although healthy lifestyle habits are not very prevalent, cardiovascular mortality remains comparatively low [44]. ## Conclusion The diagnosis of hypertension in the community environment, in patients without evidence of previous kidney or CV disease, is associated with a large increase in the risk of K/CVEs. This increased risk depends on the individual’s age at diagnosis; risk is highest if diagnosis is made before the age of 55 years and decreases with age. This diagnosis is only associated with an increase in kidney or cardiovascular or overall mortality when it occurs before age 55. Thus, the health system should increase secondary prevention measures, especially in hypertensive patients diagnosed before the age of 55, when excess risk is most evident. ## Supplementary Information Additional file 1: Generation of Cohorts. ## References 1. Zhou B, Carrillo-Larco RM, Danaei G, Riley LM, Paciorek CJ, Stevens GA. **Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants**. *Lancet* (2021.0) **398** 957-980. DOI: 10.1016/S0140-6736(21)01330-1 2. Forouzanfar MH, Liu P, Roth GA, Ng M, Biryukov S, Marczak L. **Global burden of hypertension and systolic blood pressure of at least 110 to 115mmHg, 1990–2015**. *JAMA* (2017.0) **317** 165-182. DOI: 10.1001/jama.2016.19043 3. Shihab HM, Meoni LA, Chu AY, Wang N-Y, Ford DE, Liang K-Y. **Body Mass Index and Risk of Incident Hypertension Over the Life Course: The Johns Hopkins Precursors Study**. *Circulation* (2012.0) **126** 2983-2989. DOI: 10.1161/CIRCULATIONAHA.112.117333 4. O’Brien E. **The Lancet Commission on hypertension: Addressing the global burden of raised blood pressure on current and future generations**. *J Clin Hypertens* (2017.0) **19** 564-568. DOI: 10.1111/jch.12998 5. Shah NS, Molsberry R, Rana JS, Sidney S, Capewell S, O’Flaherty M. **Heterogeneous trends in burden of heart disease mortality by subtypes in the United States, 1999–2018: observational analysis of vital statistics**. *BMJ* (2020.0) **13** m2688. DOI: 10.1136/bmj.m2688 6. Law MR, Morris JK, Wald NJ. **Use of blood pressure lowering drugs in the prevention of cardiovascular disease: Meta-analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies**. *BMJ* (2009.0) **338** 1245 7. Banach M, Bromfield S, Howard G, Howard VJ, Zanchetti A, Aronow WS. **Association of systolic blood pressure levels with cardiovascular events and all-cause mortality among older adults taking antihypertensive medication**. *Int J Cardiol* (2014.0) **176** 219-226. DOI: 10.1016/j.ijcard.2014.07.067 8. Bundy J, Li C, Stuclik P, Bu X, Kellly T, Mills KT. **Systolic Blood Pressure Reduction and Risk of Cardiovascular Disease and Mortality A Systematic Review and Network Meta-analysis**. *JAMA Cardiol* (2017.0) **2** 775-781. DOI: 10.1001/jamacardio.2017.1421 9. Yun D, Lee H, Choi W, Chang H, Son D, Lee J. **Association of optimal blood pressure with mortality in patients taking antihypertensive medications**. *J Clin Hypertens* (2020.0) **22** 2035-2043. DOI: 10.1111/jch.14034 10. Wei J, Galaviz KI, Kowalski AJ, Magee MJ, Haw JS, Narayan KMV. **Comparison of cardiovascular events among users of different classes of antihypertension medications**. *JAMA Netw Open* (2020.0) **3** e1921618. DOI: 10.1001/jamanetworkopen.2019.21618 11. Almgren T, Persson B, Wilhelmsen L, Rosengren A, Andersson OK. **Stroke and coronary heart disease in treated hypertension - A prospective cohort study over three decades**. *J Intern Med* (2005.0) **257** 496-502. DOI: 10.1111/j.1365-2796.2005.01497.x 12. Sepanlou SG, Sharafkhah M, Poustchi H, Malekzadeh MM, Etemadi A, Khademi H. **Hypertension and mortality in the Golestan Cohort Study: A prospective study of 50000 adults in Iran**. *J Hum Hypertens* (2016.0) **30** 260-7. DOI: 10.1038/jhh.2015.57 13. da Silva TLN, Klein CH, da Rocha Nogueira A, Salis LHA, de Souza E Silva NA, Bloch KV. **Cardiovascular mortality among a cohort of hypertensive and normotensives in Rio de Janeiro - Brazil - 1991-2009**. *BMC Public Health* (2015.0) **15** 623. DOI: 10.1186/s12889-015-1999-4 14. Lawlor DA, Kim L, Morris R, Amuzu A, Whincup P, Ebrahim S. **Survival with treated and well-controlled blood pressure: Findings from a prospective cohort study**. *PLoS One.* (2011.0) **6** e17792. DOI: 10.1371/journal.pone.0017792 15. Ford ES. **Trends in mortality from all causes and cardiovascular disease among hypertensive and nonhypertensive adults in the united states**. *Circulation* (2011.0) **123** 1737-1744. DOI: 10.1161/CIRCULATIONAHA.110.005645 16. Makridakis S, DiNicolantonio JJ. **Hypertension: empirical evidence and implications in 2014**. *Open Hear* (2014.0) **1** e000048. DOI: 10.1136/openhrt-2014-000048 17. Kringeland E, Tell GS, Midtbø H, Haugsgjerd TR, Igland J, Gerdts E. **Factors associated with increase in blood pressure and incident hypertension in early midlife: the Hordaland Health Study**. *Blood Press* (2020.0) **29** 267-275. DOI: 10.1080/08037051.2020.1762070 18. Kansui Y, Matsumura K, Morinaga Y, Inoue M, Kiyohara K, Ohta Y. **C-reactive protein and incident hypertension in a worksite population of Japanese men**. *J Clin Hypertens* (2019.0) **21** 524-532. DOI: 10.1111/jch.13510 19. Wang C, Yuan Y, Zheng M, Pan A, Wang M, Zhao M. **Association of Age of Onset of Hypertension With Cardiovascular Diseases and Mortality**. *J Am Coll Cardiol* (2020.0) **75** 2921-2930. DOI: 10.1016/j.jacc.2020.04.038 20. Lip S, Tan LE, Jeemon P, McCallum L, Dominiczak AF, Padmanabhan S. **Diastolic Blood Pressure J-Curve Phenomenon in a Tertiary-Care Hypertension Clinic**. *Hypertension* (2019.0) **74** 767-775. DOI: 10.1161/HYPERTENSIONAHA.119.12787 21. Bae E, Rocco MV, Lee J, Park JY, Kim YC, Yoo KD. **Impact of DBP on all-cause and cardiovascular mortality: results from the National Health and Nutrition Examination survey, 1999–2014**. *J Hypertens* (2022.0) **40** 108-116. DOI: 10.1097/HJH.0000000000002983 22. Zhou D, Xi B, Zhao M, Wang L, Veeranki SP. **Uncontrolled hypertension increases risk of all-cause and cardiovascular disease mortality in US adults: The NHANES III Linked Mortality Study**. *Sci Rep* (2018.0) **8** 1-7. PMID: 29311619 23. Domínguez-Berjón MF, Borrell C, Cano-Serral G, Esnaola S, Nolasco A, Pasarín MI. **Construcción de un índice de privación a partir de datos censales en grandes ciudades españolas (Proyecto MEDEA)**. *Gac Sanit* (2008.0) **22** 179-187. DOI: 10.1157/13123961 24. 24.World Health Organization (WHO)International Classification of Primary Care, -ICPC-22009World Health Organization. *International Classification of Primary Care, -ICPC-2* (2009.0) 25. 25.World Health Organization (WHO)ICD-10: international statistical classification of diseases and related health problems: tenth revision2004World Health Organization. *ICD-10: international statistical classification of diseases and related health problems: tenth revision* (2004.0) 26. Putter H, Fiocco M, Geskus RB. **Tutorial in biostatistics: competing risks and multi-state models**. *Stat Med* (2007.0) **26** 2389-2430. DOI: 10.1002/sim.2712 27. Williams RL. **A Note on Robust Variance Estimation for Cluster-Correlated Data**. *Biometrics* (2000.0) **56** 645-646. DOI: 10.1111/j.0006-341X.2000.00645.x 28. 28.Barnett A, Graves N. Competing risks models and time-dependent covariates. Crit Care. 2008;12(2):134. 29. 29.Fabozzi FJ, Focardi SM, Rachev ST, Arshanapalli BG. Appendix E: Model Selection Criterion: AIC and BIC. In: The Basics of Financial Econometrics. Hoboken: Wiley; 2014. p. 399–403. 30. 30.Crowther MJ, Lambert P. MULTISTATE: Stata module to perform multi-state survival analysis. Statistical Software Components S458207, Boston College Department of Economics. 2016. https://ideas.repec.org/c/boc/bocode/s458207.html. Revised 18 Jan 2023. 31. Rapsomaniki E, Timmis A, George J, Pujades-Rodriguez M, Shah AD, Denaxas S. **Blood pressure and incidence of twelve cardiovascular diseases: lifetime risks, healthy life-years lost, and age-specific associations in 1·25 million people**. *Lancet* (2014.0) **383** 1899-1911. DOI: 10.1016/S0140-6736(14)60685-1 32. Suvila K, McCabe EL, Lehtonen A, Ebinger JE, Lima JAC, Cheng S. **Early Onset Hypertension Is Associated With Hypertensive End-Organ Damage Already by MidLife**. *Hypertension* (2019.0) **74** 305-312. DOI: 10.1161/HYPERTENSIONAHA.119.13069 33. Park S, Han K, Lee S, Kim Y, Lee Y, Kang MW. **Cardiovascular or mortality risk of controlled hypertension and importance of physical activity**. *Heart* (2021.0) **107** 1472-1479. DOI: 10.1136/heartjnl-2020-318193 34. Clement FM, Chen G, Khan N, Tu K, Campbell NRC, Smith M. **Primary care physician visits by patients with incident hypertension**. *Can J Cardiol* (2014.0) **30** 653-660. DOI: 10.1016/j.cjca.2014.03.033 35. Arija V, Villalobos F, Pedret R, Vinuesa A, Jovani D, Pascual G. **Physical activity, cardiovascular health, quality of life and blood pressure control in hypertensive subjects: randomized clinical trial**. *Health Qual Life Outcomes* (2018.0) **16** 184. DOI: 10.1186/s12955-018-1008-6 36. Chan K-S, Wan EY-F, Chin W-Y, Cheng WH-G, Ho MK, Yu EY-T. **Effects of continuity of care on health outcomes among patients with diabetes mellitus and or hypertension: a systematic review**. *BMC Fam Pract* (2021.0) **22** 145. DOI: 10.1186/s12875-021-01493-x 37. Martín-Fernández J, Gómez-Gascón T, Del Cura- González M.I., Tomás-García N, Vargas-Machuca C, Rodríguez-Martínez G. **Quality of life related to health as a factor explaining the use of family medical consultation: a study on the behavioral model**. *Rev Esp Salud Publica* (2010.0) **84** 293-319. PMID: 20661529 38. Levine DM, Dixon RF, Linder JA. **Association of Structured Virtual Visits for Hypertension Follow-Up in Primary Care with Blood Pressure Control and Use of Clinical Services**. *J Gen Intern Med* (2018.0) **33** 1862-1867. DOI: 10.1007/s11606-018-4375-0 39. Franco OH, Peeters A, Bonneux L, de Laet C. **Blood Pressure in Adulthood and Life Expectancy With Cardiovascular Disease in Men and Women: Life Course Analysis**. *Hypertension* (2005.0) **46** 280-286. DOI: 10.1161/01.HYP.0000173433.67426.9b 40. Redon J, Tellez-Plaza M, Orozco-Beltran D, Gil-Guillen V, Pita Fernandez S, Navarro-Pérez J. **Impact of hypertension on mortality and cardiovascular disease burden in patients with cardiovascular risk factors from a general practice setting**. *J Hypertens* (2016.0) **34** 1075-1083. DOI: 10.1097/HJH.0000000000000930 41. King JB, Pinheiro LC, Ringel JB, Bress AP, Shimbo D, Muntner P. **Multiple social vulnerabilities to health disparities and hypertension and death in the regards study**. *Hypertension* (2022.0) **79** 196-206. DOI: 10.1161/HYPERTENSIONAHA.120.15196 42. Lawlor DA, Ebrahim S, Smith GD. **Adverse socioeconomic position across the lifecourse increases coronary heart disease risk cumulatively: Findings from the British women’s heart and health study**. *J Epidemiol Community Health* (2005.0) **59** 785-793. DOI: 10.1136/jech.2004.029991 43. de Burgos-Lunar C, Salinero-Fort MA, Cárdenas-Valladolid J, Soto-Díaz S, Fuentes-Rodríguez CY, Abánades-Herranz JC. **Validation of diabetes mellitus and hypertension diagnosis in computerized medical records in primary health care**. *BMC Med Res Methodol* (2011.0) **11** 146. DOI: 10.1186/1471-2288-11-146 44. Graciani A, León-Muñoz LM, Guallar-Castillón P, Rodríguez-Artalejo F, Banegas JR. **Cardiovascular Health in a Southern Mediterranean European Country**. *Circ Cardiovasc Qual Outcomes* (2013.0) **6** 90-98. DOI: 10.1161/CIRCOUTCOMES.112.967893
--- title: Role of age in dynamics of autoantibodies in pediatric Celiac disease authors: - Chiara Maria Trovato - Monica Montuori - Beatrice Leter - Ilaria Laudadio - Giusy Russo - Salvatore Oliva journal: Italian Journal of Pediatrics year: 2023 pmcid: PMC10037870 doi: 10.1186/s13052-023-01435-6 license: CC BY 4.0 --- # Role of age in dynamics of autoantibodies in pediatric Celiac disease ## Abstract ### Background Celiac disease (CD) is characterized by elevated serum titers of autoantibodies IgA anti-tissue transglutaminase 2 (TGA-IgA) and IgA anti-endomysial (EMA), with small bowel mucosa atrophy. We evaluated age differences between CD children exhibiting variable antibody titers at diagnosis. ### Methods CD children diagnosed between January 2014 and June 2019, according to 2012 ESPGHAN guidelines were studied. All had EMA and TGA-IgA measurements, while a proportion of them underwent esophagogastroduodenoscopy (EGD). Patients were grouped based on serum TGA-IgA titers normalized to the upper limit of normal (ULN) and differences in median age (years) assessed by analysis of variance (ANOVA) and creation of orthogonal contrasts. ### Results CD was diagnosed in 295 subjects (median age: 4.4 [IQR: 2.60–8.52]) with a biopsy sparing protocol (high titer: ≥ 10xULN) and in 204 by EGD biopsy. Of the latter, 142 (median age: 8.5 [IQR: 5.81–11.06]) and 62 (median age: 9.5 [IQR: 6.26–12.76]) had a low (< 5xULN) and a moderate (≥ 5 < 10xULN) TGA-IgA titer, respectively. Potential CD was diagnosed in 20 patients (median age: 3.6 [IQR: 2.47–6.91]). The median age was significantly lower in the no-biopsy group (ANOVA: F[3, 516] = 25.98, $p \leq .001$) than in low- and moderate titer groups ($p \leq 0.0001$), while there was no statistical difference between biopsy-sparing and potential CD groups. ### Conclusion CD patients with greatly elevated antibody titers (≥ 10xULN) were diagnosed at an earlier age than those with lower titers. This may indicate that an increase in TGA-IgA is independent of age and suggests a polarization of autoimmunity in younger individuals with higher serum antibody levels. ## Introduction Celiac disease (CD) is one of the most common autoimmune diseases worldwide, caused by gluten in genetically susceptible individuals, and presents with a variety of signs and symptoms. Like all autoimmune diseases, it can affect people of all ages, including the elderly [1, 2]. The prevalence of CD has increased dramatically in recent decades, mainly in pediatric age [3]. Characteristically, CD patients have highly specific serum autoantibodies against the major CD autoantigen, tissue transglutaminase 2 (TG2) and against circulating deamidated gliadin peptides (DGP), as well as varying degrees of small intestinal mucosa atrophy [4]. Several reports highlight the positive association between serum levels of IgA anti tissue transglutaminase 2 (TGA-IgA) and the degree of mucosal villous atrophy [5, 6]. Therefore, the latest guidelines from the European Society of Pediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN), allow the diagnosis of CD without biopsies in children with serum TGA-IgA 10 times or more the upper limit of normal (≥ 10xULN), confirmed by detection of IgA endomysial antibodies (EMA-IgA) in a second blood sample [7, 8]. In addition, these guidelines also recommend combining total IgA and TGA-IgA as an initial test, regardless of age. Different types of autoantibodies can be used in CD diagnosis, but serum TGA-IgA are the most suitable for screening purposes because of their high sensitivity and specificity. Anti-transglutaminase seroconversion can stably occur around 21 months of age [9]: before this time, an additional autoantibody, such as DGP-IgG, is usually suggested for diagnosis [10]. In patients with non-elevated titers, duodenal-jejunal biopsies are critical for CD diagnosis. There is still no clear evidence regarding the occurrence of seroconversion. To date, a few studies [11, 12] have examined age differences at diagnosis in CD children with different antibody titers. Therefore, in this study we aimed to characterize age differences between groups of CD children with different serum antibody titers. ## Study population We retrospectively enrolled all patients referred to the Pediatric Gastroenterology and Hepatology Unit at the Sapienza University Hospital Umberto I in Rome with suspected CD, between January 2014 and September 2019. Exclusion criteria included IgA deficiency and other chronic intestinal disorders such as food allergies, inflammatory bowel disease, infectious and immunological diseases, systemic disorders affecting the gut. All subjects were diagnosed according to the ESPGHAN criteria published in 2012 [7]. Patients with serum TGA-IgA < 10xULN underwent esophagogastroduodenoscopy (EGD) with multiple duodenum-jejunal biopsies, under general anesthesia or deep sedation [13]. Histological lesions were graded according to criteria of Marsh–Oberhuber (MO) [14]. TGA-IgA antibody titers were tested using commercially available ELISA kits from Eurospital (Trieste, Italy; cutoff value > 9 UA/mL). The study population was divided into groups for analysis and comparison according to the need for biopsies, mean antibody titers, and the presence or absence of duodenal mucosa atrophy, as suggested in a previously published cohort [15].Biopsy-sparing group: patients with TGA-IgA ≥ 10 ULN, serum EMA IgA positivity, genetic predisposition (presence of HLA DQ2/HLA DQ8) and presence of symptoms and signs associated with CD (diarrhea, weight loss, failure to thrive, anorexia, abdominal distention, abdominal pain, short stature, flatulence, irritability, elevated titers of liver enzymes, constipation, and anemia) diagnosed without EGD and biopsy, per ESPGAHN 2012 guidelines. Patients requiring EGD with biopsy: a) symptomatic or asymptomatic children with serum TGA-IgA ≥ 5 < 10xULN, and mucosal lesions consistent with CD at biopsy were classified as moderate titer CD (at least two prior measurements of serum TGA-IgA were consistently in this range); b) symptomatic or asymptomatic children with serum levels of TGA-IgA < 5xULN and mucosal damage compatible with CD at biopsy (at least two prior measurements of TGA-IgA were consistently in this range), defined as low titer CD.Potential CD: symptomatic children with positive autoantibodies (TGA-IgA and at least one EMA detection positive), and with normal mucosa biopsies at various sites along the duodenum. ## Statistical analysis Statistical analysis was performed using the R statistical software (R Core Team, 2019, version 3.6.1), using the tidy verse package [16] for data cleaning and presentation. Differences between the median age of the groups were examined by analysis of variance (ANOVA), through the creation of orthogonal contrasts, with a significance of 0.05 for all statistical tests. ## Results Of 1243 children referred to our Unit with suspected CD during the study period, 519 patients (median age: 6.31 years [IQR: 3.37–10.06]; 335 females) were retrospectively enrolled. Population data are summarized in Fig. 1. Of these, 295 were included in the biopsy-sparing group (median age: 4.4 [IQR: 2.60–8.52]; 196 females), 62 in the moderate titer group (median age: 9.5 [IQR: 6.26–12.76]; 38 females), 142 in the low-titer group (median age: 8.5 [IQR: 5.81–11.06]; 86 females) and finally 20 in the potential CD group (median age: 3.6 [IQR: 2.47–6.91]; 13 females). The characteristics of the patient population are summarized in the Table 1.Fig. 1Patients investigated grouped according to the diagnostic protocol (biopsy sparing and endoscopic biopsy). For each group are reported median age (years) and interquartile rangesTable 1Characteristics of the Study ParticipantsGroups of patientsBiopsy-sparingLow titer CDModerate titer CDPotential CDNumber of patients2951426220Males9956247Females196863813Age, median [IQR]4.4 [2.60–8.52]8.5 [5.81–11.06]9.5 [6.26–12.76]3.6 [2.47–6.91]Median TGA-IgA (ULN) > 102.29.031.81EMA positive (n. of patients)2951136015EMA negative (n. of patients)02925Legend: CD Celiac disease, TGA-IgA autoantibodies IgA anti-tissue transglutaminase 2, IQR interquartile ranges, ULN upper limit of normal, EMA IgA anti-endomysial Figure 2 shows a graphical representation of the median ages of the different groups. The age of patients in the biopsy-sparing group was significantly lower (ANOVA: F[3, 516] = 25.98, $p \leq 0.001$) than in the low ($p \leq 0.0001$) and moderate ($p \leq 0.0001$) titer groups. No statistically significant difference was observed between the median ages of the biopsy-sparing group and the potential CD patients ($$p \leq 0.2825$$); moreover, no difference was found between low and moderate titer CD groups (8.5 [IQR: 5.81–11.06] vs 9.5 [IQR: 6.26–12.76], respectively; $$p \leq 0.1665$$).Fig. 2Graphic representation of median ages (years) and interquartile ranges for different groups of patients As shown in the Table 1, it is worth noting that 29 and 2 of the low and moderate titers, respectively, received a diagnosis of CD despite negative EMA-IgA serum levels. Of the PCD group, 5 children were EMA negative when they underwent EGD, but all 20 PCD children had at least one EMA positive test. ## Discussion The present study, carried out on several groups of pediatric patients with CD, was aimed to analyze the age differences at diagnosis in relation to different specific autoantibody titers. We found that children diagnosed using the “biopsy-sparing protocol”, with the highest serum TGA-IgA titers (≥ 10xULN), were significantly younger than the other CD groups. To date there is still little data focusing on the association between the age at presentation of children with CD and titers of serum specific autoantibodies. Our study confirms previous results showing that CD children tend to have high TGA-IgA titers, particularly at a younger age [11]. The dynamics of autoantibody development are still unclear, and it is unclear why the levels of autoantibody markers are so different in the pediatric CD population. In 2005, Salmi et al. [ 17] found lower titers of TGA-IgA autoantibodies in EMA-negative adult CD patients: the authors suggested a possible entrapment of autoantibodies in the intestinal mucosa which would prevent them from entering the blood, due to a higher tissue avidity of autoantibodies in a long-standing disease. Interestingly, our data on higher titers of TGA-IgA autoantibodies in the youngest children appear to parallel the study by Marine et al. [ 3] showing for the first time that children, mainly the youngest, have a higher CD prevalence compared to adults. We are tempted to speculate that the difference in serum autoantibody titers might be due to a polarization of the immune system towards overproduction of TGA-IgA autoantibodies, in parallel with a concomitant reduced immunological tolerance. As is known, upon recognition of foreign antigenic peptides presented by MHC native T cells are activated and clonally expanded. Based on a recent study by Yao et al. [ 18], we believe that different types of lymphocytes could be activated by differential clonal expansion as well as autoantibody production. In the same way, cytokine expression could modulate the production of autoantibodies since they can selectively affect intraepithelial cytotoxic T cells [19]. However, this specific mechanism needs to be studied in detail before this hypothesis can be confirmed. The high prevalence of seronegative CD in adults [15, 20] supports our speculation. The relationship between age and autoantibodies at onset in other autoimmune diseases has been studied. In fact, as reported in type 1 diabetes [21], children with an early age at onset of the disease have higher levels of autoantibodies and more autoimmune diseases. Interestingly, among children at high genetic risk for type 1 diabetes, those with late onset islet autoimmunity tend to develop diabetes in adolescence or early adulthood [21, 22]; in addition, there is widespread agreement that in various autoimmune diseases (e.g. systemic lupus erythematosus, type 1 diabetes mellitus) an early age at onset can act as a negative prognostic factor for the course of the disease [23]. Interestingly, data from studies of our group of CD patients suggest that age at diagnosis is a strong predictor for the occurrence of organ-specific autoantibodies and the development of additional autoimmune diseases [24]. We cannot determine whether the dynamics of other autoimmune diseases would differ between CD children who underwent EGD and those with the highest levels of serum autoantibodies diagnosed by a biopsy-sparing protocol. However, this could be an important research topic to fully understand autoantibody’ “autonomy” with respect to other autoimmune diseases. Interestingly, we observed no statistical difference in patient age ($$p \leq 0.1665$$) between the low and the moderate titer groups: this may indicate that autoantibodies in a large proportion of subjects do not continuously increase over time. This could also bring further focus to clinical scenarios with positive low-to-moderate serum TGA-IgA levels, for which a clear and linear diagnostic work-up is not yet defined by guidelines [25]. Notably, potential CD children were significantly younger than those with biopsy-proven histological defects. This finding can be explained by the “progression of mucosal damage” [26]: indeed, due to the patchy damage to the small bowel mucosa in CD [27], it is conceivable that the damaged areas are not initially identified in some children. However, as the number of “damaged areas” increases over time, these patients can later be identified as “overt CD patients”, despite non-elevated autoantibody titers. Undoubtedly, age is one of the variables to consider when assessing the risk of progression from potential to overt CD [28]: since potential CD patients appear to be younger, a rigorous follow-up is remarkably important to intercept the potential transtition to overt CD [29]. It is worth noting that the median age was not statistically different between the biopsy-sparing and the potential CD groups, suggesting that, in autoimmunity, the peak of autoantibodies would resemble a sudden, uncontrolled storm manifesting unexpectedly and with an unpredictable autoantibody titer. ## Conclusions In summary, although our study has some limitations such as the lack of an adult control group and the small sample size enrolled by an academic tertiary care center, two groups of CD children were identified: one is characterized by individuals who quickly reach elevated autoantibody titers (with levels ≥ 10 times the upper limit of normal); the second includes subjects who produce less circulating autoantibodies and never reach high titers. The relevance of these results remains an open question for pediatric gastroenterologists. Further studies will focus on deciphering the immunological signaling pathways beyond these observations as well as understanding the dynamics of specific autoantibodies in CD children with different serum titers. In addition, the inclusion of an adult CD population would be worthwhile. ## References 1. Catassi C. **Natural history of celiac disease autoimmunity in a USA cohort followed since 1974**. *Ann Med* (2010) **42** 530-538. DOI: 10.3109/07853890.2010.514285 2. Trovato CM. **Diagnostic Value of Persistently Low Positive TGA-IgA Titers in Symptomatic Children With Suspected Celiac Disease**. *J Pediatr Gastroenterol Nutr* (2021) **72** 712-717. DOI: 10.1097/MPG.0000000000003047 3. Marine M. **The prevalence of coeliac disease is significantly higher in children compared with adults**. *Aliment Pharmacol Ther* (2011) **33** 477-486. DOI: 10.1111/j.1365-2036.2010.04543.x 4. Ludvigsson JF. **The Oslo definitions for coeliac disease and related terms**. *Gut* (2013) **62** 43-52. DOI: 10.1136/gutjnl-2011-301346 5. Donaldson MR. **Correlation of duodenal histology with tissue transglutaminase and endomysial antibody levels in pediatric celiac disease**. *Clin Gastroenterol Hepatol* (2007) **5** 567-573. DOI: 10.1016/j.cgh.2007.01.003 6. Singh P. **Titers of anti-tissue transglutaminase antibody correlate well with severity of villous abnormalities in celiac disease**. *J Clin Gastroenterol* (2015) **49** 212-217. DOI: 10.1097/MCG.0000000000000105 7. Husby S. **European Society for Pediatric Gastroenterology, Hepatology, and Nutrition guidelines for the diagnosis of coeliac disease**. *J Pediatr Gastroenterol Nutr* (2012) **54** 136-160. DOI: 10.1097/MPG.0b013e31821a23d0 8. Husby S. **European Society Paediatric Gastroenterology, Hepatology and Nutrition Guidelines for Diagnosing Coeliac Disease 2020**. *J Pediatr Gastroenterol Nutr* (2020) **70** 141-156. DOI: 10.1097/MPG.0000000000002497 9. Assandri R, Montanelli A. **Diagnosis of gluten-related enteropathy in a newborn: how and when?**. *Gastroenterol Hepatol Bed Bench* (2019) **12** 278-286. PMID: 31749915 10. Valitutti F. **Early Antibody Dynamics in a Prospective Cohort of Children At Risk of Celiac Disease**. *Am J Gastroenterol.* (2023) **118** 574-7. DOI: 10.14309/ajg.0000000000002192 11. Tiberti C. **Celiac disease-associated transglutaminase autoantibody target domains at diagnosis are age and sex dependent**. *Clin Immunol* (2003) **109** 318-324. DOI: 10.1016/j.clim.2003.08.010 12. Vivas S. **Age-related clinical, serological, and histopathological features of celiac disease**. *Am J Gastroenterol* (2008) **103** 2360-5. DOI: 10.1111/j.1572-0241.2008.01977.x 13. Corazza GR, Villanacci V. **Coeliac disease**. *J Clin Pathol* (2005) **58** 573-574. DOI: 10.1136/jcp.2004.023978 14. Oberhuber G, Granditsch G, Vogelsang H. **The histopathology of coeliac disease: time for a standardized report scheme for pathologists**. *Eur J Gastroenterol Hepatol* (1999) **11** 1185-1194. DOI: 10.1097/00042737-199910000-00019 15. Abrams JA. **Seronegative celiac disease: increased prevalence with lesser degrees of villous atrophy**. *Dig Dis Sci* (2004) **49** 546-550. DOI: 10.1023/B:DDAS.0000026296.02308.00 16. Wickham H. **Welcome to the Tidyverse**. *J Open Source Softw* (2019) **4** 1686. DOI: 10.21105/joss.01686 17. Salmi TT. **Endomysial antibody-negative coeliac disease: clinical characteristics and intestinal autoantibody deposits**. *Gut* (2006) **55** 1746-1753. DOI: 10.1136/gut.2005.071514 18. Yao Y. **T cell receptor repertoire as a potential diagnostic marker for celiac disease**. *Clin Immunol* (2021) **222** 108621. DOI: 10.1016/j.clim.2020.108621 19. Ciszewski C. **Identification of a gammac Receptor Antagonist That Prevents Reprogramming of Human Tissue-resident Cytotoxic T Cells by IL15 and IL21**. *Gastroenterology* (2020) **158** 625-637 e13. DOI: 10.1053/j.gastro.2019.10.006 20. Volta U. **Seronegative celiac disease: Shedding light on an obscure clinical entity**. *Dig Liver Dis* (2016) **48** 1018-1022. DOI: 10.1016/j.dld.2016.05.024 21. Kordonouri O. **Age-specific levels of diabetes-related GAD and IA-2 antibodies in healthy children and adults**. *J Pediatr Endocrinol Metab* (2002) **15** 47-52. DOI: 10.1515/JPEM.2002.15.1.47 22. Ozen G. **The association of autoimmune diseases with Type 1 Diabetes Mellitus in Children Depends Also by the Length of Partial Clinical Remission Phase (Honeymoon)**. *Int J Endocrinol* (2020) **2020** 2630827. DOI: 10.1155/2020/2630827 23. Amador-Patarroyo MJ, Rodriguez-Rodriguez A, Montoya-Ortiz G. **How does age at onset influence the outcome of autoimmune diseases?**. *Autoimmune Dis* (2012) **2012** 251730. PMID: 22195277 24. Tiberti C. **Screening for Type 1 Diabetes-, Thyroid-, Gastric-, and Adrenal-Specific Humoral Autoimmunity in 529 Children and Adolescents With Celiac Disease at Diagnosis Identifies as Positive One in Every Nine Patients**. *Diabetes Care* (2017) **40** e10-e11. DOI: 10.2337/dc16-2095 25. Popp A, Maki M. **Changing Pattern of Childhood Celiac Disease Epidemiology: Contributing Factors**. *Front Pediatr* (2019) **7** 357. DOI: 10.3389/fped.2019.00357 26. Jabri B, Sollid LM. **Tissue-mediated control of immunopathology in coeliac disease**. *Nat Rev Immunol* (2009) **9** 858-870. DOI: 10.1038/nri2670 27. Valitutti F. **Mapping histologic patchiness of celiac disease by push enteroscopy**. *Gastrointest Endosc* (2014) **79** 95-100. DOI: 10.1016/j.gie.2013.06.012 28. Auricchio R. **Progression of celiac disease in children with antibodies against tissue transglutaminase and normal duodenal architecture**. *Gastroenterology* (2019) **157** 413-420 e3. DOI: 10.1053/j.gastro.2019.04.004 29. Trovato CM. **The challenge of treatment in potential celiac disease**. *Gastroenterol Res Pract* (2019) **2019** 8974751. DOI: 10.1155/2019/8974751
--- title: 'Magnitude of troponin elevation in patients with biomarker evidence of myocardial injury: relative frequency and outcomes in a cohort study across a large healthcare system' authors: - Colleen K. McIlvennan - Manuel Urra - Laura Helmkamp - John C. Messenger - David Raymer - Karen S. Ream - J. Bradley Oldemeyer - Amrut V. Ambardekar - Kathleen Barnes - Larry A. Allen journal: BMC Cardiovascular Disorders year: 2023 pmcid: PMC10037877 doi: 10.1186/s12872-023-03168-0 license: CC BY 4.0 --- # Magnitude of troponin elevation in patients with biomarker evidence of myocardial injury: relative frequency and outcomes in a cohort study across a large healthcare system ## Abstract ### Background Serum troponin levels correlate with the extent of myocyte necrosis in acute myocardial infarction (AMI) and predict adverse outcomes. However, thresholds of cardiac troponin elevation that could portend to poor outcomes have not been established. ### Methods In this cohort study, we characterized all cardiac troponin elevations > 0.04 ng/mL (upper limit of normal [ULN]) from patients hospitalized with an ICD-$\frac{9}{10}$ diagnosis of AMI across our health system from 2012–2019. We grouped events into exponential categories of peak cardiac troponin and evaluated the association of these troponin categories with all-cause mortality, heart transplants, or durable left ventricular assist devices (LVAD). Patients with cardiac troponin > 10,000 × ULN were manually chart reviewed and described. ### Results There were 18,194 AMI hospitalizations with elevated cardiac troponin. Peak troponin was 1–10 × ULN in $21.1\%$, 10–100 × ULN in $34.8\%$, 100–1,000 × ULN in $30.1\%$, 1,000–10,000 × ULN in $13.1\%$, and > 10,000 × ULN in $0.9\%$ of patients. One-year mortality was 17–$21\%$ across groups, except in > 10,000 × ULN group where it was $33\%$ (adjusted hazard ratio ($99\%$CI) for > 10,000 × ULN group compared to all others: 1.86 (1.21, 2.86)). Hazards of one-year transplant and MCS were also significantly elevated in the > 10,000 × ULN group. ### Conclusions Elevation in cardiac troponin levels post AMI that are > 10,000 × ULN was rare but identified patients at particularly high risk of adverse events. These patients may benefit from clarification of goals of care and early referral for advanced heart failure therapies. These data have implications for conversion to newer high-sensitivity cardiac troponin assays whose maximum assay limit is often lower than traditional assays. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12872-023-03168-0. ## Introduction Cardiac specific troponins, including Troponin T and Troponin I, are regulatory proteins unique to the myocardium. The serologic concentration of these proteins rise and fall in the setting of myocardial injury [1]. The utility of cardiac troponin for the diagnosis of acute myocardial infarction (AMI) and other forms of myocardial injury is well established and now incorporated into the Fourth Universal Definition of Myocardial Infarction [2]. Several studies have shown a correlation between the magnitude of cardiac troponin elevation and the extent of myocyte necrosis, the development of heart failure, and the risk of mortality [3–7]. However, despite the ubiquitous use of troponins in healthcare for more than a decade, detailed analyses of troponin levels and clinical outcomes are surprisingly sparse. As such, thresholds of troponin elevation that could trigger particular clinical action have not been thoroughly explored or well developed. Patients with massive myocardial injury are at high risk for adverse outcomes, including death. However, the clinical course can be variable and difficult to predict. These patients may be eligible for durable left ventricular assist device (LVAD) or cardiac transplantation, but getting patients to such therapies usually requires timely engagement of an advanced heart failure program, sometimes with early initiation of temporary circulatory support. Late referral after massive myocardial infarction when patients have progressed to multi-organ failure can cause patients to “miss the window” for advanced therapies [8]. Given the known relationship of cardiac troponin levels to degree of cardiomyocyte injury and based on anecdotal cases of patients whose cardiac troponin level was greater than assay maximum, we hypothesized that markedly elevated peak cardiac troponin is associated with particularly high rates of adverse outcomes. Better characterizing this relationship of very high cardiac troponin levels to outcomes could improve rapid triage of in-hospital care, guide post-discharge planning, and encourage completion of advance care directives for this high-risk population. In addition, if an absolute threshold of cardiac troponin elevation could be identified beyond which outcomes were worse than that seen with LVAD and transplant, a simple rule of thumb could be created to prompt immediate consideration for engagement of advanced heart failure or palliative care consultation. Therefore, we aimed to systematically characterize all cardiac troponin elevations across a diverse health system and compare absolute cardiac troponin levels to 1-year outcomes (death, mechanical circulatory support (MCS), and heart transplant) in these populations. ## Study population and data sources UCHealth includes 12 hospitals and over 30 clinics within Colorado that cover $19\%$ of the 5.8 million Coloradans, as well as referrals from relatively rural areas in surrounding states. Major efforts were made in 2011 to convert the entire system to a single instance of the Epic electronic health record for all inpatient and outpatient activities. Since 2011, all laboratory test results, as well as diagnoses, vital signs, procedures, encounters, and deaths are reported in the UCHealth Epic record and then downloaded into the Health Data Compass virtual data warehouse (https://www.healthdatacompass.org/). All procedures were approved by the Colorado Institutional Review Board and were conducted in accordance with the principles outlined in the Declaration of Helsinki. Health Data Compass was queried for all hospital admissions to the UCHealth between 2012–2019 for patients with a diagnosis of acute myocardial infarction (AMI) based on International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9) [410, 414] and ICD-10 (I-21, I-22, I-25) codes. Based on this cohort, additional data collected were demographics (age, gender, race/ethnicity), labs obtained during AMI encounters (highest troponin, highest serum creatinine), and primary outcomes in the following year after the hospitalization event (death, MCS, heart transplant by Current Procedural Terminology codes). All hospitals used a standardized “Chest Pain Protocol” pathway, which recommended cardiac troponin testing at presentation, 3 h, and 6 h. During the study period, the cardiac troponin assay most commonly used for clinical diagnosis across hospitals was the ADIVA Centaur cardiac troponin I assay. The reference range for the assay was 0.0–0.04 ng/mL. A minority of troponin values reflected the use of alternative assays, primarily the Ortho Clinical Diagnostics VITROS 5600 (reference range 0.00–0.03 ng/mL). Upper limit of normal (ULN) was determined as above the 99th percentile of the target population. However, troponin data stored in the electronic health record did not include assay details. Within this context, we used a singular cutoff of 0.04 ng/mL for this analysis. This approach ensured patients met criteria for the Fourth Universal definition of myocardial injury (i.e. a serum cardiac troponin test result > 99th percentile of the ULN). This approximation also reflected chest pain-myocardial infarction treatment pathways at all health system sites which used a cutoff of 0.04 ng/mL regardless of assay. We limited our patient selection to age ≥ 18 years. For patients with multiple events, all events and associated cardiac troponin were considered. For the subgroup of patients with cardiac troponin > 10,000 times the ULN (> 400 ng/dL), the clinical course was further explored via manual chart review (by MU, CKM, LAA). Data obtained included MI type, procedures, and types of temporary MCS utilized, as well as confirmation of death, durable MCS, and heart transplant. ## Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting, or dissemination plans of this research. ## Statistical analysis Patients were described using counts and percentages, both overall and by cardiac troponin level. cardiac troponin was grouped into exponential categories: 1–10 × ULN (0.04–0.4 ng/mL], 10–100 × ULN (0.4–4 ng/mL], 100–1,000 × ULN (4–40 ng/mL], 1,000–1,0000 × ULN (40–400 ng/mL], and > 10,000 × ULN (> 400 ng/mL). Initial survival comparisons between groups were performed using Kaplan–Meier plots, and Kaplan–Meier estimates were used to obtain outcome rates at 1 year; death estimates account for censoring at the time of data collection for patients with recent events, while other estimates account for censoring both at the time of data collection and due to death. Hazard Ratios (HRs) and associated confidence intervals (CIs) for each troponin category compared to the reference category were obtained via Cox models, with robust sandwich covariance matrix estimates to account for the correlation among multiple events for some subjects. Analyses were repeated with the addition of several potential confounders–age, race, ethnicity, diabetes, hypertension, smoking status, and creatinine level– added to the model to obtain adjusted hazard ratios. Finally, all analyses were repeated with lowest 4 categories combined, compared with the > 10,000 × ULN (> 400 ng/mL) group, to provide hazard for this group compared to all other subjects. CIs presented are $99\%$ confidence intervals, corresponding to an alpha of 0.01. Analyses were conducted in SAS version 9.4. ## Patient characteristics Between 2012 and 2019, 27,570 hospitalizations occurred where a primary diagnosis of AMI was assigned. Of these, 8,210 ($30\%$) did not have a cardiac troponin value recorded, and 1,166 ($4\%$) had a peak cardiac troponin value ≤ 0.04, resulting in an analytic sample of 18,194 events. Some patients had multiple events in the dataset, resulting in a patient-level sample of 15,800 unique patients (Supplement). Baseline characteristics of patients at each event are reported in Table 1. Most patients were male, Caucasian, and older, with almost a third actively smoking cigarettes. Table 1Patient characteristics by troponin category at the event level (patients with more than one myocardial infarction can be included more than once)AllTnI (0.04–0.4] $$n = 3$$,$833\%$ (N)TnI (0.4–4] $$n = 6$$,$326\%$ (N)TnI (4–40] $$n = 5$$,$485\%$ (N)TnI (40–400] $$n = 2$$,$382\%$ (N)TnI 400 + $$n = 168$$% (N)Female gender$38.6\%$ [7,020]$42.4\%$ [1,625]$45.1\%$ [2,853]$34.6\%$ [1,899]$25.6\%$ [609]$20.2\%$ [34]Race White or Caucasian$75.9\%$ [13,818]$73.3\%$ [2,809]$77.3\%$ [4,889]$77.0\%$ [4,224]$74.5\%$ [1,774]$72.6\%$ [122] Black or African American$8.6\%$ [1,571]$11.7\%$ [449]$8.8\%$ [558]$7.3\%$ [402]$6.3\%$ [151]$6.5\%$ [11] Other or multiple races$13.1\%$ [2,379]$13.2\%$ [505]$12.0\%$ [757]$12.9\%$ [710]$15.9\%$ [379]$16.7\%$ [28] Unknown$2.3\%$ [426]$1.8\%$ [70]$1.9\%$ [122]$2.7\%$ [149]$3.3\%$ [78]$4.2\%$ [7]Ethnicity Non-Hispanic$87.0\%$ [15,833]$87.2\%$ [3,344]$88.3\%$ [5,587]$86.3\%$ [4,732]$85.5\%$ [2,037]$79.2\%$ [133] Hispanic$10.5\%$ [1,911]$10.7\%$ [409]$9.7\%$ [615]$10.8\%$ [593]$11.3\%$ [268]$15.5\%$ [26] Unknown$2.5\%$ [450]$2.1\%$ [80]$2.0\%$ [124]$2.9\%$ [160]$3.2\%$ [77]$5.4\%$ [9]Age < $5012.5\%$ [2,282]$11.7\%$ [448]$11.1\%$ [701]$13.6\%$ [748]$15.1\%$ [359]$15.5\%$ [26] 50–$6432.3\%$ [5,871]$30.2\%$ [1,156]$29.8\%$ [1,883]$33.8\%$ [1,853]$38.1\%$ [908]$42.3\%$ [71] 65–$7936.9\%$ [6,713]$36.0\%$ [1,381]$38.6\%$ [2,441]$36.7\%$ [2,013]$34.4\%$ [819]$35.1\%$ [59] 80 + $18.3\%$ [3,328]$22.1\%$ [848]$20.6\%$ [1,301]$15.9\%$ [871]$12.4\%$ [296]$7.1\%$ [12]Diabetes$1.9\%$ [345]$1.7\%$ [66]$2.2\%$ [140]$1.8\%$ [99]$1.6\%$ [37]$1.8\%$ [3]Hypertension$3.8\%$ [700]$3.8\%$ [146]$4.1\%$ [260]$4.0\%$ [222]$2.9\%$ [68]$2.4\%$ [4]Cigarettes$29.2\%$ [5,312]$32.3\%$ [1,238]$30.3\%$ [1,918]$26.8\%$ [1,471]$26.8\%$ [639]$27.4\%$ [46]Highest creatinine > $1.529.7\%$ [5,400]$30.6\%$ [1,173]$30.3\%$ [1,914]$29.6\%$ [1,624]$25.5\%$ [608]$48.2\%$ [81] ## Distribution of Peak troponin values Of the 18,194 events with a diagnosis of AMI and a recorded cardiac troponin above ULN, $21.1\%$ ($$n = 3$$,833) had a peak cardiac troponin 1–10 × ULN (0.04–0.4 ng/mL], $34.8\%$ ($$n = 6$$,326) had a peak cardiac troponin 10–100 × ULN (0.4–4 ng/mL], $30.1\%$ ($$n = 5$$,485) had a peak cardiac troponin 100–1,000 × ULN (4–40 ng/mL], $13.1\%$ ($$n = 2$$,382) had peak a cardiac troponin 1,000–10,000 × ULN (40–400 ng/mL], and $0.9\%$ ($$n = 168$$) had peak troponin > 10,000 × ULN (> 400 ng/mL), (Fig. 1).Fig. 1Histogram of exponential groupings of peak serum troponin elevations (ng/dL) among hospitalizations for acute myocardial infarction 2012–2019 ## Treatments and outcomes by Peak troponin At 1 year, mortality rates were highest with highest levels of cardiac troponin (Fig. 2). Of the patients with a cardiac troponin > 10,000 ULN, $33\%$ died within one year (Table 2). The other groups had mortality rates ranging from 17–$21\%$; adjusted hazard ratio (AHR) ($99\%$ CI) for the cardiac troponin > 10,000 ULN group compared to all others was 1.86 (1.21, 2.86). MCS in the year after the event increased with cardiac troponin level, highest in patients with cardiac troponin > 10,000 × ULN when reviewed using automated data extraction; the AHR ($99\%$ CI) for the cardiac troponin > 10,000 ULN group compared to all others was 5.73 (3.38, 9.70). In patients with a cardiac troponin > 10,000 × ULN, $1.6\%$ ($$n = 2$$) received durable LVAD placement and $2.5\%$ ($$n = 3$$) received a heart transplant within one year. Fewer than $1\%$ of patients in the other groups were noted to have received a LVAD or heart transplant within one year. For LVAD this difference was not statistically significant; for transplant the AHR ($99\%$ CI) for the cardiac troponin > 10,000 ULN group compared to all others was 17.04 (3.00, 96.86).Fig. 2One-year survival in patients with hospitalization for acute myocardial infarction, stratified by exponential groupings of peak serum troponin elevationTable 2One-year outcomes comparison between troponin groupsOutcome[0.04,0.4](0.4,4](4,40](40,400](400 +)HR ($99\%$ CI)For 400 + compared to other groups combinedDeath% at 1 year$20.0\%$$20.9\%$$18.2\%$$16.7\%$$32.8\%$HR ($99\%$ CI)1.06 (0.93, 1.21)1.14 (1.02, 1.27)-reference-0.93 (0.80, 1.09)2.13 (1.46, 3.13)2.03(1.40, 2.96)Adj. HR ($99\%$CI)0.95 (0.84, 1.08)1.04 (0.93, 1.16)-reference-1.08 (0.92, 1.26)1.89 (1.22, 2.92)1.86 (1.21, 2.86)Left ventricular assist device% at 1 year$0.3\%$$0.2\%$$0.4\%$$0.4\%$$1.6\%$HR ($99\%$ CI)0.84 (0.26, 2.72)0.55 (0.22, 1.39)-reference-0.97 (0.34, 2.75)4.03 (0.59, 27.50)5.01 (0.78, 32.25)Adj. HR ($99\%$CI)0.75 (0.26, 2.22)0.54 (0.22, 1.31)-reference-1.04 (0.36, 3.01)3.31 (0.46, 23.80)4.16 (0.62, 27.80)Transplant% at 1 year$0.0\%$$0.0\%$$0.3\%$$0.2\%$$2.5\%$HR ($99\%$ CI)0.00 (0.00, 0.00)0.07 (0.01, 0.82)-reference-0.70 (0.19, 2.64)8.92 (1.66, 47.86)20.96 (3.97, 110.64)Adj. HR ($99\%$CI)0.00 (0.00, 0.00)0.07 (0.01, 0.71)-reference-0.68 (0.16, 2.81)7.62 (1.32, 44.17)17.04 (3.00, 96.86)Mechanical circulatory support% at 1 year$1.2\%$$1.6\%$$3.0\%$$4.5\%$$16.2\%$HR ($99\%$ CI)0.4 (0.26, 0.63)0.53 (0.38, 0.75)-reference-1.55 (1.12, 2.15)5.78 (3.30, 10.16)7.42 (4.35, 12.68)Adj. HR ($99\%$CI)0.43 (0.27, 0.67)0.55 (0.39, 0.77)-reference-1.59 (1.14, 2.21)4.70 (2.71, 8.16)5.73 (3.38, 9.70)Survival estimates obtained via Kaplan-Meier. Hazard ratios obtained via Cox models, with robust sandwich covariance matrix estimates to account for the correlation among multiple events for some subjects. Covariates in adjusted models included age, gender, race, ethnicity, diabetes, hypertension, smoking status, and creatinine level. ## Highest troponin group For all 168 patients with cardiac troponin > 10,000 ULN (> 400 ng/dL), the clinical course was further explored via chart review (Table 3). Almost all patients were diagnosed with a Type 1 MI ($95\%$, $$n = 158$$). Most patients received a PCI within 7 days of cardiac troponin peak ($87\%$, $$n = 146$$) and only 5 patients received CABG within 7 days ($3\%$). In our manual chart review performed in this group, it was noted that 42 MCS procedures performed prior to transfer between hospitals were not captured by the automatic data extraction leading to a difference between patients requiring temporary MCS in the data extraction at $16\%$ vs $41\%$ when manual chart review was performed in this group. For those $$n = 69$$ ($41\%$) who received temporary MCS during the hospitalization, $$n = 40$$ ($24\%$) received intra-arterial balloon placement, $$n = 27$$($16\%$) received Impella support, and $$n = 10$$ ($6\%$) needed extracorporeal membrane oxygenation. Five deaths were picked up in the > 400 ng/mL group through manual chart review which were missed during automated data extraction. Table 3Outcomes obtained via chart review for 168 pts in highest troponin group only% (N) of 168 patientsType of myocardial infarction Type $194\%$ [158] Type $25\%$ [8] Unknown$1\%$ [2]Procedure (within 7 days) Percutaneous coronary intervention within 7 days$87\%$ [146] Coronary artery bypass graft within 7 days$3\%$ [5] Troponin peak occurred post-procedure$79\%$ [134]Outcomes (within 1 year)% w/ outcome (KM estimate at 1 year) Death$36\%$ Any temporary Mechanical circulatory support$41\%$ Intra-aortic balloon pump$24\%$ Impella$16\%$ Extracorporeal membrane oxygenation$6\%$ Tandem Heart$0\%$ Other$2\%$MCS types can add up to more than ‘any MCS’ because a patient can have more than one type of temporary MCSProcedures information is simple N and percent among this group. Outcomes information obtained via Kaplan–Meier estimates to account for censoring ## Discussion This retrospective observational study aimed to identify whether a threshold peak cardiac troponin level at the time of AMI could quickly identify patients at particularly high risk of subsequent adverse outcomes, and thus rapid consideration of MCS, LVAD placement, heart transplant, or involvement of palliative care. We identified that an inflection point indicating much worse outcomes exist in patients with cardiac troponin levels elevated > 10,000 times the ULN (> 400 ng/mL). Though this is an infrequent occurrence, accounting for only $1\%$ of all our cases, this group had a statistically higher incidence of death at $33\%$ vs 17–$21\%$ in all other groups (AHR ($99\%$ CI): 1.86(1.21, 2.86)) and captured all use of advanced therapies (transplant and LVAD) in the following year. Although there is a well understood relationship between the magnitude of troponin elevation and the extent of myocyte necrosis [9], the quantifiable risk of mortality and requirement of advanced therapies has not been previously explored. This study is the first to our knowledge to stratify the need for MCS, LVAD or transplant based on extent of troponin elevation during AMI. Only one other study of post PCI patients correlated very high troponin levels with death [7]. Due to the worsened outcomes in patients with a cardiac troponin > 10,000 × ULN (> 400 ng/mL), this frequently used biomarker of myocardial necrosis may serve to identify patients that would benefit from early consideration of advanced cardiac therapies. There are several triggers for patients to be referred for consideration of advanced therapies, including the I-NEED-HELP acronym [10]. Our analysis serves as possibly another “trigger” for consideration to refer patients for advanced therapies, in the context of individual preference and consideration of comorbid conditions. In the subset of our study population having a troponin > 400 ng/ dL ($$n = 168$$), one third died, more than double the rate of death after transplant or durable LVAD implantation. We therefore conclude that early consideration and referral for advanced therapies in this small sub-set of patients would be reasonable. Peak troponin was noted in the post-PCI period for most patients in the highest troponin group ($79\%$, $$n = 132$$). This group was noted to have significantly worse outcomes, including mortality, than the more modest troponin elevations. This is a noteworthy observation as there exists uncertainty as to the clinical significance of periprocedural cardiac biomarker elevation [11], with many providers considering the degree of biomarker elevation after PCI to have minimal prognostic significance. We recognize that some providers do not check troponin levels after PCI, such that the patients included here may represent a higher risk population (e.g. ongoing shock, complex PCI). Regardless, the data collected in our study, as well as another recent study [7], suggest post-procedural troponin elevation may be a clinically relevant datapoint. Troponin assays and their relationship to novel therapies for revascularization, inflammatory response, and infarct size is evolving [12]. Use of high sensitivity troponin assays (HsTrop) are being widely adopted given superior sensitivity to traditional troponin assays for early detection of acute coronary syndrome [13, 14]. Though there is improved detection of low levels of troponin with HsTrop assays, the upper range of contemporary HsTrop assays is significantly lower than traditional troponin assays (e.g., 25,000 ng/L for Siemens HsTrop and 270,270 ng/L for Beckman HsTrop, compared to 500,000 ng/L for the traditional Beckman cardiac troponin assay used for many of the patients in this study). This loss of quantitation at the upper ranges of troponin release with HsTrop may have implications for prognosis after AMI, and future iterations of HsTrop assays may need to consider optimizing both low and high ends of troponin quantitation. In the meantime, similar studies to this one with HsTrop results may be instructive. ## Limitations First, as with all retrospective studies, we were limited by the quality of available data for analysis. We used ICD codes to identify patients with AMI and may have missed patients that had cardiac troponin elevations that did not meet criteria for AMI coding. Many true peak cardiac troponins are missed as post-PCI measurement of troponins is variable. Second, automated data pulls of the electronic health record also have varying degrees of inaccuracy (e.g., 5 deaths were picked up in the > 400 ng/mL group through manual chart review which were missed during automated data extraction). Third, low event numbers for some outcomes, particularly LVAD and transplant, resulted in wide confidence intervals and limited precision in our estimates and between-group comparisons. Fourth, our estimation of hazards of MCS, transplant, and LVAD in the presence of the competing risk of death assumes that events are independent, and our hazard ratios may be altered to the extent that this assumption is not met. Fifth, we only assessed cardiac troponin I and not other cardiac troponins. The goal of our study was not to assess the relative performance of multiple assays, but rather to assess the prognostic information conferred by orders of magnitude serum troponin elevation above ULN. Last, HsTrop are currently being increasingly adopted and traditional troponin assays will likely be phased out in the future. ## Conclusion Overall, cardiac troponin > 10,000 times the ULN (> 400 ng/mL), though rare, has particularly poor outcomes in regard to requiring mechanical circulatory support and death. This serves to identify a group where clarification of goals of care and consideration for early referral for advanced therapies may prove valuable. Here we also highlight the relevance of measuring cardiac troponin until peak even if patients undergo PCI. Current HsTrop assays that are being increasingly adopted, often have lower upper limits of quantification, which may lead to a loss of prognostic information after AMI. ## Supplementary Information Additional file 1: Supplemental Table 1. Patient-level and event level demographics. ## Authors’ information Not applicable. ## References 1. Thygesen K, Mair J, Katus H, Plebani M, Venge P, Collinson P, Lindahl B, Giannitsis E, Hasin Y, Galvani M. **Recommendations for the use of cardiac troponin measurement in acute cardiac care**. *Eur Heart J* (2010) **31** 2197-2204. DOI: 10.1093/eurheartj/ehq251 2. Thygesen K, Alpert JS, Jaffe AS, Chaitman BR, Bax JJ, Morrow DA, White HD. *J Am Coll Cardiol* (2018) **72** 2231-2264. DOI: 10.1016/j.jacc.2018.08.1038 3. Christenson RH, Vollmer RT, Ohman EM, Peck S, Thompson TD, Duh SH, Ellis SG, Newby LK, Topol EJ, Califf RM. **Relation of temporal creatine kinase-MB release and outcome after thrombolytic therapy for acute myocardial infarction. TAMI Study**. *Group Am J Cardiol* (2000) **85** 543-547. DOI: 10.1016/s0002-9149(99)00808-5 4. Chia S, Senatore F, Raffel OC, Lee H, Wackers FJ, Jang IK. **Utility of cardiac biomarkers in predicting infarct size, left ventricular function, and clinical outcome after primary percutaneous coronary intervention for ST-segment elevation myocardial infarction**. *JACC Cardiovasc Interv* (2008) **1** 415-423. DOI: 10.1016/j.jcin.2008.04.010 5. Alexander JH, Sparapani RA, Mahaffey KW, Deckers JW, Newby LK, Ohman EM, Corbalán R, Chierchia SL, Boland JB, Simoons ML. **Association between minor elevations of creatine kinase-MB level and mortality in patients with acute coronary syndromes without ST-segment elevation. PURSUIT Steering Committee. Platelet Glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin Therapy**. *JAMA* (2000) **283** 347-53. DOI: 10.1001/jama.283.3.347 6. Roe MT, Peterson ED, Li Y, Pollack CV, Christenson RH, Peacock WF, Fesmire FM, Newby LK, Jesse RL, Hoekstra JW. **Relationship between risk stratification by cardiac troponin level and adherence to guidelines for non-ST-segment elevation acute coronary syndromes**. *Arch Intern Med* (2005) **165** 1870-1876. DOI: 10.1001/archinte.165.16.1870 7. Silvain J, Zeitouni M, Paradies V, Zheng HL, Ndrepepa G, Cavallini C, Feldman DN, Sharma SK, Mehilli J, Gili S. **Procedural myocardial injury, infarction and mortality in patients undergoing elective PCI: a pooled analysis of patient-level data**. *Eur Heart J* (2021) **42** 323-334. DOI: 10.1093/eurheartj/ehaa885 8. Maddox TM, Januzzi JL, Allen LA, Breathett K, Butler J, Davis LL, Fonarow GC, Ibrahim NE, Lindenfeld J, Masoudi FA. **2021 Update to the 2017 ACC expert consensus decision pathway for optimization of heart failure treatment: answers to 10 pivotal issues about heart failure with reduced ejection fraction: a report of the American College of Cardiology Solution Set Oversight Committee**. *J Am Coll Cardiol* (2021) **77** 772-810. DOI: 10.1016/j.jacc.2020.11.022 9. Byrne RA, Ndrepepa G, Braun S, Tiroch K, Mehilli J, Schulz S, Schömig A, Kastrati A. **Peak cardiac troponin-T level, scintigraphic myocardial infarct size and one-year prognosis in patients undergoing primary percutaneous coronary intervention for acute myocardial infarction**. *Am J Cardiol* (2010) **106** 1212-1217. DOI: 10.1016/j.amjcard.2010.06.050 10. Baumwol J. **“I Need Help”-A mnemonic to aid timely referral in advanced heart failure**. *J Heart Lung Transplant* (2017) **36** 593-594. DOI: 10.1016/j.healun.2017.02.010 11. Cutlip DE, Kuntz RE. **Does creatinine kinase-MB elevation after percutaneous coronary intervention predict outcomes in 2005? Cardiac enzyme elevation after successful percutaneous coronary intervention is not an independent predictor of adverse outcomes**. *Circulation* (2005) **112** 916-922. DOI: 10.1161/CIRCULATIONAHA.104.478347 12. Paolisso P, Bergamaschi L, Santulli G, Gallinoro E, Cesaro A, Gragnano F, Sardu C, Mileva N, Foa A, Armillotta A. **Infarct size, inflammatory burden, and admission hyperglycemia in diabetic patients with acute myocardial infarction treated with SGLT2-inhibitors: a multicenter international registry**. *Cardiovasc Diabetol* (2022) **21** 77. DOI: 10.1186/s12933-022-01506-8 13. Neumann JT, Twerenbold R, Ojeda F, Sörensen NA, Chapman AR, Shah ASV, Anand A, Boeddinghaus J, Nestelberger T, Badertscher P. **Application of high-sensitivity troponin in suspected myocardial infarction**. *N Engl J Med* (2019) **380** 2529-2540. DOI: 10.1056/NEJMoa1803377 14. Vasile VC, Jaffe AS. **High-sensitivity cardiac troponin for the diagnosis of patients with acute coronary syndromes**. *Curr Cardiol Rep* (2017) **19** 92. DOI: 10.1007/s11886-017-0904-4
--- title: 'Efficacy of surgical treatment on polypharmacy of elderly patients with lumbar spinal canal stenosis: retrospective exploratory research' authors: - Sota Nagai - Risa Inagaki - Takehiro Michikawa - Soya Kawabata - Kaori Ito - Kurenai Hachiya - Hiroki Takeda - Daiki Ikeda - Shinjiro Kaneko - Shigeki Yamada - Nobuyuki Fujita journal: BMC Geriatrics year: 2023 pmcid: PMC10037878 doi: 10.1186/s12877-023-03853-x license: CC BY 4.0 --- # Efficacy of surgical treatment on polypharmacy of elderly patients with lumbar spinal canal stenosis: retrospective exploratory research ## Abstract ### Background Polypharmacy is a growing public health problem occurring in all healthcare settings worldwide. Elderly patients with lumbar spinal canal stenosis (LSS) who manifest low back and neuropathic pain and have a high frequency of comorbidity are predicted to take many drugs. However, no studies have reported polypharmacy in elderly patients with LSS. Thus, we aimed to review the polypharmacy among elderly LSS patients with elective surgeries and examine how the surgical treatment reduces the polypharmacy. ### Methods We retrospectively enrolled all the patients aged ≥ 65 years who underwent spinal surgery for LSS between April 2020 and March 2021. The prescribed drugs of participants were directly checked by pharmacists in the outpatient department preoperatively and 6-month and 1-year postoperatively. The baseline characteristics were collected beside the patient-based outcomes including Roland–Morris Disability Questionnaire, Zurich Claudication Questionnaire, and Japanese Orthopaedic Association Back Pain Evaluation Questionnaire (JOABPEQ). The cutoff number of drugs for polypharmacy was defined as 6. The prescription drugs were divided into 9 categories: drugs for neuropsychiatric, cardiovascular, respiratory, digestive, endocrine metabolic, and urinary renal diseases; blood products; pain relief medication; and others. ### Results A total of 102 cases were finally analyzed, with a follow-up rate of $78.0\%$. Of the participants, the preoperative polypharmacy prevalence was $66.7\%$. The number of drugs 6-month and 1-year postoperatively was significantly less than the preoperative one. The proportions of polypharmacy at 6 months and 1 year after surgery significantly decreased to $57.8\%$ and $55.9\%$, respectively. When the prescribed drugs were divided into 9 categories, the number of drugs for pain relief and digestive diseases was significantly reduced after surgery. The multi-variable analysis revealed that a higher score in the psychological disorder of JOABPEQ was associated with 3 or more drugs decreased 1-year postoperatively (OR, 2.5; $95\%$ CI: 1.0–6.1). ### Conclusion Polypharmacy prevalence was high among elderly LSS patients indicated for lumbar spinal surgery. Additionally, our data showed that lumbar spinal surgery was effective in reducing polypharmacy among elderly LSS patients. Finally, the multi-variable analysis indicated that better psychological condition was associated with the reduction of prescribed drugs after lumbar spinal surgery. ## Background As the world moves toward an aging society, the proportion growth of the elderly is a global health and socioeconomic problem [1]. Due to population growth and aging, the number of patients with musculoskeletal disorders is increasing rapidly [1, 2]. Lumbar spinal canal stenosis (LSS), one of the common degenerative musculoskeletal disorders, is caused by a narrowing of the lumbar spinal canal that compresses the cauda equina and nerve roots. In the computed tomography study using a US community-based sample, LSS prevalence was about $19\%$ among patients in the 60s [3]. A Japanese population-based study using a self-administered questionnaire for LSS reported that the prevalence was around $11\%$ in the 70s, which increased with age [4]. The first-line treatment for LSS is conservative therapy, such as pharmacotherapy, exercise therapy, and block therapy [5]. There are many drug options for LSS, including Non-Steroidal Anti-Inflammatory Drugs (NSAIDs), opioids, serotonin-noradrenalin reuptake inhibitors (SNRIs), pregabalin/mirogabalin, prostaglandin E1 analogs (PGE1), acetaminophen, and so on [6]. On the other hand, surgical treatment, including decompression and/or fusion, is indicated for LSS patients who are refractory to conservative therapy, and the outcomes are generally favorable [7, 8]. Polypharmacy, taking multiple drugs simultaneously, is common among elderlies because they usually suffer from numerous diseases. The cutoff number for drugs in polypharmacy is not clearly defined, but commonly 5 or 6 [9, 10]. Polypharmacy is associated with the use of potentially inappropriate medications [9], resulting in increased adverse drug events (ADEs), including poor treatment adherence, reduction of physical function, decreased cognitive function, and high risk of falls [11–15]. WHO reported that polypharmacy is a growing public health problem occurring in all healthcare settings worldwide [16]. A recent study reported that degenerative lumbar spinal disorders including LSS were significantly associated with polypharmacy in elderlies with degenerative musculoskeletal disorders [17]. In particular, LSS patients with neuropathic pain and multiple comorbidities are prone to get more drugs [18, 19]. In addition, since LSS patients were at high risk of falls due to decreased motor function of the lower extremities [20], it is essential to prevent polypharmacy associated with risk of falls. However, to the best of our knowledge, no studies have reported polypharmacy in elderly patients with LSS. Additionally, the effect of surgical treatment for LSS patients on polypharmacy is also unclear. Therefore, the first aim of this study was to investigate the detail of drugs prescribed to elderly patients with elective surgeries for LSS. The second was to examine how the surgical treatment for LSS reduces the number of prescribed drugs. ## Study participants The research design of this study was a retrospective observational study. We retrospectively enrolled all the patients aged ≥ 65 years who underwent lumbar spinal surgery for LSS at our institution between April 2020 and March 2021. The follow-up period was one year. Surgical treatment was indicated for patients with clear LSS symptoms, including leg pain and numbness and neurogenic claudication, and who were refractory to conservative therapy according to the guideline [5]. Since this was an exploratory analysis of observational study, we did not estimate sample size and continued to include participants throughout the duration of the study. LSS diagnosis was confirmed using MRI, myelography, or computed tomography. Cases with fusion segments ≥ 4 were excluded before enrollment. We defined spondylolisthesis as an anterior slip of the upper vertebra ≥ $5\%$ and degenerative lumbar scoliosis as Cobb angle ≥ 10°. Cases with multiple spinal lumbar surgeries were defined as failed back surgery syndrome (FBSS). Posterior fusion was recommended for patients with spondylolisthesis and/or a posterior opening > 5° on dynamic lateral radiographs. Case with additional lumbar surgery during the follow-up period was excluded. ## Ethics approval The ethics committee of our institution granted ethical approval for this study (approval No. HM20-530). The ethics committee approved the inclusion of all eligible patients in the study unless we were contacted to opt-out. All study methods were conducted in accordance with the guidelines set out in the Declaration of Helsinki. ## Data collection Prescribed drugs of participants were directly checked by pharmacists in the outpatient department preoperatively and 6-month and 1-year postoperatively. The participants have the same physician who prescribe the drugs before and after surgery. The drug was prescribed at the discretion of the individual physician. The prescribing physician was unaware that the patient was a participant in this study. The participants with 6 medications were considered to have polypharmacy. Enrolled patients were assessed preoperatively in addition to 6-month and 1-year after the surgery for the patient-based outcomes including Roland–Morris Disability Questionnaire (RDQ), Zurich Claudication Questionnaire (ZCQ), and Japanese Orthopaedic Association Back Pain Evaluation Questionnaire (JOABPEQ). We collected the following data for each patient: age; gender; body mass index (BMI); medical history, including type 2 diabetes mellitus (DM), hypertension, hyperlipidemia, cardiovascular disease, cerebrovascular disease, and cancer; spondylolisthesis; degenerative lumbar scoliosis; FBSS; perioperative factors including surgical procedure such as decompression and/or fusion, the number of decompression level, surgical time, and surgical blood loss. The surgical treatment was considered “effective” or “not effective” according to the JOABPEQ based on the following; an increase of ≥ 20 points in the postoperative score over the preoperative one, or a preoperative score < 90 with a postoperative score ≥ 90 points [20]. ## Classification of the drugs According to the previous study [17], the prescription drugs were divided into 9 categories: drugs for neuropsychiatric, cardiovascular, respiratory, digestive, endocrine metabolic, and urinary renal diseases; blood products; pain relief medication; and others. Antithrombotic drugs were included in the category of drugs for cardiovascular diseases. Osteoporosis drugs were included in the category of drugs for endocrine metabolic diseases. Pain relief medication included NSAIDs, pregabalin/mirogabalin, opioids, SNRIs, and acetaminophen. ## Statistical analyses The data among groups were compared using Chi-squared test, McNemar’s test, or Wilcoxon signed-rank test, as appropriate. P values < 0.05 were considered to indicate statistical significance. When we performed the McNamer and Wilcoxon signed rank test more than once, a P value of 0.025 ($\frac{0.05}{2}$) was used as a statistically significance to avoid the type 1 error. To examine the independent associations of 1-year postoperative decrease of 3 or more drugs, we constructed a Poisson regression model that included age, sex, metabolic component, FBSS, surgical procedure, and preoperative score in each domain of JOABPEQ, and estimated relative risk (RR) and $95\%$ confidence intervals (CIs) for 1-year postoperative decrease of 3 or more drugs. In the Poisson regression model, JOABPEQ was selected as an explanatory variable among the three patient-based outcomes because JOABPEQ can be divided into five domains: pain disorder, lumbar function, walking ability, social life, and psychological disorder, and patients’ conditions can be evaluated in detail [21]. Because there is no clinical cutoff value for categorizing the scores in each domain of JOABPEQ, the scores of JOABPEQ were categorized in tertiles to account for the number of participants. In addition, because scores of JOABEPQ have misclassification, we divided them into groups with similar scores. The metabolic component was defined as having at least one of the following; BMI of 25 or higher, DM, hypertension, and hyperlipidemia [22]. Poisson regression was performed using the STATA16 software (Stata Corporation, College Station, TX, USA). ## Results Totally, 132 patients were enrolled in this study. A total of 29 cases were lost during the follow-up period, with a follow-up rate of $78.0\%$ (Fig. 1). One case with additional lumbar surgery during the follow-up period were excluded (Fig. 1). Finally, a total of 102 cases were analyzed. Table 1 summarizes baseline characteristics. Table 2 shows the preoperative score of RDQ, ZCQ including symptom severity and physical function, and JOABPEQ, in addition to the data for 6-month and 1-year postoperatively. Postoperative RDQ, ZCQ, and JOABPEQ scores (6-month and 1-year) were significantly improved compared to the preoperative data. The frequency of “effective” in each domain of JOABPEQ was also favorable. Fig. 1Flow diagram showing the flow of participants through each stage in this study Table 1Baseline CharacteristicsPatientsn = 102GenderMale: 56 Female: 46Age (yeas)75.5 ± 5.9BMI (kg/m2)23.8 ± 3.4Medical historyDiabetes mellitus31 ($30.4\%$)Hypertension65 ($63.7\%$)Hyperlipidemia56 ($55.0\%$)Cardiovascular disease34 ($33.3\%$)Cerebrovascular disease11 ($10.8\%$)Cancer18 ($17.6\%$)Duration of conservative therapy before surgery< 6 months28 ($27.5\%$)> = 6 months, < 1 year17 ($16.7\%$)> = 1 year, < 3 year31 ($30.4\%$)> = 3 years26 ($24.5\%$)Spondylolisthesis35 ($34.3\%$)Degenerative lumbar scoloiosis14 ($13.7\%$)FBSS8 ($7.8\%$)Surgical proceduredecompression59 ($57.8\%$)decompression + fusion43 ($42.2\%$)Surgical time (min)140.1 ± 90.5Surgical blood loss (ml)150.4 ± 168.2Decompression levels2.0 ± 0.9FBSS, Failed back surgery syndrome Table 2Valuables of patient reported outcome at baseline and follow-up after surgery ($$n = 102$$)Median (25-$75\%$tile)p value*Preoperation6POM1POYpreoperation vs. 6POMpreoperation vs. 1POYRDQ13 (7–17)6 (1–11)6 (0–12)< 0.01< 0.01ZCQSymptom severity3.3 (2.9–3.7)2.3 (1.6–2.7)2.1 (1.7–2.7)< 0.01< 0.01Physical function2.8 (2.2–3.2)1.6 (1.2–2.2)1.6 (1.2–2.2)< 0.01< 0.01JOABPEQPain disorder43 (14–71)100 (71–100)100 (43–100)< 0.01< 0.01Lumbar function50 (25–75)79 (50–100)83 (42–83)< 0.01< 0.01Walking ability21 (0–43)68 (36–100)71 (33–93)< 0.01< 0.01Social life35 (22–51)57 (39–78)59 (44–81)< 0.01< 0.01Psychological disorder44 (32–53)55 (45–69)60 (48–72)< 0.01< 0.01The number of effective case of surgical treatment on JOABPEQPain disorder71($69.6\%$)62 ($60.8\%$)Lumbar function53($52.0\%$)54 ($52.4\%$)Walking ability67($65.7\%$)69 ($67.0\%$)Social life46($45.0\%$)59 ($57.3\%$)Psychological disorder31($30.0\%$)35 ($34.0\%$)RDQ, Roland–Morris Disability Questionnaire; ZCQ, Zurich Claudication Questionnaire; JOABPEQ, JOA Back Pain Evaluation Questionnaire*Wilcoxon signed-rank test Figure 2A shows the distribution of the number of preoperative prescribed drugs among the participants. When the cutoff number of drugs was 6, the prevalence of polypharmacy was $66.7\%$. Figure 2B presents the preoperative and 6-month and 1-year postoperative number of prescribed drugs in the participants. The number of drugs 6-month and 1-year postoperatively were significantly less than the preoperative one. Figure 2C shows the preoperative and 6-month and 1-year postoperative proportion of polypharmacy. The proportions of polypharmacy at 6 months and 1 year after surgery were significantly reduced to $57.8\%$ and $55.9\%$, respectively. Fig. 2 A) The distribution of the number of preoperative prescription drugs in all the participants. B) Comparison of the preoperative and 6-month and 1-year postoperative number of prescribed drugs in the participants. C) Comparison of the preoperative and 6-month and 1-year postoperative proportion of polypharmacy Changes in the number of drugs before and after surgery were investigated in 9 categories. The number of drugs for pain relief and digestive diseases at both 6 months and 1 year after surgery was significantly lower than that at preoperation (Fig. 3). The other categories showed almost no change in the number of drugs (Fig. 3). Among the common prescribed drugs for LSS patients, we investigated changes in the proportion of patients taking NSAIDs, pregabalin/mirogabalin, opioids, SNRIs, PGE1, and acetaminophen before and after surgery (Fig. 4). In NSAIDs, the proportion of patients taking them at 1 year after surgery were significantly lower than that before surgery (Fig. 4). In pregabalin/mirogabalin, opioids, and PGE1, the proportion of patients taking them at both 6 months and 1 year after surgery were significantly lower than that before surgery (Fig. 4). Figure 5 showed the distribution of changes in the number of drugs 1 year after surgery. Of the participants, $25.5\%$ had a decrease of more than 3 drugs, $21.6\%$ had a decrease of 1–2 drugs, and $26.5\%$ had no change. In contrast, $26.5\%$ showed an increase in drug consumption. To clarify the relationship between surgical efficacy and postoperative changes in prescription drug counts, we compared the frequency of effective cases of surgical treatment in each domain of JOABPEQ between cases with an increase (I group; $$n = 27$$) and cases with a decrease (D group; $$n = 48$$) in drugs one year after surgery (Fig. 6). However, they showed no significant differences between two groups (Fig. 6). Next, because Fig. 5 showed that the participants were roughly divided into four equal parts in the distribution of changes in the number of drugs, we focused on the top $\frac{1}{4}$ which indicated 3 or more drug reductions after 1-year of surgery. Here, we used a Poisson regression model to identify the factors associated with 3 or more drug reductions after 1-year of surgery. In this analysis, the scores of JOABPEQ were categorized in tertiles because there is no clinical cutoff value for categorizing the scores in each domain of JOABPEQ. In addition, because scores of JOABEPQ have misclassification, we divided them into groups with similar scores. The multi-variable analysis revealed that a higher score in the psychological disorder in JOABPEQ was associated with 3 or more drug reduction after 1-year of surgery (RR, 2.5; $95\%$ CI: 1.0–6.1) (Table 3). Fig. 3Comparison of the preoperative and 6-month and 1-year postoperative number of prescribed drugs in 9 categories Fig. 4Comparison of the preoperative and 6-month and 1-year postoperative proportion of patients taking NSAIDs, pregabalin/mirogabalin, opioids, SNRIs, PGE1, and acetaminophen Fig. 5The distribution of changes in the number of drugs 1 year after lumbar spinal surgery Fig. 6Comparison of the frequency of effective cases of surgical treatment in each domain of JOABPEQ between cases with an increase (I group; $$n = 27$$) and cases with a decrease (D group; $$n = 48$$) in drugs one year after surgery Table 3Poisson regression model of decrease of 3 or more drugsNumber of patientsNumber of decrease of 3 or more drugsPrevalence of decrease of 3 or more drugs (%)Multi-variable model*Relative risk (RR)$95\%$ confidence intervalp-valueAge65–74471429.8Reference74-551221.80.90.5–1.90.87SexMen561526.8ReferenceWomen461123.91.00.4–2.30.99Metabolic componentsNo20840.0ReferenceYes821822.00.60.3–1.20.13FBSSNo942425.5ReferenceYes8225.00.80.3–2.30.72Surgical proceduredecompression591322.0Referencedecompression + fusion431330.21.50.7-3.00.31JOABPEQPain disorderTertile1 (< 15)38615.8ReferenceTertile2 (15–57)321031.31.60.6–4.30.32Tertile3 (> 57)321031.31.80.6–4.80.26Lumbar functionTertile1 (< 26)35925.7ReferenceTertile2 (26–67)32721.90.50.2–1.30.16Tertile3 (> 67)351028.60.50.2–1.40.17Walking abilityTertile1 [0]26519.2ReferenceTertile2 (1–29)471225.51.70.7–4.30.24Tertile3 (> 29)29931.02.40.7–8.20.15Social lifeTertile1 (< 25)341029.4ReferenceTertile2 (25–49)35617.10.50.2–1.40.19Tertile3 (> 49)331030.30.50.2–1.50.25Psychological disorderTertile1 (< 36)33618.2ReferenceTertile2 (36–50)36822.21.80.7–4.50.25Tertile3 (> 50)331236.42.61.0-6.50.045FBSS, Failed back surgery syndrome; JOABPEQ, JOA Back Pain Evaluation Questionnaire*All variables were included in this table ## Discussion For the first time, this study followed the prescribed drugs for LSS patients with lumbar spinal surgery. The frequency of polypharmacy in preoperative LSS patients was around two-thirds when the cutoff number of drugs was 6. In addition, lumbar spinal surgery significntly reduced the proportions of polypharmacy of the participants. Finally, our data indicated that psychological conditions might contribute to the postoperative reduction of prescribed drugs for LSS patients who underwent lumbar surgery. It has been reported that the ADEs, which is higher in the elderly than in the young, occurs over $10\%$ per year [23]. Since polypharmacy causes ADEs associated with drug interactions and increases the incidence of missed or mistaken drug doses, sufficient attention should be paid to polypharmacy among elderlies [11]. The elderly with degenerative musculoskeletal disorders are more likely to be diagnosed with polypharmacy because of chronic musculoskeletal pain. In a recent study, the polypharmacy prevalence (cutoff value = 6 drugs) was $52\%$ among the elderly patients with elective surgical treatment for degenerative musculoskeletal disorders [17]. Because LSS patients commonly have low back and neuropathic pain and multiple comorbidities [18], polypharmacy may be particularly common in LSS patients. Our study reported a high polypharmacy prevalence among LSS patients ($66.7\%$) who were indicated for surgical treatment; therefore, we should pay special attention to this patient group. Our results showed that $72\%$ of LSS patients had at least 1 pain relief drug, while $17\%$ had 3 or more. For LSS patients who took no pain relief medications before surgery ($28\%$), pain relief medications such as NSAIDs, pregabalin, and opioids may have already been discontinued before surgery because none of them worked. As a pain relief drug, NSAIDs have been reported to be potentially inappropriate medications for elderly patients [24]. Furthermore, NSAIDs is predicted to increase the number of drugs as the patient will require them to prevent gastrointestinal bleeding. Consistently, our results showed that $76\%$ of LSS patients had at least one drug for digestive diseases, while $19\%$ had three or more. For example, the combination of proton pump inhibitors can reduce the risk of upper gastrointestinal bleeding in elderly patients. However, although proton pump inhibitors have not been exactly defined as potentially inappropriate medications, they were reported to increase the risk of *Clostridium difficile* infections and fractures [24]. Therefore, from the viewpoint of polypharmacy, the use of NSAIDs should be especially discouraged as much as possible. Our data clearly showed that lumbar spinal surgery significantly reduced the number of prescribed drugs in LSS patients. This is mainly due to a decrease in pain relief drugs and concomitantly digestive disease drugs after surgery. In the present study, we examined the relationship between the surgical outcome of pain disorder and the change in the drug counts, but we could not find a significant relationship between them. These findings indicate that the decline in prescription drugs for LSS patients does not depend solely on surgical pain relief. The most important finding of this study was that lumbar spinal surgery significantly reduced the prevalence of LSS patient polypharmacy. Given the increasing number of patients worldwide with musculoskeletal disorders including LSS [1, 2], lumbar spinal surgery can be beneficial for the elderly LSS patients in terms of preventing ADEs such as reduction of physical function, decreased cognitive function, and high risk of falls. Lumbar spinal surgery has been reported to be effective in pain [25], motor function [26], risk of falling [20], social life [27], psychology [28], and healthy life expectancy [29] for LSS patients. Our results added one more advantage to the surgical treatment. While reducing the polypharmacy of elderly patients by lumbar spinal surgery is socioeconomically meaningful, surgical treatment, on the other hand, also raises medical expenses. Therefore, in the future, it is mandatory to make long-term health economics comparisons between LSS patients who did not undergo surgery and those who did. This study showed that among LSS patients with surgery, some had a decrease of three or more prescribed drugs after surgery, while others had an increase. We performed a multi-variable analysis to determine whether preoperative factors could predict the difference. Although the number of cases was limited and not perfect for the analysis, our results revealed that patients with a reduction of three or more drugs after surgery had relatively good psychological conditions. Considering that previous literature reported that mental health status is one of the risk factors for polypharmacy [30], the number of prescribed drugs in LSS patients with good psychological conditions may be more susceptible to the effects of surgical pain relief. Therefore, our results indicated that taking care of the psychological condition of LSS patients may facilitate the improvement of polypharmacy with lumbar spinal surgery. To increase the efficacy of lumbar spinal surgery on polypharmacy, further identification of associated factors other than psychological condition should be analyzed with larger sample size. This study has several limitations. First, data for this study were collected from a limited number of patients in a single institution. Especially, a larger sample is needed for a comprehensive multi-variable analysis. Second, more than $20\%$ of initially enrolled patients dropped out during the follow-up period. *In* general, patients who dropped out tend to have poor postoperative outcomes [31], so our data may be better as an estimate than a true result. Third, the follow-up duration of 1 year was not sufficient to assess the efficacy of lumbar spinal surgery on polypharmacy because some patients have symptoms that change more than one year after surgery. Lastly, the present study could include patients who were prescribed under appropriate decisions and management and those who were not. That is, this study could examine patients who were also diagnosed with polypharmacy but had different implications. To the best of our knowledge, however, this is the first follow-up study to assess the polypharmacy among LSS patients. This study provides useful information for LSS patients and health providers, including doctors and pharmacists. ## Conclusion This study demonstrated that the prevalence of polypharmacy was high among elderly patients with lumbar spinal surgery for LSS. In addition, our data showed that lumbar spinal surgery was effective for the reduction of polypharmacy in LSS patients. Finally, the multi-variable analysis indicated that better psychological condition was associated with the reduction of prescribed drugs after lumbar spinal surgery. ## References 1. 1.Musculoskeletal conditions. ; Accessed 2021 February 8. https://www.who.int/news-room/fact-sheets/detail/musculoskeletal-conditions. 2. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the global burden of Disease Study 2019**. *Lancet* (2020.0) **396** 1204-22. DOI: 10.1016/S0140-6736(20)30925-9 3. Kalichman L, Cole R, Kim DH, Li L, Suri P, Guermazi A. **Spinal stenosis prevalence and association with symptoms: the Framingham Study**. *Spine J* (2009.0) **9** 545-50. DOI: 10.1016/j.spinee.2009.03.005 4. Yabuki S, Fukumori N, Takegami M, Onishi Y, Otani K, Sekiguchi M. **Prevalence of lumbar spinal stenosis, using the diagnostic support tool, and correlated factors in Japan: a population-based study**. *J Orthop Sci* (2013.0) **18** 893-900. DOI: 10.1007/s00776-013-0455-5 5. 5.North American Spine Society. Evidence-based clinical guidelines for multidisciplinary spine care: diagnosis and treatment of degenerative lumbar spinal stenosis. ; 2011. Accessed 2021 January 27. http://www.spine.org/Documents/LumbarStenosis11.pdf. 6. Marcolina A, Vu K, Annaswamy TM. **Lumbar spinal stenosis and potential management with prostaglandin E1 analogs**. *Am J Phys Med Rehabil* (2021.0) **100** 297-302. DOI: 10.1097/PHM.0000000000001620 7. Kawakami M, Takeshita K, Inoue G, Sekiguchi M, Fujiwara Y, Hoshino M. **Japanese Orthopaedic Association (JOA) clinical practice guidelines on the management of lumbar spinal stenosis, 2021 - secondary publication**. *J Orthop Sci* (2022.0) **S0949–2658** 00116-6 8. Williams MG, Wafai AM, Podmore MD. **Functional outcomes of laminectomy and laminotomy for the surgical management lumbar spine stenosis**. *J Spine Surg* (2017.0) **3** 580-6. DOI: 10.21037/jss.2017.10.08 9. Masnoon N, Shakib S, Kalisch-Ellett L, Caughey GE. **What is polypharmacy? A systematic review of definitions**. *BMC Geriatr* (2017.0) **17** 230. DOI: 10.1186/s12877-017-0621-2 10. Beezer J, Al Hatrushi M, Husband A, Kurdi A, Forsyth P. **Polypharmacy definition and prevalence in heart failure: a systematic review**. *Heart Fail Rev* (2022.0) **27** 465-92. DOI: 10.1007/s10741-021-10135-4 11. Ross MSc Candidate SB, Wu PE, Atique Md Candidate A, Papillon-Ferland L, Tamblyn R, Lee TC. **Adverse drug events in older adults: review of adjudication methods in Deprescribing Studies**. *J Am Geriatr Soc* (2020.0) **68** 1594-602. DOI: 10.1111/jgs.16382 12. Lee VW, Pang KK, Hui KC, Kwok JC, Leung SL, Yu DS. **Medication adherence: is it a hidden drug-related problem in hidden elderly?**. *Geriatr Gerontol Int* (2013.0) **13** 978-85. DOI: 10.1111/ggi.12042 13. Crentsil V, Ricks MO, Xue QL, Fried LP. **A pharmacoepidemiologic study of community-dwelling, disabled older women: factors associated with medication use**. *Am J Geriatr Pharmacother* (2010.0) **8** 215-24. DOI: 10.1016/j.amjopharm.2010.06.003 14. Jyrkkä J, Enlund H, Lavikainen P, Sulkava R, Hartikainen S. **Association of polypharmacy with nutritional status, functional ability and cognitive capacity over a three-year period in an elderly population**. *Pharmacoepidemiol Drug Saf* (2011.0) **20** 514-22. DOI: 10.1002/pds.2116 15. Li SJ, Hwang HF, Yu WY, Lin MR. **Potentially inappropriate medication use, polypharmacy, and falls among hospitalized patients**. *Geriatr Gerontol Int* (2022.0) **22** 857-64. DOI: 10.1111/ggi.14473 16. 16.Medication safety in Polyharmacy. ; 2019. Accessed 2019 Nov. https://apps.who.int/iris/bitstream/handle/10665/325454/WHO-UHC-SDS-2019.11-eng.pdf. 17. Sato K, Inagaki R, Michikawa T, Kawabata S, Ito K, Morita M. **Prescription drug survey of elderly patients with degenerative musculoskeletal disorders**. *Geriatr Gerontol Int* (2022.0) **22** 121-6. DOI: 10.1111/ggi.14326 18. Ferreira Mota Colares SP, Colares GM, Freire de Carvalho J, Ewerton Maia Rodrigues C. **High prevalence of metabolic syndrome in patients with lumbar spinal stenosis: association with diabetes and elevated body mass index**. *Isr Med Assoc J* (2021.0) **23** 766-72. PMID: 34954914 19. Molist-Brunet N, Sevilla-Sánchez D, Puigoriol-Juvanteny E, Barneto-Soto M, González-Bueno J, Espaulella-Panicot J. **Improving individualized prescription in patients with multimorbidity through medication review**. *BMC Geriatr* (2022.0) **22** 417. DOI: 10.1186/s12877-022-03107-2 20. Lee BH, Kim TH, Park MS, Lim S, Park JO, Kim HS. **Comparison of effects of nonoperative treatment and decompression surgery on risk of patients with lumbar spinal stenosis falling: evaluation with functional mobility tests**. *J Bone Joint Surg Am* (2014.0) **96** e110. DOI: 10.2106/JBJS.M.00427 21. Ozaki M, Fujita N, Miyamoto A, Suzuki S, Tsuji O, Nagoshi N. **Impact of knee osteoarthritis on surgical outcomes of lumbar spinal canal stenosis**. *J Neurosurg Spine* (2019.0) **27** 1-6 22. Yamagishi K, Iso H. **The criteria for metabolic syndrome and the national health screening and education system in Japan**. *Epidemiol Health* (2017.0) **39** e2017003. DOI: 10.4178/epih.e2017003 23. Rothschild JM, Bates DW, Leape LL. **Preventable medical injuries in older patients**. *Arch Intern Med* (2000.0) **160** 2717-28. DOI: 10.1001/archinte.160.18.2717 24. Kojima T, Mizukami K, Tomita N, Arai H, Ohrui T, Eto M. **Screening tool for older persons’ appropriate prescriptions for japanese: report of the Japan Geriatrics Society Working Group on “Guidelines for medical treatment and its safety in the elderly"**. *Geriatr Gerontol Int* (2016.0) **16** 983-1001. DOI: 10.1111/ggi.12890 25. Fritsch CG, Ferreira ML, Maher CG, Herbert RD, Pinto RZ, Koes B. **The clinical course of pain and disability following surgery for spinal stenosis: a systematic review and meta-analysis of cohort studies**. *Eur Spine J* (2017.0) **26** 324-35. DOI: 10.1007/s00586-016-4668-0 26. Prasad BC, Ramesh Chandra VV, Devi BV, Chivukula SS, Pundarikakshaiah K. **Clinical, radiological, and functional evaluation of surgical treatment in degenerative lumbar canal stenosis**. *Neurol India* (2016.0) **64** 677-83. DOI: 10.4103/0028-3886.185378 27. Gates M, Tang AR, Godil SS, Devin CJ, McGirt MJ, Zuckerman SL. **Defining the relative utility of lumbar spine surgery: a systematic literature review of common surgical procedures and their impact on health states**. *J Clin Neurosci* (2021.0) **93** 160-7. DOI: 10.1016/j.jocn.2021.09.003 28. Carreon LY, Jespersen AB, Støttrup CC, Hansen KH, Andersen MO. **Is the hospital anxiety and depression scale associated with outcomes after lumbar spine surgery?**. *Global Spine J* (2020.0) **10** 266-71. DOI: 10.1177/2192568219845662 29. Fujita N, Michikawa T, Miyamoto A, Sakurai A, Otaka Y, Suzuki S. **Lumbar spinal surgery improves locomotive syndrome in elderly patients with lumbar spinal canal stenosis: a multicenter prospective study**. *J Orthop Sci* (2020.0) **25** 213-8. DOI: 10.1016/j.jos.2019.03.017 30. 30.Halli-Tierney AD, Scarbrough C, Carroll D. Polypharmacy: Evaluating Risks and Deprescribing. Am Fam Physician. 2019 Jul 1;100(1):32 – 8. 31. Hutchings A, Neuburger J, Grosse Frie K, Black N, van der Meulen J. **Factors associated with non-response in routine use of patient reported outcome measures after elective surgery in England**. *Health Qual Life Outcomes* (2012.0) **10** 34. DOI: 10.1186/1477-7525-10-34
--- title: Protocol for the SEHNeCa randomised clinical trial assesing Supervised Exercise for Head and Neck Cancer patients authors: - M. Rodriguez-Arietaleanizbeaskoa - E Mojas Ereño - MS Arietaleanizbeaskoa - G. Grandes - A Rodríguez Sánchez - V. Urquijo - I Hernando Alday - M. Dublang - G. Angulo-Garay - J. Cacicedo - Mario Rodriguez-Arietaleanizbeaskoa - Mario Rodriguez-Arietaleanizbeaskoa - Egoitz Mojas Ereño - Maria S. Arietaleanizbeaskoa - Gonzalo Grandes - Arturo Garcia-Alvarez - Aitor Coca - Nere Mendizabal - Olga del Hoyo - Javier García-Escobedo - Ángel Rodríguez Sánchez - Lucía Flores Barrenechea - Rebeca Sánchez - Virginia Urquijo - Luis Barbier Herrero - Goiztidi Díaz-Basterra - Javier Gómez-Suarez - Laura A Calles Romero - Natalia C. Iglesias-Hernandez - Iñigo Hernando Alday - Maddalen Dublang - Lina M. Ramirez-Garcia - Garazi Angulo-Garay - Silvia Dominguez-Martinez - Erreka Gil-Rey - Aitor Martinez-Aguirre - Borja Gutierrez-Santamaria - Jon Cacicedo journal: BMC Cancer year: 2023 pmcid: PMC10037879 doi: 10.1186/s12885-023-10718-4 license: CC BY 4.0 --- # Protocol for the SEHNeCa randomised clinical trial assesing Supervised Exercise for Head and Neck Cancer patients ## Abstract ### Objectives To evaluate the effectiveness of an innovative supervised exercise programme to mitigate the loss of lean body mass, functional capacity and quality of life in people with head and neck cancer, as well as to identify the optimal moment to apply it, before or after radiotherapy treatment, compared with the prescription of a physical activity plan carried out autonomously. ### Methods Patients with squamous cell carcinoma of the head and neck ($$n = 144$$), treated with radiotherapy, will be randomly assigned to one of 3 comparison groups: pre-radiotherapy supervised exercise, post-radiotherapy supervised exercise and autonomous exercise, stratifying by human papillomavirus infection and previous surgery. The exercise programme will be carried out in 36 sessions over 12 weeks, combining moderate and high intensity strength and aerobic exercises. The main outcome variable is the change in lean body mass at 6 months measured by bioimpedance, while secondary variables are functional capacity, symptoms, quality of life and adverse effects. *Longitudinal* generalised mixed models will be used for the analyses of the repeated measurements at 3, 6, and 12 months after baseline. ### Conclusions The pilot study supports the feasibility and safety of the project. However, as the programme progressed, attendance at the sessions decreased. Strategies will be necessary for increasing attendance, as well as involving the patient in their recovery and other incentives. Follow-up after treatment to assess acute/late toxicity will enable us to know the response to both the exercise programme and its adherence. ### Trial registration NCT04658706 Date and version identifier: March 1, 2023. Version 1.1 ## Introduction Head and neck cancer (HNC) was the seventh most frequently diagnosed cancer worldwide in 2020 [1]. The main risk factors that cause this type of cancer to develop are smoking, alcoholism, diet, sedentary lifestyle, obesity and human papillomavirus (HPV) infection, among others [2]. This unhealthy background added to both the disease and its treatments effects tremendously reduce cardiorespiratory capacity, lean body mass (through accelerated processes of cachexia and sarcopenia) bone mass, and provoke psychosocial distress, depression, pain, fatigue, and impaired cognitive function [3]. Between 30 and $50\%$ of HNC patients have malnutrition and it has been estimated that they may experience 5–$10\%$ weight loss during radiotherapy treatment [4, 5]. Weight loss of more than $5\%$ has been associated with a decrease in overall and specific survival for head and neck cancer. In addition, it is estimated that more than $70\%$ of weight loss can be attributed to loss of lean body mass, which leads to a decrease in strength, functional capacity and quality of life [6]. Because of that lean body mass is considered an appropriate and feasible end point in clinical trials of HNC. Additionally, sarcopenia before and after treatment is an independent factor of poor prognosis associated with decreased survival in patients undergoing radiotherapy with curative intent [7]. Moreover, sarcopenia has recently been associated with increased acute toxicity (grade ≥ 3 dysphagia) [8] and decreased progression-free survival [9]. For this reason, it is deemed appropriate to perform body composition measurements as part of the diagnosis and evaluation of interventions for HNC [10–12]. Nutritional approach is a fundamental part of support for these patients. Nutritional interventions through oral supplements and dietary advice have shown short-term benefit in patients with HNC [13]. Patients with head and neck tumours represent a group with a high prevalence of malnutrition and with very specific side effects compared to other types of tumours, which can cause loss of lean body mass [8]. Most studies have explored the impact of either nutritional supplements or physical exercise (PE) separately. Therefore, it seems necessary to explore the benefit of nutritional intervention and PE jointly in these patients and their impact on lean body mass. Up to now, randomised studies that have assessed the impact of a PE programme in these patients are scarce [14–17], and have limitations in their applicability since they have not included a control arm or the PE has not been supervised [14, 15], nor performed using a progressive overload programme to optimise lean body mass gain [16]. The most recent Head and Neck Cancer Survivorship Consensus guidelines establish that further research is needed on the implementation of physical activity promotion in routine clinical care and the optimal timing, intensity and duration of exercise intervention in relation to oncological treatment, as well as research on the links between exercise and HNC recurrence or survival [18]. In this context, this study aims to evaluate the effectiveness of an innovative supervised PE programme to mitigate the loss of lean body mass, improve functional capacity and quality of life, which are dramatically reduced in people diagnosed with head and neck cancer; to compare it with the prescription of PE carried out independently; and to identify the optimal moment to apply this programme: before (prehabilitation) or after radiotherapy treatment. ## Trial design The SEHNeCa project is a multicentre, randomised, controlled clinical trial with three arms. Patients are recruited by the radiation oncologist at the patient’s first visit to the radiation oncology department, where they are asked to participate in the programme. Four assessments are made of all the participants, regardless of the group they belong to, at 0, 3, 6 and 12 months. The baseline assessment will be performed during the same week as recruitment. After randomisation, the pre-radiotherapy group will start the supervised PE programme (at least 7–10 days before starting radiotherapy). The control and post-radiotherapy groups will initially receive physical activity guidelines that they can perform independently. The post-radiotherapy group will start supervised training 4–6 weeks after finishing radiotherapy and the control group will not have supervised training. All the patients will receive follow-up by the endocrinology service (before radiotherapy, during radiotherapy, at the end of radiotherapy and one month, 3 months, 6 months and 12 months after the end of radiotherapy) in which they will receive screening and assessment of nutritional status as well as individualised nutritional intervention following international guidelines [19], which may include nutritional advice, oral nutritional supplementation, and/or insertion of a feeding tube as required by each patient (Table 1) (Fig. 1). Table 1SEHNeCa Programme timelineFollow upD0W1W2W3W4M2M3M4M5M6M7M12M18Diagnosis and recruitmentXNutritional evaluationXXXXXRadiation oncology evaluationXXXXXXRadiotherapyXMeasurementsXXXXSupervised Exercise InterventionPRE-RADIOTHERAPY GROUPPOST-RADIOTHERAPY GROUPAutonomous Exercise InterventionControl Group and Post-radiotherapy GroupPre-radiotherapy Group and Control GroupD Day, W Week, M MonthFig. 1Flow diagram. PVS Prescribe Vida Saludable To achieve both these objectives, the radiotherapy-oncology and endocrinology departments of Basurto University Hospital, Cruces University Hospital and Galdakao Hospital have collaborated. ## Participants Inclusion and exclusion criteria (Table 2).Table 2Inclusion/Exclusion criteriaInclusion criteriaExclusion criteriaDiagnosis of carcinoma of the larynx (except T1-T2 N0 of the glottis), pharynx, oral cavity, salivary gland, paranasal sinuses, or cervical metastasis of unknown origin in stages I-II-III-IVa-IVbDecompensated heart disease, uncontrolled hypertension (SBP > 200 or DAT > 110), heart failure (NYHA > I), or constrictive pericarditisRadiotherapy (with or without concomitant chemotherapy) preceded or not by surgical treatment and performed with radical intentMedical contraindication against exercise (such as bone fractures, tendon ruptures, spondylolisthesis > $25\%$ with neurological compromise, rhabdomyolysis, herniated disc in the spine with neurological symptoms, cardiovascular comorbidities such as unstable angina, current myocarditis or thrombophlebitis); Neutropenia (< 500 mm3), severe anaemia (Hb < 8.0 g/dl), platelet count < 50.000 microL > 18 yearsRadiotherapy with palliative intentBMI > 18.5PregnancyECOG 0–1 or Karnofsky Index ≥ $80\%$Regular physical activity (150 min/week of moderate activity or 75 min/week of vigorous activity, including two or more weekly strength training sessions), as measured by the PVS questionnaireSigned consent formCarcinoma of the glottis (T1-T2, N0)T(1–2) Tumour (TNM classification), N[0] Node (TNM classification), SBP systolic blood pressure, DAT Diastolic blood pressure, NYHA New York Heart Association, Hb Haemoglobin, BMI Body Mass Index, ECOG Eastern Cooperative Oncology Group, PVS Prescribe Healthy Life ## Randomisation Randomisation is carried out on a central basis through the coordinating committee. This will be made up of the organization responsible for the study’s management and the researchers. The committee will be blind to the assignment of patients in the comparison groups and will convene in the Primary Health Research Unit of Bizkaia. The 120 patients included in the study will be randomly assigned and consequently registered and attached to one of the three study groups in a 1:1:1 ratio, through randomisation by random blocks of 3 or 6 patients. Patients are stratified according to HPV status and previous surgery yes/no. ## Evaluation Evaluation of the SEHNeCa exercise programme involves measuring the safety, adherence, and effectiveness of the programme. These analyses incorporate elements of the reach, effectiveness, adoption, implementation, and maintenance (RE-AIM) framework [20]. ## Safety The incidence and severity of any adverse events (e.g. falls or muscle strains) that occur during the health-centre based sessions will be monitored and reported by the supervising exercise physiologist/nurse using programme-specific documentation. ## Adherence Attendance at health-centre based exercise sessions and the reason for any missed sessions will be tracked throughout the programme. Further, completion of assessments at pre-programme and post-programme time points as well as follow-up questionnaires will be reported on. ## Effectiveness: primary study outcomes Change in Lean Body Mass (kg), measured by using an Inbody 770 bioimpedance analyser (In-body, Seoul, Korea). ## Anthropometry and body composition Height will be measured using a wall stadiometer, and body composition with a bioimpedance analyser. BMI (body mass index) (kg/m2), skeletal muscle mass (kg), fat free mass (kg); bone mineral content (kg), body fat mass (kg), body fat percentage (%), intra and extracellular water (l); extracellular water/total body water, visceral fat (cm2), basal metabolic rate (kcal) and phase angle (º) will be registered. Abdominal and waist circumference will be taken [21]. ## Physical function Physical function is assessed by 400-m walk tests in a 20-m corridor in a controlled environment (temperature, 20-21º C; relative humidity, 50–$55\%$; barometric pressure, 755–765 mmHg). The participants will be seated for measuring heart rate and blood pressure before and after the test. The repeat chair stand test (5 times sit to stand test), the handgrip dynamometry test and the Timed Up and Go (TUG) test will be carried out. Maximal strength in the upper and lower body will be measured in terms of the 5-repetition maximum (5RM) test (the maximum load that can be lifted five times) in chest press, seated rowing and leg extension, respectively. These assessments are to be conducted by an independent exercise physiologist not involved with administering the exercise intervention. ## Patient-reported outcomes A series of questionnaires with sound psychometric properties are to be used to assess general health and cancer-specific quality of life at baseline and 3–6-12 months thereafter. The Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36) will be used to assess general health-related quality of life status. Cancer-specific quality of life is evaluated by the core quality of life (QLQ-C-30) questionnaire developed by the European Organization for Research and Treatment of Cancer (EORTC). The Head and Neck cancer-specific questionnaire (H&N-35) is a supplementary module of the QLQ-C-30 questionnaire that includes 35 head and neck cancer-specific items. Finally, the Spanish version of the Shoulder and Pain Disability Index (SPADI) is a self-administered questionnaire that covers two aspects, one for pain and the other for functional activities, with lower scores representing less pain or disability. ## Supervised exercise intervention The programme operates throughout the year and is a free 12-week, small group exercise programme supervised by specially trained instructors including exercise physiologists, physiotherapists and nurses. Participants are required to participate in supervised exercise sessions three times a week in the exercise laboratories, (primary health-care centres and Cruces University Hospital/Biocruces Bizkaia Health Research Institute). Twice a week, the sessions include a combination of strength exercises and moderate-to-high intensity aerobic exercise, and they last approximately 1–1¼ hours. The third session does not include aerobic exercise, and lasts 30 to 45 min. Strength exercises are conducted with free-weights (to provide a quantifiable measurement of progressive overload), elastic bands and suspension training, with the intention that the patients should learn how to train on their own. [ 22]. The strength exercises target five of the major upper and lower body muscle groups, as well as core exercises. Progression from static exercises towards more dynamic exercises is encouraged, aiming to activate more muscle mass. Among the main parameters that can be modified, the actual number of repetitions performed in a set, in relation to the maximum number that can be completed (i.e., proximity to muscle failure), recently called “level of effort”, will be used to individualise the strength exercise intensity, maximise the suitability of the exercise for each patient, and optimise induced neuromuscular fatigue [23, 24].The target intensity is adjusted from 9 to 12 repetitions out of the 18 repetitions that could be completed (written as 9–12 [18]), which is equivalent to ~ $60\%$ of 1 repetition maximum (1RM) using 2 sets during the first 4 weeks, to 4 sets of 7–10 [14], which is ~ $70\%$ of 1RM in the last 4 weeks of the programme. The aerobic exercise component includes 20–25 min of at least moderate intensity for 8-min periods alternating with 2-min lower-intensity periods during the first month, moving toward higher-intensity 5-min intervals by the third month (Fig. 2). The exercise intensity zones are tailored to each patient by the estimated maximum heart rate using the Eq. ( 220—age) [25] and applying specific intensity boundaries based on the heart rate reserve (HRR), defined as the difference between the resting heart rate and maximum heart rate. % HRR has been adopted by the American College of Sports Medicine as the gold standard for indirect assessment of exercise intensity [26, 27]. Every exercise session will be monitored with a heart rate monitor, teaching the patients to self-manage their exercise sessions with respect to the prescribed target intensities; the target intensity is between 40 and $85\%$ of HRR [26, 27]. The perceived level of effort is to be recorded using the Borg Rating of Perceived Exertion (RPE) scale from 0 (rest) to 10 (maximal effort) [28], with the target intensity progressively increasing from 3 to 5–6 points (Table 3).Fig. 2SEHNeCa exercise programme. M1: Low to moderate intensity; M2: moderate intensity; M3: intensive moderate intensity; 1RM: 1 maximum repetitionTable 3Exercise intensity categoriesZonea%HRRbRPEcTraining typeHigh intensity training > 85 > 5.5High intensity training. The patient nears exhaustion and is no longer in a steady stateM360–844.5–5.5Moderate to intensive training. The patient has difficulties in talking and sweating increasesM240–593–4Moderate intensity. The patient notices an increased respiratory rateM120–391–2.5Low to moderate intensity. The patient does not notice any increase in the respiratory rateLow-intensity training < 20 < 1Low intensity. This involves daily activities requiring low levels of effortaThe intensity zones are based on the estimated maximum heart rate and applying specific intensity boundaries based on HRR [43,44]bHRR: heart rate reserve. Maximum heart rate—resting heart ratecRPE: rating of perceived exertion. Determined using the Borg scale from 0 to 10 [45] ## Autonomous exercise intervention Every patient in the control group will receive the corresponding standard treatment (i.e. radiotherapy), determined by the regular oncological consensus committee. In addition, they will receive a prescription concerning healthy lifestyles and a personalised exercise plan, together with an evaluation and individualised nutritional intervention by a physician at the Endocrinology and Nutrition department in Cruces, Basurto or Galdakao Hospitals, in the same way as for intervention groups. They will also be evaluated in the radiation oncology department with the routine follow-up (end of radiotherapy, one, three, six and twelve months after radiotherapy) to assess outcomes and toxicity due to oncological treatment according to Radiation Therapy Oncology Group and Common Terminology Criteria for Adverse Events (vs 5.0) toxicity criteria. ## Statistical analysis Differences between treatment groups in changes in outcome variables will be estimated on an intention-to-treat basis. We will use linear mixed models for longitudinal analysis of repeated measurements of continuous outcomes (SAS PROC MIXED) and generalised logistic mixed models for dichotomous outcomes (SAS PROC GLIMMIX), considering intercept and time courses as random effects and testing the significance of the interaction of time slopes for each treatment group. No imputation method will be used to handle missing data, as longitudinal mixed-effect models based on maximum likelihood estimation are more appropriate for handling missing data than common imputation methods such as the last observation carried forward, complete case analysis, or other possible forms of imputation [29]. A sample size of 40 patients per group (120 in total) will give $80\%$ power to the study to detect as significant (0.05) the difference at six months in a mean lean body mass loss of $25\%$ between the control and rehabilitation groups, and of $12\%$ between the rehabilitation and prehabilitation groups, corresponding to an estimated mean lean body mass loss from baseline of 4.8 kg ($9.1\%$) in the control group, 3.6 kg ($6.8\%$) in the rehabilitation group, and 3.2 kg ($6\%$) in the prehabilitation group (standard deviation 2,4 kg). With an anticipated drop-out rate of $20\%$, we plan to enrol 144 patients (48 per group). During these first 12 months of recruitment, the radiation oncology departments of Basurto and Cruces University Hospitals have attended 161 eligible patients, 112 met the inclusion criteria, of whom $63.4\%$ agreed to participate in the study. Therefore, we expect that the recruitment period will last 24 months. ## Quality control To ensure the quality of the study data, maximise the validity and reliability of the programme, and accurate measurement of the variables, we will take the following steps:Produce documents for the study, including operational manuals for fieldwork and forms for registering measurements and details of the intervention. Store all documentation (informed consent forms, documents containing results, etc.) in locked cabinets and on a database on a secure server. Hold regular meetings for quality control and write up progress reports every 3 months. The coordinator will contact the health centres daily, and report to the principal researcher every week. ## Discussion The SEHNeCa programme aims to increase knowledge about the effectiveness and optimal start time of a supervised PE programme specifically targeted at patients with head and neck cancer, administered before, during or after radiotherapy, carried out at the primary care facilities and Cruces University Hospital. The results obtained in the pilot study demonstrate the feasibility and safety of the project, showing that it is achievable in this type of patient. Additionally, through the pilot, good compliance with the circuit (radiation-oncologists, endocrinologists, evaluators) has been verified prior to the start of the training sessions, showing it to be feasible and achievable within the established deadlines. The pilot was only carried out with the group that performed PE prior to starting the radiotherapy treatment. This is due to the timing of the entire process, since patients must start the PE at least 7–10 days before the start of radiotherapy. However, as the treatment progressed, there was a decrease in attendance at the PE sessions, mainly caused by the effects of the treatment. For this reason, it is necessary to verify the effectiveness with a larger sample, as well as the different effects of a PE programme before and during radiotherapy. Moreover, both prehabilitation and rehabilitation programmes are limited by a lack of resources, awareness of the benefits of PE, and a lack of physical activity experts in cancer units [22]. For this reason, it is essential to have PE laboratories, with qualified professionals, in hospitals and outpatient centres, as proposed by the SEHNeCa project, so that this type of programme is more accessible to the entire targeted population and implantable in daily clinical practice. ## Strengths of the study The main strength of the study is that it is a randomised controlled clinical trial, and therefore it includes a control group with which to compare the intervention groups. In this way it will be possible to determine the effects of the SEHNeCa programme more precisely. Furthermore, in the randomised clinical trials carried out to date, no group performs the PE intervention prior to the start of treatment. Therefore, the data resulting from the study can contribute to pinpointing the optimal moment for implementation of PE during the treatment process. Follow-up after treatment to assess acute/late toxicity will allow us to know the response to both the exercise programme and its adherence. 71 patients have been included in the study since the beginning of project recruitment in January 2022. ## Limitations of the study and contingency plans One possible limitation could be the distance of the PE laboratory for some participants. Ideally, patients should be able to complete the supervised PE programme at their nearest health centre. Although we cannot definitely count on such resources, we might be able to offer availability in several centres so that they can use the laboratory of their choice. One of the challenges we face is to improve patient participation and adherence to the programme. To do this, we will try to involve the patient in their recovery, and personalise the programme. Likewise, we consider it important to seek incentives, such as patients receiving feedback on their progress in the programme. Finally, the circuit created to attract, evaluate and include the participants in the project is complex, since we have very little time to carry out the initial evaluations in radio-oncology, endocrinology and functional assessments, which could be another limitation. However, the supervision of the participants is continuous and comprehensive, also providing a large amount of data that can help the scientific dissemination of the project and for the design and implementation of new PE programmes. ## Ethical considerations This study protocol complies with the Declaration of Helsinki and its amendments, as well as with good clinical practice. The Ethics Committee of the Basque Country approved the study in the health centres, ensuring it would be implemented in compliance with the established regulations. Regarding data confidentiality, only the study researchers have access to the data of the individuals who agree to participate in the study, in compliance with the Organic Act $\frac{03}{2018}$ of December 2018, on the protection of personal data. ## References 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A. **Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries**. *CA: Cancer J Clin* (2021.0) **71** 209-49. PMID: 33538338 2. 2.SEOM. Las cifras del cáncer en España 2022 [Internet]. 2022. Available from: https://seom.org/images/LAS_CIFRAS_DEL_CANCER_EN_ESPANA_2022.pdf 3. Schmitz KH, Courneya KS, Matthews C, Demark-Wahnefried W, Galvão DA, Pinto BM. **American college of sports medicine roundtable on exercise guidelines for cancer survivors**. *Med Sci Sports Exerc* (2010.0) **42** 1409-1426. DOI: 10.1249/MSS.0b013e3181e0c112 4. Langius JAE, Bakker S, Rietveld DHF, Kruizenga HM, Langendijk JA, Weijs PJM. **Critical weight loss is a major prognostic indicator for disease-specific survival in patients with head and neck cancer receiving radiotherapy**. *Br J Cancer* (2013.0) **109** 1093-9. DOI: 10.1038/bjc.2013.458 5. Datema FR, Ferrier MB, de Baatenburg Jong RJ. **Impact of severe malnutrition on short-term mortality and overall survival in head and neck cancer**. *Oral Oncol* (2011.0) **47** 910-4. DOI: 10.1016/j.oraloncology.2011.06.510 6. Jager-Wittenaar H, Dijkstra PU, Vissink A, Langendijk JA, van der Laan BFAM, Pruim J. **Changes in nutritional status and dietary intake during and after head and neck cancer treatment**. *Head Neck* (2011.0) **33** 863-870. DOI: 10.1002/hed.21546 7. Findlay M, White K, Stapleton N, Bauer J. **Is sarcopenia a predictor of prognosis for patients undergoing radiotherapy for head and neck cancer? A meta-analysis**. *Clin Nutr* (2021.0) **40** 1711-1718. DOI: 10.1016/j.clnu.2020.09.017 8. Karavolia E, van Rijn-Dekker MI, Van den Bosch L, van den Hoek JGM, Oldehinkel E, Meijer TWH. **Impact of sarcopenia on acute radiation-induced toxicity in head and neck cancer patients**. *Radiother Oncol* (2022.0) **170** 122-128. DOI: 10.1016/j.radonc.2022.03.009 9. Nagpal P, Pruthi DS, Pandey M, Yadav A, Singh H. **Impact of sarcopenia in locally advanced head and neck cancer treated with chemoradiation: an Indian tertiary care hospital experience**. *Oral Oncol* (2021.0) **121** 105483. DOI: 10.1016/j.oraloncology.2021.105483 10. de van der Bokhorst-van Schuer Null, van der Blomberg-von Flier BM, Kuik DJ, Scholten PE, Siroen MP, Snow GB. **Survival of malnourished head and neck cancer patients can be predicted by human leukocyte antigen-DR expression and interleukin-6/tumor necrosis factor alpha response of the monocyte**. *JPEN J Parenter Enteral Nutr.* (2000.0) **24** 329-36. DOI: 10.1177/0148607100024006329 11. Platek ME, Reid ME, Wilding GE, Jaggernauth W, Rigual NR, Hicks WL. **Pretreatment nutritional status and locoregional failure of patients with head and neck cancer undergoing definitive concurrent chemoradiation therapy**. *Head Neck* (2011.0) **33** 1561-1568. DOI: 10.1002/hed.21640 12. 12.Head and Neck Guideline Steering Committee. Evidence-based practice guidelines for the nutritional management of adult patients with head and neck cancer. Sydney: Clinical Oncological Society of Australia. [Version URL: https://wiki.cancer.org.au/australiawiki/index.php?oldid=215353, cited 2023 Mar 15]. Available from: https://wiki.cancer.org.au/australia/COSA:Head_and_neck_cancer_nutrition_guidelines 13. Ravasco P, Monteiro-Grillo I, Marques Vidal P, Camilo ME. **Impact of nutrition on outcome: a prospective randomized controlled trial in patients with head and neck cancer undergoing radiotherapy**. *Head Neck* (2005.0) **27** 659-668. DOI: 10.1002/hed.20221 14. Lønbro S, Dalgas U, Primdahl H, Overgaard J, Overgaard K. **Feasibility and efficacy of progressive resistance training and dietary supplements in radiotherapy treated head and neck cancer patients–the DAHANCA 25A study**. *Acta Oncol* (2013.0) **52** 310-318. DOI: 10.3109/0284186X.2012.741325 15. Lønbro S, Dalgas U, Primdahl H, Johansen J, Nielsen JL, Aagaard P. **Progressive resistance training rebuilds lean body mass in head and neck cancer patients after radiotherapy–results from the randomized DAHANCA 25B trial**. *Radiother Oncol* (2013.0) **108** 314-319. DOI: 10.1016/j.radonc.2013.07.002 16. Capozzi LC, McNeely ML, Lau HY, Reimer RA, Giese-Davis J, Fung TS. **Patient-reported outcomes, body composition, and nutrition status in patients with head and neck cancer: Results from an exploratory randomized controlled exercise trial**. *Cancer* (2016.0) **122** 1185-1200. DOI: 10.1002/cncr.29863 17. Sandmael JA, Bye A, Solheim TS, Stene GB, Thorsen L, Kaasa S. **Feasibility and preliminary effects of resistance training and nutritional supplements during versus after radiotherapy in patients with head and neck cancer: a pilot randomized trial**. *Cancer* (2017.0) **123** 4440-4448. DOI: 10.1002/cncr.30901 18. Goyal N, Day A, Epstein J, Goodman J, Graboyes E, Jalisi S. **Head and neck cancer survivorship consensus statement from the American head and neck society**. *Laryngoscope Invest Otolaryngol* (2022.0) **7** 70-92. DOI: 10.1002/lio2.702 19. Muscaritoli M, Arends J, Bachmann P, Baracos V, Barthelemy N, Bertz H. **ESPEN practical guideline: Clin Nutr Cancer**. *Clin Nutr* (2021.0) **40** 2898-2913. DOI: 10.1016/j.clnu.2021.02.005 20. Glasgow RE, Goldstein MG, Ockene JK, Pronk NP. **Translating what we have learned into practice. Principles and hypotheses for interventions addressing multiple behaviors in primary care**. *Am J Prev Med* (2004.0) **27** 88-101. DOI: 10.1016/j.amepre.2004.04.019 21. 21.Norton, Kevin. Standards for anthropometry assessment Chapter 4 in Kinanthropometry and Exercise Physiology 4th ed, 2018. pp 68–137. 22. Landi F, Marzetti E, Martone AM, Bernabei R, Onder G. **Exercise as a remedy for sarcopenia**. *Curr Opin Clin Nutr Metab Care* (2013.0) **1** 25 23. Gorostiaga EM, Navarro-Amézqueta I, Calbet JAL, Sánchez-Medina L, Cusso R, Guerrero M. **Blood ammonia and lactate as markers of muscle metabolites during leg press exercise**. *J Strength Cond Res* (2014.0) **28** 2775-2785. DOI: 10.1519/JSC.0000000000000496 24. González-Badillo JJ, Rodríguez-Rosell D, Sánchez-Medina L, Ribas J, López-López C, Mora-Custodio R. **Short-term recovery following resistance exercise leading or not to failure**. *Int J Sports Med* (2016.0) **37** 295-304. PMID: 26667923 25. Gellish RL, Goslin BR, Olson RE, McDonald A, Russi GD, Moudgil VK. **Longitudinal modeling of the relationship between age and maximal heart rate**. *Med Sci Sports Exerc* (2007.0) **39** 822-829. DOI: 10.1097/mss.0b013e31803349c6 26. Mezzani A, Hamm LF, Jones AM, McBride PE, Moholdt T, Stone JA. **Aerobic exercise intensity assessment and prescription in cardiac rehabilitation: a joint position statement of the European association for cardiovascular prevention and rehabilitation, the American association of cardiovascular and pulmonary rehabilitation and the Canadian association of cardiac rehabilitation**. *Eur J Prev Cardiol* (2013.0) **20** 442-467. DOI: 10.1177/2047487312460484 27. Gil-Rey E, Quevedo-Jerez K, Maldonado-Martin S, Herrero-Román F. **Exercise intensity guidelines for cancer survivors: a comparison with reference values**. *Int J Sports Med* (2014.0) **35** e1-9. DOI: 10.1055/s-0034-1389972 28. Borg G, Ljunggren G, Ceci R. **The increase of perceived exertion, aches and pain in the legs, heart rate and blood lactate during exercise on a bicycle ergometer**. *Eur J Appl Physiol Occup Physiol* (1985.0) **54** 343-349. DOI: 10.1007/BF02337176 29. Wood AM, White IR, Hillsdon M, Carpenter J. **Comparison of imputation and modelling methods in the analysis of a physical activity trial with missing outcomes**. *Int J Epidemiol* (2005.0) **34** 89-99. DOI: 10.1093/ije/dyh297
--- title: Rational design of GDP‑d‑mannose mannosyl hydrolase for microbial l‑fucose production authors: - Cong Fu - Xuexia Xu - Yukang Xie - Yufei Liu - Min Liu - Ai Chen - Jenny M. Blamey - Jiping Shi - Suwen Zhao - Junsong Sun journal: Microbial Cell Factories year: 2023 pmcid: PMC10037897 doi: 10.1186/s12934-023-02060-y license: CC BY 4.0 --- # Rational design of GDP‑d‑mannose mannosyl hydrolase for microbial l‑fucose production ## Abstract ### Background l‑*Fucose is* a rare sugar that has beneficial biological activities, and its industrial production is mainly achieved with brown algae through acidic/enzymatic fucoidan hydrolysis and a cumbersome purification process. Fucoidan is synthesized through the condensation of a key substance, guanosine 5′‑diphosphate (GDP)‑l‑fucose. Therefore, a more direct approach for biomanufacturing l‑fucose could be the enzymatic degradation of GDP‑l‑fucose. However, no native enzyme is known to efficiently catalyze this reaction. Therefore, it would be a feasible solution to engineering an enzyme with similar function to hydrolyze GDP‑l‑fucose. ### Results Herein, we constructed a de novo l‑fucose synthetic route in *Bacillus subtilis* by introducing heterologous GDP‑l‑fucose synthesis pathway and engineering GDP‑mannose mannosyl hydrolase (WcaH). WcaH displays a high binding affinity but low catalytic activity for GDP‑l‑fucose, therefore, a substrate simulation‑based structural analysis of the catalytic center was employed for the rational design and mutagenesis of selected positions on WcaH to enhance its GDP‑l‑fucose‑splitting efficiency. Enzyme mutants were evaluated in vivo by inserting them into an artificial metabolic pathway that enabled B. subtilis to yield l‑fucose. WcaHR36Y/N38R was found to produce 1.6 g/L l‑fucose during shake‑flask growth, which was $67.3\%$ higher than that achieved by wild‑type WcaH. The accumulated l‑fucose concentration in a 5 L bioreactor reached 6.4 g/L. ### Conclusions In this study, we established a novel microbial engineering platform for the fermentation production of l‑fucose. Additionally, we found an efficient GDP‑mannose mannosyl hydrolase mutant for L‑fucose biosynthesis that directly hydrolyzes GDP‑l‑fucose. The engineered strain system established in this study is expected to provide new solutions for l‑fucose or its high value‑added derivatives production. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12934-023-02060-y. ## Introduction L‑Fucose (6‑deoxy‑l‑galactose) is an indispensable component of mammalian glycans and glycolipids, and the O‑fucosylation of glycans and glycolipids is biologically important due to its effect on protein function [1]. l‑Fucose also exhibits unique functional properties with regard to metabolic regulation, immunological responses and pathogen resistance. It has been reported that l‑fucose decreased fat accumulation and hepatic triglycerides in mice fed a high‑fat diet, helping to restore healthy intestinal composition and function [2]. l‑*Fucose is* known to suppress pathogen virulence, protecting the host against infection and inflammation [3], and l‑fucose‑containing glycan motifs alter pro‑ and anti‑inflammatory immunoglobulin G activities, which indicates their potential application in immunotherapy [4]. Countless studies on the physiological mechanisms of l‑fucose have led to its application in the treatment of human diseases. Research has found that ∼$20\%$ of the population has a Fut2 defect, which is linked to diseases with potential connections to the gut microbiota [5, 6], while the presence of l‑fucose enhances fitness by supporting beneficial gut bacteria [7]. Interestingly, l‑fucose is also a common constituent of extracellular polysaccharide (EPS) secreted by many gram‑negative pathogens [8]. Oral l‑fucose supplementation has been reported as a promising therapy for congenital glycosylation disorders, improving growth and cognitive skills in children with GFUS‑CDG (GDP‑l‑fucose synthase‑Congenital disorders of glycosylation) [9]. l‑Fucose can also be applied as the substrate for producing 2′‑fucosyllactose (2′‑FL) and 3‑fucosyllactose (3‑FL), two key oligosaccharides in human milk [10, 11]. A recent study reported the chemical synthesis of l‑fucose‑containing colanic acid hexasaccharide [12], and it is believed that other valuable fucosylated glycans can be synthesized from l‑fucose.l‑Fucose exists in the form of sulfated l‑fucose‑rich polysaccharides called fucoidans in brown algae, seaweeds and echinoderms [13]. Industrial‑level l‑fucose production is achieved by laborious and costly extraction and purification processes using brown algae as the crude material [14]. Other routes for fabricating l‑fucose have been developed to meet the increasing demand for this chemical. For example, l‑fucose can be chemically converted from d‑mannose [15], and it can also be extracted from the enzymatic hydrolysate of l‑fucose‐rich microbial EPS [16] or bio converted from rarer monosaccharides, such as l‑fuculose [17]. Synthetic biology has greatly expanded our ability to competitively produce high‑value chemicals through the rational design of metabolic networks and microbial strain engineering [18, 19]. Pipeline design and construction for l‑fucose bioproduction from sustainable carbon sources seems to be a more promising approach. Previously engineered *Escherichia coli* was developed to produce l‑fucose from in situ synthesized 2′‐fucosyllactose (2′‐FL) by recombinant 2′‐fucosyltransferase and α‐l‑fucosidase [20]. As demonstrated in Fig. 1, such a microbial l‑fucose synthetic route, involving the addition and discharge of a lactose molecule, would be redundant if guanosine 5′‑diphosphate (GDP)‑l‑fucose could be directly degraded into GDP and L‑fucose; however, no enzyme has been identified as being capable of such a catalytic reaction. A known enzyme with a similar function is E. coli GDP‑mannose mannosyl hydrolase (GDPMH, also known as WcaH). wcaH is part of an operon encoding enzymes for the synthesis of colanic acid, a l‑fucose‑rich extracellular polysaccharide. WcaH is speculated to regulate the cellular level of GDP‑d‑mannose by catalyzing the hydrolysis of GDP‑D‑mannose to GDP and d‑mannose in the presence of Mg2+ to limit the carbon flux channeled toward GDP‑l‑fucose formation [21]. Interestingly, the acceptable substrates of WcaH are GDP‑d‑mannose, GDP‑d‑glucose and GDP‑l‑fucose, all of which bind to the catalytic center of the enzyme through 12 hydrogen bonds [22], although for GDP‑l‑fucose, the kcat of WcaH is low, with a high Km, suggesting that GDP‑l‑fucose might be a potent competitive inhibitor of GDP‑d‑mannose hydrolysis [23]. Thus, WcaH is potentially a good candidate for the direct hydrolysis of GDP‑l‑fucose to yield l‑fucose (Fig. 1).Fig. 1Schematic for L‑fucose production and application. Previously reported L‑fucose production methods are shown in blue font, and the synthesis method reported in this work is shown in red font. LacY: lactose permease; FucT: 2′‐fucosyltransferase; AFU: α‐L‐fucosidase; WcaH: guanosine 5’‑diphosphate (GDP)‑mannose mannosyl hydrolase Regarding the main applications of l‑fucose in the food or pharmaceutical fields, among the common bacterial hosts, *Bacillus subtilis* is favored over E. coli since the latter produces lipopolysaccharides (LPS) that can trigger aberrant inflammation. B. subtilis is regarded as generally recognized as safe (GRAS) and has become a model chassis host cell for industrial applications due to its robustness and metabolic versatility [24]. A variety of valuable compounds, including small molecular compounds, proteins and biopolymers, have been produced with engineered B. subtilis microbial platforms [25–27]. However, B. subtilis lacks the GDP‑l‑fucose generation pathway that E. coli naturally possesses. To achieve enzymatic l‑fucose production from GDP‑l‑fucose, an E. coli‑derived GDP‑l‑fucose generation pathway has to be transferred into B. subtilis ATCC 6051a, and similar work has been accomplished for microbial 2’‑FL production [28]. To produce l‑fucose with engineered B. subtilis, this work began with rationally designing WcaH mutants based on the docking model of wild‑type WcaH, followed by expressing the GDP‑l‑fucose synthetic pathway together with WcaH or its mutants. This is the first report of microbial l‑fucose production through GDP‑l‑fucose hydrolysis. ## Construction of a de novo pathway for l‑fucose synthesis in B. subtilis The host B. subtilis 6051a (namely, B. subtilis 164) used in this study is a well‑studied chassis strain with a clear genetic background and is easy to cultivate at a high density [29]. In addition, B. subtilis is generally recognized as safe (GRAS) by the FDA and is suitable for the production of food processing enzymes or food additives [26, 30–32]. The pyrogen‑free nature of B. subtilis also offers advantages in product separation and purification. To assess whether B. subtilis was able to accumulate l‑fucose during cultivation, a B. subtilis growth experiment was performed. As shown in Additional file 1: Fig. S1, no colony was formed when cells were plated on solid medium containing l‑fucose as the sole carbon source, indicating that B. subtilis 164 lacks the capability to catabolize l‑fucose. Next, to facilitate the convenient genetic manipulation of B. subtilis 164, we improved the transformation efficiency by integrating comk (B. subtilis endogenous gene for the regulation of genetic competence and DNA uptake) into the genome of host cells under the control of a mannitol‑inducible PmtlA promoter, and the generated strain was designated 164 M. However, there are no efficient expression kits for B. subtilis 164 M. To facilitate the efficient expression of target pathway, we attempted to transplant a T7 expression system into the 164 M genome at the aprE locus by integrating a DNA cassette expressing T7 RNA polymerase (T7 RNAP), which generated 164MT. In some gram‑negative intestinal bacteria, GDP‑l‑fucose is used as the l‑fucose donor to synthesize extracellular polysaccharides or lipopolysaccharides. For example, colanic acid (M‑antigen EPS) produced by E. coli has a specialized operon for expressing the enzymes required to perform glycosyl‑transferring reactions and generate GDP‑l‑fucose from d‑mannose, including WcaH. However, l‑fucose accumulation has not been reported, probably due to the low catalytic efficiency of WcaH for GDP‑l‑fucose. In addition, endogenous fucI, rhaA, and araA may also metabolize trace l‑fucose even if it is accidentally produced. Nevertheless, the E. coli pathway that enables GDP‑l‑fucose production together with WcaH or its mutants was transferred into B. subtilis. In contrast with E. coli, there is no known l‑fucose metabolizing pathway in B. subtilis, which would be a major advantage for l‑fucose accumulation. As shown in Fig. 2, the GDP‑l‑fucose formation pathway was cumulatively expressed by two artificial operons that were then engineered into 164MT. The first operon consisted of manA, manB and manC. manA encodes a native mannose‑6‑phosphate (M6P) isomerase that transforms fructose‑6‑phosphate (F6P) to M6P, which is then used as the substrate of phosphomannomutase (ManB) and catalyzed to mannose‑1‑phosphate (M1P). M1P is further converted into GDP‑D‑mannose (GDP‑Man) through the catalysis of mannose‑1‑phosphate guanyltransferase (ManC). The second operon contained coding sequences of GDP‑mannose 4, 6‑dehydratase (Gmd) and GDP‑fucose synthetase (WcaG). Gmd and WcaG convert GDP‑Man to GDP‑L‑fucose. Both operons were cloned behind the T7 promoter, and manA/manB/manC cassette was inserted at the manA locus in 164MT, resulting in 164MTM. Next, gmd/wcaG cassette was integrated at the xylA locus in the 164MTM genome, leading to 164GF, which was then transformed by the WcaH‑expressing plasmid pMK4‑T7wcaH. The transformant bearing pMK4‑T7wcaH was designated 164GFpWT. As the control, pMK4‑T7 was transformed into 164GF, generating 164GFpC (Fig. 2).Fig. 2Schematic diagram of the main metabolic pathways involving L‑fucose generation and the genotypes of B. subtilis strains leading to 164GFpWT, a strain capable of L‑fucose production after incorporating a vector expressing WcaH. G3P: glycerol‑3‑phosphate; DHAP: dihydroxy acetone‑phosphate; GAP: glyceraldehyde‑3‑phosphate; FBP: fructose‑1,6‑diphosphate; F6P: fructose‑6‑phosphate; M6P: mannose‑6‑phosphate; M1P: mannose‑1‑phosphate; GDP‑Man: GDP‑D‑mannose; GDP‑Fuc: GDP‑L‑fucose Both 164GFpC and 164GFpWT were cultivated in shake flasks containing Luria‑Bertani (LB) medium for 96 h, with glycerol supplemented as an extra carbon source. The cell‑free broth was collected as a sample for high‑performance liquid chromatography (HPLC) analysis during growth. As shown in Fig. 3A, a compound with the same retention time as l‑fucose was found in the culture broth of 164GFpWT but not in the culture broth of 164GFpC. The compound was further identified to be l‑fucose following liquid chromatography‑mass spectrometry (LC‑MS) analysis, with m/z values of 164.10 (l‑fucose), 182.16 (l‑fucose·H2O), and 147.08 (l‑fucose‑OH‑), corresponding to the molecular ion of l‑fucose (Additional file 1: Fig. S2). The l‑fucose accumulation in the broth of 164GFpWT reached 436.1 ± 27.5 mg/L after 96 h of cultivation (Fig. 3B). The cellular lysate of the cultures was also sampled and analyzed by HPLC, and almost no L‑fucose was found, indicating that nearly all L‑fucose was excreted out of the cells. WcaH can also break the native substrate GDP‑d‑mannose to yield d‑mannose, but d‑mannose was not detected in the fermentation broth by HPLC analysis. This fact is supposed to be caused by two reasons, first, B. subtilis can metabolize d‑mannose through mannokinase (or hexokinase) and enter glycolysis pathway; second, l‑fucose produced competitively inhibit the native catalytic reaction that should have yield d‑mannose. Herein, a de novo l‑fucose synthetic pathway was successfully constructed in a B. subtilis strain that was capable of l‑fucose production during growth. In contrast to the reported l‑fucose‑producing E. coli system [22], the engineered B. subtilis bypassed 2′‑FL formation and breakage. It is interesting that there has never been any report demonstrating that E. coli was able to synthesize l‑fucose utilizing its native GDP‑l‑fucose‑forming pathway and wild‑type (WT) WcaH, indicating that in addition to the l‑fucose metabolizing enzymes, sophisticated regulation mechanisms may also exist to guide GDP‑l‑fucose only toward polysaccharide synthesis, suggesting that the delicate engineering of E. coli native pathways is another feasible way to produce l‑fucose. Fig. 3Detection of L‑fucose by high‑performance liquid chromatography (HPLC). A HPLC profiles of cell‑free broths from the control strain and the WcaH‑expressing strain. B L‑fucose accumulated during 164GFpWT fermentation, as detected by HPLC ## Modeling‑based mutation of the l‑Fucose C6 methyl binding sites of WcaH WcaH is a type of MutT nucleoside triphosphate pyrophosphohydrolase belonging to a protein family of Nudix hydrolases. WcaH binds with Mg2+/Mn2+ and the triphosphate moiety of the substrate in a novel loop‑helix‑loop motif that is conserved in the MutT family of proteins, and its homology modeling has been established by using published X‑ray structures as templates [33]. The chemical difference between GDP‑L‑fucose and GDP‑d‑mannose is the inclusion of a methyl group on the C6 of L‑fucose instead of a hydroxymethyl group in d-mannose (Fig. 4A). Structural analysis of the complex of WcaH combined with GDP‑d‑mannose or GPD‑l‑fucose revealed that the R36 of WcaH is responsible for binding the hydroxyl group on C6 of GDP‑d‑mannose. In addition, there is a hydrogen bond between R36 and the C6 hydroxyl, and R36 can form a salt bridge with the second phosphate radical on GDP, which helps to stabilize the intermediate state formed during catalysis. Moreover, GDP‑l‑fucose is not able to form an efficient connection with R36, which may account for the low catalytic activity of WcaH for GDP‑l‑fucose. To potentially enhance the GDP‑l‑fucose‑hydrolyzing activity of WcaH, R36 can be replaced by a hydrophobic amino acid, which supposedly stabilizes the methyl group on the C6 of l‑fucose. For that purpose, R36 was substituted for aromatic amino acids, such as R36Y, R36F, and R36W. Another strategy is to form a stabler salt bridge between the enzyme and GDP‑l‑fucose in the intermediate state by changing R36 to an aliphatic amino acid, such as isoleucine (R36I) or leucine (R36L), which might provide a more hydrophobic microenvironment for the methyl group on the C6 of l‑fucose. Fig. 4Modeling of WcaH, the R36 site mutations and the effects on L‑fucose production by B. subtilis strains. A Structure of WcaH binding with GDP‑L‑fucose (green) and GDP‑D‑mannose (pink) and residue R36. B L‑fucose accumulated in the broths of different strains. C Comparison of L‑fucose produced by strains expressing wild‑type WcaH‑ or R36‑modified enzymes Thus, 164GF strain harboring vectors expressing WcaHR36Y, WcaHR36F, WcaHR36L, WcaHR36I, and WcaHR36W were constructed and tested for L‑fucose production with growth experiments. As shown in Fig. 4B, l-fucose accumulated gradually and reached its maximum level after 96 h of fermentation. The highest titers of l-fucose during 96 h cultivation by WcaHR36Y, WcaHR36F, WcaHR36L, WcaHR36I, and WcaHR36W reached 609.3 mg/L, 514.5 mg/L, 310.7 mg/L, 357.7 mg/L, and 425.3 mg/L, respectively, compared to 436.1 mg/L achieved by WT WcaH. Thus, $39.7\%$ and $18.0\%$ increases in l-fucose were achieved by WcaHR36Y and WcaHR36F, respectively (Fig. 4C). Surprisingly, the GDP‑l-fucose‑hydrolyzing abilities of WcaHR36L and WcaHR36I seemed to be dramatically lowered. ## Enzyme design around the phosphate radical of GDP Following the investigation of the interaction between GDP‑l-fucose and the WcaH catalytic center, amino acid residues enclosing the phosphate radical were further projected and rationally designed based on WcaHR36Y and WcaHR36F. We speculated that the changes occurred on the binding sites on phosphate moiety may change the position or the orientation of the substrate in the pocket, so that it may lead to improved catalytic efficiency. Then N38R/K, L65R/K and F102R/K were designed and deduced to be able to stabilize the intermediate state of the enzyme/substrate. We speculated that those alkaline amino acids could form salt bridges or hydrogen bonds with the phosphate radical of GDP. Among them, L65R/K might act on the first phosphate radical in GDP, while N38R/K and F102R/K might stabilize the second phosphate radical. The molecular mechanics‑generalized born surface area (MM‑GBSA) method was used to calculate the binding free energy to evaluate the thermodynamic stability. As shown in Fig. 5B, most of the mutant enzymes were more stable than WT WcaH with respect to the binding of GDP‑l-fucose, except for Y102R performed on R36Y or R36F (Fig. 5B); therefore, most of those mutants exhibited higher catalytic activity for the substrate. Fig. 5Mutation of the site around the phosphate radical of GDP‑L‑fucose. A Structure of the WcaH active center around the phosphate groups. B MM/GBSA‑based free binding energy prediction of the enzymes upon binding GDP‑L‑fucose. C L‑fucose production of WcaHR36Y and its derived mutants. D L‑fucose production of WcaHR36F and its derived mutants Based on the modeling analysis, N38R/K, L65R/K, or F102R/K was introduced into WcaHR36Y or WcaHR36F, and the l-fucose production potential of the resultant mutants was evaluated thereafter through growth experiments. As shown in Fig. 5C and D, the l-fucose titers after 96 h of cultivating WcaHR36Y/F102K, WcaHR36Y/F102R, WcaHR36Y/L65K, WcaHR36Y/L65R, WcaHR36Y/N38K, and WcaHR36Y/N38R reached 625.6 mg/L, 561.5 mg/L, 683.8 mg/L, 634.8 mg/L, 652.9 mg/L, and 777.3 mg/L, respectively. Moreover, the L‑fucose titers of WcaHR36F/F102K, WcaHR36F/F102R, WcaHR36F/L65K, WcaHR36F/L65R, WcaHR36F/N38K, and WcaHR36F/N38R were determined to be 445.9 mg/L, 306.6 mg/L, 544.8 mg/L, 464.8 mg/L, 562.9 mg/L, and 602.1 mg/L, respectively. Notably, the double mutants derived from WcaHR36Y had comparatively higher l-fucose titers than their counterparts derived from WcaHR36F, which is consistent with the results observed from the related single mutations. Among all of the mutants, WcaHR36Y/N38R demonstrated the best l-fucose production performance, with a $27.6\%$ increase compared to that of the single mutant WcaHR36Y. Modeling and simulating the structures of enzyme‑substrate complexes has become a common method in enzyme engineering, and structure‑guided design at the active site can not only enhance the catalytic capacity of enzymes but also facilitate gain‑of‑function for the synthesis of new products in many studies [34–36]. In this manner, we have successfully developed a “better” GDP‑l-fucose‑hydrolyzing enzyme by rational design and the site‑directed mutagenesis of WT WcaH. The amino acid residue at position 36 was confirmed to be a key residue for substrate binding. The nonpolar amino acid side chain generated by mutation at position 36 may form a suitable fucosyl‑binding pocket for GDP‑l-fucose. Amino acid residues around the phosphate group of GDP‑l-fucose were also found to affect the activity of WcaH. Consequently, when the asparagine residue at position 38 was changed into a basic amino acid arginine residue, the activity of engineered WcaH for GDP‑l-fucose was further increased. In addition to the rational design of key binding sites reported in this work, more mutagenesis libraries involving other structural sites should be considered and investigated to further increase the enzymatic efficiency. ## Systematic improvement of the cellular platform to improve l‑Fucose production Recombinant T7 RNAP is one of most important biological applications [37], and the construction of a T7 RNAP/T7 promoter‑based heterologous expression system in B. subtilis is vital for l-fucose production. To investigate the impact of the level of T7 RNAP on heterologous peptides as well as the l-fucose synthesis efficiency, different promoters that drive T7 RNAP expression were evaluated systematically. To achieve this, the coding DNA fragment of T7P was integrated at the aprE locus under PaprE, the native aprE promoter, and three other promoters, including the constitutive promoter P43, PxylA, a xylose‑inducible promoter, and PrpsF, that specifically and efficiently remained active during the logarithmic phase [38], producing strains that included 164MT, 164MCT, 164MCX, and 164MCR, respectively (Fig. 6A). Next, plasmid pMK4‑T7‑gfp was constructed and transformed into those strains. To assess the promoter strength, GFP‑expressing strains were cultivated in 96‑well microplates or in 250 ml shaking flasks containing liquid LB medium. Green fluorescent protein (GFP) fluorescence was automatically monitored by a microplate reader between 0–960 min (Fig. 6B) or manually measured using a spectrometer between 0–60 h (Fig. 6C).Fig. 6Evaluation of the expression strength of T7 RNA polymerase on the level of GFP transcribed under a T7 promoter. A Schematic diagram of the evaluation system, including the placement of different promoters for the transcription of T7 RNAP and the construction and assay of GFP‑expressing strains. B Automatic monitoring of the fluorescence intensity of cultures incubated in 96‑well microplates. C Fluorescence intensity of strains cultivated in shaking flasks. D L‑fucose production of strains equipped with T7 RNAP driven by P43 (164MCTGF) or PxylA (164MCXGF) As shown in Fig. 6B, during early logarithmic growth (0–120 min), P43 and PxylA led to the other two promoters in the expression of GFP. During shake‑flask growth (Fig. 6C), P43 remained active after 12 h during the whole cultivation period; however, its accumulated GFP fluorescence was surpassed by that of PxylA after 48 h, whereas PrpsF exhibited the lowest promoter efficiency. Based on these results, we speculated that T7 RNAP is better transcribed by PxylA or P43 when all of the components of the artificial synthesis pathway are placed under the T7 promoter. Therefore, growth experiments were performed using the engineered B. subtilis strains 164MCTGF and 164MCXGF that harboring pMK4-R36Y/N38R, and the accumulated l-fucose titers were monitored during growth. As seen in Fig. 6D, after 120 h of cultivation, 164MCXGF produced 1600.4 mg/L of L‑fucose, which was $16.4\%$ higher than that produced by 164MCTGF (1375.4 mg/L). ## Fed‑batch fermentation for l‑Fucose production To evaluate the feasibility of l-fucose scale‐up production, fed‑batch fermentation was further performed. Respectively, 164MCTGF pR36Y/N38R and 164MCXGF pR36Y/N38R were fermented in bioreactors in preliminary study, and their productivity was quite close. For the need of inducer‑free during growth, we use 164MCTGF pR36Y/N38R for further optimizing fermentation study. 164MCTGF pR36Y/N38R was cultivated for 96 h in 5 L bioreactors, each containing 2.5 L of LB supplemented with 800 g/L glycerol as the feeding source. As shown in Fig. 7, the l-fucose yield increased progressively, with the highest L‑fucose titer reaching 6.4 g/L within 96 h, which was 4 times that achieved during shake‑flask growth. The highest cell density was achieved at 24 h, with an OD600 value of 131.7. Glycerol was replenished into the bioreactor during fermentation and maintained at more than 10 g/L, and 18.4 g/L glycerol was left unconsumed in the broth at 96 h. The strain produced 6.4 g/L l-fucose within 96 h, reaching a productivity of 0.06 g/L/h during that period. Fig. 7Concentration profiles of L‑fucose, glycerol and biomass during the fed‑batch growth of 164MCTGF Therefore, by introducing rationally designed WcaH into an engineered B. subtilis strain, for the first time, we created a new fermentation biotechnology for l-fucose production that was independent of the chemical or enzymatic lysis of fucoidan or fucosyllated oligosaccharides. While fucosylation is considered to be an important modification that endows molecules with diverse structures and biological activities, L‑fucose itself can be used for pharmaceutical, medical nutrition and cosmetic applications. This novel bioproduction route paves the foundation for large‑scale l-fucose fermentation production and commercialization in the future. ## Conclusion In summary, we engineered B. subtilis for the production of a valuable monosaccharide, l-fucose, from an economic carbon source combined with the rational design of GDP‑d‑mannose mannosyl hydrolase and de novo construction of the related pathway. This work is the first to report l-fucose biosynthesis via an artificial pathway that directly hydrolyzes GDP‑L‑fucose. The expression system and pathway constructed in this study not only provide a sustainable method for food‑grade l-fucose production but also provide a platform for the synthesis of other valuable sugars derived from l-fucose. ## Strains, plasmids and medium The strains and plasmids used in this study are listed in Additional file 1: Table S1. The primers used in this study are listed in Additional file 1: Table S2. B. subtilis ATCC 6051a was purchased from the American Type Culture Collection (ATCC) and used for genetic manipulation in this study. E. coli DH5α was used for vector construction, propagation and preservation, and its genomic DNA was applied as the template for the polymerase chain reaction (PCR) preparation of DNA fragments derived from E. coli. pMK4, a gift from the Bacillus Genetic Stock Center (BGSC), was used as the backbone to construct B. subtilis expression vectors. For recombinant expression, wcaH (NCBI Gene ID: 946,559) was cloned behind a T7 promoter on pMK4‑T7. LB medium (yeast extract 5 g/L, tryptone 10 g/L, NaCl 10 g/L) was used for the cultivation of E. coli and B. subtilis. Routine growth experiments were performed at 37 ℃ with shaking rotors at 200 rpm, and chloramphenicol (10 mg/L), ampicillin (100 mg/L) and erythromycin (10 mg/L) were added to the medium as needed. ## Growth experiments For shake‑flask growth, a B. subtilis single colony was inoculated into 5 mL of liquid LB and incubated at 37 ℃ overnight with shaking at 200 rpm. Then, 1 mL of grown culture was transferred to 50 mL of modified LB (LB with extra glycerol, 20 g/L) in a $\frac{250}{500}$ mL shake flask and grown at 37 ℃ with shaking at 200 rpm. Samples were collected at regular time intervals for HPLC and LC–MS analyses. For fed‑batch fermentation, an overnight culture of 50 mL LB grown in a 500 mL shake flask was transferred to a 5 L bioreactor with 3.0 L of fermentation medium that contained 20 g/L glycerol, 5 g/L yeast extract, 20 g/L tryptone, 2 g/L MgSO4·6H2O, 10 g/L (NH4)2HPO4, and 5 g/L KCl. The medium pH was maintained at 6.5 using aqueous ammonia as an alkali supplement and 500 g/L glycerol as a carbon source supplement. The growth temperature was set at 37 °C, with the ventilation volume set at 3.0 vvm and an initial stirring speed of 500 rpm. ## Genetic manipulations Vector DNA construction was mostly accomplished by using the ClonExpress II One Step Cloning Kit (Vazyme Biotech Co., Ltd.). Linear DNA fragments for homologous recombination were constructed via the overlapping PCR technique. For instance, the PmtlA‑comk cassette contained a copy of comk assembled behind the mtlA promoter, homologous arms (UnprE and DnprE) for recombination and an erythromycin resistance gene (ermC) for selection (Fig. 1). Such a five‑component fragment (UnprE, ermC, PmtlA, comk and DnprE) was prepared by two rounds of PCR. First, PCR with 2 × Phanta Flash Master Mix (Vazyme Biotech Co., Ltd.) was performed to prepare each fragment component. For preparation of the flanking fragments (UnprE and DnprE) of approximately 1000 bp each, PmtlA and comk, the B. subtilis 6051a genome was used as the template, while the ermC fragment was obtained by PCR using the plasmid pMD19T‑aea as the template [26]. Next, all PCR products were treated with QuickCut™ DpnI (TaKaRa Biotechnology Co., Ltd.) and pooled together at the proper mole ratio (usually 1:1 between every two fragments) for overlapping PCR that generated the PmtlA‑comk cassette. Other linear DNAs, such as P43‑T7P, PxylA‑T7P, PrpsF‑T7P, PT7‑manC‑manB‑manA, or PT7‑gmd‑wcaG, were all prepared by similar strategies. Recombination‑mediated genetic manipulations were performed by incubation of linear DNAs with corresponding B. subtilis competent cells, which was prepared as described by Ji et al., with the modification of $1\%$ mannitol being used as the inducer instead of xylose [32]. Usually, 100 ng of plasmid or 1‑3 µg of linear DNA fragment was applied and incubated with 0.5 ml competence‑induced cell suspension for 120 min at 37 ℃. Then, 100 μL of regenerated cells were plated on solid medium and incubated overnight at 37 ℃. Colonies formed on selective plates were picked for colony PCR. *The* generated mutant strain was again transformed with plasmid pMK4‑cre to excise the selective marker ermC from the genome. ## WcaH modeling and docking The WT WcaH model was built through Prime in Cite Schrodinger 2020‑3 (https://www.schrodinger.com/) using PDB structures 2I8T [22] and 2GT4 [39] as templates. Mg2+ and GDP‑mannose from the templates were kept in the WT WcaH model. Glide XP [40] was used in Cite Schrodinger 2020‑3 to simulate and dock GDP‑l-fucose and GDP‑mannose to the WT WcaH model. Modeling of WcaH mutants docked with GDP‑l-fucose was performed according to the conformation of WT WcaH bound with GDP‑l-fucose. Prime MM‑GBSA [41] (with the OPLS‑3e force field) was used to calculate the binding free energy for all docked complexes. ## Analytical and quantitative methods The cell biomass was monitored by measuring the OD600 values of the cell suspension with an ultraviolet spectrophotometer, and the growth curve was fitted with a sigmoidal model. Detection of l-fucose and other related metabolites in the fermentation broth was conducted using a Shimadzu 20AVP HPLC system (Shimadzu Corp.) equipped with an RID‑10A refractive index detector. To prepare cell‑free samples, cultures were centrifuged at 12,000 ×g for 10 min and filtered using PTFE filters with an aperture span of 0.22 µm. The samples were then loaded into an Aminex HPX‑87H column (300 × 7.8 mm) (Bio‑Rad, USA) and separated by 5 mM H2SO4 solution as the mobile phase at a flow rate of 0.6 mL/min at 65 ℃. The L‑fucose concentration was calculated according to the calibrated standard curve. All data are representative of parallel independent experiments and presented as the means ± standard deviations (SDs). Statistical differences among the means of two or more groups ($p \leq 0.05$) were determined using one‑way analysis of variance. ## Supplementary Information Additional file 1: Figure S1. B. subtilis growth experiment when supplied with A glucose B xylose C fucose D glycerol as the sole carbon source. Figure S2. L-Fucose standard A and fermentation supernatant B were analyzed by LC-MS. Table S1. The strains and plasmids used in this study. Table S2. The primers used in this study. ## References 1. Holdener BC, Haltiwanger RS. **Protein O-fucosylation: structure a-nd function**. *Curr Opin Struct Biol* (2019) **56** 78-86. DOI: 10.1016/j.sbi.2018.12.005 2. Wu GY, Niu MW, Tang WL, Hu JJ, Wei GQ, He ZK. **l-Fu-cose ameliorates high-fat diet-induced obesity and hepatic steatosis in mice**. *J Transl Med* (2016). DOI: 10.1186/s12967-018-1718-x 3. Pickard JM, Chervonsky AV. **Intestinal Fucose as a mediator of host-microbe symbiosis**. *J Immunol* (2015) **194** 5588-5593. DOI: 10.4049/jimmunol.1500395 4. Quast I, Lunemann JD. **Fc glycan-modulated immunoglobulin g effector functions**. *J Clin Immunol* (2014) **34** S51-S55. DOI: 10.1007/s10875-014-0018-3 5. Kelly RJ, Rouquier S, Giorgi D, Lennon GG, Lowe JB. **Sequence and expression of a candidate for the human Secretor blood group alpha(1,2) fucosyltransferase gene (FUT2) Homozygosity for an enzyme-inactivating nonsense mutation commonly correlates with the non-secretor phenotype**. *J Biol Chem* (1995) **270** 4640-9. DOI: 10.1074/jbc.270.9.4640 6. Smyth DJ, Cooper JD, Howson JMM, Clarke P, Downes K, Mistry T. **FUT2 nonsecretor status links type 1 diabetes susceptibility and resistance to infection**. *Diabetes* (2011) **60** 3081-3084. DOI: 10.2337/db11-0638 7. Koda Y, Tachida H, Pang H, Liu YH, Soejima M, Ghaderi AA. **Contrasting patterns of polymorphisms at the ABO-secretor gene (FUT2) and plasma alpha(1,3) fucosyltransferase gene (FUT6) in human populations**. *Genetics* (2001) **158** 747-756. DOI: 10.1093/genetics/158.2.747 8. Yang BY, Gray JSS, Montgomery R. **Extracellular polysaccharide of erwinia-chrysanthemi ECH6**. *Int J Biol Macromol* (1994) **16** 306-312. DOI: 10.1016/0141-8130(94)90061-2 9. Feichtinger RG, Hullen A, Koller A, Kotzot D, Grote V, Rapp E. **A spoonful of L-fucose-an efficient therapy for GFUS-CDG, a new glycosylation disorder**. *Embo Mol Med* (2021). DOI: 10.15252/emmm.202114332 10. Chin YW, Seo N, Kim JH, Seo JH. **Metabolic engineering of**. *Biotechnol Bioeng* (2016) **113** 2443-2452. DOI: 10.1002/bit.26015 11. Choi YH, Park BS, Seo JH, Kim BG. **Biosynthesis of the human milk oligosaccharide 3-fucosyllactose in metabolically engineered Escherichia coli via the salvage pathway through increasing GTP synthesis and beta-galactosidase modification**. *Biotechnol Bioeng* (2019) **116** 3324-3332. DOI: 10.1002/bit.27160 12. Zhang X, Xu P, Yu B. **Chemical synthesis of a colanic acid hexasaccharide**. *Org Lett* (2022). DOI: 10.1021/acs.orglett.2c03116 13. Fitton JH, Stringer DN, Karpiniec SS. **Therapies from fucoidan: an update**. *Mar Drugs* (2015) **13** 5920-5946. DOI: 10.3390/md13095920 14. Flowers HM. **Chemistry and biochemistry of D- and L-fucose**. *Adv Carbohydr Chem Biochem* (1981) **39** 279-345. DOI: 10.1016/S0065-2318(08)60208-5 15. Gesson JP, Jacquesy JC, Mondon M, Petit P. **A short synthesis of L-fucose and analogs from D-mannose**. *Tetrahedron Lett* (1992) **33** 3637-3640. DOI: 10.1016/s0040-4039(00)92522-x 16. Roca C, Alves VD, Freitas F, Reis MAM. **Exopolysaccharides enriched in rare sugars: bacterial sources, production, and applications**. *Front Microbiol* (2015). DOI: 10.3389/fmicb.2015.00288 17. Kim IJ, Kim DH, Nam KH, Kim KH. **Enzymatic synthesis of l-fucose from L-fuculose using a fucose isomerase from**. *Biotechnol Biofuels* (2019). DOI: 10.1186/s13068-019-1619-0 18. Gramelsberger G. **The simulation approach in synthetic biology**. *Stud Hist Philos Biol Biomed Sci* (2013) **44** 150-157. DOI: 10.1016/j.shpsc.2013.03.010 19. Purnick PEM, Weiss R. **The second wave of synthetic biology: from modules to systems**. *Nat Rev Mol Cell Biol* (2009) **10** 410-422. DOI: 10.1038/nrm2698 20. Liu JJ, Lee JW, Yun EJ, Jung SM, Seo JH, Jin YS. **L-Fucose production by engineered**. *Biotechnol Bioeng* (2019) **116** 904-911. DOI: 10.1002/bit.26907 21. Legler PM, Lee HC, Peisach J, Mildvan AS. **Kinetic and magnetic resonance studies of the role of metal ions in the mechanism of Escherichia coli GDP-mannose mannosyl hydrolase, an unusual nudix enzyme**. *Biochemistry* (2002) **41** 4655-4668. DOI: 10.1021/bi012118d 22. Zou YZ, Li C, Brunzelle JS, Nair SK. **Molecular basis for substrate selectivity and specificity by an LPS biosynthetic enzyme**. *Biochemistry* (2007) **46** 4294-4304. DOI: 10.1021/bi061056u 23. Xia ZY, Azurmendi HF, Lairson LL, Withers SG, Gabelli SB, Bianchet MA. **Mutational, structural, and kinetic evidence for a dissociative mechanism in the GDP-mannose mannosyl hydrolase reaction**. *Biochemistry* (2005) **44** 8989-8997. DOI: 10.1021/bi050583v 24. Fischer E, Sauer U. **Large-scale in vivo flux analysis shows rigidity and suboptimal performance of Bacillus subtilis metabolism**. *Nat Genet* (2005) **37** 636-640. DOI: 10.1038/ng1555 25. Westbrook AW, Ren X, Oh J, Moo-Young M, Chou CP. **Metabolic engineering to enhance heterologous production of hyaluronic acid in**. *Metab Eng* (2018) **47** 401-413. DOI: 10.1016/j.ymben.2018.04.016 26. Ji MH, Li SJ, Chen A, Liu YH, Xie YK, Duan HY. **A wheat bran inducible expression system for the efficient production of alpha-L-arabinofuranosidase in**. *Enzym Microbial Technol* (2021). DOI: 10.1016/j.enzmictec.2020.109726 27. Sauer U, Hatzimanikatis V, Bailey JE, Hochuli M, Szyperski T, Wuthrich K. **Metabolic fluxes in riboflavin-producing**. *Nat Biotechnol* (1997) **15** 448-452. DOI: 10.1038/nbt0597-448 28. Ji MH, Liu YF, Xie SQ, Fu C, Liu M, Shi JP. **De novo synthesis of 2'-fucosyllactose in engineered**. *Process Biochem* (2022) **120** 178-185. DOI: 10.1016/j.procbio.2022.06.007 29. Jeong H, Sim YM, Park S-H, Choi S-K. **Complete genome sequence of**. *Genome Announc* (2015). DOI: 10.1128/genomeA.00532-15 30. Skolpap W, Scharer JM, Douglas PL, Moo-Young M. **Fed-batch optimization of alpha-amylase and protease-producing Bacillus subtilis using Markov chain methods**. *Biotechnol Bioeng* (2004) **86** 706-717. DOI: 10.1002/bit.20079 31. Widner B, Thomas M, Sternberg D, Lammon D, Behr R, Sloma A. **Development of marker-free strains of**. *J Ind Microbiol Biotechnol* (2000) **25** 204-212. DOI: 10.1038/sj.jim.7000051 32. Ji MH, Liu YH, Wu HY, Li SJ, Duan HY, Shi JP. **Engineering**. *J Ind Microbiol Biotechnol* (2021). DOI: 10.1093/jimb/kuab024 33. Gabelli SB, Azurmendi HF, Bianchet MA, Amzel LM, Mildvan AS. **X-ray, NMR, and mutational studies of the catalytic cycle of the GDP-mannose mannosyl hydrolase reaction**. *Biochemistry* (2006) **45** 11290-11303. DOI: 10.1021/bi061239g 34. Bi HP, Qu G, Wang S, Zhuang YB, Sun ZT, Liu T. **Biosynthesis of a rosavin natural product in**. *Metab Eng* (2022) **69** 15-25. DOI: 10.1016/j.ymben.2021.10.010 35. Fisher AK, Freedman BG, Bevan DR, Senger RS. **A review of metabolic and enzymatic engineering strategies for designing and optimizing performance of microbial cell factories**. *Comput Struct Biotechnol J* (2014) **11** 91-99. DOI: 10.1016/j.csbj.2014.08.010 36. Zhu T, Cui YL, Geng WC, Liu GX, Jiang HF, Li RF. **Creating an unusual glycine-rich motif in a peptide amidase leads to versatile Protein C-terminal traceless functionalization**. *Acs Catalysis* (2022). DOI: 10.1021/acscatal.2c01456 37. Wang WY, Li YWB, Wang YQ, Shi C, Li CM, Li Q. **Bacteriophage T7 transcription system: an enabling tool in synthetic biology**. *Biotechnol Adv* (2018) **36** 2129-2137. DOI: 10.1016/j.biotechadv.2018.10.001 38. Nijland R, Lindner C, van Hartskamp M, Hamoen LW, Kuipers OP. **Heterologous production and secretion of**. *J Biotechnol* (2007) **127** 361-372. DOI: 10.1016/j.jbiotec.2006.07.014 39. Gabelli SB, Azurmendi HF, Bianchet MA, Amzel LM, Mildvan AS. **X-ray, NMR, and mutational studies of the catalytic cycle of the GDP-mannose mannosyl hydrolase reaction**. *Biochemistry* (2006) **45** 11290-11303. DOI: 10.1021/bi061239g 40. Friesner RA, Murphy RB, Repasky MP, Frye LL, Greenwood JR, Halgren TA. **Extra precision glide: docking and scoring incorporating a model of hydrophobic enclosure for protein-ligand complexes**. *J Med Chem* (2006) **49** 6177-6196. DOI: 10.1021/jm051256o 41. Jacobson MP, Pincus DL, Rapp CS, Day TJF, Honig B, Shaw DE. **A hierarchical approach to all-atom protein loop prediction**. *Proteins-Struct Funct Bioinform* (2004) **55** 351-367. DOI: 10.1002/prot.10613
--- title: Mediation effects of DNA methylation and hydroxymethylation on birth outcomes after prenatal per- and polyfluoroalkyl substances (PFAS) exposure in the Michigan mother–infant Pairs cohort authors: - Rebekah L. Petroff - Raymond G. Cavalcante - Elizabeth S. Langen - Dana C. Dolinoy - Vasantha Padmanabhan - Jaclyn M. Goodrich journal: Clinical Epigenetics year: 2023 pmcid: PMC10037903 doi: 10.1186/s13148-023-01461-5 license: CC BY 4.0 --- # Mediation effects of DNA methylation and hydroxymethylation on birth outcomes after prenatal per- and polyfluoroalkyl substances (PFAS) exposure in the Michigan mother–infant Pairs cohort ## Abstract ### Background Per- and polyfluoroalkyl substances (PFAS) are chemicals that are resistant to degradation and ubiquitous in our environments. PFAS may impact the developing epigenome, but current human evidence is limited to assessments of total DNA methylation. We assessed associations between first trimester PFAS exposures with newborn DNA methylation, including 5-methylcytosine (5-mC) and 5-hydroxymethylcytosine (5-hmC). DNA methylation mediation of associations between PFAS and birth outcomes were explored in the Michigan Mother Infant Pairs cohort. Nine PFAS were measured in maternal first trimester blood. Seven were highly detected and included for analysis: PFHxS, PFOA, PFOS, PFNA, PFDA, PFUnDA, and MeFOSAA. Bisulfite-converted cord blood DNA ($$n = 141$$) and oxidative-bisulfite-converted cord blood ($$n = 70$$) were assayed on Illumina MethylationEPIC BeadChips to measure total DNA methylation (5-mC + 5-hmC) and 5-mC/5-hmC. Correcting for multiple comparisons, beta regressions were used to assess associations between levels of PFAS and total methylation, 5-mC, or 5-hmC. Nonlinear mediation analyses were used to assess the epigenetic meditation effect between PFAS and birth outcomes. ### Results PFAS was significantly associated with total methylation (q < 0.05: PFHxS—12 sites; PFOS—19 sites; PFOA—2 sites; PFNA—3 sites; PFDA—4 sites). In 72 female infants and 69 male infants, there were sex-specific associations between five PFAS and DNA methylation. 5-mC and 5-hmC were each significantly associated with thousands of sites for PFHxS, PFOS, PFNA, PFDA, PFUnDA, and MeFOSAA (q < 0.05). Clusters of 5-mC and 5-hmC sites were significant mediators between PFNA and PFUnDA and decreased gestational age (q < 0.05). ### Conclusions This study demonstrates the mediation role of specific types of DNA methylation on the relationship between PFAS exposure and birth outcomes. These results suggest that 5-mC and 5-hmC may be more sensitive to the developmental impacts of PFAS than total DNA methylation. ### Supplementary Information The online version contains supplementary material available at 10.1186/s13148-023-01461-5. ## Introduction Gestational exposure to toxicants can negatively impact birth outcomes and have lasting effects on child and adult health, including adverse effects on neurodevelopment, growth, adiposity, and metabolism [1, 2]. One group of toxicants concerning to the health of pregnant women and children are per- and polyfluoroalkyl substances (PFAS), a class of over 12,000 unique chemicals [3] that are widely found in products including cookware, carpet, and food packaging because of their resistance to stains, water, and grease [4, 5]. PFAS have also been used in aqueous film-forming foams used for fire suppression at airports and military bases, leading to the contamination of the surrounding environment and nearby drinking water [6, 7]. PFAS are highly persistent and have accordingly been detected in maternal or umbilical cord plasma or serum in birth cohorts across the United States of America [8–11], Spain [12], China [13], Taiwan [14], Japan [15], and more. Most research to date has measured legacy PFAS, including perfluorohexanesulphonic acid (PFHxS), perfluorooctanesulfonic acid (PFOS), perfluorooctanoic acid (PFOA), and perfluorononanoic acid (PFNA), reporting near ubiquitous detection of all four chemicals in pregnant participants. In other studies, these exposures have been connected with a variety of adverse birth outcomes, including preterm birth or shorter gestational length [16–18], lower birth weight [11, 17–19], and hypertensive disorders of pregnancy, such as preeclampsia [20]. Concerningly, the health effects of prenatal PFAS exposures appear to extend beyond birth, with longitudinal studies reporting links between gestational PFAS and childhood adiposity/metabolic health. In Project Viva ($$n = 876$$), girls who had higher prenatal exposure to PFHxS, PFOS, PFOA, and PFNA also had increased mid-childhood adiposity [10]. In the HOME study ($$n = 212$$), early gestational PFHxS and PFOA concentrations were associated with higher central adiposity and increased risk for overweight/obesity at 12 years of age [21]. One major mechanism by which PFAS may be causing birth and later childhood health effects is via epigenetic perturbations. Epigenetic marks are mitotically heritable modifications to DNA and chromatin that control the expression of genes without altering the DNA sequence [22]. During embryogenesis, the epigenome is highly vulnerable to dysregulation, due to post-fertilization epigenetic erasure and post-implantation reprogramming [22]. Any epigenetic disruption during this early developmental stage can be passed on to all subsequent cells across tissue types. One mechanism of epigenetic regulation that is stable across time is DNA methylation at cytosine-guanine (CpG) dinucleotides [23]. DNA methylation at so-called CpG sites (5-methylcytosine or 5-mC) can undergo oxidation to hydroxymethylation (5-hydroxymethylcytosine or 5-hmC) [24, 25] via TET enzymes [26, 27]. While 5-hmC is less abundant than 5-mC, it is a stable DNA modification, with detectable levels in the mammalian brain, liver, kidney, testes, placenta, colon, blood, and embryonic stem cells [28–32]. Like 5-mC, 5-hmC undergoes dynamic changes during early gestation that may persist throughout the lifespan, but the functional roles of 5-hmC and 5-mC in gene regulation seem to be distinct [33]. Still, both types of methylation are independently essential in processes of cell differentiation, fetal growth, and nervous system development and function from early life through adolescence [34–38]. PFAS exposures in human, animal (rodents and zebrafish), and in vitro models have broadly been linked with differences in DNA methylation (for reviews, see Kim et al. [ 39] and Perng et al. [ 40]). Specifically, human prenatal exposures to PFHxS, PFOS, PFOA, PFNA, and perfluorodecanoic acid (PFDA) have been associated with differences in newborn or childhood blood methylation in six different epidemiological studies assessed via Infinium arrays [40–46]. Across these cohorts, only one statistically significant gene, TNXB, was replicated in two studies [45, 46], but differential methylation of genes and enriched gene pathways related to developmental processes, adiposity, metabolism, and neurological function were identified in most studies. There is also evidence for sex-specific associations in two of three studies that considered these relationships [44–46]. However, these epidemiological studies only measured total DNA methylation; the commonly used bisulfite-treatment methods do not distinguish between 5-mC and 5-hmC. In vitro studies have shown that PFAS can disrupt the regulation of oxidating TET genes [27, 47], suggesting that PFAS could broadly alter 5-hmC. Thus, it is important to assess the hydroxymethylome in studies of gestational PFAS exposure. The present study aimed to identify genes in umbilical cord blood DNA that are differentially methylated and/or hydroxymethylated by first trimester exposures to PFAS and assess how these epigenetic differences mediate relationships between PFAS and adverse birth outcomes in the Michigan Mother–Infant Cohort (MMIP). We hypothesized that gestational exposures to well-studied, legacy PFAS (PFHxS, PFOS, PFOA, and PFNA), as well as additional, less-studied PFAS (PFDA, perfluoroundecanoic acid (PFUnDA), 2-(N-methyl-perfluorooctane sulfonamido) acetic acid (MeFOSAA)) would be associated with altered newborn DNA total methylation, 5-mC, and 5-hmC. We also hypothesized that some associations would be specific to assigned sex at birth. ## Cohort characteristics After samples from the cohort were assessed for quality (Fig. 1), demographic data suggested that maternal variables were largely similar between the entire cohort, those with passing total methylation data, and those with individual level 5-mC and 5-hmC data (Table 1). In those with passing total methylation and those with individual 5-mC/5-hmC data, maternal age was an average of 31.8 years, mean baseline weight was between 69 and 70 kg, and average baseline BMI ranged from 25.5 to 25.8. Participants were largely married, never-smokers, and self-identified as White, non-Hispanic. There were slight differences between 5-mC/5-hmC and both the entire cohort and those with passing total methylation in marriage status, smoking status, and self-reported race and ethnicity ($p \leq 0.05$).Fig. 1Schematic diagram of study. 309 pregnant people were recruited in the first trimester and 288 remained in the study and had data collected at the time of birth. Among these, 173 provided a cord blood sample for epigenetic analysis at delivery. 206 of these families also had PFAS measured on their first-trimester plasma samples. Analytes of 9 PFAS were measured (Additional file 1: Table S1). Two PFAS were dropped from analysis due to poor detection (> $80\%$ of samples below the limit of detection, BLOD). Another two PFAS (PFUnDA and MeFOSAA) were converted into categorical variables, detected or not detected, due to their moderate detection (> $40\%$ and < $80\%$ of samples BLOD). Five PFAS (PFHxS, PFOS, PFOA, PFNA, PFDA) were treated as continuous concentration measures in analysis, with < $40\%$ of samples BLOD. From dyads that had PFAS, 155 had EPIC data on total methylation, and 90 had 5-hydroxymethylcytosine (5-hmC)/5-methylcytosine (5-mC) EPIC data. For total methylation, 141 samples and 744,926 probes passed quality control (QC). For 5-hmC/5-mC, 70 samples and 528,389 probes passed QC and screening criteria. Abbreviations: 5-hmC: 5-hydroxymethylcytosine; 5-mC: 5-methylcytosine; BLOD: below the limit of detection; MeFOSAA: 2-(N-methyl-perfluorooctane sulfonamido) acetic acid; MMIP: Michigan Mother Infant Pairs; PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOA: perfluorooctanoic acid; PFOS: perfluorooctanesulfonic acid; PFUnDA: perfluoroundecanoic acid; QC: quality controlTable 1Cohort demographicsMaternal characteristicsMean (SD) or percent (count)Entire cohort $$n = 309$$*Total methylation $$n = 1415$$-mC + 5-hmC $$n = 70$$Age (years)31.53 (4.2)31.84 (4.1)31.80 (4.4)Parity (count)1.05 (1.0)1.01 (0.9)0.90 (0.8)Pre-pregnancy weight (kg)69.28 (15.2)70.65 (16.9)69.10 (15.7)Pre-pregnancy BMI25.42 (5.3)25.80 (5.7)25.51 (5.4)Marital statusMarried$81.6\%$ [252]$80.1\%$ [113]$75.7\%$ [53]Single$16.8\%$ [52]$19.1\%$ [27]$22.9\%$ [16]Unknown$1.6\%$ [5] < $1\%$ [1] < $1\%$ [1]Smoking statusNever$69.6\%$ [215]$75.1\%$ [106]$68.6\%$ [48]Quit before Pregnancy$12.6\%$ [39]$13.4\%$ [19]$12.9\%$ [9]Quit during Pregnancy$2.3\%$ [7]$2.8\%$ [4]$4.3\%$ [3]Smoked < 11 cigarettes/day$2.9\%$ [9]$2.1\%$ [3]$4.3\%$ [3]Unknown$12.6\%$ [39]$6.4\%$ [9]$10.0\%$ [7]Ethnicity and raceWhite, Non-Hispanic$81.9\%$ [253]$81.6\%$ [115]$72.9\%$ [51]Black or African American, Non-Hispanic$6.5\%$ [20]$6.4\%$ [9]$10.0\%$ [7]Hispanic$2.6\%$ [8]$2.1\%$ [3]$4.3\%$ [3]Asian, Non-Hispanic$4.5\%$ [14]$2.8\%$ [4]$2.9\%$ [2]Native American, Non-Hispanic$0.6\%$ [2] < $1\%$ [1]$1.4\%$ [1]Hawaiian or Pacific Islander$0.6\%$ [2]$0\%$ [0]$0\%$ [0]Multi-Ethnic/Racial$1.0\%$ [3]$2.1\%$ [3]$1.4\%$ [1]Unknown$2.3\%$ [7]$4.3\%$ [6]$5.7\%$ [4]Infant Characteristicsn = 288Sex Male$48.3\%$ [147]$48.9\%$ [69]$52.9\%$ [37] Female$51.0\%$ [139]$51.1\%$ [72]$47.1\%$ [33]Gestational Age at Birth (days)274.49 (12.5)276.73 (8.1)275.86 (9.1)Birthweight (grams)3414.3 (533.0)3435.0 (482.1)3351.79 (518.9)Fenton Z-Score0.09 (0.91)0.04 (0.91)−0.11 (1.0)PFAS Exposures^$$n = 206$$PFHxS (µg/L)3.40 (2.0)3.19 (1.6)3.08 (1.7)PFOS (µg/L)5.73 (2.9)5.25 (1.7)5.04 (1.8)PFOA (µg/L)1.35 (0.9)1.14 (1.9)1.14 (1.9)PFNA (µg/L)0.41 (0.2)0.36 (1.8)0.37 (1.8)PFDA (µg/L)0.16 (0.1)0.13 (1.8)0.12 (1.8)PFUnDA (% above the LOD)$41.3\%$ [85]$35.5\%$ [50]$38.6\%$ [27]MeFOSAA (% above the LOD)$34\%$ [70]$37.6\%$ [53]$42.9\%$ [30]5-hmC: 5-hydroxymethylcytosine; 5-mC: 5-methylcytosine; BMI: body mass index; LOD: limit of detection; MeFOSAA: 2-(N-methyl-perfluorooctane sulfonamido) acetic acid; PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOA: perfluorooctanoic acid; PFOS: perfluorooctanesulfonic acid; PFUnDA: perfluoroundecanoic acid.*This includes all women originally enrolled in the first trimester. Note that many were lost to follow-up or provided incomplete data. Of these, 288 stayed in the study and had a live, singleton birth at the study hospital (2 were missing information on infant sex). A subset of 206 participants with archived first trimester samples were selected for the PFAS analysis.^Numeric PFAS reported in geometric means. ## PFAS exposure assessment PFHxS, PFOS, PFOA, and PFNA were highly detected, with > $89\%$ of measurements above the LODs (Additional file 1: Table S1). PFDA was well detected, with $60\%$ of measurements above the LOD (Additional file 1: Table S1). All five of these PFAS were treated as numeric variables in analysis. Geometric mean concentrations were 3.2 µg/L (GSD: 1.6) for PFHxS; 5.3 µg/L (GSD: 1.7) for PFOS; 1.1 µg/L (GSD: 1.9) for PFOA; 0.37 µg/L (GSD: 1.8) for PFNA; and 0.12 µg/L (GSD: 1.8) for PFDA (Additional file 1: Table S1, Fig. S2). PFUnDA and MeFOSAA were moderately detected, with $35.5\%$ and $37.6\%$ of samples above the LOD (Additional file 1: Table S1). These were treated as categorical variables in the final analysis. Intercorrelation of PFAS (Additional file 1: Fig. S3) showed that there were significant correlations between PFHxS and PFOS (r2: 0.29, $p \leq 0.001$) and PFOA (r2: 0.49, $p \leq 0.001$); PFOS and PFDA (r2: 0.52, $p \leq 0.001$); and PFNA and PFOS (r2: 0.77, $p \leq 0.001$), PFOA (r2: 0.58, $p \leq 0.001$), and PFDA (r2: 0.73, $p \leq 0.001$). Using Chi-squared tests, PFUnDA and MeFOSAA were not related to other PFAS ($p \leq 0.05$). Parity was negatively correlated with concentrations of PFOS (r2: −0.40, $p \leq 0.001$) and PFOA (r2: −0.29, $p \leq 0.001$). Maternal age and pre-pregnancy BMI were not correlated with any PFAS ($p \leq 0.05$). Race and ethnicity, smoking status, and marital status each had associations with 1–2 PFAS: self-reported race as African American or Black (but not other self-reported race or ethnicities or missing self-reported race or ethnicity) was associated with PFUnDA below the LOD (χ2: 64, $p \leq 0.05$); smoking was associated with MeFOSAA below the LOD(χ2: 169, $p \leq 0.05$); and singleness was associated with MeFOSAA below the LOD (χ2: 188, $p \leq 0.001$). Due to these results, self-reported race was included in models as a variable for reporting African American or Black as a proxy for specific racism and racist policies that influence PFAS exposure burden in this self-identified group. ## PFAS and total DNA methylation In the models for DNA methylation, parity, self-reported race (as Black or African American), and smoking status were considered true confounders. Marital status was not included in the final model, as this variable has not been traditionally associated with effects or differences in infant DNA methylation. Beta regression models for total methylation across 744,926 CpG sites were fit for each PFAS, and genomic inflation factors for each model suggested minimal p-value inflation (Additional file 1: Table S3). Site-specific differences in total methylation were found for all PFAS modeled as continuous concentrations, including in 12 sites for PFHxS, 19 sites for PFOS, 2 sites for PFOA, 3 sites for PFNA, and 4 sites for PFDA (q < 0.05, Fig. 2, see Tables 2 and 3). Of these, total methylation of several CpG sites overlapped between PFAS—cg15429214 in an intergenic region of chromosome 22 was negatively associated with PFOS and PFNA; cg20360148 on the ATG2A (autophagy related 2A) gene was positively associated with PFOS, PFNA, and PFDA; and cg26157972 on an intergenic region of chromosome 5 was negatively associated with PFOA, PFOS, PFNA, and PFDA (Table 3). PFHxS was also positively associated with total methylation at two sites near the transcription start site of GTPBP3 (GTP binding protein 3, mitochondrial), but no other PFAS had significant sites in this gene. Fig. 2Significant total methylation sites ($$n = 141$$). Each row represents an individual PFAS. On the left, Manhattan plots describe the chromosomal location of all sites and corresponding Benjamini–Hochberg corrected p-value (q-value), with the sites significantly associated with PFAS labeled and noted with solid triangles. On the right, volcano plots depict all PFAS-site associations by effect estimate representing differences for each log-fold unit change in PFAS concentration or categorical PFAS detection and Benjamini–Hochberg corrected p-value (q-value), with the statistically significant sites noted in solid triangles. Dashed lines on both represent q-value = 0.05. Abbreviations: PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOA: perfluorooctanoic acid; PFOS: perfluorooctanesulfonic acidTable 2Number of Sites Significantly Associated with PFAS by Model (q < 0.05)Total Methylation ($$n = 141$$)Sex Interaction ($$n = 141$$)Females ($$n = 72$$)Males ($$n = 69$$)PFHxS12841781PFOS19987810PFOA2100PFNA3510PFDA4422PFUnDA0222MeFOSAA0100Counts of the significant number of DNA methylation sites related to each PFAS exposure. Sex interaction models were used to select the sites with evidence for sex-specific relationships. Models were then run at these sites, stratified by infant sex. The following models were fit, where bolded term indicates the estimate of interest generating the counts above:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} {\text{Total}}\;{\text{Methylation}}\;{\text{at}}\;{\text{744}},{\text{926}}\;{\text{sites}} & = \beta _{0} + \varvec{\beta} _{1} {\mathbf{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Smoking}} + \beta _{4} {\text{Race}} \\ & \quad + \beta _{5} {\text{CD}}4{\text{T}} + \beta _{6} {\text{CD}}8{\text{T}} + \beta _{7} {\text{GranCell}} \\ & \quad + \beta _{8} n{\text{RBC}} + \beta _{9} {\text{PC}}1 + \beta _{{10}} {\text{PC}}2 + \beta _{{12}} {\text{Sex}} \\ \end{aligned}$$\end{document}TotalMethylationat744,926sites=β0+β1PFAS+β2Parity+β3Smoking+β4Race+β5CD4T+β6CD8T+β7GranCell+β8nRBC+β9PC1+β10PC2+β12Sex\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} {\text{Sex}}\;{\text{Interaction}}\;{\text{Methylation}}\;{\text{at}}\;{\text{744}},{\text{926}}\;{\text{sites}} = & \beta _{0} + \beta _{1} {\text{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Smoking}} + \beta _{4} {\text{Race}} \\ & \quad + \beta _{5} {\text{CD}}4{\text{T}} + \beta _{6} {\text{CD}}8{\text{T}} + \beta _{7} {\text{GranCell}} + \beta _{8} n{\text{RBC}} \\ & \quad + \beta _{9} {\text{PC}}1 + \beta _{{10}} {\text{PC}}2 + \beta _{{12}} {\text{Sex}} + \varvec{\beta }_{{{\mathbf{13}}}} {\mathbf{Sex*PFAS}} \\ \end{aligned}$$\end{document}SexInteractionMethylationat744,926sites=β0+β1PFAS+β2Parity+β3Smoking+β4Race+β5CD4T+β6CD8T+β7GranCell+β8nRBC+β9PC1+β10PC2+β12Sex+β13Sex∗PFAS\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} {\text{Male}}\;{\text{or}}\;{\text{Female}}\;{\text{Methylation}}\;{\text{at}}\;{\text{Sites}}\;{\text{with}}\;{\text{Sex}}\;{\text{Interaction}} & = \beta _{0} + \varvec{\beta }_{{\mathbf{2}}} {\mathbf{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Smoking}} + \beta _{4} {\text{Race}} \\ & \quad + \beta _{5} {\text{CD}}4{\text{T}} + \beta _{6} {\text{CD}}8{\text{T}} + \beta _{7} {\text{GranCell}} \\ & \quad + \beta _{8} n{\text{RBC}} + \beta _{9} {\text{PC}}1 + \beta _{{10}} {\text{PC}}2 \\ \end{aligned}$$\end{document}MaleorFemaleMethylationatSiteswithSexInteraction=β0+β2PFAS+β2Parity+β3Smoking+β4Race+β5CD4T+β6CD8T+β7GranCell+β8nRBC+β9PC1+β10PC2CD4T: CD4 T lymphocytes; CD8T: CD8 T lymphocytes; GranCell: granulated cells; MeFOSAA: 2-(N-methyl-perfluorooctane sulfonamido) acetic acid; nRBC: nucleated red blood cells; PC: principal component representing batch effects; PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOA: perfluorooctanoic acid; PFOS: perfluorooctanesulfonic acid; PFUnDA: perfluoroundecanoic acidTable 3Sites with significant associations between PFAS and total methylation (q < 0.05, $$n = 141$$)PFASIllumina CpG NameEstimateSELower $95\%$ CIUpper $95\%$ CIp-valueBHq-valueChromosome: PositionUCSC Gene NamePFHxScg05645702−0.6920.094−0.876−0.5081.71E−116.37E−06chr2:242190905HDLBPcg032826860.5330.0880.3610.7051.56E−080.003chr4:177116826SPATA4cg21555240−0.5980.109−0.812−0.3841.93E−070.016chr5:89734950cg08313040 < 0.001 < 0.001 < 0.001 < 0.0016.70E−394.99E−33chr6:33092243HLA-DPB2cg11172857−0.6590.109−0.873−0.4451.73E−080.003chr6:168956903SMOC2cg18850427 < 0.001 < 0.001 < 0.001 < 0.0012.82E−070.021chr7:4009036SDK1cg02124746−0.6660.124−0.909−0.4233.54E−070.024chr9:91615137S1PR3cg00267172−0.4790.079−0.634−0.3241.14E−080.003chr10:11487791cg20139630−0.2220.040−0.300−0.1441.48E−070.014chr15:40476003BUB1Bcg02249969−0.2800.050−0.378−0.1821.11E−070.012chr17:1969338SMG6cg081273480.1970.0350.1280.2661.04E−070.012chr19:17448311GTPBP3cg268530930.6160.1160.3890.8434.67E−070.029chr19:17448469GTPBP3PFOScg129094550.5950.0950.4090.7814.91E−090.001chr1:32458635cg018486600.5090.0860.3400.6782.45E−080.003chr2:68269960C1Dcg26157972−0.6760.107−0.886−0.4664.49E−090.001chr5:1049232cg216850540.4490.0830.2860.6122.96E−070.018chr5:169810494KCNMB1cg097819870.5390.0970.3490.7291.58E−070.012chr6:4828434CDYLcg196371770.001 < 0.001 < 0.0010.0032.64E−080.003chr6:109417087C6orf182cg168323020.4430.0820.2820.6043.30E−070.018chr6:169689336cg173418790.3970.0740.2520.5424.22E−070.021chr7:75690308MDH2cg071076330.3180.0590.2020.4343.45E−070.018chr8:40960365cg209431550.5420.1030.3400.7446.33E−070.028chr10:118031958GFRA1cg069315910.3850.0730.2420.5285.66E−070.026chr10:118980094cg063549840.5870.0900.4110.7631.41E−094.60E−04chr10:128211107C10orf90cg027194270.002 < 0.001 < 0.0010.0041.85E−094.60E−04chr11:2151725INS-IGF2; IGF2cg203601480.6000.0920.4200.7801.69E−094.60E−04chr11:64685078ATG2Acg097932690.4320.0840.2670.5979.61E−070.040chr12:105348269cg239146940.1860.0340.1190.2532.51E−070.017chr12:132832250GALNT9cg112801850.4440.0800.2870.6011.42E−070.012chr16:237270cg186800350.2670.0520.1650.3691.08E−060.042chr19:6818326VAV1cg15429214−0.4480.078−0.601−0.2956.43E−080.006chr22:43166281PFOAcg26157972−0.5690.097−0.759−0.3793.95E−080.029chr5:1049232cg06537609−0.1400.025−0.189−0.0911.09E−070.040chr5:176217086PFNAcg26157972−0.7110.109−0.925−0.4971.39E−090.001chr5:1049232cg203601480.5520.0940.3680.7364.08E−080.015chr11:64685078ATG2Acg15429214−0.4240.075−0.571−0.2771.06E−070.026chr22:43166281PFDAcg26157972−0.6020.089−0.776−0.4285.30E−101.32E−04chr5:1049232cg203601480.4910.0760.3420.6402.12E−093.95E−04chr11:64685078ATG2Acg036472330.003 < 0.0010.0010.0052.33E−118.68E−06chr11:117387430DSCAML1cg03958076 < 0.001 < 0.001 < 0.001 < 0.0011.25E−349.30E−29chr22:41304942XPNPEP3Because beta regressions (logit link functions) were used to model differences in DNA methylation, estimates and SEs for methylation differences representing each log-fold unit change in PFAS concentration or categorical PFAS detection can be approximated by exp(estimate), exp(SE), or exp(CI)BH: Benjamini–Hochberg; CI: confidence interval; CpG: cytosine-guanine site where methylation was measured; PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOA: perfluorooctanoic acid; PFOS: perfluorooctanesulfonic acid; SE: standard error When comparing these estimates to previously published associations between prenatal PFAS and neonatal total methylation, there were not strong similarities (Additional file 1: Fig. S4). There was no overlap in significant CpG sites by PFAS identified in the present and former studies at q < 0.05 (Additional file 2: Table S1). At a raw p-value < 0.05 for the present study, only a few associations replicated in the same direction as previous PFAS epigenome-wide association studies: PFOS with increased methylation in ANO3 (cg05146852) [42] and PFNA with decreased methylation in HIF1A (cg04509825) and TTL (cg03065329) and increased methylation in PTGIS (cg27059136) and USP19 (cg01673931) [41]. When examining the sex-specific differences associated with each PFAS, all PFAS had at least one site with a significant sex interaction (q < 0.05, Table 2 and Additional file 2: Table S2). After stratifying by female and male infants and running models for these sites, PFHxS, PFOS, PFNA, PFDA, and PFUnDA all had significant sex-specific associations with total methylation in at least one CpG site (q < 0.05, Tables 2, Additional file 1: Tables S4, S5). In females, PFHxS had 17 significant sites, PFOS had 78 sites, PFNA had 1 site, PFNA had 2 sites, and PFUnDA had 2 sites (Additional file 1: Table S4). In males, PFHxS had 81 significant sites, PFOS had 10 sites, PFDA had 2 sites, and PFUnDA had 2 sites (Additional file 1: Table S5). Sex interactions in PFHxS and PFOS were primarily driven by males and females, respectively. Within each sex, most of the significant CpG sites were isolated to unique genes, with the exception of three genes with multiple CpG sites in females that were associated with PFOS (C2orf78, chromosome 2 open reading frame 78; SPATS2L, spermatogenesis associated serine rich 2 Like; RAP1GAP2, RAP1 GTPase activating protein 2), four genes with multiple CpG sites in males that were associated with PFHxS (SPATA4, spermatogenesis associated 4; AGPAT1, 1-acylglycerol-3-phosphate O-acyltransferase 1; RNF5, ring finger protein 5; RNF5P1, RNF5-pseudogene 1), and one gene that was associated with both PFHxS and PFOS in males (S1PR3, sphingosine-1-phosphate receptor 3). AGPAT1, RNF5, and RNF5P1 are located near each other on chromosome 6 and were highly associated with PFHxS in males, potentially representing a region of interest that is sensitive to total methylation changes in males. Specific region-wide analyses or pathway analyses, however, were unable to be conducted in any total methylation analysis, as there were too few significant sites overall. For total DNA methylation ($$n = 141$$), the relationship between each PFAS with 744,926 loci were assessed using beta regressions as outlined above. Results were considered to be statistically significant with a BH false discovery rate (FDR) cut-off of $q = 0.05.$ To examine potential sex-specific effects, additional models were run including an interaction term for sex x PFAS. CpG sites with significant interaction terms at a BH-corrected q-value of 0.05 were then stratified by sex. ## PFAS and 5-mC and 5-hmC Using the model outlined above for the interaction of PFAS with type of methylation (5-mC and 5-hmC), beta regression models across 528,389 CpG sites in the genome were fit for each PFAS ($$n = 70$$), and genomic inflation factors for each model suggested minimal inflation (Additional file 1: Table S6). After filtering sites that had an interaction by methylation type, over 15,000 sites were identified for stratification (q < 0.2), including 105 for PFHxS; 1,516 for PFOS; 637 for PFOA; 2,281 for PFNA; 8,054 for PFDA; 3,103 for PFUnDA; and 272 for MeFOSAA (Table 4, Additional file 2: Table S3). Each of these sites was stratified for methylation type, and 5,036 and 13,376 of these sites had significant associations between a PFAS with 5-mC or 5-hmC, respectively (q < 0.05, see Table 4). Within each PFAS, there were more sites associated with differences in 5-hmC as compared to 5-mC (Additional file 2: Tables S4 and S5). The majority of significant sites had decreased 5-hmC (75 sites for PFHxS, 1,289 for PFOS; 1,534 for PFNA; 7,234 for PFDA; 2,367 for PFUnDA; and 229 for MeFOSAA) and increased 5-mC (64 for PFHxS; 23 for PFOS; 812 for PFNA; 3,455 for PFDA; 338 for PFUnDA; and 140 for MeFOSAA, see Fig. 3). For all PFAS but PFOA, there were more significant associations with 5-hmC, when compared to 5-mC (Table 4).Table 4Number of significant associations between PFAS with 5-mC and/or 5-hmC ($$n = 70$$)5-mC/5-hmC Interaction(q < 0.2)5-mC Only(q < 0.05)5-hmC Only(q < 0.05)PFHxS10580101PFOS1516241388PFOA63710PFNA22818791684PFDA805435277432PFUnDA31033612507MeFOSAA272164264Counts of the number of CpG sites with statistically significant associations between each PFAS and methylation ($$n = 70$$ for all analyses). 5-mC/5-hmC-interaction models were used to select the sites to stratify by each type of methylation. The following models were fit, where bolded term indicates the estimate of interest generating the counts above:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} 5 - {\text{mC}}\;{\text{or}}\;5 - {\text{hmC}}\;{\text{proportion}}\;{\text{at}}\;{\text{528}},{\text{389}}\;{\text{Sites}} & = \beta _{0} + \beta _{1} {\text{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Smoking}} + \beta _{4} {\text{Race}} \\ & \quad + \beta _{5} {\text{sex}} + \beta _{6} {\text{CD}}4{\text{T}} + \beta _{7} {\text{CD}}8{\text{T}} + \beta _{8} {\text{GranCell}} \\ & \quad + \beta _{9} n{\text{RBC}} + \beta _{{10}} {\text{PC}}1 + \beta _{{11}} {\text{PC}}2 + \beta _{{12}} {\text{Type}} \\ & \quad + \varvec{\beta }_{{{\mathbf{13}}}} {\mathbf{Type*PFAS}} + [1|ID] \\ \end{aligned}$$\end{document}5-mCor5-hmCproportionat528,389Sites=β0+β1PFAS+β2Parity+β3Smoking+β4Race+β5sex+β6CD4T+β7CD8T+β8GranCell+β9nRBC+β10PC1+β11PC2+β12Type+β13Type∗PFAS+[1|ID]\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$5 - {\text{mC}}\;{\text{or}}\;5 - hmC\;{\text{Methylation}}\;{\text{at}}\;{\text{Sites}}\;{\text{with}}\;{\text{Type}} - {\text{Interaction}} \, = \, \beta _{0} + \varvec{\beta }_{{\mathbf{1}}} {\mathbf{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Smoking}} + \beta _{4} {\text{Race}} + \beta _{5} {\text{CD4T}} + \beta _{6} {\text{CD}}8{\text{T}} + \beta _{7} {\text{GranCell}} + \beta _{8} n{\text{RBC}} + \beta _{9} {\text{PC}}1 + \beta _{{10}} {\text{PC}}$$\end{document}5-mCor5-hmCMethylationatSiteswithType-Interaction=β0+β1PFAS+β2Parity+β3Smoking+β4Race+β5CD4T+β6CD8T+β7GranCell+β8nRBC+β9PC1+β10PCCD4T: CD4 T lymphocytes; CD8T: CD8 T lymphocytes; GranCell: granulated cells; MeFOSAA: 2-(N-methyl-perfluorooctane sulfonamido) acetic acid; nRBC: nucleated red blood cells; PC: principal component representing batch effects; PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOA: perfluorooctanoic acid; PFOS: perfluorooctanesulfonic acid; PFUnDA: perfluoroundecanoic acidFig. 3Number and direction of significant associations between PFAS and 5-methylcytosine (5-mC) and 5-hydroxymethylcytosine (5-hmC) after stratification of loci with a significant PFAS: type of methylation interaction ($$n = 70$$). A shows all sites with a significant association between each PFAS and 5-mC (q < 0.05) and B shows the 5-hmC sites (q < 0.05). Counts above the horizontal line at zero represent the number of sites that were positively associated with each PFAS, and counts below the horizontal line represent the number of sites that were negatively associated with each PFAS. Abbreviations: MeFOSAA: 2-(N-methyl-perfluorooctane sulfonamido) acetic acid; PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOS: perfluorooctanesulfonic acid; PFUnDA: perfluoroundecanoic acid Accordingly, there was little overlap in CpG sites or genes between PFAS for 5-mC; only the gene RPS6KA2 (ribosomal protein S6 kinase A2) had overlap between more than 3 PFAS (PFNA, PFDA, PFUnDA, and MeFOSAA, see Additional file 1: Fig. S5). There were some CpG sites that overlapped between 2 PFAS, but no CpG was universally associated with all PFAS (Fig. 4A). Similarly, there were several genes with significant differences in 5-hmC that were shared among PFAS, but no single gene was shared among all PFAS (Fig. 4B). Hydroxymethylation in three genes was associated with at least five of the seven PFAS (see Fig. 4C), including SHANK2 (SH3 and multiple ankyrin repeat domains 2), PARD3 (par-3 family cell polarity regulator), and MYH9 (myosin heavy chain 9). There were 88 other genes that had differences in hydroxymethylation associated with four PFAS, with the majority (73 of 88) shared between PFOS, PFNA, PFDA, and PFUnDA (Fig. 4B, Additional file 2: Table S4). For individual PFAS, sites with differences in 5-hmC were similarly distributed across gene locations (Additional file 1: Fig. S6), but there were some discrepancies in the distribution of sites in relation to CpG islands (Additional file 1: Fig. S7), with proportionately fewer significant 5-hmC sites in actual CpG islands, when compared to all sites included on the EPIC array. Fig. 4Overlap for significant 5-hydroxymethylcytosine (5-hmC, q < 0.05) sites (A) and genes (B) by PFAS ($$n = 70$$). Each plot and Venn diagram inset shows the overlap between all PFAS in the present study. C Lists the three genes that overlapped between at least five PFAS. β corresponds the beta regression coefficient estimate representing exp(estimate) differences for each log-fold unit change in PFAS concentration or categorical PFAS detection, p represents the uncorrected p-value, and q represents the Benjamini–Hochberg corrected q-value. Abbreviations: MeFOSAA: 2-(N-methyl-perfluorooctane sulfonamido) acetic acid; MYH9: myosin heavy chain 9; PARD3: Par-3 family cell polarity regulator; PFAS: per-/polyfluoroalkyl substances; PFHxS: perfluorohexanesulphonic acid; PFDA: perfluorodecanoic acid; PFNA: perfluorononanoic acid; PFOS: perfluorooctanesulfonic acid; PFUnDA: perfluoroundecanoic acid; SHANK2: SH3 and multiple ankyrin repeat domains 2 Regional differences in 5-hmC were identified using results from hydroxymethylation models for each PFAS associated with at least 100 sites. PFDA was the only PFAS with significant regions (q < 0.05). Top regional differences are reported in Table 5.Table 5Top regional 5-hmC differences related to PFDA exposureChromosomePosition startPosition endp-valueBH q−valueNumber probesGene name147780564477839258.74E−076.92E−044TRABD2B697600110976029652.82E−069.56E−044ENSG00000271860.91977582997765072.50E−075.24E−043CLSTN11278682326786832674.41E−075.24E−043488736370887383961.60E−067.20E−043FAM13A7978718697891874.58E−060.00123679762944797651191.09E−050.00231761431976175631.53E−050.0023CAMTA1619619236196201881.54E−050.0023LNC-LBCS412758255127611411.58E−050.0023695632160956329371.64E−050.0023584773408847741041.78E−050.0023293797164937981432.18E−050.002234964918496509273.28E−050.00243ENSG00000287117.1799653239996559233.68E−050.00243CYP3A5773708261737111185.92E−050.00283STX1A872631834726319371.19E−040.00333KCNB2182669840826728494.25E−040.00493ENSG00000233290.4332696525326992806.69E−040.00553CNOT107469749646988807.72E−040.00583FOXK1Describes the chromosomal locations of significant regions of 5-hmC related to PFDA exposures. p represents the uncorrected p-value, and q represents the Benjamini–Hochberg (BH) corrected q-value KEGG pathway analysis was conducted using results from hydroxymethylation models for each PFAS associated with at least 100 sites. While no pathway met a q < 0.05 cutoff for enrichment, each PFAS had several pathways that were of interest ($p \leq 0.05$ with at least 2 differentially hydroxymethylated genes in the pathway, Additional file 1: Table S7). Within this analysis, there were several overlapping functions and specific pathways that were associated with differential hydroxymethylated genes across many PFAS. *In* general, the most common classifications that were associated with hydroxymethylation differences with any PFAS were in neuroendocrine system pathways. Within specific KEGG pathways, the glutamatergic synapse pathway was enriched among genes associated with PFHxS and PFOS; Huntington disease was associated with genes from the PFOS and PFNA models; insulin secretion, gonadotropin-releasing hormone (GnRH) secretion, and high-affinity IgE receptor (FcεRI) signaling was enriched among genes from the PFNA and PFDA models ($p \leq 0.05$). ## PFAS and birth outcomes: mediation by epigenetics Across all families with passing total DNA methylation data ($$n = 141$$), there were several relationships suggestive of significance ($p \leq 0.1$) between birth outcomes (gestational age at birth, or Fenton z-scores) and PFAS exposures, when controlling for the necessary confounders for statistical mediation analyses (Table 6). When controlling for parity, self-reported race, and smoking status, concentrations of PFHxS were related to decreased Fenton z-score (β = −0.25, $$p \leq 0.036$$). There were also negative relationships of both PFNA (β = −0.31, $$p \leq 0.089$$) and PFUnDA (β = −0.46, $$p \leq 0.019$$) with gestational age (measured in weeks).Table 6Associations between PFAS and birth outcomes ($$n = 141$$)PFASGestational age (weeks)Fenton Z-score: size-for-gestational age, corrected by sexPFHxSβ = 0.04 ± 0.21p = 0.85β = −0.25 ± 0.16p = 0.036PFOSβ = −0.160 ± 0.19p = 0.40β = −0.12 ± 0.15p = 0.41PFOAβ = −0.056 ± 0.17p = 0.75β = −0.036 ± 0.14p = 0.79PFNAβ = −0.31 ± 0.19p = 0.089β = −0.016 ± 0.15p = 0.91PFDAβ = −0.27 ± 0.17p = 0.114β = 0.13 ± 0.13p = 0.383PFUnDAβ = −0.46 ± 0.19p = 0.019β = 0.010 ± 0.16p = 0.95MeFOSAAβ = 0.15 ± 0.20p = 0.45β = −0.059 ± 0.16p = 0.71Relationships with a p-value < 0.1 were considered for mediationMeFOSAA—2-(N-methyl-perfluorooctane sulfonamido) acetic acid; PFAS—per-/polyfluoroalkyl substances; PFHxS—perfluorohexanesulphonic acid; PFDA—perfluorodecanoic acid; PFNA—perfluorononanoic acid; PFOA—perfluorooctanoic acid; PFOS—perfluorooctanesulfonic acid; PFUnDA—perfluoroundecanoic acid. Coefficients representing differences for each log-fold unit change in PFAS concentration or categorical PFAS detection and p-values from linear regressions assessing the relationship between PFAS and gestational age, birthweight, or Fenton z-score, using the following linear regressions:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{Gestational}}\;{\text{Age}} = \beta _{0} + \beta _{1} {\text{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Race}} + \beta _{4} {\text{Smoking}}$$\end{document}GestationalAge=β0+β1PFAS+β2Parity+β3Race+β4Smoking\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\text{Fenton}}\;Z - {\text{Score}} = \beta _{0} + \beta _{1} {\text{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Race}} + \beta _{4} {\text{Smoking}}$$\end{document}FentonZ-Score=β0+β1PFAS+β2Parity+β3Race+β4Smoking To reduce the number of methylation sites assessed for mediation, only significant sites (q < 0.05) from total methylation, 5-mC, and 5-hmC analyses associated with the three birth outcome-associated PFAS (PFHxS, PFNA, and PFUnDA) were considered (Additional file 2: Table S6). For PFHxS, five 5-mC sites and six 5-hmC sites were related to Fenton z-score ($p \leq 0.05$) and selected for mediation analysis. For PNFA, two total methylation CpG sites, 37 5-mC sites, and 26 5-hmC sites were related to gestational age and selected for mediation. For PFUnDA, there were 60 sites for 5-mC and 57 sites for 5-hmC that were considered for mediation on gestational age. *For* gene-wise sites as well as the collective group of sites within a methylation type and for an individual PFAS, kernel machine regression was used to assess the nonlinear, gene-wise mediation effects of methylation on the relationship between the PFAS and birth outcomes. After applying Benjamini-Hochberg (BH) corrections to the p-values, no single gene met a q < 0.05 for any exposure (Additional file 2: Table S7). *Two* genes with one CpG site each (INADL, or the PATJ crumbs cell polarity complex component, $q = 0.07$; LOC100506023, $q = 0.07$) were suggestive of a nonlinear mediation effect of 5-hmC between PFNA and gestational age. When assessing the total mediation of combined effects of all CpG sites included for each relationship (e.g., each PFAS, birth effect, and type of methylation), there were significant mediation effects of 37 5-mC sites and 26 5-hmC sites between PFNA exposure and gestational age ($q = 1.28$E−05 and $q = 1.28$E−05), as well as both 60 5-mC sites and 57 5-hmC sites between PFUnDA exposure and gestational age ($q = 0.026$ and $q = 7.15$E−04; see Table 7). Many of the genes these mediating CpG sites were in functions related to either cell proliferation and viability or apoptosis and cell death. Among potential mediators, there were several genes shared among the methylation types for PFNA (VTI1B, vesicle transport through interaction with T-SNAREs 1B, and LOC100506023) and PFUnDA (RPIA, ribose 5-phosphate isomerase A, GTF3C2, general transcription factor IIIC subunit 2, SDK1, sidekick cell adhesion molecule 1, TLE3, TECPR2, tectonin beta-propeller repeat containing 2, ERN2, endoplasmic reticulum to nucleus signaling 2, LOC284395). Similarly, there was one gene shared across exposures for 5-mC (AKR7A3, aldo–keto reductase family 7 member A3), and there were two genes shared for 5-hmC (HEATR3, HEAT repeat containing 3, and GSDMA, gasdermin A).Table 7Mediation by methylation across multiple lociExposure, mediator, and outcomeNumber of CpG Sitesp-ValueBenjamini–Hochberg q-valuePFHxS, 5-mC, and Fenton-Z-Score50.03930.136PFHxS, 5-hmC, and Fenton-Z-Score60.2410.241PFNA, Total methylation, and gestational age20.009130.0738PFNA, 5-mC, and gestational age371.93E-071.28E-05PFNA, 5-hmC, and gestational age263.72E-071.28E-05PFUnDA, 5-mC, and gestational age604.57E-040.0263PFUnDA, 5-hmC, and gestational age576.22E-067.15E-04Kernel machine regression results using all significant CpG sites meeting the following filter criteria: CpG related to PFAS exposure (q < 0.05), PFAS related to birth outcome ($p \leq 0.1$), and CpG related to birth outcome ($p \leq 0.05$). *Individual* gene-wise results are described in Additional file 2: Table S7. $$n = 141$$ for total methylation and $$n = 70$$ for 5-mC and 5-hmC ## Discussion PFAS are widespread environmental contaminants that are actively impacting human health, with known effects on reproduction, immune and metabolic function, and development (for review, see Fenton et al. [ 48]). One mechanism that may underlie these effects is aberrant epigenetic programming, which has been observed in laboratory models and in human epidemiological cohorts. In our investigation of associations between PFAS and DNA methylation, we posited that any epigenetic differences may mediate the relationship of PFAS and adverse birth outcomes. Across our analyses, our hypothesis was largely supported; we found significant relationships between PFAS and DNA methylation (total, 5-mC, and 5-hmC), as well as PFAS and birth outcomes (decreased gestational age and Fenton z-scores for size at birth). Even more, we found a significant mediation effect of both 5-mC and 5-hmC (but not total methylation) on the relationship between both PFNA and PFUnDA and decreased gestational age at birth, demonstrating the mediation effects of not just general epigenetic differences, but specific types of DNA methylation marks, on the relationship between developmental PFAS exposure and birth outcomes. These results strengthen the known evidence of the relationship of developmental PFAS exposure and early-life epigenetic differences [40–46], despite variations in both the study populations and PFAS exposure levels. One of the most frequently reported genes differentially methylated in epigenome-wide developmental studies of PFAS exposure is TNXB (tenascin-XB) [45, 46]. Presently, differences in only either 5-mC or 5-hmC, and not total methylation, were observed in this gene, suggesting that there are likely subtle differences in the type of methylation that could contribute to discordantly methylated genes observed across study populations. TNXB is also highly represented on Illumina arrays, which could contribute to commonalities that were previously reported [40]. These differences may be further exacerbated by sex-specific differences in epigenetics, which we were presently unable to assess in 5-mC or 5-hmC due to small sample sizes. Future studies should prioritize studies that are large enough to investigate these potentially important sex-specific effects in specific methylation types. Presently and to the best of our knowledge, this is the hallmark study investigating prenatal PFAS exposure and 5-mC/5-hmC specifically. Results reported here show striking patterns of reductions in 5-hmC with concomitant increases in 5-mC, across six of the seven PFAS included. Differences in methylation that broadly occur across regions or within regulatory elements are more likely to be associated with gene expression changes [49, 50]. In a regional analysis of 5-hmC, we did find some genes that had broad regions of differential hydroxymethylation. These regions were often associated with regulatory elements, such as known gene enhancers or histone modifications. Compared to 5-mC, 5-hmC is proposed to be more closely linked to histone modifications and gene regulation [51]. Our data broadly supports this, but the EPIC array only covers loci in a small portion of known regulatory elements [52]. Follow-up with other methods, such as hydroxymethylated DNA immunoprecipitation sequencing (hMeDIP-Seq) or nano-hmC-Seal, could better investigate 5-hmC differences in important regulatory elements. Some of these differences were also mediators of the relationship between PFAS and decreased gestational age, with trends towards mediating PFAS-related decreases in size at birth. These two birth outcomes may be early indicators of adverse neurocognitive [53, 54] and behavioral/emotional [55] effects later in life. In the brain, 5-hmC is thought to modulate mammalian postnatal neurodevelopment, with marked increases from early postnatal stages to adulthood [56]. Pathway analyses of 5-hmC genes suggested that there were differences in the status of genes related to general cell processes, as well as functions in the endocrine, immune, and nervous systems. While the evidence of the cognitive effects of prenatal PFAS exposure is presently mixed [57–61], other health effects resulting from gestational PFAS exposure are well documented [10, 19, 20, 48]. Given enrichment of pathways relevant to neurological function, the endocrine system, and insulin secretion among the PFAS-associated genes, the role of 5-hmC in the development of long-term adverse health outcomes is an essential area for future investigation. PFAS may disrupt epigenetic programming through the widespread dysregulation of epigenetic machinery and/or other PFAS mechanisms of toxicity. A primary mechanism of interest is PFAS-induced oxidative stress [20], which has been widely documented in vitro [62–65] and in human epidemiological cohorts [66–68]. While some researchers have suggested that this relationship could lead to genotoxicity or cytotoxicity [63], others have observed PFAS-induced oxidative stress without any evidence of either of these effects [64]. Seminal work with mice and cells demonstrated that oxidative stress may directly alter TET enzymes that are responsible for the formation of 5-hmC, leading to widespread decreases in 5-hmC across the genome [51]. Changing TET activity may be a compensatory mechanism to combat the deleterious effects of oxidative stress that may also be connected with alterations in the hydroxymethylation of noncoding RNAs that could contribute to epigenetic regulation. Interestingly, many other researchers have noted differences in noncoding RNAs that were related to PFAS exposure [69–72]. While we did not investigate these other epigenetic regulators, there were many genes encoding noncoding RNAs with significant associations between PFAS and 5-hmC. This complex interplay of epigenetic machinery, oxidative stress, and endpoints (i.e., 5-hmC) is an essential area to understand the molecular mechanisms underpinning toxicity by environmental contaminants. Overall, results in this manuscript detail a compelling and complex interplay of early-life exposures to PFAS, specific differences in DNA methylation types, and adverse birth outcomes. The link to birth outcomes is particularly important for public health, as this cohort represents a group of mothers and infants who were healthy at birth (full term, no known complications). Because cases of extreme preterm birth or other severe birth outcomes were excluded from the cohort, our results indicate that there are subtle, but important relationships between PFAS, epigenetics, and birth outcomes. Exploration was limited by small sample sizes available for these measures, as well as the time points and tissue types available for assessment. One time point of particular interest to the present study is that of exposure; PFAS were measured in blood samples from early maternal pregnancy which is known to have levels much higher than fetal tissues [73]. These levels, however, remain consistent and/or increase in fetal tissues over time. Because we were measuring DNA modifications that are rewritten in the early embryonic stages, additional research is needed to clarify the maternal–fetal kinetics of these PFAS and the relationship of kinetics with epigenetic differences and birth outcomes. Additionally, while we were able to delineate PFAS-related effects between specific methylation types (5-mC and 5-hmC), we were unable to assess the sex-specific effects in these markers. Our methods also selected for the probes with higher levels of 5-hmC and those that had an interaction term between 5-mC and 5-hmC, which could be contributing to the large number of significant sites observed in the separate 5-mC and 5-hmC analyses. As total methylation related to PFAS has sex-specific differences and the Illumina EPIC array selects for sites that may not best represent important environmentally induced changes [74], additional work should prioritize large enough sample sizes and appropriate methods needed to confirm these results. Due to the small sample size, we also selected for other precision variables, such as cell types. Research with larger sample sizes that allow for the inclusion of other important precision variables is warranted. In this manuscript, we were able to assess seven unique PFAS, but thousands more exist that humans may be exposed to. There has been some speculation that different PFAS moieties may have varying mechanisms of action, additive/multiplicative effects, or cumulating effects that should be considered [75, 76]. Finally, while there has been evidence of total methylation differences related to PFAS exposures across many types of cohorts [41–46], our cohort was rather demographically homogeneous and did not capture the full implications of health disparities caused by racism and structural policies. In particular, we found that PFAS exposure was only associated with self-reported race as Black or African American, which likely does not capture all structural factors that affect public health and birth outcomes. Going forward, research should continue to expand on the significant findings here, while also addressing these limitations, to best understand the health effects from these important environmental contaminants. This is the first report to our knowledge of widespread 5-hmC alterations by prenatal PFAS exposure. These results were observed in a healthy birth cohort with modest PFAS exposures, suggesting that the developmental epigenetic impacts from PFAS may be sufficiently concerning in the general population, especially in populations with higher exposure burdens. As we continue to develop our understanding of PFAS, the long-term impacts of the demonstrated relationship should be carefully considered. ## Study population and sample collection The MMIP is a birth cohort study based out of the University of Michigan Von Voigtlander Women’s Hospital, which recruited participants from 2010 to 2019 [77, 78]. Participants were recruited during their first prenatal visit if they were at least 18 years old, had a singleton pregnancy, were between 8 and 14 weeks gestation, and intended to deliver at the University of Michigan Hospital. All participants provided informed, written consent prior to study enrollment. The University of Michigan Medical School Institutional Review Board approved all study procedures (Approval no. HUM00017941). Venous blood samples were collected during the first trimester prenatal visit by trained phlebotomists. Blood samples were collected in BD Vacutainer EDTA-preserved tubes (Becton, Dickinson and Company) and were centrifuged at 1000 relative centrifugal force for 10 min. Separated plasma was then stored at -80 ºC until analysis. Shortly after delivery, cord blood was collected via standard venipuncture into EDTA-preserved tubes (Becton, Dickinson and Company) and PaxGene DNA and RNA tubes (PreAnalytix). EDTA tubes were centrifuged, and plasma was aliquoted into tubes. Plasma, DNA tubes, and RNA tubes were frozen at -80 ºC until further processing. At that time, samples were selected for laboratory analysis, 309 participants had been recruited during the first trimester, of which 288 had remained in the study and given birth at the University of Michigan Hospital; 206 of these participants that had an adequate volume of first trimester plasma collected and were selected for PFAS analysis. Among these, DNA methylation was completed on those with cord blood samples preserved for DNA isolation. Those that passed stringent quality control (see below, DNA Methylation Processing, $$n = 141$$) were included (see study schematic in Fig. 1). ## Survey and birth outcome data At the first trimester visit, maternal weight and BMI were ascertained from the medical records and a survey was administered to collect information on demographics, including race, ethnicity, age, parity, marital status, and smoking status. For analysis, smoking status was considered as binary of either smoking reported in pregnancy or not. Race was collected as self-described racial and ethnic identities. For analysis, race was included as proxy for racism and racist policies that are related to health. Parity was considered as a numerical variable. At delivery, data were collected from the health records on infant sex, gestational age, and birth anthropometry. Gestational age was recorded as the healthcare provider’s best estimate from either the last menstrual period or ultrasound, as recommended by the American Congress of Obstetricians and Gynecologists. Fenton z-scores were calculated from infant sex, birthweight, and gestational age using the PediTools website (https://peditools.org/fenton2013/ [79]). ## PFAS analysis Concentrations of nine PFAS were quantified in first trimester maternal plasma samples through the NSF International laboratory (Ann Arbor, MI). The measured PFAS were: 2-(N-methyl-perfluorooctane sulfonamido) acetic acid (MeFOSAA), perfluorooctanesulfonamide (PFOSA), perfluorohexanesulphonic acid (PFHxS), perluoroheptanoic acid (PFHpA), perfluorooctanoic acid (PFOA), perfluorooctanesulfonic acid (PFOS), perfluorodecanoic acid (PFDA), perfluorononanoic acid (PFNA), and perfluoroundecanoic acid (PFUnDA). Concentrations were measured via a method based off the US Centers for Disease Control and Prevention (CDC) Polyfluoroalkyl Chemicals Method Laboratory Procedure 6304.1 [80]. This method uses on-line solid phase extraction coupled with high-performance liquid chromatography–isotope dilution tandem mass spectrometry. Analysis was performed using a Thermo Scientific Transcend TXII Turbulent Flow system (ThermoFisher Scientific) interfaced with Thermo Scientific Quantiva triple quadrupole mass spectrometer (ThermoFisher Scientific) using MRM in negative mode. The method incorporates calibration curve checks and known standards interspersed with study samples to ensure accuracy and precision. The limits of detection (LOD) were established by replicate injections of low concentration standards (Additional file 1: Table S1). The laboratory was part of the National Institute of Health Children’s Health Exposure Analysis Resource network (NIH CHEAR) at the time and participated in inter-lab quality control and quality assurance [81]. ## DNA methylation analysis DNA was isolated from nucleated cells of cord blood (leukocytes and nucleated red blood cells) via a Paxgene DNA isolation kit (PreAnalytix) according to the manufacturer’s protocol. DNA concentrations were quantified via the Quant-IT Picogreen double stranded DNA assay (ThermoFisher Scientific). DNA methylation profiles were assessed with the Infinium MethylationEPIC BeadChip (Illumina), which covers over 850,000 unique methylation sites (CpG sites) [82]. DNA was bisulfite converted using the EZ DNA Methylation kit (Zymo Research) and assayed on the BeadChip. As bisulfite conversion quantifies total methylation and cannot distinguish between 5-mC and 5-hmC, this traditional analysis can be considered ‘total methylation’ (5-mc + 5hmC). To profile 5-mC and 5-hmC individually in a subset of samples ($$n = 90$$ before quality control), parallel bisulfite conversion and oxidative bisulfite conversion was performed using the Nugen TrueMethyl oxBS Module (NuGEN Technologies, Inc.). In this protocol, samples are oxidized, converting 5-hmC to 5- formylcytosine (5-fC), which is then converted to uracil following bisulfite treatment, leaving only 5-mC as cytosine residues. Following both bisulfite and oxidative bisulfite treatments, samples were randomized across chips and chip positions, hybridized to BeadChips, and signals were read at the University of Michigan Advanced Genomics Core. For samples with traditional bisulfite conversion, the final data consist of average betas representing the proportion of total methylated cytosine (5-mC + 5-hmC) for each site. For sample aliquots undergoing oxidative treatment, final data consist of betas representing the proportion of 5-mC only. This procedure has been used to generate 5-hmC data from BeadChips by other epidemiological studies [83–85]. Instead of simply subtracting 5-mC from total DNA methylation at each CpG site to estimate 5-hmC levels (which can result in negative values in hypomethylated sites), the Maximum Likelihood Methylation Levels (MLML) method, available in the R package MLML2R, was used to estimate 5-hmC [86, 87]. The computationally efficient MLML method accepts data from bisulfite sequencing or Infinium arrays and simultaneously estimates 5-mC, 5-hmC, and 5-C (unmethylated) proportions at each loci using an algorithm that does not allow negatives or summed scores over 1 [87]. Prior to performing MLML, data were preprocessed together as described below, with the exception of quantile normalization. Estimated 5-hmC and 5-mC from the MLML method were used in analyses. ## DNA methylation processing Quality control (QC) and preprocessing (i.e., normalization) of the array data following both treatments was completed in R (version > 4.1) using Bioconductor packages (minfi, Enmix) [88–92]. Briefly, raw image IDAT files for all samples were read into R. In individual IDAT files for each treated DNA sample, samples with poor coverage (< 3 beads), samples with > $5\%$ of sites failed, and samples of which predicted sex did not match reported sex and/or genetic background did not match that of matching maternal samples were removed from the dataset. In either the bisulfite treated group or the oxidative bisulfite group, probes with poor detection ($p \leq 1$e−16 when compared with background) were removed. After these steps, 31,434 CpG sites and 39 samples were removed from all datasets. Additional probes with SNPs in the CpG or single base extension site, probes known to be cross-reactive [52, 93], CH probes, and probes in the X/Y chromosomes were removed. Probes that had high intra-sample variability (> $5\%$ difference, based on 20 replicated samples) were also removed. The same probes were removed from both bisulfite treated and oxidative bisulfite treated samples, yielding 744,926 high-quality probes in all datasets. For replicated samples, one sample that passed all QC checks was randomly selected for inclusion in analysis. For those passing QC, correction and normalization was completed on each group (bisulfite data and oxidative bisulfite data describing both 5-hmC and 5-mC). Background correction with out-of-band (oob) and dye bias correction with RELIC were applied [89]. Quantile normalization was applied on each probe type and color channel separately [90]. Cell type proportions were estimated using an algorithm based on a reference dataset from seven sorted cord blood cell types [94, 95]. Surrogate variable analysis was performed on data from the control probes to create variables that best estimate the technical variability of samples [96]. For total methylation, 141 unique MMIP samples with PFAS data passed QC. Out of these, 70 also had a matching oxidative bisulfite converted sample that passed QC. ## Statistical analysis Analyses were performed in R, version > 4.1 [97]. PFAS plasma concentration distributions were evaluated for normality and any potential outliers. Individual concentrations below the analytical limit of detection (LOD, see Additional file 1: Table S1) were replaced with a value of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$LOD/\sqrt{2}$$\end{document}LOD/2. The LOD for all PFAS was 0.1 µg/L. Any PFAS with > $80\%$ of samples below the LOD were not included in any further statistical analyses (PFOSA and PFHpA). PFAS with 40–$80\%$ of samples below the LOD were dichotomized (above the LOD vs. below the LOD) and treated as a categorical variable in subsequent analyses (PFUnDA and MeFOSAA). PFAS that had < $40\%$ below the LOD remaining were treated as continuous variables (PFHxS, PFOA, PFOS, PFNA, and PFDA), consistent with recommended handling of PFAS data by NHANES [98]. PFAS concentrations for these variables were natural log-transformed prior to analysis. Correlations of all seven PFAS included in this analysis were examined to determine the relationship between the exposures of interest. Descriptive statistics were computed for all continuous and categorical variables, and differences between cohort groups (entire cohort, those with passing total methylation, and those with passing 5-mC/5-hmC data) were compared using one-way ANOVAs. A directed acyclic graph (DAG) was constructed to identify confounders of interest in the relationship between PFAS and DNA methylation (Fig. 5). Infant sex, gestational age, and maternal characteristics (early pregnancy BMI, age, parity, race and ethnicity, smoking status, marital status) were initially considered as potential confounders, and correlations of these confounders with PFAS were assessed using either Pearson’s correlations for numeric variables or Chi-squared tests for categorical variables. Technical variables including estimated cord blood cell type proportions and batch variables were the main predictors of variability within the DNA methylation and DNA hydroxymethylation data, as determined by surrogate variable analysis in the R package ChAMP [99, 100].Fig. 5A directed acyclic graph (DAG) used to identify real and theoretical confounders in the relationship between PFAS exposure and newborn DNA methylation and hydroxymethylation. Rectangles represent variables included as exposures or outcomes. Ovals represent confounding or control variables that were assessed for model inclusion. Thick solid lines show the relationship considered for mediation analysis. Filled ovals represent all precision, confounding, or control variables included in the models. Other maternal variables (unfilled oval on bottom of DAG) were considered for inclusion (age, early pregnancy body mass index (BMI), marital status, and income). Within our dataset, only one variable (self-reported race as a proxy for racism and racist policies) showed evidence as a true confounder (associated with both exposure and outcome); other maternal variables were not included in the models. Infant sex was not considered as true confounder, as only gestational age and Fenton z-scores (already adjusted for age and sex) were assessed as outcomes To minimize confounding bias and the impact of cell-type and batch effects while not overfitting methylation data, ten variables were selected for inclusion in the final model. Infant sex, parity, smoking status, and self-reported race as African American or Black (as a proxy of racism and racist policies that influence PFAS exposure burden) were significantly associated with at least one PFAS ($p \leq 0.05$) and considered true confounders in this study population in the main relationship of interest (association with both PFAS and DNA methylation or hydroxymethylation). The top three cell type proportions that were significantly associated with at least one PFAS were selected for model inclusion (CD4 + T cells, CD8 + T cells, and granulated cells). Nucleated red blood cells were also selected for model inclusion, as this type of cell is uniquely common in neonatal blood, and may significantly impact the methylation profiles in cord blood [101]. Finally, the top two component variables (PCs from the surrogate variable analysis described above) to account for batch effects were included in the models. For all models, methylation was regressed on individual PFAS exposures, as continuous PFAS concentrations (natural log transformed PFHxS, PFOA, PFOS, PFNA, PFDA) or categorical PFAS (above the LOD and below the LOD for PFUnDA and MeFOSAA). Any outliers for continuous measures of PFAS were kept in analyses and believed to represent real exposure data. For all analyses, beta regressions, which are designed to explicitly model continuous proportional data, were fit using normalized beta values which represent the proportion of methylation at each CpG site (between 0 and 1). The GAMLSS R package [102] was used to regress beta values at each CpG site on the individual PFAS, adjusting for parity (numeric; 1–4), reported smoking during pregnancy (any versus none); self-described race as African American or Black (versus any other); infant sex, estimated cell type proportions for granulocytes, CD4 + T, CD8 + T, and nucleated red blood cells; and PC1 and PC2 representing technical/batch effects. Data for parity, infant sex, cell types, and PCs were available in all participants. Nine individuals were missing smoking status and six individuals were missing race. Missingness in these categories was found to be unrelated to exposures, so an imputation method was applied. In brief, the distributions of the complete data were defined, and random samples were drawn from these distributions. This method utilizes the first step of multiple imputation, but it does not run the analysis multiple times (due to the extreme computational load required for this analysis). In this case, sampled missing variables were imputed with the majority category by chance. Model inflation was assessed using genomic inflation factors (lambdas), comparing all raw p-values from each model to an expected distribution. Results from each model were considered after applying a BH procedure. CpG sites were annotated with data available from the IlluminaHumanMethylationEPICanno.ilm10b2.hg19 R package. ## PFAS and DNA hydroxymethylation For 5-mC and 5-hmC ($$n = 70$$), an approach proposed by Kochmanski et al. [ 103] was used. Because hydroxymethylation is biologically and methodologically linked with methylation, interdependence precludes independently modeling these values. Instead, paired data can be evaluated to assess site level differences in methylation and hydroxymethylation. Estimated 5-hmC and 5-mC data from the MLML method were concatenated into a single matrix, resulting in two observations for each individual, with replicated phenotype data. An additional term (Type) was added to delineate if an observation was 5-mC or 5-hmC data. Because 5-hmC does not uniquely occur throughout the genome, any CpG site with a total (5-mC + 5-hmC) methylation of < 0.1 was excluded from analysis, yielding a total of 528,389 sites. For these 140 observations from 70 mother–infant pairs, associations with PFAS were tested using beta regressions in GAMLSS, with a random-effect for the ID, and allowing the methylation type to have independent Φ (identity link functions):\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} 5 - mC\;{\text{or}}\;5 - hmC\;{\text{proportion}} & = \beta _{0} + \beta _{1} {\text{PFAS}} + \beta _{2} {\text{Parity}} + \beta _{3} {\text{Smoking}} + \beta _{4} {\text{Race}} + \beta _{5} {\text{sex}} \\ & \quad + \beta _{6} {\text{CD}}4{\text{T}} + \beta _{7} {\text{CD}}8{\text{T}} + \beta _{8} {\text{GranCell}} + \beta _{9} n{\text{RBC}} \\ & \quad + \beta _{{10}} {\text{PC}}1 + \beta _{{11}} {\text{PC}}2 + \beta _{{12}} {\text{Type}} \\ & \quad + \beta _{{13}} {\text{Type}}*{\text{PFAS}} + [1|{\text{ID}}] \\ \end{aligned}$$\end{document}5-mCor5-hmCproportion=β0+β1PFAS+β2Parity+β3Smoking+β4Race+β5sex+β6CD4T+β7CD8T+β8GranCell+β9nRBC+β10PC1+β11PC2+β12Type+β13Type∗PFAS+[1|ID] CpG sites with a BH-corrected interaction term of with q-value < 0.2 were than stratified by methylation type; associations between PFAS with 5-mC and 5-hmC were then modeled separately at these loci. Genomic inflation values were calculated to assess potential p-value inflation. A significance cutoff of q-value < 0.2 was used at this stage to reduce the number of tests and limit the number of sites with separate 5-mC and 5-hmC modeling. Within only these sites, a beta regression model identical to the model for total methylation (above) was fit for either 5-mC or 5-hmC. A q-value < 0.05 was used to identify either 5-mC or 5-hmC sites that were significantly associated with the PFAS of interest. Sex-stratified analyses were not included in this analysis, as the sample size for 5-mC and 5-hmC specific data was limited. ## Posthoc methylation assessments To better understand the public health implications of any significant relationships between PFAS, total methylation, 5-mC and 5-hmC, several post hoc assessments were conducted. For the total methylation analysis without sex-stratification, a correlation analysis was used to compare the directionality of the coefficients to results from any previously reported study that examined PFAS exposure and genome-wide total methylation differences in early life [41–46]. Sex-stratified results were not compared. For models with more than 1000 significant sites (in 5-hmC analyses), regional differences were assessed using ipDMR [104], using 310 bp bins. For models with more than 100 significant sites (in 5-hmC analyses), KEGG pathways were assessed using the methylGSA package in R. ## Epigenetic mediation assessment Because birth outcomes have previously been associated with PFAS exposure, differences in DNA methylation were considered as potential mediators in the exposure to outcome pathway. To assure assumptions for mediation analyses were met in this study population [105], the direct relationship between PFAS exposure and birth outcomes, including gestational age and Fenton z-score adjusted size-for-gestational age at birth was computed in R using linear regressions that controlled for parity, race and smoking status in pregnancy. Mediation analyses were conducted using a nonlinear, kernel machine regression, which was specifically designed for epigenetic studies [106, 107]. To meet the assumptions for mediation, only those relationships with effects suggestive of significance were considered for mediation. Relationships were screened first at q < 0.05 for the association between PFAS and any type of methylation; then $p \leq 0.1$ for PFAS and birth outcomes; and finally, $p \leq 0.05$ for any type of methylation and birth outcomes. Gene-wise CpG sites meeting these criteria were included as mediators. Additionally, the complete groups of all sites related to any single PFAS and birth outcome were also assessed. BH q-values were applied to all comparisons by PFAS, and a q-value < 0.05 was considered as statistically significant. *Significant* genes’ functions were identified using the human causal genes in Ingenuity Pathway Analysis’s BioProfiler tool (Qiagen). ## Supplementary Information Additional file 1. Supplemental Tables and FiguresAdditional file 2. Supplemental Excel Materials. ## References 1. Braun JM. **Early-life exposure to EDCs: role in childhood obesity and neurodevelopment**. *Nat Rev Endocrinol* (2017.0) **13** 161-173. DOI: 10.1038/nrendo.2016.186 2. Meeker JD. **Exposure to environmental endocrine disruptors and child development**. *Arch Pediatr Adolesc Med* (2012.0) **166** E1-E7. PMID: 22664748 3. 3.Enviromental Protection Agency. PFASMASTER [Available from: https://comptox.epa.gov/dashboard/chemical-lists/pfasmaster. 4. Boronow KE, Brody JG, Schaider LA, Peaslee GF, Havas L, Cohn BA. **Serum concentrations of PFASs and exposure-related behaviors in African American and non-Hispanic White women**. *J Expo Sci Environ Epidemiol* (2019.0) **29** 206-217. DOI: 10.1038/s41370-018-0109-y 5. Schaider LA, Balan SA, Blum A, Andrews DQ, Strynar MJ, Dickinson ME. **Fluorinated compounds in U.S. fast food packaging**. *Environ Sci Technol Lett* (2017.0) **4** 105-11. DOI: 10.1021/acs.estlett.6b00435 6. Boone JS, Vigo C, Boone T, Byrne C, Ferrario J, Benson R. **Per- and polyfluoroalkyl substances in source and treated drinking waters of the United States**. *Sci Total Environ* (2019.0) **653** 359-369. DOI: 10.1016/j.scitotenv.2018.10.245 7. Hu XC, Andrews DQ, Lindstrom AB, Bruton TA, Schaider LA, Grandjean P. **Detection of poly- and perfluoroalkyl substances (PFASs) in U.S. drinking water linked to industrial sites, military fire training areas, and wastewater treatment plants**. *Environ Sci Technol Lett* (2016.0) **3** 344-50. DOI: 10.1021/acs.estlett.6b00260 8. Ashley-Martin J, Dodds L, Arbuckle TE, Bouchard MF, Fisher M, Morriset AS. **Maternal concentrations of perfluoroalkyl substances and fetal markers of metabolic function and birth weight**. *Am J Epidemiol* (2017.0) **185** 185-193. PMID: 28172036 9. Kingsley SL, Eliot MN, Kelsey KT, Calafat AM, Ehrlich S, Lanphear BP. **Variability and predictors of serum perfluoroalkyl substance concentrations during pregnancy and early childhood**. *Environ Res* (2018.0) **165** 247-257. DOI: 10.1016/j.envres.2018.04.033 10. Mora AM, Oken E, Rifas-Shiman SL, Webster TF, Gillman MW, Calafat AM. **Prenatal exposure to perfluoroalkyl substances and adiposity in early and mid-childhood**. *Environ Health Perspect* (2017.0) **125** 467-473. DOI: 10.1289/EHP246 11. Starling AP, Adgate JL, Hamman RF, Kechris K, Calafat AM, Ye X. **Perfluoroalkyl substances during pregnancy and offspring weight and adiposity at birth: examining mediation by maternal fasting glucose in the healthy start study**. *Environ Health Perspect* (2017.0) **125** 067016. DOI: 10.1289/EHP641 12. Matilla-Santander N, Valvi D, Lopez-Espinosa MJ, Manzano-Salgado CB, Ballester F, Ibarluzea J. **Exposure to perfluoroalkyl substances and metabolic outcomes in pregnant women: evidence from the spanish inma birth cohorts**. *Environ Health Perspect* (2017.0) **125** 117004. DOI: 10.1289/EHP1062 13. Tian Y, Zhou Y, Miao M, Wang Z, Yuan W, Liu X. **Determinants of plasma concentrations of perfluoroalkyl and polyfluoroalkyl substances in pregnant women from a birth cohort in Shanghai**. *China Environ Int* (2018.0) **119** 165-173. DOI: 10.1016/j.envint.2018.06.015 14. Liu CY, Chen PC, Lien PC, Liao YP. **Prenatal perfluorooctyl sulfonate exposure and Alu DNA hypomethylation in cord blood**. *Int J Environ Res Public Health* (2018.0) **15** 1066. DOI: 10.3390/ijerph15061066 15. Kobayashi S, Azumi K, Goudarzi H, Araki A, Miyashita C, Kobayashi S. **Effects of prenatal perfluoroalkyl acid exposure on cord blood IGF2/H19 methylation and ponderal index: The Hokkaido Study**. *J Expo Sci Environ Epidemiol* (2017.0) **27** 251-259. DOI: 10.1038/jes.2016.50 16. Ferguson KK, O'Neill MS, Meeker JD. **Environmental contaminant exposures and preterm birth: a comprehensive review**. *J Toxicol Environ Health B Crit Rev* (2013.0) **16** 69-113. DOI: 10.1080/10937404.2013.775048 17. Meng Q, Inoue K, Ritz B, Olsen J, Liew Z. **Prenatal exposure to perfluoroalkyl substances and birth outcomes; an updated analysis from the danish national birth cohort**. *Int J Environ Res Public Health.* (2018.0) **15** 1832. DOI: 10.3390/ijerph15091832 18. Sagiv SK, Rifas-Shiman SL, Fleisch AF, Webster TF, Calafat AM, Ye X. **Early-pregnancy plasma concentrations of perfluoroalkyl substances and birth outcomes in project viva: Confounded by pregnancy hemodynamics?**. *Am J Epidemiol* (2018.0) **187** 793-802. DOI: 10.1093/aje/kwx332 19. Bach CC, Bech BH, Brix N, Nohr EA, Bonde JP, Henriksen TB. **Perfluoroalkyl and polyfluoroalkyl substances and human fetal growth: a systematic review**. *Crit Rev Toxicol* (2015.0) **45** 53-67. DOI: 10.3109/10408444.2014.952400 20. Erinc A, Davis MB, Padmanabhan V, Langen E, Goodrich JM. **Considering environmental exposures to per- and polyfluoroalkyl substances (PFAS) as risk factors for hypertensive disorders of pregnancy**. *Environ Res* (2021.0) **197** 111113. DOI: 10.1016/j.envres.2021.111113 21. Liu Y, Li N, Papandonatos GD, Calafat AM, Eaton CB, Kelsey KT. **Exposure to per- and polyfluoroalkyl substances and adiposity at age 12 years: evaluating periods of susceptibility**. *Environ Sci Technol* (2020.0) **54** 16039-16049. DOI: 10.1021/acs.est.0c06088 22. Reik W, Dean W, Walter J. **Epigenetic reprogramming in mammalian development**. *Science* (2001.0) **293** 1089-1093. DOI: 10.1126/science.1063443 23. Jones PA. **Functions of DNA methylation: islands, start sites, gene bodies and beyond**. *Nat Rev Genet* (2012.0) **13** 484-492. DOI: 10.1038/nrg3230 24. Sadakierska-Chudy A, Kostrzewa RM, Filip M. **A comprehensive view of the epigenetic landscape part I: DNA methylation, passive and active DNA demethylation pathways and histone variants**. *Neurotox Res* (2015.0) **27** 84-97. DOI: 10.1007/s12640-014-9497-5 25. Bachman M, Uribe-Lewis S, Yang X, Williams M, Murrell A, Balasubramanian S. **5-Hydroxymethylcytosine is a predominantly stable DNA modification**. *Nat Chem* (2014.0) **6** 1049-1055. DOI: 10.1038/nchem.2064 26. Rashid F, Ramakrishnan A, Fields C, Irudayaraj J. **Acute PFOA exposure promotes epigenomic alterations in mouse kidney tissues**. *Toxicol Rep* (2020.0) **7** 125-132. DOI: 10.1016/j.toxrep.2019.12.010 27. Wen Y, Mirji N, Irudayaraj J. **Epigenetic toxicity of PFOA and GenX in HepG2 cells and their role in lipid metabolism**. *Toxicol In Vitro* (2020.0) **65** 104797. DOI: 10.1016/j.tiv.2020.104797 28. Nestor CE, Ottaviano R, Reddington J, Sproul D, Reinhardt D, Dunican D. **Tissue type is a major modifier of the 5-hydroxymethylcytosine content of human genes**. *Genome Res* (2012.0) **22** 467-477. DOI: 10.1101/gr.126417.111 29. Globisch D, Munzel M, Muller M, Michalakis S, Wagner M, Koch S. **Tissue distribution of 5-hydroxymethylcytosine and search for active demethylation intermediates**. *PLoS ONE* (2010.0) **5** e15367. DOI: 10.1371/journal.pone.0015367 30. Szwagierczak A, Bultmann S, Schmidt CS, Spada F, Leonhardt H. **Sensitive enzymatic quantification of 5-hydroxymethylcytosine in genomic DNA**. *Nucleic Acids Res* (2010.0) **38** e181. DOI: 10.1093/nar/gkq684 31. Wu H, D'Alessio AC, Ito S, Wang Z, Cui K, Zhao K. **Genome-wide analysis of 5-hydroxymethylcytosine distribution reveals its dual function in transcriptional regulation in mouse embryonic stem cells**. *Genes Dev* (2011.0) **25** 679-684. DOI: 10.1101/gad.2036011 32. Stroud H, Feng S, Morey Kinney S, Pradhan S, Jacobsen SE. **5-Hydroxymethylcytosine is associated with enhancers and gene bodies in human embryonic stem cells**. *Genome Biol* (2011.0) **12** R54. DOI: 10.1186/gb-2011-12-6-r54 33. Lopez V, Fernandez AF, Fraga MF. **The role of 5-hydroxymethylcytosine in development, aging and age-related diseases**. *Ageing Res Rev* (2017.0) **37** 28-38. DOI: 10.1016/j.arr.2017.05.002 34. Jobe EM, Zhao X. **DNA methylation and adult neurogenesis**. *Brain Plast* (2017.0) **3** 5-26. DOI: 10.3233/BPL-160034 35. Spiers H, Hannon E, Schalkwyk LC, Bray NJ, Mill J. **5-hydroxymethylcytosine is highly dynamic across human fetal brain development**. *BMC Genomics* (2017.0) **18** 738. DOI: 10.1186/s12864-017-4091-x 36. Vogel Ciernia A, LaSalle J. **The landscape of DNA methylation amid a perfect storm of autism aetiologies**. *Nat Rev Neurosci* (2016.0) **17** 411-423. DOI: 10.1038/nrn.2016.41 37. Stoyanova E, Riad M, Rao A, Heintz N. **5-Hydroxymethylcytosine-mediated active demethylation is required for mammalian neuronal differentiation and function**. *Elife.* (2021.0) **10** e66973. DOI: 10.7554/eLife.66973 38. Shen Y, Zhou S, Zhao X, Li H, Sun J. **Characterization of genome-wide dna methylation and hydroxymethylation in mouse arcuate nucleus of hypothalamus during puberty process**. *Front Genet* (2020.0) **11** 626536. DOI: 10.3389/fgene.2020.626536 39. Kim S, Thapar I, Brooks BW. **Epigenetic changes by per- and polyfluoroalkyl substances (PFAS)**. *Environ Pollut* (2021.0) **279** 116929. DOI: 10.1016/j.envpol.2021.116929 40. Perng W, Nakiwala D, Goodrich JM. **What happens in utero does not stay in utero: a review of evidence for prenatal epigenetic programming by per- and polyfluoroalkyl substances (PFAS) in infants, children, and adolescents**. *Curr Environ Health Rep.* (2022.0) **10** 35-44. DOI: 10.1007/s40572-022-00387-z 41. Liu Y, Eliot MN, Papandonatos GD, Kelsey KT, Fore R, Langevin S. **Gestational perfluoroalkyl substance exposure and DNA methylation at birth and 12 years of age: a longitudinal epigenome-wide association study**. *Environ Health Perspect* (2022.0) **130** 37005. DOI: 10.1289/EHP10118 42. Miura R, Araki A, Miyashita C, Kobayashi S, Kobayashi S, Wang SL. **An epigenome-wide study of cord blood DNA methylations in relation to prenatal perfluoroalkyl substance exposure: the Hokkaido study**. *Environ Int* (2018.0) **115** 21-28. DOI: 10.1016/j.envint.2018.03.004 43. Kingsley SL, Kelsey KT, Butler R, Chen A, Eliot MN, Romano ME. **Maternal serum PFOA concentration and DNA methylation in cord blood: a pilot study**. *Environ Res* (2017.0) **158** 174-178. DOI: 10.1016/j.envres.2017.06.013 44. Leung YK, Ouyang B, Niu L, Xie C, Ying J, Medvedovic M. **Identification of sex-specific DNA methylation changes driven by specific chemicals in cord blood in a Faroese birth cohort**. *Epigenetics* (2018.0) **13** 290-300. DOI: 10.1080/15592294.2018.1445901 45. Starling AP, Liu C, Shen G, Yang IV, Kechris K, Borengasser SJ. **Prenatal exposure to per- and polyfluoroalkyl substances, umbilical cord blood DNA methylation, and cardio-metabolic indicators in newborns: the healthy start study**. *Environ Health Perspect* (2020.0) **128** 127014. DOI: 10.1289/EHP6888 46. Robinson SL, Zeng X, Guan W, Sundaram R, Mendola P, Putnick DL. **Perfluorooctanoic acid (PFOA) or perfluorooctane sulfonate (PFOS) and DNA methylation in newborn dried blood spots in the Upstate KIDS cohort**. *Environ Res* (2021.0) **194** 110668. DOI: 10.1016/j.envres.2020.110668 47. Ahmad S, Wen Y, Irudayaraj JMK. **PFOA induces alteration in DNA methylation regulators and SARS-CoV-2 targets Ace2 and Tmprss2 in mouse lung tissues**. *Toxicol Rep* (2021.0) **8** 1892-1898. DOI: 10.1016/j.toxrep.2021.11.014 48. Fenton SE, Ducatman A, Boobis A, Dewitt JC, Lau C, Ng C. **Per- and polyfluoroalkyl substance toxicity and human health review: current state of knowledge and strategies for informing future research**. *Environ Toxicol Chem* (2021.0) **40** 606-630. DOI: 10.1002/etc.4890 49. Suzuki MM, Bird A. **DNA methylation landscapes: provocative insights from epigenomics**. *Nat Rev Genet* (2008.0) **9** 465-476. DOI: 10.1038/nrg2341 50. Moore LD, Le T, Fan G. **DNA methylation and its basic function**. *Neuropsychopharmacology* (2013.0) **38** 23-38. DOI: 10.1038/npp.2012.112 51. Delatte B, Jeschke J, Defrance M, Bachman M, Creppe C, Calonne E. **Genome-wide hydroxymethylcytosine pattern changes in response to oxidative stress**. *Sci Rep* (2015.0) **5** 12714. DOI: 10.1038/srep12714 52. Pidsley R, Zotenko E, Peters TJ, Lawrence MG, Risbridger GP, Molloy P. **Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling**. *Genome Biol* (2016.0) **17** 208. DOI: 10.1186/s13059-016-1066-1 53. Leijon I, Ingemansson F, Nelson N, Wadsby M, Samuelsson S. **Reading deficits in very low birthweight children are associated with vocabulary and attention issues at the age of seven**. *Acta Paediatr* (2016.0) **105** 60-68. DOI: 10.1111/apa.13094 54. Orchinik LJ, Taylor HG, Espy KA, Minich N, Klein N, Sheffield T. **Cognitive outcomes for extremely preterm/extremely low birth weight children in kindergarten**. *J Int Neuropsychol Soc* (2011.0) **17** 1067-1079. DOI: 10.1017/S135561771100107X 55. Hayes B, Sharif F. **Behavioural and emotional outcome of very low birth weight infants–literature review**. *J Matern Fetal Neonatal Med* (2009.0) **22** 849-856. DOI: 10.1080/14767050902994507 56. Szulwach KE, Li X, Li Y, Song C-X, Wu H, Dai Q. **5-hmC–mediated epigenetic dynamics during postnatal neurodevelopment and aging**. *Nat Neurosci* (2011.0) **14** 1607-1616. DOI: 10.1038/nn.2959 57. Oh J, Schmidt RJ, Tancredi D, Calafat AM, Roa DL, Hertz-Picciotto I. **Prenatal exposure to per- and polyfluoroalkyl substances and cognitive development in infancy and toddlerhood**. *Environ Res* (2021.0) **196** 110939. DOI: 10.1016/j.envres.2021.110939 58. Oh J, Shin HM, Nishimura T, Rahman MS, Takahashi N, Tsuchiya KJ. **Perfluorooctanoate and perfluorooctane sulfonate in umbilical cord blood and child cognitive development: Hamamatsu Birth Cohort for Mothers and Children (HBC Study)**. *Environ Int* (2022.0) **163** 107215. DOI: 10.1016/j.envint.2022.107215 59. Skogheim TS, Villanger GD, Weyde KVF, Engel SM, Suren P, Oie MG. **Prenatal exposure to perfluoroalkyl substances and associations with symptoms of attention-deficit/hyperactivity disorder and cognitive functions in preschool children**. *Int J Hyg Environ Health* (2020.0) **223** 80-92. DOI: 10.1016/j.ijheh.2019.10.003 60. Vuong AM, Yolton K, Xie C, Dietrich KN, Braun JM, Webster GM. **Prenatal and childhood exposure to poly- and perfluoroalkyl substances (PFAS) and cognitive development in children at age 8 years**. *Environ Res* (2019.0) **172** 242-248. DOI: 10.1016/j.envres.2019.02.025 61. Varsi K, Torsvik IK, Huber S, Averina M, Brox J, Bjorke-Monsen AL. **Impaired gross motor development in infants with higher PFAS concentrations**. *Environ Res* (2022.0) **204** 112392. DOI: 10.1016/j.envres.2021.112392 62. Jiao X, Liu N, Xu Y, Qiao H. **Perfluorononanoic acid impedes mouse oocyte maturation by inducing mitochondrial dysfunction and oxidative stress**. *Reprod Toxicol* (2021.0) **104** 58-67. DOI: 10.1016/j.reprotox.2021.07.002 63. Wielsoe M, Long M, Ghisari M, Bonefeld-Jorgensen EC. **Perfluoroalkylated substances (PFAS) affect oxidative stress biomarkers in vitro**. *Chemosphere* (2015.0) **129** 239-245. DOI: 10.1016/j.chemosphere.2014.10.014 64. Crebelli R, Caiola S, Conti L, Cordelli E, De Luca G, Dellatte E. **Can sustained exposure to PFAS trigger a genotoxic response? A comprehensive genotoxicity assessment in mice after subacute oral administration of PFOA and PFBA**. *Regul Toxicol Pharmacol* (2019.0) **106** 169-177. DOI: 10.1016/j.yrtph.2019.05.005 65. Pierozan P, Cattani D, Karlsson O. **Tumorigenic activity of alternative per- and polyfluoroalkyl substances (PFAS): mechanistic in vitro studies**. *Sci Total Environ* (2022.0) **808** 151945. DOI: 10.1016/j.scitotenv.2021.151945 66. Lin CY, Lee HL, Hwang YT, Su TC. **The association between total serum isomers of per- and polyfluoroalkyl substances, lipid profiles, and the DNA oxidative/nitrative stress biomarkers in middle-aged Taiwanese adults**. *Environ Res* (2020.0) **182** 109064. DOI: 10.1016/j.envres.2019.109064 67. Liu X, Chen D, Wang B, Xu F, Pang Y, Zhang L. **Does low maternal exposure to per- and polyfluoroalkyl substances elevate the risk of spontaneous preterm birth? A nested case-control study in China**. *Environ Sci Technol* (2020.0) **54** 8259-8268. DOI: 10.1021/acs.est.0c01930 68. Omoike OE, Pack RP, Mamudu HM, Liu Y, Strasser S, Zheng S. **Association between per and polyfluoroalkyl substances and markers of inflammation and oxidative stress**. *Environ Res* (2021.0) **196** 110361. DOI: 10.1016/j.envres.2020.110361 69. Li W, He Q-Z, Wu C-Q, Pan X-Y, Wang J, Tan Y. **PFOS disturbs BDNF-ERK-CREB signalling in association with increased MicroRNA-22 in SH-SY5Y Cells**. *Biomed Res Int* (2015.0) **2015** 1-10 70. Sonkar R, Kay MK, Choudhury M. **PFOS modulates interactive epigenetic regulation in first-trimester human trophoblast cell line HTR-8/SV**. *Chem Res Toxicol* (2019.0) **32** 2016-2027. DOI: 10.1021/acs.chemrestox.9b00198 71. Dong H, Curran I, Williams A, Bondy G, Yauk CL, Wade MG. **Hepatic miRNA profiles and thyroid hormone homeostasis in rats exposed to dietary potassium perfluorooctanesulfonate (PFOS)**. *Environ Toxicol Pharmacol* (2016.0) **41** 201-210. DOI: 10.1016/j.etap.2015.12.009 72. Wang F, Liu W, Jin Y, Wang F, Ma J. **Prenatal and neonatal exposure to perfluorooctane sulfonic acid results in aberrant changes in miRNA expression profile and levels in developing rat livers**. *Environ Toxicol* (2015.0) **30** 712-723. DOI: 10.1002/tox.21949 73. Mamsen LS, Bjorvang RD, Mucs D, Vinnars MT, Papadogiannakis N, Lindh CH. **Concentrations of perfluoroalkyl substances (PFASs) in human embryonic and fetal organs from first, second, and third trimester pregnancies**. *Environ Int* (2019.0) **124** 482-492. DOI: 10.1016/j.envint.2019.01.010 74. Gunasekara CJ, MacKay H, Scott CA, Li S, Laritsky E, Baker MS. **Systemic interindividual epigenetic variation in humans is associated with transposable elements and under strong genetic control**. *Genome Biol* (2023.0) **24** 2. DOI: 10.1186/s13059-022-02827-3 75. Ojo AF, Peng C, Annamalai P, Megharaj M, Ng JC. **Toxicity assessment of historical aqueous film-forming foams (AFFFs) using cell-based assays**. *Environ Pollut.* (2022.0) **2022** 119806. DOI: 10.1016/j.envpol.2022.119806 76. Ojo AF, Peng C, Ng JC. **Combined effects and toxicological interactions of perfluoroalkyl and polyfluoroalkyl substances mixtures in human liver cells (HepG2)**. *Environ Pollut* (2020.0) **263** 114182. DOI: 10.1016/j.envpol.2020.114182 77. Goodrich JM, Ingle ME, Domino SE, Treadwell MC, Dolinoy DC, Burant C. **First trimester maternal exposures to endocrine disrupting chemicals and metals and fetal size in the Michigan Mother-Infant Pairs study**. *J Dev Orig Health Dis* (2019.0) **10** 447-458. DOI: 10.1017/S204017441800106X 78. LaBarre JL, McCabe CF, Jones TR, Song PX, Domino SE, Treadwell MC. **Maternal lipodome across pregnancy is associated with the neonatal DNA methylome**. *Epigenomics* (2020.0) **12** 2077-2092. DOI: 10.2217/epi-2020-0234 79. Chou JH, Roumiantsev S, Singh R. **PediTools electronic growth chart calculators: applications in clinical care, research, and quality improvement**. *J Med Internet Res* (2020.0) **22** e16204. DOI: 10.2196/16204 80. Kuklenyik Z, Needham LL, Calafat AM. **Measurement of 18 perfluorinated organic acids and amides in human serum using on-line solid-phase extraction**. *Anal Chem* (2005.0) **77** 6085-6091. DOI: 10.1021/ac050671l 81. Kannan K, Stathis A, Mazzella MJ, Andra SS, Barr DB, Hecht SS. **Quality assurance and harmonization for targeted biomonitoring measurements of environmental organic chemicals across the Children's Health Exposure Analysis Resource laboratory network**. *Int J Hyg Environ Health* (2021.0) **234** 113741. DOI: 10.1016/j.ijheh.2021.113741 82. Moran S, Arribas C, Esteller M. **Validation of a DNA methylation microarray for 850,000 CpG sites of the human genome enriched in enhancer sequences**. *Epigenomics* (2016.0) **8** 389-399. DOI: 10.2217/epi.15.114 83. Hernandez Mora JR, Sanchez-Delgado M, Petazzi P, Moran S, Esteller M, Iglesias-Platas I. **Profiling of oxBS-450K 5-hydroxymethylcytosine in human placenta and brain reveals enrichment at imprinted loci**. *Epigenetics* (2018.0) **13** 182-191. DOI: 10.1080/15592294.2017.1344803 84. Houseman EA, Johnson KC, Christensen BC. **OxyBS: estimation of 5-methylcytosine and 5-hydroxymethylcytosine from tandem-treated oxidative bisulfite and bisulfite DNA**. *Bioinformatics* (2016.0) **32** 2505-2507. DOI: 10.1093/bioinformatics/btw158 85. Stewart SK, Morris TJ, Guilhamon P, Bulstrode H, Bachman M, Balasubramanian S. **oxBS-450K: a method for analysing hydroxymethylation using 450K BeadChips**. *Methods* (2015.0) **72** 9-15. DOI: 10.1016/j.ymeth.2014.08.009 86. 86.Kiihl SF, Martinez-Garrido MJ, Domingo-Relloso A, Bermudez J, Tellez-Plaza M. MLML2R: an R package for maximum likelihood estimation of DNA methylation and hydroxymethylation proportions. Stat Appl Genet Mol Biol. 2019;18(1). 87. Qu J, Zhou M, Song Q, Hong EE, Smith AD. **MLML: consistent simultaneous estimates of DNA methylation and hydroxymethylation**. *Bioinformatics* (2013.0) **29** 2645-2646. DOI: 10.1093/bioinformatics/btt459 88. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD. **Minfi: a flexible and comprehensive bioconductor package for the analysis of Infinium DNA methylation microarrays**. *Bioinformatics* (2014.0) **30** 1363-1369. DOI: 10.1093/bioinformatics/btu049 89. Xu Z, Langie SA, De Boever P, Taylor JA, Niu L. **RELIC: a novel dye-bias correction method for Illumina Methylation BeadChip**. *BMC Genomics* (2017.0) **18** 4. DOI: 10.1186/s12864-016-3426-3 90. Xu Z, Niu L, Li L, Taylor JA. **ENmix: a novel background correction method for Illumina HumanMethylation450 BeadChip**. *Nucleic Acids Res* (2016.0) **44** e20. DOI: 10.1093/nar/gkv907 91. Xu Z, Niu L, Taylor JA. **The ENmix DNA methylation analysis pipeline for Illumina BeadChip and comparisons with seven other preprocessing pipelines**. *Clin Epigenet* (2021.0) **13** 216. DOI: 10.1186/s13148-021-01207-1 92. Fortin JP, Triche TJ, Hansen KD. **Preprocessing, normalization and integration of the Illumina HumanMethylationEPIC array with minfi**. *Bioinformatics* (2017.0) **33** 558-560. DOI: 10.1093/bioinformatics/btw691 93. McCartney DL, Walker RM, Morris SW, McIntosh AM, Porteous DJ, Evans KL. **Identification of polymorphic and off-target probe binding sites on the Illumina Infinium MethylationEPIC BeadChip**. *Genom Data* (2016.0) **9** 22-24. DOI: 10.1016/j.gdata.2016.05.012 94. Bakulski KM, Feinberg JI, Andrews SV, Yang J, Brown S. **DNA methylation of cord blood cell types: applications for mixed cell birth studies**. *Epigenetics* (2016.0) **11** 354-62. DOI: 10.1080/15592294.2016.1161875 95. 95.Andrews S, Bakulski KM. FlowSorted.CordBlood.450k: Illumina 450k data on sorted cord blood cells. R package version 1240. 2022. 96. Teschendorff AE, Zhuang J, Widschwendter M. **Independent surrogate variable analysis to deconvolve confounding factors in large-scale microarray profiling studies**. *Bioinformatics* (2011.0) **27** 1496-1505. DOI: 10.1093/bioinformatics/btr171 97. 97.R Core Team. R: A language and environment for statistical computing Vienna, Austria: R Foundation for Statistical Computing; 2021 [Available from: https://www.R-project.org/. 98. 98.CDCFourth National Report on Human Exposure to Environmental Chemicals2009AtlantaCDC. *Fourth National Report on Human Exposure to Environmental Chemicals* (2009.0) 99. Teschendorff AE, Menon U, Gentry-Maharaj A, Ramus SJ, Gayther SA, Apostolidou S. **An epigenetic signature in peripheral blood predicts active ovarian cancer**. *PLoS ONE* (2009.0) **4** e8274. DOI: 10.1371/journal.pone.0008274 100. Tian Y, Morris TJ, Webster AP, Yang Z, Beck S, Feber A. **ChAMP: updated methylation analysis pipeline for Illumina BeadChips**. *Bioinformatics* (2017.0) **33** 3982-3984. DOI: 10.1093/bioinformatics/btx513 101. Hermansen MC. **Nucleated red blood cells in the fetus and newborn**. *Archiv Dis Childhood Fetal Neonat Ed* (2001.0) **84** F211-F215. DOI: 10.1136/fn.84.3.F211 102. Rigby RA, Stasinopoulos DM. **Generalized additive models for location, scale and shape (with discussion)**. *J Roy Stat Soc: Ser C (Appl Stat)* (2005.0) **54** 507-554 103. Kochmanski J, Savonen C, Bernstein AI. **A novel application of mixed effects models for reconciling base-pair resolution 5-methylcytosine and 5-hydroxymethylcytosine data in neuroepigenetics**. *Front Genet* (2019.0) **10** 801. DOI: 10.3389/fgene.2019.00801 104. Xu Z, Xie C, Taylor JA, Niu L. **ipDMR: identification of differentially methylated regions with interval**. *Bioinformatics* (2021.0) **37** 711-713. DOI: 10.1093/bioinformatics/btaa732 105. Valeri L, Vanderweele TJ. **Mediation analysis allowing for exposure–mediator interactions and causal interpretation: theoretical assumptions and implementation with SAS and SPSS macros**. *Psychol Methods* (2013.0) **18** 137-150. DOI: 10.1037/a0031034 106. Zhao N, Zhan X, Huang Y-T, Almli LM, Smith A, Epstein MP. **Kernel machine methods for integrative analysis of genome-wide methylation and genotyping studies**. *Genet Epidemiol* (2018.0) **42** 156-167. DOI: 10.1002/gepi.22100 107. Fang R, Yang H, Gao Y, Cao H, Goode EL, Cui Y. **Gene-based mediation analysis in epigenetic studies**. *Brief. Bioinform* (2021.0) **22** bbaa113. DOI: 10.1093/bib/bbaa113
--- title: SERPINE1 and its co-expressed genes are associated with the progression of clear cell renal cell carcinoma authors: - Lingyu Guo - Tian An - Ziyan Wan - Zhixin Huang - Tie Chong journal: BMC Urology year: 2023 pmcid: PMC10037920 doi: 10.1186/s12894-023-01217-6 license: CC BY 4.0 --- # SERPINE1 and its co-expressed genes are associated with the progression of clear cell renal cell carcinoma ## Abstract ### Background Clear cell renal cell carcinoma(ccRCC) is a frequently occurring malignant tumor of the urinary system. Despite extensive research, the regulatory mechanisms underlying the pathogenesis and progression of ccRCC remain largely unknown. ### Methods We downloaded 5 ccRCC expression profiles from the Gene Expression Omnibus (GEO) database and obtained the list of differentially expressed genes (DEGs). Using String and Cytoscape tools, we determined the hub genes of ccRCC, and then analyzed their relationship with ccRCC patient survival. Ultimately, we identified SERPINE1 as a prognostic factor in ccRCC. Meanwhile, we confirmed the role of SERPINE1 in 786-O cells by cell transfection and in vitro experiments. ### Results Our analysis yielded a total of 258 differentially expressed genes, comprising 105 down-regulated genes and 153 up-regulated genes. Survival analysis of SERPINE1 expression in The Cancer Genome Atlas (TCGA) confirmed its association with the increase of tumor grade, lymph node metastasis, and tumor stage, as well as with shorter survival. Furthermore, we found that SERPINE1 expression levels were associated with CD8 + T cells, CD4 + T cells, B cells, macrophages, neutrophils, and dendritic cells. Cell experiments showed that knockdown SERPINE1 expression could inhibit the proliferation, migration and invasion of ccRCC cells. Among the co-expressed genes with the highest correlation, ITGA5, SLC2A3, SLC2A14, SHC1, CEBPB, and ADA were overexpressed and associated with shorter overall survival (OS) in ccRCC. ### Conclusions In this study, we identified hub genes that are strongly related to ccRCC, and highlights the potential utility of overexpressed SERPINE1 and its co-expressed genes could be used as prognostic and diagnostic biomarkers in ccRCC. ### Supplementary Information The online version contains supplementary material available at 10.1186/s12894-023-01217-6. ## Introduction Renal cell carcinoma (RCC) is a malignancy that originates in the renal epithelium, accounting for 2–$3\%$ of adult malignancies [1]. Clear cell renal cell carcinoma(ccRCC) is the most common subtype, comprising over $80\%$ of RCC cases and is associated with the worst prognosis [2]. Although surgical resection is effective for many ccRCC patients, approximately one-third of patients still experience metastasis or postoperative recurrence, leading to ineffective treatment [3, 4]. Current clinical guidelines rely on tumor size, stage, and grade, which lack molecular biomarkers. However, even patients deemed low risk by these clinical guidelines can still experience disease progression and tumor recurrence [5]. The absence of reliable clinical biomarkers makes early diagnosis and treatment of ccRCC challenging, contributing to the high mortality rate of ccRCC [6–8]. While some studies have identified specific molecules contribute to the risk stratification of ccRCC, their clinical application may be too costly and complex [9, 10]. Nevertheless, immune-related genes have been found to be associated with ccRCC prognosis, offering a potential complement to the existing staging system [11, 12]. Thus, further revealing the biomolecular mechanism of ccRCC and identifying valuable diagnostic and prognostic biomarkers may improve the treatment of ccRCC patients. Serpin Family E Member 1 (SERPINE1), also known as plasminogen activator inhibitor type 1 (PAI-1), is a member of the serine proteinase inhibitor (serpin) superfamily [13, 14]. It is a fibrinolytic inhibitor that involved in several human malignancies[15, 16] and is also an important component of innate antiviral immunity. Previous investigations have reported that SERPINE1 is involved in the regulatory process of various types of tumors, including gastric adenocarcinoma [14], glioma [17], bladder cancer [18], and lung cancer [19]. In gastric cancer, SERPINE1 has been found to regulate the EMT process and promote tumor progression [14], while in colon cancer, it can regulate tumor microenvironment and immune cell infiltration [20]. Furthermore, high expression of SERPINE1 in lung cancer is associated with poor prognosis [19]. However, it is still unclear whether SERPINE1 regulates the development of ccRCC. Microarray and RNA sequencing technology provide high-throughput genome sequencing data, and bioinformatics analysis enables efficient analysis of gene expression [21]. The GEO and TCGA databases provide valuable information on disease gene expression, helping to explore key signaling pathways and important molecular mechanisms of disease [22]. In this study, we selected 5 microarray data sets from the GEO database to identify DEGs and ccRCC hub genes. Using bioinformatics methods, we aimed to explore the relationship between SERPINE1 and clinical characteristics of ccRCC patients. Additionally, we performed a series of in vitro experiments to verify the biological functions of SERPINE1. ## Screening the differentially expressed genes We selected a total of 5 GEO databases for further research, including GSE15641, GSE16441, GSE40435, GSE53000, and GSE71963 [23–27]. To explore the molecular variation in ccRCC occurrence and development, all selected datasets included ccRCC tumor tissue samples and corresponding non-tumor tissue samples. GSE15641 from the GPL96 Affymetrix Human Genome U133A Array contains 23 normal and 32 ccRCC samples. GSE16441 from the GPL6480 Agilent-014850 Whole Human Genome Microarray contains 17 ccRCC tumors and 17 corresponding non-tumor samples. GSE40435 from GPL10558 Illumina HumanHT-12 V4.0 expression beadchip includes 101 pairs of ccRCC tumors and adjacent non-tumor renal tissue. GSE53000 from GPL6244 Affymetrix Human Gene 1.0 ST Array includes 56 ccRCC tumor samples and 6 normal samples. GSE71963 from GPL6480 Agilent-014850 Whole Human Genome Microarray includes 16 normal kidney tissues and 32 ccRCC tissues. We used the online analysis tool GEO2R (www.ncbi.nlm.nih.gov/geo/geo2r) to identify differentially expressed genes in ccRCC. Thresholds for analysis were set as p-value < 0.05 and |log fold change (FC)|≥ 1. Then we obtained a summary of differential genes in different datasets by drawing a Venn diagram. Finally, we identified 258 DEGs in ccRCC, of which 153 genes were up-regulated and 105 were down-regulated. ## Hub genes analysis To evaluate the biological functions and signaling pathways of the obtained DEGs, we used the Database for Annotation, Visualization and Integrated Discovery (DAVID) (https://david.ncifcrf.gov/). The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed [28–30]. To examine the correlations between hub gene proteins, the STRING database (https://cn.string-db.org/) was used to build a protein–protein interaction (PPI) network [31]. Cytoscape software (version 3.9.0) was used to visualize and analyze the PPI network [32]. The Cytoscape MCODE plug-in was used to identify key modules from the entire PPI network [33]. Subsequently, the cytoHubba plug-in was used to screen for possible hub genes from the identified genes. To investigate the potential role of hub genes in the disease progression of ccRCC, gene expression and prognostic correlation analysis were performed. We used the Kaplan–Meier Plotter online analysis tool (https://kmplot.com/analysis/) to determine the prognostic association of hub genes with ccRCC [34]. ## Immune infiltration analysis The TIMER database (https://cistrome.shinyapps.io/timer/) was used to explore the association between SERPINE1 expression and immune infiltrates in ccRCC [35]. We also employed the cBioPortal (http://www.cbioportal.org/) and GEPIA (http://gepia.cancer-pku.cn/) web tool to calculate the relationships between SERPINE1 and tumor immune infiltration gene markers [36]. ## Cell line generation and cell culture. The 786-O cell line was obtained from the American Type Culture Collection (ATCC) and cultured in RPMI-1640 medium supplemented with $1\%$ penicillin/streptomycin and $10\%$ fetal bovine serum. Cells were incubated at 37 °C in a $5\%$ CO2 incubator. ## Cell transfection To down-regulate the expression of SERPINE1, we obtained the si-SERPINE1 sequence from GenePharma Co. Ltd, which was as follows: 5′-UGAACUUGUUGGUCUGAGCTT-3′. The negative control siRNA sequence (si-NC) was 5′-ACGUGACACGUUCGGAGAATT-3′. The transfection experiment was performed using Lipofectamine 2000 (Thermo Fisher Scientific, Inc.) following the product instructions meticulously. ## Proliferation assay To investigate the impact of SERPINE1 on tumor cell proliferation, we conducted a CCK8 assay. 48 h post-transfection with siRNA, 786-O cells were harvested, suspended and then seeded on 96-well plates (Corning Inc, Corning, NY, USA) at a density of 1000 cells/well. At the time of the assay, CCK-8 reagent was added to each well according to the instruction. Cell viability was measured every 24 h by detecting the optical density (OD) values at the wavelength of 450 nm. ## Cell migration and invasion assay For the cell migration and invasion assays, we used transwell chambers from Corning Inc. The transfected cells were suspended in serum-free medium and seeded into the upper chamber at a density of 2 × 104 cells per well, while 600 ul of complete medium was added to the lower compartment. After incubating the cells for 24 h in a cell incubator, we used $4\%$ paraformaldehyde to fix the invaded cells for 30 min, followed by staining with $0.1\%$ crystal violet for 20 min. Residual cells that did not migrate were carefully removed. We calculated and photographed the cells under a microscope. For the cell invasion experiment, the bottom of the transwell chamber was covered with matrigel glue before cell seeding, and the remaining steps were the same as those for the cell migration experiment. ## Wound-healing assay After transfection for 48 h, the 786-O cells were harvested and seeded into a 6-well plate at a density of 5 × 104 cells/ml. When the cells reached more than $80\%$ confluence, a straight wound was created in the center of each well using a 200 μl pipette tip. The cells were then washed twice with PBS to remove any floating cells. After washing, the cells were recorded as 0 h and incubated in a medium containing $2\%$ FBS for further culture. After 24 h, the cells were observed and photographed under a microscope to evaluate the degree of wound healing. ## SERPINE1 related genes analysis The cBioPortal website was adopted to select SERPINE1′s co-expressed genes based on 538 samples from TCGA-KIRC [37]. UALCAN (http://ualcan.path.uab.edu/) and Kaplan–Meier plotter databases were used to further analyze the expression and prognostic value of SERPINE1 related genes in ccRCC [38]. ## Statistical analysis The statistical analyses for this study were conducted using SPSS 22.0 (SPSS, Inc.), R (version 4.0.2), and GraphPad Prism (version 8.0) software. For survival analysis, patients were divided based on the median value of gene expression. ROC curves were generated using the pROC package to evaluate the diagnostic value of these genes. The unpaired Student's t-test was employed to compare the results of the two groups. Chi-square test was used to compare clinical information between groups. Spearman's correlation analysis was used to detect correlations between genes. All experiments were performed in triplicate and presented as mean ± standard deviation (SD). A $p \leq 0.05$ was considered statistically significant. ## Identification of DEGs in ccRCC We analyzed 5 ccRCC GEO profile datasets and identified a total of 258 DEGs, with 153 genes highly expressed in tumor tissues and 105 genes with low expression (Fig. 1A). The volcano maps illustrate the differential expression of genes in each dataset (Fig. 1B–F). To further understand the biological processes associated with these DEGs, the DAVID database was used to annotate these robust DEGs, and performed GO and KEGG analyses. In the BP portion of GO analysis, the DEGs were mainly enriched in the ‘Oxidation–reduction process’ and ‘Response to drug’ (Fig. 2A). In the CC portion of the analysis, the DEGs were primarily concentrated in the ‘Plasma membrane’ and ‘Plasma membrane’(Fig. 2B). The main molecular function performed by DEGs was’Protein binding’(Fig. 2C). Through KEGG analysis, we learned that the DEGs were mainly involved in ‘PI3K-Akt signaling pathway’, ‘Focal adhesion’, and ‘Biosynthesis of antibiotics’ (Fig. 2D). These results suggest that these DEGs play a crucial role in the disease progression of ccRCC and are closely related to the onset and recurrence of the tumor. Fig. 1Differentially expressed genes in ccRCC. A Venn diagram of differentially expressed genes in different datasets. B GSE15641 data, C GSE16441 data, D GSE40435 data, E GSE53000 data, and F GSE71963 data. The red and blue dots represent up-regulated and down-regulated genes. The threshold for |logFC|≥ 1 and $p \leq 0.05.$ Black points represent genes with no significant difference. FC: fold change; GEO: Gene Expression Omnibus; DEGs; differentially expressed genesFig. 2DEGs functional enrichment analysis. A Biological processes, B Cellular components, and C Molecular functions. D Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis ## Hub genes selection and analysis The Cytoscape software was used to construct a PPI network of DEGs (Fig. 3A). Subsequently, CytoHubba plug-in of Cytoscape software was used to identify hub genes of ccRCC according to the MCC and EPC algorithms (Fig. 3B, C). The overlapped hub genes in the results of the two methods included VEGFA, EGF, TIMP1, MMP9, VWF, and SERPINE1(Fig. 3D). Then, we utilized the MCODE plug-in to identify the most significant module from the entire PPI network. The resulting significant module consisted of 25 nodes and 137 edges (Table 1). The expression levels and prognostic value of these 6 hub genes was studied. The expression levels of MMP9, VWF, VEGFA, SERPINE1, and TIMP1 in ccRCC tissues were higher than those in normal tissues, while the expression of EGF in tumor tissues was lower (Fig. 4A). Furthermore, higher expression of MMP9, SERPINE1, and TIMP1 suggested shorter OS in ccRCC patients, while higher expression of VWF suggested longer OS in ccRCC patients ($p \leq 0.05$) (Fig. 4B).Fig. 3PPI network construction and hub genes analysis. A PPI network. B and C The top 10 hub genes were screened using the Cytoscape software plugin cytoHubba. D Venn diagram of hub genes. The red and blue represent up-regulated and down-regulated genesTable 1Significant modules from the whole PPI networkModuleScoreNodesEdgesNode IDs111.41725137PGF, LAPTM5, TYROBP, RAC2, CD53, VWF, PLG, FGFR1, CYBB, ARHGDIB, JAK2, EGF, HRG, SERPINE1, HMOX1, EDN1, CORO1A, PLEK, CXCR4, TIMP1, ANGPT2, FGF1, ITGB2, CD48, IL10RA26.4621442ANGPTL4, PPARGC1A, ENO2, HK2, FBP1, PFKP, GK, PCK2, PYGL, ALDOC, HMGCS2, SCD, PCK1, HADH361130CP, LOX, STC2, LGALS1, CCND1, CDK3, MMP9, DCN, ESR1, CAV1, CA945.7781026SLC12A3, CLCNKA, NPHS2, UMOD, SCNN1G, KCNJ1, SCNN1A, SLC34A1, CLCNKB, WNK1551330VCAN, TGFBI, C3, CCL20, P4HA1, COL1A1, COL4A1, IGFBP3, PTGER3, ITGA5, COL5A2, ACKR3, PLOD264.8612GBP2, TAP1, PSMB9, HLA-F, FCGR1B, HLA-B73.714813PCLAF, TYMS, TOP2A, SLC1A3, NUSAP1, PVALB, CALB1, FABP78333CAVIN3, EHD2, CAV29333SLCO2B1, ABCC3, SLC22A610333GABARAPL1, ANK3, ANK2Fig. 4Expression and prognosis analysis of hub genes in ccRCC. A Expression of hub genes. B Relationship between hub genes expression levels and ccRCC OS. *** $p \leq 0.001.$ OS, overall survival Metabolic reprogramming is a complex process that involves various metabolic pathways such as aerobic glycolysis, fatty acid metabolism, and the utilization of tryptophan, glutamine, and arginine [39]. Enrichment analysis of the DEGs also suggested that tumor metabolic reprogramming plays an important role in ccRCC. Among the above three genes, we focused on SERPINE1 due to its reported association with tumor metabolism [40]. However, the specific mechanism of how SERPINE1 regulating the progression of ccRCC is still not well studied. Correlation between SERPINE1 expression and ccRCC progression. To further explore the role of SERPINE1 in ccRCC progression, the relationship between SERPINE1 expression and other clinicopathological parameters of ccRCC were studied. The expression of SERPINE1 increased gradually with the increase of tumor grade, lymph node metastasis, and tumor stage (Fig. 5A–C). Further analysis of TCGA-KIRC data revealed that the expression level of SERPINE1 was significantly associated with patients' gender, T stage, N stage, pathologic stage, and histological grade (Table 2).Fig. 5Correlation between SERPINE1 expression level and clinicopathological parameters. Correlation between SERPINE1 expression level and A tumor grade, B metastasis status, C cancer stage in ccRCC samples (ULCAN). *** $p \leq 0.001.$Table 2Clinical information of patients with clear cell renal cell carcinoma in the TCGACharacteristicsLow expression of SERPINE1High expression of SERPINE1p-valuen270271Age, n (%)0.111605111 < = 60125 ($23.1\%$)144 ($26.6\%$) > 60145 ($26.8\%$)127 ($23.5\%$)Gender, n (%)8.16029E-06 Female118 ($21.8\%$)69 ($12.8\%$) Male152 ($28.1\%$)202 ($37.3\%$)Pathologic T stage, n (%)0.009964032 T1&T2189 ($34.9\%$)161 ($29.8\%$) T3&T481 ($15\%$)110 ($20.3\%$)Pathologic N stage, n (%)0.018326564 N0134 ($51.9\%$)108 ($41.9\%$) N14 ($1.6\%$)12 ($4.7\%$)Pathologic M stage, n (%)0.24744979 M0215 ($42.3\%$)214 ($42.1\%$) M134 ($6.7\%$)45 ($8.9\%$)Pathologic stage, n (%)0.015713843 Stage I&Stage II179 ($33.3\%$)153 ($28.4\%$) Stage III&Stage IV89 ($16.5\%$)117 ($21.7\%$)Histologic grade, n (%)0.008439461 G1&G2139 ($26.1\%$)111 ($20.8\%$) G3&G4125 ($23.5\%$)158 ($29.6\%$) ## Relationship between SERPINE1 and immune characteristics In recent years, immune checkpoint inhibitors have been considered effective against a wide range of tumors, and research on tumor immune checkpoints has provided many important targets for tumor immunotherapy [41]. To explore the significance of SERPINE1 in tumor immunotherapy, we investigated the relationship between SERPINE1 expression and more than 40 key immune-related genes. Interestingly, in ccRCC, SERPINE1 expression levels were significantly correlated with 20 of our selected immune checkpoint marker genes, including CD86, TNFRSF14, TNFRSF18, and CD80 (Fig. 6A). Therefore, these results show that the SERPINE1 gene may play an important role in tumor immunity. According to the TIMER database, we found that SERPINE1 in ccRCC affected tumor-infiltrating immune cells. We first analyzed the correlation between SERPINE1 expression and tumor-infiltrating immune cells, which suggested that the SERPINE1 was significantly correlated with the immune infiltration of ccRCC (Fig. 6B–F). As shown in Fig. 6B, the expression of SERPINE1 was negatively correlated with B cells (R = − 0.214, $$p \leq 3.65$$e − 06). In addition, the expression of SERPINE1 was significantly positively correlated with various immune cell infiltrates, including CD4 + T cells ($R = 0.094$, $$p \leq 4.37$$e − 02), CD8 + T cells ($R = 0.127$, $$p \leq 6.15$$e − 03), myeloid dendritic cells ($R = 0.151$, $$p \leq 1.18$$e − 03), and macrophages ($R = 0.207$, $$p \leq 7.51$$e − 6). The table shows the correlation between SERPINE1 and immune cell-related markers on the cBioPortal website (Table 3). Therefore, SERPINE1 may play an important role in regulating the immune cell infiltration in ccRCC.Fig. 6Correlation analysis of SERPINE1 expression with immune checkpoint genes and tumor immune-associated cells. A *Correlation analysis* between SERPINE1 expression levels and immune checkpoint gene levels in ccRCC. B–F SERPINE1 expression was positively closely related with B cells, CD4 + T cells, CD8 + T cells, dendritic cells, and macrophages in ccRCC (TIMER)Table 3Correlation analysis between SERPINE1 and biomarkers of immune cells in ccRCC determined by GEPIA databaseImmune cellBiomarkerSpearman's Correlationp-valueB cellCD190.2595929481.13E-09CD79A0.2249381421.49E-07CD8 + T cellCD8A0.1103955090.010683103CD8B0.071503040.098827444CD4 + T cellCD40.2474844456.79E-09M1 macrophageNOS20.0387468750.371526712IRF5 − 0.1008829220.019714615PTGS20.3355695571.61E-15M2 macrophageCD1630.4461778031.74E-27VSIG40.2949432783.53E-12MS4A4A0.3472652271.40E-16NeutrophilCEACAM8 − 0.0228886480.597672419ITGAM0.1164641190.007056392CCR70.3096360632.50E-13Dendritic cellHLA-DPB10.0913579620.034805998HLA-DQB10.0535225960.216900999HLA-DRA0.1080002810.012518625HLA-DPA10.081940730.05845524CD1C0.010076170.816301262NRP10.2568309881.72E-09ITGAX0.1450592540.000773698MonocyteCSF1R0.1803695962.7562E-05CD860.2134639176.40E-07TregFOXP30.3027281488.83E-13CCR80.2228840051.95E-07TGFB10.4788319155.89E-32 ## Analysis of SERPINE1 related genes in ccRCC We conducted an analysis of TCGA-KIRC data stored in the cBioPortal database to identify genes associated with SERPINE1 expression in ccRCC. From this analysis, we selected the 10 genes that were most highly associated with SERPINE1 expression, namely ITGA5, SLC2A3, ELL2, ABL2, MT2A, SLC2A14, XIRP1, SHC1, CEBPB, and ADA (Table 4). Of these genes, ITGA5 exhibited the highest correlation with SERPINE1 expression, with a Pearson’s Correlation $R = 0.69$ and Spearman’s Correlation $R = 0.66.$ Furthermore, we analyzed the biological functions of these co-expressed genes and found that they may serve as biomarkers involved in tumor occurrence and progression. To investigate their expression and prognostic value in ccRCC and normal renal tissues, we explored the expression and survival data in the UALCAN and Kaplan–Meier plotter databases. The results showed that 9 of the 10 co-expressed genes, including ITGA5, SLC2A3, ELL2, ABL2, SLC2A14, XIRP1, SHC1, CEBPB, and ADA, were significantly upregulated in ccRCC tissues (Fig. 7A–J). In addition, the expression levels of 6 out of the 10 co-expressed genes were significantly correlated with disease OS, including ITGA5, SLC2A3, MT2A, SHC1, CEBPB, and ADA (Fig. 8A–J).Table 4Co-expression genes of SERPINE1 in ccRCC by cBioPortalCorrelated GeneCytobandPearson's CorrelationSpearman's Correlationp-Valueq-ValueITGA512q13.130.690.6567930043.19E-676.39E-63SLC2A312p13.310.620.6057764768.61E-558.62E-51ELL25q150.620.6040812022.04E-541.36E-50ABL21q25.20.580.6011208579.10E-544.56E-50MT2A16q130.620.5923223317.07E-522.83E-48SLC2A1412p13.310.530.5916079551.00E-513.34E-48XIRP13p22.20.520.5805868791.91E-495.47E-46SHC11q21.30.60.5775116988.00E-492.00E-45CEBPB20q13.130.580.5695243213.06E-476.82E-44ADA20q13.120.590.5675097267.56E-471.51E-43Fig. 7Expression of SERPINE1-related genes in ccRCC. A–J The expression levels of ITGA5, SLC3A3, ELL2, ABL2, MT2A, SLC2A14, XIRP1, SHC1, CEBPB, and ADA in ccRCC (ULCAN). *** $p \leq 0.001$Fig. 8The prognostic value of SERPINE1-related genes. A–J The expression levels of ITGA5, SLC3A3, ELL2, ABL2, MT2A, SLC2A14, XIRP1, SHC1, CEBPB, and ADA were correlated with OS in ccRCC (Kaplan–Meier Plotter). OS, overall survival To evaluate the diagnostic value of SERPINE1 and its co-expressed genes in ccRCC, we conducted receiver operating characteristic (ROC) curve analysis on the gene expression data downloaded from TCGA. The results demonstrated that SERPINE1 expression had high diagnostic value in ccRCC with an area under the curve (AUC) of 0.789 (Fig. 9). Moreover, the expression levels of SERPINE1 co-expressed genes also showed high diagnostic significance with ITGA5(AUC = 0.925), SLC2A3(AUC = 0.887), ELL2(AUC = 0.705), ABL2(AUC = 0.834), MT2A(AUC = 0.556), SLC2A14(AUC = 0.827), XIRP1(AUC = 0.701), SHC1(AUC = 0.925), CEBPB(AUC = 0.817), and ADA(AUC = 0.941). In summary, nearly all of the related genes (9 out of 10) exhibited promising diagnostic potential. Fig. 9The diagnostic value of SERPINE1-related genes. ROC, receiver operating characteristic curve. AUC, area under the curve ## Knockdown of SERPINE1 influenced ccRCC cell proliferation, migration, and invasion After transfecting 786-O cells with si-RNA transfection for 72 h, cell proteins were extracted for a Western Blot experiment to detect the expression level of SERPINE1 protein. The results indicated that SERPINE1 expression level in the si-RNA transfected 786-O cells was significantly lower than that in the si-NC group (Fig. 10A). In this study, CCK8 assay was used to evaluate the effect of knockdown SERPINE1 expression on the proliferation of 786-O cells. The results showed that the proliferation ability of 786-O cells decreased with the decrease of SERPINE1expression ($p \leq 0.001$, Fig. 10B). These results suggest that low SERPINE1 expression can inhibit the proliferation of ccRCC cells. To explore the effect of SERPINE1 expression on the migration and invasion ability of ccRCC cells, a transwell experiment was designed and implemented. The results suggested that the migration and invasion of 786-O cells were significantly reduced after SERPINE1 expression was down-regulated (Fig. 10C). In addition, cell scratch assay results showed that downregulation of SERPINE1 reduced the wound healing ability of 786-O cells (Fig. 10D). Overall, our in vitro experiments indicated that the high expression of SERPINE1 in ccRCC was associated with the development and metastasis of ccRCC.Fig. 10Analysis of SERPINE1 in ccRCC. A Western blot analysis of SERPINE1 expression in 786-O cells after transfection of SERPINE1 siRNA. B OD value of si-SERPINE1 cells. C Wound healing assay of si-SERPINE1 cells. D Migratory and invasive potential of si-SERPINE1 cells. * $p \leq 0.05$, **$p \leq 0.01$, ***$p \leq 0.001$ ## Discussion In this study, we identified hub genes in ccRCC that have potential as biomarkers through data mining of the GEO database. We then analyzed the expression level and prognostic value of these hub genes in ccRCC tissues based on the TCGA database. Our results showed that three genes (MMP9, SERPINE1, and TIMP1) were overexpressed in ccRCC, and high expression of these genes suggested poor prognosis. TIMP1 has been shown to be involved in the EMT process of ccRCC and can enhance tumor cells metastasis [42]. MMP9 can regulate the immune infiltration in ccRCC and affect tumor progression [43]. Studies have shown that SERPINE1 is involved in glucose and lipid metabolism [40, 44], which may be associated with the metabolic reprogramming of ccRCC. After combining clinical data downloaded from TCGA, we found that SERPINE1 expression was closely correlated with clinical indicators such as cancer stage and survival. Importantly, our further analysis revealed that SERPINE1 expression was associated with several immune biomarkers, indicating that SERPINE1 may have therapeutic value as a target of immunotherapy for ccRCC. SERPINE1 is known to inhibit tissue plasminogen activator (tPA) and urokinase-type plasminogen activator (PLAU) [16]. As a PLAU inhibitor, it can participate in the regulation of cell adhesion and spreading regulatory process [45]. Although the exact role of of SERPINE1 in disease and its association with the disease are not fully understood, researchers have shown that growth factors, cytokines, and hormones can regulate its expression [46]. Higgins [47] studied the expression of SERPINE1 in squamous cell carcinoma and found that TGF-β1 induces the high expression of SERPINE1 at the early stage of the tumor, with such high expression distributed on the invasive front of the tumor. Other findings have suggested interactions between SERPINE1 and the EGFR/MEK/Rho-ROCK signaling pathway [48]. Additionally, studies have shown that high SERPINE1 expression can help cells resist programmed cell death and regulate cell survival by activating Akt and ERK signaling pathways or inhibiting Fas/FasL-dependent apoptosis [49, 50]. Despite these findings, the specific mechanism by which SERPINE1 regulates tumor remains unknown. In this study, we used si-RNA transfection to knockdown the expression of SERPINE1 in the ccRCC cell line 786-O, and we found that inhibition of SERPINE1 expression reduced the proliferation, migration, and invasion of ccRCC cells. In addition, we explored genes co-expressed with SERPINE1 in ccRCC to provide a basis for explaining the regulation ability of SERPINE1 in ccRCC. Through co-expression analysis using the cBioPortal database, we identified a set of co-expressed genes with SERPINE1 in ccRCC. These co-expressed genes were found to be involved in transmembrane transporter activity, epidermal growth factor receptor binding, and other pathways. Interestingly, 9 out of the 10 co-expressed genes, including ITGA5, SLC2A3, ELL2, ABL2, SLC2A14, XIRP1, SHC1, CEBPB, and ADA, were significantly upregulated in ccRCC tissues. Additionally, the expression levels of 6 out of the 10 co-expressed genes were significantly correlated with disease OS, including ITGA5, SLC2A3, MT2A, SHC1, CEBPB, and ADA. Previous studies have demonstrated that these co-expressed genes play important roles in the regulation of a variety of tumors. For example, ITGA5 has been shown to ptomote the occurrence and development of colorectal cancer [51], SLC2A3 can promote macrophage infiltration by glycolysis reprogramming in gastric cancer [52]. Zhao et al. [ 53] study showed that MT2A promotes oxaliplatin resistance in colorectal cancer cells, SHC1 is a key driver of breast cancer initiation [54], CEBPB can regulate stemness and chemo-resistance of gastric cancer [55], and ADA tends to have a positive association with breast cancer [56]. Based on these findings, we believe that SERPINE1 and its co-expressed genes are widely involved in the occurrence and development of ccRCC, and may have high diagnostic and therapeutic value. However, further in vivo and in vitro studies are necessary to support these hypotheses and develop them as potential diagnostic and therapeutic targets for ccRCC. There are still some limitations in this study. Firstly, clinical samples of ccRCC were not available, and thus, tissue-level expression of SERPINE1 could not be evaluated. Secondly, although some findings were validated by in vitro experiments, further experiments are necessary to strengthen the results. Thirdly, the roles of SERPINE1 co-expressed genes in ccRCC have not been experimentally confirmed. This needs to be explored and demonstrated in future studies (Additional file 1). ## Conclusion In conclusion, this study has demonstrated the high expression of SERPINE1 in ccRCC and its correlation with various clinicopathological features, indicating the potential diagnostic and therapeutic value of SERPINE1 in ccRCC. ## Supplementary Information Additional file 1. Supplement figure 1. Raw immunoblot data for images in Fig. 10A. (A). Immunoblot anti-SERPINE1 of ccRCC cells. ( B). Immunoblot anti-GAPDH of ccRCC cells. ## References 1. Siegel RL, Miller KD, Fuchs HE, Jemal A. **Cancer statistics, 2021**. *CA Cancer J Clin* (2021.0) **71** 7-33. DOI: 10.3322/caac.21654 2. Shao Y, Wu B, Jia W, Zhang Z, Chen Q, Wang D. **Prognostic value of pretreatment neutrophil-to-lymphocyte ratio in renal cell carcinoma: a systematic review and meta-analysis**. *BMC Urol* (2020.0) **20** 90. DOI: 10.1186/s12894-020-00665-8 3. Lalani AA, McGregor BA, Albiges L, Choueiri TK, Motzer R, Powles T, Wood C, Bex A. **Systemic treatment of metastatic clear cell renal cell carcinoma in 2018: current paradigms, use of immunotherapy, and future directions**. *Eur Urol* (2019.0) **75** 100-110. DOI: 10.1016/j.eururo.2018.10.010 4. Siska PJ, Beckermann KE, Rathmell WK, Haake SM. **Strategies to overcome therapeutic resistance in renal cell carcinoma**. *Urol Oncol* (2017.0) **35** 102-110. DOI: 10.1016/j.urolonc.2016.12.002 5. Ljungberg B, Bensalah K, Canfield S, Dabestani S, Hofmann F, Hora M, Kuczyk MA, Lam T, Marconi L, Merseburger AS. **EAU guidelines on renal cell carcinoma: 2014 update**. *Eur Urol* (2015.0) **67** 913-924. DOI: 10.1016/j.eururo.2015.01.005 6. Kim SH, Park WS, Chung J. **Tumour heterogeneity in triplet-paired metastatic tumour tissues in metastatic renal cell carcinoma: concordance analysis of target gene sequencing data**. *J Clin Pathol* (2019.0) **72** 152-156. DOI: 10.1136/jclinpath-2018-205456 7. Bedke J, Gauler T, Grunwald V, Hegele A, Herrmann E, Hinz S, Janssen J, Schmitz S, Schostak M, Tesch H. **Systemic therapy in metastatic renal cell carcinoma**. *World J Urol* (2017.0) **35** 179-188. DOI: 10.1007/s00345-016-1868-5 8. Atkins MB, Tannir NM. **Current and emerging therapies for first-line treatment of metastatic clear cell renal cell carcinoma**. *Cancer Treat Rev* (2018.0) **70** 127-137. DOI: 10.1016/j.ctrv.2018.07.009 9. Rini B, Goddard A, Knezevic D, Maddala T, Zhou M, Aydin H, Campbell S, Elson P, Koscielny S, Lopatin M. **A 16-gene assay to predict recurrence after surgery in localised renal cell carcinoma: development and validation studies**. *Lancet Oncol* (2015.0) **16** 676-685. DOI: 10.1016/S1470-2045(15)70167-1 10. Yu J, Mao W, Sun S, Hu Q, Wang C, Xu Z, Liu R, Chen S, Xu B, Chen M. **Identification of an m6A-related lncRNA signature for predicting the prognosis in patients with kidney renal clear cell carcinoma**. *Front Oncol* (2021.0) **11** 663263. DOI: 10.3389/fonc.2021.663263 11. Liao G, Wang P, Wang Y. **Identification of the prognosis value and potential mechanism of immune checkpoints in renal clear cell carcinoma microenvironment**. *Front Oncol* (2021.0) **11** 720125. DOI: 10.3389/fonc.2021.720125 12. Hua X, Chen J, Su Y, Liang C. **Identification of an immune-related risk signature for predicting prognosis in clear cell renal cell carcinoma**. *Aging* (2020.0) **12** 2302-2332. DOI: 10.18632/aging.102746 13. Zhang Q, Lei L, Jing D. **Knockdown of SERPINE1 reverses resistance of triplenegative breast cancer to paclitaxel via suppression of VEGFA**. *Oncol Rep* (2020.0) **44** 1875-1884. PMID: 33000256 14. Yang JD, Ma L, Zhu Z. **SERPINE1 as a cancer-promoting gene in gastric adenocarcinoma: facilitates tumour cell proliferation, migration, and invasion by regulating EMT**. *J Chemother* (2019.0) **31** 408-418. DOI: 10.1080/1120009X.2019.1687996 15. Lopez-Legarrea P, Mansego ML, Zulet MA, Martinez JA. **SERPINE1, PAI-1 protein coding gene, methylation levels and epigenetic relationships with adiposity changes in obese subjects with metabolic syndrome features under dietary restriction**. *J Clin Biochem Nutr* (2013.0) **53** 139-144. DOI: 10.3164/jcbn.13-54 16. Freeberg MAT, Farhat YM, Easa A, Kallenbach JG, Malcolm DW, Buckley MR, Benoit DSW, Awad HA. **Serpine1 Knockdown enhances MMP activity after flexor tendon injury in mice: implications for adhesions therapy**. *Sci Rep* (2018.0) **8** 5810. DOI: 10.1038/s41598-018-24144-1 17. Wu DM, Wang S, Wen X, Han XR, Wang YJ, Fan SH, Zhang ZF, Shan Q, Lu J, Zheng YL. **MircoRNA-1275 promotes proliferation, invasion and migration of glioma cells via SERPINE1**. *J Cell Mol Med* (2018.0) **22** 4963-4974. DOI: 10.1111/jcmm.13760 18. Sun X, Cai Y, Hu XH, Mo M, Zhao C, He W, Li YL. **Long noncoding RNA MAFG-AS1 facilitates bladder cancer tumorigenesis via regulation of miR-143-3p/SERPINE1 axis**. *Transl Cancer Res* (2020.0) **9** 7214-7226. DOI: 10.21037/tcr-20-1971 19. Kong HJ, Kwon EJ, Kwon OS, Lee H, Choi JY, Kim YJ, Kim W, Cha HJ. **Crosstalk between YAP and TGF beta regulates SERPINE1 expression in mesenchymal lung cancer cells**. *Int J Oncol* (2021.0) **58** 111-121. DOI: 10.3892/ijo.2020.5153 20. Wang S, Pang L, Liu Z, Meng X. **SERPINE1 associated with remodeling of the tumor microenvironment in colon cancer progression: a novel therapeutic target**. *BMC Cancer* (2021.0) **21** 767. DOI: 10.1186/s12885-021-08536-7 21. Yang S, Gao K, Li W. **Identification of hub genes and pathways in glioblastoma by bioinformatics analysis**. *Oncol Lett* (2019.0) **17** 1035-1041. PMID: 30655863 22. Bris C, Goudenege D, Desquiret-Dumas V, Charif M, Colin E, Bonneau D, Amati-Bonneau P, Lenaers G, Reynier P, Procaccio V. **Bioinformatics tools and databases to assess the pathogenicity of mitochondrial DNA variants in the field of next generation sequencing**. *Front Genet* (2018.0) **9** 632. DOI: 10.3389/fgene.2018.00632 23. Jones J, Otu H, Spentzos D, Kolia S, Inan M, Beecken WD, Fellbaum C, Gu X, Joseph M, Pantuck AJ. **Gene signatures of progression and metastasis in renal cell cancer**. *Clin Cancer Res* (2005.0) **11** 5730-5739. DOI: 10.1158/1078-0432.CCR-04-2225 24. Liu H, Brannon AR, Reddy AR, Alexe G, Seiler MW, Arreola A, Oza JH, Yao M, Juan D, Liou LS. **Identifying mRNA targets of microRNA dysregulated in cancer: with application to clear cell renal cell carcinoma**. *BMC Syst Biol* (2010.0) **4** 51. DOI: 10.1186/1752-0509-4-51 25. Wozniak MB, Le Calvez-Kelm F, Abedi-Ardekani B, Byrnes G, Durand G, Carreira C, Michelon J, Janout V, Holcatova I, Foretova L. **Integrative genome-wide gene expression profiling of clear cell renal cell carcinoma in Czech Republic and in the United States**. *PLoS ONE* (2013.0) **8** e57886. DOI: 10.1371/journal.pone.0057886 26. Gerlinger M, Horswell S, Larkin J, Rowan AJ, Salm MP, Varela I, Fisher R, McGranahan N, Matthews N, Santos CR. **Genomic architecture and evolution of clear cell renal cell carcinomas defined by multiregion sequencing**. *Nat Genet* (2014.0) **46** 225-233. DOI: 10.1038/ng.2891 27. Takahashi M, Tsukamoto Y, Kai T, Tokunaga A, Nakada C, Hijiya N, Uchida T, Daa T, Nomura T, Sato F. **Downregulation of WDR20 due to loss of 14q is involved in the malignant transformation of clear cell renal cell carcinoma**. *Cancer Sci* (2016.0) **107** 417-423. DOI: 10.1111/cas.12892 28. Huang DW, Sherman BT, Lempicki RA. **Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources**. *Nat Protoc* (2009.0) **4** 44-57. DOI: 10.1038/nprot.2008.211 29. Kanehisa M. **The KEGG database**. *Novart Fdn Symp* (2002.0) **247** 91-103. DOI: 10.1002/0470857897.ch8 30. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K, Dwight SS, Eppig JT. **Gene ontology: tool for the unification of biology. The gene ontology consortium**. *Nat Genet* (2000.0) **25** 25-29. DOI: 10.1038/75556 31. 31.Szklarczyk D, Franceschini A, Wyder S, Forslund K, Heller D, Huerta-Cepas J, Simonovic M, Roth A, Santos A, Tsafou KP et al: STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res 2015, 43(Database issue):D447–452. 32. Smoot ME, Ono K, Ruscheinski J, Wang PL, Ideker T. **Cytoscape 2.8: new features for data integration and network visualization**. *Bioinformatics* (2011.0) **27** 431-432. DOI: 10.1093/bioinformatics/btq675 33. Bandettini WP, Kellman P, Mancini C, Booker OJ, Vasu S, Leung SW, Wilson JR, Shanbhag SM, Chen MY, Arai AE. **MultiContrast Delayed Enhancement (MCODE) improves detection of subendocardial myocardial infarction by late gadolinium enhancement cardiovascular magnetic resonance: a clinical validation study**. *J Cardiovasc Magn Reson* (2012.0) **14** 83. DOI: 10.1186/1532-429X-14-83 34. Gyorffy B. **Survival analysis across the entire transcriptome identifies biomarkers with the highest prognostic power in breast cancer**. *Comput Struct Biotec* (2021.0) **19** 4101-4109. DOI: 10.1016/j.csbj.2021.07.014 35. Li T, Fu J, Zeng Z, Cohen D, Li J, Chen Q, Li B, Liu XS. **TIMER2.0 for analysis of tumor-infiltrating immune cells**. *Nucleic Acids Res* (2020.0) **48** W509-W514. DOI: 10.1093/nar/gkaa407 36. Tang Z, Li C, Kang B, Gao G, Li C, Zhang Z. **GEPIA: a web server for cancer and normal gene expression profiling and interactive analyses**. *Nucleic Acids Res* (2017.0) **45** W98-W102. DOI: 10.1093/nar/gkx247 37. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, Jacobsen A, Byrne CJ, Heuer ML, Larsson E. **The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data**. *Cancer Discov* (2012.0) **2** 401-404. DOI: 10.1158/2159-8290.CD-12-0095 38. Chandrashekar DS, Bashel B, Balasubramanya SAH, Creighton CJ, Ponce-Rodriguez I, Chakravarthi B, Varambally S. **UALCAN: a portal for facilitating tumor subgroup gene expression and survival analyses**. *Neoplasia* (2017.0) **19** 649-658. DOI: 10.1016/j.neo.2017.05.002 39. Wettersten HI, Aboud OA, Lara PN, Weiss RH. **Metabolic reprogramming in clear cell renal cell carcinoma**. *Nat Rev Nephrol* (2017.0) **13** 410-419. DOI: 10.1038/nrneph.2017.59 40. Bai L, Wang W, Xiang Y, Wang S, Wan S, Zhu Y. **Aberrant elevation of GDF8 impairs granulosa cell glucose metabolism via upregulating SERPINE1 expression in patients with PCOS**. *Mol Ther Nucleic Acids* (2021.0) **23** 294-309. DOI: 10.1016/j.omtn.2020.11.005 41. Su Y, Fu J, Du J, Wu B. **First-line treatments for advanced renal-cell carcinoma with immune checkpoint inhibitors: systematic review, network meta-analysis and cost-effectiveness analysis**. *Ther Adv Med Oncol* (2020.0) **12** 1758835920950199. DOI: 10.1177/1758835920950199 42. Shi SN, Qin X, Wang S, Wang WF, Zhu YF, Lin Y, Zhou ZL, Shi BK, Liu XG. **Identification of potential novel differentially-expressed genes and their role in invasion and migration in renal cell carcinoma**. *Aging* (2020.0) **12** 9205-9223. DOI: 10.18632/aging.103192 43. Xu T, Gao S, Liu J, Huang Y, Chen K, Zhang X. **MMP9 and IGFBP1 regulate tumor immune and drive tumor progression in clear cell renal cell carcinoma**. *J Cancer* (2021.0) **12** 2243-2257. DOI: 10.7150/jca.48664 44. Hu Q, Peng J, Chen X, Li H, Song M, Cheng B, Wu T. **Obesity and genes related to lipid metabolism predict poor survival in oral squamous cell carcinoma**. *Oral Oncol* (2019.0) **89** 14-22. DOI: 10.1016/j.oraloncology.2018.12.006 45. Yang K, Zhang S, Zhang D, Tao Q, Zhang T, Liu G, Liu X, Zhao T. **Identification of SERPINE1, PLAU and ACTA1 as biomarkers of head and neck squamous cell carcinoma based on integrated bioinformatics analysis**. *Int J Clin Oncol* (2019.0) **24** 1030-1041. DOI: 10.1007/s10147-019-01435-9 46. Ghosh AK, Murphy SB, Kishore R, Vaughan DE. **Global gene expression profiling in PAI-1 knockout murine heart and kidney: molecular basis of cardiac-selective fibrosis**. *PLoS ONE* (2013.0) **8** e63825. DOI: 10.1371/journal.pone.0063825 47. Higgins PJ. **Balancing AhR-dependent pro-oxidant and Nrf2-responsive anti-oxidant pathways in age-related retinopathy: is SERPINE1 expression a therapeutic target in disease onset and progression?**. *J Mol Genet Med* (2014.0) **8** 101. PMID: 25237384 48. Freytag J, Wilkins-Port CE, Higgins CE, Higgins SP, Samarakoon R, Higgins PJ. **PAI-1 mediates the TGF-beta1+EGF-induced "scatter" response in transformed human keratinocytes**. *J Invest Dermatol* (2010.0) **130** 2179-2190. DOI: 10.1038/jid.2010.106 49. Pavon MA, Arroyo-Solera I, Tellez-Gabriel M, Leon X, Viros D, Lopez M, Gallardo A, Cespedes MV, Casanova I, Lopez-Pousa A. **Enhanced cell migration and apoptosis resistance may underlie the association between high SERPINE1 expression and poor outcome in head and neck carcinoma patients**. *Oncotarget* (2015.0) **6** 29016-29033. DOI: 10.18632/oncotarget.5032 50. Yao H, He G, Chen C, Yan S, Lu L, Song L, Vijayan KV, Li Q, Xiong L, Miao X. **PAI1: a novel PP1-interacting protein that mediates human plasma's anti-apoptotic effect in endothelial cells**. *J Cell Mol Med* (2017.0) **21** 2068-2076. DOI: 10.1111/jcmm.13127 51. 51.Yu M, Chu ST, Fei BY, Fang XD, Liu Z: O-GlcNAcylation of ITGA5 facilitates the occurrence and development of colorectal cancer. Exp Cell Res 2019, 382(2–3). 52. 52.Yao XX, He ZK, Qin CLT, Deng XQ, Bai L, Li GX, Shi JL: SLC2A3 promotes macrophage infiltration by glycolysis reprogramming in gastric cancer. Cancer Cell Int 2020, 20(1). 53. Zhao ZC, Zhang GJ, Li WD. **MT2A promotes oxaliplatin resistance in colorectal cancer cells**. *Cell Biochem Biophys* (2020.0) **78** 475-482. DOI: 10.1007/s12013-020-00930-5 54. Wright KD, Miller BS, El-Meanawy S, Tsaih SW, Banerjee A, Geurts AM, Sheinin Y, Sun Y, Kalyanaraman B, Rui H. **The p52 isoform of SHC1 is a key driver of breast cancer initiation**. *Breast Cancer Res* (2019.0) **21** 74. DOI: 10.1186/s13058-019-1155-7 55. Wu H, Liu B, Chen Z, Li G, Zhang Z. **MSC-induced lncRNA HCP5 drove fatty acid oxidation through miR-3619-5p/AMPK/PGC1alpha/CEBPB axis to promote stemness and chemo-resistance of gastric cancer**. *Cell Death Dis* (2020.0) **11** 233. DOI: 10.1038/s41419-020-2426-z 56. Afshar AS, Nematpour FS, Meshkani M, Khafi A. **Growth inhibition of human breast cancer cells and down-regulation of ODC1 and ADA genes by Nepeta binaloudensis**. *Rev Bras Farmacogn* (2017.0) **27** 84-90. DOI: 10.1016/j.bjp.2016.07.005
--- title: Experimental diets dictate the metabolic benefits of probiotics in obesity authors: - Ida Søgaard Larsen - Béatrice S.-Y. Choi - Bandik Föh - Nanna Ny Kristensen - Adia Ouellette - Rune Falkenberg Haller - Peter Bjarke Olsen - Delphine Saulnier - Christian Sina - Benjamin A. H. Jensen - André Marette journal: Gut Microbes year: 2023 pmcid: PMC10038044 doi: 10.1080/19490976.2023.2192547 license: CC BY 4.0 --- # Experimental diets dictate the metabolic benefits of probiotics in obesity ## ABSTRACT Growing evidence supports the use of probiotics to prevent or mitigate obesity-related dysmetabolism and non-alcoholic fatty liver disease (NAFLD). However, frequent reports of responders versus non-responders to probiotic treatment warrant a better understanding of key modifiers of host–microbe interactions. The influence of host diet on probiotic efficacy, in particular against metabolic diseases, remains elusive. We fed C57BL6/J mice a low fat reference diet or one of two energy-matched high fat and high sucrose diets for 12 weeks; a classical high fat diet (HFD) and a customized fast food-mimicking diet (FFMD). During the studies, mice fed either obesogenic diet were gavaged daily with one of two probiotic lactic acid bacteria (LAB) strains previously classified as Lactobaccillus, namely Limosilactobacillus reuteri (L. reuteri)or Lacticaseibacillus paracaseisubsp. paracasei (L. paracasei), or vehicle. The tested probiotics exhibited a reproducible efficacy but dichotomous response according to the obesogenic diets used. Indeed, L. paracaseiprevented weight gain, improved insulin sensitivity, and protected against NAFLD development in mice fed HFD, but not FFMD. Conversely, L. reuteri improved glucoregulatory capacity, reduced NAFLD development, and increased distal gut bile acid levels associated with changes in predicted functions of the gut microbiota exclusively in the context of FFMD-feeding. We found that the probiotic efficacy of two LAB strains is highly dependent on experimental obesogenic diets. These findings highlight the need to carefully consider the confounding impact of diet in order to improve both the reproducibility of preclinical probiotic studies and their clinical research translatability. ## GRAPHICAL ABSTRACT ## Introduction Since 1975 the global prevalence of overweight and obesity has nearly tripled to ~$40\%$ of the adult population1. Obesity is a leading risk factor of cardiometabolic multimorbidity2 including type 2 diabetes3,4 and nonalcoholic fatty liver disease (NAFLD)5, also known as metabolic associated fatty liver disease (MAFLD)6, which affects ~$25\%$ of the population worldwide7. The etiology of obesity is highly complex, involving positive energy balance resulting from poor dietary habits and sedentarity8 but also features hereditary and socioeconomic traits9,10. Although lifestyle interventions such as dietary restrictions may facilitate acute weight loss, this tends to decrease over time resulting in recurrent weight gain11–14. Pharmacological therapies against obesity with proven clinical efficacy are emerging, however adverse effects such as nausea are frequently reported among users15. Surgical strategies, most notably gastric bypass, are more successful yet highly invasive and not spared from potential perioperative complications16. Probiotics – defined as “live microorganisms which when administered in adequate amounts confer a health benefit on the host”17 - exhibit promising preclinical potential to reduce obesity, NAFLD, and insulin resistance. Probiotics are therefore considered a noninvasive yet potentially highly effective therapeutic approach against obesity and associated metabolic diseases18,19. Still, evidence from clinical trials is inconsistent and often of low quality, as the probiotic field is loosely regulated and evidence required for approval by authorities highly dependent on geographical region20,21. This emphasizes a critical need for an enhanced molecular understanding of strain-specific capabilities22 including potential adverse effects, as reported recently for probiotic use in cancer immunotherapy23. Human studies with probiotic supplementation further reveal responders and non-responders as a common trait, both in terms of colonization and general host effects24,25. Whether the lack of consistent probiotic efficacy in previous studies relates to methodological, genetic and/or environmental factors, such as dietary patterns, remains largely unknown24,25. A limited number of studies have reported efficacy discrepancies of both lactic acid bacteria (LAB) strains and other taxa in mice depending on the diet they consumed. Indeed, L. helveticus R0052 induced different effects on anxiety-like behavior depending on whether the mice consumed a fiber-rich chow or high-fat Western diet26, which was also the case for L. plantarum WCFS1’s effects on bacterial gene expression and host intestinal inflammation27. On a similar note, Prevotella copri appears to exhibit opposing effects on glucose metabolism in both humans and mice depending on dietary challenge, being either a fiber-rich low fat diet28 or a low fiber, high fat diet29. Importantly, these studies were confounded by dietary differences in energy and fiber content. To our knowledge, no studies have yet examined probiotic efficacy in different but energy-matched obesogenic diets resulting in distinct metabolic phenotypes and disease severity in preclinical murine models. This is critically important to optimize the clinical translatability of preclinical research in the probiotic field. In this study, we therefore evaluated the efficacy of two LAB strains in mice fed either a standard HFD commonly used in the field or a customized diet designed to exacerbate NAFLD and nonalcoholic steatohepatitis (NASH). We demonstrate that the efficacy of the tested bacterial strains on preventing obesity, insulin resistance, and NAFLD development is entirely dependent on the obesogenic diet used, underlining the critical importance of considering the composition of experimental diets in preclinical studies for evaluating the effects of probiotics. Our work further highlights the need to carefully assess dietary patterns and nutritional factors as potential confounders of probiotic efficacy in clinical studies. ## Animals study design and diets Three independent studies with identical experimental outlines were carried out using male C57BL6/J mice (Jackson Lab) co-housed three mice per cage. Twelve days prior to the start of the study period the mice were changed from a chow diet to a compositionally defined low fat diet (LFD) during acclimatization. Experimental groups were stratified based on similar average and variance of baseline body weight, fat mass, and additionally for Study 1, 5 h fasting blood was sampled for determinations of plasma glucose and insulin levels before starting the experimental diets (described in a later section) without disturbing the social hierarchy within cages (Figure S1A-E). Study period was set to 12 weeks, allowing for significant weight gain, glucoregulatory disturbances and NAFLD onset30. During the 12-week study period mice were fed experimental compositional defined diets (CDD) from eight weeks of age receiving either a commonly used high fat diet (HFD, D12451, provided by Ssniff Spezialdiäten (Study 1) or Research diets (Study 2), a customized fast food-mimicking diet (FFMD, D12079 mod. provided by Ssniff Spezialdiäten (Study 1 and 3) or continued on the LFD by Ssniff Spezialdiäten consisting of dietary sources matched to the FFMD (Figure 1a, Table 1). Feed intake was monitored and exchanged three times a week and body weight was measured weekly. Body composition was assessed by magnetic resonance (MR) scan (Minispec LF290 NMR analyzer, Bruker) prior to, and then every four weeks during the study. Energy uptake was measured at week 11 of the study by monitoring the feed intake and fecal excretion during 24 h in clean cages with bedding followed by energy excretion using a calorimeter. Due to co-caging energy in- and uptake was only possible to assess at cage-level and presented as average/mouse/cage with n-size equaling the number of cages per group. From the start of feeding experimental diets, mice were orally gavaged daily 4 h into the light cycle with 100 µL of either vehicle (PBS, Gibco) or 108 CFU Limosilactobacillus reuteri, formerly known as Lactobacillus reuteri, DSM 32,910 (L. reuteri) or Lacticaseibacillus paracasei subsp. paracasei, formerly classified as Lactobacillus paracasei subsp. tolerans, DSM 32,851 (L. paracasei) freshly thawed from glycerol stocks stored at −80°C and the live cells diluted in PBS (Gibco). After twelve weeks, the mice were anaesthetized with isoflourane (Fresenius Kabi) and euthanized by cervical dislocation in alternating order after 5 h of fasting from 2 h into the light cycle. Cardiac puncture was performed using a 25 G needle and 1 mL syringe coated with EDTA (Sigma-Aldrich). Blood was transferred to Eppendorf tubes containing 1 μL of DPP IV inhibitor (Millipore) and 1 μL of a protease inhibitor cocktail (Sigma) and kept on ice. The samples were centrifuged at 1000 rcf at 4°C for 10 min and the plasma was placed on dry ice and transferred to −80°C storage until further processing. Figure 1.Body composition and insulin sensitivity are diet-dependently affected by probiotic strains. a) Experimental outline for all three studies with 12 weeks of feeding either High Fat Diet (HFD), Fast Food-mimicking Diet (FFMD) or Low Fat reference Diet (LFD) (matched to the FFMD in nutritional sources) with indicated total energy content in MJ/kg and energy distribution of macronutrients in percentage. Shades indicate mixed nutritional sources of the macronutrients. During the study, the mice were orally gavaged daily with either 108 CFU of either Limosilactobacillus reuteri (L. reuteri), Lacticaseibacillus paracasei subsp. paracasei (L. paracasei), or an equal volume of vehicle (PBS). b) Body weight in grams at the end of the study period. c) Fat mass in grams assessed by MR scan at the end of the study period. d) Energy uptake per cage monitored over 24 h as the inverse percentage of energy excretion compared to energy intake assessed in week 11 of the study. e) Accumulated energy intake per cage during the study period as average intake per mouse. f) Plasma 5 h fasting insulin assessed in week 10 of the study. g) Blood glucose levels of 5 h fasted mice in week 10 of the study. h) Blood glucose values from Study 1 during oral glucose tolerance test (oGTT) with 1.5 µg/g lean mass dextrose in week 10. i) *Corresponding plasma* insulin levels during oGTT. j) *Corresponding plasma* C-peptide levels during oGTT. b-g) Bars indicate group mean ± SEM. Points represent individual mice; individual studies are identified by point shapes where circles indicate Study 1, triangles Study 2, and squares Study 3. Dashed line indicates LFD group mean with interquartile range shown in gray. Asterisks indicate fdr-corrected q-values <0.05 using linear mixed effects models comparing indicated groups. h-j) Lines indicate group mean ± SEM. Asterisks indicate p-values <0.05 using two-way ANOVA with multiple comparisons between vehicle-treated groups or Lactobacillu treated group to vehicle-treated group fed the same diet with Bonferroni post-hoc test. b, c, f, g) LFD+Vehicle $$n = 21$$, HFD+Vehicle $$n = 20$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 20$$, FFMD+Vehicle $$n = 18$$, FFMD+L.Reuteri $$n = 21$$, FFMD+L. paracasei $$n = 20$.$ d-e) LFD+Vehicle $$n = 7$$, HFD+Vehicle $$n = 7$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 7$$, FFMD+Vehicle $$n = 6$$, FFMD+L.Reuteri $$n = 7$$, FFMD+L. paracasei $$n = 7$.$ h-j) LFD+Vehicle $$n = 8$$, HFD+Vehicle $$n = 9$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 9$$, FFMD+Vehicle $$n = 9$$ (i-j $$n = 8$$ due to insufficient plasma volume from one mouse), FFMD+L.Reuteri $$n = 9$$, FFMD+L. paracasei $$n = 8$.$ Table 1.Composition of diets. ProteinFat Low Fat Diet ControlFast Food Mimicking DietHigh fat dietIngredientcontentcontentUmixed Proteinmixed Protein45 kcal% FatProduct No. S9552-E026S9552-E024S9552-E060Casein87<$0.6\%$3,0003,00025,000Poultry protein$7110\%$8,5008,500——Soyprotein isolate880,$4\%$4,9004,900——Wheat gluten, Vital76,51,$7\%$2,3002,300——Egg yolk$3856\%$1,6001,600——Egg white790,$1\%$1,6001,600——Whey powder, $71\%$ Lactose11,$91\%$2,9002,900——L-Cystine72,2——%0,3000,3000,300Corn starch————%45,2605,2007,000Maltodextrin————%8,0008,00011,000Sucrose 1)————%7,00024,40020,050Fructose 1)————%——6,500——Cellulose powder————%5,0005,0005,700Vitamin premixture 2)————%1,0001,0001,000Mineral premixture + additions 2)————%4,3004,3006,000Choline Cl ($50\%$)————%0,2900,2900,200L-Carnitine————%0,0500,050——Cholesterol————%——0,1500,150Butylated hydroxytoluene————%0,0100,010——Butter fat, dehydrated————%2,10018,800——Corn oil————%1,6201,100——Linseed oil————%0,2700,100——Pork lard————%————20,800Soybean oil————%————2,800Proximate contents Crude protein %17,217,222,100Crude fat %5,821,823,600Crude fibre %5,05,15,700Crude ash %4,84,85,400Starch %43,55,06,700Dextrin %7,97,910,900Sugar (total) %10,234,021,100Energy (Atwater) MJ/kg15,719,219,300kcal% Protein 181520,000kcal% Fat 144345,000kcal% Carbohydrates 684235,000Lysine %0,950,951,83Methionine %0,430,430,78Methionine & Cystine %0,940,941,18Threonine %0,620,620,97Tryptophan %0,190,190,29Arginine %0,910,910,86Histidine %0,370,370,67Valine %0,860,861,54Isoleucine %0,710,711,25Leucine %1,231,232,19Phenylalanine %0,740,741,14Phe + Tyr %1,291,292,31Glycine %0,960,960,47Glutamic acid %3,003,004,97Aspartic acid %1,361,361,64Proline %1,201,202,53Serine %0,820,821,32Alanine %0,840,840,66Fatty acids, % in the diet C 4:0 0,080,72——C 6:0 0,050,47——C 8:0 0,030,26——C 10:0 0,060,560,03C 12:0 0,070,640,05C 14:0 0,242,000,29C 16:0 1,155,505,34C 18:0 0,361,972,92C 20:0 0,020,040,07C 16:1 0,130,430,62C 18:1 1,685,169,42C 18:2 1,211,213,45C 18:3 0,170,170,37Cholesterol mg/kg~440~2340~1580Calcium %0,820,820,92Phosphorus %0,580,570,64Sodium %0,390,380,20Magnesium %0,10,10,23Potassium %0,80,80,97Iron mg/kg7272168Manganese mg/kg222295Zinc mg/kg424265Copper mg/kg131313Iodine mg/kg0,310,311,16Selenium mg/kg0,200,200,18Vitamin A IU/kg15.90020.50015.000Vitamin D3 IU/kg1.5501.5501.500Vitamin E mg/kg156159150Vitamin C mg/kg303030Vitamin K mg/kg202020Vitamin B1 mg/kg262625Vitamin B2 mg/kg181816Vitamin B6 mg/kg171716Vitamin B12 µg/kg323230Niacin mg/kg535347Pantothenic acid mg/kg595955Folic acid mg/kg161616Biotin µg/kg355355300Choline mg/kg1.3801.380920 The weight of the liver, white adipose tissues (WATs), and cecum were measured, and the tissues immediately frozen in liquid nitrogen and stored at −80°C until further processing. Tissues were dissected by the same operator and taken in the same order for all mice in each study. The small intestine (from stomach to cecum) and colon (from cecum to rectum) were kept on a Plexiglas plate cooled by underlying wet ice throughout the handling time. Duodenum was considered the first 5 cm of the small intestine and the remaining small intestine tissue was divided into three parts of equal length. The proximal 3 cm of the distal $\frac{1}{3}$ of the small intestine was discarded and the remaining tissue categorized as ileum. Content of the small intestine, cecum, and colon were isolated by mechanical pressure, frozen on dry ice, and subsequently stored at −80°C. Tissues from the ileum and colon were snap-frozen in liquid nitrogen and stored at −80°C. The animal studies were conducted in accordance with Canadian Council on Animal Care under the Laval University license 2017–086. ## Assessment of glucose homeostasis, oral glucose tolerance, glucose-stimulated insulin, and C-peptide levels Study 1 baseline glucose homeostasis was assessed in 5 h fasted mice the day of study initiation. Blood glucose was measured by sampling blood from the tail vein using a glucometer (OneTouch Vario Flex, LifeScan) and plasma sampled using EDTA prepared capillary tubes (Sarstedt) kept on ice until blood samples were centrifuged for 10 min at 1000 rcf at 4°C. Plasma insulin was quantified by Mouse Ultrasensitive Insulin ELISA (Alpco) using the manufacturer’s protocol. An oral glucose tolerance test was carried out in week 10 of the experimental protocol. Mice were fasted from 2 h into the light cycle for 5 h still receiving their usual oral daily gavage. Fasting blood glucose and plasma insulin were measured as described for baseline glucose homeostasis prior to oral gavage with 1.5 µg/g (Study 1 and 3) or 2.0 µg/g (Study 2) lean mass of dextrose. Blood glucose was measured from tail vein puncture at time points 15, 30, 60, 90, and 120 min after dextrose challenge. Blood samples for quantification of plasma insulin levels were sampled in EDTA prepared capillary tubes (Sarstedt) at time points 0, 15, 30, 60, and 120 min and for C-peptide at time points 0, 15, and 30 min post oral glucose challenge. Mice were subcutaneously injected with 0.5 mL saline (Hospira) after the procedure for rehydration. Blood samples were centrifuged for 10 min at 1000 rcf at 4°C. Plasma insulin was quantified by Mouse ultrasensitive Insulin ELISA (Alpco) and C-peptide by Mouse C-peptide ELISA kit (Crystal Chem) following the manufacturers’ protocols. ## Liver triacylglycerol (TG) quantification TG was quantified in snap-frozen liver tissue stored at −80°C until cryo-grinding in liquid nitrogen and 50 ± 5 mg tissue added 0.9 mL of a 2:1 chloroform:methanol solution and homogenized 1 min at 50 os/sec using a TissueLyser LT (Qiagen) with beads. The content was transferred to new tubes and added 0.3 mL methanol and incubated for 2 h with rotation. The tubes were centrifuged for 15 min at 4000 rcf at 10°C followed by transfer of 0.825 mL supernatant to new tubes and 0.4 mL chloroform was added. The samples were vortexed, 0.275 mL of $0.73\%$ NaCl was added and samples were vortexed again for 30 sec followed by centrifugation for 3 min at 4000 rcf at 10°C. The upper phase was aspirated and the tubes rinsed with 0.8 mL Folch solution (Chloroform:Methanol:NaCl $0.58\%$ (3:48:47)) repeatedly for three washes followed by addition of 2–5 drops of methanol until the solution appeared clear after vortexing. The liquid was evaporated using liquid nitrogen and the extracted lipids were resuspended in a solution of isopropanol with $10\%$ TritonX-100 and TG levels quantified by Infinity Triglycerides Reagent (ThermoFisher) and calculated from serial diluted standard curves. ## Lipidomics of liver tissue and plasma Liver and plasma lipids were extracted in a solvent consisting 89.9:10:0.1 of Heptan/IPA/HCOOH and analyzed by liquid chromatography-mass spectrometry (LC-MS) (Thermo Scientific). The MS was operated in negative and positive ion mode electrospray ionization mode. Data was collected from 80 to 1500 m/z. Lipids were normalized to sample median by log transformation and Pareto scaling using MetaboAnalyst31, which was additionally used to generate heatmaps of the top 40 or 100 features of each measure as indicated in figure legends. ## Histopathological scoring of liver tissue Liver tissue (lobus dexter medialis hepatis) was collected and fixed in $4\%$ paraformaldehyde for 3 d. After paraffin embedding 3 µm sections were prepared and stained with hematoxylin and eosin (H&E) or Sirius Red. NAFLD activity was assessed blindly in H&E-stained cross-sections using an established scoring system for murine NAFLD32. In short, the levels of hepatocellular hypertrophy, macrovesicular, and microvesicular steatosis were determined at 40× to 100× magnification relative to the total liver area and scored as described in Supplementary Table S1. Hepatic inflammation was scored by counting the number of inflammatory foci per 1 mm2 using a 100× magnification. Inflammatory foci were defined as clusters (not rows) of at least five inflammatory cells and the mean of five random fields was used for scoring, as previously described32. Histopathological categories were based on previously described methodology32 where liver slides are categorized with NAFLD if the steatosis score (based on the sum of micro-, macrovesicular steatosis, and hepatocellular hypertrophy scores) ≥1 with an inflammation score of 0. Mice are categorized with NASH if steatosis score ≥1 and an inflammation score ≥1. Livers are considered healthy if the steatosis score = 0 regardless of the inflammation score. Fibrosis was blindly assessed by the *Ishak fibrosis* stage using Sirius Red-stained cross-sections following previously published methods33. The scoring system introduced by Ishak et al. allows for a sensitive scoring of fibrosis that takes quantity and location of fibrosis into account34 and is described further in Supplementary Table S2. ## Cytokine quantification in liver and adipose tissue Cryo-grinded liver tissues were washed by adding 200 µL ice-cold PBS and protease inhibitors (P8340 Sigma-Aldrich), centrifugation at 16,000 × g for 20 sec at 4°C and aspiration of the liquidphase. This wash step was repeated a total of 10 times. Afterwards 400 µL T-PER Tissue Protein Extraction Buffer (ThermoFisher) and beads were added and vortexed for 2 × 1 min at 50 os/sec. Samples were centrifuged at 4,000 rcf for 1 min at 4°C and the supernatants transferred to new tubes that were centrifuged again at 13,000 rcf for 10 min at 4°C. Snap-frozen mWAT was broken into pieces of 100–200 mg tissue then added acid washed zirconium beads (OPS Diagnostics) while kept frozen. mWAT samples were homogenized in ice-cold lysis buffer (PBS+$1\%$ IGEPAL CA-630 (Sigma) plus protease inhibitors (P8340 Sigma-Alrich)) added 500 µL/100 mg tissue using a Bead Mill homogenizer (VWR) and continuously rotated at 4°C for 30 min. Samples were centrifuged at 12,000 rcf for 15 min at 4°C, supernatants transferred to clean tubes to repeat the centrifugation for 10 min. Protein concentration from liver and mWAT extracts were measured in triplicates using Pierce BCA Protein Assay kit (ThermoFisher) following the manufacturer’s instructions. Cytokine levels in samples normalized to the protein concentration were measured using ProcartaPlex Immunoassay (ThermoFisher) for liver samples or Bio-Plex Pro Mouse Th17 Cytokine Assay (Bio-Rad) for mWAT samples following the manufacturers’ instructions on the Bio-Plex 200 system (Bio-Rad). ## Bile acid (BA) characterization in liver tissue and cecum content BAs in liver tissue and cecum content were extracted with methanol and 5 µL was injected onto a reverse-phased chromatography (1290 Infinity II, Agilent) and mass spectrometer (6546 Q-TOF/MS, Agilent). The MS was operated in negative ion mode electrospray ionization. Data were collected from 80 to 1600 m/z. The chromatographic part of the LC-MS-system was set up with a CSH-peptide C18 column (Waters) using a gradient system with Eluent A: water with $0.1\%$ formic acid and Eluent B: Acetonitrile with $0.1\%$ formic acid. A flow of 0.250 mL/min was used, starting at $99.5\%$ Eluent A lowering to $1\%$ within 19 min followed by a new gradient endpoint $99.5\%$ Eluent A at 25.5 min. The gradient was kept isocratic until 30 min and the eluent returned to initial conditions at 26 min. All gradients were linear. Results were related to serial diluted standard curves. ## Tissue gene expression by quantitative reverse transcription PCR (RT-qPCR) RNA from snap-frozen ileum and liver tissues were extracted using the Quick-RNA Miniprep Plus kit (Zymo Research). cDNA was synthesized from 1.5 µg ileum or 2 µg liver RNA using High-Capacity cDNA Reverse Transcriptase kit (Applied Bioscience) following the manufacturer’s protocol. qPCR was carried out using 4 µl of cDNA, 5 µl of Advanced qPCR MasterMix (Wisent Bioproducts) and 0.5 µl of each primer (diluted at a concentration of 10 µM) in a total reaction volume of 10 µL with the following cycle setting: 95°C for 2 minutes, (95°C for 20 sec, 61.5°C for 20 sec, 72°C for 20 sec) x 40 ending with a melting curve: 65°C to 95°C. The targets were evaluated in each tissue and accepted in case of a unified peak from the melting curve and amplification efficiency of $100\%$±11 and R2>0.99 from a standard curve. Relative expression was calculated by 2ΔCq of target Cq to 18S Cq of the sample accepting replicates with coefficient of variation <0.02. Target primer sequences are reported in Supplementary Table S3. ## Microbiota profiling by 16S rRNA gene amplicon sequencing Fecal samples were collected prior to the start of the studies and every 4 weeks throughout the study period at the same time as body composition assessment 22–24 h after the latest oral gavage with probiotic strains or vehicle. During necropsy 2–5 h after the latest oral gavage contents of small intestine and colon were collected. All samples were immediately placed on dry ice and stored at −80°C until DNA extraction using the NucleoSpin 96 soil kit (Macherey-Nagel) following the manufacturer’s instructions. Stocks of the administered LAB strains were processed similarly to the remaining samples. The 16S rRNA gene was amplified over 25 cycles using primers for the V3-V4 region with Illumina adaptors (S-D-Bact-0341-b-S-17: 5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3‘and S-D-Bact-0785-a-A-21: 5’-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3’) 35. Indices were added in a second PCR over 8 cycles with unique primer combinations using the Nextera XT Index Kit V2 (Illumina). The samples were pooled and cleaned using AMPure XP beads (Beckman Coulter) and the library was sequenced on an Illumina MiSeq desktop sequencer using the MiSeq Reagent Kit V3 (Illumina) for 2 × 300 bp paired-end sequencing. *The* generation of an amplicon sequence variant (ASV) table was carried out using usearch version 10.0.240. Primer binding regions were removed with fastx_truncate and reads were filtered to contain less than one error per read. The quality filtered reads were denoised with unoise3. ASV abundance was calculated by mapping with usearch global using a $97\%$ identity threshold. The phylogenetic tree was made by aligning the 16S sequences with mafft, and the tree was inferred by FastTree. Taxonomical classification was done with the qiime classifier (qiime2–2019.4) trained on the Silva database (Silva_132) as previously described36. The dataset had a median of 45,468, a mean of 46,400 reads per sample with a standard deviation of 19,177 and including samples with a minimum of 10,000 reads. Alpha diversity measures and analysis of differential abundances were calculated based on rarefied data. Principal coordinate analysis (PCoA) of microbiota composition was carried out using weighted UniFrac distances. Functions of the microbiota were predicted using the Kyoto Encyclopedia of Genes and Genomes (KEGG) orthology generated by Picrust2 and illustrated with PCoAs using Bray-Curtis distances. Specific changes in regulated predicted pathways were tested using DeSeq2 between groups of interest reporting p-value and false discovery (fdr)-corrected q-values. ## Correcting for multiple hypothesis testing For all analyses described below and throughout the manuscript, nominal p-values were corrected for multiple hypothesis testing to fdr (q-values) using the Benjamini–Hochberg method37. ## Software and Statistical analysis The statistical analysis is specified in the figure legend of each figure. Statistical analysis by linear mixed-effects models was carried out in R using the lmerTest package38 including study, batch, and cage as nested random effects. Statistical significance was illustrated by asterisks based on fdr-corrected q-values. Two-way analysis of variance (ANOVA) were conducted to analyze the effects of treatments and diet in addition to Bonferroni multiple comparison test to respective vehicle-treated control on the same diet. Categorical data was tested for statistical differences by Kruskal-Wallis and Mann–Whitney tests with Dunn’s post hoc for multiple comparisons testing. Statistical analysis of lipidomics data was carried out by one-way ANOVA with Tukey’s post-hoc analysis in the MetaboAnalyst software. The graphical abstract was created with BioRender.com. Bar charts were made using GraphPad Prism 9. Analysis of community-based statistical differences on gut microbiota composition and its predicted functions were carried out by permutational multivariate analysis of variance (PERMANOVA) by the R package vegan39 using weighted UniFrac or Bray-Curtis distances, respectively, reporting an average p-value of 100 permutations. The glmer.nb function from the lme4 package40 fitting a generalized linear mixed-effects model (GLMM) for the negative binomial family building on glmer was used for the analysis of differential abundance of bacterial phylae on rarefied data. The top 100 most abundant taxa on each taxonomy rank were used in the analysis. Cage was used as random effect in the model. Multiple testing correction was done with the fdr method. To provide full transparency of reported findings, parameters assessed in independent animal studies were shown in the same graph with experiment-dependent point shapes, as indicated in figure legends when appropriate, allowing the reader to visually assess experiment-to-experiment variation. To avoid unwarranted power inflations, we used linear mixed-effects model as described above, including study, batch, and cages as nested random effects and only reports fdr-corrected q-values. Collectively, these precautions and strict statistical analysis enhance experimental reproducibility, and thus study validity. Study 1 sample size (n) = 9 mice per group except for FFMD + L. paracasei group with $$n = 8$.$ Study 2 $$n = 11$$ mice per group. Study 3 $$n = 12$$ mice per group except for FFMD + Vehicle group with $$n = 9$.$ Asterisks identify statistically significant fdr-corrected q-values where * = <0.05, ** = <0.01 and *** = <0.001. ## Experimental diet determines probiotic efficacy on body composition and insulin sensitivity We ensured equal starting points between the experimental groups and fed the mice one of two energy-matched high fat, high sucrose diets or a low fat reference diet (LFD) for 12 weeks with daily oral gavage of either L. reuteri, L. paracasei, or vehicle (Figure 1a, S1A-E, Table 1). Despite similar feed energy content, vehicle-treated FFMD-fed mice exhibited enhanced body weight-, fat- and lean mass gain compared to their classical HFD-fed counterparts (Figure 1b,c, Fig S1F). This trait was closely associated with an increase in energy uptake (Figure 1d) and a similar trend for energy intake ($$p \leq 0.02$$, $q = 0.11$, Figure 1e). Of note, fat percentage, subcutaneous, epididymal, and retroperitoneal white adipose tissues (WATs) were equally increased by FFMD and HFD, whereas FFMD augmented visceral mesenteric WAT deposition to a higher degree than HFD resulting in higher total mWAT inflammation as revealed by the increased levels of several pro-inflammatory cytokines in the tissue, which were otherwise not significantly influenced by probiotic treatments (Fig S1G-N). L. paracasei reduced body weight and fat mass gain in HFD-, but not in FFMD-fed mice (Figure 1b,c S1G). This effect of L. paracasei on body composition was not explained by a reduction of energy intake or uptake (Figure 1d,e). We next investigated the effect of the probiotic strains on glucose homeostasis after 10 weeks of feeding the obesogenic diets. Importantly, while the classical HFD impaired glucose homeostasis and increased both blood glucose and glucose-stimulated insulin secretion (GSIS) during glucose challenge compared to LFD-fed mice, FFMD-feeding significantly aggravated all of the above (Figure 1f–j). Strikingly, FFMD-fed mice more than doubled their fasting insulin levels compared to HFD-fed counterparts (Figure 1f). Moreover, although fasting blood glucose was unaffected by the type of obesogenic diet, FFMD-fed mice failed to return to baseline blood glucose levels two hours post challenge accompanied by a 2–3 fold increase in GSIS, suggesting an early onset of frank diabetes in the latter dietary model (Figure 1g–j). Notably, L. reuteri lowered fasting insulin levels and improved glucose tolerance, with a concomitant trend toward diminished GSIS based on both insulin and c-peptide excursions exclusively in FFMD-fed mice (Figure 1f–j, S1M-N). In contrast, L. paracasei improved glucose tolerance in HFD-fed mice after high dose (Fig S1O-P) but not low dose (Figure 1h–j) glucose challenge. Collectively, these data demonstrates that the metabolic benefits of two Lactobacilli strains are highly dependent on the energy-dense obesogenic diets used. ## NAFLD development was diet-dependently mitigated by probiotic strains As a major glucoregulatory organ, we next assessed hepatic lipid accumulation. Both liver weight per se and TG concentration were >3-fold higher in FFMD-fed mice compared to their HFD-fed counterparts (Figure 2a–c). In agreement with enhanced liver TGs, FFMD feeding altered the hepatic lipid profile (Fig S2A), and increased the clinical marker of liver damage, plasma ALT, associating with a borderline reduction in AST/ALT ratio ($$p \leq 0.02$$, $q = 0.09$, Figure 2d,e S2C). FFMD aggravated both histological prevalence and severity of diet-induced NAFLD, with ~$90\%$ of the animals in the FFMD + Vehicle group having histopathologically validated NASH, contrasting with only ~$45\%$ of NASH observed in HFD + Vehicle mice (Figure 2f–g). Accordingly, FFMD-fed mice developed significantly increased hepatocellular hypertrophy and microvesicular steatosis, as well as excessive hepatic inflammation and moderate fibrosis (Figure 2h–n, S2D-E) all pointing toward aggravated liver damage41 as compared to their HFD-fed counterparts. Of note, and in line with previous publications demonstrating that reference CDD stimulate modest fat accretion42,43, ~$5\%$ of the LFD-fed mice exhibited mild NAFLD (Figure 2g) possibly linked to incorporation of sucrose in the LFD diet to match the nutrient composition of obesogenic diets. Figure 2.Nonalcoholic fatty liver disease (NAFLD) mitigation by probiotic strains depends on diet. a) Images of total livers from necropsy from representative mice in vehicle-gavaged dietary groups of Study 1 weighting respectively 1.01 g (LFD), 1.58 g (HFD) and 3.37 g (FFMD). b) Liver weight in grams at the end of the study. c) Liver triacylglyceride concentration as µg/mg tissue. d) Alanine transaminase (ALT) levels in plasma at the end of the study. e) Aspartate transaminase (AST) to ALT ratio in plasma. f) *Histological steatosis* score from 0–7 blindly assessed in H&E stained liver slides. g) Histopathological NAFLD categories as relative distribution per group of: No NAFLD, NAFLD, and NASH based on steatosis and inflammation scores of H&E stained liver slides. h) Hepatocellular hypertrophy score from 0–3 blindly assessed in H&E-stained liver slides. i) *Macrovesicular steatosis* score from 0–3 blindly assessed in H&E-stained liver slides. j) *Microvesicular steatosis* score from 0–3 blindly assessed in H&E-stained liver slides. k) Average number of inflammatory foci in 5 high power fields (HPF). l) *Histological fibrosis* assessed by Ishak scoring of Sirius Red-stained liver slides from Study 1. m) *Relative* gene expression of collagen type I alpha I chain (Col1a1) in liver tissue by qPCR. n) H&E and Sirius Red stained liver slides from Study 1 selected as representative by having the median value of the group in the histological steatosis score. b-f, h-k) Point shape represent individual mice in three individual studies where circles indicate Study 1, triangles Study 2, squares Study 3. b-f, k-m) Dashed line indicates mean of LFD group with interquartile range shown in gray. b-e, m) Bars indicate group mean ± SEM. Asterisks indicate fdr-corrected q-values <0.05 using linear mixed effects models comparing indicated groups. f, h-l) Bars indicate group median with interquartile range. Asterisks indicate p-values <0.05 by Kruskal–Wallis test with multiple comparisons between vehicle-treated groups or LAB group to vehicle-treated group fed the same diet with Dunn’s post-hoc test. b-e) LFD+Vehicle $$n = 21$$, HFD+Vehicle $$n = 20$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 20$$, FFMD+Vehicle $$n = 18$$, FFMD+L.Reuteri $$n = 21$$, FFMD+L. paracasei $$n = 20$.$ f-k) LFD+Vehicle $$n = 21$$, HFD+Vehicle $$n = 20$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 20$$, FFMD+Vehicle $$n = 20$$, FFMD+L.Reuteri $$n = 21$$, FFMD+L. paracasei $$n = 20$.$ l-m) LFD+Vehicle $$n = 8$$, HFD+Vehicle $$n = 9$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 9$$, FFMD+Vehicle $$n = 9$$, FFMD+L.Reuteri $$n = 9$$, FFMD+L. paracasei $$n = 8$.$ L. paracasei prevented diet-induced hepatic lipid accumulation in 17 out of 21 HFD-fed mice, presenting TG levels similar to those of LFD-fed reference mice (Figure 2c). A similar trajectory was noted for the hepatic lipid profile, where a subset of the HFD + L. paracasei group clustered together with their LFD-fed counterparts (Fig S2B). Additionally, the plasma AST/ALT ratio approached a normalization in HFD-fed mice receiving L. paracasei compared to the vehicle control (Figure 2e, $$p \leq 0.04$$, $q = 0.09$). The hepatic lipid clustering of the HFD + L. paracasei group fully recapitulated the histological steatosis score and NAFLD distribution (Figure 2f–g, S2B). Thus, HFD + L. paracasei mice clustering together with LFD-fed reference mice had a blinded steatosis score of 0-1, whereas those clustering together with HFD + vehicle control mice obtained a steatosis score ≥4, hence corroborating that NAFLD severity and not dietary feeding dictated hepatic lipid profile (Figure 2f–g, S2B). Both L. reuteri and L. paracasei, on the other hand, partly prevented the massively increased liver tissue mass and TG concentration in FFMD-fed mice (Figure 2b,c). When assessing the histological steatosis score, we observed that L. paracasei reduced steatosis development exclusively in HFD-fed mice, while only L. reuteri was able to significantly reduce this phenotype in FFMD-fed mice (Figure 2f). The selective NAFLD-reducing effect of L. paracasei in HFD-fed animals was reflected by a strong trend of diminished hepatocellular hypertrophy ($$p \leq 0.06$$) and macrovesicular steatosis (Figure 2h–n) hence mirroring the observed hepatic TG concentrations (Figure 2c). The selective amelioration of NAFLD by L. reuteri in FFMD-fed mice was mainly associated with reduced hepatocellular hypertrophy and a partial improvement of microvesicular steatosis (Figure 2h,j). L. paracasei reduced inflammatory foci counts in HFD-fed mice supported by a similar tendency in hepatic cytokine levels (Figure 2k, S2D-E, PERMANOVA $$p \leq 0.089$$), pinpointing the beneficial NAFLD ameliorating capabilities of L. paracasei in HFD-, but not FFMD-fed mice. ## L. reuteri diet-dependently altered intestinal bile acid levels HFD and FFMD feeding distinctively impacted the plasma lipid profile (Figure 3a), where a majority of the altered lipids were regulated similarly to that observed in the liver (Fig S2A). Circulating levels of total cholesterol (Figure 3b) and TG (Fig S3A, $$p \leq 0.01$$, $q = 0.06$) were increased by FFMD compared to HFD while total free fatty acids (FA) were not significantly altered by the background diet (Fig S3B). A subset of the HFD-fed mice receiving L. paracasei clustered with the LFD reference group (Figure 3a), as seen with hepatic lipids (Fig S2B). L. reuteri significantly increased the relative abundance of FA 18:2 exclusively in FFMD-fed mice (Figure 3c). Figure 3.L. paracasei and L.Reuteri diet-dependently affect liver lipids and cecum bile acids (BAs). a) Heatmap of the 40 most discriminating plasma lipids from end of study 1 assesed by lipidomics. Experimental groups are indicated by color and samples are distributed by hierical clustering. b) Total cholesterol levels in plasma. Point shape represent individual mice in three individual studies where circles indicate Study 1, triangles Study 2, squares Study 3. LFD+Vehicle $$n = 21$$, HFD+Vehicle $$n = 19$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 20$$, FFMD+Vehicle $$n = 18$$, FFMD+L.Reuteri $$n = 21$$, FFMD+L. paracasei $$n = 20$.$ c) *Relative plasma* levels of free fatty acid (FA) 18:2 measured by lipidomics. d) Total measured BA concentration in cecum content. e) Concentration of unconjugated BA in cecum content as fold change to the LFD group. f) Concentration of conjugated cholic and chenodeoxycholic acids (CA and CDCA) in cecum content as fold change to the LFD group. g) Concentration of conjugated muricholic acids (MCA) in cecum content as fold change to the LFD group. h) Relative Fibroblast growth factor 15 (Fgf15) gene expression in ileum tissue by RT-qPCR. i) Ratio of measured conjugated to unconjugated BAs in cecum content. b-l) Bars indicate group mean ± SEM. Points represent individual data points. Asterisks indicate fdr-corrected q-values <0.05 using linear mixed effects models comparing indicated groups. b-d, h-i) Dashed line indicate mean of LFD group with interquartile range shown in gray. c-l) LFD+Vehicle $$n = 9$$ (except C where $$n = 8$$), HFD+Vehicle $$n = 9$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 9$$, FFMD+Vehicle $$n = 9$$ (except E where $$n = 8$$), FFMD+L.Reuteri $$n = 9$$ (except C where $$n = 8$$), FFMD+L. paracasei $$n = 8$.$ Due to the known influence of bile acids (BAs) on the gut-liver axis by the enterohepatic pathway we quantified specific BAs in liver tissue. L. paracasei tended to increase total liver BA levels in HFD-fed mice ($$p \leq 0.02$$, $q = 0.11$) and the BA precursor THCA (Fig S3C-D) associated with borderline higher levels of an unconjugated and and taurine-conjugated BAs (Fig S3E-F, $p \leq 0.05$, q < 0.1). Over $90\%$ of BAs excreted into the intestine are recirculated in the small intestine44 prompting us to quantify the remaining BAs in the cecum, which did not differ in weight between the groups (Fig S3G). FFMD increased the levels of unconjugated BAs in this intestinal segment compared to HFD as well as secondary BAs without affecting the total cecal BA concentration (Figure 3d,e, S3H). These increases were associated with higher expression of fibroblast growth factor (Fgf)15 in ileum tissue (Figure 3g), which is upregulated by BAs through intestinal Fxr44. The indication of enhanced activation of the Fxr-Fgf15 enterohepatic pathway did, however, not result in repression of Cyp7a1 expression (Fig S3I), the rate-limiting enzyme of BA synthesis in the liver44 thus suggesting a dysregulation of the pathway by FFMD feeding. The probiotics did not significantly alter Fgf15 or host-defense protein RegIIIγ gene expression in ileum tissue on the respective diets (Figure 3H, S3J). Despite the tendency of augmented hepatic BA levels in L. paracasei treated HFD-fed mice (Fig S3C-F), we did not observe changes in cecal BA levels in this group (Figure 3e–g). Conversely, L. reuteri increased cecal (Figure 3D) but not hepatic (Fig S3C) BA levels exclusively in FFMD-fed mice. Notably, only absolute abundance of conjugated BAs were affected by L. reuteri (Figure 3e–i), suggesting a reduced re-uptake of the conjugated BA as proposed for other L. reuteri strains45,46 although this was not mirrored by altered gene expression of BA transporter Asbt in ileum tissue (Fig S3K). Investigation of the strains’ genomic makeup revealed that L. reuteri, but not L. paracasei, habored the bile salt hydrolase (Bsh) gene (PFAM PF02275), potentially driving the changes in intestinal BA content in a diet-dependent manner. ## Predicted functions of the gut microbiota correlated with disease severity, not macronutrient composition, and were modulated by probiotics We collected fecal samples at weeks 0, 4, 8, and 12 to assess longitudinal changes in gut microbiota composition as a function of intervention and diet. At termination, we further collected the content of small intestine and colon. Microbiota composition was markedly affected by dietary macronutrient composition (FFMD and LFD clustering separately from HFD), in colonic and fecal samples, suggesting that dietary nutrients, such as proteins, carbohydrate or lipid sources, rather than lipid proportion or the occurrence of obesity determined gut microbiota composition (Table 1, Figure 4a–c S4A-C). This was despite equal baseline microbiota composition and with a transient increase of fecal alpha diversity measures from FFMD and LFD (Fig S4A, E-H, Table 2). Figure 4.*Gut microbiota* composition and predicted function are affected by L. reuteri in combination with FFMD. a) Principal coordinate analysis (PCoA) of microbiota composition using weighted UniFrac distances of small intestine content sampled at the end of Study 1. Centroids indicate group average illustrated with $50\%$ CI and smaller points indicate individual data points. b) as a for colon content. c) as a for fecal samples collected at week 12 of the study. d) Relative abundance in % of *Proteobacteria phylum* in small intestine-, colon content, and fecal samples at the end of Study 1. Points indicate individual data points. Asterisks indicate p-values <0.05 comparing the indicated groups. e) Distribution of amplicon sequence variants (ASVs) as relative abundances identified in Limosilactobacillus reuteri DSM 32,910 (L. reuteri) or Lacticaseibacillus paracasei subsp. paracasei DSM 32,851 (L. paracasei) stocks. f) Relative abundance in % of ASV15, ASV5 and ASV31, respectively, from the administered Lactobacillus strains in small intestine-, colon content, and fecal samples collected 2-5 h (small intestine and colon) or 22-24 h after latest oral gavage at the end of the study. Points indicate individual data points and bars group mean with interquartile range. g) PCoA of predicted functions of the microbiota by KEGG orthology using Bray-Curtis distances of small intestine content sampled at the end of Study 1. Centroids indicate group average illustrated with $50\%$ CI and smaller points indicate individual data points. h) as g for colon content. i) as g for fecal samples collected at week 12 of the study.j) Relative changes in predicted 2-methylcitrate cycle I BioCyc ID PWY0–42 in fecal samples from the end of Study 1. k) as j for fatty acid salvage BioCyc ID PWY-7094. l) as j for glycosis pathway BioCyc ID GLYCOLYSIS-TCA-GLYOX-BYPASS. a-d, f-i) LFD+Vehicle Small int. and fecal $$n = 6$$ colon $$n = 9$$, HFD+Vehicle Small int. and fecal $$n = 9$$ colon $$n = 8$$, HFD+L.Reuteri $$n = 9$$, HFD+L.Paracasei $$n = 9$$, FFMD+Vehicle $$n = 9$$, FFMD+L.Reuteri $$n = 9$$, FFMD+L. paracasei $$n = 8$.$ Table 2.PERMANOVA p-values. Microbiota compositionPredicted functions Small intestineColonFecesSmall intestineColonFecesHFD+Vehicle vs FFMD+Vehicle0.008**0.001***0.004**0.006**0.017*0.009**HFD+Vehicle vs HFD+L. reuteri0.200.820.300.470.840.26HFD+Vehicle vs HFD+L. paracasei0.420.620.690.360.570.17FFMD+Vehicle vs FFMD+L. reuteri0.140.080.100.04*0.01*0.04*FFMD+Vehicle vs FFMD+L. paracasei0.640.820.880.550.440.99 The tested probiotics did not significantly alter the microbiota composition in small intestine or colon content sampled 2–5 h after the latest oral gavage (Figure 4a,b, Table 2). The fecal microbiota, sampled 22−24 h after the latest gavage, was however transiently modified by L. paracasei in HFD-fed mice (Fig S4A, D). L. reuteri temporarily changed FFMD-induced fecal microbiota composition (Fig S4B, D), which tended to be modified by the end of the study (Figure 4a–c, Table 2) and was significantly modified in a replicated study (Fig S4F-H). Investigation of differential abundances at all phylogenetic levels robustly modified by diet in the different studies revealed an increase in abundance of the phylum Proteobacteria by FFMD in colonic and fecal samples compared to LFD ($p \leq 0.001$, Figure 4e, S4I-J). The FFMD-induced Proteobacteria abundance in the distal intestinal segments was numerically reduced by L. reuteri and to a lesser degree L. paracasei in reproduced studies (Figure 4e, S4I-J). To investigate whether colonization of the administered LAB strains in the intestinal tract was affected by diet we assessed the ASVs originating from the administered L. reuteri (ASV15) and L. paracasei (ASV5 and ASV31, Figure 4f). As expected, these ASVs were found in the small intestine of mice 2–5 hours after receiving the corresponding strain (Figure 4g). No change from background signal was detected in colon content 2–5 hours, or feces 22–24 hours after oral gavage (Figure 4g, Fig S4N), indicating negligible colonic colonization by the applied probiotics. We next assessed the predicted functions of the gut microbiota by assessing KEGG orthology. Interestingly, while colon and fecal microbiota composition reflected ingested nutrients (Figure 4a–c, the predicted microbiota functionality more closely resembled the pathology of experimental mice (Figure 4g–i, Table 3). As such, LFD-fed mice clustered in close vicinity to HFD-fed mice, despite being fed a diet matched to the FFMD (Table 1, Figure 4g–i). This observation was further supported by L. paracasei treated HFD-fed mice who exhibited an intermediary disease phenotype and likewise clustered between the LFD-fed mice and their vehicle-treated HFD-fed counterparts (Figure 4g–i). Despite the lack of clear and reproducible effects on gut microbiota composition from the gavage by the LAB strains there was a significant change in FFMD-induced predicted functions of the microbiota by L. reuteri in all assessed intestinal segments at termination (Figure 4g–i, Table 2), again mirroring their partial protection against FFMD-induced liver pathology. The specific predicted pathways of fecal samples affected by L. reuteri included a reduction in the methylcitrate cycle (Table 3, Figure 4l) transforming propionic acid to pyruvate and succinate. Functionally, we did not detect differences in cecal propionic acid or other short-chain fatty acids (SCFAs) from the administered strains (Fig S4O-P). Interestingly, we observed additional changes in predicted functional pathways in fecal samples by L. reuteri including a normalization in fatty acid salvage and glycolysis pathways, which were most pronounced in FFMD-fed mice (Figure 4k,l, Table 3). Table 3.Differences in predicted pathways of fecal microbiota week 12 in Study 1 (Top 20). PathwayDescriptionlog2FoldChangepvaluepadjustHFD+Vehicle vs FFMD+Vehicle1LEU-DEG2-PWYL-leucine degradation I−2,834.52235874873049e-151.39288649460899e-122PWY0–1479tRNA processing−2,841.12750188714142e-131.73635290619778e-113HEMESYN2-PWYHeme biosynthesis II (anaerobic)−1,944.69645288826021e-134.82169163194715e-114PWY0–1261Anhydromuropeptides recycling−2,316.85443809664332e-135.27791733441536e-115P105-PWYTCA cycle IV (2-oxoglutarate decarboxylase)−2,546.65923813052144e-114.10209068840121e-96PWY-5188Tetrapyrrole biosynthesis I (from glutamate)0,922.34558165434382e-101.03205592791128e-87PWY-5189Tetrapyrrole biosynthesis II (from glycine)0,932.20030620768975e-101.03205592791128e-88GLYCOLYSIS-TCA-GLYOX-BYPASSSuperpathway of glycolysis, pyruvate dehydrogenase, TCA, and glyoxylate bypass−2,494.63352300454785e-101.78390635675092e-89TCA-GLYOX-BYPASSSuperpathway of glyoxylate bypass and TCA−2,526.03798864270785e-102.06633389106002e-810PWY0–422-methylcitrate cycle I−2,471.17409073188659e-93.61619945421069e-811PWY-57472-methylcitrate cycle II−2,432.3556397125142e-96.59579119503976e-812PWY-7094Fatty acid salvage−2,268.0741308642353e-92.07236025515373e-713PWY-6876Isopropanol biosynthesis6,828.99444248769771e-92.130990989393e-714PWY-7234Inosine-5’−phosphate biosynthesis III−1,852.73971483486545e-86.02737263670398e-715PWY-7328Superpathway of UDP-glucose-derived O-antigen building blocks biosynthesis1,093.92276440853005e-88.05474291884836e-716PWY-7315dTDP-N-acetylthomosamine biosynthesis1,555.19495687424118e-81E–0617PWY-621Sucrose degradation III (sucrose invertase)0,891.22052004182342e-72,21E–0618FASYN-INITIAL-PWYSuperpathway of fatty acid biosynthesis initiation (E. coli)−2,441.92247421873961e-72,75E–0619PWY-5971Palmitate biosynthesis II (bacteria and plants)−2,331.91446278376084e-72,75E–0620PWY-5989Stearate biosynthesis II (bacteria and plants)−2,432.02205214857467e-72,75E–06FFMD+Vehicle vs FFMD+L.reuteri1LEU-DEG2-PWYL-leucine degradation I9,8372475.84578403775772e-161.80050148362938e-132PWY-57472-methylcitrate cycle II8,499313.0266289339534e-133.58192550679826e-113PWY-70942-methylcitrate cycle I3,3380534.65185130753021e-133.58192550679826e-114PWY0–422-methylcitrate cycle I8,4732364.0885414191369e-133.58192550679826e-115ALL-CHORISMATE-PWYSuperpathway of chorismate metabolism21,165911.72357006274882e-71,06E–056DENITRIFICATION-PWYNitrate reduction I (denitrification)−6,926752.26429166798132e-71,16E–057PWY-922Mevalonate pathway I3,1559061,61E–067,08E–058PWY-5910Superpathway of geranylgeranyldiphosphate biosynthesis I (via mevalonate)3,0967264,25E–050,0016359ASPASN-PWYSuperpathway of L-aspartate and L-asparagine biosynthesis−0,291710,0002090,00715310P105-PWYTCA cycle IV (2-oxoglutarate decarboxylase)1,5206260,0002740,00843311P125-PWYSuperpathway of (R,R)-butanediol biosynthesis1,4773180,0004580,01283712ANAGLYCOLYSIS-PWYGlycolysis III (from glucose)−0,242730,0017310,02383213CALVIN-PWYCalvin-Benson-Bassham cycle−0,32460,0029040,02383214COA-PWYCoenzyme A biosynthesis I−0,185620,0033470,02383215GLCMANNANAUT-PWYSuperpathway of N-acetylglucosamine, N-acetylmannosamine and N-acetylneuraminate degradation−0,520690,0030930,02383216GLYCOLYSIS-TCA-GLYOX-BYPASSSuperpathway of glycolysis, pyruvate dehydrogenase, TCA, and glyoxylate bypass1,3741910,0013270,02383217GLYOXYLATE-BYPASSGlyoxylate cycle1,1794040,0031390,02383218HSERMETANA-PWYL-methionine biosynthesis III1,3312390,0030380,02383219NONMEVIPP-PWYMethylerythritol phosphate pathway I−0,275090,0033780,02383220NONOXIPENT-PWYPentose phosphate pathway (non-oxidative branch)−0,296010,0026450,023832 ## Discussion We used mice fed two distinct but energy-matched high fat and high sucrose diets to test the preventive effects of two LAB strains on obesity and related metabolic disturbances. While both HFD and FFMD-induced obesity and compromised glucoregulatory capacity in 12 weeks, FFMD led to enhanced ectopic fat distribution to visceral fat and liver. This induced major phenotypic differences between the HFD and FFMD, notably on body weight gain, energy harvesting, glucose homeostasis, enterohepatic circulation of BAs as well as gut microbiota composition and predicted functionality. Accordingly, the greatest diet-specific impact of the customized FFMD was observed on NAFLD features including hepatocellular morphology, liver lipid, as well as inflammatory profiles and fibrosis. The FFMD-induced liver pathology culminated in significant fibrotic remodeling in >$75\%$ of the mice, hence contrasting the results in HFD-fed mice where no fibrotic bridging was observed and only 1 of 9 mice exhibited fibrosis expansion of portal areas. These diet-specific effects can all potentially affect the ability of probiotics to alleviate obesity-related dysmetabolism and related diseases such as NAFLD. This has major physiological significance and clinical relevance given that the choice of experimental diets in preclinical investigation, and the dietary habits of subjects enrolled in clinical trials, are thus likely to influence the efficacy of probiotics. This could lead to the dismissal of promising probiotic candidates for mitigating obesity and related metabolic disorders due to false-negative results in preclinical trials, a lack of clinical translatability when advancing to clinical phases where dietary macronutrient composition is not specifically controlled for, or – worst case – lead to the endorsement of probiotics with limited or potentially even unfavorable effects in combination with certain diets not adequately recapitulated in preceeding studies20,23. In our study, L. paracasei reduced obesity development associated with diminished NAFLD exclusively in the context of HFD-feeding without changing energy balance. As a result, $45\%$ of the HFD-mice receiving L. paracasei had no histopathological signs of NAFLD, which was not prevented in any other HFD-fed mouse in the study. L. reuteri on the other hand improved glucose homeostasis and the more severe NAFLD phenotype only in FFMD-fed mice. This was associated with increased plasma FA 18:2 levels, reported to inversely correlate with human hepatocellular carcinoma (HCC)47 and a murine NASH-HCC induction 48. L. reuteri additionally modified the FFMD-induced BA profile by augmenting the amount of conjugated BAs in cecum. Elevated BA levels could point toward increased energy excretion although such indications were not captured by the bomb calometry measures employed in the current study. Augmented secondary, i.e. dehydroxylated, BAs have also recently been shown to enhance the presence of peripherally induced regulatory T cells46,49,50, hence dampening mucosal immunity; a trait that would be beneficial in the context of diet-induced obesity51. Future studies in either microbiota depleted or germ-free mice are thus warranted to elucidate if and how L. reuteri induced BA modifications may affect host immunity. Although it remains uncertain what caused the diet-dependent effects on probiotic efficacy, possible explanations could relate to direct dietary influence where the administered strains rely on utilization of specific macronutrients. This is exemplified by FFMD-feeding leading to major increases in predicted metabolic functions of the gut microbiota; a traits that was normalized by L. reuteri in FFMD-fed mice. It should be noted, though, that while the metabolic benefits of L. paracasei associated with reduced weight gain and fat accretion in HFD-fed mice, the improved insulin sensitivity of FFMD-fed animals co-treated with L. reuteri was not linked to diminished weight gain, strongly pointing toward dissimilar mechanisms in probiotic efficacy. Importantly, the diet-dependent efficacy of probiotic candidates were highly reproducible across studies and persisted after rigorous statistical testing, cf. our strict approaches adjusting for individual studies, study-batches, potential co-caging and random effects, as well as correcting for potential false discovery rates as outlined in the method section. We also observed that the diet composition, rather than the amount of dietary fat and high glycemic carbohydrates, was the main driver of gut microbial community structures. This was examplified by a reproducible FFMD-induced augmentation of *Proteobacteria phylum* abundances, which has previously been linked to epithelial damage via disruption in anaerobiosis52 and NASH53. Interestingly, we and others previously observed Proteobacteria as the dominant bacterial phylum in extra-intestinal tissues of morbidly obese individuals with T2D compared to normoglycemic weight-matched counterparts54,55. Predicted functions of the microbiota, however, followed disease phenotype rather than diet composition. As such, the tested probiotic strains had negligible impact on microbiota composition, but not its predicted functions, in the sampled intestinal segments and showed minimal signs of colonization of the gastro-intestinal tract, supporting previous reports56. In the study from Zmora et al, individual responses to probiotics were suggested to originate from genetic or baseline microbiota differences naturally found in humans24. Both factors were accounted for in our studies, where we show dietary differences alone were sufficient in determining probiotic efficacy. The diet-dependent trait of the strains is warranted to interogate in future studies, just as a combination therapy may alleviate diets dependency. To this end, a combination of the L. reuteri and L. paracasei strains is currently investigated in an ongoing randomized, double-blinded, placebo-controlled clinical trial in prediabetic subjects (ClinicalTrials.gov identifier NCT04767789). Differences in probiotic efficacy have been reported for L. helveticus R0052, L. plantarum WCFS1, and P. copri26–29 when comparing lean, chow-fed mice to obese, Western diet-fed mice. Our study demonstrates that strain-specific efficacies of probiotics on body composition, insulin resistance, and NAFLD were reproducible, yet highly influenced by the type of obesogenic diet, and thus substantially expanding the understanding of dietary influence on probiotic efficacy. While future studies are warranted to elucidate the generic implication of our findings, the dietary interactions described here may likely extend beyond the tested probiotic strains. Potentially, this could aid in explaining empiric observations of responders versus non-responders to probiotics in general and might indicate a need for pre-treatment stratification of patients. ## Limitations The study was limited by the use of 16S rRNA gene amplicons for evaluation of the gut microbiota with lower resolution of microbial species compared to shotgun sequencing and relying on assumptions to predict microbiota functionality. While the macronutrients of the diets are relatively similar between the HFD and FFMD, the nutrient sources are vastly different. Future lines of research should therefore elucidate the precise dietary components driving the reported phenotypic discrepanices. ## Disclosure statement This study was partly funded by Novozymes A/S (NZ) which provided the bacterial strains and was granted a patent related to the findings (WO20084051). NZ contributed to study design and parts of the analyses but had no role in data integration, interpretation, and conclusion. ## Author Contributions BAHJ, ISL, NNK, AM and BSYC conceived and designed the studies. ISL, BAHJ, and BSYC carried out the in vivo studies. ISL, BSYC, BF, AO, RFH, PBO, and DS generated data. BAHJ designed the diets and supervised all parts of the study; AM and CS supervised parts of the study. ISL integrated the data and wrote the first drafts of the manuscript, which was revised by all the authors, who agreed upon the submitted manuscript. ## Data availability statement The data that support the findings of this study are available in Mendeley Data at doi: http://doi.org/$\frac{10.17632}{72}$j8tkm36g.1 ## Supplementary material Supplemental data for this article can be accessed online at https://doi.org/$\frac{10.1080}{19490976.2023.2192547}$ ## References 1. **Obesity and overweight fact sheet**. (2016) 2. Kivimäki M, Kuosma E, Ferrie JE, Luukkonen R, Nyberg ST, Alfredsson L, Batty GD, Brunner EJ, Fransson E, Goldberg M. **Overweight, obesity, and risk of cardiometabolic multimorbidity: pooled analysis of individual-level data for 120 813 adults from 16 cohort studies from the USA and Europe**. *Lancet Public Heal* (2017) **2** e277-20. DOI: 10.1016/S2468-2667(17)30074-9 3. Fonseca VA.. **Defining and characterizing the progression of type 2 diabetes**. *Diabetes Care* (2009) **32** S151-156. DOI: 10.2337/dc09-S301 4. Chatterjee S, Khunti K, Davies MJ. **Type 2 diabetes**. *Lancet* (2017) **389** 2239-2251. DOI: 10.1016/S0140-6736(17)30058-2 5. Fabbrini E, Sullivan S, Klein S. **Obesity and nonalcoholic fatty liver disease: biochemical, metabolic, and clinical implications**. *Hepatology* (2010) **51** 679-689. DOI: 10.1002/hep.23280 6. Eslam M, Sanyal AJ, George J, Sanyal A, Neuschwander-Tetri B, Tiribelli C, Kleiner DE, Brunt E, Bugianesi E, Yki-Järvinen H. **MAFLD: a consensus-driven proposed nomenclature for metabolic associated fatty liver disease**. *Gastroenterology* (2020) **158** 1999-2014.e1. DOI: 10.1053/j.gastro.2019.11.312 7. Younossi ZM, Koenig AB, Abdelatif D, Fazel Y, Henry L, Wymer M. **Global epidemiology of nonalcoholic fatty liver disease—meta-analytic assessment of prevalence, incidence, and outcomes**. *Hepatology* (2016) **64** 73-84. DOI: 10.1002/hep.28431 8. Hall KD, Guo J. **Obesity energetics: body weight regulation and the effects of diet composition**. *Gastroenterology* (2017) **152** 1718-1727.e3. DOI: 10.1053/j.gastro.2017.01.052 9. Stunkard AJ, Harris JR, Pedersen NL, McClearn GE. **The body-mass index of twins who have been reared apart**. *N Engl J Med* (1990) **322** 1483-1487. DOI: 10.1056/NEJM199005243222102 10. Tyrrell J. **Height, body mass index, and socioeconomic status: mendelian randomisation study in UK Biobank**. *BMJ* (2016) **352** i582. DOI: 10.1136/bmj.i582 11. Venditti EM, Bray GA, Carrion-Petersen ML, Delahanty LM, Edelstein SL, Hamman RF, Hoskin MA, Knowler WC, Ma Y. **First versus repeat treatment with a lifestyle intervention program: attendance and weight loss outcomes**. *Int J Obes* (2008) **32** 1537-1544. DOI: 10.1038/ijo.2008.134 12. Varkevisser RDM, van Stralen MM, Kroeze W, Ket JCF, Steenhuis IHM. **Determinants of weight loss maintenance: a systematic review**. *Obes Rev* (2019) **20** 171-211. DOI: 10.1111/obr.12772 13. Thonusin C, Shinlapawittayatorn K, Chattipakorn SC, Chattipakorn N. **The impact of genetic polymorphisms on weight regain after successful weight loss**. *Br J Nutr* (2020) **124** 809-823. DOI: 10.1017/S0007114520001968 14. Dulloo AG, Montani J-P. **Pathways from dieting to weight regain, to obesity and to the metabolic syndrome: an overview**. *Obes Rev* (2015) **16** 1-6. DOI: 10.1111/obr.12250 15. Khalil H, Ellwood L, Lord H, Fernandez R. **Pharmacological treatment for obesity in adults: an umbrella review**. *Ann Pharmacother* (2020) **54** 691-705. DOI: 10.1177/1060028019898912 16. Arterburn DE, Telem DA, Kushner RF, Courcoulas AP. **Benefits and risks of bariatric surgery in adults**. *JAMA* (2020) **324** 879. DOI: 10.1001/jama.2020.12567 17. **Probiotics in food - Health and nutritional properties and guidelines for evaluation. Report of a Joint FAO/WHO expert consultation on evaluation of health and nutritional properties of probiotics in food including powder milk with live lactic acid bacter**. (2001) 18. Delzenne NM, Neyrinck AM, Bäckhed F, Cani PD. **Targeting gut microbiota in obesity: effects of prebiotics and probiotics**. *Nat Rev Endocrinol* (2011) **7** 639-646. DOI: 10.1038/nrendo.2011.126 19. Wiciński M, Gębalski J, Gołębiewski J, Malinowski B. **Probiotics for the treatment of overweight and obesity in humans—a review of clinical trials**. *Microorganisms* (2020) **8** 1148. DOI: 10.3390/microorganisms8081148 20. **Probiotics: elixir or empty promise?**. *Lancet Gastroenterol Hepatol* (2019) **4** 81. DOI: 10.1016/S2468-1253(18)30415-1 21. Hill C, Guarner F, Reid G, Gibson GR, Merenstein DJ, Pot B, Morelli L, Canani RB, Flint HJ, Salminen S. **Expert consensus document: the international scientific association for probiotics and prebiotics consensus statement on the scope and appropriate use of the term probiotic**. *Nat Rev Gastroenterol Hepatol* (2014) **11** 506-514. DOI: 10.1038/nrgastro.2014.66 22. Krautkramer KA, Fan J, Bäckhed F. **Gut microbial metabolites as multi-kingdom intermediates**. *Nat Rev Microbiol* (2020) 77-94. DOI: 10.1038/s41579-020-0438-4 23. Spencer CN, McQuade JL, Gopalakrishnan V, McCulloch JA, Vetizou M, Cogdill AP, Khan MAW, Zhang X, White MG, Peterson CB. **Dietary fiber and probiotics influence the gut microbiome and melanoma immunotherapy response**. *Science* (2021) **374** 1632-1640. DOI: 10.1126/science.aaz7015 24. Zmora N, Zilberman-Schapira G, Suez J, Mor U, Dori-Bachash M, Bashiardes S, Kotler E, Zur M, Regev-Lehavi D, Brik RBZ. **Personalized gut mucosal colonization resistance to empiric probiotics is associated with unique host and microbiome features**. *Cell* (2018) **174** 1388-1405.e21. DOI: 10.1016/j.cell.2018.08.041 25. Reid G, Gaudier E, Guarner F, Huffnagle GB, Macklaim JM, Munoz AM, Martini M, Ringel-Kulka T, Sartor BR, Unal RR. **Responders and non-responders to probiotic interventions**. *Gut Microbes* (2010) **1** 200-204. DOI: 10.4161/gmic.1.3.12013 26. Ohland CL, Kish L, Bell H, Thiesen A, Hotte N, Pankiv E, Madsen KL. **Effects of Lactobacillus helveticus on murine behavior are dependent on diet and genotype and correlate with alterations in the gut microbiome**. *Psychoneuroendocrinology* (2013) **38** 1738-1747. DOI: 10.1016/j.psyneuen.2013.02.008 27. Tachon S, Lee B, Marco ML. **Diet alters probiotic Lactobacillus persistence and function in the intestine**. *Environ Microbiol* (2014) **16** 2915-2926. DOI: 10.1111/1462-2920.12297 28. Kovatcheva-Datchary P, Nilsson A, Akrami R, Lee Y, De Vadder F, Arora T, Hallen A, Martens E, Björck I, Bäckhed F. **Dietary fiber-induced improvement in glucose metabolism is associated with increased abundance of prevotella**. *Cell Metab* (2015) **22** 971-982. DOI: 10.1016/j.cmet.2015.10.001 29. Pedersen HK, Gudmundsdottir V, Nielsen HB, Hyotylainen T, Nielsen T, Jensen BAH, Forslund K, Hildebrand F, Prifti E, Falony G. **Human gut microbes impact host serum metabolome and insulin sensitivity**. *Nature* (2016) **535** 376-381. DOI: 10.1038/nature18646 30. Damgaard MTF, Pærregaard SI, Søgaard I, Agerholm M, Paulson JN, Treebak JT, Sina C, Holm JB, Kristiansen K, Jensen BAH. **Age-dependent alterations of glucose clearance and homeostasis are temporally separated and modulated by dietary fat**. *J Nutr Biochem* (2018) **54**. DOI: 10.1016/j.jnutbio.2017.09.026 31. Chong J, Wishart DS, Xia J. **Using MetaboAnalyst 4.0 for comprehensive and integrative metabolomics data analysis**. *Curr Protoc Bioinforma* (2019) **68** 1-128. DOI: 10.1002/cpbi.86 32. Liang W, Menke AL, Driessen A, Koek GH, Lindeman JH, Stoop R, Havekes LM, Kleemann R, van den Hoek AM. **Establishment of a general NAFLD scoring system for rodent models and comparison to human liver pathology**. *Plos One* (2014) **9** 1-17. DOI: 10.1371/journal.pone.0115922 33. Ishak K, Baptista A, Bianchi L, Callea F, De Groote J, Gudat F, Denk H, Desmet V, Korb G, MacSween RNM. **Histological grading and staging of chronic hepatitis**. *J Hepatol* (1995) **22** 696-699. DOI: 10.1016/0168-8278(95)80226-6 34. Goodman ZD. **Grading and staging systems for inflammation and fibrosis in chronic liver diseases**. *J Hepatol* (2007) **47** 598-607. DOI: 10.1016/j.jhep.2007.07.006 35. Klindworth A, Pruesse E, Schweer T, Peplies J, Quast C, Horn M, Glöckner FO. **Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies**. *Nucleic Acids Res* (2013) **41** 1-11. DOI: 10.1093/nar/gks808 36. Larsen IS, Jensen BAH, Bonazzi E, Choi BSY, Kristensen NN, Schmidt EGW, Süenderhauf A, Morin L, Olsen PB, Hansen LBS. **Fungal lysozyme leverages the gut microbiota to curb DSS-induced colitis**. *Gut Microbes* (2021) **13**. DOI: 10.1080/19490976.2021.1988836 37. Benjamini Y, Hochberg Y. **Controlling the false discovery rate: a practical and powerful approach to multiple testing**. *J R Stat Soc Ser B* (1995) **57** 289-300. DOI: 10.1111/j.2517-6161.1995.tb02031.x 38. Kuznetsova A, Brockhoff PB, Christensen RHB. **lmerTest package: tests in linear mixed effects models**. *J Stat Softw* (2017) **82** 1-26. DOI: 10.18637/jss.v082.i13 39. Oksanen J. **Vegan: community ecology package**. (2017). DOI: 10.4135/9781412971874.n145 40. Bates D, Mächler M, Bolker B, Walker S. **Fitting linear mixed-effects models using lme4**. *J Stat Softw* (2014) **67**. DOI: 10.18637/jss.v067.i01 41. Tandra S, Yeh MM, Brunt EM, Vuppalanchi R, Cummings OW, Ünalp-Arida A, Wilson LA, Chalasani N. **Presence and significance of microvesicular steatosis in nonalcoholic fatty liver disease**. *J Hepatol* (2011) **55** 654-659. DOI: 10.1016/j.jhep.2010.11.021 42. Chassaing B, Miles-Brown J, Pellizzon M, Ulman E, Ricci M, Zhang L, Patterson AD, Vijay-Kumar M, Gewirtz AT. **Lack of soluble fiber drives diet-induced adiposity in mice**. *Am J Physiol - Gastrointest Liver Physiol* (2015) **309** G528-541. DOI: 10.1152/ajpgi.00172.2015 43. Daniel N, Rossi Perazza L, Varin TV, Trottier J, Marcotte B, St-Pierre P, Barbier O, Chassaing B, Marette A. **Dietary fat and low fi ber in puri fi ed diets differently impact the gut-liver axis to promote obesity-linked metabolic impairments**. *Am J Physiol Gastrointest Liver Physiol* (2021) **320**. DOI: 10.1152/ajpgi.00028.2021 44. Inagaki T, Choi M, Moschetta A, Peng L, Cummins CL, McDonald JG, Luo G, Jones SA, Goodwin B, Richardson JA. **Fibroblast growth factor 15 functions as an enterohepatic signal to regulate bile acid homeostasis**. *Cell Metab* (2005) **2** 217-225. DOI: 10.1016/j.cmet.2005.09.001 45. Lu M, Sun J, Zhao Y, Zhang H, Li X, Zhou J, Dang H, Zhang J, Huang W, Qi C. **Prevention of high-fat diet-induced hypercholesterolemia by lactobacillus reuteri Fn041 through promoting cholesterol and bile salt excretion and intestinal mucosal barrier functions**. *Front Nutr* (2022) **9** 1-14. DOI: 10.3389/fnut.2022.851541 46. Zhang C, Fang R, Lu X, Zhang Y, Yang M, Su Y, Jiang Y, Man C. **Lactobacillus reuteri J1 prevents obesity by altering the gut microbiota and regulating bile acid metabolism in obese mice**. *Food Funct* (2022) **13** 6688-6701. DOI: 10.1039/D1FO04387K 47. Qiu J-F, Zhang K-L, Zhang X-J, Hu Y-J, Li P, Shang C-Z, Wan J-B. **Abnormalities in plasma phospholipid fatty acid profiles of patients with hepatocellular carcinoma**. *Lipids* (2015) **50** 977-985. DOI: 10.1007/s11745-015-4060-6 48. Vlock EM, Karanjit S, Talmon G, Farazi PA. **Reduction of polyunsaturated fatty acids with tumor progression in a lean non-alcoholic steatohepatitis- associated hepatocellular carcinoma mouse model**. *J Cancer* (2020) **11** 5536-5546. DOI: 10.7150/jca.48495 49. Paik D, Yao L, Zhang Y, Bae S, D’agostino GD, Zhang M, Kim E, Franzosa EA, Avila-Pacheco J, Bisanz JE. **Human gut bacteria produce ΤΗ17-modulating bile acid metabolites**. *Nature* (2022). DOI: 10.1038/s41586-022-04480-z 50. Campbell C, McKenney PT, Konstantinovsky D, Isaeva OI, Schizas M, Verter J, Mai C, Jin W-B, Guo C-J, Violante S. **Bacterial metabolism of bile acids promotes generation of peripheral regulatory T cells**. *Nature* (2020). DOI: 10.1038/s41586-020-2193-0 51. Jensen BAH, Holm JB, Larsen IS, von Burg N, Derer S, Sonne SB, Pærregaard SI, Damgaard MV, Indrelid SA, Rivollier A. **Lysates of Methylococcus capsulatus Bath induce a lean-like microbiota, intestinal FoxP3+RORγt+IL-17+ Tregs and improve metabolism**. *Nat Commun* (2021) **12**. DOI: 10.1038/s41467-021-21408-9 52. Litvak Y, Byndloss MX, Tsolis RM, Bäumler AJ. **Dysbiotic Proteobacteria expansion: a microbial signature of epithelial dysfunction**. *Curr Opin Microbiol* (2017) **39** 1-6. DOI: 10.1016/j.mib.2017.07.003 53. Zhu L, Baker SS, Gill C, Liu W, Alkhouri R, Baker RD, Gill SR. **Characterization of gut microbiomes in nonalcoholic steatohepatitis (NASH) patients: a connection between endogenous alcohol and NASH**. *Hepatology* (2013) **57** 601-609. DOI: 10.1002/hep.26093 54. Anhê FF, Jensen BAH, Varin TV, Servant F, Van Blerk S, Richard D, Marceau S, Surette M, Biertho L, Lelouvier B. **Type 2 diabetes influences bacterial tissue compartmentalisation in human obesity**. *Nat Metab* (2020). DOI: 10.1038/s42255-020-0178-9 55. Massier L, Chakaroun R, Tabei S, Crane A, Didt KD, Fallmann J, von Bergen M, Haange S-B, Heyne H, Stumvoll M. **Adipose tissue derived bacteria are associated with inflammation in obesity and type 2 diabetes**. *Gut* (2020) 1-11. DOI: 10.1136/gutjnl-2019-320118 56. Kristensen NB, Bryrup T, Allin KH, Nielsen T, Hansen TH, Pedersen O. **Alterations in fecal microbiota composition by probiotic supplementation in healthy adults: a systematic review of randomized controlled trials**. *Genome Med* (2016) **8**. DOI: 10.1186/s13073-016-0300-5
--- title: Clostridioides difficile aggravates dextran sulfate solution (DSS)-induced colitis by shaping the gut microbiota and promoting neutrophil recruitment authors: - Danfeng Dong - Tongxuan Su - Wei Chen - Daosheng Wang - YiLun Xue - Qiuya Lu - Cen Jiang - Qi Ni - Enqiang Mao - Yibing Peng journal: Gut Microbes year: 2023 pmcid: PMC10038061 doi: 10.1080/19490976.2023.2192478 license: CC BY 4.0 --- # Clostridioides difficile aggravates dextran sulfate solution (DSS)-induced colitis by shaping the gut microbiota and promoting neutrophil recruitment ## ABSTRACT Clostridioides difficile is a pathogen contributing to increased morbidity and mortality of patients with inflammatory bowel disease (IBD). To determine how C. difficile affects the severity of colitis, we constructed a dextran sulfate solution-induced colitis model challenged with C. difficile. Without antibiotic administration, C. difficile led to transient colonization in mice with colitis, but still significantly enhanced disease severity as assessed by weight loss, histopathological damages, and inflammatory cytokine concentrations. Because this effect is independent of toxin production as shown by infection with a non-toxigenic strain, we focused on changes in the gut microbiota. The microbiota altered by C.difficile, featured with reduced proportions of g_Prevotellaceae_UCG-001 and g_Muribaculaceae, were confirmed to contribute to disease severity in colitis mice via fecal microbiota transplantations. The inflamed colon showed neutrophil accumulation by flow cytometric analysis and myeloperoxidase immunochemical staining. There was enrichment of upregulated genes in leukocyte chemotaxis or migration as shown by RNA sequencing analysis. The isolated neutrophils from C. difficile-infected mice with colitis showed a robust migratory ability and had enhanced expression of cytokines and chemokines. We observed a detrimental role of neutrophils in the progress of disease by hindering neutrophil recruitment with the CXCR2 inhibitor SB225002. Furthermore, neutrophil recruitment appeared to be regulated by interleukin (IL)-1β, as inhibition of IL-1β production by MCC950 markedly ameliorated inflammation with decreased neutrophil accumulation and neutrophil-derived chemokine expression. In conclusion, our study provides information on the complicated interaction between microbiota and immune responses in C. difficile-induced inflammation in mice with colitis. Our findings could help determine potential therapeutic targets for patients with IBD concurrent with C. difficile infection. ## Introduction Clostridioides difficile is a gram-positive, spore-forming, anaerobic bacterium, which is the major cause of antibiotic-associated infections in health-care facilities. Perturbation of intact microbiota or the loss of indigenous microorganisms renders individuals susceptible to C. difficile colonization, causing diseases, such as diarrhea and pseudomembranous colitis, or even death.1–3 Although antibiotic treatment is the most common risk factor for C. difficile infection (CDI), other factors are also recognized, especially the preexistence of inflammatory bowel disease (IBD).4–6 IBDs, such as Crohn’s disease (CD) and ulcerative colitis (UC), are characterized by excessive intestinal inflammation, accompanied by aberrant immune responses and dysbiosis of the gut microbiota. Patients with IBD are approximately five times more likely to develop CDI.7 CDI results in longer hospitalizations, increased escalation and readmission rates, a higher colectomy risk, and increased mortality of patients with IBD.8 These findings highlight the importance of studying IBD with concurrent CDI. Regarding how IBD status affects subsequent infection by C.difficile, previous studies reported that intestinal inflammation and distinct intestinal microbiota may lead to susceptibility to CDI in the Helicobacter hepaticus-induced colitis mouse model.9 Furthermore, increased Th17 cells and Th17-related cytokines induced by colitis are responsible for the subsequent severity of CDI.10 However, the causes of the increased severity of IBD by C. difficile remain poorly understood. Generally, the progression of IBD is a complicated process with a combination of genetic and environmental factors, and host – microbe interactions.11 Dysbiosis of the gut microbiota, defined as decreased microbial diversity and an imbalance between potentially protective and pathogenic microorganisms, may play a major role in the pathogenesis of IBD.12,13 A few clinical studies have reported the effect of C. difficile on the microbiota of patients with IBD and showed pronounced intestinal dysbiosis in patients with concomitant IBD and CDI.14 However, whether this dysbiosis is truly a cause or a consequence remains unestablished. In fragile gut environment with a disturbed microbial community, immune cells have prominent roles in maintaining intestinal hemostasis via releasing cytokines and chemokines. This in turn impairs intestinal barrier integrity and perpetuates gut inflammation.11 Neutrophils, as a first responder, are implicated in the development of colitis.15 Neutrophils communicate with gut microbes by sensing microbiota-derived components through toll-like receptors and inflammasome signaling pathways or responding to metabolites, such as short-chain fatty acids, through histone deacetylases and G protein-coupled receptors.15–18 After neutrophils are recruited to an inflamed colon, they defend pathogens by releasing antimicrobial peptides and reactive oxygen species, forming neutrophil extracellular traps and producing inflammatory cytokines and chemokines.19 Studies have shown that mice depleted of neutrophils exhibit aggravated intestinal inflammation,20 while excessive accumulation of neutrophils leads to the persistence of inflammatory responses and epithelial damage.21 These studies suggest a controversial role of neutrophils in the pathogenesis of colitis. This study aimed to investigate how C. difficile affects the severity of dextran sulfate solution (DSS)-induced colitis in mice. We found that transient colonization of C. difficile in mice with colitis changed the gut microbiome community and increased neutrophil infiltration by upregulating multiple migration genes. The altered microbiota due to C. difficile was responsible for the disease severity and promoted neutrophils to express higher levels of proinflammatory cytokines and chemokines. The robust neutrophil infiltration was probably regulated by increased interleukin (IL)-1β levels because hindering the production of IL-1β by the inhibitor MCC950 significantly alleviated colitis induced by C. difficile in DSS mice. Our findings may help to better understand how intestinal inflammatory responses are driven by C. difficile in colitis and help provide evidence for potential therapeutic targets in patients with IBD and a risk of CDI. ## Exposure to C. difficile increases the disease severity of DSS-induced colitis First, we examined the effect of C. difficile on gut inflammation in mice with experimental colitis. We constructed a 7-day mouse model of DSS-induced colitis, followed by oral gavage of C. difficile on day 7 (Figure 1a). DSS mice challenged with the pathogen (DSSCD group) showed more severe colitis, presenting more weight loss and higher disease activity scores (Figure 1b,c). In line with the phenotype, a histological examination of colonic tissues from the DSSCD group showed increased histological scores as shown by extensive inflammation, marked epithelial disruption, and severe crypt collapse (Figure 1d). A damaged intestinal epithelial barrier in the DSSCD group was also determined by a reduced abundance of goblet cells using alcian blue staining (Figure 1d) and decreased expression levels of muc2 and cldn2 (Figure 1e), which are critical components of the gut barrier. Thirteen proinflammatory cytokines were measured using fluorescence-encoded beads and analyzed by a flow cytometer. The proinflammatory cytokines monocyte-chemoattractant protein (MCP)-1, IL-6, and IL-1β were remarkedly elevated at the protein and mRNA levels in the DSSCD group (Figure 1f,g, Supplementary Figure S1), along with increased serum IL-6 concentrations (Figure 1h). In contrast, a challenge with C. difficile in control mice (CD group) did not result in any intestinal damage or enhanced inflammatory responses (Figure 1b-d). To determine whether the pathological effect of C. difficile on colitis was distinctive in this mouse model, we also investigated other common pathogens, namely Escherichia coli, Staphylococcus aureus, Enterococcus faecalis, and Candida albicans, but C. difficile had the greatest effect as shown by weight loss (Supplementary Figure S2). Intriguingly, when we measured the C. difficile burden from fresh colonic contents, it was comparable between DSS-treated and untreated mice at 6, 12, 24, or 48 hours post infection (Figure 1i, j). The number of C. difficile cells peaked at 6 hours, and they were barely detected after 48 hours, along with similar trends detected in the transcripts of the tcdB gene. These data suggested that, regardless of DSS treatment, oral gavage of C. difficile led to a non-sustained presentation of this pathogen, which could be considered as transient colonization. This condition aggravated the disease severity and induced excessive inflammatory responses of colitis. Figure 1.Increased disease severity in mice with DSS challenged with C. difficile. ( a) Schematic outlining the timing and treatment regiments for each group ($$n = 5$$–10/group). Mice were culled and samples for further analysis were collected at 2 days post-infection. ( b, c) Weight loss (b) and disease activity scores (c) in each group are shown, and significant differences between the DSS and DSSCD groups are indicated on the corresponding day (*$P \leq 0.05$, **$P \leq 0.01$). ( d) Representative images of H&E and alcian blue staining of colonic tissue (left) and histological scores (right). Scale bar, 200 μm. ( e – h) Relative expression levels of the gut barrier proteins MUC2 and CLDN2 (e) and the cytokines MCP1, IL-6, and IL-1β (g) were examined by real-time qPCR using actin as an internal control. Protein concentrations of MCP-1, IL-6, and IL-1β in the colon (f) and serum (h) were detected using flow cytometry. Colonic cytokine production (f) was normalized to the total protein concentration. ( i and j) the C. difficile burden in the colonic contents of mice in different groups at different time points post-infection was measured by CFU counting through culturing (i) and real-time qPCR analysis of tcdB relative expression (j). Each dot indicates an individual mouse. Data are shown as the mean ± standard deviation (SD) and represent at least three independent experiments. Statistical analysis between the groups was performed by the Mann – Whitney test. * $P \leq 0.05$, **$P \leq 0.01$; NS, not significant. ## Transient colonization of C. difficile alters the gut microbiota in mice with colitis Toxins A and B were the main virulent factors for C. difficile. To determine whether pathogen-mediated inflammation was dependent on toxin production, we challenged DSS mice with VPI10463 (a hypervirulent strain) and a non-toxigenic strain (NTCD) isolated from the clinic. The NTCD strain behaved similarly to the VPI10463 strain, leading to increased disease severity and robust inflammatory responses as indicated by weight loss, colonic tissue histology and histological scores, and IL-6 concentrations (Figure 2a–d). This implied that the pathogenic effect of C.difficile on colitis might be independent of toxin production. Because of the importance of the gut microbiota in regulating intestinal immune responses associated with the pathophysiology of colitis, we performed 16S rRNA and internal transcribed spacer (ITS) rDNA region sequencing to characterize the bacterial and fungal community structures, respectively. After confirming the sufficient depth of sequencing coverage via rarefaction curves (Supplementary Figure S3a), we conducted beta diversity analysis of principle coordinates analysis (PCoA) of the Bray – Curtis distance to assess the variability among the groups. The fecal samples showed that bacterial communities in the DSSCD and DSS groups were clearly separated ($R = 0.6652$, $$P \leq 0.001$$), while those in the CD and control groups showed a similar microbiota structure (Figure 2e). Additionally, microbial richness and diversity were remarkably reduced in the DSS and DSSCD groups as shown by alpha diversity measured with the ACE and Shannon indices (Suplementary Figure S3a). We did not observe any significant changes in the fungal community among the four groups, as shown by beta or alpha diversity determined by PCoA ($R = 0.116$, $$P \leq 0.414$$) or the ACE or Shannon index (Suplementary Figure S3b, c). These data suggest that C. difficile colonization affects the gut microbiota, especially the bacterial community structure. Figure 2.The gut microbiota is altered in mice with colitis challenged with C. difficile. ( a – d) the effect of C. difficile on mice with colitis was independent of toxin production, as indicated by weight loss (a), H&E staining (b), histological scores (c), and production of IL-6 as shown by ELISA (d). Scale bar, 200 μm. ( e and f) PcoA at the OTU level for samples from feces (e) and colonic contents (f), and the plots were based on the Bray – Curtis distance. The horizontal and vertical axes represent inter-sample variations. Each point represents an individual sample, and different colors refer to different groups. ( g) Cladograms were generated by linear discriminant analysis effect size analysis to detect the differences in bacterial taxa between the DSSCD and DSS groups. Circles show phylogenetic levels from the phylum to the genus. To screen out differentially abundant taxa, the linear discriminant analysis threshold score was set to>4.0. Red and blue bars indicate taxa enrichment in the DPI48H_DSS and DPI48H_DSSCD groups, respectively. ( h) Correlations between IL-6 levels and relative abundance of g_prevotellaceae_ucg001 (OTU288) in DSSCD groups (DPI6H_DSSCD and DPI48_DSSCD) were analyzed by Spearman’s correlation. Each dot represents a value from an individual mouse. Data are expressed as the mean±sd. Statistical differences between groups were assessed by the Mann – Whitney test. * $P \leq 0.05.$ We then investigated the dissimilarity of bacterial communities between the DSSCD and DSS groups using colonic contents collected at 6 and 48 hours post-infection, grouped as DPI6H_DSSCD, DPI48H_DSSCD, DPI6H_DSS, and DPI48H_DSS. As shown by rarefaction curves, sequences were sufficient to analyze the bacterial diversity of the samples (Suplementary Figures 3d). PCoA analysis by the Bray-Curtis distance and weighted unifrac metrics both showed that transient colonization of C. difficile was a major driver of community similarity because mice challenged with C. difficile were clustered together and were clearly distinguishable from DSS mice (Figure 2f, Supplementary FigureS 3f). Additionally, the communities in the DPI6H_DSS and DPI48H_DSS were distinct, indicating different disease statuses of colitis. Consistently, alpha diversity including Shannon and ACE indices as well as Faith’s phylogenetic diversity and species richness also showed comparable indices in the C. difficile challenged group (DPI6H_DSSCD and DPI48H_DSSCD), and these indices were slightly higher than those in the DSS group (DPI6H_DSS and DPI48H_DSS) (Supplementary Figure S3e). We conducted linear discriminant analysis of the effect size (linear discriminant analysis > 4.0) to detect the predominant taxonomic differences between the groups. At 6 hours post-infection, f_Lachnospiraceae and f_Oscillospirales were more abundant in the DSSCD group than in the DSS group (Supplementary Figures 3 g). However, after 48 hours post-infection, the proportions of p_Bacteroidota, mainly g_Prevotellaceae_UCG-001, g_Muribaculaceae, and g_Alloprevotella were lower, while those of p_Firmicutes, including g_Ruminococcaeceae, g_Anearoplasma, and g_Harryflintia were relatively enriched in the DSSCD group compared with the DSS group (Figure 2g). The Wilcoxon rank sum test between the DPI48H_DSSCD and DPI48H_DSS groups at the operational taxonomic unit (OTU) level showed that OUT288 and OTU287 were the top enriched species that were different between the groups, and they corresponded to g_Prevotellaceae_UCG001 and f_Muribaculaceae, respectively (Supplementary Figure S3h). To identify the prominent genus associated with inflammatory indices, we conducted a correlation analysis between MCP-1 and IL-6 concentrations and the relative abundance of genera. Among the most 20 abundant species at the OTU level, OTU288 and OTU287 showed significant negative correlations with the production of IL-6 and MCP-1 (Supplementary S3i). In multiple regression analysis models including treatment groups as variables, OTU288 was detected correlated with IL-6 levels independently ($$P \leq 0.041$$) in the DSSCD groups (DPI6H_DSSCD and DPI48H_DSSCD) (r2 = 0.5006, $$P \leq 0.0221$$) (Figure 2h), while OTU287 was found significant ($$P \leq 0.037$$) in DSS groups (DPI6H_DSS and DPI48H_DSS) (Supplementary S3j). OTU288 was also negatively correlated with the number of OTU reads of C. difficile (OTU418) (r2 = 0.2, $$P \leq 0.031$$) (Supplementary Figure S2k). Therefore, the gut bacterial communities that showed less abundance of g_Prevotellaceae_UCG001 driven by C. difficile were probably associated with the enhanced inflammatory status of colitis. Figure 3.Altered gut microbiota in DSSCD group contributes to the severity of colitis. ( a) Experimental design of FMT. The mice were pretreated with an antibiotic cocktail for 3 weeks, followed by $2\%$ DSS treatment and oral gavage of fecal samples derived from DSSCD and DSS group every other day during modeling. ( b – e) the severity of colitis in the mouse transplanted fecal microbiota in the DSSCD and DSS groups was assessed by body weight loss (b), H&E and alcian blue staining (c), histological scores (d), and IL-6 levels from colonic tissue (e). Scale bars, 200 μm. ( f) the relative abundance of OTU288 and OTU287 was quantified by real-time qPCR with normalization to total bacterial (16S rRNA). ( g – h) the effect of C. difficile on mice with colitis and antibiotic pretreatment was evaluated by weight loss (g), histological scores from H&E staining (h), and colonic IL-6 production by ELISA (i). Scale bars, 200 μm. Data are shown as the mean±sd and represent at least three independent experiments. Statistical analysis between groups was conducted by the Mann – Whitney test. * $P \leq 0.05$, **$P \leq 0.01.$ ## C. difficile-induced changes in the microbiota contribute to the severity of colitis To further determine the causality between C. difficile-related changes in the gut microbiota and disease severity, we transferred gut microbiota collected from DSSCD and DSS groups to microbiota-depleted mice through fecal microbiota transfer (FMT) assays (Figure 3a), grouped as FMT_DSSCD and FMT_DSS group, respectively. Donor drafts confirmed the absence of C.difficile and the differentially abundant microbes mainly g_Prevotellaceae_UCG001 (OTU288) by 16s rRNA sequencing (data not shown). Mice in the FMT_DSSCD group showed significantly more weight loss, higher histological scores calculated by infiltration of immune cells, intestinal epithelial integrity, and crypt structure, less mucin staining, and higher IL-6 concentrations than those in the FMT_DSS group (Figure 3b–e). Consistent with the production of cytokines, the relative DNA burden of OTU288 was obviously decreased in the FMT_DSSCD group, although no significant difference was found in OTU287 between the FMT_DSSCD and FMT_DSS groups. To determine if the microbiota is a unique factor affecting disease severity, we pretreated mice with antibiotics before DSS to lessen the effect on the gut microbiome. We then performed a C. difficile challenge on antibiotic-treated mice and found exacerbating disease, as shown by weight loss, tissue damage, and local inflammation (Figure 3g–i). This finding indicated that the gut microbiota was not the only factor affecting disease severity. Collectively, our findings suggested that the gut microbiota affected by C. difficile contributed to the aggravated colitis. ## C. difficile promotes colonic neutrophil accumulation The intestinal microbiota is considered a critical regulator of intestinal immunity. To better understand immune responses to C. difficile-related alteration of the microbiota in DSS mice, we characterized the accumulation of relevant immune cells, such as macrophages, neutrophils, and CD4+ T cell subtypes, from the lamina propria. The proportions of macrophages (CD11b+F$\frac{4}{80}$+) and CD4+ T cells, along with the main CD4+T subsets Th17 (CD4+IL17+) and Treg (CD4+Foxp3+), were comparable between DSSCD and DSS groups (Supplementary Figure S4a-d). Noticeably, we detected a robust increase in the proportion of neutrophils (CD11b+Ly6G+) driven by C. difficile in mice with colitis (Figure 4a), which was further suggested by myeloperoxidase (MPO) staining of colonic tissue (Figure 4b). Therefore, we focused on neutrophils to investigate how they are regulated by persistent colitis induced by C. difficile. Figure 4.Mice with colitis challenged with C. difficile show increased neutrophil infiltration and enhanced expression of migratory genes. ( a) Representative flow plots (left) and relative quantification (right) of the proportions of CD11b+Ly6G+ in CD45 cells. ( b) Representative images of immunohistochemical MPO staining. Scale bars, 200 μm. ( c) *Upregulated* genes in the DSSCD group relative to the DSS group were enriched for GO functional analysis. The top 10 pathways are shown in a bubble plot. The size and color of the bubble represent gene numbers enriched in each pathway and the respective enrichment significance. ( d) a heatmap depicts upregulated gene expression from the most enriched pathways between the DSSCD and DSS groups. ( e) a gene expression interaction network was generated by STRING and shown by Cytoscape. Red and blue rectangles represent up- and downregulated genes in the DSSCD group, respectively. The size of the rectangles indicates their betweenness centrality value. ( f) Neutrophils were isolated from bone marrow in the DSSCD and DSS groups. The number of migrated neutrophils in the lower chamber was counted using trypan blue staining. ( g) the mRNA transcripts of IL-6, IL-1β, CXCL2, and CCL4 in neutrophils isolated from the DSSCD and DSS groups were detected by real-time qPCR. Gene expression was normalized to β-actin. Data are shown as the mean±sd. Each dot indicates an individual mouse. Data (a, b, f, and g) are representative of at least three independent experiments. Statistical analysis between the groups was performed by the Mann – Whitney test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ To determine how the host responds to C. difficile-induced inflammation leading to the accumulation of neutrophils, we performed RNA sequencing and compared the gene expression of colonic tissues between the DSSCD and DSS groups. Overall, there were 290 upregulated and 206 downregulated genes that were differentially expressed in the DSSCD group relative to the DSS group (Supplementary Figure S5a). Samples from each group were distinctly clustered by these differentially expressed genes, which suggested that host responses in colitis were changed by C. difficile (Supplementary Figure S5b). We then performed enrichment analysis of differentially expressed gene sets using Gene Ontology (GO) biological terms. The pathways of downregulated genes were mostly enriched in immune responses related to monocytes, such as monocyte chemotaxis, lymphocyte migration, and mononuclear cell migration, and interferon beta responses (Supplementary Figure S5c). Strikingly, the enrichment of upregulated genes was mainly involved in neutrophil migration and chemotaxis, as well as cytokine secretion (Figure 4c), which was particularly consistent with phenotypes of neutrophil infiltration. Furthermore, we identified upregulated genes from the most enriched pathways and plotted them in a heatmap (Figure 4d). The transcription levels of genes responsible for neutrophil migration, such as the inflammasome pathway genes nlrp3 and Il1-β, and chemokines such as cxcl2, cxcr2, ccl3, and ccl4, were significantly increased (Figure 4d). Moreover, the protein interaction network of differentially expressed genes by STRING analysis showed that IL-1β might be the core protein interacting with differential expression genes (DEGs) (Figure 4e). Based on the above-mentioned findings of upregulated genes related to neutrophil recruitment, we aimed to determine the migratory capacity of neutrophils isolated from DSSCD and DSS groups using trans-well assays. More cells from the DSSCD group than those from the DSS group were transmitted from the upper chamber, which was attached to the reverse side of the membrane (Supplementary Figure S5d) and moved to the bottom (Figure 4f). These cells displayed robust trafficking ability. Moreover, neutrophils isolated from the DSSCD group also showed remarkably higher Il-6, Il1-β, Cxcl2, and Ccl4 mRNA expression than those from the DSS group, indicating a highly proinflammatory effect. Taken together, these findings suggest that C. difficile-aggravated colitis is characterized by accumulated neutrophils, and it shows a strong ability of migration and proinflammatory mediator production. ## An altered microbiota by C. difficile leads to the production of proinflammatory cytokines and chemokines in neutrophils and macrophages Neutrophils, which are a critical component of innate immunity, produce inflammatory cytokines and chemokines in an inflamed intestine by sensing microbial components and metabolites. Therefore, cross-talk between neutrophils and microbiota is considered essential in the gut microenvironment. We investigated whether the altered microbiota by C. difficile contributes to increased concentrations of inflammatory mediators in vitro. Fecal contents from DSSCD and DSS groups were collected and subjected to stimulation of neutrophils isolated from mice with colitis. We found that the proinflammatory cytokines IL-6, MCP-1, and IL-1β, and the chemokines Chemokine (C-X-C motif) ligand 2(CXCL2) and C-C Motif Chemokine Ligand 4 (CCL4) were significantly more highly expressed in neutrophils upon stimulation of the microbiota in DSSCD mice than in DSS mice (Figure 5a). Neutrophils can also be activated by these cytokines and chemokines released by macrophages. Therefore, we also assessed the effect of the microbiota on THP-1 cells, which is a human monocytic cell line differentiated into macrophages, in the presence of phorbol-12-myristate-13-acetate (PMA). As expected, the microbiota in the DSSCD group showed remarkably stimulated THP-1 cells, which expressed higher IL-6, IL-1β and MCP-1, CXCL2, and CCL4 concentrations. Taken together, these data demonstrated that the gut microbiota affected by C. difficile promoted the expression of genes related to inflammation and neutrophil migration. Figure 5.*Gut microbiota* in DSSCD mice induces neutrophils and THP-1 cells, leading to higher levels of proinflammatory cytokines and chemokines. Neutrophils (a) and THP-1 cells pretreated with PMA (b) were stimulated in vitro in the presence or absence of microbiota suspensions prepared from different groups for 3 hours. The mRNA levels of IL-6, IL-1β, MCP-1 CXCL2, and CCL4 were measured by real-time qPCR and normalized to β-actin. Statistical analysis was performed by the Mann – Whitney test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ Results are representative of at least three independent experiments. ## Inhibition of neutrophil migration attenuates colitis in DSS mice challenged with C. difficile Because DSS mice challenged with C. difficile showed increased infiltration of neutrophils triggered by an altered gut microbiota, we then aimed to determine the role of leukocyte migration in the process of colitis. CXCR2 is a chemokine receptor expressed in neutrophils, and it plays a crucial role in recruiting neutrophils and activating related chemokines. We then constructed a colitis model with C.difficile using the CXCR2 inhibitor SB225002 to hinder the migratory capacity of neutrophil (Figure 6a). The mice treated with SB225002 developed milder colitis post-infection of C. difficile, with less weight loss, less tissue damage as shown by hematoxylin and eosin (H&E) staining, and lower histopathological scores (Figure 6b, d). Consistently, we observed decreased numbers of MPO-positive cells by immunostaining (Figure 6c) and a reduced proportion of neutrophils (CD11b+Ly6G+) by flow cytometry, accompanied by decreased neutrophil-derived IL-6 and CXCL2 expression (Figure 6e, f). Interestingly, blockage of CXCR2 did not affect the colonic transcription levels of Il-6, Il-1β, Cxcl2, Mcp-1, or Ccl4 (Figure 6g, Supplementary Figure S6b), or the intestinal abundance of OTU288 and OTU287 (Figure 6h). However, the CXCR2 inhibitor SB225002 did not affect the diseases of colitis alone (Figure 6a, Supplementary FigureS 6a). Our data suggested that the inhibition of neutrophil migration effectively alleviated C. difficile-mediated disease in mice with colitis, without changing levels of inflammation cytokines and the proportions of predominant species. Figure 6.Inhibition of neutrophil migration by SB225002 alleviates colitis driven by C. difficile. ( a) Schematic diagram of the CXCR2 inhibition model. After DSS mice were challenged with or without C. difficile on day 7, the mice were injected i.P. with the CXCR2 inhibitor SB225002 at a dose of 1 mg/kg/mouse on days 7 and 8, and PBS was used as a negative control. ( b – d) the severity of colitis was assessed by weight loss (b), H&E and MPO staining of colonic sections (c), and histological scores (d). Scale bars, 200 μm. ( e and f) Representative flow cytometric plots (e) and a histogram (f) quantifying the proportions of CD11b+Ly6G+ in CD45+ cells. ( g and h) Expression of IL-6, IL-1β, and CXCL2 mRNA was detected by real-time qPCR in isolated neutrophils (g) and colonic tissue (h) with or without inhibitor treatment. ( i) Colonic contents were collected to detect the relative abundance of OTU288 and OTU287 as shown by real-time qPCR. DNA expression was normalized to total bacteria (16S rRNA). Significant differences were determined by the Mann – Whitney test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ Data are shown as the mean±sd and are representative of at least three independent experiments. i.P., intraperitoneally; DPI, days post-infection; NS, no significance. ## IL-1β secretion may lead to neutrophil accumulation during colitis challenged with C. difficile Since IL-1β was involved in leukocyte chemotaxis in GO biological terms, played a main role in interacting with DEGs as shown by protein interaction analysis (Figure 4e), and its production was not affected by neutrophil accumulation (Figure 6g), we then hypothesized to examine whether IL-1β is implicated in regulating neutrophil recruitment and intestinal inflammation during colitis. We constructed a colitis mouse model with or without a C. difficile challenge using MCC950, which is a selective NLRP3 inflammasome inhibitor, to inhibit the production of IL-1β in vivo (Figure 7a). MCC950 treatment significantly diminished the pathogenic effect of C. difficile on colitis, as evaluated by milder weight loss and histological pathology of colonic tissues (Figure 7b–d), but it did not affect the severity of colitis alone (Supplementary Figure S7). As expected, reduced neutrophil infiltration in the colon with less MPO-positive cells was observed (Figure 7c). In addition, colonic transcription of IL-1β, IL-6, and CXCL2 was consistently lower (Figure 7e), along with a decreased production of CXCL2 in neutrophils with MCC950 treatment (Figure 7f). We also detected the relative expression of OTU288 and OTU287, but no changes were found with MCC950 treatment. Taken together, our findings suggested that elevated IL-1β expression driven by C. difficile in colitis contributed to neutrophil infiltration and intestinal inflammatory responses. Figure 7.Inhibition of IL-1β by MCC950 ameliorates the severity of colitis challenged with C. difficile. ( a) Schematic diagram of the MCC950 inhibition model. After DSS mice were challenged with C. difficile on day 7, they were injected i.P. with MCC950 at a dose of 20 mg/kg/mouse on days 7 and 8, and PBS was used as a negative control. ( b – d) Disease severity was assessed by weight loss (b), H&E staining (c), and histological sores (d). ( e and f) Expression of IL-1β, IL-6, and CXCL2 mRNA in colonic tissue (e) and CXCL2 mRNA in isolated neutrophils (f) were detected by real-time qPCR and normalized to β-actin. ( g) the relative DNA burden of OTU288 and OTU297 was measured by real-time qPCR and normalized to total bacteria (16s rRNA). Significant differences were determined by the Mann – Whitney test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ Data are shown as the mean±sd and are representative of at least three independent experiments. i.P., intraperitoneally; DPI, days post-infection; NS, no significance. ## Discussion C. difficile has a pathogenic effect on patients with IBD, with severe symptoms and poor clinical outcomes. The association between IBD and CDI regarding pathogenic susceptibility, changes in the microbiota, and host immune responses is so complicated that applying an appropriate murine model is necessary to clarify this association. Researchers have constructed a colitis model concurrent with CDI by antibiotic administration to disturb the indigenous microbiota and make it susceptible for colonization.10,22 Antibiotics are not a necessary prescription for patients with IBD and whether they are an independent risk factor for concurrent CDI remains controversial.23,24 Therefore, the pretreatment of antibiotics in mouse models is unlikely to represent the clinical situation. Therefore, we challenged mice with DSS-induced colitis with C. difficile in the absence of antibiotics. Notably, we found that although the C. difficile challenge to non-antibiotic colitis failed to lead to persistent colonization, it still considerably aggravated the disease status. The minor difference between our study and Zhou et al. ’s report, which showed no worsened histopathological damage, might be attributed to the different concentrations of DSS used for colitis.22 Moreover, studies have shown that intestinal inflammation along with an altered microbiota in colitis are beneficial for C. difficile colonization in the IL10−/− colitis mouse model.9 However, in our study, we found that treatment with DSS did not significantly affect the colonization status of C. difficile, but mice with only colitis challenged with C. difficile developed much more severe intestinal inflammation. This finding suggested that the underlying intestinal microenvironment of colitis, other than C. difficile itself, was a prerequisite for disease progression. Additionally, it warns that exposure to a C. difficile-contaminated environment puts patients with IBD at risk of severe symptoms, even if they are non-persistent colonizers. By comparing to other commonly detected gut pathogens, C.difficile had a greater pathogenic effect on colitis, emphasizing the importance to make further investigations on the mechanisms by which C.difficile aggravated colitis. Accumulating studies have indicated that the gut microbiota participates in the pathogenesis of IBD.25 Patients with IBD and CDI harbor more-pronounced intestinal dysbiosis, but the observed features do not clearly indicate whether CDI expedites the intestinal dysbiosis or patients with disturbed gut microbiota are more prone to CDI.14,26 *In this* study using a dynamic analysis of bacterial communities post-infection, we found that a surge of C. difficile at 6 hours increased the microbial diversity with a higher abundance of Lachnospirales and Oscillospiraceae (previously known as Ruminococcaceae). Lachnospiraceae and Oscillospiraceae families are considered potentially protective, producing short-chain fatty acids that provide energy sources to colonic enterocytes and secondary bile acids to hinder the growth of C. difficile.26–29 Similar results were also described from clinical studies, which showed that the emergence of C. difficile might cause a beneficial transient change in bacterial taxa.30,31 Subsequently, at the height of the disease (48 hours post-infection) when C. difficile cells had already been cleared, bacterial communities in DSSCD mice were characterized by a lower abundance of g_Prevotellaceae_UCG-001 (OTU288) and g_Muribaculaceae (OTU287), which belong to the Bacteroidetes phylum. The declined abundance of g_Prevotellaceae_UCG001 was independently related to the disease severity and showed a highly inflammatory effect on neutrophils and macrophages in colitis mice with C.difficile. Members of Bacteroidetes degrade polysaccharides,32 supplying nutrients to other residents in the gut and producing metabolites to exert an anti-inflammatory effect.33 Therefore, we propose that they have a protective effect on colitis. However, we could not determine the exact role of either genus. Collectively, the influx of C. difficile might induce protective microbiota responses, but the microbiota still accelerates gut inflammation in colitis. Disturbed gut microbial communities, accompanied by uncontrolled host immune responses, are major drivers of disease complications.11 An inflamed colon driven by C. difficile in the colitis mouse model shows accumulation of neutrophils, in addition to enriched gene sets associated with the migration ability. Consequently, researchers have paid attention to the function of neutrophils. In patients with IBD, the degree of neutrophil infiltration and the reappearance of neutrophils in the intestinal mucosa are thought to represent the disease activity and clinical remission, respectively.34,35 During the whole process of DSS modeling, inhibiting neutrophil infiltration daily by blocking CXCR2, a common ligand for neutrophil functionality, was reported to decrease proinflammatory cytokine production and reduced intestinal damages36. However, since neutrophil infiltration and inflammatory cytokine production in DSS models were initiated one-day post treatment37, blockade of CXCR2 when colitis was developed presently showed limited effect on inflammation status induced by DSS, but significantly alleviated disease severity induced by C.difficile. This finding indicated a detrimental role of colonic neutrophils in the present colitis mouse model challenged with C. difficile, despite their reported protective role in attacking invasion of pathogens.38 MPO staining of colonic tissues reflected reduced levels of reactive oxygen species, which are considered hazardous to tissues by oxidative DNA damage to epithelial cells.39 Neutrophil-derived cytokines and chemokines were also decreased, which are closely related to the initiation and continuation of inflammation by regulating innate and adaptive immune responses.40 CXCR2 blockade probably inhibited its cross-talk with other immune cells. Moreover, the unchanged levels of colonic inflammatory mediators, such as IL-6, IL-1β, and CXCL2, indicated neutrophils’ downstream act in the intestinal inflammations. Using RNA sequencing analysis, we mapped the differentially expressed genes enriched in leukocyte chemotaxis or migration, especially IL-1β and CXCL2, CXCR2, CCL3, and CCL4. Furthermore, IL-1β was a core gene, which showed the most intricate interactions with other genes, and may be an upstream regulator of neutrophil recruitment. Though IL-1β is important for repair of intestinal epithelial cell and reconstitution of the epithelial barrier41, the excessive production of IL-1β may exacerbate colon inflammation and is associated with intestinal inflammation of IBD and the pathogenesis of CDI.42–44 However, the role of IL-1β in IBD with CDI is undetermined. Our study showed a central role of IL-1β in regulating neutrophil functions and the colonic inflammatory status in C.difficile-driven inflammations in colitis. MCC950 was considered to reduce IL-1β release by blocking NLRP3-dependent ASC oligomerization and NLRP3 inflammasome assembly45, meanwhile it could also decrease the mRNA expressions of IL-1β46. Since IL-1β expression could be regulated by IL-1β itself and multiple cytokines47, the reduced transcriptional levels by MCC950 could be attributed to the inhibited secretion of IL-1β or IL-1β downstream cytokines. Inhibiting IL-1β production by MCC950 failed to relieve the symptoms of colitis when it was used after colitis induced but significantly ameliorated the colitis with subsequent C.difficile invasion. Hopefully, IL-1β antagonists may be therapeutic agents for patients with IBD and a risk of CDI. How IL-1β is affected by C. difficile in colitis needs to be clarified. C. difficile in cells and toxins can trigger the activation of inflammasomes,44,48,49 releasing IL-1β and mediating intestinal inflammation. The altered microbiota affected by C. difficile in our study also promoted IL-1β production in vitro. Since the colonization of C. difficile in our model is transient, whether IL-1β was induced by C. difficile directly, indirectly by the microbial community, or both synergistically is unknown. IL-1β might regulate the neutrophils by driving Th17 differentiation to induce emergency granulopoiesis or by upregulating the neutrophil chemoattractant to enhance chemotaxis.50,51 *In this* study, IL-1β appeared to regulate the expression of several chemokines, such as CXCL2, which is a ligand for CXCR2 involved in triggering neutrophil function and recruitment. Recently, Pavlidis et al. predicted IL-1β as an important driver of neutrophil-active chemokines via IL-22 pathways through bioinformatics analysis.52. Nevertheless, the precise mechanism regarding how IL-1β is involved in C. difficile-induced inflammation in colitis requires further investigation. In summary, our study shows the pathogenic effect of transient colonization of C. difficile in mice with DSS-induced colitis. We conducted a comprehensive investigation on the interaction between the gut microbiota and host immune responses. We highlight the importance of changes in the microbiota, IL-1β production, and neutrophil accumulation, and propose a possible pathway of microbiota – IL-1β–neutrophil regulation in C. difficile-driven gut inflammation in DSS-induced colitis. Our findings will hopefully be helpful in determining potential therapeutic targets to treat patients with IBD concurrent with CDI. ## Bacterial strains and culture conditions C. difficile VPI10463 (ATCC 43,255), *Escherichia coli* (ATCC 25,922), *Staphylococcus aureus* (ATCC 25,923), *Enterococcus faecalis* (ATCC 29,212), Candida albicans (ATCC 90,028), and THP-1 cells (ATCC TIB-202) were purchased from the American Type Culture Collection (ATCC). The non-toxigenic strain NTCD was isolated clinically and identified as negative for toxin genes, such as tcdA, tcdB, cdtA, and cdtB, previously in our laboratory.53 THP-1 cells were cultured in RPMI 1640 medium supplemented with $10\%$ fetal bovine serum (Gibco, USA) and 50 mg/L penicillin/streptomycin (Gibco, USA). C. difficile strains were cultured in brain heart infusion (Oxoid Ltd., USA) broth supplemented with 5 g/L yeast extract (Oxoid Ltd.), $0.1\%$ L-cysteine (Sangon Biotech, China), and $0.1\%$ sodium taurocholate (Sangon Biotech, China) for 48 hours at 37°C anaerobically. The other bacterial strains and C. albicans were cultured in Luria-Bertani (Sangon Biotech, China) and yeast peptone dextrose broth (Sangon Biotech, China), respectively, for 24 hours aerobically. The liquid culture was centrifuged at 1500 g for 5 minutes, and the pellet was washed twice with sterile phosphate-buffered saline (PBS). The inoculum was adjusted to approximately 5 × 106 Colony Forming Units (CFU)/ml. ## DSS-induced colitis and pathogen challenge Male mice aged 6 weeks were purchased from Charles River (Beijing, China) and housed at a constant temperature of 20°C–22°C with a 12-hour light–dark circle under specific pathogen-free conditions. The mice were acclimatized for 1 week before modeling. The experimental schema is shown in Figure 1a. To induce acute colitis, the mice were administered $2\%$ (w/v) DSS (molecular weight: 36000–50,000 Da, MP Biomedicals) in drinking water for 5 days, followed by the administration of regular water. To construct IBD concurrent with C. difficile or other pathogen models, the mice were administered with 106 CFUs by oral gavage in random order on day 7. Four groups were used for further monitoring, namely the DSS (DSS treated mice without C.difficile challenge), DSSCD (DSS treated mice challenged with C. difficile), control (DSS untreated mice without C. difficile administration), and CD (DSS untreated mice challenged with C. difficile only) groups. To perform CXCR2 and NLRP3 inhibition in the in vivo model, the DSSCD group were administered the CXCR2 selective inhibitor SB225002 (Sigma Aldrich, USA) (1 mg/kg/mouse) or the NLRP3 selective inhibitor MCC950 (Selleck, USA) (20 mg/kg/mouse) intraperitoneally (i.p.) on days 7 and 8 (Figures 6a and 7a). During the modeling, the mice were weighed and scored daily. Disease activity scores were assessed by weight loss, stool consistency, and bleeding as described previously54.Samples including stools, serum, and colonic tissue and contents were collected on day 9, at 2 days post-C. difficile treatment, unless specified otherwise. To monitor the burden of C. difficile in vivo, the mice were scarified at 6, 12, 24, and 48 hours post-infection, and colonic contents were freshly collected for further analysis. All animal experiments were approved by the Ethics Committee of Ruijin Hospital, Shanghai Jiaotong University School of Medicine. ## Quantification of the C. difficile burden To quantify the C. difficile burden, colonic contents were suspended in ethanol, serially diluted, and plated anaerobically at 37°C on brain heart infusion broth supplemented with 5 g/L yeast extract, $0.1\%$ L-cysteine, $0.1\%$ sodium taurocholate, 16 mg/L cefoxitin (BBI Life Science, USA) and 500 mg/L D-cycloserine (BBI Life Science, USA). After 48 hours of incubation, CFUs were counted and normalized to the stool weight. ## Histological analysis Tissue samples were harvested from the colon, rinsed gently with PBS to remove colonic content, and fixed in $4\%$ paraformaldehyde. Embedded samples were stained with H&E and alcian blue, and MPO immunostaining was also performed55. Scores ranging from 0 to 3 were used to assess epithelial disruption, crypt architecture, and the degree and range of inflammatory cell infiltration56. This assessment was made by at least two independent, blinded observers. Images were captured by a Nikon DS-F2 microscope. ## Tissue protein and cytokine analysis Colonic tissues were isolated and homogenized in 1 ml of PBS with a protease inhibitor cocktail (Roche, USA). The homogenate was centrifuged at 12,000 g for 5 minutes, and the supernatants were collected and stored at−80°C. The protein concentrations were measured using the Enhanced BCA Protein Assay Kit (Beyotime, China). The production of proinflammatory cytokines, namely IL-23, IL-1⍺, interferon-γ, tumor necrosis factor-⍺, MCP-1, IL-12, IL-1β, IL-10, IL-6, IL-27, IL-17A, interferon-β, and Granulocyte-macrophage colony-stimulating factor (GM-CSF), was detected by flow cytometry using the FACSCantoII (BD, USA) with fluorescence-encoded beads in accordance with the manufacturer’s instructions of the LEGENDplex Mouse Inflammation Panel (Biolegend, USA). The production of IL-6 was also measured by the enzyme-linked immunosorbent assay (ELISA) kit (Biolegend, USA). Cytokine concentrations were normalized to the total protein concentration. ## DNA extraction and real-time quantitative polymerase chain reaction DNA was extracted from fresh colonic contents using the TIANamp Stool DNA Kit (Tiangen Biotech, China) in accordance with the manufacturer’s instructions. The expression levels of the tcdB gene and the relative abundance of OTU288 (g_Prevotellaceae_UCG-001) and OTU287 (g_Muribaculaceae) were quantified by performing real-time quantitative polymerase chain reaction (qPCR) using TB Green™ Premix Ex Taq™ (Takara, Japan) on a Light Cycler 480 Real-Time PCR system (Roche). The data were normalized to total bacteria (16S rRNA). The primer sequences are listed in Supplementary Table S1. ## RNA extraction and real-time qPCR Total RNA from tissue or cell samples was extracted with TRIzol Reagent (Invitrogen, USA). Complementary DNA was synthesized by the PrimeScript RT Reagent Kit (Takara). Real-time qPCR was conducted as described above, using β-actin as the internal control. The primers are listed in Supplementary Table S1. ## Lamina propria cell isolation and flow cytometry To isolate mononuclear cells in the lamina propria, tissues were incubated with PBS supplemented with 1 mM DL-dithiothreitol (Sigma Aldrich, USA) for 30 minutes in an incubator at 37°C. This was followed by another 30 minutes’ treatment in PBS supplemented with 30 mM Ethylenediaminetetraacetic acid (EDTA) (BBI Life Science, China). The tissues were then cut into small pieces and digested in the RPMI 1640 medium (Gibco, USA) with $10\%$ fetal bovine serum (Gibco, USA), 200 U/ML collagenase (Sigma Aldrich), and 150 μg/ML deoxyribonuclease (Sigma Aldrich, USA) for 1 hour. After the cells were strained through 40-μm filters, they were obtained in the $40\%$ to $80\%$ interface of Percoll (Sigma Aldrich, USA) by density gradient centrifugation. To perform flow cytometric analysis, 106 cells were resuspended in fluorescence-activated cell sorting buffer comprising PBS with $0.5\%$ bovine serum albumin (BBI Life Science, China) and 2 mM EDTA. After blocking the Fc receptor with anti-mouse CD$\frac{16}{32}$ and staining with the viability dye Zombie (Biolegend, USA), the following antibodies were stained for surface markers: FITC-CD45.2, PerCP/Cyanine5.5-CD11b, PerCP/Cyanine5.5-CD4, APC-CD25, PE-Ly-6 G, and APC-F$\frac{4}{80.}$ To perform Treg analysis, cells were fixed and permeabilized with Fix/Perm solution (Biolegend, USA) for 30 minutes after staining with surface markers, and then incubated with PE-Foxp3 antibody. To detect IL-17A, cells were stimulated with cell activation cocktail (Biolegend, USA) in RPMI 1640 medium for 5 hours, followed by surface marker labeling. The cells were then treated with a transcription factor staining buffer set (BD Biosciences, USA) in accordance with the manufacturer’s protocols and were stained with PE-IL-17A. All antibodies used in this study were purchased from Biolegend, USA. Flow cytometric data were acquired on a BD Canto II flow cytometer (BD Biosciences) and were analyzed using Flowjo software 10.0. ## 16s rRNA and ITS1 sequencing Genomic DNA was extracted from colonic contents as described above. The DNA quality and concentration were determined with a NanoDrop ND-2000 spectrophotometer (Thermo Scientific, USA). DNA was sent to Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China) for amplification, sequencing, and data processing as described previously.31 Briefly, the hypervariable V3–V4 regions were amplified with the primer pairs 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806 R (5′-GGACTACHVGGGTWTCTAAT-3′) for 16s rRNA sequencing. The fungal ITS1 was amplified with the primer pairs ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2R (5′-GCTGCGTTCTTCAT CGATGC-3′) for fungal sequencing. Amplicons were paired-end sequenced on an Illumina MiSeq PE300 platform (San Diego, CA) in accordance with the standard protocols. Raw FASTQ files were de-multiplexed using an in-house per script, quality-filtered by fastp version 0.19.6, and then merged by FLASH version 1.2.7. The optimized sequences were clustered into OTUs with $97\%$ sequence similarity and removed chimera using UPARSE 7.157. The taxonomy of each OTU representative sequence was analyzed by RDP Classifier version 2.2 against a 16S rRNA gene database (e.g., Silva v138) or the Targeted Host Fungi ITS1 database using a confidence threshold of 0.7. All bioinformatic analyses were performed by the Majorbio Cloud Platform (https://cloud.majorbio.com/). ## RNA sequencing and transcriptomic analysis RNA derived from colonic tissue was extracted as described above. An RNA sequencing transcriptome library was prepared by following the TruSeqTM RNA sample preparation kit from Illumina. Briefly, mRNA was isolated in accordance with the polyA selection method by oligo(dT) beads and then fragmented by fragmentation buffer. Double-stranded cDNA was synthesized using the SuperScript double-stranded cDNA synthesis kit (Invitrogen) with random hexamer primers. Libraries were then size selected for cDNA target fragments of 300 base pairs on $2\%$ Low Range Ultra Agarose followed by PCR amplification using Phusion DNA polymerase (NEB, USA) for 15 PCR cycles. After quantification by TBS380, paired-end RNA sequencing was performed with the Illumina HiSeq xten/NovaSeq 6000 sequencer. The raw paired-end reads were trimmed, and the poor quality reads were removed by SeqPrep (https://github.com/jstjohn/SeqPrep) and Sickle (https://github.com/najoshi/sickle), respectively, with default parameters. Clean reads were separately aligned to reference the genome with the orientation mode using HISAT2 (http://ccb.jhu.edu/software/hisat2/index.shtml) software and assembled by StringTie (https://ccb.jhu.edu/software/stringtie/index.shtml?t=example). Gene abundance was quantified by RSEM (http://deweylab.biostat.wisc.edu/rsem/), and a differential expression analysis was performed using differentially expressed genes with |log2Fold Change (FC)|>1. In addition, a functional enrichment analysis of GO was performed to identify which differentially expressed genes were significantly enriched in GO terms at a Bonferroni-corrected p value ≤ 0.05 compared with the whole transcriptomic background. This analysis was carried out by Goatools (https://github.com/tanghaibao/Goatools). All bioinformatic analyses were performed using the Majorbio Cloud Platform. ## Antibiotic treatment and FMT To clear the gut microbiota, mice received antibiotic cocktails containing 1 g/L ampicillin, 1 g/L metronidazole, 0.5 g/L vancomycin, and 1 g/L neomycin for 2 weeks in drinking water. This was followed by another week of administration of 2 g/L streptomycin, 0.17 g/L ciprofloxacin, 0.125 g/L gentamicin, and 1 g/L bacteriocin. After antibiotic treatment, stool samples of mice were collected and cultured on Columbia blood agar plate (Chromagar, China) anaerobically and aerobically to confirm microbiota depletion. Subsequently, the mice were modeled with $2\%$ DSS as described above. Fecal transplant samples were collected at day 9, 2 days post infection, prepared using colonic contents pooled from DSSCD and DSS group donor mice ($$n = 5$$–10/group) and stored at −80°C. 16s rRNA sequencing was analyzed for transplant donor samples to assess the microbial composition and confirm the absence of C.difficile. On the day of transplantation, samples were resuspended in sterile PBS at a concentration of 100 mg/mL, and the supernatants were collected after centrifugation at 800 rpm for 3 minutes. Transplantation should be completed with fresh supernatants by oral gavage within 10 mins to minimize changes in microbial compositions58. Antibiotic-treated mice were administered 200 μL of PBS suspensions in each mouse by oral gavage every other day from the start of DSS modeling. ## Neutrophil isolation, transmigration assay, and cellular stimulation Neutrophils were collected from mouse bone marrow as described previously.59 Briefly, bone marrow cells were harvested in the RPMI 1640 medium supplemented with $10\%$ fetal bovine serum and $1\%$ penicillin/streptomycin. The neutrophils were purified from the interface between Hisotopague 1119 (Sigma Aldrich, USA) and Histopaque 1077 (Sigma Aldrich, USA) by density gradient centrifugation. The cells were counted with trypan blue and identified by flow cytometry staining with anti-CD45, anti-CD11b, and anti-Ly6G. In the transmigration assay, approximately 2 × 105 cells in 200 μl of medium were seeded above the transmigration membrane, while culture medium supplemented with $20\%$ fetal bovine serum was added to the basolaterial side. After incubation for 24 hours, migrated cells in the lower chamber were collected for counting. Migrated cells on the membrane of the basolateral side were fixed in $4\%$ paraformaldehyde for 30 minutes, followed by staining with $0.1\%$ crystal violet (Sangon Biotech). Images were acquired by optical microscopy. To perform stimulation assays, microbiota samples from the DSS and DSSCD groups were prepared in the same manner as that for FMT donors. THP-1 cells were seeded on 12-well plates and incubated with 100 ng/ML PMA (Sigma Aldrich) for 1 day, and then cultured by fresh medium without PMA for another 24 hours. Approximately, 106 isolated neutrophil cells or THP-1 cells in each well were incubated with 50 μl of microbiota PBS suspension for 3 hours. The cells were collected and subjected for further analysis. ## Statistical analysis Statistical analyses were conducted using SPSS software version 16.0, and $P \leq 0.05$ was considered significant. Data were generated using Graphpad Prism software version 8.0 and R software version 4.0.5. Statistical differences between the two groups were assessed using Welch’s t-test or the Mann–Whitney test depending on whether the data showed normal distribution Rarefaction curves were generated to assess the sufficiency of sequence reads to describe the bacterial diversity and rarified to lowest OTUs per sample. Alpha diversity including the Shannon and ACE indices as well as Faith’s phylogenetic diversity and species richness were calculated at the OTU level and compared among the groups using the Student’s t-test or paired t-test. The difference in PCoA of the Bray – Curtis distance and weighted unifrac metrics was compared by Adonis analysis. The predominant phyla or genera of the linear discriminant analysis effect size were compared using the Wilcoxon rank-sum test or Wilcoxon signed-rank test. Correlations between species relative abundance and cytokine levels were calculated using Spearman’s analysis. Since the elevated cytokines could also be affected by treatment groups, we performed multiple-regression analysis with results from four groups (DPI6H_DSSCD, DPI48H_DSSCD, DPI6H_DSS, and DPI48H_DSS), DSSCD groups (DPI6H_DSSCD and DSS48H_DSSCD), and DSS groups (DPI6H_DSS and DPI48H_DSS) separately, involving abundances of OTU288 and OTU287, treatment groups, and levels of IL-6 as variables using SPSS version 24.0. ## Disclosure statement No potential conflict of interest was reported by the authors. ## Data availability statement The sequence dataset generated from this study has been deposited in the NCBI database under BioProject number PRJNA897872 in https://www.ncbi.nlm.nih.gov/bioproject/. The data that support the findings of this study are available from the corresponding author upon reasonable request. ## Supplementary material Supplemental data for this article can be accessed online at https://doi.org/$\frac{10.1080}{19490976.2023.2192478.}$ ## References 1. Schnizlein MK, Young VB.. **Capturing the environment of the Clostridioides difficile infection cycle**. *Nat Rev Gastroenterol Hepatol* (2022) **19** 508-23. DOI: 10.1038/s41575-022-00610-0 2. Baumgartner M, Pfrunder-Cardozo KR, Hall AR. **Microbial community composition interacts with local abiotic conditions to drive colonization resistance in human gut microbiome samples**. *Proc Biol Sci* (2021) **288** 20203106. DOI: 10.1098/rspb.2020.3106 3. Rodriguez C, Van Broeck J, Taminiau B, Delmée M, Daube G. **Clostridium difficile infection: early history, diagnosis and molecular strain typing methods**. *Microb Pathog* (2016) **97** 59-78. DOI: 10.1016/j.micpath.2016.05.018 4. Kucharzik T, Ellul P, Greuter T, Rahier JF, Verstockt B, Abreu C, Albuquerque A, Allocca M, Esteve M, Farraye FA. **ECCO guidelines on the prevention, diagnosis, and management of infections in inflammatory bowel disease**. *J Crohns Colitis* (2021) **15** 879-913. DOI: 10.1093/ecco-jcc/jjab052 5. Ananthakrishnan AN. **Clostridium difficile infection: epidemiology, risk factors and management**. *Nat Rev Gastroenterol Hepatol* (2011) **8** 17-26. DOI: 10.1038/nrgastro.2010.190 6. Rao K, Higgins PD. **Epidemiology, diagnosis, and management of clostridium difficile infection in patients with inflammatory bowel disease**. *Inflamm Bowel Dis* (2016) **22** 1744-1754. DOI: 10.1097/mib.0000000000000793 7. Singh H, Nugent Z, Yu BN, Lix LM, Targownik LE, Bernstein CN. **Higher incidence of clostridium difficile infection among individuals with inflammatory bowel disease**. *Gastroenterology* (2017) **153** 430-438.e2. DOI: 10.1053/j.gastro.2017.04.044 8. Balram B, Battat R, Al-Khoury A, D’aoust J, Afif W, Bitton A, Lakatos PL, Bessissow T. **Risk factors associated with clostridium difficile infection in inflammatory bowel disease: a systematic review and meta-analysis**. *J Crohns Colitis* (2019) **13** 27-38. DOI: 10.1093/ecco-jcc/jjy143 9. Abernathy-Close L, Barron MR, George JM, Dieterle MG, Vendrov KC, Bergin IL, Young VB. **Intestinal inflammation and altered gut microbiota associated with Inflammatory bowel disease render mice susceptible to clostridioides difficile colonization and infection**. *mBio* (2021) **12** e0273320. DOI: 10.1128/mBio.02733-20 10. Saleh MM, Frisbee AL, Leslie JL, Buonomo EL, Cowardin CA, Ma JZ, Simpson ME, Scully KW, Abhyankar MM, Petri WA. **Colitis-induced Th17 cells increase the risk for severe subsequent clostridium difficile infection**. *Cell Host Microbe* (2019) **25** 756-765 e5. DOI: 10.1016/j.chom.2019.03.003 11. Neurath MF. **Targeting immune cell circuits and trafficking in inflammatory bowel disease**. *Nat Immunol* (2019) **20** 970-979. DOI: 10.1038/s41590-019-0415-0 12. Lavelle A, Sokol H. **Gut microbiota-derived metabolites as key actors in inflammatory bowel disease**. *Nat Rev Gastroenterol Hepatol* (2020) **17** 223-237. DOI: 10.1038/s41575-019-0258-z 13. Ott SJ, Musfeldt M, Wenderoth DF, Hampe J, Brant O, Fölsch UR, Timmis KN, Schreiber S. **Reduction in diversity of the colonic mucosa associated bacterial microflora in patients with active inflammatory bowel disease**. *Gut* (2004) **53** 685-693. DOI: 10.1136/gut.2003.025403 14. Sokol H, Jegou S, McQuitty C, Straub M, Leducq V, Landman C, Kirchgesner J, Le Gall G, Bourrier A, Nion-Larmurier I. **Specificities of the intestinal microbiota in patients with inflammatory bowel disease and Clostridium difficile infection**. *Gut Microbes* (2018) **9** 55-60. DOI: 10.1080/19490976.2017.1361092 15. Uhlig HH, Powrie F. **Translating immunology into therapeutic concepts for inflammatory bowel disease**. *Annu Rev Immunol* (2018) **36** 755-781. DOI: 10.1146/annurev-immunol-042617-053055 16. Hayashi F, Means TK, Luster AD. **Toll-like receptors stimulate human neutrophil function**. *Blood* (2003) **102** 2660-2669. DOI: 10.1182/blood-2003-04-1078 17. Furusawa Y, Obata Y, Fukuda S, Endo TA, Nakato G, Takahashi D, Nakanishi Y, Uetake C, Kato K, Kato T. **Commensal microbe-derived butyrate induces the differentiation of colonic regulatory T cells**. *Nature* (2013) **504** 446-450. DOI: 10.1038/nature12721 18. Koh A, De Vadder F, Kovatcheva-Datchary P, Bäckhed F. **From dietary fiber to host physiology: short-chain fatty acids as key bacterial metabolites**. *Cell* (2016) **165** 1332-1345. DOI: 10.1016/j.cell.2016.05.041 19. Zhang D, Frenette PS. **Cross talk between neutrophils and the microbiota**. *Blood* (2019) **133** 2168-2177. DOI: 10.1182/blood-2018-11-844555 20. Kühl AA, Kakirman H, Janotta M, Dreher S, Cremer P, Pawlowski NN, Loddenkemper C, Heimesaat MM, Grollich K, Zeitz M. **Aggravation of different types of experimental colitis by depletion or adhesion blockade of neutrophils**. *Gastroenterology* (2007) **133** 1882-1892. DOI: 10.1053/j.gastro.2007.08.073 21. Fournier BM, Parkos CA. **The role of neutrophils during intestinal inflammation**. *Mucosal Immunol* (2012) **5** 354-366. DOI: 10.1038/mi.2012.24 22. Zhou F, Hamza T, Fleur AS, Zhang Y, Yu H, Chen K, Heath JE, Chen Y, Huang H, Feng H. **Mice with inflammatory bowel disease are susceptible to clostridium difficile infection with severe disease outcomes**. *Inflamm Bowel Dis* (2018) **24** 573-582. DOI: 10.1093/ibd/izx059 23. Voth E, Solanky D, Loftus EV, Pardi DS, Khanna S. **Novel risk factors and outcomes in inflammatory bowel disease patients with Clostridioides difficile infection**. *Therap Adv Gastroenterol* (2021) **14** 1756284821997792. DOI: 10.1177/1756284821997792 24. Yanai H, Nguyen GC, Yun L, Lebwohl O, Navaneethan U, Stone CD, Ghazi L, Moayyedi P, Brooks J, Bernstein CN. **Practice of gastroenterologists in treating flaring inflammatory bowel disease patients with clostridium difficile: antibiotics alone or combined antibiotics/immunomodulators?**. *Inflamm Bowel Dis* (2011) **17** 1540-1546. DOI: 10.1002/ibd.21514 25. Xavier RJ, Podolsky DK. **Unravelling the pathogenesis of inflammatory bowel disease**. *Nature* (2007) **448** 427-434. DOI: 10.1038/nature06005 26. Vacca M, Celano G, Calabrese FM, Portincasa P, Gobbetti M, De Angelis M. **The controversial role of human gut lachnospiraceae**. *Microorganisms* (2020) **8**. DOI: 10.3390/microorganisms8040573 27. Gu X, Sim JXY, Lee WL, Cui L, Chan YFZ, Chang ED, Teh YE, Zhang AN, Armas F, Chandra F. **Gut Ruminococcaceae levels at baseline correlate with risk of antibiotic-associated diarrhea**. *iScience* (2022) **25** 103644. DOI: 10.1016/j.isci.2021.103644 28. Lee YJ, Arguello ES, Jenq RR, Littmann E, Kim GJ, Miller LC, Ling L, Figueroa C, Robilotti E, Perales M-A. **Protective factors in the intestinal microbiome against clostridium difficile infection in recipients of allogeneic hematopoietic stem cell transplantation**. *J Infect Dis* (2017) **215** 1117-1123. DOI: 10.1093/infdis/jix011 29. Morrison DJ, Preston T. **Formation of short chain fatty acids by the gut microbiota and their impact on human metabolism**. *Gut Microbes* (2016) **7** 189-200. DOI: 10.1080/19490976.2015.1134082 30. Vincent C, Miller MA, Edens TJ, Mehrotra S, Dewar K, Manges AR. **Bloom and bust: intestinal microbiota dynamics in response to hospital exposures and Clostridium difficile colonization or infection**. *Microbiome* (2016) **4** 12. DOI: 10.1186/s40168-016-0156-3 31. Wang D, Dong D, Wang C, Cui Y, Jiang C, Ni Q, Su T, Wang G, Mao E, Peng Y. **Risk factors and intestinal microbiota: clostridioides difficile infection in patients receiving enteral nutrition at intensive care units**. *Crit Care* (2020) **24** 426. DOI: 10.1186/s13054-020-03119-7 32. Thomas F, Hehemann JH, Rebuffet E, Czjzek M, Michel G. **Environmental and gut bacteroidetes: the food connection**. *Front Microbiol* (2011) **2** 93. DOI: 10.3389/fmicb.2011.00093 33. Zafar H, Saier MH. **Gut bacteroides species in health and disease**. *Gut Microbes* (2021) **13** 1-20. DOI: 10.1080/19490976.2020.1848158 34. Brazil JC, Louis NA, Parkos CA. **The role of polymorphonuclear leukocyte trafficking in the perpetuation of inflammation during inflammatory bowel disease**. *Inflamm Bowel Dis* (2013) **19** 1556-1565. DOI: 10.1097/MIB.0b013e318281f54e 35. Magro F, Doherty G, Peyrin-Biroulet L, Svrcek M, Borralho P, Walsh A, Carneiro F, Rosini F, de Hertogh G, Biedermann L. **ECCO position paper: harmonization of the approach to ulcerative colitis histopathology**. *J Crohn's Colitis* (2020) **14** 1503-1511. DOI: 10.1093/ecco-jcc/jjaa110 36. Zhu F, He H, Fan L, Ma C, Xu Z, Xue Y, Wang Y, Zhang C, Zhou G. **Blockade of CXCR2 suppresses proinflammatory activities of neutrophils in ulcerative colitis**. *Am J Transl Res* (2020) **12** 5237-5251. PMID: 33042416 37. Yan Y, Kolachala V, Dalmasso G, Nguyen H, Laroui H, Sitaraman SV, Merlin D. **Temporal and spatial analysis of clinical and molecular parameters in dextran sodium sulfate induced colitis**. *PLoS One* (2009) **4** e6073. DOI: 10.1371/journal.pone.0006073 38. Zhou G, Yu L, Fang L, Yang W, Yu T, Miao Y, Chen M, Wu K, Chen F, Cong Y. **CD177 + neutrophils as functionally activated neutrophils negatively regulate IBD**. *Gut* (2018) **67** 1052-1063. DOI: 10.1136/gutjnl-2016-313535 39. Pereira C, Grácio D, Teixeira JP, Magro F. **Oxidative stress and DNA damage: implications in inflammatory bowel disease**. *Inflamm Bowel Dis* (2015) **21** 2403-2417. DOI: 10.1097/mib.0000000000000506 40. Mantovani A, Cassatella MA, Costantini C, Jaillon S. **Neutrophils in the activation and regulation of innate and adaptive immunity**. *Nat Rev Immunol* (2011) **11** 519-531. DOI: 10.1038/nri3024 41. Bersudsky M, Luski L, Fishman D, White RM, Ziv-Sokolovskaya N, Dotan S, Rider P, Kaplanov I, Aychek T, Dinarello CA. **Non-redundant properties of IL-1α and IL-1β during acute colon inflammation in mice**. *Gut* (2014) **63** 598-609. DOI: 10.1136/gutjnl-2012-303329 42. Shouval DS, Biswas A, Kang YH, Griffith AE, Konnikova L, Mascanfroni ID, Redhu NS, Frei SM, Field M, Doty AL. **Interleukin 1β mediates intestinal inflammation in mice and patients with interleukin 10 receptor deficiency**. *Gastroenterology* (2016) **151** 1100-1104. DOI: 10.1053/j.gastro.2016.08.055 43. Coccia M, Harrison OJ, Schiering C, Asquith MJ, Becher B, Powrie F, Maloy KJ. **IL-1β mediates chronic intestinal inflammation by promoting the accumulation of IL-17A secreting innate lymphoid cells and CD4+ Th17 cells**. *J Exp Med* (2012) **209** 1595-1609. DOI: 10.1084/jem.20111453 44. Ng J, Hirota SA, Gross O, Li Y, Ulke–Lemee A, Potentier MS, Schenck LP, Vilaysane A, Seamone ME, Feng H. **Clostridium difficile toxin–Induced inflammation and intestinal injury are mediated by the inflammasome**. *Gastroenterology* (2010) **139** 542-552. DOI: 10.1053/j.gastro.2010.04.005 45. Coll RC, Hill JR, Day CJ, Zamoshnikova A, Boucher D, Massey NL, Chitty JL, Fraser JA, Jennings MP, Robertson AAB. **MCC950 directly targets the NLRP3 ATP-hydrolysis motif for inflammasome inhibition**. *Nat Chem Biol* (2019) **15** 556-559. DOI: 10.1038/s41589-019-0277-7 46. Perera AP, Fernando R, Shinde T, Gundamaraju R, Southam B, Sohal SS, Robertson AAB, Schroder K, Kunde D, Eri R. **MCC950, a specific small molecule inhibitor of NLRP3 inflammasome attenuates colonic inflammation in spontaneous colitis mice**. *Sci Rep* (2018) **8** 8618. DOI: 10.1038/s41598-018-26775-w 47. Fenton MJ. **Review: transcriptional and post-transcriptional regulation of interleukin 1 gene expression**. *Int J Immunopharmacol* (1992) **14** 401-411. DOI: 10.1016/0192-0561(92)90170-p 48. Liu YH, Chang YC, Chen LK, Su P-A, Ko W-C, Tsai Y-S, Chen Y-H, Lai H-C, Wu C-Y, Hung Y-P. **The ATP-P2X(7) signaling axis is an essential sentinel for intracellular clostridium difficile pathogen-induced inflammasome activation**. *Front Cell Infect Microbiol* (2018) **8** 84. DOI: 10.3389/fcimb.2018.00084 49. Xu H, Yang J, Gao W, Li L, Li P, Zhang L, Gong YN, Peng X, Xi JJ, Chen S. **Innate immune sensing of bacterial modifications of Rho GTPases by the Pyrin inflammasome**. *Nature* (2014) **513** 237-241. DOI: 10.1038/nature13449 50. Yoshimura T, Matsushima K, Oppenheim JJ, Leonard EJ. **Neutrophil chemotactic factor produced by lipopolysaccharide (LPS)-stimulated human blood mononuclear leukocytes: partial characterization and separation from interleukin 1 (IL 1)**. *J Immunol* (1987) **139** 788-793. DOI: 10.4049/jimmunol.139.3.788 51. Pyrillou K, Burzynski LC, Clarke MCH. **Alternative pathways of IL-1 activation, and its role in health and disease**. *Front Immunol* (2020) **11** 613170. DOI: 10.3389/fimmu.2020.613170 52. Pavlidis P, Tsakmaki A, Pantazi E, Li K, Cozzetto D, Digby- Bell J, Yang F, Lo JW, Alberts E, Sa ACC. **Interleukin-22 regulates neutrophil recruitment in ulcerative colitis and is associated with resistance to ustekinumab therapy**. *Nat Commun* (2022) **13** 5820. DOI: 10.1038/s41467-022-33331-8 53. Dong D, Zhang L, Chen X, Jiang C, Yu B, Wang X, Peng Y. **Antimicrobial susceptibility and resistance mechanisms of clinical Clostridium difficile from a Chinese tertiary hospital**. *Int J Antimicrob Agents* (2013) **41** 80-84. DOI: 10.1016/j.ijantimicag.2012.08.011 54. Taghipour N, Molaei M, Mosaffa N, Rostami-Nejad M, Asadzadeh Aghdaei H, Anissian A, Azimzadeh P, Zali MR. **An experimental model of colitis induced by dextran sulfate sodium from acute progresses to chronicity in C57BL/6: correlation between conditions of mice and the environment**. *Gastroenterol Hepatol Bed Bench* (2016) **9** 45-52. PMID: 26744614 55. Fischer AH, Jacobson KA, Rose J, Zeller R. **Hematoxylin and eosin staining of tissue and cell sections**. *CSH Protoc* (2008) **2008**. DOI: 10.1101/pdb.prot4986 56. Dieleman LA, Palmen MJ, Akol H, Bloemena E, Peña AS, Meuwissen SGM, Van Rees EP. **Chronic experimental colitis induced by dextran sulphate sodium (DSS) is characterized by Th1 and Th2 cytokines**. *Clin Exp Immunol* (1998) **114** 385-391. DOI: 10.1046/j.1365-2249.1998.00728.x 57. Rc E. **UPARSE: highly accurate OTU sequences from microbial amplicon reads**. *Nat Methods* (2013) **10** 996-998. DOI: 10.1038/nmeth.2604 58. Wu J, Wei Z, Cheng P, Qian C, Xu F, Yang Y, Wang A, Chen W, Sun Z, Lu Y. **Rhein modulates host purine metabolism in intestine through gut microbiota and ameliorates experimental colitis**. *Theranostics* (2020) **10** 10665-10679. DOI: 10.7150/thno.43528 59. Swamydas M, Luo Y, Dorf ME, Lionakis MS. **Isolation of mouse neutrophils**. *Curr Protoc Immunol* (2015) **110** 20 1-3 20 15. DOI: 10.1002/0471142735.im0320s110
--- title: 'Evaluating the causal relationship between five modifiable factors and the risk of spinal stenosis: a multivariable Mendelian randomization analysis' authors: - Bangbei Wan - Ning Ma - Weiying Lu journal: PeerJ year: 2023 pmcid: PMC10038083 doi: 10.7717/peerj.15087 license: CC BY 4.0 --- # Evaluating the causal relationship between five modifiable factors and the risk of spinal stenosis: a multivariable Mendelian randomization analysis ## Abstract ### Background Spinal stenosis is a neurological disorder related to the compression of the spinal cord or nerve roots, and its incidence increases yearly. We aimed to use Mendelian randomization (MR) to investigate the causal relationship between several modifiable risk factors and the risk of spinal stenosis. ### Methods We obtained genome-wide association study summary data of large-sample projects (more than 100,000 individuals) from public databases. The data were associated with traits, including years of schooling (educational attainment) from the IEU OpenGWAS Project, smoking behavior (never vs. initiation) from the IEU OpenGWAS Project, body mass index (BMI) from the UK Biobank, length of mobile phone use from the UK Biobank, time spent watching television (TV) from the UK Biobank, and spinal stenosis from FinnGen biobank. Spinal stenosis was used as the outcome, whereas the other four traits were used as exposures. Inverse variance weighted (IVW) regressions were used as a primary to estimate the causal-effect size. Several sensitive analyses (including consistency, heterogenicity, and pleiotropy analyses) were conducted to test the stability and reliability of causal estimates. ### Results Univariable MR analyses showed that genetically predicted higher educational attainment (IVW; odds ratio (OR) = 0.606; $95\%$ confidence interval (CI): 0.507–0.724; $$P \leq 3.37$$ × 10−8) and never smoking (IVW; OR = 1.388; $95\%$ CI [1.135–1.697]; $$P \leq 0.001$$) were negatively correlated with the risk of spinal stenosis. Meanwhile, a higher BMI (IVW; OR = 1.569; $95\%$ CI [1.403–1.754]; $$P \leq 2.35$$ × 10−8), longer time spent using a mobile phone (IVW; OR = 1.895; $95\%$ CI [1.306–2.750]; $$P \leq 0.001$$), and watching TV (IVW; OR = 1.776; $95\%$ CI [1.245–2.532]; $$P \leq 0.002$$) were positively associated with the risk of spinal stenosis. Multivariable MR analysis indicated that educational attainment (IVW; OR = 0.670; $95\%$ CI [0.465–0.967]; $$P \leq 0.032$$) and BMI (IVW; OR = 1.365; $95\%$ CI [1.179–1.580]; $$P \leq 3.12$$ × 10−5) were independently causally related to the risk of spinal stenosis. ### Conclusion Our findings supported the potential causal associations of the five factors (educational attainment, smoking behavior, BMI, length of mobile phone use, and watching TV) with the risk for spinal stenosis. While replication studies are essential, these findings may provide a new perspective on prevention and intervention strategies directed toward spinal stenosis. ## Introduction Spinal stenosis is a common painful chronic disease caused by spinal-cord compression and seriously affects the quality of life of affected people (Jensen et al., 2021; Melancia, Francisco & Antunes, 2014). Spinal stenosis is divided into three types based on the narrow location, including lumbar, cervical, and thoracic spinal stenoses. The lumbar and cervical spinal stenoses are the most common types. Epidemiological investigations indicate that lumbar stenosis affects more than 103 million persons worldwide, especially older populations (Katz et al., 2022). Activity modification, analgesia, and physical therapy are the first-line treatments for spinal stenosis, but their curative effect is poor, and many patients require further therapy by surgery. Based on such harmful influence, early seeking out risk factors and implementing interventions are crucial to reducing the risk of spinal stenosis. Spinal stenosis is an intervertebral-disc degeneration disease, and its etiologies remain unclear. Existing evidence has suggested that some bad behaviors and diseases can increase the risk for spinal stenosis (Bagley et al., 2019; Karsenty et al., 1985). The length of mobile phone use and the time spent watching television are all critical indicators for indirectly reflecting sedentary behavior and physical inactivity. Importantly, sedentary behavior and physical inactivity are detrimental to public health and can induce some the disorders, such as obesity, diabetes, and cardiovascular disease (Arocha Rodulfo, 2019; Tomkins-Lane et al., 2013). Additionally, some observational studies have reported that overweight or obesity is positively associated with the risk of spinal stenosis and may trigger factors affecting spinal stenosis (Fortin et al., 2017; Hirano et al., 2013). Furthermore, a prospective cohort study with a large sample has shown an association of smoking behavior with the risk of spinal stenosis, as well as a dose correlation (Knutsson et al., 2018). The above studies demonstrate an association between several factors and the risk of spinal stenosis; however, causal inferences are lacking. Therefore, we propose a hypothesis that these factors may have a causal relationship with the risk of spinal stenosis. A dependable causality is advantageous and provides robust supporting evidence in drawing up and implementing public-health policies. Mendelian randomization (MR) is a newly emerging field that uses genetic variants as instrumental variables to investigate the causal relationship between exposure(s) and outcome(s) (Emdin, Khera & Kathiresan, 2017; Holmes, Ala-Korpela & Smith, 2017). In the present work, we investigated the causal relationship of several modifiable factors (years of schooling, smoking behavior (never vs. initiation), body mass index (BMI), length of mobile phone use, and time spent watching television (TV)) with the risk of spinal stenosis by using two-sample MR. The inverse variance weighted (IVW) algorithm was adopted as a primary computing method to assess the causal-effect size. Given the harmfulness of spinal stenosis to public health, determining causality between potential risk factors and spinal stenosis is essential to aid the establishment of prevention strategies. ## Study design and data sources The three critical basic assumptions and overall study design flowchart of the MR are displayed in Fig. 1. In this work, the exposures- and outcome-related genome-wide association study (GWAS) summary data of European populations were extracted from an available public database (the IEU Open GWAS database: https://gwas.mrcieu.ac.uk/) and used to perform MR analysis. The data details were as follows. Educational attainment (years of schooling, standard deviation (SD): 4.2 years)-related summary data with a GWAS sample size of 766 345 individuals were derived from the IEU OpenGWAS Project (Lee et al., 2018). BMI (SD: 4.75 kg/m2)-related genetic data were from a GWAS meta-analysis involving 681 275 individuals and extracted from the UK Biobank (Yengo et al., 2018). Smoking behavior (never vs. initiation)-related statistical data were from a GWAS meta-analysis with 311 629 cases and 321 173 controls and extracted from the IEU OpenGWAS Project (Liu et al., 2019). Using mobile (length of mobile phone use)-related summary-level data of 456 972 volunteers were derived from a GWAS analysis of the UK Biobank. Watching TV (time spent watching TV; SD: 1.62 h/day)-related summary genetic data of 437 887 individuals were also from a GWAS analysis of the UK Biobank. The spinal stenosis GWAS summary statistical data, which involved 9,169 spinal stenosis cases and 164 682 controls, were extracted from the FinnGen research project. **Figure 1:** *Flowchart of univariable and multivariable MR analyses.(A) Univariable MR. MR has three fundamental assumptions. (1) Relevance assumption: the genetic variants (instrumental variables) must be strongly correlated with exposure(s) (P < 5 × 10−8) (r2 < 0.001 and distance > 10 000 kb, the SNPs in pairwise linkage disequilibrium). (2) Independence assumption: no unmeasured confounders of the correlations existed between genetic variants and outcome(s). (3) Exclusion restriction assumption: the genetic variants influenced the outcome(s) only via exposure(s). (B) Multivariable MR: MR, Mendelian randomization; educational attainment, years of schooling; BMI, body mass index; smoking behavior, smoking behavior (never vs. initiation); using mobile phone, length of mobile phone use; watching TV, time spent watching television; and SNP, single nucleotide polymorphism.* ## Univariable and multivariable MR analyses Univariable inverse variance weighted (IVW) regression was used as a principal algorithm to estimate causal-effect size from exposure(s) to outcome. We identified single-nucleotide polymorphisms (SNPs) independently correlated with exposures (educational attainment, smoking behavior, BMI, using mobile, and watching TV) and used them as instrumental variables to assess the causal relationship between exposure and outcome. The independent SNPs were selected according to the following parameters: (a) $P \leq 5.0$ × 10−8 was regarded as a statistically significant threshold for a strong correlation between SNPs and exposure; and (b) r2 < 0.001 and distance >10,000 kb among SNPs in pairwise linkage disequilibrium (LD) were deemed the independent threshold. The F statistic was utilized to estimate the instrument strength and its computing method, as described in a previous study (Pierce, Ahsan & Vanderweele, 2011). An F statistic >10 was considered as having no weak instrument bias. Next, we conducted a series of sensitivity analyses to validate the robustness and reliability of the univariate MR analyses. The MR Steiger test was used to inspect the correctness of causal hypotheses in the MR analyses. Four methods including the MR–Egger (Bowden, Davey Smith & Burgess, 2015), maximum likelihood (Xue, Shen & Pan, 2021), MR–pleiotropy residual sum outlier (MR-PRESSO) (Verbanck et al., 2018), and robust adjusted profile score (MR-RAPS) (Zhao et al., 2020) were adopted to prove the consistency of causal hypothesis in IVW analysis. The statistical power of univariable MR analyses was computed using an available online tool (https://shiny.cnsgenomics.com/mRnd/) (Brion, Shakhbazov & Visscher, 2013). A power greater than $80\%$ was deemed as excellent statistical evidence. Cochran’s Q statistics in the IVW and MR–Egger models were used to assess the heterogeneity of SNPs. $P \leq 0.05$ was deemed to indicate significant heterogeneity. The MR–PRESSO, MR–Egger, and IVW approaches were utilized to identify and remove potential outliers that can cause underlying pleiotropy. MR–Egger regression was used to determine whether a potential pleiotropy existed in univariable MR analysis. The leave-one-out permutation method was used to examine whether an existing single SNP can alter the pooled effect of all SNPs in IVW analysis. The MR Steiger test was utilized to determine whether the causal assumption was correct. Finally, considering the importance of the five factors for the risk of spinal stenosis, we further included the five factors to conduct a multivariable MR analysis (Lu, Wan & Sun, 2022) to identify the independent exposure(s). All statistical analyses of MR were conducted using the TwoSampleMR (version 0.5.6) (Hemani, Tilling & Davey Smith, 2017) and MRPRESSO (version 1.0) (Verbanck et al., 2018) packages in R software (version 4.1.2; R Core Team, 2021). ## Univariable and multivariable MR analysis First, to clarify the potential causal relationship between each exposure (educational attainment, smoking behavior, BMI, mobile phone use, and watching TV) and outcome (spinal stenosis), we performed univariable MR analyses using IVW regression. The results from univariable MR analyses showed that a 1-SD increase in years of schooling was correlated with a $39.40\%$ reduction in the spinal stenosis risk (IVW; odds ratio (OR) = 0.606, $95\%$ confidence interval (CI) [0.507–0.724]; $$P \leq 3.37$$ ×10−8) (Fig. 2); Smoking was associated with a $38.30\%$ increase in the spinal stenosis risk (IVW; OR = 1.388; $95\%$ CI [1.135–1.697]; $$P \leq 0.001$$) (Fig. 2); A 1-SD increase in BMI was correlated with a $56.90\%$ rise in the risk of spinal stenosis (IVW; OR = 1.569; $95\%$ CI [1.403–1.754]; $$P \leq 2.35$$ ×10−15) (Fig. 2); Longer time spent watching TV (IVW; OR = 1.776; $95\%$ CI [1.245–2.532]; $$P \leq 0.002$$) and using mobile phone (IVW; OR = 1.895; $95\%$ CI [1.306–2.750]; $$P \leq 0.001$$) were significantly associated with $77.60\%$ and $89.50\%$ increase in the risk of spinal stenosis, respectively (Fig. 2). Results of univariable MR analyses showed no potential weak-instrument bias (all F statistics >10). Moreover, the power value of each analysis was almost $100\%$, indicating outstanding reliability (Fig. 2). All included SNPs are exhibited in Tables S1–S5. **Figure 2:** *Forest plot showing univariable MR analysis.OR, odds ratio; CI, confidence interval; P-het, P value for heterogeneity using Cochran Q test; P-intercept, P value for MR-Egger intercept; P-Steiger, P value for MR–Steiger test; IVW, inverse variance weighted; MR–PRESSO, Mendelian randomization–pleiotropy residual sum outlier; MR–RAPS, robust adjusted profile score; SNP, single-nucleotide polymorphism; educational attainment, years of schooling; BMI, body mass index; smoking behavior, smoking behavior (never vs. initiation); using mobile phone, length of mobile phone use; and watching TV, time spent watching television.* Sensitivity analyses were subsequently conducted to examine the stability and dependability of the univariable studies. Results from the four methods (MR–Egger, maximum likelihood, MR–PRESSO, and MR-RAPS) were almost similar to the IVW estimates (Fig. 3). Results of heterogeneity analyses indicated certain heterogeneities among the four univariable analyses (educational attainment, smoking behavior, BMI, and time spent watching TV) (Fig. 2). The heterogeneities may have originated from Mendel’s law of independent assortment rather than existing pleiotropy (Lewis & Simpson, 2022; Qi, 2009). The MR–Egger regressions indicated no unbalanced horizontal pleiotropy in the MR analyses (all P−intercept >0.05) (Fig. 2). The iterative leave-one-out test displayed no single SNP that influenced the univariable results (all $P \leq 0.01$) (Tables S6–S10). Results of the MR Steiger test showed that all causal assumptions were correct (all $P \leq 0.01$) (Fig. 2). **Figure 3:** *Scatter plots indicated the IVW regression direction tested by four methods.MR, Mendelian randomization; SNP, single-nucleotide polymorphism; IVW, inverse variance weighted; MR–PRESSO, Mendelian randomization–pleiotropy residual sum outlier; MR–RAPS, robust adjusted profile score; educational attainment, years of schooling; BMI, Body mass index; smoking behavior, smoking behavior (never vs. initiation); using mobile phone, length of mobile phone use; watching TV, time spent watching television. (Figure created by Zhi Zhou).* Finally, considering the importance of the five factors for the risk of spinal stenosis, we also conducted multivariable MR analysis to reduce the effect of confoundings and identify the independent exposure(s). We observed that educational attainment and BMI were independently associated with the risk of spinal stenosis in multivariable MR (Fig. 4). Results suggested a direct causal relationship between the two factors (educational attainment and BMI) and the risk of spinal stenosis. **Figure 4:** *Forest plot displaying multivariable MR analysis.OR, odds ratio; CI, confidence intervals; educational attainment, years of schooling; BMI body mass index; smoking behavior, smoking behavior (never vs. initiation); using mobile phone, length of mobile phone use; watching TV, time spent watching television.* ## Discussion We used GWAS summary-level data from large-sample studies to investigate the causal link between the five factors (educational attainment, smoking behavior, BMI, length of mobile phone use, and time spent watching TV) and the risk of spinal stenosis. We observed a causal association between the five factors and the risk of spinal stenosis; educational attainment and BMI were independently correlated with the risk of spinal stenosis. The three factors (smoking behavior, length of mobile phone use, and time spent watching TV) may impact the risk of spinal stenosis by regulating other factors, such as obesity and degenerative changes in the spine. Spinal stenosis is characterized by chronic back pain and occurs in aging populations (Austevoll et al., 2021). The people affected by spinal stenosis are gradually becoming younger, thereby seriously affecting their quality of life (Katz et al., 2022; Lurie & Tomkins-Lane, 2016). Therefore, ascertaining risk factors and mapping out effective public strategies to prevent spinal stenosis is necessary and urgent. Poor lifestyle behaviors are apparent risk factors for spinal stenosis, whereas changes in the underlying molecular mechanism are intrinsic pathogeny for spinal stenosis (Byvaltsev et al., 2022). Previous studies have reported that a higher educational level is a protective factor for many diseases, such as type 2 diabetes (Agardh et al., 2011), coronary heart disease (Falkstedt & Hemmingsson, 2011), osteoporosis (Ho, Chen & Woo, 2005), etc. Although a recent study has also shown a spinal-stenosis risk difference in the populations with different educational levels, the difference vanishes if controlling for other confoundings (age, gender, and BMI) (Yabuki et al., 2013). Similarly, our work indicated that educational attainment was negatively correlated with the risk of spinal stenosis in univariable MR analysis. Interestingly, when adjusting for the other four factors, educational attainment still independently influences the risk of spinal stenosis. Statistical evidence of several sensitivity analyses also strongly supported our findings. Based on the above findings, increasing educational attainment may be a very beneficial strategy to attenuate the risk of spinal stenosis. Smoking behavior and BMI were risk factors for spinal stenosis (Bagley et al., 2019). A recent study involving 331 941 individuals revealed that individuals with smoking behavior have a higher risk of spinal stenosis, and the risk increases with increased smoking dose (Knutsson et al., 2018). Likewise, our result supported a causal association between smoking and the risk of spinal stenosis. Smoking behavior increased the risk for spinal stenosis by $38.80\%$ in univariable MR analysis. Still, multivariable MR analysis showed that smoking was not an independent risk factor for spinal stenosis after correcting four factors (educational attainment, BMI, length of mobile phone use, and time spent watching TV). These results suggested that smoking could impact the risk of spinal stenosis by mediating other factors. BMI is also one of the risk factors for spinal stenosis. In a previous cohort study including 364 467 individuals, a higher BMI is found to be associated with a higher risk of spinal stenosis (Knutsson et al., 2015). In our work, the causal inference of univariable investigation showed that increased BMI was associated with a high risk of spinal stenosis. Additionally, the multivariable MR analysis result denoted a direct causal relationship between BMI and the risk of spinal stenosis. Collectively, these findings preliminarily indicated that losing weight and stopping smoking may be very helpful in preventing spinal stenosis. Using a mobile phone or watching TV for a long time is terrible behavior and closely related to public health (Foreman et al., 2021; Ghavamzadeh, Khalkhali & Alizadeh, 2013; Henschel, Gorczyca & Chomistek, 2020; Ikinci Keles & Uzun Sahin, 2021; Madhav, Sherchand & Sherchan, 2017). No literature has reported the relationship of using a mobile phone or watching TV for a long time with the risk of spinal stenosis. In the present work, we revealed for the first time an indirect causal relationship between the two modifiable factors (using a mobile phone and watching TV for a long time) and the risk of spinal stenosis. Our work has some strengths and limitations. One of the strengths was our use of GWAS summary-level data from recent extensive sample studies, and the included SNPs were more comprehensive. Another strength was our use of several sensitivity analyses including heterogeneity tests, pleiotropy, and robustness assessment to improve the results’ reliability and stability. Last but not least, our work makes up for the observational study deficiency that lacks causal inference. Existing limitations are also inevitable in our work. First, we used the GWAS data derived from European ancestry populations; whether the findings can ultimately be generalized to non-European ancestry populations remains unclear. Second, although our work primarily revealed the causal relationship of five modifiable risk factors with the risk of spinal stenosis, the causal effect’s underlying mechanisms remain unexplained. Third, data limitations mean we cannot achieve a stratified analysis for parameters such as gender and age. In conclusion, we used the extensive GWAS summary data to investigate the association of five factors (educational attainment, smoking behavior, BMI, length of mobile phone use, and time spent watching TV) with the risk of spinal stenosis via MR methods. We found that when correcting other confoundings, a direct causal relationship existed between two factors (educational attainment and BMI) and the risk of spinal stenosis, whereas an indirect causal correlation existed between the other three factors (smoking behavior, length of mobile phone use, and time spent watching TV) and the risk of spinal stenosis. These findings provided preliminary evidence to support the fact that elevating educational attainment and reducing BMI can help attenuate the risk of spinal stenosis. Moreover, we need to explore possible mediators before implementing interventions for these three factors (smoking behavior, length of mobile phone use, and time spent watching TV) to reduce the risk of spinal stenosis. Our findings provided insights into drawing up public policies to prevent spinal stenosis. ## References 1. Agardh E, Allebeck P, Hallqvist J, Moradi T, Sidorchuk A. **Type 2 diabetes incidence and socio-economic position: a systematic review and meta-analysis**. *International Journal of Epidemiology* (2011) **40** 804-818. DOI: 10.1093/ije/dyr029 2. Arocha Rodulfo JI. **Sedentary lifestyle a disease from xxi century**. *Clínica e Investigación en Arteriosclerosis* (2019) **31** 233-240. DOI: 10.1016/j.arteri.2019.04.004 3. Austevoll IM, Hermansen E, Fagerland MW, Storheim K, Brox JI, Solberg T, Rekeland F, Franssen E, Weber C, Brisby H, Grundnes O, Algaard KRH, Boker T, Banitalebi H, Indrekvam K, Hellum C. **Decompression with or without Fusion in Degenerative Lumbar Spondylolisthesis**. *New England Journal of Medicine* (2021) **385** 526-538. DOI: 10.1056/NEJMoa2100990 4. Bagley C, MacAllister M, Dosselman L, Moreno J, Aoun SG, El Ahmadieh TY. **Current concepts and recent advances in understanding and managing lumbar spine stenosis**. *F1000Research* (2019) **8** 137. DOI: 10.12688/f1000research.16082.1 5. Bowden J, Davey Smith G, Burgess S. **Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression**. *International Journal of Epidemiology* (2015) **44** 512-525. DOI: 10.1093/ije/dyv080 6. Brion MJ, Shakhbazov K, Visscher PM. **Calculating statistical power in Mendelian randomization studies**. *International Journal of Epidemiology* (2013) **42** 1497-1501. DOI: 10.1093/ije/dyt179 7. Byvaltsev VA, Kalinin AA, Hernandez PA, Shepelev VV, Pestryakov YY, Aliyev MA, Giers MB. **Molecular and genetic mechanisms of spinal stenosis formation: systematic review**. *International Journal of Molecular Sciences* (2022) **23** 13479. DOI: 10.3390/ijms232113479 8. Emdin CA, Khera AV, Kathiresan S. **Mendelian randomization**. *Journal of the American Medical Association* (2017) **318** 1925-1926. DOI: 10.1001/jama.2017.17219 9. Falkstedt D, Hemmingsson T. **Educational level and coronary heart disease: a study of potential confounding from factors in childhood and adolescence based on the Swedish 1,969 conscription cohort**. *Annals of Epidemiology* (2011) **21** 336-342. DOI: 10.1016/j.annepidem.2010.12.005 10. Foreman J, Salim AT, Praveen A, Fonseka D, Ting DSW, Guang He M, Bourne RRA, Crowston J, Wong TY, Dirani M. **Association between digital smart device use and myopia: a systematic review and meta-analysis**. *Lancet Digit Health* (2021) **3**. PMID: 34625399 11. Fortin M, Lazary A, Varga PP, Battie MC. **Association between paraspinal muscle morphology, clinical symptoms and functional status in patients with lumbar spinal stenosis**. *European Spine Journal* (2017) **26** 2543-2551. DOI: 10.1007/s00586-017-5228-y 12. Ghavamzadeh S, Khalkhali HR, Alizadeh M. **TV viewing, independent of physical activity and obesogenic foods, increases overweight and obesity in adolescents**. *Journal of Health, Population and Nutrition* (2013) **31** 334-342. DOI: 10.3329/jhpn.v31i3.16825 13. Hemani G, Tilling K, Davey Smith G. **Orienting the causal relationship between imprecisely measured traits using GWAS summary data**. *PLOS Genetics* (2017) **13**. DOI: 10.1371/journal.pgen.1007081 14. Henschel B, Gorczyca AM, Chomistek AK. **Time spent sitting as an independent risk factor for cardiovascular disease**. *American Journal of Lifestyle Medicine* (2020) **14** 204-215. DOI: 10.1177/1559827617728482 15. Hirano K, Imagama S, Hasegawa Y, Muramoto A, Ishiguro N. **Impact of spinal imbalance and BMI on lumbar spinal canal stenosis determined by a diagnostic support tool: cohort study in communityliving people**. *Archives of Orthopaedic and Trauma Surgery* (2013) **133** 1477-1482. DOI: 10.1007/s00402-013-1832-4 16. Ho SC, Chen YM, Woo JL. **Educational level and osteoporosis risk in postmenopausal Chinese women**. *American Journal of Epidemiology* (2005) **161** 680-690. DOI: 10.1093/aje/kwi047 17. Holmes MV, Ala-Korpela M, Smith GD. **Mendelian randomization in cardiometabolic disease: challenges in evaluating causality**. *Nature Reviews Cardiology* (2017) **14** 577-590. DOI: 10.1038/nrcardio.2017.78 18. Ikinci Keles A, Uzun Sahin C. **Exposure to electromagnetic field, cell phone use behaviors, SAR values, and changes in health following exposure in adolescent university students**. *Arquivos de Neuro-Psiuiatria* (2021) **79** 139-148. DOI: 10.1590/0004-282X-ANP-2020-0283 19. Jensen RK, Harhangi BS, Huygen F, Koes B. **Lumbar spinal stenosis**. *BMJ* (2021) **373** n1581. DOI: 10.1136/bmj.n1581 20. Karsenty G, Lacour B, Ulmann A, Pierandrei E, Drueke T. **Phosphate fluxes in isolated enterocytes from vitamin D replete and vitamin D deficient rats–early effects of calcitriol**. *Pflugers Archiv. European Journal of Physiology* (1985) **403** 151-155. DOI: 10.1007/BF00584093 21. Katz JN, Zimmerman ZE, Mass H, Makhni MC. **Diagnosis and management of lumbar spinal stenosis: a review**. *Journal of the American Medical Association* (2022) **327** 1688-1699. DOI: 10.1001/jama.2022.5921 22. Knutsson B, Mukka S, Wahlstrom J, Jarvholm B, Sayed-Noor AS. **The association between tobacco smoking and surgical intervention for lumbar spinal stenosis: a cohort study of 331,941 Workers**. *The Spine Journal* (2018) **18** 1313-1317. DOI: 10.1016/j.spinee.2017.11.018 23. Knutsson B, Sanden B, Sjoden G, Jarvholm B, Michaelsson K. **Body mass index and risk for clinical lumbar spinal stenosis: a cohort study**. *Spine* (2015) **40** 1451-1456. DOI: 10.1097/BRS.0000000000001038 24. Lee JJ, Wedow R, Okbay A, Kong E, Maghzian O, Zacher M, Nguyen-Viet TA, Bowers P, Sidorenko J, Karlsson Linner R, Fontana MA, Kundu T, Lee C, Li H, Li R, Royer R, Timshel PN, Walters RK, Willoughby EA, Yengo L, Alver M, Bao Y, Clark DW, Day FR, Furlotte NA, Joshi PK, Kemper KE, Kleinman A, Langenberg C, Magi R, Trampush JW, Verma SS, Wu Y, Lam M, Zhao JH, Zheng Z, Boardman JD, Campbell H, Freese J, Harris KM, Hayward C, Herd P, Kumari M, Lencz T, Luan J, Malhotra AK, Metspalu A, Milani L, Ong KK, Perry JRB, Porteous DJ, Ritchie MD, Smart MC, Smith BH, Tung JY, Wareham NJ, Wilson JF, Beauchamp JP, Conley DC, Esko T, Lehrer SF, Magnusson PKE, Oskarsson S, Pers TH, Robinson MR, Thom K, Watson C, Chabris CF, Meyer MN, Laibson DI, Yang J, Johannesson M, Koellinger PD, Turley P, Visscher PM, Benjamin DJ, Cesarini D. **Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals**. *Nature Genetics* (2018) **50** 1112-1121. DOI: 10.1038/s41588-018-0147-3 25. Lewis RG, Simpson B. **Genetics, autosomal dominant**. *StatPearls* (2022) 26. Liu M, Jiang Y, Wedow R, Li Y, Brazel DM, Chen F, Datta G, Davila-Velderrain J, McGuire D, Tian C, Zhan X, Psychiatry HA-I, Choquet H, Docherty AR, Faul JD, Foerster JR, Fritsche LG, Gabrielsen ME, Gordon SD, Haessler J, Hottenga JJ, Huang H, Jang SK, Jansen PR, Ling Y, Magi R, Matoba N, McMahon G, Mulas A, Orru V, Palviainen T, Pandit A, Reginsson GW, Skogholt AH, Smith JA, Taylor AE, Turman C, Willemsen G, Young H, Young KA, Zajac GJM, Zhao W, Zhou W, Bjornsdottir G, Boardman JD, Boehnke M, Boomsma DI, Chen C, Cucca F, Davies GE, Eaton CB, Ehringer MA, Esko T, Fiorillo E, Gillespie NA, Gudbjartsson DF, Haller T, Harris KM, Heath AC, Hewitt JK, Hickie IB, Hokanson JE, Hopfer CJ, Hunter DJ, Iacono WG, Johnson EO, Kamatani Y, Kardia SLR, Keller MC, Kellis M, Kooperberg C, Kraft P, Krauter KS, Laakso M, Lind PA, Loukola A, Lutz SM, Madden PAF, Martin NG, McGue M, McQueen MB, Medland SE, Metspalu A, Mohlke KL, Nielsen JB, Okada Y, Peters U, Polderman TJC, Posthuma D, Reiner AP, Rice JP, Rimm E, Rose RJ, Runarsdottir V, Stallings MC, Stancakova A, Stefansson H, Thai KK, Tindle HA, Tyrfingsson T, Wall TL, Weir DR, Weisner C, Whitfield JB, Winsvold BS, Yin J, Zuccolo L, Bierut LJ, Hveem K, Lee JJ, Munafo MR, Saccone NL, Willer CJ, Cornelis MC, David SP, Hinds DA, Jorgenson E, Kaprio J, Stitzel JA, Stefansson K, Thorgeirsson TE, Abecasis G, Liu DJ, Vrieze S. **Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use**. *Nature Genetics* (2019) **51** 237-244. DOI: 10.1038/s41588-018-0307-5 27. Lu L, Wan B, Sun M. **Mendelian randomization identifies age at menarche as an independent causal effect factor for gestational diabetes mellitus**. *Diabetes, Obesity and Metabolism* (2022) **25** 248-260. DOI: 10.1111/dom.14869 28. Lurie J, Tomkins-Lane C. **Management of lumbar spinal stenosis**. *BMJ* (2016) **352** h6234. DOI: 10.1136/bmj.h6234 29. Madhav KC, Sherchand SP, Sherchan S. **Association between screen time and depression among US adults**. *Preventive Medicine Reports* (2017) **8** 67-71. DOI: 10.1016/j.pmedr.2017.08.005 30. Melancia JL, Francisco AF, Antunes JL. **Spinal stenosis**. *Handbook of Clinical Neurology* (2014) **119** 541-549. DOI: 10.1016/B978-0-7020-4086-3.00035-7 31. Pierce BL, Ahsan H, Vanderweele TJ. **Power and instrument strength requirements for Mendelian randomization studies using multiple genetic variants**. *International Journal of Epidemiology* (2011) **40** 740-752. DOI: 10.1093/ije/dyq151 32. Qi L. **Mendelian randomization in nutritional epidemiology**. *Nutrition Reviews* (2009) **67** 439-450. DOI: 10.1111/j.1753-4887.2009.00218.x 33. R Core Team (2021) R Core Team R: a language and environment for statistical computing2021Version 4.1.2R Foundation for Statistical ComputingViennahttps://www.r-project.org. (2021) 34. Tomkins-Lane CC, Lafave LM, Parnell JA, Krishnamurthy A, Rempel J, Macedo LG, Moriartey S, Stuber KJ, Wilson PM, Hu R, Andreas YM. **The spinal stenosis pedometer and nutrition lifestyle intervention (SSPANLI) randomized controlled trial protocol**. *BMC Musculoskeletal Disorders* (2013) **14** 322. DOI: 10.1186/1471-2474-14-322 35. Verbanck M, Chen CY, Neale B, Do R. **Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases**. *Nature Genetics* (2018) **50** 693-698. DOI: 10.1038/s41588-018-0099-7 36. Xue H, Shen X, Pan W. **Constrained maximum likelihood-based Mendelian randomization robust to both correlated and uncorrelated pleiotropic effects**. *American Journal of Human Genetics* (2021) **108** 1251-1269. DOI: 10.1016/j.ajhg.2021.05.014 37. Yabuki S, Fukumori N, Takegami M, Onishi Y, Otani K, Sekiguchi M, Wakita T, Kikuchi S, Fukuhara S, Konno S. **Prevalence of lumbar spinal stenosis, using the diagnostic support tool, and correlated factors in Japan: a population-based study**. *Journal of Orthopaedic Science* (2013) **18** 893-900. DOI: 10.1007/s00776-013-0455-5 38. Yengo L, Sidorenko J, Kemper KE, Zheng Z, Wood AR, Weedon MN, Frayling TM, Hirschhorn J, Yang J, Visscher PM. **Meta-analysis of genome-wide association studies for height and body mass index in approximately 700000 individuals of European ancestry**. *Human Molecular Genetics* (2018) **27** 3641-3649. DOI: 10.1093/hmg/ddy271 39. Zhao Q, Wang J, Hemani G, Bowden J. **Statistical inference in two-sample summary-data Mendelian randomization using robust adjusted profile score**. *The Annals of Statistics* (2020) **48** 1742-1769. DOI: 10.1214/19-AOS1866
--- title: Identifying effects of genetic obesity exposure on leukocyte telomere length using Mendelian randomization authors: - Bangbei Wan - Ning Ma - Cai Lv journal: PeerJ year: 2023 pmcid: PMC10038084 doi: 10.7717/peerj.15085 license: CC BY 4.0 --- # Identifying effects of genetic obesity exposure on leukocyte telomere length using Mendelian randomization ## Abstract ### Background Observational studies have shown that obesity is closely associated with leukocyte telomere length (LTL). However, the causal relationship between obesity and LTL remains unclear. This study investigated the causal relationship between obesity and LTL through the Mendelian randomization approach. ### Materials and Methods The genome-wide association study (GWAS) summary data of several studies on obesity-related traits with a sample size of more than 600,000 individuals were extracted from the UK Biobank cohort. The summary-level data of LTL-related GWAS (45 6,717 individuals) was obtained from the IEU Open GWAS database. An inverse-variance-weighted (IVW) algorithm was utilized as the primary MR analysis method. Sensitivity analyses were conducted via MR-Egger regression, IVW regression, leave-one-out test, MR-pleiotropy residual sum, and outlier methods. ### Results High body mass index was correlated with a short LTL, and the odds ratio (OR) was 0.957 ($95\%$ confidence interval [CI] 0.942–0.973, $$p \leq 1.17$$E−07). The six body fat indexes (whole body fat mass, right leg fat mass, left leg fat mass, right arm fat mass, left arm fat mass, and trunk fat mass) were consistently inversely associated with LTL. Multiple statistical sensitive analysis approaches showed that the adverse effect of obesity on LTL was steady and dependable. ### Conclusion The current study provided robust evidence supporting the causal assumption that genetically caused obesity is negatively associated with LTL. The findings may facilitate the formulation of persistent strategies for maintaining LTL. ## Introduction Telomeres are DNA-protein complexes located on chromosome ends and play a critical role in guaranteeing the stability and integrality of chromosomes. An abnormal change in telomere length is closely related to human health. Leukocyte telomere length (LTL) is a quickly gauging indicator compared with other tissues. A previous study showed that LTL was closely related to the telomere length of tissues and served as an indicator that reflects telomere length in other tissues (Takubo et al., 2002). Mounting evidence has shown that extreme long LTL is strongly correlated with disease occurrence and development (Antwi & Petersen, 2018; Ayora et al., 2022; D’Mello et al., 2015). The findings of a meta-analysis suggested that short LTL is associated with a high risk of coronary heart disease (Haycock et al., 2014). Likewise, a recent meta-analysis reported that LTL is inversely related to the risk of atrial fibrillation, especially in men (Zheng et al., 2022). Furthermore, aberrant LTL is associated with a variety of diseases, such as major depressive disorder (Pisanu et al., 2020), female fertility (Michaeli et al., 2022), complications of type 2 diabetes mellitus (Testa et al., 2011), Barrett’s esophagus (Risques et al., 2007), and non-obstructive azoospermia (Yang et al., 2018). Hence, screening risk factors leading to the aberrant LTL in the early stage is vital to disease prevention. Obesity is a severe public health problem worldwide and the main triggering factor for diseases (Adom et al., 2017; Cheng et al., 2020; Wang et al., 2021). Substantial evidence has shown that obesity is associated with abnormal change in LTL (Buxton et al., 2011; Zhang et al., 2021). A negative correlation between obesity and LTL was reported in a recent meta-analysis (Zadeh et al., 2021), and the finding of a recent observational study has shown that obesity is inversely associated with LTL in Asian children (Ooi et al., 2021). Moreover, a meta-analysis including 63 studies and involving 119,439 populations revealed a negative correlation between obesity and LTL (Mundstock et al., 2015). Despite that considerable evidence suggests that obesity is negatively associated with LTL, a large heterogenicity has been found in different studies, and the causal relationship between obesity and LTL remains unclear. Therefore, we propose a hypothesis that obesity may be a potential causal risk factor to cause LTL changes. Mendelian randomization (MR) study is a powerful epidemiological approach for investigating the causality between exposure and outcome by using genetic variants as instrumental variables (Burgess, Butterworth & Thompson, 2013). Given that the allocation of genetic variants is a randomization process and is not affected by extraneous postnatal factors (Emdin, Khera & Kathiresan, 2017), MR controlling residual confounding factors is the same as a randomized controlled trial (RCT) (Davies, Holmes & Davey Smith, 2018; Ference, Holmes & Smith, 2021). In the present work, a two-sample MR study was conducted to investigate the causal relationship between obesity and LTL. The inverse-variance weighted (IVW) method was used as the primary analysis algorithm to assess the potential causation. We are seeing the perniciousness of obesity in human health. Clarifying obesity’s potential causal influence on LTL is beneficial in drawing up strategies for preventing diseases. As far as we know, this is the first MR study investigating the causality between obesity-related traits and LTL. Nevertheless, more molecular experiments are necessary to investigate further the potential mechanism for the effect of obesity on LTL. ## Study design The summary-level data of obesity-related traits, including body mass index (BMI), whole body fat mass, leg fat mass (right), leg fat mass (left), arm fat mass (right), arm fat mass (left), and trunk fat mass were extracted from the IEU Open GWAS database (https://gwas.mrcieu.ac.uk/). The summary statistical data of GWAS associated with LTL were also obtained from the IEU Open GWAS database. The data were from large sample studies and were used in conducting the two-sample MR investigation and exploring the causal relationship between obesity-related traits and LTL. ## Assumptions of MR study The fundamental assumptions and design of the MR investigation are shown in Fig. 1. [ 1] Relevance assumption: *The* genetic variants (instrumental variables) must be strongly correlated with obesity-related traits (exposure). [ 2] Independence assumption: no unpredictable confounders of the correlations between genetic variants and LTL (outcome) are present. [ 3] Exclusion restriction assumption: the genetic variants influence LTL only via obesity-related traits. ## Data sources The GWAS summary statistical data of obesity-related traits were obtained from European populations. The information included whole body fat mass with 330,762 participants, leg fat mass (right) with 331,293 populations, leg fat mass (left) with 331,275 populations, arm fat mass (right) with 331,226 individuals, arm fat mass (left) with 331,164 people, and trunk fat mass with 331,093 volunteers; all above obesity-related traits were obtained from UK Biobank cohort of the Neale lab. Single-nucleotide polymorphisms (SNPs) associated with obesity-related traits were identified as instrumental variables based on these parameters: $p \leq 5$ × 10−8 as a genome-wide statistical significance, independence among SNPs in linkage disequilibrium (r2 < 0.001; clump window, 10,000 kb). The F statistic was used in assessing the instrumental variables’ power in the MR and the computational method described by a previous study (Pierce, Ahsan & Vanderweele, 2011). F statistic > 10 was defined as the minimum required threshold. The GWAS summary data associated with LTL from European ancestry (472,174 individuals) were downloaded from the IEU Open GWAS database. All participants in the obesity-related trait research projects were not screened for the LTL cohort. ## Statistical analysis The IVW was used as a primary analysis algorithm in estimating the causal effect size of obesity-related traits on LTL (Burgess, Butterworth & Thompson, 2013). The MR-Egger (Bowden, Davey Smith & Burgess, 2015), maximum likelihood (Xue, Shen & Pan, 2021), MR-pleiotropy residual sum outlier (MR-PRESSO) (Verbanck et al., 2018), and robust adjusted profile score (MR-RAPS) (Zhao et al., 2020) algorithms were utilized in validating the reliability and robustness for the causal relationship between obesity traits and LTL. An available online tool (https://shiny.cnsgenomics.com/mRnd/) (Brion, Shakhbazov & Visscher, 2013) was used to calculate the statistical power of MR analysis, and a power greater than $80\%$ was deemed as a good value. The directionality that obesity-related traits cause LTL was verified via the MR Steiger test (Hemani, Tilling & Davey Smith, 2017). $p \leq 0.05$ indicated statistical significance. **Figure 1:** *Schematic diagram of MR investigating the causal relationship between obesity and LTL.The instrumental variable (IV) assumptions: (1) the IVs must be strongly associated with obesity-related traits (p < 5 × 10−8); (2) the IVs must not be correlated with any unmeasured confounders of obesity-related traits vs. LTL relationship; (3) the IVs should only affect the risk of LTL via obesity-related traits. SNPs, single-nucleotide polymorphisms; LTL, leukocyte telomere length; IVW, inverse-variance-weighted.* ## Sensitivity analysis The heterogeneity of SNPs was inspected via the IVW method and MR-Egger regression. MR-PRESSO, MR-Egger, and IVW approaches were used in identifying and removing outliers. The MR-Egger algorithms were utilized to detect potential pleiotropy. A single SNP’s influence on the total effect of IVW was evaluated through leave-one-out permutation analysis. All MR analyses were conducted using TwoSampleMR (version 0.5.6) and MRPRESSO packages in R software (version 4.1.2; R Core Team, 2021). ## Causality between BMI and LTL To understand the causal relationship between obesity and LTL preliminarily, we analyzed the effect of BMI on LTL by using the IVW method in the two-sample MR study. After outliers were removed, all 268 independent SNPs associated with BMI were included in the MR for the computation of the pool effect size of BMI on LTL (File S1). The result of the IVW approach showed that a genetically determined 1-SD increase in BMI was correlated with decreased LTL, and the odds ratio (OR) was 0.957 ($95\%$ confidence interval [CI] = 0.942–0.973, $$p \leq 1.17$$E−07; Table 1 and Fig. 2A). In the analysis, the F statistic of the SNPs was 61.0, and the statistical power was $100\%$, indicating the absence of weak-instrument bias and high credibility, respectively (Table 1). A series of sensitivity analysis approaches was used in verifying the robustness and reliability of the above result. First, as shown in Fig. 2B, four methods (MR-Egger, maximum likelihood, MR-PRESSO, and MR-RAPS) consistently exhibited causal direction from BMI to LTL. The result suggested that the causality from BMI to LTL was stable. Second, the results from Cochran’s Q test in the IVW model (p-het = 4.76E−07) and MR-Egger model (p-het = 4.48E−07) indicated heterogeneity among the SNPs, which may have been caused by the random allocation of alleles. Third, the statistical result of the MR-Egger intercepts indicated no directional pleiotropy in the MR analysis (p-intercept = 0.448). Fourth, leave-one-out analysis showed that no single SNP significantly influenced the causal association between BMI and LTL ($p \leq 0.05$; File S2). Finally, statistical evidence from the MR Steiger test suggested that BMI influencing LTL was a correct causal direction ($p \leq 0.001$; Table 1). ## Causality between six body fat indexes and LTL Previous studies reported some bias when only BMI was used as an indicator for measuring obesity and estimating the correlation between obesity and diseases (Nimptsch, Konigorski & Pischon, 2019; Piche, Tchernof & Despres, 2020; Vecchie et al., 2018) because BMI cannot reflect body fat distribution. Therefore, to avoid the bias of causality, six body fat indexes (whole body fat mass, right leg fat mass, left leg fat mass, right arm fat mass, left arm fat mass, and trunk fat mass) were used to confirm the casual assumption of obesity impact on LTL. After outliers were deleted, 247, 242, 245, 240, 236, and 247 independently available SNPs were associated with whole body fat mass, right leg fat mass, left leg fat mass, right arm fat mass, left arm fat mass, and trunk fat mass, respectively (File S1). The SNPs were then utilized in proving the genetically predicted causality between obesity and LTL. The result of the IVW algorithm indicated that enhanced body fat indexes are causally associated with a decrease in LTL. The effect size of six body fat indexes’ influences on LTL were as follows: whole body fat mass: OR = 0.949 ($95\%$ CI [0.932–0.967]; $$p \leq 3.69$$E−08), right leg fat mass: OR = 0.932 ($95\%$ CI [0.910–0.954]; $$p \leq 6.98$$E−09), left leg fat mass: OR = 0.931 ($95\%$ CI [0.910–0.952]; $$p \leq 3.79$$E−10), right arm fat mass: OR = 0.955 ($95\%$ CI [0.939–0.972]; $$p \leq 3.67$$E−07), left arm fat mass: OR = 0.949 ($95\%$ CI [0.932–0.965]; $$p \leq 4.63$$E−09), and trunk fat mass: OR = 0.964 ($95\%$ CI [0.947–0.981]; $$p \leq 5.53$$E−05; Table 1). In the analyses, the F statistics of the SNPs were larger than 10.0 (range: 58.8–61.1), indicating the absence of potential weak instrument bias, and all statistical power rates were approximately $100\%$, indicating high credibility (Table 1). Likewise, sensitivity analysis methods were used in testing the robustness and reliability of the above result. First, MR-Egger, maximum likelihood, MR-PRESSO, and MR-RAPS were utilized in validating the stability of the causal hypothesis. The results uniformly showed that the causality between per body fat index and LTL was stable (Figs. 3A–3F). We analyzed the heterogeneity of per body fat index-related SNPs by using Cochran’s Q test in the IVW model and the MR-Egger model. The results suggested heterogeneity among SNPs (all p-het < 0.05; Table 1). Furthermore, uncorrelated horizontal pleiotropy was detected via the MR-Egger method, and the result showed that the MR analyses had no uncorrelated horizontal pleiotropy (all p-intercept < 0.01; Table 1). Moreover, we used the iterative leave-one-out analysis method to determine whether a single SNP significantly modifies the pool effect value of the IVW. The result indicated that no single SNP significantly disrupted the combined effect of IVW (all $p \leq 0.05$; File S2). Finally, the causal hypothesis of the six body fat indexes’ influences on LTL was examined using the MR Steiger algorithm. The results suggested that the six body fat indexes affecting LTL was the correct causal direction (all $p \leq 0.001$). **Figure 3:** *Scatter plots of the correlation between six body fat indexes and LTL.(A) Whole body fat mass; (B) right leg fat mass; (C) left leg fat mass; (D) right arm fat mass; (E) left arm fat mass; (F) Trunk fat mass. MR, Mendelian randomization; SNP, single-nucleotide polymorphism; MR-PRESSO, MR-pleiotropy residual sum outlier; MR-RAPS, robust adjusted profile score.* ## Discussion In the present work, we first used the extensive sample GWAS summary data of European populations to investigate the causal relationship between obesity-related features and LTL comprehensively. Results show that genetically predicted higher BMI, whole body fat mass, right leg fat mass, left leg fat mass, right arm fat mass, left arm fat mass, and trunk fat mass were all causally associated with a lower LTL. These findings are the first to prove that obesity was a key causal risk factor that led to a decrease in LTL. Previous studies have shown that aberrant LTL can increase the risk of many diseases (Codd et al., 2013; Purdue-Smithe et al., 2021; Sanders & Newman, 2013). Thus, protecting LTL may be an essential way of preventing illness. Obesity is a highly prevalent disease and seriously harms people (Conway & Rene, 2004). Although substantial evidence supports that obesity is inversely correlated with LTL, the strength of the correlation is uneven, and the causality between obesity and LTL remains unexplained. This situation might be associated with the use of different indicators in measuring obesity. A meta-analysis included 16 original studies to investigate the correlation between BMI and LTL and found a negative association between BMI and LTL in adults (Muezzinler, Zaineddin & Brenner, 2014). Similarly, a cross-sectional study with 1000 participants found that BMI and LTL has a negative association (Muezzinler et al., 2016). A negative association between BMI and LTL was reported in a cross-sectional study with 35,096 individuals (Williams et al., 2016). A recent large meta-analysis including 87 studies and involving 146,114 individuals investigated the association between BMI and LTL in different age categories, gender, and ethnicity; a negative correlation between BMI and LTL was found, especially in younger participants (Gielen et al., 2018). In the current study, the standardized data of LTL was used to analyze the causal relationship between BMI and LTL and found BMI’s impact on LTL was a correct causal direction (OR = 0.957 [$95\%$ CI [0.942–0.973]], $$p \leq 1.17$$E−07). The finding initially illustrated the causal relationship between obesity and LTL. The statistical evidence of several sensitivity analyses confirmed that our result was stable and reliable. Despite that BMI remains to be a primary indicator for assessing obesity in clinical works, bias occurs when obesity is measured with BMI alone because BMI does not reflect body fat distribution (Antonopoulos et al., 2016). Body fat mass may be a better and more direct indicator for describing obesity (Kesztyus, Lampl & Kesztyus, 2021; Lee & Giovannucci, 2019). A previous observational study with a small sample (45 women) reported body fat mass was negatively correlated with LTL (Shin & Lee, 2016). Similarly, the finding of an observational study including 145 healthy term-born infants indicated that body fat mass was negatively associated with LTL (De Fluiter et al., 2021). Although these observational studies suggested a negative association between fat mass and LTL, the evidence was limited because they only used a single fat mass to estimate the causality between obesity and LTL. Therefore, to accurately assess the causal relationship between obesity and LTL, we used six body fat mass indexes (whole body fat mass, right leg fat mass, left leg fat mass, right arm fat mass, left arm fat mass, and trunk fat mass) to evaluate obesity’s influence on LTL. We found that the indexes were all inversely associated with LTL. These results again proved that the causal direction from obesity to LTL was correct. The present study overcomes the limitations of traditional observational studies that neither well control for unmeasured confounding nor prove causality between exposure and outcome (Davies, Holmes & Davey Smith, 2018). ## Limitations The current study possesses some shortcomings. First, obesity was divided into four types: metabolically healthy obesity, metabolically obese normal weight phenotype, normal weight obese syndrome, and sarcopenic obesity (Vecchie et al., 2018). Whether our result applies to all types is unknown. Second, we failed to achieve stratification analysis according to age and gender. Third, genetic variation is only one of the factors causing exposure changes, and many more remain to be further explored, such as environmental and epigenetic factors. Finally, given that the MR study was conducted in European ancestry, whether it can be popularized in non-European ancestry need to be investigated further. ## Conclusions In conclusion, using the extensive GWAS summary data, we implemented two-sample MR investigations to examine the causal relationship between BMI and LTL. We identified potential causal effects of several obesity-related traits on LTL. Our results suggested that genetically predicted obesity is inversely associated with LTL. In addition, given that the lifelong adverse effects of obesity on LTL are due to genetic variants, our findings may be useful in formulating persistent strategies for maintaining LTL and promoting health. ## References 1. Adom T, Puoane T, De Villiers A, Kengne AP. **Prevalence of obesity and overweight in African learners: a protocol for systematic review and meta-analysis**. *BMJ Open* (2017) **7**. DOI: 10.1136/bmjopen-2016-013538 2. Antonopoulos AS, Oikonomou EK, Antoniades C, Tousoulis D. **From the BMI paradox to the obesity paradox: the obesity-mortality association in coronary heart disease**. *Obesity Reviews* (2016) **17** 989-1000. DOI: 10.1111/obr.12440 3. Antwi SO, Petersen GM. **Leukocyte telomere length and pancreatic cancer risk: updated epidemiologic review**. *Pancreas* (2018) **47** 265-271. DOI: 10.1097/MPA.0000000000000995 4. Ayora M, Fraguas D, Abregu-Crespo R, Recio S, Blasco MA, Moises A, Derevyanko A, Arango C, Diaz-Caneja CM. **Leukocyte telomere length in patients with schizophrenia and related disorders: a meta-analysis of case-control studies**. *Molecular Psychiatry* (2022) **27** 2968-2975. DOI: 10.1038/s41380-022-01541-7 5. Bowden J, Davey Smith G, Burgess S. **Mendelian randomization with invalid instruments: effect estimation and bias detection through egger regression**. *International Journal of Epidemiology* (2015) **44** 512-525. DOI: 10.1093/ije/dyv080 6. Brion MJ, Shakhbazov K, Visscher PM. **Calculating statistical power in Mendelian randomization studies**. *International Journal of Epidemiology* (2013) **42** 1497-1501. DOI: 10.1093/ije/dyt179 7. Burgess S, Butterworth A, Thompson SG. **Mendelian randomization analysis with multiple genetic variants using summarized data**. *Genetic Epidemiology* (2013) **37** 658-665. DOI: 10.1002/gepi.21758 8. Buxton JL, Walters RG, Visvikis-Siest S, Meyre D, Froguel P, Blakemore AI. **Childhood obesity is associated with shorter leukocyte telomere length**. *The Journal of Clinical Endocrinology and Metabolism* (2011) **96** 1500-1505. DOI: 10.1210/jc.2010-2924 9. Cheng F, Luk AO, Tam CHT, Fan B, Wu H, Yang A, Lau ESH, Ng ACW, Lim CKP, Lee HM, Chow E, Kong AP, Keech AC, Joglekar MV, So WY, Jenkins AJ, Chan JCN, Hardikar AA, Ma RCW. **Shortened relative leukocyte telomere length is associated with prevalent and incident cardiovascular complications in type 2 diabetes: analysis from the Hong Kong diabetes register**. *Diabetes Care* (2020) **43** 2257-2265. DOI: 10.2337/dc20-0028 10. Codd V, Nelson CP, Albrecht E, Mangino M, Deelen J, Buxton JL, Hottenga JJ, Fischer K, Esko T, Surakka I, Broer L, Nyholt DR, Mateo Leach I, Salo P, Hagg S, Matthews MK, Palmen J, Norata GD, O’Reilly PF, Saleheen D, Amin N, Balmforth AJ, Beekman M, De Boer RA, Bohringer S, Braund PS, Burton PR, De Craen AJ, Denniff M, Dong Y, Douroudis K, Dubinina E, Eriksson JG, Garlaschelli K, Guo D, Hartikainen AL, Henders AK, Houwing-Duistermaat JJ, Kananen L, Karssen LC, Kettunen J, Klopp N, Lagou V, Van Leeuwen EM, Madden PA, Magi R, Magnusson PK, Mannisto S, McCarthy MI, Medland SE, Mihailov E, Montgomery GW, Oostra BA, Palotie A, Peters A, Pollard H, Pouta A, Prokopenko I, Ripatti S, Salomaa V, Suchiman HE, Valdes AM, Verweij N, Vinuela A, Wang X, Wichmann HE, Widen E, Willemsen G, Wright MJ, Xia K, Xiao X, Van Veldhuisen DJ, Catapano AL, Tobin MD, Hall AS, Blakemore AI, Van Gilst WH, Zhu H, consortium CA, Erdmann J, Reilly MP, Kathiresan S, Schunkert H, Talmud PJ, Pedersen NL, Perola M, Ouwehand W, Kaprio J, Martin NG, Van Duijn CM, Hovatta I, Gieger C, Metspalu A, Boomsma DI, Jarvelin MR, Slagboom PE, Thompson JR, Spector TD, Van der Harst P, Samani NJ. **Identification of seven loci affecting mean telomere length and their association with disease**. *Nature Genetics* (2013) **45** 422-427. DOI: 10.1038/ng.2528 11. Conway B, Rene A. **Obesity as a disease: no lightweight matter**. *Obesity Reviews* (2004) **5** 145-151. DOI: 10.1111/j.1467-789X.2004.00144.x 12. Davies NM, Holmes MV, Davey Smith G. **Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians**. *BMJ* (2018) **362** k601. DOI: 10.1136/bmj.k601 13. De Fluiter KS, Codd V, Denniff M, Kerkhof GF, Van Beijsterveldt I, Breij LM, Samani NJ, Abrahamse-Berkeveld M, Hokken-Koelega ACS. **Longitudinal telomere length and body composition in healthy term-born infants during the first two years of life**. *PLOS ONE* (2021) **16**. DOI: 10.1371/journal.pone.0246400 14. D’Mello MJ, Ross SA, Briel M, Anand SS, Gerstein H, Pare G. **Association between shortened leukocyte telomere length and cardiometabolic outcomes: systematic review and meta-analysis**. *Circulation-Cardiovascular Genetics* (2015) **8** 82-90. DOI: 10.1161/CIRCGENETICS.113.000485 15. Emdin CA, Khera AV, Kathiresan S. **Mendelian randomization**. *JAMA* (2017) **318** 1925-1926. DOI: 10.1001/jama.2017.17219 16. Ference BA, Holmes MV, Smith GD. **Using mendelian randomization to improve the design of randomized trials**. *Cold Spring Harbor Perspectives in Medicine* (2021) **11** a040980. DOI: 10.1101/cshperspect.a040980 17. Gielen M, Hageman GJ, Antoniou EE, Nordfjall K, Mangino M, Balasubramanyam M, De Meyer T, Hendricks AE, Giltay EJ, Hunt SC, Nettleton JA, Salpea KD, Diaz VA, Farzaneh-Far R, Atzmon G, Harris SE, Hou L, Gilley D, Hovatta I, Kark JD, Nassar H, Kurz DJ, Mather KA, Willeit P, Zheng YL, Pavanello S, Demerath EW, Rode L, Bunout D, Steptoe A, Boardman L, Marti A, Needham B, Zheng W, Ramsey-Goldman R, Pellatt AJ, Kaprio J, Hofmann JN, Gieger C, Paolisso G, Hjelmborg JBH, Mirabello L, Seeman T, Wong J, Van der Harst P, Broer L, Kronenberg F, Kollerits B, Strandberg T, Eisenberg DTA, Duggan C, Verhoeven JE, Schaakxs R, Zannolli R, Reis RMRDos, Charchar FJ, Tomaszewski M, Mons U, Demuth I, Iglesias Molli AE, Cheng G, Krasnienkov D, D’Antono B, Kasielski M, McDonnell BJ, Ebstein RP, Sundquist K, Pare G, Chong M, Zeegers MP, Group T. **Body mass index is negatively associated with telomere length: a collaborative cross-sectional meta-analysis of 87 observational studies**. *The American Journal of Clinical Nutrition* (2018) **108** 453-475. DOI: 10.1093/ajcn/nqy107 18. Haycock PC, Heydon EE, Kaptoge S, Butterworth AS, Thompson A, Willeit P. **Leucocyte telomere length and risk of cardiovascular disease: systematic review and meta-analysis**. *BMJ* (2014) **349** g4227. DOI: 10.1136/bmj.g4227 19. Hemani G, Tilling K, Davey Smith G. **Orienting the causal relationship between imprecisely measured traits using GWAS summary data**. *PLOS Genetics* (2017) **13**. DOI: 10.1371/journal.pgen.1007081 20. Kesztyus D, Lampl J, Kesztyus T. **The weight problem: overview of the most common concepts for body mass and fat distribution and critical consideration of their usefulness for risk assessment and practice**. *International Journal of Environmental Research and Public Health* (2021) **18** 11070. DOI: 10.3390/ijerph182111070 21. Lee DH, Giovannucci EL. **The obesity paradox in cancer: epidemiologic insights and perspectives**. *Current Nutrition Reports* (2019) **8** 175-181. DOI: 10.1007/s13668-019-00280-6 22. Michaeli J, Smoom R, Serruya N, El Ayoubi H, Rotshenker-Olshinka K, Srebnik N, Michaeli O, Eldar-Geva T, Tzfati Y. **Leukocyte telomere length correlates with extended female fertility**. *Cells* (2022) **11** 513. DOI: 10.3390/cells11030513 23. Muezzinler A, Mons U, Dieffenbach AK, Butterbach K, Saum KU, Schick M, Stammer H, Boukamp P, Holleczek B, Stegmaier C, Brenner H. **Body mass index and leukocyte telomere length dynamics among older adults: results from the ESTHER cohort**. *Experimental Gerontology* (2016) **74** 1-8. DOI: 10.1016/j.exger.2015.11.019 24. Muezzinler A, Zaineddin AK, Brenner H. **Body mass index and leukocyte telomere length in adults: a systematic review and meta-analysis**. *Obesity Reviews* (2014) **15** 192-201. DOI: 10.1111/obr.12126 25. Mundstock E, Sarria EE, Zatti H, Mattos Louzada F, Kich Grun L, Herbert Jones M, Guma FT, Mazzola In Memoriam J, Epifanio M, Stein RT, Barbe-Tuana FM, Mattiello R. **Effect of obesity on telomere length: systematic review and meta-analysis**. *Obesity* (2015) **23** 2165-2174. DOI: 10.1002/oby.21183 26. Nimptsch K, Konigorski S, Pischon T. **Diagnosis of obesity and use of obesity biomarkers in science and clinical medicine**. *Metabolism* (2019) **92** 61-70. DOI: 10.1016/j.metabol.2018.12.006 27. Ooi DSQ, Dorajoo R, Gurung RL, Dehghan R, Lim YY, Ho CWL, Tay V, Karuppiah V, Loke KY, Lim SC, Liu JJ, Sng AA, Lee YS. **Association of leukocyte telomere length with obesity-related traits in Asian children with early-onset obesity**. *Pediatric Obesity* (2021) **16**. DOI: 10.1111/ijpo.12771 28. Piche ME, Tchernof A, Despres JP. **Obesity phenotypes, diabetes, and cardiovascular diseases**. *Circulation Research* (2020) **126** 1477-1500. DOI: 10.1161/CIRCRESAHA.120.316101 29. Pierce BL, Ahsan H, Vanderweele TJ. **Power and instrument strength requirements for Mendelian randomization studies using multiple genetic variants**. *International Journal of Epidemiology* (2011) **40** 740-752. DOI: 10.1093/ije/dyq151 30. Pisanu C, Tsermpini EE, Skokou M, Kordou Z, Gourzis P, Assimakopoulos K, Congiu D, Meloni A, Balasopoulos D, Patrinos GP, Squassina A. **Leukocyte telomere length is reduced in patients with major depressive disorder**. *Drug Development Research* (2020) **81** 268-273. DOI: 10.1002/ddr.21612 31. Purdue-Smithe AC, Kim K, Andriessen VC, Pollack AZ, Sjaarda LA, Silver RM, Schisterman EF, Mumford SL. **Preconception leukocyte telomere length and pregnancy outcomes among women with demonstrated fecundity**. *Human Reproduction* (2021) **36** 3122-3130. DOI: 10.1093/humrep/deab201 32. R Core Team (2021) R Core Team R: a language and environment for statistical computing2021Version 4.1.2ViennaR Foundation for Statistical Computinghttps://www.r-project.org. (2021) 33. Risques RA, Vaughan TL, Li X, Odze RD, Blount PL, Ayub K, Gallaher JL, Reid BJ, Rabinovitch PS. **Leukocyte telomere length predicts cancer risk in Barrett’s esophagus**. *Cancer Epidemiology, Biomarkers & Prevention* (2007) **16** 2649-2655. DOI: 10.1158/1055-9965.EPI-07-0624 34. Sanders JL, Newman AB. **Telomere length in epidemiology: a biomarker of aging, age-related disease, both, or neither?**. *Epidemiologic Reviews* (2013) **35** 112-131. DOI: 10.1093/epirev/mxs008 35. Shin YA, Lee KY. **Low estrogen levels and obesity are associated with shorter telomere lengths in pre- and postmenopausal women**. *Journal of Exercise Rehabilitation* (2016) **12** 238-246. DOI: 10.12965/jer.1632584.292 36. Takubo K, Izumiyama-Shimomura N, Honma N, Sawabe M, Arai T, Kato M, Oshimura M, Nakamura K. **Telomere lengths are characteristic in each human individual**. *Experimental Gerontology* (2002) **37** 523-531. DOI: 10.1016/s0531-5565(01)00218-2 37. Testa R, Olivieri F, Sirolla C, Spazzafumo L, Rippo MR, Marra M, Bonfigli AR, Ceriello A, Antonicelli R, Franceschi C, Castellucci C, Testa I, Procopio AD. **Leukocyte telomere length is associated with complications of type 2 diabetes mellitus**. *Diabetic Medicine* (2011) **28** 1388-1394. DOI: 10.1111/j.1464-5491.2011.03370.x 38. Vecchie A, Dallegri F, Carbone F, Bonaventura A, Liberale L, Portincasa P, Fruhbeck G, Montecucco F. **Obesity phenotypes and their paradoxical association with cardiovascular diseases**. *European Journal of Internal Medicine* (2018) **48** 6-17. DOI: 10.1016/j.ejim.2017.10.020 39. Verbanck M, Chen CY, Neale B, Do R. **Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases**. *Nature Genetics* (2018) **50** 693-698. DOI: 10.1038/s41588-018-0099-7 40. Wang Y, Zhao L, Gao L, Pan A, Xue H. **Health policy and public health implications of obesity in China**. *The Lancet Diabetes & Endocrinology* (2021) **9** 446-461. DOI: 10.1016/S2213-8587(21)00118-2 41. Williams DM, Palaniswamy S, Sebert S, Buxton JL, Blakemore AI, Hypponen E, Jarvelin MR. **25-Hydroxyvitamin D concentration and leukocyte telomere length in young adults: findings from the Northern Finland birth cohort 1966**. *American Journal of Epidemiology* (2016) **183** 191-198. DOI: 10.1093/aje/kwv203 42. Xue H, Shen X, Pan W. **Constrained maximum likelihood-based Mendelian randomization robust to both correlated and uncorrelated pleiotropic effects**. *American Journal of Human Genetics* (2021) **108** 1251-1269. DOI: 10.1016/j.ajhg.2021.05.014 43. Yang Q, Luo X, Bai R, Zhao F, Dai S, Li F, Zhu J, Liu J, Niu W, Sun Y. **Shorter leukocyte telomere length is associated with risk of nonobstructive azoospermia**. *Fertility and Sterility* (2018) **110** 648-654 e641. DOI: 10.1016/j.fertnstert.2018.05.008 44. Zadeh FA, Bokov DO, Yasin G, Vahdat S, Abbasalizad-Farhangi M. **Central obesity accelerates leukocyte telomere length (LTL) shortening in apparently healthy adults: a systematic review and meta-analysis**. *Critical Reviews in Food Science and Nutrition* (2021) 1-10. DOI: 10.1080/10408398.2021.1971155 45. Zhang Y, Xu Z, Yang Y, Cao S, Lyu S, Duan W. **Association between weight change and leukocyte telomere length in US adults**. *Frontiers in Endocrinology* (2021) **12** 650988. DOI: 10.3389/fendo.2021.650988 46. Zhao Q, Wang J, Hemani G, Bowden J, Small DSJTAoS. **Statistical inference in two-sample summary-data Mendelian randomization using robust adjusted profile score**. *Annals of Statistics* (2020) **48** 1742-1769. DOI: 10.1214/19-AOS1866 47. Zheng Y, Zhang N, Wang Y, Wang F, Li G, Tse G, Liu T. **Association between leucocyte telomere length and the risk of atrial fibrillation: an updated systematic review and meta-analysis**. *Ageing Research Reviews* (2022) **81** 101707. DOI: 10.1016/j.arr.2022.101707
--- title: Using Patient Profiles for Sustained Diabetes Management Among People With Type 2 Diabetes authors: - Shang-Jyh Chiou - Yen-Jung Chang - Chih-Dao Chen - Kuomeng Liao - Tung-Sung Tseng journal: Preventing Chronic Disease year: 2023 pmcid: PMC10038094 doi: 10.5888/pcd20.220210 license: CC BY 4.0 --- # Using Patient Profiles for Sustained Diabetes Management Among People With Type 2 Diabetes ## Abstract ### Introduction Our objective was to evaluate the association between patient profiles and sustained diabetes management (SDM) among patients with type 2 diabetes. ### Methods We collected HbA1c values recorded from 2014 through 2020 for 570 patients in a hospital in Taipei, Taiwan, and calculated a standard level based on an HbA1c level less than $7.0\%$ to determine SDM. We used patients’ self-reported data on diabetes self-care behaviors to construct profiles. We used 8 survey items to perform a latent profile analysis with 3 groups (poor management, medication adherence, and good management). After adjusting for other determining factors, we used multiple regression analysis to explore the relationship between patient profiles and SDM. ### Results The good management group demonstrated better SDM than the poor management group (β = 0.183; $$P \leq .003$$). Using the most recent HbA1c value and the 7-year average of HbA1c values as the outcome, we found lower HbA1c values in the good management group than in the poor management group (β = −0.216 [$$P \leq .01$$] and −0.217 [$$P \leq .008$$], respectively). ### Conclusion By using patient profiles, we confirmed a positive relationship between optimal patient behavior in self-care management and SDM. Patients with type 2 diabetes exhibited effective self-care management behavior and engaged in more health care activities, which may have led to better SDM. In promoting patient-centered care, using patient profiles and customized health education materials could improve diabetes care. ## What is already known on this topic? Successful diabetes management relies on optimal and acceptable patient behavior. Self-efficacy and self-management are essential factors in diabetes-related health behavior. ## What is added by this report? We used the HbA1c level of less than $7.0\%$ to assess the outcome of diabetes control and sustained diabetes management. Latent profile analysis is a novel approach for conceptualizing patient profiles and assessing patient behavior in diabetes control. ## What are the implications for public health practice? Using patient health profiles can make medical teams aware of patients’ diabetes care and provide incentives for better patient behaviors. Better behaviors lead to patients’ optimal adherence to diabetes care, and subsequently, better health outcomes. ## Introduction In the US and around the world, diabetes is a serious public health issue. This metabolic disease is the leading cause of blindness, kidney failure, myocardial infarction, and stroke (1–3). Effective patient self-management of diabetes includes not only working with a medical team but also performing self-care behaviors. In clinical settings, physicians ideally need to consider many clinical or behavioral aspects of diabetes care, such as obesity, comorbidities, age, race, sex, gender, disease duration, life expectancy, and quality of life, to make decisions about treatment [4,5]. In reality, clinical practitioners and their patients with diabetes can best manage only a few factors. Using one-size-fits-all guidelines in diabetes care may have limited effects [6,7]. Furthermore, in diabetes care, patients who practice self-care management behaviors, which incorporate the core values of self-efficacy and self-regulation (8–10), may improve glycemic control through adherence to the 4 major components of diabetes control: medication adherence, diet, exercise, and the self-monitoring of blood glucose (SMBG). Adequate patient behavior in diabetes care is crucial to reaching optimal health outcomes, and integrated approaches are necessary for a long-lasting effect [11]. Diabetes management is enhanced through behaviors such as exercise, smoking cessation, and eating a healthy diet. Diabetes is a lifetime disease; when patients follow behavioral guidelines inconsistently, they rarely reach optimal goals [12]. Integrating diabetes care into their daily lives is often challenging; therefore, patients’ undesirable behaviors are not likely to improve, and appropriate behaviors are unlikely to occur and be maintained over time [13,14]. In addition, comprehensive diabetes management plans include factors related to the disease process and patient-management behaviors. To experience optimal quality of life, patients with type 2 diabetes must control their condition by following guidelines on medication adherence, healthy diet, regular physical activity, and blood glucose monitoring, which lowers the risk of macro or micro complications [15,16]. Hence, the American Diabetes Association emphasizes the importance of self-care management [17]. Therefore, an approach that involves behavioral science to understand patient behavior for alleviating barriers and psychological conditions is imperative in diabetes care. Some patients can sustain the 4 major components of diabetes control; however, some may only perform efficiently in certain tasks. We are particularly interested in whether different patient subgroups have different levels of risks of future complications. We hypothesized that information from patient profiles classified by multiple behaviors may not only identify groups at high risk of complications but may also be used to customize health education materials. In this patient-centered era, the use of patient profiles and customization of health education materials may be the most efficient way to help patients improve their behaviors. Traditionally, a variable-oriented approach, such as a regression model, has been used to address the relationship between HbA1c levels and patient behavior–related factors. However, this method cannot easily handle multiple factors simultaneously, thus creating potential for type I errors [18]. Additionally, researchers have had to make complicated decisions, such as how many variables to include in a model. To address these challenges, we used latent profile analysis (LPA), a participant-oriented approach that provides rich yet concise information needed to determine a patient’s effort in diabetes control. Physicians may concentrate only on the most recent HbA1c level to categorize a patient’s diabetes control and decide on treatment. For the diabetes control proxy, we used sustained diabetes management (SDM) from the long-term standard HbA1c level. Using patient profiles and SDM is a novel way of evaluating diabetes management, especially an approach that involves behavioral science, wherein components of profiles in diabetes care are constructed. LPA is an ideal technique to address complex conceptualization (such as patient profiles) for developing typologies based on data [19]. The use of patient profiles could help clinical practitioners be aware of patient behaviors in diabetes care; it could also enable health authorities to provide incentives that encourage improved patient behaviors, thus leading to optimal adherence to diabetes care and enhanced health outcomes. Hence, our study aimed to explore the associations between patient profiles and SDM in patients with type 2 diabetes. ## Methods This study was conducted in the Department of Metabolism at a regional hospital in northern Taipei from November 2019 through May 2020. Trained staff members used structured questionnaires to conduct face-to-face interviews of patients with type 2 diabetes after their medical consultations. Before the interview, all patients provided written informed consent that included use of their biomarker data (HbA1c) from the hospital’s health information system. All patients participating in the survey had type 2 diabetes and answered the questionnaire voluntarily; patients with mental disorders or cognitive impairments or who were unable to provide informed consent or participate in the survey were excluded. We included 570 patients in the analysis. After the survey, we compensated participants with an NT$100 gift card. The institutional review board of National Taiwan University and the hospital approved this study. ## Dependent variable: SDM We obtained the 2014–2020 medical records from the hospital’s health information system of participants who provided written consent. At their physician’s office for a scheduled appointment, participants had an HbA1c test. Each patient had multiple medical records for HbA1c values from 2014 to 2020. We calculated the standard level based on HbA1c criteria (<$7.0\%$). We used the following equation to calculate SDM: SDM = number of HbA1c measurements less than $7.0\%$ divided by the total number of HbA1c events. For example, 1 patient had HbA1c values of 7.1, 7.0, 6.9, 7.1, and 6.7; two of the 5 values were considered standard because they were less than $7.0\%$. The number of standard values ($$n = 2$$) was divided by the number of data points ($$n = 5$$) to obtain the SDM: $\frac{2}{5}$ = 0.4. Therefore, 0.4 indicated SDM during the study period. ## Independent variable The questionnaire asked about such characteristics as age, sex, education level, diabetes diagnosis date, and diabetes care–associated variables. Such variables included the self-reported level of health education received by the participants from the medical staff (1–10, with 10 being the highest level); the diabetes management self-efficacy scale in Chinese (DMSES–C); the treatment self-regulation questionnaire on diabetes (TSRQd); a self-report assessment of the self-management of diabetes control pertaining to medication, healthy diet, SMBG, and regular exercise (scale of 1 to 5, with 5 being the highest level of control for all 4 measures); and 2 self-reported items on health status. The original DMSES–C and TSRQd have 20 and 19 items, respectively. They were translated from English into Chinese with acceptable validity and reliability. After consulting with the translation team and experts in using the DMSES-C and TSRQd, we used shorter versions of the DMSES-C (shortened to 11 items) and the TSRQd (shortened to 15 items), which excluded some items from the 2 questionnaires as described elsewhere [20]. On the basis of the original design [21,22], the user can categorize the TSRQd into 2 dimensions, namely autonomous regulatory style (which involves conducting an activity for the enjoyment inherent in engaging in the behavior itself) and controlled regulatory style (which includes behavior motivated by contingencies not inherent to the activity itself), which we defined as TSRQd–A and TSRQd–C, respectively. The questionnaire asked patients the following health status–related question: “As compared to the past 12 months, how would you evaluate your current health status: better, neutral, or worse?” We also scored their self-reported health status from 0 to 100, with 100 indicating the best. ## Statistical analysis In the last 2 decades, LPA has been used extensively in the social sciences; additionally, it has been applied in the medical field to cluster individuals into subgroups and unveil hidden patterns of association, such as different risk groups or social support levels. We used 8 survey items on patients’ self-report of diabetes self-care behaviors for constructing patient profiles in the LPA. Eight items included scores for self-assessment in health education; medication self-management, healthy diet, SMBG, regular exercise; self-efficacy (DMSES–C), and self-regulation evaluation (TSRQd–A and TSRQd–C). After comparing Akaike information criterion, Bayesian information criterion, and entropy at 3, 4, and 5 class levels (Table 1), we decided to use the following 3 groups in the LPA model: poor management, medication adherence, and good management (Figure). Of the 570 patients in our sample, $6\%$ ($$n = 35$$) were in the poor management group, $21\%$ ($$n = 117$$) were in the medication adherence group, and $73\%$ ($$n = 418$$) were in the good management group (Table 2). The good management group had the highest mean values for all survey items. In contrast, the medication adherence group had the lowest mean values for all survey items, except medication self-management. The poor management group self-reported a slightly higher mean value than the medication adherence group for most items. We summarized information on patient profiles derived from the LPA. We used χ2 tests and 1-way analysis of variance to initially evaluate patient demographic data and the distribution of survey item responses by LPA subgroup. We used multiple regression analysis to estimate the likelihood of SDM by LPA subgroup with the other determining factors (Model 1). Considering that medical personnel were familiar with the HbA1c value as a marker for diabetes control, we used the most recent HbA1c value (Model 2) and the 7-year average of HbA1c values (Model 3) to conduct the regression model again for a sensitivity analysis. Subsequently, with the same determining factors, we compared the results of Models 2 and 3 with those of Model 1. Education, health status, and the LPA subgroups were transformed into dummy variables in the regression models. We used SAS version 9.3.1 (SAS Institute, Inc) and SPSS 20.0 (IBM Corporation) to analyze all data. The significance level was set at.05. ## Results The good management group had the oldest mean age (63.1 y), the highest scores for health status (76.9 vs 70.1 [poor management group] and 69.5 [medication adherence group]), the longest diabetes duration (11.9 y vs 9.3 y [poor management group] and 10.3 y [medication adherence group]), and the highest proportion of patients with a standard HbA1c rate of ≥0.5 ($70.1\%$ vs $51.3\%$ [poor management group] and $55.7\%$ [medication adherence group]) (Table 3). Moreover, the good management group had better behaviors for diabetes control (Table 2) than the other 2 groups. Differences in sex, education level, and health status were not significant. **Table 3** | Characteristic | Poor management group (n = 35) | Medication adherence group (n = 117) | Good management group (n = 418) | P valueb | | --- | --- | --- | --- | --- | | Sex | Sex | Sex | Sex | Sex | | Male | 21 (60.0) | 70 (59.8) | 253 (60.5) | .99 | | Female | 14 (40.0) | 47 (40.2) | 165 (39.5) | .99 | | Age, mean (SD), y | 55.3 (12.4) | 58.3 (14.4) | 63.1 (11.9) | <.001 | | Education | Education | Education | Education | Education | | Primary school | 4 (11.4) | 22 (18.8) | 88 (21.1) | .65 | | Junior high school | 4 (11.4) | 18 (15.4) | 52 (12.4) | .65 | | Senior high school | 14 (40.0) | 31 (26.5) | 116 (27.8) | .65 | | College and above | 13 (37.1) | 46 (39.3) | 162 (38.8) | .65 | | Health status compared with previous 12 months | Health status compared with previous 12 months | Health status compared with previous 12 months | Health status compared with previous 12 months | Health status compared with previous 12 months | | Worse | 6 (17.1) | 20 (17.1) | 65 (15.6) | .94 | | Neutral | 20 (57.1) | 63 (53.8) | 220 (52.8) | .94 | | Better | 9 (25.7) | 34 (29.1) | 132 (31.7) | .94 | | Health status score, mean (SD)c | 69.5 (12.0) | 70.1 (13.4) | 76.9 (11.9) | <.001 | | Diabetes duration, mean (SD), y | 9.3 (6.7) | 10.3 (7.3) | 11.9 (7.8) | <.001 | | HbA1c standard level | HbA1c standard level | HbA1c standard level | HbA1c standard level | HbA1c standard level | | <0.5 | 14 (48.3) | 43 (44.3) | 109 (29.9) | .007 | | ≥0.5 | 15 (51.3) | 54 (55.7) | 256 (70.1) | .007 | In Model 1, the good management group and medication adherence group were more likely than the poor management group to achieve better SDM (β = 0.183 [$$P \leq .003$$] and 0.120 [$$P \leq .07$$], respectively) (Table 4). Patients with a longer diabetes duration had lower SDM (β = −0.015; $P \leq .001$). In Models 2 and 3, which used the most recent HbA1c value and the 7-year average for HbA1c values, patients with a longer diabetes duration had higher HbA1c values (Model 2 β = 0.184, $P \leq .001$; Model 3 β = 0.208; $P \leq .001$). Conversely, we found lower HbA1c values among older patients (Model 2 β = −0.165, $$P \leq .002$$; Model 3 β = −0.274, $P \leq .001$) and the good management group (Model 2 β = −0.216, $$P \leq .01$$; Model 3 β = −0.217, $$P \leq .008$$). **Table 4** | Characteristic | Model 1,a β (95% CI) [P value] | Model 2,b standardized β (P value) | Model 3,c standardized β (P value) | | --- | --- | --- | --- | | Male sex | 0.045 (−0.014 to 0.103) [.13] | 0.030 (.52) | −0.005 (.91) | | Age | 0.008 (0.005 to 0.011) [<.001] | −0.165 (.002) | −0.274 (<.001) | | Education | Education | Education | Education | | Primary school | Reference | Reference | Reference | | Junior high school | 0.030 (−0.070 to 0.131) [.56] | 0.045 (.41) | 0.031 (.55) | | Senior high school | 0.132 (0.047 to 0.217) [.002] | −0.015 (.81) | −0.113 (.06) | | College and above | 0.154 (0.069 to 0.239) [<.001] | −0.071 (.28) | −0.179 (.005) | | Health status compared with previous 12 months | Health status compared with previous 12 months | Health status compared with previous 12 months | Health status compared with previous 12 months | | Worse | Reference | Reference | Reference | | Neutral vs worse | 0.036 (−0.055 to 0.126) [.44] | −0.036 (.59) | −0.052 (.42) | | Better vs worse | 0.093 (0.011 to 0.174) [.03] | −0.114 (.08) | −0.143 (.02) | | Health status score | 0.001 (−0.001 to 0.004) [.33] | −0.031 (.52) | −0.068 (.15) | | Diabetes duration | −0.015 (−0.018 to −0.011) [<.001] | 0.184 (<.001) | 0.208 (<.001) | | Patient profile group | Patient profile group | Patient profile group | Patient profile group | | Poor management | Reference | Reference | Reference | | Medication adherence | 0.120 (−0.009 to 0.250) [.07] | −0.136 (.10) | −0.155 (.054) | | Good management | 0.183 (0.062 to 0.303) [.003] | −0.216 (.01) | −0.217 (.008) | ## Discussion Our study used patient profiles to show that enhanced self-assessment in diabetes care, including diet, medication, exercise, and SMBG, self-efficacy, and self-regulation, may lead to improved SDM. The moderate association between patient profiles and SDM demonstrates a novel way to manage the manifest indicator of diabetes control from multiple years and classify patient behaviors in summary profiles derived from multiple dimensions. Apparently, when patients with type 2 diabetes had better self-care behaviors, they had a greater likelihood of having acceptable HbA1c levels (defined by a $7.0\%$ cut point). In addition, patients with more motivation to engage in health promotion and health care behaviors (autonomous or self-determined) had better outcomes in SDM. Our results are consistent with the results of previous studies [23,24]. The main goal of the 4 major components of diabetes control (diet, medication, exercise, and SMBG) is to maintain HbA1c at an optimal level to reduce the risk of complications, such as retinopathy, nephropathy, neuropathy, and stroke [25]. Moreover, according to social cognitive and self-determination theory, patients believe they can execute the behaviors necessary for producing and maintaining performance outcomes in accordance with the demands of diabetes care [26,27]. The Association of Diabetes Care & Education Specialists has provided an evidence-based model to help patients improve the behaviors necessary for diabetes self-management and increase their self-efficacy toward such self-care activities [28]. Although patients themselves play an important role in diabetes management, they may not change all their self-management behaviors to align with suggested standards. For example, low- and middle-income individuals have exhibited inadequate self-care behaviors because of the extensive dietary restrictions required and the suggestions for SMBG [29]. Previous studies listed unsurprising barriers to adequate self-care behaviors, such as a lack of motivation or inadequate knowledge and skills [30,31]. The integration of self-efficacy [28,32], self-regulation, and other factors with adherence to self-care behaviors among patients with diabetes could be complicated for the health education teams who are making immediate decisions in a limited amount of time, especially in a clinical setting. Therefore, we used 8 survey items to depict different patient profiles; this succinct questionnaire may help health care providers capture data on patient characteristics and their diabetes-related self-care behaviors. Patients in the poor management group had the lowest values for most self-care behaviors; however, they had slightly better values for self-regulation and health education. These patients were defined as having ineffective management. Although patients in the medication adherence group performed poorly in most self-care behaviors, they had the highest score in medication adherence and were described as demonstrating medication adherence. Patients in the good management group performed appropriately in all self-care behaviors. The good management group had better SDM than the other 2 groups, indicating that diabetes management should include medication as well as a healthy diet and physical activity. However, in our research, the use of patient profiles derived from LPA insufficiently captured the characteristics that reflect different patient behaviors. For example, the average scores of healthy diet and SMBG in the poor management and medication adherence groups were very close. Possible explanations include our small sample size, the limited number of dimensions, and the homogeneity of the items in the self-care evaluation to extract the subgroup information from the LPA model. Further studies should either adopt the diversity of patient behaviors based on the conception of behavioral science for validation or use a large database to create patient health profiles from LPA. Such information could help physicians analyze patient behaviors in diabetes care and develop customized diabetes control plans. Nonetheless, our novel method of using patient profiles in a clinical setting is beneficial. By evaluating patient behaviors through the use of limited questions in the decision-making process, physicians were not only able to review biomarker tendencies but also obtain a snapshot of behaviors from patient profiles. Thus, this approach helped them create diabetes management suggestions that their patients can understand. The American Diabetes Association suggests that health care providers offer diabetes self-management education and support (DSMES) that considers a patient’s confidence and self-efficacy behaviors as well as family and social support [14]. However, many contributing factors can hinder behavior change in diabetes management, ranging from motivation, skills, and resources to social support and the environment. Being able to address all possible factors is ideal; however, in doing so, the enormous complexity of diabetes and the one-size-fits-all behavior change can be overwhelming for patients. Thus, tailored strategies that can help them overcome modifiable barriers are needed. By using patient profiles, we found that most (>$70\%$) patients achieved favorable measures of diabetes control. However, the remaining participants relied on medication to control their diabetes and focused on health education to change their diet, exercise, and SMBG behaviors. Although a balanced method for diabetes management is the optimal approach, we can also consider using an easy-scale survey of the major components of self-care diabetes behavior. Through such an investigation, health education teams can provide additional resources to help patients overcome barriers to improving diabetes self-management. Medical teams may be more familiar with appraising biomarkers, such as the most recent HbA1c values in the medical record or the fluctuation of values over 3 to 5 years, than with assessing patients’ self-reported self-management. Interestingly, using either the most recent HbA1c value or the 7-year average of the HbA1c values leads to the same conclusion and demonstrates the potential of applying patient profiles in diabetes care. A patient-oriented approach, such as LPA, can be a better alternative for understanding diabetes management behavior as a totally functioning, not separate, variable, in terms of whole-system properties. The patient-oriented approach used in our study produced a condensed summary from a single categorical variable, as good at predicting an outcome as the original variables, and it also bypassed the difficulty of testing the interaction on empirical data in a variable-oriented approach. Health care providers may use such information obtained from patient profiles to help patients adjust their lifestyles according to customized suggestions. However, we need additional evidence to demonstrate and develop a practical approach (eg, a checklist) in clinical settings for diabetes management. ## Limitations This study has several limitations. First, although the use of patient profiles is a novel approach, it might not deal perfectly with multiple factors together; this shortcoming would be especially applicable when some factors are excessively homogenous in the population of interest. Our study used 8 items in the LPA model, including self-assessment in health education, self-management of medication, healthy diet, SMBG, regular exercise, self-efficacy, and self-regulation. Additionally, risk behaviors such as alcohol consumption, smoking, and family and environmental factors should be considered. Second, selection bias may have been possible. Compared with patients who did not participate, survey participants may have been more aware of their diabetes condition and more willing to comply with suggested diabetes care behaviors. We could have considered using a randomized trial or including all patients with type 2 diabetes in the hospital’s Department of Metabolism to obtain richer information. Third, the use of the standard HbA1c level is not common in clinical settings; furthermore, its accuracy has not been validated in diabetes care management. Finally, our survey did not consider several important factors, such as motivation or health literacy; thus, we did not examine their function in patient profiles. Our study had information on education level, which is strongly associated with literacy, and the TSRQd can be used for measuring motivation. However, because of insufficient measurements of these factors, we would be cautious about extending our explanations about them. Furthermore, we realize that diverse constructions in patient profiles may lead to different concepts of patient behaviors in diabetes management; it may be essential for future studies to consider multidisciplinary dimensions in patient behaviors to develop useful tools in diabetes care. ## Conclusions Using patient profiles derived from LPA confirmed the positive relationship between optimal patient behaviors in self-care management and SDM. Patients with type 2 diabetes exhibited good self-care management behaviors and confidence in these behaviors; moreover, greater engagement in health care behaviors may lead to improved SDM. However, additional information is required to validate the application of patient profiles in diabetes care in clinical settings. In promoting patient-centered care, the use of patient profiles with customized health education materials is a worthwhile approach to diabetes care. ## References 1. Bourne RR, Stevens GA, White RA, Smith JL, Flaxman SR, Price H. **Causes of vision loss worldwide, 1990–2010: a systematic analysis.**. *Lancet Glob Health* (2013) **1** e339-49. DOI: 10.1016/S2214-109X(13)70113-X 2. Saran R, Li Y, Robinson B, Ayanian J, Balkrishnan R, Bragg-Gresham J. **US Renal Data System 2014 annual data report: epidemiology of kidney disease in the United States.**. *Am J Kidney Dis* (2015) **66** S1-305, S1–305. DOI: 10.1053/j.ajkd.2015.05.001 3. Ling W, Huang Y, Huang Y-M, Fan R-R, Sui Y, Zhao H-L. **Global trend of diabetes mortality attributed to vascular complications, 2000-2016.**. *Cardiovasc Diabetol* (2020) **19** 182. DOI: 10.1186/s12933-020-01159-5 4. Anderson BJ, Laffel LM, Domenger C, Danne T, Phillip M, Mazza C. **Factors associated with diabetes-specific health-related quality of life in youth with type 1 diabetes: the Global TEENs Study.**. *Diabetes Care* (2017) **40** 1002-9. DOI: 10.2337/dc16-1990 5. Nam S, Chesla C, Stotts NA, Kroon L, Janson SL. **Barriers to diabetes management: patient and provider factors.**. *Diabetes Res Clin Pract* (2011) **93** 1-9. DOI: 10.1016/j.diabres.2011.02.002 6. Kruger DF. **Diabetes treatment options: one size does NOT fit all.**. *Clin Diabetes* (2012) **30** 41-3. DOI: 10.2337/diaclin.30.2.41 7. Rushforth B, McCrorie C, Glidewell L, Midgley E, Foy R. **Barriers to effective management of type 2 diabetes in primary care: qualitative systematic review.**. *Br J Gen Pract* (2016) **66** e114-27. DOI: 10.3399/bjgp16X683509 8. King DK, Glasgow RE, Toobert DJ, Strycker LA, Estabrooks PA, Osuna D. **Self-efficacy, problem solving, and social-environmental support are associated with diabetes self-management behaviors.**. *Diabetes Care* (2010) **33** 751-3. DOI: 10.2337/dc09-1746 9. D’Souza MS, Karkada SN, Parahoo K, Venkatesaperumal R, Achora S, Cayaban ARR. **Self-efficacy and self-care behaviours among adults with type 2 diabetes.**. *Appl Nurs Res* (2017) **36** 25-32. DOI: 10.1016/j.apnr.2017.05.004 10. Ahmad Sharoni SK, Abdul Rahman H, Minhat HS, Shariff-Ghazali S, Azman Ong MH. **The effects of self-efficacy enhancing program on foot self-care behaviour of older adults with diabetes: a randomised controlled trial in elderly care facility, Peninsular Malaysia.**. *PLoS One* (2018) **13**. DOI: 10.1371/journal.pone.0192417 11. Quirk H, Blake H, Tennyson R, Randell TL, Glazebrook C. **Physical activity interventions in children and young people with type 1 diabetes mellitus: a systematic review with meta-analysis.**. *Diabet Med* (2014) **31** 1163-73. DOI: 10.1111/dme.12531 12. Edelman SV, Polonsky WH. **Type 2 diabetes in the real world: the elusive nature of glycemic control.**. *Diabetes Care* (2017) **40** 1425-32. DOI: 10.2337/dc16-1974 13. Hood KK, Hilliard M, Piatt G, Ievers-Landis CE. **Effective strategies for encouraging behavior change in people with diabetes.**. *Diabetes Manag (Lond)* (2015) **5** 499-510. DOI: 10.2217/dmt.15.43 14. **Facilitating behavior change and well-being to improve health outcomes: standards of medical care in diabetes — 2021.**. *Diabetes Care* (2021) **44** S53-72. DOI: 10.2337/dc21-S005 15. Jing X, Chen J, Dong Y, Han D, Zhao H, Wang X. **Related factors of quality of life of type 2 diabetes patients: a systematic review and meta-analysis.**. *Health Qual Life Outcomes* (2018) **16** 189. DOI: 10.1186/s12955-018-1021-9 16. Ausili D, Bulgheroni M, Ballatore P, Specchia C, Ajdini A, Bezze S. **Self-care, quality of life and clinical outcomes of type 2 diabetes patients: an observational cross-sectional study.**. *Acta Diabetol* (2017) **54** 1001-8. DOI: 10.1007/s00592-017-1035-5 17. **Summary of revisions: standards of medical care in diabetes — 2020.**. *Diabetes Care* (2020) **43** S4-6. DOI: 10.2337/dc20-Srev 18. Herzog MH, Francis G, Clarke A. (2019) 19. Pastor DA, Barron KE, Miller BJ, Davis SL. **A latent profile analysis of college students’ achievement goal orientation.**. *Contemp Educ Psychol* (2007) **32** 8-47. DOI: 10.1016/j.cedpsych.2006.10.003 20. Chiou SJ, Liao K, Lin KC, Lin W. **Using patient health profile evaluation for predicting the likelihood of retinopathy in patients with type 2 diabetes: a cross-sectional study using latent profile analysis.**. *Int J Environ Res Public Health* (2022) **19** 6084. DOI: 10.3390/ijerph19106084 21. Mohn J, Graue M, Assmus J, Zoffmann V, Thordarson H, Peyrot M. **The effect of guided self-determination on self-management in persons with type 1 diabetes mellitus and HbA**. *BMJ Open* (2017) **7**. DOI: 10.1136/bmjopen-2016-013295 22. Williams GC, Freedman ZR, Deci EL. **Supporting autonomy to motivate patients with diabetes for glucose control.**. *Diabetes Care* (1998) **21** 1644-51. DOI: 10.2337/diacare.21.10.1644 23. Mottalib A, Salsberg V, Mohd-Yusof B-N, Mohamed W, Carolan P, Pober DM. **Effects of nutrition therapy on HbA1c and cardiovascular disease risk factors in overweight and obese patients with type 2 diabetes.**. *Nutr J* (2018) **17** 42. DOI: 10.1186/s12937-018-0351-0 24. Bweir S, Al-Jarrah M, Almalty AM, Maayah M, Smirnova IV, Novikova L. **Resistance exercise training lowers HbA1c more than aerobic training in adults with type 2 diabetes.**. *Diabetol Metab Syndr* (2009) **1** 27. DOI: 10.1186/1758-5996-1-27 25. Little RR, Rohlfing CL. **The long and winding road to optimal HbA1c measurement.**. *Clin Chim Acta* (2013) **418** 63-71. DOI: 10.1016/j.cca.2012.12.026 26. Bierhoff E, Grönniger J, Steppan U, Junginger T. **[Transperitoneal and extraperitoneal replacement of the aortic bifurcation — comparison of intra- and postoperative complications]**. *Vasa Suppl* (1988) **23** 153-6. PMID: 3206332 27. Bandura A. **Self-efficacy: toward a unifying theory of behavioral change.**. *Psychol Rev* (1977) **84** 191-215. DOI: 10.1037/0033-295X.84.2.191 28. Karimy M, Koohestani HR, Araban M. **The association between attitude, self-efficacy, and social support and adherence to diabetes self-care behavior.**. *Diabetol Metab Syndr* (2018) **10** 86. DOI: 10.1186/s13098-018-0386-6 29. Mogre V, Johnson NA, Tzelepis F, Shaw JE, Paul C. **A systematic review of adherence to diabetes self-care behaviours: evidence from low- and middle-income countries.**. *J Adv Nurs* (2019) **75** 3374-89. DOI: 10.1111/jan.14190 30. Mogre V, Johnson NA, Tzelepis F, Paul C. **Barriers to diabetic self-care: a qualitative study of patients’ and healthcare providers’ perspectives.**. *J Clin Nurs* (2019) **28** 2296-308. DOI: 10.1111/jocn.14835 31. Adu MD, Malabu UH, Malau-Aduli AEO, Malau-Aduli BS. **Enablers and barriers to effective diabetes self-management: A multi-national investigation.**. *PLoS One* (2019) **14**. DOI: 10.1371/journal.pone.0217771 32. Tan CCL, Cheng KKF, Hwang SW, Zhang N, Holroyd E, Wang W. **Effect of a diabetes self-efficacy enhancing program on older adults with type 2 diabetes: a randomized controlled trial.**. *Clin Nurs Res* (2020) **29** 293-303. DOI: 10.1177/1054773818792480
--- title: 'Determinants of Self-Care and Home-Based Management of Hypertension: An Integrative Review' authors: - Kennedy Diema Konlan - Jinhee Shin journal: Global Heart year: 2023 pmcid: PMC10038107 doi: 10.5334/gh.1190 license: CC BY 4.0 --- # Determinants of Self-Care and Home-Based Management of Hypertension: An Integrative Review ## Abstract ### Introduction: Patients with hypertension should perform diverse self-care activities that incorporate medication adherence and lifestyle modification, such as no smoking or alcohol, weight reduction, a low-salt diet, increased physical activity, increased self-monitoring, and stress reduction, for effective management at home. ### Aim: This systematic review assessed and synthesized the factors that are associated with self-care and home-based management of hypertension. ### Methods: The search of the articles incorporated the population, intervention, comparison, and outcome (PICO) framework. The literature was searched in four databases (PubMed, the Cumulative Index to Nursing and Allied Health Literature [CINAHL], Embase, and Web of Science) until 2022. The articles retrieved and searched from the reference list [531] were transported to EndNote version 20, and duplicates [19] were identified and removed to produce 512 titles. Following the eventual title, abstracts, and full-text screening, 13 articles were appropriate for this study. The narrative and thematic data analysis were used to analyze and integrate the data. ### Results: The analysis showed five themes were associated with home-based self-care and blood pressure (BP) control among patients diagnosed with hypertension. These themes that emerged were [1] the prevalence of control of BP, [2] sociodemographic factors, [3] treatment-related factors, [4] knowledge of management, and [5] knowledge of the prevention of risk factors of hypertension. The demographic factors influencing home-based self-care for hypertension were gender, age, and socioeconomic status. In contrast, the treatment factors were duration of hypertension treatment, medication burden, and medication adherence. Other factors that influenced self-care were inadequate knowledge of BP management, follow-up care, and risk factors of hypertension. ### Conclusion: Hypertension self-care interventions must incorporate individual, societal, and cultural perspectives in increasing knowledge and improving home-based hypertension management. Therefore, well-designed clinical and community-dwelling interventions should integrate personal, social, and cultural perspectives to improve behavior in the home management of hypertension by increasing knowledge and self-efficacy. ## Introduction Hypertension disease presents a challenge to patients, as they are expected to institute measures at home to ensure effective self-care and management practices. Patients with hypertension perform diverse activities that can be described as self-care activities for effective disease management [1]. The focus of hypertension self-care management must incorporate medication adherence and lifestyle modification (no smoking or alcohol, weight reduction, low-salt diet, and increased physical activity), increased self-monitoring of blood pressure (BP), and stress reduction [2]. However, self-care management in hypertensive patients must be lifelong, even though it is usually challenging and overwhelming because the patient lacks experience in self-management and the necessary knowledge, tools, and support [3]. Self-care is described as individual actions directed toward self or the environment to regulate individual functioning to improve health, reduce risk, and avoid related complications, as well as ensure general well-being [45]. It is important to note that self-care behavior among hypertension patients relates to BP control and prevents related complications [5]. Patient self-care and home-based management of hypertension positively affect clinical outcomes and reduce the occurrence of stroke and related cardiovascular disease [25]. Home-based measures are important in the patient’s likelihood of avoiding complications and effective improvement [246]. Self-care adherence is low among adults with hypertension [7] because patients are often unwilling to make recommended behavioral changes [89], especially in settings where lay health knowledge is averagely low. Globally, the prevalence of hypertension has increased and shifted from developed to developing countries over the past 40 years [10]. Nearly three-quarters of hypertensive patients live in developing countries [11]. Awareness of high BP management is deficient in developing countries [1213]. Many hypertensive patients are ignorant of their disease and treatment due to lack of health care access, distrust of Western medicine, and inadequate health literacy [1214]. Although adherence to self-care and management is an important part of patient management to achieve hypertension treatment goals, self-care in the African population remains poor [1516]. Therefore, it is very important to identify factors that influence hypertension self-care and management at home. The self-care and home management of hypertension, like other chronic diseases, involve various behavioral changes that require optimal and effective medication adherence, self-efficacy, and prevention of complications [6817]. Recent systematic reviews on self-care management of hypertension focused on the use of electronic-based technologies [181920], the influence of self-care efficacy [6], and health promotion interventions for the prevention of hypertension [161721]. However, to the best of our knowledge, no systematic review in recent times has specifically focused on the factors that influence home-based and self-care management of hypertension. This systematic review assessed and thematically synthesized the factors that are associated with self-care and home-based management of hypertension. ## Study framework The search of the articles incorporated the population, intervention, comparison, and outcome (PICO) framework. The population was patients diagnosed with hypertension and referred to home care or self-care management. There were no specific comparisons. The outcome was improved hypertension status or a reduction in BP. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework was incorporated in reporting the studies. Using the PRISMA guidelines and checklist allowed for reproducibility, transparency, and clarity in reporting the findings [2223]. The phases of the entire study included identifying the research question, identifying relevant studies, quality appraisal, and selecting studies for inclusion. ## Identification of the research question Home and self-care management of hypertension is an essential factor that can promote the health of people diagnosed with the disease, especially in the poorest parts of the world, like sub-Saharan Africa. Therefore, interventions must be tailored toward addressing the specific determinants of hypertension self-care management. The specific research question was, what are the determinants of self-care and home management of hypertension in low-resource settings in Africa? ## Identification of relevant studies The literature search was done in four databases (PubMed, the Cumulative Index to Allied Health and Nursing Research [CINAHL], Embase, and Web of Science) up until 2022. In developing the search terms, the medical subject heading (MeSH) terms were used as the bases as well as free terms and wildcards incorporated with the appropriate Boolean operators. The search keywords were modified with the appropriate keywords to reflect the specific requirement of each database. The key terms used were first generated in PubMed and subsequently modified to sweet for each database. These primary keywords from PubMed using the PICO framework include population as “Hypertension”[MeSH] OR “High Blood Pressure” OR “Heart Disease”; Interventions as (“Self-Care”[MeSH] OR “Home Care Services”[MeSH] OR “Disease Management”[MeSH] OR “Therapeutics”[MeSH]); Comparison (Determinant* OR Predictor* OR Related OR Associated); and Outcome as “Medication Adherence” OR “Dietary Restriction” OR “Improved Physical Activity” OR “Increased Knowledge” OR “Improved Awareness.” These keywords were combined with the appropriate Boolean operators and MeSH terms where it was applicable. ## Selection of studies After conducting the initial search, each title, abstract, and full text was screened based on predetermined inclusion criteria and the identified key terms. Also, the references of selected articles were also searched to identify any additional articles that would be relevant to this study. The articles retrieved from the database search and reference list [531] were transported to EndNote version 20, and duplicates [19] were identified and removed to produce 512 titles. Following the eventual title, abstracts, and full-text screening, the total number of appropriate articles for this study was 13. ## Selection criteria Only studies that identified the factors associated with home management hypertension, regardless of the study setting, were included. The studies excluded were protocols, systematic studies (hospital service factors) that hinder diagnosis and self-monitoring of hypertension, and the factors related to the service provider (e.g., nurse, doctor), like shortage of items. ## Quality appraisal The two authors independently used the Mixed Methods Appraisal Tool (MMAT) version 2018 to assess each study for quality [2425]. The assessment results using this tool were compared between the two researchers for similarities. Where differences were found between the results from the authors, it was discussed until a consensus was achieved. The MMAT is a quality assessment tool that appraises the methodological quality of qualitative, quantitative, and mixed methods studies. The quantitative section appropriate for appraising the articles selected for this review assessed the aim of the study, study design, methodology, recruitment of study participants, and data collection methods, including analysis, presentation of results, and discussion, as well as the conclusions. Based on the assessment, the studies can be individually rated as high, moderate, or low quality. However, the subsequent views of Hong et al. 2018 emphasized that the researcher does not need to assess the overall quality of studies using this categorization [25]. In their paper, the authors strongly recommended that a detailed presentation of the appraised findings should be done. The components of the screening questions from the MMAT include whether there were (a) the presence of clear research questions and (b) collected data addressing the research question. Based on the consensus gained during the appraisal, all the articles received an affirmative response to the two screening questions above. The appraisal sections that were relevant for this study were the descriptive quantitative section that assessed the following criteria: [1] relevance of sampling strategy, [2] representativeness of sample to the target population, [3] appropriateness of measurements, [4] risk of nonresponse bias, and [5] appropriateness of statistical analysis. All the studies ($$n = 11$$) met the criteria of the appropriateness of the sampling strategy to answer the research question except two studies [2627]. One study [28] assessed methodological quality using the qualitative section. The results showed an affirmation of all the qualitative appraisal questions. Another study [29] was assessed under the mixed methods section and was also affirmative to all the questions. ## Data extraction and analysis First, the two authors developed, discussed, and accepted a matrix to ensure comprehensiveness in the extracted data. The key parameters extracted from each study were the year, population and sample, study settings, outcome variable, main determinants of home care and self-care, and key findings. Second, the information of each study was transformed into narrative statements to enable the use of the thematic synthesis approach for data analysis [3031]. Third, in this approach, codes were generated from the narratives formed; the codes were then linked to form subthemes, while similar themes were coalesced to form the main themes. The main themes incorporated in informing the report of this study were [1] the prevalence of control of BP, [2] sociodemographic factors, [3] treatment-related factors, [4] knowledge of management, and [5] knowledge of prevention of risk factors of hypertension. ## Results Through the advanced search of electronic databases using predetermined keywords [529] and a reference list of identified studies [2], about 531 studies were identified, and 19 were identified as duplicates in the endnotes. The duplicates were removed, and all 512 titles were screened through title and abstract. Based on a predetermined inclusion criterion, 41 studies were full articles, and 13 were identified as appropriate for this study. The process of article selection is shown in Figure 1. **Figure 1:** *PRISMA flow diagram for selection of studies.* ## Study characteristics The studies were conducted in various countries, including Tanzania [27], Ethiopia [3233343536], Ghana [37], Sierra Leone [29], Cameroon [38], Eritrea [28], South Africa [2639], and Kenya [40]. The various studies used diverse study designs. The characteristics of the studies used are shown in Table 1. The designs that were adopted in identifying the factors associated with home-based self-care management of hypertension in sub-Saharan Africa were cross-sectional [2627343738], hospital-based cross-sectional [333536], nationally representative cross-sectional data [3940], retrospective cohort study [32], retrospective chart review [29], and qualitative study designs [28]. **Table 1** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Unnamed: 5 | | --- | --- | --- | --- | --- | --- | | AUTHOR | MAIN GOAL | SETTING | DESIGN | SAMPLE | DATA ANALYSIS | | Maginga et al., 2015 | Determine factors associated with BP control among adults attending a hypertension clinic | Bugando Medical Centre, Tanzania | Cross-sectional study | 300 hypertension patients, selected consecutively | Fisher’s exact, Wilcoxon rank-sum test, univariable and multivariable logistic regression | | Berhe et al., 2017 | Examined determinants of achieving BP control and treatment intensification in patients with uncontrolled BP | Six public hospitals, Ethiopia | A retrospective cohort study | 897 adult ambulatory hypertension patients | Descriptive statistics and multivariable logistic regression | | Labata et al., 2019 | Assessed predictors of self-care practices among hypertensive patients | Jimma University Specialized Hospital, Ethiopia | Hospital-based cross-sectional study | 341 adult hypertensive patients | Descriptive statistics and multivariate logistics regression | | Niriayo et al., 2019 | Assessed the rate of adherence to self-care behaviors and associated factors among hypertensive patients | Ayder Comprehensive Specialized Hospital, Ethiopia | Cross-sectional study | 276 ambulatory hypertensive patients | Univariable and binary logistic regression | | Berhe et al., 2020 | Assessed the prevalence and factors associated with uncontrolled hypertension among adults | Mekelle public hospitals, Tigray, Ethiopia | Hospital-based cross-sectional study | 396 hypertensive patients, systematic random sampling | Bivariable and multivariable logistic regression | | Gebremichael et al., 2019 | Assessed self-care practices and associated factors among hypertensive patients | Ayder Comprehensive Specialized Hospital, Ethiopia | Hospital-based cross-sectional study | 320 hypertension patients, simple random sampling | Descriptive statistics, logistics, and multivariate regression | | Okai et al., 2020 | Assessed the patient-level factors that influence hypertension control | Two hospitals in Accra, Ghana | Cross-sectional study | 360 hypertensive patients | Chi-square tests and logistic regression | | Herskind et al., 2019 | Assessed an initiative conducted by two health clinics to begin treatment of hypertension among patients | Two clinics, Sierra Leone | Retrospective chart review and survey | 487 records of patients and 68 hypertension patients’ convenience sample | Descriptive statistics | | Adidja et al., 2018 | Determine the rate and factors associated with nonadherence to antihypertensive pharmacotherapy, the association between nonadherence and BP control | Buea Health District, Cameroon | Community-based cross-sectional study | 183 adults, stratified cluster sampling | Descriptive, chi-square, Fisher’s exact test, t-test, multivariable logistic regression | | Gebrezgi et al., 2017 | Identified barriers and facilitated hypertension management from the perspective of the patients | Asmara, Eritrea | Qualitative study | 48 individual in-depth interviews and 2 FGD | Thematic analysis | | Ware et al., 2019 | Investigated traditional risk factors alongside other health and sociodemographic indicators to determine predictors of hypertension prevalence and management | South Africa | Cross-sectional of a nationally representative cohort | WHO-SAGE South Africa Wave 1 recruited 4,223 respondents from selected probability sampled | Chi-square, Mann-Whitney U test, t-tests, logistic regression | | Mohamed et al., 2018 | Estimated the prevalence of hypertension, awareness, treatment, and control | Kenya | A national cross-sectional household survey study | 4,485 data from the 2015 Kenya STEPs survey, randomly selected | Descriptive statistics, multiple logistic regression, bivariate logistic regression | | Adeniyi et al., 2016 | Examined the sociodemographic and clinical determinants of uncontrolled hypertension among individuals living with T2DM in the rural communities | Mthatha, South Africa. | Cross-sectional study | 265 individuals living with T2DM and hypertension | Univariate and multivariate logistic regression | ## Thematic Data Analysis The synthesis showed that five key themes were associated with home-based self-care and BP control among patients. These key terms that emerged were [1] the prevalence of control of BP, [2] sociodemographic factors, [3] treatment-related factors, [4] knowledge of management, and [5] knowledge of prevention of risk factors of hypertension. The distribution of key findings is presented in Table 2. **Table 2** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | | --- | --- | --- | --- | | AUTHOR | OUTCOME AND MEASUREMENT | KEY DETERMINANTS OF HYPERTENSION CONTROL | KEY FINDINGS | | Maginga et al., 2015 | Medication adherence: MMAS-4Knowledge: Self-developed pretested questionnaire | Good knowledge (OR = 2.50, 95% CI, 1.00–6.10, p = 0.047)Attitudes (OR = 2.70, 95% CI, 1.00–7.10, p = 0.004)Practices (OR = 5.40, 95% CI, 2.30–13.0, p < 0.001) | Patients (47.7%) had controlled hypertension.Obesity and higher medication costs were associated with decreased control.There was high adherence (56.0%) to medication.Participants had moderate scores for knowledge (41.0%), attitudes (45.3%), and practices (49.3%). | | Berhe et al., 2017 | Medication adherence: MMASOther determinants: Self-developed questionnaire | Treatment at general hospitals (OR = 1.89, 95% CI 1.26–2.83)Previously uncontrolled BP (OR = 0.30, 95% CI 0.21–0.43)Treatment regimens with diuretics (OR = 0.68, 95% CI 0.50–0.94)Age (OR = 0.99, 95% CI = 0.98–1.00) | BP was controlled in 37.0%, and treatment was intensified for 23.0% of patients with uncontrolled BP.The antihypertensive medication adherence rate (MMAS ≥ 7) was 40.0% and 57.0% for the lower cutoff (MMAS ≥ 6). | | Labata et al., 2019 | Hypertension self-care practices: Adapted H-SCALE questionnaire | Normal weight (AOR = 1.82, 95% CI = 1.07–3.09) a predictor of medication usageGood self-efficacy (AOR = 2.58, 95% CI 1.47–0.52) a predictor of a low-salt dietFemale predictor physical activity (AOR = 0.51, 95% CI 0.30–0.88) and nonsmoking (AOR = 3.62, 95% CI 1.21–10.85) | 61.9%, 30.5%, 44.9%, 88.3%, 93.5%, and 56.9% were adherent to medication, low-salt diet, physical activity, alcohol abstinence, nonsmoking, and weight management, respectively.Adequate knowledge of hypertension was 2.58 times more likely, and females were less likely to adhere to physical activity. | | Niriayo et al., 2019 | Self-care behaviors: H-SCALEBeliefs about medication: Belief about medicine questionnaire (BMQ) | Rural resident (AOR = 0.45, 95% CI: 0.21–0.97)Comorbidity (AOR = 0.16, 95% CI 0.08–0.31)Negative medication belief (AOR = 0.25, 95% CI 0.14–0.46) | Antihypertensive medications adherent (48.2%) and recommended physical activity (44.9%)Female (AOR = 1.97, 95% CI 1.03–3.75) and lack of knowledge on self-care (AOR = 0.07, 95% CI 0.03–0.16) were associated with alcohol abstinence and a low-salt diet. | | Berhe et al., 2020 | Adherence to self-care activities: H-SCALE | Age ≥ 50 years (AOR = 2.33, 95% CI 1.25, 4.35)Nonadherence to antihypertensive medication (AOR = 1.82, 95% CI 1.08–3.04)Nonadherence to physical exercise (AOR = 1.79, 95% CI 1.13–2.83)Nonadherence to low-salt diet (AOR = 1.98, 95% CI 1.18–3.31)Nonadherence to weight management (AOR = 2.06, 95% CI 1.31–3.23) | Prevalence of uncontrolled hypertension was found to be 48.6%.26.1%, 59.1%, 73.9%, and 38.6% of hypertensive patients were nonadherent to medication, physical exercise, low-salt diet, and weight management, respectively. | | Gebremichael et al., 2019 | Self-care practice: H-SCALEKnowledge: Hypertension evaluation of lifestyle and management (HELM) scale | Sex (AOR = 2.25, 95% CI 1.09–4.65)Age (AOR = 3.26, 95% CI 1.03–10.35)Educational status (AOR = 4.20, 95% CI 1.30–13.55)Disease duration (AOR = 3.12, 95% CI 1.20–8.10)BP status (AOR = 2.72, 95% CI 1.25–5.92)Knowledge (AOR = 6.19, 95% CI 2.90–13.21) | Good self-care practice was only found among 20.3% of patients.Adherence to not smoking, antihypertensive medication, alcohol abstinence, dietary management, physical exercise, and weight management was found to be 99.1%, 74.1%, 67.2%, 63.1%, 49.4%, and 40.6%, respectively. | | Okai et al., 2020 | Blood pressure control: Pretested self-developed questionnaire with expert opinion | Sex (AOR = 3.53, 95% CI 1.73–7.25)Educational at junior high school (AOR = 3.52, 95% CI 1.72–7.22)Senior and junior high school (AOR = 2.64, 95% CI 1.40–6.66 and AOR = 3.06, 95% CI 1.03–6.67)Comorbidity (AOR = 2.41, 95% CI 1.32– 4.42)Increased pill burden (AOR = 0.27, 95% CI 0.10–0.73)Length of diagnosis of 2–5 years (AOR = 0.32, 95% CI 0.18–0.57) | No comorbidities (18.0%) had achieved hypertension control.Dyslipidemia (8.9%) had controlled hypertension (p < 0.006).Taking a higher number of antihypertensive pills per day was also associated with a reduced likelihood of attaining hypertension control.Most patients reported forgetfulness, side effects of medication, and high pill burden as reasons for missing their medications. | | Herskind et al., 2019 | Medication adherence: Medication possession ratio | Patients were most likely to cite transportation (81.0%), financial burden (69.0%), and schedule conflicts with work or other prior commitments (25.0%) as barriers to care. | Forgetfulness (12.0%) and lack of symptoms (9.0%) were challenges that patients reported facing in attending follow-up appointments.Home visits (13.0%), outreach (13.0%), and phone or mobile reminders (12.0%) were strategies to improve adherence. | | Adidja et al., 2018 | Medication adherence: MMAS | Forgetfulness (AOR = 7.90, 95% CI 3.00–20.80)Multiple daily doses (AOR = 2.50, 95% CI 1.20–5.60)Financial constraints (AOR = 2.80, 95% CI 1.10–6.90)Adverse drug effects (AOR = 7.60, 95% CI 1.70–33.0) | Participants (67.7%) were nonadherent to medications.BP was controlled in only 21.3% of participants and was better in those who were adherent to medication (47.5%, p < 0.010). | | Gebrezgi et al., 2017 | Facilitators and barriers to self-care: Self-developed interview guide and focus group discussion guide | | Individual factors: economic barriers, stress, nonadherence to medications due to the use of traditional remedies, and difficulties and misconceptions about following physical activity guidelines influenced self-care.Individual knowledge, family, and government support were important factors to the patient’s success in the personal hypertension management. | | Ware et al., 2019 | Predictors of hypertension prevalence and management: World Health Survey (WHS, 2002–2004; 70 countries) | Waist-to height ratio > 0.5 and diabetes comorbidity were the most significant predictors of hypertension presence, awareness, and treatment.Women and individuals reporting lower salt use were more likely to be aware of and treated for hypertension. | Older age, larger waist-to-height ratio, lower levels of education, and diabetes comorbidity were also predictive of individuals with hypertension being aware of their status.Older age, female sex, larger waist-to-height ratio, diabetes comorbidity, lower levels of education, and not adding salt to food at the table were predictive of current antihypertensive medication use. | | Mohamed et al., 2018 | World Health Organization’s STEPs survey methodology tool | Among those aware, only 26.9% were on treatment, and 51.7% among those on treatment had achieved blood pressure control.Factors associated with hypertension were older age, higher BMI, and harmful use of alcohol. | The overall age-standardized prevalence for hypertension was 24.5%.Only 15.6% were aware of their elevated blood pressure.Factors associated with awareness were older age (p = 0.013) and being male (p < 0.001). | | Adeniyi et al., 2016 | Uncontrolled hypertension: Self-developed questionnaire | Unemployed status (p < 0.001)Excessive alcohol intake (p = 0.007)Consumption of a Western-type diet (p < 0.001) | Independent determinants of uncontrolled hypertension were unemployment, current excessive drinker of alcohol and adherence to Western-type diet. | ## Prevalence of control of BP Two studies specifically assessed the prevalence of hypertension control during their respective studies. It was shown that hypertension control prevalence during the study periods was $47.7\%$ [27] and $37.0\%$ [32]. ## Sociodemographic factors associated with BP control Several sociodemographic variables were assessed to determine the factors that influence hypertension control. The key subthemes assessed included gender (sex), age, and socioeconomic status. ## Gender as a predictor of BP The gender of hypertension patients was identified as an important factor associated with the ability to control BP [3436]. Some of the studies showed that females (AOR = 2.25, $95\%$ CI 1.09–4.65, $$p \leq 0.028$$) were found 2.25 times [36], 3.55 times (AOR = 3.55, $95\%$ CI 1.72–7.22) more likely to have good self-care practice than males [37]. Also, men were identified to have reduced awareness of BP status [263940]. Men had significantly reduced odds of being aware of their hypertensive status (AOR = 0.35, $95\%$ CI 0.22–0.56) compared to women [40]. Another important predictor was that females had an increased tendency to be adherent to antihypertensive medication use [26273439]. Females had a higher tendency of BP control behavior, such as weight management (AOR = 0.46, $95\%$ CI 0.23–0.92) and physical activity (AOR = 0.22, $95\%$ CI 0.12–0.40) than males [34]. Females were also identified to have higher medication adherence and BP control [38]. However, some studies did not identify any significant difference in BP control across genders [262733]. Control of BP was not significantly different between men and women [27]. Female respondents were less likely to adhere to physical activity (AOR = 0.51, $95\%$ CI 0.30–0.88) and nonsmoking (AOR = 3.62, $95\%$ CI 1.21–10.85) behavior [33]. ## Age as a predictor of BP control The age of hypertension patients was identified as an important predictor of the ability to control BP [323435363839]. Uncontrolled hypertension was 2.3 times (AOR = 0.19, $95\%$ CI 0.06–0.61) higher among patients above 50 years compared to those above 60 [34]. Also, patients above 50 years were less adherent to weight management than younger individuals (18–35 years) [34]. Self-care practices of participants also improved with advancing age [36]. Patients under 65 years were 3.26 times (AOR = 3.26, $95\%$ CI 1.03–10.35, $$p \leq 0.044$$) more likely to have good self-care practice than patients above 65 years [36]. Also, other studies showed that increasing age was associated with the tendency to be medication adherent and consequently maintain BP control [3436]. ## The socioeconomic situation as a predictor of BP control The socioeconomic status of hypertension patients was an important factor associated with BP control [2734]. The major economic challenge associated with controlling BP was the medication cost and treatment [27]. The higher medication cost was associated with decreased odds of BP control at the study visit [27]. Odds of control (OR = 0.80, $95\%$ CI 0.70–0.95, $$p \leq 0.010$$) decreased by $20\%$ for every 10,000 Tanzanian shillings (TZS) spent on medication [27]. The place of residence influences the ability of respondents to adhere to medications [34]. Rural residents were less adherent to their medication (AOR = 0.45, $95\%$ CI 0.21–0.97) than urban dwellers [34]. Other factors that predicted BP control and the self-care ability of respondents were the educational levels, as those who had at least a college level of education were found to be 4.205 times (AOR = 4.20, $95\%$ CI 1.30–13.55, $$p \leq 0.016$$) more likely to have good self-care practice than those who were unable to read and write [36]. Participants with greater social support were 2.81 times (AOR = 2.81, $95\%$ CI 1.20–6.53) more likely to adhere to a low-salt diet than their counterparts [33]. Other important socioeconomic factors that predicted control of hypertension were forgetfulness (AOR = 7.90, $95\%$ CI 3.00–20.80, $p \leq 0.001$) and lack of finances (AOR = 2.80, $95\%$ CI 1.10–6.90, $$p \leq 0.024$$) [34]. ## Treatment-related factors The treatment-related factors associated with home-based and self-care management of hypertension had three themes: [1] duration of treatment of hypertension, [2] pill-related factors, and [3] medication adherence. ## Duration of treatment Several important key factors influence the ability to continue treatment and ensure BP control [323637]. Some of these factors included the duration of disease and treatment, significantly influencing the ability to initiate and maintain self-care [3638] and control BP [3238]. Less than four years of disease duration was 3.12 times (AOR = 3.12, $95\%$ CI 1.20–8.10, $$p \leq 0.019$$) more likely to practice good self-care than those with less than < 2 years of disease duration [36]. ## Medication-related factors The number of medications the hypertension patient took also influenced their ability to control and maintain self-care [2637]. When hypertension patients took three to four antihypertensive pills daily, the odds of having a controlled BP were reduced by $68.0\%$ (AOR = 0.32, $95\%$ CI 0.18–0.57) compared to those who took one to two pills [37]. When patients take multiple antihypertensive agents, it improves their tendency to have controlled BP [26]. Also, multiple daily doses (AOR = 2.50, $95\%$ CI 1.20–5.60, $$p \leq 0.020$$) and drug side effects (AOR = 7.00, $95\%$ CI 1.70–33.6, $$p \leq 0.007$$) were independent predictors of nonadherence after controlling for potential confounders in multivariate analysis [38]. Other pill-related factors that influence the control of BP were follow-up at general hospitals (OR = 1.89, $95\%$ CI 1.26–2.83), inadequately controlled BP at prior visits (OR = 0.30, $95\%$ CI 0.21–0.43), longer treatment duration per year (OR = 1.04, $95\%$ CI 1.01–1.06), and prescribed diuretics (OR = 0.68, $95\%$ CI 0.50–0.94) [32]. ## Medication adherence to BP control The level of medication adherence was determined to be high [273234] and was measured using the Morisky Medication Adherence Scale [MMAS] [2732]. Most patients ($56.0\%$) had high adherence to medication on the MMAS-4 [27]. The antihypertensive medication adherence rate (MMAS ≥ 7) was $40.0\%$ and $57.0\%$ for the lower cutoff (MMAS ≥ 6) [32]. Patients with high medication adherence to the MMAS-4 had increased odds (OR = 18.8, $95\%$ CI 7.80–45.40, $p \leq 0.001$) of control relative to those with low adherence at the study visit [27]. Patients with a negative medication belief were less likely to adhere to their medication (AOR = 0.25, $95\%$ CI 0.14–0.46) than those with a positive medication belief [34]. Adult participants were adherent to the prescribed antihypertensive medications ($48.2\%$) and the recommended level of physical activity ($44.9\%$) [34]. Patients who were nonadherent to prescribed antihypertensive drugs were two times (AOR = 1.82, $95\%$ CI 1.08–3.04) more likely to have uncontrolled hypertension than those who were adherent [35]. Participants with poor self-efficacy (AOR = 0.40, $95\%$ CI 0.22–0.73) were less likely to adhere to medication usage than participants with good self-efficacy [33]. Regarding specific practices, rarely or never taking medications as prescribed (OR = 0.10, $95\%$ CI 0.50–0.20, $p \leq 0.001$) was associated with decreased hypertension [HTN] control [27]. ## Knowledge as a predictor of BP control An important determinant of BP control was the level of knowledge of hypertension and control measures [273436]. This was because moderate (OR = 1.80, $95\%$ CI 1.01–3.20, $$p \leq 0.046$$) and good knowledge (OR = 2.10, $95\%$ CI 1.00–4.50, $$p \leq 0.049$$) of HTN had increased odds of control [27]. Also, good knowledge was found 6.19 times (AOR = 6.19, $95\%$ CI 2.90–13.21, $p \leq 0.001$) more positively associated with good self-care practice than poor knowledge [36]. Hypertension patients who knew the negative effects of salt ($94.9\%$), alcohol ($81.5\%$), and smoking ($87.7\%$) and the positive effect of physical exercise ($54.7\%$) had positive actions toward hypertension control [34]. Hypertension patients who were not knowledgeable about SCBs were less adherent to weight management (AOR = 0.13, $95\%$ CI 0.03–0.57) and alcohol abstinence (AOR = 0.07, $95\%$ CI 0.03–0.16) compared to those who were knowledgeable [34]. Overall, $82.2\%$ of the participants were knowledgeable about the impact of the Self-care behavior [SCB] on hypertension control [34]. Those with moderate (OR = 2.80, $95\%$ CI 1.20–6.40, $$p \leq 0.020$$) and good (OR = 3.00, $95\%$ CI 1.30–7.00, $$p \leq 0.010$$) attitudes had increased odds of hypertension control [27]. Hypertension patients having adequate knowledge of hypertension were 2.58 times (AOR = 2.58, $95\%$ CI 1.12–5.94) more likely to adhere to practicing physical activity [33]. ## Knowledge of prevention of risk factors of hypertension Several risk factors of hypertension were identified to influence the tendency to have good BP control. Some of these key risk factors include salt intake, practices, weight control, avoiding smoking and alcohol, and the presence of other comorbidities. ## The level of salt in the diet was associated with BP control The level of salt intake was an important predictor of hypertension control among patients, as nonadherents to a low-salt diet were two times (AOR = 1.98, $95\%$ CI 1.18–3.31) more likely to develop uncontrolled hypertension [35]. Also, when individuals’, especially women’s, knowledge of hypertension was identified to be low, their chances of having a high-salt diet increased [39]. Another important factor was adding salt to meals (OR = 0.40, $95\%$ CI, 0.20–0.60, $$p \leq 0.001$$) as an important factor for BP control [27]. ## Behavioral measures as a predictor of BP control The conscious behavioral measures instituted by individuals were an important predictor of BP control among hypertension patients [273236]. Controlled BP was found 2.7 times (AOR = 2.72, $95\%$ CI 1.25–5.92, $$p \leq 0.011$$) more associated with good self-care practice than uncontrolled BP [36]. Those who had moderate (OR = 4.80, $95\%$ CI 2.40–9.40, $p \leq 0.001$) and good (OR = 11.00, $95\%$ CI 5.00–24.20, $p \leq 0.001$) practices to prevent hypertension also had increased odds of controlled BP [27]. More severe hypertension stages, stage II hypertension (OR = 0.17, $95\%$ CI 0.09–0.35) and stage I hypertension (OR = 0.34, $95\%$ CI 0.17–0.67), were associated with more difficulty in achieving the target BP [32]. ## Weight as a predictor of BP control Hypertension patients who were more conscious of controlling body weight had a higher chance of BP control [3335]. The odds (AOR = 2.06, $95\%$ CI 1.31–3.23) of uncontrolled hypertension were twice as high among those with nonadherent weight management [35]. Normal weight patients were 1.82 times (AOR = 1.82, $95\%$ CI 1.07–3.09) more likely to adhere to medication usage practice than overweight respondents [3335], while normal weight respondents were 2.22 times more likely (AOR = 2.21, $95\%$ CI 1.21–4.04) to practice weight management [33]. Hypertension patients with good self-efficacy were 2.60 times more likely (AOR = 2.58, $95\%$ CI 1.41–4.73) to maintain their weight than poor self-efficacy [33]. Also, rarely or never adhering to normal weight control advice (OR = 0.40, $95\%$ CI 0.20–0.60, $p \leq 0.001$) was a predictor of poor hypertension control [27]. Patients who did not adhere to physical exercise were 1.8 times (AOR = 1.79, $95\%$ CI 1.13–2.83) more likely to have uncontrolled hypertension compared to those who adhered to physical exercise [35]. ## Smoking and alcohol intake are predictors of high BP Alcohol intake was identified as an important factor that influenced the likelihood of having hypertension control [2733]. This was because drinking alcohol (OR = 0.30, $95\%$ CI 0.10–0.70, $$p \leq 0.006$$) was a significant factor in BP control [2733]. Females were more likely to adhere to alcohol abstinence (AOR = 1.97, $95\%$ CI 1.03–3.75) and nonsmoking behavior (AOR = 6.33, $95\%$ CI 1.80–22.31) than males [34]. ## Comorbidities as a predictor of BP control The level and presence of comorbidities were identified to be associated with hypertension patients’ ability to have controlled BP [343739]. Patients with comorbidities were also less adherent to their medication (AOR = 0.16, $95\%$ CI 0.08–0.31) than those without comorbidities [34]. Also, a $69.0\%$ (AOR = 0.31, $95\%$ CI 0.11–0.89) reduction in the odds of having controlled hypertension was identified among patients who suffered from dyslipidemia as a comorbidity [37]. Elevated waist-to-height ratio and diabetes diagnosis were the most significant predictors of hypertension and being aware of hypertension status [39]. ## Discussion This review identified and integrated the factors associated with home-based self-care management of hypertension. In chronic diseases like hypertension, home-based management is cardinal in improving a patient’s outcome and the ability to avoid complications and limit the progress of the disease [414243]. Hypertension patients must identify and institute home measures that help them improve their BP levels and avoid related complications. The studies on hypertension home-based management are largely cross-sectional and limited to a specific culture or geographical location. Therefore, we were motivated to identify the factors associated with home-based management of hypertension to identify and integrate measures to improve client outcomes. We identified the prevalence of BP control to range from $21.3\%$ [38] to $47.7\%$ [27] among hypertension patients. This study highlighted the diverse components that must be considered when interventions are implemented for home-based self-care management of hypertension. Hypertension control through self-care at home was reported in similar systematic reviews to be influenced by diverse factors [41]. An important point of this home management involves measuring BP [4243]. Self-care management of hypertension was instrumental when it was noted that intervention efficacy is most felt when the patient is used as the change agent [44]. Poor self-management and medication adherence were identified to negatively influence hypertension control among patients [9]. In this study, we identified multiple interactive factors that influence home-based self-care and management of BP. These factors include sociodemographic, treatment-related, and knowledge of hypertension management and risk factors. Divergent and multiple nature of the factors that are identified to influence hypertension control in Africa warrants commensurate measures to eliminate the associated repercussions of the disease. The culmination of these interventions must focus on reducing hypertension risk, ensuring medication adherence, and promoting appropriate lifestyle modifications. The divergent factors of multilevel, multicomponent interventions will ensure and promote comprehensive solutions (Mills et al., 2018), especially in lower- and middle-income countries like those in sub-Saharan Africa. This strategy is imperative to limit the challenges imposed by high BP. The sociodemographic factors that influence hypertension control through the home-based self-care ability of patients were gender (female), age (older), place of residence (urban), educational level (educated), and socioeconomic status (high). These sociodemographic characteristics are predictive of hypertension as explanatory variables [16174045464748], yet they were identified to influence the ability of patients to control hypertension. While little can be done about sociodemographic factors as a risk to hypertension (nonmodifiable risk), identifying these factors gives a category of where the emphasis should be placed on preventing and controlling hypertension. Home management education must prioritize categories that have inadequate skills in home management. To promote home-based self-care management of hypertension in lower- and middle-income countries, intervention studies must focus on identifying the influence of each factor on the care of the patient and their ability to institute home-based management of hypertension. Increasing knowledge on control measures, including limiting risk factors, and improving medication adherence through health education will be central in assisting patients in improving home-based care. In a related systematic review and meta-analysis, this study demonstrated that the number of pills and intake duration strongly influence adherence to medication at home, especially among patients with hypertension [49]. Patients who were noted to take antihypertensive medications over a long period with few medications were said to have higher adherence levels. These factors were important in other systematic reviews [495051]. It is important that when patients are discharged from the hospital to home for self-care management, only relevant (few) medications with limited pill dosage requirements will promote a positive attitude toward care and ensure hypertension control. It was noted that coordinated interventions used in managing hypertension that limit pill number and frequency among patients and increase knowledge and lifestyle changes are useful [5253]. Other important themes that emerged were the level of knowledge on hypertension and the presence of risk factors (modifiable risk factors) of hypertension. These modifiable risk factors are mainly centered on knowledge levels, smoking, alcohol, BMI, level of salt intake, and the presence of comorbidities. It was shown that even though this factor predicts the presence of hypertension, it also influences the level of treatment patients adopt during home management. The impact of these modifiable risk factors on the ability to control hypertension, especially in lower- and middle-income countries, was also documented in a previous study [545556]. To improve knowledge and, at the same time, eliminate the risk associated with hypertension, interactive technological methods will increase the likelihood of behavior change, which is implicated in the inability to achieve successful home-based self-care management [1252]. Research and policy making should streamline intervention studies that will improve home-based and self-care management of hypertension, especially in low-resource settings. In this review, we identified that important interventions that will improve self-care and home management of hypertension should focus on improving self-efficacy, increasing knowledge, and targeting specific attitudes of patients to improve adherence. Also, intervention research methods must focus on eliminating the risk factors of hypertension and segregating patients based on demographic categorization (like age, gender, socioeconomic status, and level of education) for implementation. ## Conclusion This systematic review identified and synthesized the factors associated with home-based and self-care management of hypertension. These factors are complex and multi-sectoral to improve the lives of people with hypertension; multi-sectoral approaches are their force required. Hypertension self-care interventions in lower resource settings must incorporate individual, societal, and cultural perspectives in increasing knowledge and improving home-based hypertension management. In this review, it could be seen that the use of technology-based interventions for improving home-based self-care management of hypertension is limited. This warrants the use of technology-based intervention (including social media networks and phone-based text messaging) that could improve teaching, coaching, and monitoring and give feedback to patients, especially when they engage in home-based self-care management of hypertension. Also, well-designed clinical experimental studies use complex interventions to increase knowledge and self-efficacy and improve behaviors toward home management of hypertension. ## Data Accessibility Statement All data from which the conclusions of this study were made are included in this manuscript, and no data is deposited in any public database. ## Additional File The additional file for this article can be found as follows: ## Competing Interests The author has no competing interests to declare. ## Author Contributions All the authors contributed substantially to the conception, design, search, and acquisition of data, extraction, and synthesis. All the researchers also contributed significantly to drafting and revising the manuscript for critically important intellectual content and gave final approval for publication of this manuscript in your esteemed journal. ## References 1. Salim H, Lee PY, Sazlina SG, Ching SM, Mawardi M, Shamsuddin NH. **The self-care profiles and its determinants among adults with hypertension in primary health care clinics in Selangor, Malaysia**. *PLOS One* (2019) **14**. DOI: 10.1371/journal.pone.0224649 2. Lin P-H, Appel LJ, Funk K, Craddick S, Chen C, Elmer P. **The PREMIER intervention helps participants follow the Dietary Approaches to Stop Hypertension dietary pattern and the current Dietary Reference Intakes recommendations**. *Journal of the American Dietetic Association* (2007) **107** 1541-1551. DOI: 10.1016/j.jada.2007.06.019 3. Bayliss EA, Steiner JF, Fernald DH, Crane LA, Main DS. **Descriptions of barriers to self-care by persons with comorbid chronic diseases**. *Annals of Family Medicine* (2003) **1** 15-21. DOI: 10.1370/afm.4 4. Han H-R, Lee H, Commodore-Mensah Y, Kim M. **Development and validation of the hypertension self-care profile: A practical tool to measure hypertension self-care**. *Journal of Cardiovascular Nursing* (2014) **29**. DOI: 10.1097/JCN.0b013e3182a3fd46 5. Schroeder K, Fahey T, Ebrahim S. **How can we improve adherence to blood pressure–lowering medication in ambulatory care? Systematic review of randomized controlled trials**. *Archives of Internal Medicine* (2004) **164** 722-732. DOI: 10.1001/archinte.164.7.722 6. Tan FCJH, Oka P, Dambha-Miller H, Tan NC. **The association between self-efficacy and self-care in essential hypertension: A systematic review**. *BMC Family Practice* (2021) **22** 1-12. DOI: 10.1186/s12875-021-01391-2 7. Zare S, Rezaee R, Aslani A, Shirdeli M, Kojuri J. **Moving toward community based telehealth services using mhealth for hypertensive patients**. *International Journal of Technology Assessment in Health Care* (2019) **35** 379-383. DOI: 10.1017/S0266462319000655 8. Motlagh SFZ, Chaman R, Sadeghi E, Eslami AA. **Self-care behaviors and related factors in hypertensive patients**. *Iranian Red Crescent Medical Journal* (2016) **18**. DOI: 10.5812/ircmj.35805 9. Lu X, Yang H, Xia X, Lu X, Lin J, Liu F. **Interactive mobile health intervention and blood pressure management in adults: A meta-analysis of randomized controlled trials**. *Hypertension* (2019) **74** 697-704. DOI: 10.1161/HYPERTENSIONAHA.119.13273 10. Zhou B, Bentham J, Di Cesare M, Bixby H, Danaei G, Cowan MJ. **Worldwide trends in blood pressure from 1975 to 2015: A pooled analysis of 1479 population-based measurement studies with 191 million participants**. *Lancet* (2017) **389** 37-55. DOI: 10.1016/S0140-6736(16)31919-5 11. Mallat SG, Samra SA, Younes F, Sawaya M-T. **Identifying predictors of blood pressure control in the Lebanese population—A national, multicentric survey—I-PREDICT**. *BMC Public Health* (2014) **14** 1-9. DOI: 10.1186/1471-2458-14-1142 12. van de Vijver S, Akinyi H, Oti S, Olajide A, Agyemang C, Aboderin I. **Status report on hypertension in Africa—Consultative review for the 6th Session of the African Union Conference of Ministers of Health on NCD’s**. *Pan African Medical Journal* (2014) **16**. DOI: 10.11604/pamj.2013.16.38.3100 13. Agyemang C, Bruijnzeels MA, Owusu-Dabo E. **Factors associated with hypertension awareness, treatment, and control in Ghana, West Africa**. *Journal of Human Hypertension* (2006) **20** 67-71. DOI: 10.1038/sj.jhh.1001923 14. Kayima J, Wanyenze RK, Katamba A, Leontsini E, Nuwaha F. **Hypertension awareness, treatment and control in Africa: A systematic review**. *BMC Cardiovascular Disorders* (2013) **13** 1-11. DOI: 10.1186/1471-2261-13-54 15. Buda ES, Hanfore LK, Fite RO, Buda AS. **Lifestyle modification practice and associated factors among diagnosed hypertensive patients in selected hospitals, South Ethiopia**. *Clinical Hypertension* (2017) **23** 1-9. DOI: 10.1186/s40885-017-0081-1 16. Konlan KD, Afam-Adjei CJ, Afam-Adjei C, Oware J, Appiah TA, Konlan KD. **Practice and sociodemographic factors influencing self-monitoring of blood pressure in Ghanaians with hypertension**. *International Journal of Chronic Diseases* (2020) **2020** 6016581. DOI: 10.1155/2020/6016581 17. Konlan KD, Lee H, Lee M, Kim Y, Lee H, Abdulai JA. **Risk factors associated with the incidence and prevalence of hypertension in Ghana: An integrated review (2016–2021)**. *International Journal of Environmental Health Research* (2022) 1-16. DOI: 10.1080/09603123.2022.2076814 18. Jamshidnezhad A, Kabootarizadeh L, Hoseini SM. **The effects of smartphone applications on patients self-care with hypertension: A systematic review study**. *Acta Informatica Medica* (2019) **27** 263. DOI: 10.5455/aim.2019.27.263-267 19. Ma Y, Cheng HY, Cheng L, Sit JW. **The effectiveness of electronic health interventions on blood pressure control, self-care behavioural outcomes and psychosocial well-being in patients with hypertension: A systematic review and meta-analysis**. *International Journal of Nursing Studies* (2019) **92** 27-46. DOI: 10.1016/j.ijnurstu.2018.11.007 20. Yue N, Claes S. **Wearable sensors for smart abnormal heart rate detection during skiing**. *Internet Technology Letters* (2021) **4**. DOI: 10.1002/itl2.230 21. Shin J, Konlan KD, Mensah E. **Health promotion interventions for the control of hypertension in Africa, a systematic scoping review from 2011 to 2021**. *PLOS One* (2021) **16**. DOI: 10.1371/journal.pone.0260411 22. Panic N, Leoncini E, De Belvis G, Ricciardi W, Boccia S. **Evaluation of the endorsement of the preferred reporting items for systematic reviews and meta-analysis (PRISMA) statement on the quality of published systematic review and meta-analyses**. *PLOS One* (2013) **8**. DOI: 10.1371/journal.pone.0083138 23. Moher D, Liberati A, Tetzlaff J, Altman DG, Altman D, Antes G. **Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement (Chinese edition)**. *Journal of Chinese Integrative Medicine* (2009) **7** 889-896. DOI: 10.3736/jcim20090918 24. Hong QN, Gonzalez-Reyes A, Pluye P. **Improving the usefulness of a tool for appraising the quality of qualitative, quantitative and mixed methods studies, the Mixed Methods Appraisal Tool (MMAT)**. *Journal of Evaluation in Clinical Practice* (2018) **24** 459-467. DOI: 10.1111/jep.12884 25. Hong QN, Fàbregues S, Bartlett G, Boardman F, Cargo M, Dagenais P. **The Mixed Methods Appraisal Tool (MMAT) version 2018 for information professionals and researchers**. *Education for Information* (2018) **34** 285-291. DOI: 10.3233/EFI-180221 26. Adeniyi OV, Yogeswaran P, Longo-Mbenza B, Goon DT. **Uncontrolled hypertension and its determinants in patients with concomitant type 2 diabetes mellitus (T2DM) in rural South Africa**. *PLOS One* (2016) **11**. DOI: 10.1371/journal.pone.0150033 27. Maginga J, Guerrero M, Koh E, Holm Hansen C, Shedafa R, Kalokola F. **Hypertension control and its correlates among adults attending a hypertension clinic in Tanzania**. *Journal of Clinical Hypertension* (2016) **18** 207-216. DOI: 10.1111/jch.12646 28. Gebrezgi MT, Trepka MJ, Kidane EA. **Barriers to and facilitators of hypertension management in Asmara, Eritrea: Patients’ perspectives**. *Journal of Health, Population and Nutrition* (2017) **36** 1-7. DOI: 10.1186/s41043-017-0090-4 29. Herskind J, Holmes D, Bacher K, Zelasko J. **The outpatient management of hypertension at two Sierra Leonean health centres: A mixed-method investigation of follow-up compliance and patient-reported barriers to care**. *African Journal of Primary Health Care and Family Medicine* (2020) **12** 1-7. DOI: 10.4102/phcfm.v12i1.2222 30. Pluye P, Hong QN. **Combining the power of stories and the power of numbers: Mixed methods research and mixed studies reviews**. *Annual Review of Public Health* (2014) **35** 29-45. DOI: 10.1146/annurev-publhealth-032013-182440 31. Clarke V, Braun V, Hayfield N, Smith JA. *Qualitative psychology: A practical guide to research methods* (2015) 222-248 32. Berhe DF, Taxis K, Haaijer-Ruskamp FM, Mulugeta A, Mengistu YT, Mol PG. **Hypertension treatment practices and its determinants among ambulatory patients: Retrospective cohort study in Ethiopia**. *BMJ Open* (2017) **7**. DOI: 10.1136/bmjopen-2016-015743 33. Labata BG, Ahmed MB, Mekonen GF, Daba FB. **Prevalence and predictors of self care practices among hypertensive patients at Jimma University Specialized Hospital, Southwest Ethiopia: Cross-sectional study**. *BMC Research Notes* (2019) **12** 1-8. DOI: 10.1186/s13104-019-4125-3 34. Niriayo YL, Ibrahim S, Kassa TD, Asgedom SW, Atey TM, Gidey K. **Practice and predictors of self-care behaviors among ambulatory patients with hypertension in Ethiopia**. *PLOS One* (2019) **14**. DOI: 10.1371/journal.pone.0218947 35. Aberhe W, Mariye T, Bahrey D, Zereabruk K, Hailay A, Mebrahtom G. **Prevalence and factors associated with uncontrolled hypertension among adult hypertensive patients on follow-up at Northern Ethiopia, 2019: Cross-sectional study**. *Pan African Medical Journal* (2020) **36**. DOI: 10.1155/2020/8896990 36. Gebremichael GB, Berhe KK, Beyene BG, Gebrekidan KB. **Self-care practices and associated factors among adult hypertensive patients in Ayder Comprehensive Specialized Hospital, Tigray, Ethiopia, 2018**. *BMC Research Notes* (2019) **12** 1-6. DOI: 10.1186/s13104-019-4502-y 37. Okai DE, Manu A, Amoah EM, Laar A, Akamah J, Torpey K. **Patient-level factors influencing hypertension control in adults in Accra, Ghana**. *BMC Cardiovascular Disorders* (2020) **20** 1-7. DOI: 10.1186/s12872-020-01370-y 38. Adidja NM, Agbor VN, Aminde JA, Ngwasiri CA, Ngu KB, Aminde LN. **Non-adherence to antihypertensive pharmacotherapy in Buea, Cameroon: A cross-sectional community-based study**. *BMC Cardiovascular Disorders* (2018) **18** 1-9. DOI: 10.1186/s12872-018-0888-z 39. Ware LJ, Chidumwa G, Charlton K, Schutte AE, Kowal P. **Predictors of hypertension awareness, treatment and control in South Africa: Results from the WHO-SAGE population survey (Wave 2)**. *Journal of Human Hypertension* (2019) **33** 157-166. DOI: 10.1038/s41371-018-0125-3 40. Mohamed SF, Mutua MK, Wamai R, Wekesah F, Haregu T, Juma P. **Prevalence, awareness, treatment and control of hypertension and their determinants: results from a national survey in Kenya**. *BMC Public Health* (2018) **18** 1-10. DOI: 10.1186/s12889-018-6052-y 41. Mills KT, Bundy JD, Kelly TN, Reed JE, Kearney PM, Reynolds K. **Global disparities of hypertension prevalence and control: A systematic analysis of population-based studies from 90 countries**. *Circulation* (2016) **134** 441-450. DOI: 10.1161/CIRCULATIONAHA.115.018912 42. Roerecke M, Kaczorowski J, Tobe SW, Gmel G, Hasan OS, Rehm J. **The effect of a reduction in alcohol consumption on blood pressure: A systematic review and meta-analysis**. *Lancet Public Health* (2017) **2** e108-e20. DOI: 10.1016/S2468-2667(17)30003-8 43. Türker F, Sahın T, Oral A, Koçak E, Türker BÇ, Niğdelioğlu A. **Evaluation of predisposing metabolic risk factors for portopulmonary hypertension in patients with NASH cirrhosis**. *International Journal of General Medicine* (2022) **15** 859-865. DOI: 10.2147/IJGM.S339474 44. Kaboli PJ, Howren MB, Ishani A, Carter B, Christensen AJ, Vander Weg MW. **Efficacy of patient activation interventions with or without financial incentives to promote prescribing of thiazides and hypertension control: A randomized clinical trial**. *JAMA Network Open* (2018) **1**. DOI: 10.1001/jamanetworkopen.2018.5017 45. Craig LS, Gage AJ, Thomas AM. **Prevalence and predictors of hypertension in Namibia: A national-level cross-sectional study**. *PLOS One* (2018) **13**. DOI: 10.1371/journal.pone.0204344 46. Mamdouh H, Alnakhi WK, Hussain HY, Ibrahim GM, Hussein A, Mahmoud I. **Prevalence and associated risk factors of hypertension and pre-hypertension among the adult population: Findings from the Dubai Household Survey, 2019**. *BMC Cardiovascular Disorders* (2022) **22** 1-9. DOI: 10.1186/s12872-022-02457-4 47. Okello S, Muhihi A, Mohamed SF, Ameh S, Ochimana C, Oluwasanu AO. **Hypertension prevalence, awareness, treatment, and control and predicted 10-year CVD risk: A cross-sectional study of seven communities in East and West Africa (SevenCEWA)**. *BMC Public Health* (2020) **20** 1-13. DOI: 10.1186/s12889-020-09829-5 48. Manios Y, Karatzi K, Moschonis G, Ioannou G, Androutsos O, Lionis C. **Lifestyle, anthropometric, sociodemographic and perinatal correlates of early adolescence hypertension: The Healthy Growth Study**. *Nutrition, Metabolism and Cardiovascular Diseases* (2019) **29** 159-169. DOI: 10.1016/j.numecd.2018.10.007 49. Parati G, Kjeldsen S, Coca A, Cushman WC, Wang J. **Adherence to single-pill versus free-equivalent combination therapy in hypertension: A systematic review and meta-analysis**. *Hypertension* (2021) **77** 692-705. DOI: 10.1161/HYPERTENSIONAHA.120.15781 50. Reeves L, Robinson K, McClelland T, Adedoyin CA, Broeseker A, Adunlin G. **Pharmacist interventions in the management of blood pressure control and adherence to antihypertensive medications: A systematic review of randomized controlled trials**. *Journal of Pharmacy Practice* (2021) **34** 480-492. DOI: 10.1177/0897190020903573 51. Hermida RC, Mojón A, Hermida-Ayala RG, Smolensky MH, Fernández JR. **Extent of asleep blood pressure reduction by hypertension medications is ingestion-time dependent: Systematic review and meta-analysis of published human trials**. *Sleep Medicine Reviews* (2021) **59** 101454. DOI: 10.1016/j.smrv.2021.101454 52. Abel WM, Spikes T, Greer DB. **A qualitative study: Hypertension stigma among black women**. *Journal of Cardiovascular Nursing* (2021) **36** 96-103. DOI: 10.1097/JCN.0000000000000759 53. Salam A, Atkins ER, Hsu B, Webster R, Patel A, Rodgers A. **Efficacy and safety of triple versus dual combination blood pressure-lowering drug therapy: A systematic review and meta-analysis of randomized controlled trials**. *Journal of Hypertension* (2019) **37** 1567-1573. DOI: 10.1097/HJH.0000000000002089 54. Debon R, Coleone JD, Bellei EA, De Marchi ACB. **Mobile health applications for chronic diseases: A systematic review of features for lifestyle improvement**. *Diabetes & Metabolic Syndrome: Clinical Research & Reviews* (2019) **13** 2507-2512. DOI: 10.1016/j.dsx.2019.07.016 55. Debon R, Bellei EA, Biduski D, Volpi SS, Portella MR. **Effects of using a mobile health application on the health conditions of patients with arterial hypertension: A pilot trial in the context of Brazil’s family health strategy**. *Scientific Reports* (2020) **10** 1-10. DOI: 10.1038/s41598-020-63057-w 56. Yusuf S, Joseph P, Rangarajan S, Islam S, Mente A, Hystad P. **Modifiable risk factors, cardiovascular disease, and mortality in 155 722 individuals from 21 high-income, middle-income, and low-income countries (PURE): A prospective cohort study**. *Lancet* (2020) **395** 795-808. DOI: 10.1016/S0140-6736(19)32008-2
--- title: 'Exploring Medical Students’ Learning Around Uncertainty Management Using a Digital Educational Escape Room: A Design-based Research Approach' authors: - Jenny Moffett - Dara Cassidy - Naoise Collins - Jan Illing - Marco Antonio de Carvalho Filho - Harold Bok journal: Perspectives on Medical Education year: 2023 pmcid: PMC10038110 doi: 10.5334/pme.844 license: CC BY 4.0 --- # Exploring Medical Students’ Learning Around Uncertainty Management Using a Digital Educational Escape Room: A Design-based Research Approach ## Abstract ### Introduction: Medical professionals meet many transitions during their careers, and must learn to adjust rapidly to unfamiliar workplaces and teams. This study investigated the use of a digital educational escape room (DEER) in facilitating medical students’ learning around managing uncertainty in transitioning from classroom to clinical placement. ### Methods: We used design-based research to explore the design, build, and test of a DEER, as well as gain insight into how these novel learning environments work, using Community of Inquiry (CoI) as a guiding conceptual framework. This study represented a mixed methods pilot test of a prototype DEER. Twenty-two medical students agreed to participate, and data were collected through qualitative (i.e., focus groups, game-play observations) and quantitative (i.e., questionnaires) methods. ### Results: Eighty-two per cent of participants agreed or strongly agreed that the DEER supported their learning around uncertainty. Participants offered diverse examples of how the game had facilitated new insights on, and approaches to, uncertainty. With respect to the learning environment, multiple indicators and examples of the three domains of CoI – cognitive, teaching and social presence – were observed. ### Discussion: Our findings suggested that DEERs offer a valuable online learning environment for students to engage with complex and emotion-provoking challenges, such as those experienced at transitions. The study also suggested that CoI can be applied to the design, implementation, and evaluation of DEER learning environments, and we have proposed a set of design principles that may offer guidance here. ## Introduction Medical professionals meet many transitions during their careers, and must learn to adjust rapidly to unfamiliar workplaces and new teams. Such profound changes begin in medical school; an early and important example of this is the transition from pre-clinical to clinical training. This step into ‘real-world’ medicine represents an exciting and rewarding time for medical students [1]. However, it is also a step into the unknown, with the potential to evoke experiences of stress and uncertainty [23]. Although many supports exist which address the knowledge and practical skills needed for clinical placements (e.g., special-purpose courses, clinical skills training), these can fall short in preparing students for ‘the dynamics of a new environment, which itself is unstable’ [3, p.566]. With healthcare practice becoming increasingly complex and unpredictable [4], it is important to better prepare students to engage with dynamic clinical learning environments. In recent years, there has been an increased interest in how medical professionals manage uncertainty, both at transitions and more generally [5]. The evidence highlights that health professionals’ responses to uncertainty can influence their decision-making skills [6], attitudes to patients [7], career choices [8], and experiences of work-related stress [910]. More recent research also suggests that it may be possible to train medical students to prepare for uncertainty [11]. Clinical debriefs, simulations, and peer-to-peer conversations have been proposed as pedagogical approaches that may help students to better manage the uncertainty of clinical practice [1213]; however, there is little empirical research in this domain. This study explores the use of a type of simulation-based educational game known as an escape room to facilitate medical students’ learning around uncertainty experienced at the transition from classroom into clinical settings. Escape rooms are ‘live-action team-based game where players discover clues, solve puzzles, and accomplish tasks in one or more rooms in order to accomplish a specific goal… in a limited amount of time’ [14]. Educational escape rooms have rapidly become popular within health professions’ education [15]. A variety of studies have explored the capacity of escape rooms to facilitate learning in clinical [1617181920] and professional skills [212223] domains. Research, however, is at an early stage with relatively little known about how learning takes place within these novel environments [24]. Escape rooms can be held within face-to-face or virtual learning environments where, in the latter case, they are referred to as digital educational escape rooms (DEERs). In this study, we built a DEER in order to explore how this learning environment might be used to facilitate medical students’ learning around uncertainty, as well as to gain more general insight as to how escape rooms work. We selected the Community of Inquiry (CoI) model [25] as a guiding conceptual framework, and a lens with which to investigate the DEER learning environment. CoI is a widely studied online learning model [2627], that can help researchers to conceptualise ‘the educational transaction and processes of learning’ in online settings [28, p.9]. The framework (Table 1) proposes that meaningful online learning arises through the development of three overlapping domains [29]: **Table 1** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | | --- | --- | --- | | ELEMENTS | CATEGORIES | INDICATORS | | Cognitive | Triggering event | Having a sense of puzzlement | | Cognitive | | | | Cognitive | Exploration | Exchanging information | | Cognitive | | | | Cognitive | Integration | Connecting ideas | | Cognitive | | | | Cognitive | Resolution | Applying new ideas | | Teaching | Design and organisation | Setting curriculum and methods | | Teaching | | | | Teaching | Facilitation of discourse | Sharing personal meaning | | Teaching | | | | Teaching | Direct instruction | Focusing discussion | | Social | Open communication | Enabling risk-free communication | | Social | | | | Social | Group cohesion | Encouraging collaboration | | Social | | | | Social | Affective expression | Expressing emotions, camaraderie | CoI adopts a collaborative-constructivist stance [33], making it a framework of particular interest for the team-based DEER learning environment [3435]. However, there is limited empirical research here too. Thus, our research questions for this study were: To explore these research questions, we used a design-based research (DBR) approach. DBR is ‘a systematic but flexible methodology aimed to improve educational practices through iterative analysis, design, development, and implementation, based on collaboration among researchers and practitioners in real-world settings’ [36]. A key tenet of DBR is that it holds dual goals: the research should facilitate the development of a specific innovation or intervention, whilst also testing and refining theories to gain insight into complex learning environments [37]. Although there is great variety in how DBR is implemented, this approach typically involves four stages: analysis of the problem, design of solutions, testing and iteration, and reflection [38]. In this study, we used DBR to design, build and test our DEER in an online setting whilst simultaneously furthering our understanding of the applications of CoI in this context. ## Study design DBR involves the development and evaluation of multiple prototypes. An initial prototype DEER underwent evaluation [39] and data from that design cycle was used to inform the build for this second prototype (Figure 1). The current study explores a design cycle where the second prototype escape room was pilot-tested using a convergent parallel mixed methods study design [40]. We used qualitative (i.e., focus groups, game-play observations) and quantitative (i.e., questionnaires) data collection methods, with an emphasis on the qualitative strand [41]. Ethical approval for the study was granted by the RCSI Research and Ethics Committee, RCSI University of Medicine and Health Sciences (ID 202103004). **Figure 1:** *Data from a preliminary design cycle was used to inform the build for a second prototype (adapted from McKenney & Reeves, 2012. [42]).* ## Context The study took place at RCSI University of Medicine and Health Sciences, a culturally diverse institution with over 4,000 students from 90 different countries. The university offers a direct-entry medical programme with two pre-clinical (Years 1–2) and three clinical years (Years 3–5). Our study population consisted of students enrolled in Year 2 of the programme in advance of their commencing clinical placements. All students within this cohort were eligible to participate, and recruitment was promoted via university email and social media. An incentive to take part, entry into a draw for a book voucher, was offered. Study participants were invited to play a prototype DEER in October 2021. This prototype had been build using draft principles derived from the first design cycle and a review of the CoI research literature (Table 2). The DEER was designed to be played by small groups (4–5) of students, and it was intended that students would work together to overcome ambiguity, solve puzzles and ‘escape’ a fictional creepy hospital [39]. The DEER consisted of ten puzzles, including numerical, word-based, logic, and general knowledge formats, and three in-game reflections, which were built on an interactive content authoring platform (Genially; Madrid, Spain). Individual puzzles were designed to align with sources of uncertainty in healthcare that have been identified by Han et al. [ 43]. This meant that participants met puzzles which involved managing complex information, recognising ambiguity, and working with the different outcomes that can emerge in medicine (i.e., patient gets better, or patient does not). Although participants could follow different pathways within the DEER, all groups needed to complete a final, culminating ‘meta-puzzle’ to complete the game. **Table 2** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | | --- | --- | --- | | COI PRESENCE | DESIGN PRINCIPLES | REFERENCES | | Cognitive | Use an engaging storyline that evokes curiosity for learnersAlign escape room puzzles with educational learning outcomesProvide challenging puzzles that provoke shared reflection | Garrison, 2016 [44]; Garrison, 1999 [45]; Redmond, 2014 [46] | | Teaching | Provide clear instructions to learners before gameUse facilitation skills to establish a safe, supportive learning environmentOffer scaffolded support to learners throughout (e.g., pre-brief, hint strategy, technical support, de-brief) | Cheng, 2020 [47]; Garrison, 2016 [44]; McKerlich, 2007 [48]; Shea, 2010 [49] | | Social | Use web-conferencing software with breakout room capability to facilitate small group interactionsEmploy collaborative rather than competitive game strategies (e.g., escape against clock rather than ‘first team to escape wins’)Use puzzles to evoke emotions such as confusion and excitement | Fayram, 2017 [50]; Garrison, 2016 [44]; Lowenthal, 2014 [51]; Moallem, 2015 [52] | Prior to game-play, participants were given details of the DEER, a participant information sheet and a consent form (Figure 2). On the day of game-play, participants joined the session via Microsoft Teams (Redmond, WA, USA). There was a short introduction, or pre-brief, before participants were asked to join breakout rooms and begin the activity. The pre-brief aimed to establish psychological safety by providing clear instructions for game play, as well emphasising the fun element and the availability of help for overcoming “roadblocks” encountered during the game [47]. Each group was allocated 50 minutes of game play, and participants were directed to play as a team, appointing leaders to ‘share screens’ and input answers. After the allocated time, breakout rooms were closed, and a de-brief with the full cohort of students was held. The de-brief was designed to allow participants an opportunity to disclose and discuss the uncertainties that arose for them, as well as other experiences that they felt were important. The de-brief also offered a space for the participants to engage in shared reflection around the key learning outcomes from game play, including the in-game reflections. Finally, an email with uncertainty management resources and a link to the DEER was sent to participants after the session. **Figure 2:** *Flow chart of the study design.* ## Qualitative data collection Qualitative data were collected during game-play and immediately afterwards through focus group discussions. Game-play and break-out rooms were video-recorded and the audio component transcribed. Text from the session web-chat as well as observational data (e.g., the actions of the participants) were also recorded. The focus group discussions, facilitated by experienced researchers using a pre-determined question guide (Appendix A), were also video-recorded and the audio transcribed. Focus group participants were also invited to submit text responses via a digital noticeboard, Padlet, to ensure everyone had the opportunity to provide feedback (Padlet; San Francisco, CA, USA), which were collected. ## Quantitative data collection Quantitative data were collected before and after the game-play session through use of pre- and post-intervention questionnaires, via an online survey platform (SurveyMonkey; San Mateo, CA, USA). Data collection was intended to capture any impact of game-play on participants’ uncertainty tolerance. The pre-intervention questionnaire (Appendix B) consisted of: the Intolerance of Uncertainty Scale (Short Form) (IUS-12), a 12-item questionnaire which assesses individuals’ perceptions of uncertainty and which has previously demonstrated high internal consistency (α = 0.91) with medical student cohorts [53]; the Tolerance for Ambiguity (TFA) Scale, a 7-item questionnaire which assesses individuals’ tolerance of general uncertainty in life and which has demonstrated acceptable internal consistency with cohorts of medical students (α = 0.75) [54]; and, a set of demographic questions. The post-intervention questionnaire (Appendix C) consisted of repeats of the IUS-12 and TFA, alongside a 12-item escape room perception survey adapted from Eukel et al. [ 55]. ## Qualitative data analysis Two separate qualitative data analyses were carried out. The first analysis explored the focus group transcriptions and digital noticeboard text. Here, data were combined and organised using NVivo 12 (QSR International; Melbourne, Australia), and examined using a reflexive thematic analysis approach [56]. The researchers used an initial inductive step to understand the experiences of the students in relation to the escape room. JM listened to the audio data, and then read and re-read the transcribed recordings. JM then created initial codes, which were specifically related to participants’ perspectives of using a DEER to facilitate learning around uncertainty. JM then applied a subsequent step of deductive analysis whereby the data was examined with respect to the social, cognitive and teaching presences of CoI. Following several passes through the data, themes were identified, refined and re-organised before final agreement with the research team (JM, DC & JI). The second analysis explored the game-play transcriptions, web-chat and qualitative observational data. Here, JM and DC used a CoI instrument adapted from McKerlich & Anderson [48] to examine the data. This involved viewing the session videos twice, reading and re-reading the session transcripts and web-chat text, before discussing and documenting indicators and examples of social, cognitive and teaching presences. The researchers drew on existing CoI research [495157] to help define boundaries around the presences. ## Quantitative data analysis Quantitative data were analysed in two stages. First, the pre- and post-intervention surveys items were analysed. Internal consistency was assessed by calculating Cronbach’s coefficient alpha for each [58], and a Shapiro-Wilks test was used to assess the normality of the resulting data. A paired-design t-test was used to determine if there was a significant difference between the scores on the IUS-12 scale and the TFA scale pre- versus post-intervention. A separate test was carried out for each of the scales and alpha was set at 0.05. Second, a one-sample t-test and descriptive statistics were used to explore responses to the escape room perception survey. The perception survey was measured on a five-point Likert scale ranging from ‘1 = strongly disagree’ to ‘5 = strongly agree’, with items 9 and 10 of the survey reverse-scored. The one-sample t-test assessed whether students’ mean (SD) perception score was significantly different to the mean value of the scale, ‘3 = not agree nor disagree’. All statistical analyses were carried out using STATA statistical package version 17 (StataCorp; Texas, USA). ## Reflection As a final stage of data analysis, the research team (HB, MDF, JM and DC) met to discuss the data in relation to the initial DEER design principles. The researchers examined the data through the lens of the CoI framework and engaged in shared reflection, with the aim of co-constructing an updated set of design principles. Following analysis of the data and a process of shared reflection, the research team co-constructed a list of revised design principles for DEERs that are underpinned by the CoI framework (Table 4). **Table 4** | Unnamed: 0 | Unnamed: 1 | | --- | --- | | COI PRESENCE | DESIGN PRINCIPLES | | Cognitive | Use an engaging storyline that evokes curiosity for learners | | Cognitive | | | Cognitive | Explicitly align escape room puzzles with meaningful/purposeful learning outcomes | | Cognitive | | | Cognitive | Provide challenging puzzles aligned with learners’ developmental levels which provoke shared reflection | | Teaching | Open the game with a pre-brief which provides clear instructions, encourages engagement and establishes a safe, supportive and playful learning environment | | Teaching | | | Teaching | During the game, maintain learner engagement through responsive facilitation (e.g., technical support), and effective game design (e.g., hint strategy) | | Teaching | | | Teaching | After the game, use a debrief to help learners to make sense of the activity, facilitating the resolution phase of cognitive presence as well as emotional closure for learners | | Teaching | | | Teaching | Encourage engagement and peer learning through consideration of small group size and composition, and team-work strategy | | Teaching | | | Teaching | Assist learners who are not familiar with each other to build rapport (e.g., through introductions and ice-breakers) | | Teaching | | | Teaching | Ensure that game play and the ‘rules of engagement’ align with the intended cognitive process, learners’ behaviour, and learning outcomes | | Social | Use web-conferencing software with breakout room capability to facilitate small group interactions | | Social | | | Social | Employ complementary game strategies, from social collaboration to healthy competition, optimising learners’ engagement | | Social | | | Social | Use puzzles to evoke emotions that increase arousal and positively impact on cognitive presence | ## Results Our results are organised in two sections. First, we report findings that relate to our first and second research questions, i.e. exploring the use of a DEER in relation to medical students’ learning around, and tolerance of, uncertainty. Second, we report findings that relate to our third research question (i.e., investigating the CoI as a framework of relevance in understanding DEER learning environments). Twenty-two second year pre-clinical undergraduate medical students (10 female and 12 male students) agreed to participate in the study. Participant quotes, with details on focus group, gender and participant number (e.g., FG1F1), have been provided. ## Qualitative data Ten participants (4 female; 6 male students) participated in two focus group discussions. Data analysis of the focus groups identified two themes that related to the participants’ perspectives on using a DEER to facilitate learning around uncertainty: affective experiences of uncertainty, and building approaches to uncertainty. The participants highlighted that the DEER learning environment provided multiple opportunities for affective experiences of uncertainty. They noted that playing the game felt inherently uncertain due to the challenges of the puzzles and the ambiguous clues. Others felt unsure about what the game would entail, and whether it would represent a good use of their time. Further to this, participants reported uncertainty in relation to working with new and unfamiliar colleagues. Some participants expressed self-doubt and a sense of vulnerability in relation to their abilities (i.e., whether or not they would be able to complete the game, or contribute to the team). ‘ I don’t know if I need tonnes of outside knowledge and all? I don’t want to be the weak person, throwing out stuff that’s completely left field and not at all correct.’ ( FG1F1) One group reported experiences of uncertainty due to a technology breakdown (i.e., lagging internet connection). Overall, participants spoke of uncertainty in terms of a variety of different emotional states including anxiety, frustration, curiosity and excitement. ‘ I’ve never actually ever come across something like this escape room… I was pretty curious and anxious, like what it is we will actually do?’ ( FG1M1) The participants discussed ways in which the DEER had helped them to think differently about uncertainty. They highlighted new strategies in managing uncertainty, such as adopting a team approach (i.e., harnessing different perspectives). The validation and support of others helped them to propose ideas and solutions, despite feeling unsure. ‘ A lot of moments I was confused and didn’t know what to do and they backed me up. Individually, we didn’t know everything. This is something we all need to learn, it’s an important student experience. It was like a metaphor for diagnosing patients.’ ( FG1M2) Participants also reported that the game had helped them to engage with critical thinking and creative approaches to problem solving. Others alluded to shifts from negative to more positive mind-sets around uncertainty. ‘ There will be times when we will be uncertain so it shouldn’t be a factor that makes us feel uncomfortable. It should be a motivating factor to learn more.’ ( FG1M3) However, not all participants agreed that they had learned about uncertainty. Some felt that the puzzles did not reflect the uncertainty experienced in real-world, clinical practice. Others commented that the learning was not linked to their course work, and thus seemed less relevant to them. These views were predominant within the group who had experienced technology problems. ‘I just feel like, have we really learnt anything by playing the game?’ ( FG2F1) ## Quantitative data Sixteen participants ($\frac{16}{22}$, $73\%$ of the study cohort) completed both the pre-intervention and post-intervention questionnaires. The reliability was high for the IUS-12 scale (Cronbach’s alpha = 0.89) and acceptable for the TFA-scale (Cronbach’s alpha = 0.74). The data were found to be normally distributed on the Shapiro-Wilks test. No significant difference in Intolerance of Uncertainty ($t = 0$, df = 15, p-value = 1) nor Tolerance of Ambiguity (t = –0.81, df = 15, p-value = 0.43) was detected between the pre-intervention and post-intervention groups. With respect to the escape room perceptions survey, 17 participants submitted responses ($77\%$ of the study cohort) (Table 3). The mean perception value for the cohort ($m = 3.99$ +/– 0.59 sd) on a five-point evaluation scale was significantly higher than the neutral point [3] of the evaluation scale ($t = 6.98$, df = 16, $p \leq 0.01$). This suggests that the students’ perceived learning through the escape room was strongly positive. **Table 3** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Unnamed: 5 | Unnamed: 6 | | --- | --- | --- | --- | --- | --- | --- | | ITEM | MEAN (SD) | STRONGLY DISAGREE (%) | DISAGREE (%) | NEUTRAL (%) | AGREE (%) | STRONGLY AGREE (%) | | 1. The escape room encouraged me to think about material in a new way | 3.7 (0.3) | 5.9 | 5.9 | 23.5 | 41.2 | 23.5 | | 2. I would recommend this activity to other students | 4.5 (0.2) | 0 | 5.9 | 0 | 35.3 | 58.8 | | 3. I learned from my peers during the uncertainty escape room | 4.3 (0.1) | 0 | 0 | 5.9 | 58.8 | 35.3 | | 4. The escape room was an effective way to review the topic of uncertainty | 3.6 (0.3) | 11.8 | 11.8 | 0 | 58.8 | 17.6 | | 5. The escape room was an effective way to learn new information related to uncertainty | 3.5 (0.3) | 5.9 | 17.6 | 11.8 | 52.9 | 11.8 | | 6. I learn better in a game format than in a lecture | 4.6 (0.1) | 0 | 0 | 0 | 41.2 | 58.8 | | 7. The escape room was an effective way to assist my learning around managing uncertainty | 3.7 (0.3) | 5.9 | 11.8 | 5.9 | 58.8 | 17.6 | | 8. I feel I was able to engage with my teammates to learn new material | 4.0 (0.2) | 0 | 5.9 | 5.9 | 70.6 | 17.6 | | 9. It was difficult for me to focus on learning because I was feeling stressed or overwhelmed | 3.9 (0.2)* | 29.4 | 41.2 | 17.6 | 11.8 | 0 | | 10. The non-educational portions (e.g., puzzles, etc.) distracted me from learning about uncertainty | 3.4 (0.3)* | 11.8 | 47.0 | 17.6 | 11.8 | 11.8 | | 11. I prefer assembling information from a variety of sources when learning new material | 4.1 (0.2) | 0 | 5.9 | 11.8 | 47.0 | 35.3 | | 12. In general, I enjoy playing games (video games, board games, social media games, etc.) | 4.8 (0.1) | 0 | 0 | 0 | 23.5 | 76.5 | The majority of participants ($$n = 14$$, $82\%$) agreed or strongly agreed that the escape room was an effective way to assist their learning around managing uncertainty. Ninety-four per cent of the participants agreed or strongly agreed that they had learned from their peers during the game-play session. Finally, $94\%$ of participants agreed or strongly agreed that they would recommend the game to other students. ## CoI as a framework of relevance in understanding DEER learning environments Data collected during the focus group discussions and the game-play sessions were categorised according to the presences of CoI: cognitive, teaching and social [25]. These data are presented in Appendix D. ## Focus group data Within the focus groups, participants highlighted several aspects of the escape room experience that appeared to be consistent with CoI. With respect to social presence, they reported that the game provided a warm environment that supported team interaction. They felt validated, supported and motivated by each other during game play, and reported a wide range of affective experiences including: curiosity, enjoyment, excitement, fun, pride, relief, satisfaction, annoyance, anxiety, confusion, exasperation, and frustration (Table 3). With respect to teaching presence, participants noted the role of the instructor in: setting the tone for the game; establishing team collaboration; offering clear instructions; providing guidance and technical help; supporting insights around uncertainty; and re-emphasising the game’s learning outcomes (Table 3). One aspect of the game’s design that evoked mixed opinions was the ‘race against the clock’ time strategy. Some participants reported that the time pressure added to the fun, and helped them to establish trust within their team quickly. Others said that time pressure caused them to rush through the game, sometimes progressing without fully understanding a topic. With respect to cognitive presence, there were relatively fewer comments. Although many participants reported that the game had involved them in cognitive effort, there appeared to be variation in how deeply they engaged with the puzzles. Many participants commented that the in-game reflective activities broke their sense of flow and immersivity within the game. ## Game-play data Qualitative data collected during game-play also highlighted multiple indicators and examples of cognitive, teaching and social presence within the DEER (Table 3). With respect to cognitive presence, participants seemed to share information, connect ideas and test theories with each other. Cognitive presence appeared to be most salient during puzzle-solving interactions. Teaching presence was observed in the planning and organisation of the DEER as well as through facilitation of discourse and direct instruction, which could be subdivided into facilitator and peer categories. Teaching presence related to the facilitator was dominant in the pre- and de-brief sections, whereas teaching presence related to the participants was dominant within the breakout rooms. Social presence was observed during all stages of the session with multiple examples of open communication, group cohesion and affective expression. With regards to the latter, many overt expressions of uncertainty were observed within the peer interactions. ## Discussion This study sought to explore medical students’ perspectives on the use of a DEER to facilitate learning around managing uncertainty at the transitions from classroom to clinical placement, and what impact, if any, a DEER has on students’ uncertainty tolerance. Our findings suggest that DEERs generate an engaging online learning environment that allows medical students to meet with uncertainty in a safe and constructive manner. Many of these uncertainties appear to resonate with those experienced by medical students at clinical transitions (i.e., making sense of ambiguous information, engaging in decision-making under time pressure, and building trust quickly with unfamiliar people). Although at least some of the uncertainty was evoked through the novelty of the DEER, which may decrease as students become more acquainted with such strategies, the game seemed to provoke relevant affective states and offer a supportive environment that facilitated shared disclosure. Our findings also suggest that the DEER had facilitated learning around uncertainty management. The majority of students perceived that the DEER had assisted their learning, whilst the focus group discussions revealed examples of students’ insights and approaches to managing uncertainty. For example, students reported that they held a better understanding of the different strengths and perspectives a team can bring to meet a challenging situation, again a finding that translates well into the clinical setting. However, not all students enjoyed, or perceived that they had learned from, the DEER. For example, students that had encountered technology problems during game-play were less positive about the experience overall. This highlights that issues such as internet access and digital skills represent an important challenge for DEERs in comparison to physical escape rooms. Furthermore, quantitative data analysis found no evidence that the DEER had had an impact on the students’ uncertainty tolerance. It may be that a once-off intervention or a short interval between measurement was insufficient to detect a change in students’ responses. The small cohort of this pilot study makes it difficult to draw firm conclusions. We also set out to explore whether or not the CoI framework could facilitate our understanding of DEER learning environments, and, if so, what indicators of social presence, teaching presence and cognitive presence exist. Our findings strongly suggest that CoI has a natural resonance with DEER learning environments, and that the framework can shed light on how learning takes place in such novel online settings. We also found evidence of cognitive, teaching and social presences that we will discuss in relation to the existing literature below. Social presence, which relates to open communication, emotional expression, and group cohesion [51], was widely evident within the participants’ interactions. The DEER seemed to encourage rapid rapport and trust building, and despite some initial hesitation about playing the game with unfamiliar individuals, they quickly settled into teamwork. This was particularly apparent in the breakout rooms where, in the absence of the instructor, participants engaged in supportive, informal and humour-filled verbal communication. This finding supports previous CoI research [59, p.6], which suggests that ‘synchronous communications can be especially useful in quickly establishing, building and modeling social presence.’ There were also many, varied expressions of affective experiences during game play. Aside from uncertainty, students reported feeling enjoyment, humour, curiosity and pride, as well as anxiety and frustration. These findings support evidence that DEERs can offer learners opportunities ‘to deal with and overcome intense negative emotions, in particular fear or disgust, to move forward’ [60, p.16], which may be particularly useful in preparing medical students for ‘emotion-laden’ clinical experiences [61, p.198]. Teaching presence was also evident within the escape room environment, with different aspects apparent at different stages of the game. For example, teaching presence centred on the instructor during preparation for the game and within the pre- and de-brief sections. Teaching presence centred on peer interaction was most apparent in the small-group breakout rooms. This finding underlines a view within CoI research that ‘the term for this component of the CoI is ‘teaching’ and not ‘teacher’ presence. This provides room for, and encourages, students to take a positive and visible role in the learning of their peers’ [59, p.7], The extension of teaching presence to embrace students as teachers has been proposed as a ‘vital question’ which should be addressed as the CoI model matures [62, p.27] Our findings suggest that DEERs can provide a valuable learning environment for peer learning which may help student to understand the salience of ‘building relationships with staff, peers or near-peers’ in clinical settings [3, p.566]. ‘ Students as teachers’ also hints at a potential for DEER activities be scaled up, offering an effective vehicle for active learning in online, large group classrooms. To do so, it may be helpful for educational game designers to consider including opportunities for students to take on instructional roles when planning game-play strategies. Indicators and examples of cognitive presence were also apparent within the DEER, although fewer in number. This is not surprising considering that cognitive presence, which represents a critical-thinking process that switches ‘between the public shared world and the private reflective world’ [25, p.21], can be hard to observe. Here, it appeared that the emotional arousal elicited by the puzzles drew most students into a cycle of cognitive activity. At times this activity seemed aligned with the deep processes involved in cognitive presence but, at others, it seemed more superficial. It is worth highlighting that lively interaction may be present in a learning environment, but if it does not support participants to integrate ideas into meaningful constructs, it does not represent the existence of cognitive presence [63]. This finding may be due to the design of this specific DEER, i.e. here the aim was to provoke experiences of, and reflections on, uncertainty, rather than present content material that provoked deeper cognitive processing. Nonetheless, our results suggest that strong alignment of game-play and puzzle content with learning outcomes is advisable. Other elements of the game design also seemed to impact on cognitive presence. For example, the in-game reflective activities encouraged some students to engage in shared reflection, whilst triggering annoyance and frustration for others. Furthermore, the game’s time strategy seemed to impact on the students’ approaches to puzzles in different ways. Some groups found the time limit exciting, whilst others experienced it as pressure, causing them to skip over the activities. This tension between achieving game goals and engaging in deep, reflective learning in a time-constrained game environment has been highlighted in the literature [6465]. Thus, whilst our findings suggest that DEERs offer advantages in keeping learners ‘on-task’ in the online setting, care must be taken to ensure that puzzles and game-play align with intended learner behaviour and meaningful learning outcomes, which award students with a ‘sense of purpose’. For example, a limited-time strategy that encourages students to ‘race to the finish’ might be useful for exploring a clinical scenario where quick action is required (e.g., managing sepsis); however, the sense of urgency this evokes may divert students away from the sustained communication required for cognitive presence [65]. ## Limitations and future research Our study population represented a small convenience sample of medical students. It is likely that our participants were inherently interested in educational escape games, and a larger cohort of participants may have led to different findings. A larger sample size would also be helpful in identifying any statistically significant changes between the pre- and post-intervention questionnaire responses. To deepen our understanding of how the CoI framework can be used in the design and implementation of DEERs, we recommend that further research is carried out in different contexts, with different DEER formats and diverse populations of students. A future prototype of this DEER will be incorporated into the medical programme at RCSI University of Medicine and Health Sciences, providing an opportunity to test our proposed design principles, and to evaluate the scalability of the intervention in a large group classroom. More broadly, this study highlights the opportunities provided by DBR in supporting the development of educational resources, alongside gaining insight as to how these operate within specific learning environments. DBR may be of specific interest to health professions’ educators who wish to investigate the application of innovations such as virtual reality, augmented reality and artificial intelligence within real-world settings. ## Conclusion Overall, our study suggests that DEERs offer a suitable learning environment for medical students to engage with complex, team-based and emotion-provoking challenges, such as those experienced in the transition from pre-clinical to clinical training. Our findings also support the value of CoI as a lens through which the DEER learning environment can be explored. The framework has highlighted important considerations in the advancement of this specific prototype, as well as offering more general guidance with respect to the cultivation of engaging, collaborative DEER learning environments. We concur with McKerlich and Anderson’s [48, p.48] assertion that CoI offers a valuable way to ‘describe and assess educational experiences and contexts’. As research around game-based learning and simulation games expands, these approaches are likely to gain ground on more traditional pedagogical methods in health professions’ education. ## Funding Information This project was supported by funding from the Irish Medical Council/INHED Research in Medical Education (RIME) Award, and from the RCSI StEP (Student Engagement and Partnership) Framework. ## Competing Interests The authors have no competing interests to declare. ## References 1. Teunissen PW, Westerman M. **Opportunity or threat: the ambiguity of the consequences of transitions in medical education**. *Med Educ* (2011) **45** 51-9. DOI: 10.1111/j.1365-2923.2010.03755.x 2. Surmon L, Bialocerkowski A, Hu W. **Perceptions of preparedness for the first medical clerkship: a systematic review and synthesis**. *BMC Med Educ* (2016) **16** 1-1. DOI: 10.1186/s12909-016-0615-3 3. Atherley A, Dolmans D, Hu W. **Beyond the struggles: a scoping review on the transition to undergraduate clinical training**. *Med Educ* (2019) **53** 559-70. DOI: 10.1111/medu.13883 4. Plsek PE, Greenhalgh T. **The challenge of complexity in health care**. *BMJ* (2001) **323** 625-8. DOI: 10.1136/bmj.323.7313.625 5. Patel P, Hancock J, Rogers M, Pollard SR. **Improving uncertainty tolerance in medical students: A scoping review**. *Med Educ* (2022). DOI: 10.1111/medu.14873 6. Strout TD, Hillen M, Gutheil C. **Tolerance of uncertainty: A systematic review of health and healthcare-related outcomes**. *Patient Educ Couns* (2018) **101** 1518-37. DOI: 10.1016/j.pec.2018.03.030 7. Wayne S, Dellmore D, Serna L. **The association between intolerance of ambiguity and decline in medical students’ attitudes toward the underserved**. *Acad Med* (2011) **86** 877-82. DOI: 10.1097/ACM.0b013e31821dac01 8. Caulfield M, Andolsek K, Grbic D, Roskovensky L. **Ambiguity tolerance of students matriculating to US medical schools**. *Acad Med* (2014) **89** 1526-32. DOI: 10.1097/ACM.0000000000000485 9. Lally J, Cantillon P. **Uncertainty and ambiguity and their association with psychological distress in medical students**. *Acad Psychiatry* (2014) **38** 339-44. DOI: 10.1007/s40596-014-0100-4 10. Simpkin AL, Khan A, West DC. **Stress from uncertainty and resilience among depressed and burned out residents: a cross-sectional study**. *Acad Pediatr* (2018) **18** 698-704. DOI: 10.1016/j.acap.2018.03.002 11. Russel SM, Geraghty JR, Renaldy H, Thompson TM, Hirshfield LE. **Training for professional uncertainty: Socialization of medical students through the residency application process**. *Acad Med* (2021) **96** S144-50. DOI: 10.1097/ACM.0000000000004303 12. Papanagnou D, Ankam N, Ebbott D, Ziring D. **Towards a medical school curriculum for uncertainty in clinical practice**. *Med Educ Online* (2021) **26** 1972762. DOI: 10.1080/10872981.2021.1972762 13. Stephens GC, Sarkar M, Lazarus MD. **‘A whole lot of uncertainty’: A qualitative study exploring clinical medical students’ experiences of uncertainty stimuli**. *Med Educ* (2022) **56** 736-46. DOI: 10.1111/medu.14743 14. Nicholson S. (2015) 15. Neetoo A, Zary N, Zidoun Y. **Escape rooms in Medical Education: Deductive Analysis of Designs, Applications and Implementations**. (2021). DOI: 10.20944/preprints202107.0644.v1 16. Guckian J, Eveson L, May H. **The great escape? The rise of the escape room in medical education**. *Future Healthc J* (2020) **7** 112. DOI: 10.7861/fhj.2020-0032 17. Elliott N, Nguyen M, Fritzges J. **Escape this Emergency Room: Simulation Education During a Pandemic**. *West J Emerg Med* (2022) **23** 18. Kinio AE, Dufresne L, Brandys T, Jetty P. **Break out of the classroom: the use of escape rooms as an alternative teaching strategy in surgical education**. *J Surg Educ* (2019) **76** 134-9. DOI: 10.1016/j.jsurg.2018.06.030 19. Cates AL, Krueger J, Simpson SE, Stobart-Gallagher M. **Comparing the effectiveness of a virtual toxicology escape room at two emergency medicine residencies**. *Cureus* (2020) **12**. DOI: 10.7759/cureus.11262 20. Dimeo SP, Astemborksi C, Smart J, Jones EL. **A virtual escape room versus lecture on infectious disease content: effect on resident knowledge and motivation**. *West J Emerg Med* (2022) **23** 9. DOI: 10.5811/westjem.2021.12.54010 21. Wu C, Wagenschutz H, Hein J. **Promoting leadership and teamwork development through escape rooms**. *Med Educ* (2018) **52** 561-2. DOI: 10.1111/medu.13557 22. Zhang XC, Lee H, Rodriguez C. *Cureus* (2018) **10**. DOI: 10.7759/cureus.2256 23. Dittman JM, Maiden K, Matulewicz AT. **A flexible customizable virtual escape room approach for interprofessional learners**. *J Interprof Educ Pract* (2021) **24** 100455. DOI: 10.1016/j.xjep.2021.100455 24. Makri A, Vlachopoulos D, Martina RA. **Digital escape rooms as innovative pedagogical tools in education: a systematic literature review**. *Sustainability* (2021) **13** 4587. DOI: 10.3390/su13084587 25. Garrison DR, Anderson T, Archer W. **Critical thinking, cognitive presence, and computer conferencing in distance education**. *Am J Distance Educ* (2001) **15** 7-23. DOI: 10.1080/08923640109527071 26. Bozkurt A, Akgun-Ozbek E, Yilmazel S. **Trends in distance education research: A content analysis of journals 2009–2013**. *International Review of Research in Open and Distributed Learning* (2015) **16** 330-63. DOI: 10.19173/irrodl.v16i1.1953 27. Castellanos-Reyes D. **20 years of the community of inquiry framework**. *TechTrends* (2020) **64** 557-60. DOI: 10.1007/s11528-020-00491-7 28. Akyol Z, Garrison DR. **Understanding cognitive presence in an online and blended community of inquiry: Assessing outcomes and processes for deep approaches to learning**. *Br J Educ Technol* (2011) **42** 233-50. DOI: 10.1111/j.1467-8535.2009.01029.x 29. Garrison DR, Arbaugh JB. **Researching the community of inquiry framework: Review, issues, and future directions**. *Internet High Educ* (2007) **10** 157-72. DOI: 10.1016/j.iheduc.2007.04.001 30. Anderson T, Liam R, Garrison DR, Archer W. **Assessing teaching presence in a computer conferencing context**. (2001) 31. Rourke L, Anderson T, Garrison DR, Archer W. **Assessing social presence in asynchronous text-based computer conferencing**. *The Journal of Distance Education/Revue de l’ducation Distance* (1999) **14** 50-71 32. Garrison DR, Arbaugh JB. **Researching the community of inquiry framework: Review, issues, and future directions**. *Internet High Educ* (2007) **10** 157-72. DOI: 10.1016/j.iheduc.2007.04.001 33. Garrison R. **Implications of online and blended learning for the conceptual development and practice of distance education**. *International Journal of E-Learning & Distance Education/Revue internationale du e-learning et la formation à distance* (2009) **23** 93-104 34. Abdul Rahim AS. **Mirror Mirror on the Wall: Escape a Remote Virtual Stereochemistry Lab Together**. *J Chem Educ* (2022) **99** 2160-7. DOI: 10.1021/acs.jchemed.2c00050 35. Lyman K. **Implementation of a virtual escape room mock code to comply with social distancing**. *Nurse Educ Today* (2022) **110** 105273. DOI: 10.1016/j.nedt.2022.105273 36. Wang F, Hannafin MJ. **Design-based research and technology-enhanced learning environments**. *Educ Technol Res Dev* (2005) **53** 5-23. DOI: 10.1007/BF02504682 37. Dolmans DH, Tigelaar D. **Building bridges between theory and practice in medical education using a design-based research approach: AMEE Guide No. 60**. *Med Teach* (2012) **34** 1-0. DOI: 10.3109/0142159X.2011.595437 38. Reeves T. *Educational design research* (2006) 64-78. DOI: 10.4324/9780203088364-13 39. Moffett J, Cassidy D. **Building a digital educational escape room using an online design-thinking process**. *Online Learn* (2023) 40. Creswell JW, Clark VL. (2017) 41. DeCuir-Gunby JT. **Mixed methods research in the social sciences**. *Best Practices in Quantitative Methods* (2008) **1** 125-36. DOI: 10.4135/9781412995627.d11 42. McKenney SE, Reeves TC. (2012). DOI: 10.4324/9780203818183 43. Han PK, Klein WM, Arora NK. **Varieties of uncertainty in health care: a conceptual taxonomy**. *Med Decis Making* (2011) **31** 828-38. DOI: 10.1177/0272989X10393976 44. Garrison DR. (2016) 45. Garrison DR, Anderson T, Archer W. **Critical inquiry in a text-based environment: Computer conferencing in higher education**. *Internet High Educ* (1999) **2** 87-105. DOI: 10.1016/S1096-7516(00)00016-6 46. Redmond P. **Reflection as an indicator of cognitive presence**. *E-Learning and Digital Media* (2014) **11** 46-58. DOI: 10.2304/elea.2014.11.1.46 47. Cheng A, Kolbe M, Grant V. **A practical guide to virtual debriefings: communities of inquiry perspective**. *Adv Simul* (2020) **5** 1-9. DOI: 10.1186/s41077-020-00141-1 48. McKerlich R, Anderson T. **Community of inquiry and learning in immersive environments**. *Online Learn* (2007) **11**. DOI: 10.24059/olj.v11i4.22 49. Shea P, Hayes S, Vickers J. **Online instructional effort measured through the lens of teaching presence in the community of inquiry framework: A re-examination of measures and approach**. *International Review of Research in Open and Distributed Learning* (2010) **11** 127-54. DOI: 10.19173/irrodl.v11i3.915 50. Fayram J. (2017) 51. Lowenthal PR, Dunlap JC. **Problems measuring social presence in a community of inquiry**. *E-Learning and Digital Media* (2014) **11** 19-30. DOI: 10.2304/elea.2014.11.1.19 52. Moallem M. **The impact of synchronous and asynchronous communication tools on learner self-regulation, social presence, immediacy, intimacy and satisfaction in collaborative online learning**. *Online J Distance Educ Elearn* (2015) **3** 55-77 53. Carleton RN, Norton MP, Asmundson GJ. **Fearing the unknown: A short version of the Intolerance of Uncertainty Scale**. *J Anxiety Disord* (2007) **21** 105-17. DOI: 10.1016/j.janxdis.2006.03.014 54. Geller G, Faden RR, Levine DM. **Tolerance for ambiguity among medical students: implications for their selection, training and practice**. *Soc Sci Med* (1990) **31** 619-24. DOI: 10.1016/0277-9536(90)90098-D 55. Eukel HN, Frenzel JE, Cernusca D. **Educational gaming for pharmacy students–design and evaluation of a diabetes-themed escape room**. *Am J Pharm Educ* (2017) **81**. DOI: 10.5688/ajpe8176265 56. Clarke V, Braun V, Hayfield N. **Thematic analysis**. *Qualitative psychology: A practical guide to research methods* (2015) **222** 248 57. Garrison DR, Kanuka H. **Blended learning: Uncovering its transformative potential in higher education**. *Internet High Educ* (2004) **7** 95-105. DOI: 10.1016/j.iheduc.2004.02.001 58. Taber KS. **The use of Cronbach’s alpha when developing and reporting research instruments in science education**. *Res Sci Educ* (2018) **48** 1273-96. DOI: 10.1007/s11165-016-9602-2 59. Anderson T. **How communities of inquiry drive teaching and learning in the digital age**. *North Contact* (2017) **1** 1-6 60. Krekhov A, Emmerich K, Rotthaler R, Krueger J. **Puzzles Unpuzzled: Towards a Unified Taxonomy for Analog and Digital Escape Room Games**. *Proceedings of the ACM on Human-Computer Interaction* (2021) **5** 1-24. DOI: 10.1145/3474696 61. Watling C, Driessen E, van der Vleuten CP, Lingard L. **Learning from clinical work: the roles of learning cues and credibility judgements**. *Med Educ* (2012) **46** 192-200. DOI: 10.1111/j.1365-2923.2011.04126.x 62. Stenbom S. **A systematic review of the Community of Inquiry survey**. *Internet High Educ* (2018) **39** 22-32. DOI: 10.1016/j.iheduc.2018.06.001 63. Garrison DR, Cleveland-Innes M. **Facilitating cognitive presence in online learning: Interaction is not enough**. *Am J Distance Educ* (2005) **19** 133-48. DOI: 10.1207/s15389286ajde1903_2 64. Veldkamp A, van de Grint L, Knippels MC, van Joolingen WR. **Escape education: A systematic review on escape rooms in education**. *Educ Res Rev* (2020) **31** 100364. DOI: 10.1016/j.edurev.2020.100364 65. Van Gaalen AE, Brouwer J, Schönrock-Adema J. **Gamification of health professions education: a systematic review**. *Adv Health Sci Educ Theory Pract* (2021) **26** 683-711. DOI: 10.1007/s10459-020-10000-3
--- title: The Minderoo-Monaco Commission on Plastics and Human Health authors: - Philip J. Landrigan - Hervé Raps - Maureen Cropper - Caroline Bald - Manuel Brunner - Elvia Maya Canonizado - Dominic Charles - Thomas C. Chiles - Mary J. Donohue - Judith Enck - Patrick Fenichel - Lora E. Fleming - Christine Ferrier-Pages - Richard Fordham - Aleksandra Gozt - Carly Griffin - Mark E. Hahn - Budi Haryanto - Richard Hixson - Hannah Ianelli - Bryan D. James - Pushpam Kumar - Amalia Laborde - Kara Lavender Law - Keith Martin - Jenna Mu - Yannick Mulders - Adetoun Mustapha - Jia Niu - Sabine Pahl - Yongjoon Park - Maria-Luiza Pedrotti - Jordan Avery Pitt - Mathuros Ruchirawat - Bhedita Jaya Seewoo - Margaret Spring - John J. Stegeman - William Suk - Christos Symeonides - Hideshige Takada - Richard C. Thompson - Andrea Vicini - Zhanyun Wang - Ella Whitman - David Wirth - Megan Wolff - Aroub K. Yousuf - Sarah Dunlop journal: Annals of Global Health year: 2023 pmcid: PMC10038118 doi: 10.5334/aogh.4056 license: CC BY 4.0 --- # The Minderoo-Monaco Commission on Plastics and Human Health ## Abstract ### Background: Plastics have conveyed great benefits to humanity and made possible some of the most significant advances of modern civilization in fields as diverse as medicine, electronics, aerospace, construction, food packaging, and sports. It is now clear, however, that plastics are also responsible for significant harms to human health, the economy, and the earth’s environment. These harms occur at every stage of the plastic life cycle, from extraction of the coal, oil, and gas that are its main feedstocks through to ultimate disposal into the environment. The extent of these harms not been systematically assessed, their magnitude not fully quantified, and their economic costs not comprehensively counted. ### Goals: The goals of this Minderoo-Monaco Commission on Plastics and Human Health are to comprehensively examine plastics’ impacts across their life cycle on: [1] human health and well-being; [2] the global environment, especially the ocean; [3] the economy; and [4] vulnerable populations—the poor, minorities, and the world’s children. On the basis of this examination, the Commission offers science-based recommendations designed to support development of a Global Plastics Treaty, protect human health, and save lives. ### Report Structure: This Commission report contains seven Sections. Following an Introduction, Section 2 presents a narrative review of the processes involved in plastic production, use, and disposal and notes the hazards to human health and the environment associated with each of these stages. Section 3 describes plastics’ impacts on the ocean and notes the potential for plastic in the ocean to enter the marine food web and result in human exposure. Section 4 details plastics’ impacts on human health. Section 5 presents a first-order estimate of plastics’ health-related economic costs. Section 6 examines the intersection between plastic, social inequity, and environmental injustice. Section 7 presents the Commission’s findings and recommendations. ### Plastics: Plastics are complex, highly heterogeneous, synthetic chemical materials. Over $98\%$ of plastics are produced from fossil carbon- coal, oil and gas. Plastics are comprised of a carbon-based polymer backbone and thousands of additional chemicals that are incorporated into polymers to convey specific properties such as color, flexibility, stability, water repellence, flame retardation, and ultraviolet resistance. Many of these added chemicals are highly toxic. They include carcinogens, neurotoxicants and endocrine disruptors such as phthalates, bisphenols, per- and poly-fluoroalkyl substances (PFAS), brominated flame retardants, and organophosphate flame retardants. They are integral components of plastic and are responsible for many of plastics’ harms to human health and the environment. Global plastic production has increased almost exponentially since World War II, and in this time more than 8,300 megatons (Mt) of plastic have been manufactured. Annual production volume has grown from under 2 Mt in 1950 to 460 Mt in 2019, a 230-fold increase, and is on track to triple by 2060. More than half of all plastic ever made has been produced since 2002. Single-use plastics account for 35–$40\%$ of current plastic production and represent the most rapidly growing segment of plastic manufacture. Explosive recent growth in plastics production reflects a deliberate pivot by the integrated multinational fossil-carbon corporations that produce coal, oil and gas and that also manufacture plastics. These corporations are reducing their production of fossil fuels and increasing plastics manufacture. The two principal factors responsible for this pivot are decreasing global demand for carbon-based fuels due to increases in ‘green’ energy, and massive expansion of oil and gas production due to fracking. Plastic manufacture is energy-intensive and contributes significantly to climate change. At present, plastic production is responsible for an estimated $3.7\%$ of global greenhouse gas emissions, more than the contribution of Brazil. This fraction is projected to increase to $4.5\%$ by 2060 if current trends continue unchecked. ### Plastic Life Cycle: The plastic life cycle has three phases: production, use, and disposal. In production, carbon feedstocks—coal, gas, and oil—are transformed through energy-intensive, catalytic processes into a vast array of products. Plastic use occurs in every aspect of modern life and results in widespread human exposure to the chemicals contained in plastic. Single-use plastics constitute the largest portion of current use, followed by synthetic fibers and construction. Plastic disposal is highly inefficient, with recovery and recycling rates below $10\%$ globally. The result is that an estimated 22 Mt of plastic waste enters the environment each year, much of it single-use plastic and are added to the more than 6 gigatons of plastic waste that have accumulated since 1950. Strategies for disposal of plastic waste include controlled and uncontrolled landfilling, open burning, thermal conversion, and export. Vast quantities of plastic waste are exported each year from high-income to low-income countries, where it accumulates in landfills, pollutes air and water, degrades vital ecosystems, befouls beaches and estuaries, and harms human health—environmental injustice on a global scale. Plastic-laden e-waste is particularly problematic. ### Environmental Findings: Plastics and plastic-associated chemicals are responsible for widespread pollution. They contaminate aquatic (marine and freshwater), terrestrial, and atmospheric environments globally. The ocean is the ultimate destination for much plastic, and plastics are found throughout the ocean, including coastal regions, the sea surface, the deep sea, and polar sea ice. Many plastics appear to resist breakdown in the ocean and could persist in the global environment for decades. Macro- and micro-plastic particles have been identified in hundreds of marine species in all major taxa, including species consumed by humans. Trophic transfer of microplastic particles and the chemicals within them has been demonstrated. Although microplastic particles themselves (>10 µm) appear not to undergo biomagnification, hydrophobic plastic-associated chemicals bioaccumulate in marine animals and biomagnify in marine food webs. The amounts and fates of smaller microplastic and nanoplastic particles (MNPs <10 µm) in aquatic environments are poorly understood, but the potential for harm is worrying given their mobility in biological systems. Adverse environmental impacts of plastic pollution occur at multiple levels from molecular and biochemical to population and ecosystem. MNP contamination of seafood results in direct, though not well quantified, human exposure to plastics and plastic-associated chemicals. Marine plastic pollution endangers the ocean ecosystems upon which all humanity depends for food, oxygen, livelihood, and well-being. ### Human Health Findings: Coal miners, oil workers and gas field workers who extract fossil carbon feedstocks for plastic production suffer increased mortality from traumatic injury, coal workers’ pneumoconiosis, silicosis, cardiovascular disease, chronic obstructive pulmonary disease, and lung cancer. Plastic production workers are at increased risk of leukemia, lymphoma, hepatic angiosarcoma, brain cancer, breast cancer, mesothelioma, neurotoxic injury, and decreased fertility. Workers producing plastic textiles die of bladder cancer, lung cancer, mesothelioma, and interstitial lung disease at increased rates. Plastic recycling workers have increased rates of cardiovascular disease, toxic metal poisoning, neuropathy, and lung cancer. Residents of “fenceline” communities adjacent to plastic production and waste disposal sites experience increased risks of premature birth, low birth weight, asthma, childhood leukemia, cardiovascular disease, chronic obstructive pulmonary disease, and lung cancer. During use and also in disposal, plastics release toxic chemicals including additives and residual monomers into the environment and into people. National biomonitoring surveys in the USA document population-wide exposures to these chemicals. Plastic additives disrupt endocrine function and increase risk for premature births, neurodevelopmental disorders, male reproductive birth defects, infertility, obesity, cardiovascular disease, renal disease, and cancers. Chemical-laden MNPs formed through the environmental degradation of plastic waste can enter living organisms, including humans. Emerging, albeit still incomplete evidence indicates that MNPs may cause toxicity due to their physical and toxicological effects as well as by acting as vectors that transport toxic chemicals and bacterial pathogens into tissues and cells. Infants in the womb and young children are two populations at particularly high risk of plastic-related health effects. Because of the exquisite sensitivity of early development to hazardous chemicals and children’s unique patterns of exposure, plastic-associated exposures are linked to increased risks of prematurity, stillbirth, low birth weight, birth defects of the reproductive organs, neurodevelopmental impairment, impaired lung growth, and childhood cancer. Early-life exposures to plastic-associated chemicals also increase the risk of multiple non-communicable diseases later in life. ### Economic Findings: Plastic’s harms to human health result in significant economic costs. We estimate that in 2015 the health-related costs of plastic production exceeded $250 billion (2015 Int$) globally, and that in the USA alone the health costs of disease and disability caused by the plastic-associated chemicals PBDE, BPA and DEHP exceeded $920 billion (2015 Int$). Plastic production results in greenhouse gas (GHG) emissions equivalent to 1.96 gigatons of carbon dioxide (CO2e) annually. Using the US Environmental Protection Agency’s (EPA) social cost of carbon metric, we estimate the annual costs of these GHG emissions to be $341 billion (2015 Int$). These costs, large as they are, almost certainly underestimate the full economic losses resulting from plastics’ negative impacts on human health and the global environment. All of plastics’ economic costs—and also its social costs—are externalized by the petrochemical and plastic manufacturing industry and are borne by citizens, taxpayers, and governments in countries around the world without compensation. ### Social Justice Findings: The adverse effects of plastics and plastic pollution on human health, the economy and the environment are not evenly distributed. They disproportionately affect poor, disempowered, and marginalized populations such as workers, racial and ethnic minorities, “fenceline” communities, Indigenous groups, women, and children, all of whom had little to do with creating the current plastics crisis and lack the political influence or the resources to address it. Plastics’ harmful impacts across its life cycle are most keenly felt in the Global South, in small island states, and in disenfranchised areas in the Global North. Social and environmental justice (SEJ) principles require reversal of these inequitable burdens to ensure that no group bears a disproportionate share of plastics’ negative impacts and that those who benefit economically from plastic bear their fair share of its currently externalized costs. ### Conclusions: It is now clear that current patterns of plastic production, use, and disposal are not sustainable and are responsible for significant harms to human health, the environment, and the economy as well as for deep societal injustices. The main driver of these worsening harms is an almost exponential and still accelerating increase in global plastic production. Plastics’ harms are further magnified by low rates of recovery and recycling and by the long persistence of plastic waste in the environment. The thousands of chemicals in plastics—monomers, additives, processing agents, and non-intentionally added substances—include amongst their number known human carcinogens, endocrine disruptors, neurotoxicants, and persistent organic pollutants. These chemicals are responsible for many of plastics’ known harms to human and planetary health. The chemicals leach out of plastics, enter the environment, cause pollution, and result in human exposure and disease. All efforts to reduce plastics’ hazards must address the hazards of plastic-associated chemicals. ### Recommendations: To protect human and planetary health, especially the health of vulnerable and at-risk populations, and put the world on track to end plastic pollution by 2040, this Commission supports urgent adoption by the world’s nations of a strong and comprehensive Global Plastics Treaty in accord with the mandate set forth in the March 2022 resolution of the United Nations Environment Assembly (UNEA). International measures such as a Global Plastics Treaty are needed to curb plastic production and pollution, because the harms to human health and the environment caused by plastics, plastic-associated chemicals and plastic waste transcend national boundaries, are planetary in their scale, and have disproportionate impacts on the health and well-being of people in the world’s poorest nations. Effective implementation of the Global Plastics Treaty will require that international action be coordinated and complemented by interventions at the national, regional, and local levels. This Commission urges that a cap on global plastic production with targets, timetables, and national contributions be a central provision of the Global Plastics Treaty. We recommend inclusion of the following additional provisions: This Commission encourages inclusion in the Global Plastic Treaty of a provision calling for exploration of listing at least some plastic polymers as persistent organic pollutants (POPs) under the Stockholm Convention. This Commission encourages a strong interface between the Global Plastics Treaty and the Basel and London Conventions to enhance management of hazardous plastic waste and slow current massive exports of plastic waste into the world’s least-developed countries. This Commission recommends the creation of a Permanent Science Policy Advisory Body to guide the Treaty’s implementation. The main priorities of this Body would be to guide Member States and other stakeholders in evaluating which solutions are most effective in reducing plastic consumption, enhancing plastic waste recovery and recycling, and curbing the generation of plastic waste. This Body could also assess trade-offs among these solutions and evaluate safer alternatives to current plastics. It could monitor the transnational export of plastic waste. It could coordinate robust oceanic-, land-, and air-based MNP monitoring programs. This Commission recommends urgent investment by national governments in research into solutions to the global plastic crisis. This research will need to determine which solutions are most effective and cost-effective in the context of particular countries and assess the risks and benefits of proposed solutions. Oceanographic and environmental research is needed to better measure concentrations and impacts of plastics <10 µm and understand their distribution and fate in the global environment. Biomedical research is needed to elucidate the human health impacts of plastics, especially MNPs. ### Summary: This Commission finds that plastics are both a boon to humanity and a stealth threat to human and planetary health. Plastics convey enormous benefits, but current linear patterns of plastic production, use, and disposal that pay little attention to sustainable design or safe materials and a near absence of recovery, reuse, and recycling are responsible for grave harms to health, widespread environmental damage, great economic costs, and deep societal injustices. These harms are rapidly worsening. While there remain gaps in knowledge about plastics’ harms and uncertainties about their full magnitude, the evidence available today demonstrates unequivocally that these impacts are great and that they will increase in severity in the absence of urgent and effective intervention at global scale. Manufacture and use of essential plastics may continue. However, reckless increases in plastic production, and especially increases in the manufacture of an ever-increasing array of unnecessary single-use plastic products, need to be curbed. Global intervention against the plastic crisis is needed now because the costs of failure to act will be immense. ## Section 1—Introduction Plastic is the signature material of our age. It has contributed to improvements in human health, extensions in longevity, and growth of the global economy. It has supported some of the most significant advances of modern civilization in fields as diverse as construction, electronics, aerospace, and medicine. Medical breakthroughs that could never have occurred in the absence of plastic include intravenous tubing, oropharyngeal airways, flexible endoscopes, and artificial heart valves. Plastics are used in food packaging and in the manufacture of furniture, toys, clothing, and athletic goods. It is now clear, though, that the benefits provided by plastics have come at great cost to human health, the environment, and the economy. These harms have been suspected for decades [12], and their great magnitude is now becoming increasingly apparent. The main driver of these harms has been massive increases in plastic production from under 2 megatons (Mt) per year in 1950 to more than 400 Mt today [3456]. Half of all plastic ever produced has been made since 2002. Sharp increases in plastic production coupled with very low rates of recovery and recycling—less than $10\%$ globally—have led to the accumulation of over 6 gigatons (Gt) of plastic waste in the earth’s environment and wide-scale human exposure to plastics’ and plastic-associated chemicals [78]. Plastic manufacture significantly contributes to climate change and is at present responsible for about $3.7\%$ of greenhouse gas (GHG) emissions [910111213], a contribution that is projected to increase to $4.5\%$ by 2060 if current trends continue unchecked [14]. The volume of plastic and plastic-associated chemicals in the earth’s environment has become so great as to exceed global capacity for assessment, monitoring, and response and may be approaching a point where it could irreversibly damage the planet’s support systems [1516]. The ocean has been badly damaged by plastics [17181920212223]. An estimated 4.8–12.7 Mt of plastic waste entered the marine environment from the land in 2010 alone [24]. Macroplastics (the bottles, barrels, packaging materials, and fishing gear that litter beaches, kill marine animals, and accumulate in vast mid-ocean gyres) are the most visible component of ocean plastic pollution. An additional quantity of plastic debris in the ocean consists of chemical-laden microplastic and nanoplastic particles (MNPs) and fibers, mostly formed through the degradation of plastic waste. Many plastics appear able to resist degradation and could persist in the environment for many decades [17]. Plastic is responsible for disease, disability, and premature death at every stage of its life cycle, from extraction of the coal, oil, and gas that are its feedstocks, through transport, manufacture, refining, consumption, recycling, combustion, and disposal into the environment [2526]. Coal miners and oil field and fracking workers suffer high rates of injury, traumatic death, lung disease, cardiovascular disease, and cancer [2728]. Plastic production workers suffer high rates of cancer and lung disease. Residents of “fenceline” communities—those adjacent to plastic manufacturing plants—experience high rates of premature birth, low birth weight, childhood leukemia, asthma, chronic obstructive pulmonary disease, cardiovascular disease, vehicular injuries, and mental health problems [2930313233]. Plastics’ harms fall disproportionately on the world’s poorest and most vulnerable people. Children are especially susceptible [33]. Research into plastics’ health impacts has been fragmented and has not comprehensively examined these impacts across the plastic life cycle [272829303132]. Until recently, most studies of plastics’ effects have appeared in agency publications [1822] or in oceanographic and environmental journals, where they are not often seen by physicians or public health professionals. Plastics’ contributions to the global burden of disease have not been quantified. The health-related economic costs of plastic production, use, and disposal have not been counted. ## The Minderoo-Monaco Commission on Plastics and Human Health The enormous and inadequately charted scale of plastics’ effects on human health and the economy, and our recognition that these impacts will worsen inexorably if current trends continue unchecked, led us to form the Minderoo-Monaco Commission on Plastics and Human Health [35]. This interdisciplinary *Commission is* composed of scientists, clinicians, and policy analysts from around the world. It is coordinated by the Global Observatory on Planetary Health at Boston College. This *Commission is* committed to educating physicians, nurses, public health workers, the press, civil society, and the global public about the full magnitude of plastics’ hazards. An additional goal is to inform the work of policy makers, government leaders, and international organizations as they strive to fulfill the urgent call of the United Nations Environment Assembly (UNEA) to curb plastic pollution and its unsustainable impacts by negotiating a legally binding Global Plastics Treaty [3637]. The Commission presents a comprehensive analysis of plastics’ impacts on human health across its life cycle. We examine direct health impacts, including those caused by chemicals used in plastics, as well as indirect health effects mediated through plastics’ damage to the terrestrial, freshwater, and marine ecosystems. We estimate plastics’ health-related economic costs and construct a framework to support further expansion of these economic analyses [38]. We consider the ethical and moral implications of the unending production, consumption, and disposal of plastics, which fall disproportionately on the poor and disadvantaged, especially in the Global South [3940]. We identify knowledge gaps and research needs [41]. The Commission offers science-based solutions to slow the pace of plastic production, reduce the accumulation of plastic waste, and protect human and planetary health. Our strongest recommendation is that pursuant to the 2022 mandate of the UNEA, the world’s nations develop and implement a strong, comprehensive, and legally binding Global Plastics Treaty that ensures urgent action and effective interventions across the entire life cycle of plastics [3642]. It is essential that this Treaty covers all components of plastics, including its myriad chemicals, and that it includes measures to protect the health and well-being of the vulnerable and at-risk populations most seriously harmed by the current patterns of plastic production, use, and disposal [43]. This Commission recommends that a central provision of the Global Plastics Treaty should be a global cap on plastic production guided by targets and timetables and supported by national commitments. Our additional recommendations speak to the need to eliminate unnecessary uses of plastics, especially single-use plastics [44]; the need for plastic manufacturers to take full financial responsibility for their products across their life cycle—extended producer responsibility (EPR) [45]; the need for reductions in the complexity of plastics; the need for health-protective standards for plastic-associated chemicals; and the possibility of listing plastics as persistent organic pollutants (POPs) under the Stockholm Convention. While gaps remain in knowledge about plastics’ harms to human health and the global environment, the evidence available today unequivocally shows that these harms are already great and indicates that they will increase in magnitude and worsen in severity in the absence of urgent intervention [34]. The manufacture and use of essential plastics may continue, but increases in plastic production, especially endless increases in the manufacture of an ever-increasing array of unnecessary single-use plastic products, need to be curbed. Key ingredients for future successes in curbing plastic production, reducing plastic waste, and protecting human and planetary health are an educated and engaged citizenry; champions from medicine, science, public health, and sustainable development; strengthened international agencies; and courageous, visionary, and ethically grounded leaders at every level of government who heed the science and act responsibly to protect human health and preserve the earth, our common home [46]. ## Introduction The plastic life cycle is long, complex, and far from circular and encompasses three major phases: production, use, and disposal (Figure 2.1) [4748]. Plastic production spans multiple countries and continents [1249], resulting in extensive transboundary pollution and externalization of adverse environmental and human health impacts across national borders [5051]. Box 2.1 describes our plastic age. **Figure 2.1:** *The plastic life cycle. The plastic life cycle is long and complex spanning multiple countries. There are three major phases. During production, carbon feedstocks – derived 99% from coal, oil and gas – are transformed through energy-intensive, catalytic processes into a vast array of products. Plastic is used in virtually every aspect of modern life and has provided many benefits. Single-use plastic constitutes the largest market share followed by synthetic fibers, building and construction, transport, electrical, agriculture and medical. Recycling is minimal. Disposal involves landfilling as well as controlled and uncontrolled burning. Plastic-laden e-waste is particularly problematic. Transnational environmental leakage of chemicals and plastic waste occurs throughout the life cycle resulting in extensive pollution and health hazards.Credit: Designed in 2022 by Will Stahl-Timmins.* In the 75 years since large-scale production began in the aftermath of World War II, more than 8,300 Mt of plastic have been manufactured [3]. Annual production has grown from under 2 Mt in 1950 to 460 Mt in 2019 [45]. Plastic production is on track to treble by 2060 [14]. In this Section, we describe the plastic life cycle and the hazards that plastics pose to human health at each phase of this life cycle. We define health hazards as factors that increase the risk of an adverse health outcome following exposure [52]. Detailed descriptions of the resulting adverse health outcomes and some of their underlying mechanisms are presented in Section 4. The ocean stabilizes Earth’s climate, provides essential ecosystem services, generates oxygen to the atmosphere, and produces protein to feed billions of people. Concern about the negative impacts of plastics (particularly macroplastics and plastic debris) in the ocean first arose in the 1960s with the finding that seabirds were ingesting substantial quantities [488]. In the 1970s, reports of MPs in the ocean and estuaries began to appear [1489490491492493]. Concerns about the negative effects of chemicals added to plastics during manufacture resulted from the discovery that intravenous materials stored in PVC bags contained phthalates that had leached from the bags [494]. Ultimately, a personal account published in 2003 describing plastics floating in the North Pacific Ocean far from land, in the presumed “pristine” open ocean environment, raised alarm bells among environmentalists and the public [495]. Although this description of a “garbage patch” resulted in some misperceptions—namely, that of an immense floating island of plastic trash rather than the widespread dispersal of primarily tiny plastic particles—the report of sizeable amounts of plastic debris so far from land was shocking. Widespread plastic debris is, however, only a symptom of a larger problem. A year later, a seminal paper [2] describing MPs as small as 20 µm in beach sediments and plankton samples going back to the 1960s propelled this new subfield of environmental science, which grew with exceptional speed, as indicated by the number of scientific articles on plastic pollution in the ocean and the environment published in the 20 years since [496]. There are now >4,500 publications on the sources, occurrence, distribution, fate, and impacts of plastics (all types) in the ocean, and the implications for human health. Excellent summaries can be found in recent reports [23497498499500501]. In preparing this Section of the Commission report, we drew information from these and other reviews [1941502503504505506507508509510511512513514515516517518] as well as from recent primary literature considering the issues of plastics and plastic-derived chemicals in the ocean, their impacts, and their potential to be transferred to humans. Despite the abundance of information, our understanding of the significance and potential for impacts of ocean plastics, including macroplastics but especially small MNPs, is fragmentary. Yet, given the continued growth in the manufacture of plastics, as detailed in Section 2, the amounts entering the ocean and the potential for harm are certain to increase. Recognizing the great and growing magnitude of the plastic pollution problem, the ECHA recently concluded that, even in the face of incomplete scientific information, there is sufficient evidence of risk to recommend legal actions to reduce the inputs of MPs to the environment [519]. In addition, the world’s nations adopted a resolution in 2022 under the auspices of UNEA to establish a Global Plastics Treaty by the end of 2024 to end global plastic pollution [520]. Thus, as we move toward measures to address the plastics problem, we must continue to seek a greater understanding of the behavior and impacts of plastics in the ocean to better inform the search for solutions (Section 7). The diversity and complexity of the materials that fall under the category “plastics” are major obstacles to a complete understanding of plastic’s behavior and impacts in the ocean. Thus, as noted in Section 2 and by others [82043521522523524525], plastics found in the environment (including the marine environment) collectively include an enormous variety of polymers, sizes (macroplastics and MNPs ranging over at least nine orders of magnitude), shapes, colors, added or sorbed chemicals (thousands), surface chemistries, and other features. Adding to the complexity are the physical, chemical, and biological transformations that occur after the plastics enter the environment [526]. The distribution, fate, and impacts of plastics are strongly influenced by their properties, which vary tremendously in a continuum across the range of features mentioned above. This complexity severely limits our ability to generalize from results of studies examining one or a few types of plastics. In particular, NPs have properties and behavior that are distinct from those of both MPs and engineered nanomaterials [527528529]. Indeed, it has been suggested that different types of plastics should be considered separate materials [525]. One promising approach to managing the complexity of plastics is through the use of continuous probability distributions of plastic properties, which focuses attention on the plastics that are most relevant for assessing exposure (including human exposure) and risk [522524530]. Plastics contaminate the environment on a global scale, occurring nearly ubiquitously in terrestrial, aquatic (marine and freshwater), atmospheric, and built environments. Here, we focus on plastics in aquatic environments (primarily the ocean), which have been extensively investigated and which have potential impacts on human health both directly (e.g., through consumption of seafood) and indirectly (e.g., through degradation of ecosystems). Readers interested in the presence, fate, and effects of plastics in terrestrial and atmospheric environments are referred to recent reviews of these topics [275531532533534]. In focusing on aquatic environments, especially the ocean, we diverge from the organizational structure of “production – use – disposal” that is used in other Sections of this report. We do not detail impacts from production of fossil carbon feedstocks in the ocean, although there are impacts associated with obtaining feedstocks in offshore oil production, as there are on land. Sections 2 and 4 detail human impacts from feedstock production. We recognize that overall, as with all oil produced, a small percentage of the oil produced offshore will be used in plastic production. Nonetheless, impacts of feedstock production on life in the ocean derive from spills during transport such as the Exxon Valdez or accidents such as Deepwater Horizon. Impacts of oil in the sea have been reviewed substantially for many years [535]. There may also be impacts of sounds from well operation or from tanker vessels [536537538]. The over-riding concerns with plastics in the ocean are more with distribution, degradation, impacts in the ocean, and also with how plastics in the ocean may impact humans. Plastic endangers human health and causes disease, disability, and premature death at every stage of its long and complex life cycle—from extraction of the coal, oil, and gas that are its main feedstocks; to transport, manufacture, refining, use, recycling, and combustion; and finally to reuse, recycling, and disposal into the environment [12]. Children are particularly vulnerable (see Box 4.1). The purpose of this section, which parallels the structure of Section 2, is to trace plastics’ health impacts at each stage across its life cycle. These impacts are summarized in Figure 4.1. **Figure 4.1:** *Health impacts of plastic. Plastic threatens and harms human health at every stage of its life cycle. COPD is chronic obstructive pulmonary disease.Credit: Designed in 2022 by Will Stahl-Timmins.* The harms to human health and the environment caused by the production, use, and disposal of plastics impose significant economic costs on individuals, society, and governments. These include costs of illness; productivity losses resulting from disease, disability, and premature death; and costs resulting from damages to ecosystems. The goal of this section is to quantify the burden of disease and estimate the health-related costs associated with the production, use, and disposal of plastics. To undertake these analyses, we relied on three data elements: data on exposures to plastic-related hazards in at-risk populations; data on incidence and prevalence of plastic-related health outcomes in these populations; and dose-response functions relating plastic-associated hazards to adverse health outcomes. We first estimate the costs of health impacts associated with plastic production—for example, the costs of occupational disease and premature death in workers producing plastics, and the health-related costs of air pollution exposure in communities adjacent to oil and gas extraction and plastic production facilities. Next, we estimate some of the health-related costs associated with plastic disposal. We also present estimates of the costs of mortality and other impacts associated with emissions of CO2 and other GHGs from plastics production (Figure 5.1). **Figure 5.1:** *Health costs of plastic. Plastic causes significant harm to humans as well as the environment across all stages of its life cycle. Quantifying the human health disease burden and economic costs associated with plastic production, use, and disposal is a complex, and at times difficult, endeavor. Conducting high-quality epidemiological studies will greatly benefit this emerging field of research. PPP, purchasing power parity; PM2.5, particulate matter with a diameter of 2.5 micrometers or less; CO2, carbon dioxide; Gt, Gigatons; CO2e, carbon dioxide equivalent; DEHP, di(2-ethylhexyl) phthalate; PBDE, polybrominated diphenyl ether; BPA, bisphenol A.Credit: Designed in 2022 by Will Stahl-Timmins.* Estimating the disease burden and the economic costs associated with plastics use is more challenging. Most of this disease burden appears to result from exposures to plastic additive chemicals, and the epidemiological literature on these additives, many of which are endocrine disruptors, carcinogens, and neurotoxicants, is still in its infancy, although rapidly expanding. Estimates of population exposures to these chemicals exist primarily for countries in Europe and North America. We are therefore able to estimate and value the health impacts of some EDCs and neurotoxic chemicals found in plastics for these countries but not for the rest of the world. We sought additionally to quantify the burden and costs of disease and death caused by MNP particles. Epidemiologic studies of morbidity and mortality associated with MNP exposures are, however, rare, due to limited measurement of these particles in the general population. We therefore do not quantify or value these health effects. We note that the economic costs of harms to human health and the environment caused by plastics are often not borne by plastics manufacturers or fossil fuel companies. Instead, most of these costs are externalized and are borne by governments, businesses, and individual citizens, locally as well as globally. Plastics’ health consequences fall disproportionately on the poor, minorities, the marginalized, and people in the Global South. Groups at especially high risk of disease, disability, and death caused by plastic, its feedstocks, its components, and its waste are people of color; Indigenous populations; fossil fuel extraction workers; chemical and plastic production workers; informal waste and recovery workers; persons living in “fenceline” communities adjacent to fossil fuel extraction, plastic production, and plastic waste facilities; and children (see, e.g., Figure 6.1 and Figure 6.2). **Figure 6.1:** *The impact of plastic on social and environmental justice.Credit: Designed in 2022 by Will Stahl-Timmins.* **Figure 6.2:** *Informal waste pickers – transit storage site in Ogun State, Nigeria.Credit: Adetoun Mustapha and Korede Out.* The purpose of this section of the Minderoo-Monaco Commission on Plastics and Human *Health is* to increase understanding of these disproportionate impacts and devise solutions that identify and address the needs of vulnerable groups. This requires looking at the problem of plastic pollution through the lenses of both social justice and the more specific issue of environmental justice (SEJ). See Box 6.1. Plastics are ubiquitous in modern societies. They have supported breakthroughs in fields as diverse as medicine, electronics, aerospace, construction, food packaging, and sports. It is now clear, however, that current patterns of plastic production, use, and disposal are not sustainable and are responsible for significant harms to human health, the environment, and the economy as well as for deep societal injustices. While there remain gaps in knowledge about plastics’ harms and uncertainties about their full magnitude, the evidence available today demonstrates unequivocally that these harms are already great and that they will increase in magnitude and severity in the absence of urgent and effective intervention [34]. Manufacture and use of essential plastics may continue. But reckless increases in plastic production, especially increases in the manufacture of an ever-increasing array of unnecessary single-use plastic products, that take no heed of health or environmental consequences must be curbed. Global intervention against the plastic crisis is needed now, because the costs of failure to act will be immense. The good news is that many of plastics’ harms can be avoided via better practices of production, design of alternative, less toxic materials, and decreased consumption. Plastics’ harms to human health, the environment, and the global economy can be mitigated by building on the same cost-effective strategies that international bodies and governments at every level have used for 50 years to prevent and control air, water, soil, and ocean pollution [211380]. Contrary to the oft-heard tropes that pollution is the unavoidable price of progress and that pollution control destroys economies, a review by the Lancet Commission on Pollution and Health clearly demonstrates that actions taken by governments to prevent and control pollution have, in fact, yielded large positive returns on investment [9]. Thus, every dollar invested in air pollution control in the USA since passage of the Clean Air Act in 1970 has yielded a return of $30 (USD) (range, $4–88) [1492]. These gains resulted from the substantially increased economic productivity of a healthier, longer-lived population and from reductions in the costs of health care associated with pollution-related disease. Likewise, the removal of lead from gasoline in the USA reduced children’s blood lead levels by $95\%$ and has generated an estimated economic benefit of $200 billion (USD) in each year’s annual birth cohort since 1980—an aggregate benefit in the past 40 years of over $8 trillion USD [1493]. This large economic gain resulted from the population-wide increase in children’s cognitive function (IQ scores), creativity, and productivity that followed widescale reduction in lead exposure. ## Box 2.1 Our Plastic Age: From Bright Dawn to Current Reality. Plastics are lightweight, versatile, durable, and inexpensive materials that have transformed our daily lives to an extent unimaginable when the first plastics were developed in the 19th century. By the early 20th century, the wide range of potential applications and benefits of plastics were rapidly becoming apparent [53]. The authors of a Pelican reference book of the time offered the following thoughts on plastics’ future: *It is* a world free from moth and rust, and full of color, a world largely built of synthetic materials made from the most universally distributed substances, a world in which nations are more and more independent of localized naturalized resources, a world in which man, like a magician, makes what he wants for almost every need out of what is beneath and around him [54]. By the 1950s, all of the main synthetic plastics in use today, including polyvinyl chloride, polypropylene, polyethylene, polyethylene terephthalate, and polystyrene, had been formulated [55]. Greatly reduced costs were associated with their mass production and led to a shift in plastic use away from predominantly durable goods to the dawn of a new era—disposable living—where for our convenience items made from a range of materials, including plastics, could be used just once, with the environmental costs of this practice being externalized by both producer and consumer. At that time, annual global plastics production was around 5 Mt, and while the concept of throwaway living was inherently wasteful, the quantities of plastic waste produced were still quite small [53]. Indeed, in 1955, Life magazine featured a compelling image with an article titled “Throwaway Living: Disposable Items Cut Down Household Chores” (Figure 2.2). **Figure 2.2:** *Throwaway Living – Disposable items cut down household chores. A family tossing single-use products through the air illustrating how society has turned into a disposable society with throwaway products, New York, NY, July 7 1955 (Life Magazine August 1st, 1955).Credit: Peter Stackpole/The LIFE Picture Collection/Shutterstock.* The world’s population has tripled since the 1950s, but production of plastics has increased over 70-fold in the same period. It now exceeds 460 Mt per year, with $35\%$–$40\%$ being the production of single-use items. This increase in production coupled with the popularity of single-use items provides a pathway facilitating the rapid transfer of nonrenewable fossil oil and gas reserves into short-lived products and thence to highly persistent waste that is rapidly accumulating in both managed systems and as litter in the natural environment [256]. In short, our plastic age is causing unprecedented damage; it is depleting natural resources, impacting economies, and negatively affecting human health [5758]. So substantial are the effects of plastic production, use, and disposal that they are considered, alongside climate change and habitat destruction, as substantive negative drivers of the Anthropocene—a geological period during which human activity is the dominant influence on climate and the environment [59]. It is clear that our current linear use of a fossil carbon resource, via plastics, to waste is no longer sustainable. If we are to continue to benefit from the opportunities offered by plastics, current practices need to change [5758]. Levels of concern are so substantive that alongside conventions on biodiversity and climate, the United Nations has recently voted to support a Global Treaty on Plastics, UNEA 5.2. What then are the solutions? To the present, our understanding of solutions has been largely limited to the waste hierarchy advocated on the first Earth Day in 1970 of the three R’s—reduce, reuse, and recycle—alongside a few other more recent R’s, like redesign [60] and refuse [61]. For plastics, reduce appears the most promising of these solutions, especially given the health threats and need for precaution. However, given the many societal and environmental benefits conveyed by plastic [5662], there will be a limit to the number of plastic products we can reduce by elimination. For example, single-use packaging is often highlighted as the epitome of wastefulness. Yet, that same packaging protects products and extends the life of food and drink, thus reducing food and other wastage, which is another major environmental challenge [5662]. To retain the benefits offered by our plastic age, without the largely unintended side effects of the linear business model developed in the 1950s, we rapidly need to understand how to use plastics more responsibly. We need granular evidence on which solutions will be most effective in which contexts [60]. Plastic pollution is a wicked problem, and we are now at a crossroads of opportunity to address it. UNEA 5.2 is an immense achievement [60]. Delivering on its ambition will require robust evidence from transdisciplinary research and stakeholder collaborations to indicate the most, and the least, appropriate interventions. While we need to act with urgency, as we move toward solutions, it is important to recognize that robust scientific evidence and evaluation will be just as critical as it was in raising awareness of the issue itself [60]. Success in the next chapter of our plastic age rests with us all—the public, governments, international organizations, and industry. ## Production—Markets and market trends Plastics are high-volume, highly heterogeneous manufactured materials based on chemical polymers. They are versatile, useful, durable, and low cost. More than 8 Gt (8 billion tonnes) of plastics have been produced over the past seven decades. Annual production has grown from 2 Mt in 1950 to 460 Mt in 2019, a 230-fold increase [5]. In the past two decades, global plastic production has increased almost exponentially, with an average annual growth rate since 2012 of $4.6\%$ [63]. More than half of the 8.3 Gt of plastic produced since 1950 have been manufactured since 2002 [64]. The Organisation for Economic Co-operation and Development (OECD) predicts that global plastic use will continue to increase from 460 Mt in 2019 to 1,231 Mt in 2060 [14]. Overall production and market trends are summarized in Figure 2.3. **Figure 2.3:** *Plastic life cycle: Production and market predictions. Large volumes of plastic have been made since production started in 1950s with continuing predicted increases. Mt, Megatons; Gt, Gigatons. References: [1](Organisation for Economic Co-operation and Development (OECD), 2022a); [2](European Environment Agency (EEA), 2021); [3](Cabernard et al., 2022); [4](Wiesinger, Wang and Hellweg, 2021); [5](Charles, Kimman and Saran, 2021); [6](Geyer, 2020); [7](Organisation for Economic Co-operation and Development (OECD), 2022b); [8](Lebreton and Andrady, 2019).Credit: Designed in 2022 by Will Stahl-Timmins.* Single-use and short-lived plastics account for $35\%$–$40\%$ of current plastic production [65], and this fraction is growing very rapidly. The production of single-use plastics is predicted to increase by $30\%$ (70 Mt) between 2021 and 2025 [65]. Estimates of plastic consumption vary by region and product type. China and the US account for $20\%$ and $18\%$, respectively, of current plastic demand [5]. Annual consumption of single-use plastics varies widely, for example, being 50–60 kg per person in Australia and the US; 30–45 kg in South Korea, the UK, Japan, and Saudi Arabia; and 5–30 kg in Germany, Thailand, Turkey, and India [65]. The largest future increases in plastic use are anticipated to occur in the emerging economies of sub-Saharan Africa and Asia [5]. The main driver of recent explosive increases in plastic production and the manufacture of single-use plastics is a massive pivot in investment by multinational fossil carbon corporations. These vertically integrated companies, which produce coal, oil, and gas as well as plastics and petrochemicals, are pivoting away from fossil fuel production in response to growing demand for renewable energy while also expanding their capacity for polymer production [136566]. The companies principally responsible for this pivot include Sinopec (+$36\%$ growth in plastic production), ExxonMobil (+$35\%$), and PetroChina (+$38\%$), with even greater expansions being anticipated for Russian-owned SIBUR (+$240\%$), Oman Oil Refineries and Petroleum (+$269\%$), and Indian HPLC-Mittal (+$343\%$) [65]. An additional driver of recent increases in plastic production is growing demand from emerging economies. Only about $30\%$ of the plastic produced since 1950 is still in use [3]. Most plastics used today are virgin plastics, as opposed to recycled plastics, with polypropylene (PP) ($16\%$), fibers ($13\%$), high-density polyethylene (HDPE) ($12\%$), and low-density polyethylene (LDPE) ($12\%$) predominating [5]. Of the virgin plastics produced globally in 2020, $52\%$ were produced in Asia ($32\%$ in China and $3\%$ in Japan), followed by North America ($19\%$), Europe ($17\%$), the Middle East and Africa ($7\%$), and Latin America ($4\%$) [512]. Plastic recycling rates are very low, about $9\%$ globally [5]—much lower than recycling rates for glass (European Union [EU] ~$75\%$), paper (EU ~$70\%$), and aluminum (EU ~$65\%$) [676869]. The combination of rapid growth in the production of numerous different polymer types coupled with low recycling rates has resulted in the generation of enormous quantities of plastic waste. Cumulative global production of plastic waste since 1950 is estimated to be 5.8 Gt [3]. The quantity of plastic production can be contextualized by considering that plastic-carbon ($83\%$ of plastic mass, on average) equaled total carbon in global human biomass in 1962 (0.06 Pg-C, petagram = 1015 g) and, assuming current trends (i.e., cubic growth), is predicted to equal global blue carbon mass (14-Pg-C) in 2035 and total carbon in global bacterial biomass (70-Pg-C) by 2095 [7071]. The Stockholm Environment Institute has developed the concept of “planetary boundaries” to define a “safe operating space for humanity,” the conditions necessary for human societies to survive and thrive on earth. Plastics and chemicals are now included in this framework along with climate change and biodiversity loss [72]. The Stockholm Environment Institute has recently concluded that production and environmental dissemination of plastics and petrochemicals have increased so rapidly and uncontrollably that they have outstripped global capacity for assessment and monitoring. Risk is high that pollution by plastics and chemicals could—like climate change and biodiversity loss—lead to abrupt, nonlinear, and catastrophic disruption of the earth’s operating systems; destabilize modern societies; and endanger human survival [16]. ## Conventional Fossil carbon—coal, oil, and natural gas—is the primary feedstock for more than $98\%$ of global plastic production [126673]. Counting use for both plastic feedstocks and for energy consumed during plastic production, coal contributed by far the largest proportion of this fossil carbon ($67\%$ in 2015) compared to oil ($23\%$) and natural gas ($10\%$) [13]. Global estimates of the total amount of fossil fuel used to make plastic are, by contrast, difficult to derive because of flexibility in consumption of feedstocks by the petrochemical industry [74], a lack of detailed data on this aspect of industrial consumption, and regional differences in the mix of fossil fuels used to generate energy versus other products [1374]. However, it was estimated that, in 2015, $3.8\%$ of the fossil carbon was used globally, both as feedstock and energy, to produce plastic [13]. Similarly, the World Economic Forum estimated that plastics’ share of global oil consumption was $6\%$ in 2014 and is predicted to increase to $20\%$ by 2050 [774]. For the US, the Energy Information Administration’s Short-Term Energy Outlook has forecast continuing increases in crude oil production from 12.3 million barrels/day in 2019 to 12.8 million barrels/day in 2024 [75]. Similarly, a report from British Petroleum predicts robust increases in the noncombusted use of oil, gas, and coal driven by “particularly strong growth in plastics, despite increasing regulation on the use of plastics [76].” Coal production has significant environmental impacts. Virtually all the elements of the periodic table occur in coal, and 20 are present in quantities sufficient to be of environmental concern, namely arsenic, cadmium, chromium, mercury, lead, selenium, boron, fluorine, manganese, molybdenum, nickel, beryllium, copper, thorium, uranium, vanadium, zinc, barium, cobalt, and antimony [77]. On- and offshore oil production generates significant air pollution as a result of emissions from drilling equipment, hydrocarbons escaping from wells, flaring of natural gas, emissions from transport vehicles [78], and intentional use of chemicals [7980]. Workers face significant occupational risks, including fire and explosions, particularly during offshore drilling operations that are vulnerable to blowouts and involve handling heavy equipment [78]. Contaminants similar to those used in hydraulic fracturing, or fracking, are found in water produced during the oil extraction process and include benzene, toluene, ethylbenzene, and xylene (BTEX) as well as toxic metal(loid)s such as arsenic, cadmium, chromium, and mercury [78]. The oil industry also uses millions of tons of barium sulfate as a densifying additive in drilling fluids to seal the space around the drill bit [81], which can be solubilized by other common chemicals, thereby creating hazardous waste [78]. PFAS are used for a wide range of applications in the mining and petroleum industry, including foaming and antifoaming, oil field mapping, and as wetting agents and surfactants in equipment such as pipelines, seals, conveyor belts, and electrical insulation [82]. In the Niger delta, oil production and gas flaring from over 100 wells has resulted in extensive water pollution, largely as a result of oil spills [83]. Analyses at nine sites revealed widespread toxic metal contamination with substantial exposures of populations at all ages [83]. The drilling rig failure of British Petroleum’s Deepwater Horizon in 2010 in the Gulf of Mexico resulted in the deaths of 11 workers and the largest oil spill in US history; 4.9 million barrels of crude oil were released over three months [84]. There were also multiple human health impacts, including large quantities of dispersants and their aerosolization being associated with toxicological effects such as obesogenicity and illness as well as potential increases in harmful algal blooms, mercury exposure, and pathogenic Vibrio bacteria. Due to their heavy reliance on natural resources, Gulf communities were vulnerable to high levels of disruption to their livelihoods, including contaminated fish and shellfish, psychological impacts, and institutional distrust [84]. Natural gas production is associated with multiple health hazards. Natural gas, largely methane and ethane, can be extracted using conventional or unconventional well pads. Conventional methods extract gas from permeable reservoirs (e.g., sandstone) by moving gas to the surface without the need to pump. In contrast, unconventional methods extract gas from shale reservoirs, which have low permeability, through horizontal drilling combined with staged hydraulic fracturing. Compared to unconventional gas extraction, conventional gas wells have resulted in considerably greater releases of methane to the environment per unit production [85]. The risk of fire and explosion is a potential occupational health and safety hazard for both conventional and unconventional gas wells [86]. While the majority of research on health hazards related to natural gas extraction has been focused on unconventional wells, multiple contaminants have also been detected in groundwater sites near conventional gas production regions. For example, a unique suite of contaminants (e.g., BTEX) was detected in a potable drinking water well from a well pad in a conventional gas region [87]. Also, high levels of endocrine-disrupting activities were detected in the samples from this well pad, including antagonist activities for androgen, glucocorticoid, and progesterone receptors and significant agonist activity for estrogen receptors [88]. ## Unconventional Unconventional oil and gas extraction, termed fracking, engineering jargon for hydraulic fracturing, is a technology used to recover large volumes of hydrocarbons (oil and gas) trapped in fine-grained, porous, low-permeable rock formations such as shale [89]. Industrial fracking opens the natural fault lines and fissures in these geologic formations using a large volume of fracturing fluid (also called “slick water”) that is injected into the rock at high pressure along with proppants (gritty material with uniformly sized particles like sand) to hold the fissures open [89]. Fracking is a chemical-intensive procedure [90]. Slick water includes chemical additives such as polyacrylamide and oxidizers (e.g., ammonium, potassium sodium salt of peroxydisulfate) to reduce viscosity, biocides to prevent degradation of polymer-containing fluids, and hydrochloric acid to clean the bores after fracking, as well as an array of other chemicals with numerous functions [8991]. Although chemical additives make up a small percentage of fracking fluid by volume ($0.5\%$–$2\%$), the massive quantity of fluid required to crack shale leads to the use of vast amounts of chemicals. An average injection of fracking fluid can total approximately 18,500 kg of additives per frack per well [92]. The chemicals used in fracking are not only abundant but also diverse. Over 1,000 chemicals have been identified in fracking fluids and/or wastewater [90], many of which are known or suspected carcinogens [93]. The fracking process generates multiple hazards, including airborne emissions of methane and other pollutants, groundwater contamination, induced seismicity, and the flammable and explosive nature of the extracted oils [8994]. In the US alone, producers have drilled over 52,000 shale gas wells. Failures of well casings, leakage from aboveground storage, emissions from gas-processing equipment, and the large number of heavy transport vehicles moving through previously isolated rural communities have been endemic, contributing to environmental contamination and human exposure [95]. ## Air pollution Fracking releases large volumes of air pollutants from over a dozen processes and sources [94]. Air pollutants commonly emitted from fracking include particulate matter (PM) and volatile organic compounds (VOCs) [95], especially methane [96]. Air pollutants can be emitted through direct releases, such as venting and flaring; from “fugitive” leaks; or from other junctures in the fracking infrastructure, such as wastewater pits, dehydrators, and pipelines [97]. Air pollutants can also originate from the volatilization of fracture fluid components and from proppant injection. A 2011 study noted that $37\%$ of chemicals used in US operations could escape to ambient air [93]. An examination of material safety data sheets and chemical databases such as TOXNET indicates that fracking chemicals are associated with a range of adverse health impacts, including respiratory disease; injury to the eyes, skin, sensory organs, and gastrointestinal (GI) tract; cardiovascular disease; and kidney and liver damage. Eighty-one percent of the fracking chemicals used have the potential to damage the brain and nervous system [93]. Air pollutants emitted from fracking operations can travel great distances [94]. For example, an assessment of air quality 1.1 km from a fracking site in western Colorado over the course of a year before as well as during and after its development and operation [98] detected a wide range of chemicals in every sample during the study, including VOCs (methane, ethane, propane, and toluene) and carbonyls (formaldehyde and acetaldehyde). Chemicals with the highest concentrations (in order) across the sampling period were methane, methylene chloride, methanol, ethanol, acetone, and propane [98]. Concentrations were highest during initial drilling and did not increase during fracking [98], but 30 of the detected chemicals affect the endocrine system. Concentrations of fracking-related air pollutants are also highest in proximity to fracking sites, and residents living within 0.8 km of wells are at greater risk for adverse health effects than residents living farther away [99]. Volatile hydrocarbons emitted from fracking operations can be photochemically oxidized in the atmosphere in the presence of nitrogen oxides (NOx) to form ground-level ozone. Ozone concentrations in several drilling regions have exceeded the current eight-hour national ambient air quality standard (less than 71 parts per billion) set by the US Environmental Protection Agency (EPA) [100] and have been shown to extend into nearby residential communities [3194]. Ozone levels above the air quality standard are considered harmful to human health, and both short- and long-term exposures to ozone are associated with adverse health effects, including preterm birth [101] and increased respiratory [102], cause-specific [103], and all-cause [104] mortality. Air pollutants produced in coal, gas, and oil extraction include particulate matter (PM), NOx, sulfur oxides (SOx), carbon monoxide, hydrogen sulfide, and volatile C5-C9 hydrocarbons. NOx and SOx are respiratory irritants. Carbon monoxide and hydrogen disulfide can cause sudden death at high concentrations and chronic neurobehavioral impairment at lower levels of exposure [966]. ## Water pollution From 2012 to 2014, 116 billion liters of water were used annually in the US alone to extract shale gas compared to 66 billion liters/year for unconventional oil extraction [105]. Once the slick water and proppant are injected and as the fracking process proceeds, a large proportion of this water returns to the surface, as “flowback” water; this mainly consists of hydraulic fracturing fluids and “formation water” (i.e., naturally occurring water from rock formations such as shale), which is associated with high rates of oil and gas production. With time, the fracturing fluids are depleted, and oil and gas production decreases; now the “produced” water is composed almost entirely of formation water [105]. Because they are contaminated with VOCs and other organic compounds, neither flowback nor produced water can be returned to drinking water supplies [106]. A systematic evaluation of 1,021 chemicals identified in fracking fluids and/or wastewater for potential reproductive and developmental toxicity found that data were available for only $24\%$ of these chemicals, $65\%$ of which presented potential adverse reproductive and developmental effects [90]. For example, BTEX, which are VOCs and known hazardous toxins, have been detected in wastewater from shale oil and gas fields in concentrations ranging from 96.7 μg/L to 9 mg/L [107]. Fracking wastewater also contains contaminants such as radionuclides (e.g., radium and radon) and metal(loid)s (e.g., cadmium, copper, zinc, chromium, lead, arsenic), which are drawn from the shale formation and flow to the surface in quantities capable of presenting significant hazards [108109110]. In theory, wastewater from the fracking process can be safely contained in tanks, disposal wells, and pits. In actual practice, however, this water and its chemical additives, toxic metals, and radioactive isotopes escape into local surface water and groundwater supplies to cause extensive contamination [106]. Spills and leaks into the environment can occur at every point in the fracking process—from blowouts, storms, and flooding events; unlined wastewater evaporation pits; accidents during transportation of chemicals and wastewater; outflows during mixing, pumping, and at numerous points during active fracking; and mechanical failure, all enabling seepage [110111]. Toxic pollutants, including benzene and other hydrocarbons, have been detected in surface and groundwater [95] near fracking sites, and higher levels of estrogenic, antiestrogenic, or antiandrogenic activities have been reported in surface water and groundwater from a drilling-dense region of Colorado compared to reference sites [87]. Leakage of hazardous chemicals from oil and gas extraction can result in pollution of surface water and groundwater with exposures of workers as well as community residents [31]. Chemical pollutants detected in water supplies near drill sites include 1,3-butadiene, tetrachloroethane, benzene, methane, ethane, propane, and toluene as well as methanol, ethanol, formaldehyde, and acetaldehyde [985987]. Many of these chemicals are associated with human health effects, such as cancer, skin and eye irritations, and GI effects [93988989990]. Indeed, exposure to contaminated water from oil and gas extraction has been shown to cause several health impacts, including neurological, GI, and dermatological effects [991]. ## Production—Transport of oil and natural gas Following its extraction, crude oil and natural gas are moved from the wellhead to the refinery and beyond using various means of maritime, pipeline, rail, and road transportation. Factors such as cost, speed, and distance largely determine the mode of transport. ## Maritime transport Due to their low cost and relative safety, tankers are among the most viable options for the transport of large volumes of crude oil and natural gas (liquefied petroleum gas, liquefied natural gas (LNG), and compressed natural gas), as well as their products and derivatives [112]. Tankers are used to transport $50\%$–$60\%$ of the world’s crude oil supply and are thus indispensable for the international oil trade [113]. As a proportion of international maritime trade, tankers have increased steadily over the past 40 years [114]. The development of ultra-large crude oil carriers has almost doubled deadweight tonnage from 0.337 billion in 1980 to 0.601 billion in 2020 [115]. Barges are also used to transport smaller quantities of crude oil and natural gas on inland waterways, and they are used together with tugboats for bulk transport to and from remote locations, such as offshore drilling rigs [116]. Maritime transport of crude oil and natural gas is hazardous, with diverse and far-reaching environmental and human health consequences when operations go awry. Between 1997 and 2007, there were 24 major oil spills, with releases ranging from, for example, 37,000 tons (Exxon Valdez, 1989) to 287,000 tons (Atlantic Empress, 1979) and shoreline impacts from 1 km to 3,000 km [117]. Oil spills threaten the lives of crew members and can also result in economic loss and resource damage, along with serious negative consequences for affected aquatic and terrestrial ecosystems and coastal communities [118119120121122]. The effects of seven major oil tanker spills on human health between 1989 (Exxon Valdez, Alaska) and 2002 (Prestige, Spain) have been reviewed. Epidemiological studies have found both acute and chronic health impacts [123]. Acute toxicity was seen among individuals involved in the cleanup and included headaches, sore throats and eyes, and respiratory symptoms, all known toxicological effects of oil. Genotoxicity assessment revealed DNA damage. Long-term adverse neurological effects, including stress and generalized anxiety disorder, were also reported [123]. The fraction of the oil involved in these spills that was destined for plastic manufacture is unknown. The physical properties of LNG determine its hazards, which include fire and explosion, vapor clouds, rollover, freezing liquid, rapid phase transition, and pool fire [124]. Hazards to the health of individuals on board LNG carriers include gas leaks (which can result in asphyxiation), cryogenic burns, and confined space hazards. LNG spillages can ignite, resulting in pool fires, which can generate damaging levels of thermal radiation that threaten the surrounding environment [125] and release combustion products that are hazardous to human health, including smoke, carbon monoxide, and carbon dioxide (CO2), among others [126]. ## Pipelines Pipelines require significant outlay to construct because they are fitted with multiple valves, pump/compressor stations, communications systems, and meters, but once built, they are one of the most efficient, cost-effective, and safest means of moving crude oil, petroleum products, and natural gas across vast distances [127]. Offshore pipelines are typically more expensive and difficult to build than land-based pipelines, which can be installed either aboveground or belowground [128]. Currently, the most extensive pipeline network is in the US [127]. In 2013, approximately 61,000 and 320,000 miles (that is, ~99,000 and 515,000 km) of the US pipeline network was dedicated to the transport of crude oil and gas, respectively [129]. Pipelines have a lower spill incident and fatality rate per Gt-miles of oil transported compared to other means used to transport crude oil. However, pipeline failures, deteriorating infrastructure, human error, and natural disasters can all result in major pipeline breaches, which can have severe and long-lasting impacts on public health, the environment, and regional economies [130131]. Gas pipelines are also associated with their own unique hazards. In addition to accidents and unintended leaking (fugitive emissions), intended releases (blowdowns) are used to relieve pressure, with plumes reaching 30–60 meters into the air. Although little is known of their full chemical content, these emissions are known to include methane, ethane, benzene (a known human carcinogen), toluene, xylene, 1,3-butadiene, and other compounds that are either known or suspected endocrine disruptors [132]. Bayesian modeling and simulation suggest causal effects between pipeline emissions and both thyroid cancer and leukemia [133]. Mechanical failure of welded seams, external mechanical damage, and severe internal corrosion of gas transmission pipelines also occur and result in explosions, fires, and the formation of large craters [134]. Sudden ground shifts caused by rapid thawing and freezing, even in mild climates, can induce severe subterranean forces [135]. In Massachusetts, in 2018, a series of explosions and fires resulted from the overpressurization by an order of magnitude of a low-pressure pipeline; causation was attributed to lack of hazard analysis and safety review [136]. In sub-Saharan Africa, gas explosions are largely unreported, but a 2014 review revealed 28 separate pipeline explosions that killed 1,756 people (injuries were seldom reported in Nigeria, Kenya, Ghana, Sierra Leone, and Tanzania), with the most common causes of the explosions being intentional via theft, vandalism, or military activity [137]. ## Rail transport Rail transport is used to transport crude oil and natural gas to refineries as well as refined products to downstream destinations, particularly when existing pipeline capacities are insufficient or do not link to required locations [138139140]. Rail has traditionally been considered a safe and efficient means of transporting oil [139]. However, increased use of railroads for the transportation of crude oil and a spate of highly publicized railroad accidents have revived concerns about the safety of transportation of crude oil by rail and its potential impacts on the environment [138141]. Derailment is of particular concern, as it may result in oil spills that release a large amount of environmentally harmful and flammable materials, which in turn may cause fires or explosions, environmental degradation, property damage, injury, or loss of life, as well as potential long-term health and psychosocial impacts [142143144]. ## Road transport Crude oil, natural gas, and refined petroleum products are also transported via roads. Although the highly distributed nature of road transport and the many different fossil carbon products that are transported make attribution to plastic production challenging, the proportion of fossil carbon feedstocks used to make plastic is predicted to increase, for example, to $20\%$ of global oil consumption by 2050 [7]. *In* general, truck drivers are regularly exposed to numerous occupational hazards, including whole body vibrations, noise, climatic factors, ergonomic hazards, and psychosocial hazards, such as stress, fatigue, and poor lifestyle habits (e.g., cigarette smoking, poor diet, lack of exercise) [145]. They are also exposed to PM, nitrous oxide, and carbon monoxide from diesel exhaust fumes. Traffic accidents also pose the risk of injury, permanent disability, or death [145]. India lacks crude oil resources, yet it is the third-largest global consumer of petroleum and the fourth-largest refiner [146]. However, India’s pipeline network is limited (16,226 km), and, as a consequence, the petroleum and chemical supply chain is almost entirely dependent on road transport [146]. Of the 705 road accidents reported between 2014 and 2019, $8\%$ involved chemicals, the remainder being fuel; health hazards were not reported [146]. India also has a large plastics industry, with 30,000 processing units and 2,000 exporters, that is currently worth US$37.8 billion and is projected to increase to US$126 billion in the next four to five years [147]. For this sector, improvements are needed to improve safety and reduce hazards. ## Production—Chemical and petrochemical industry Plastics are highly complex materials. Thousands of chemicals are used to manufacture many different polymers and enhance production and processing; thousands more are added to the polymer matrix to provide properties such as color, flexibility, or fire resistance and to enhance production processes [8]. Additional chemicals are inadvertently included (nonintentionally added substances [NIAS]; see Box 2.4) in the manufacture of plastics or are formed and released during its use [148149], both of which result in the presence of environmental contaminants in the final product [150151]. The chemical and petrochemical industry is the second-largest global manufacturing industry [152]. Global chemical production has increased 50-fold since 1950, with projections of a tripling between 2010 and 2050 [153]. It is estimated that over 235,000 chemicals with individual Chemical Abstracts Service Registry Numbers (or CAS numbers) have been registered in national or regional industrial chemical inventories, with a further 120,000 registered without revealing their assigned CAS numbers [36]. Plastics are one part of the complex industrial chemical web that uses 927 Mt per year of fossil fuel–based feedstocks (natural gas, natural gas liquids, naphtha, coal, and refinery feedstocks) to produce 820 Mt per year of chemical products [154]. Plastics make up the majority of these products ($40\%$), comprising thermoplastics (222 Mt, $27\%$) and thermoset plastics, fibers, and elastomers (107 Mt, $13\%$). The remainder include nitrogen-based fertilizers (274 Mt, $33\%$); solvents, additives, and explosives (107 Mt, $13\%$); and other categories (109 Mt, $13\%$) [154]. Environmental impacts from oil refineries include toxic air and water emissions, releases of chemicals, hazardous waste disposal, and thermal and noise pollution. Hazardous air pollutants include BTEX and n-heptane, while hydrocarbons, sulfur dioxide (SO2), and PM contribute to acid rain [78]. In the US, oil refineries are major industrial sources of SO2 and VOC emissions [78]. Impacts extend beyond the refineries. For example, in some US states, oil and gas wastewater is used to deice roads and control road dust. This wastewater can contain salt, radium, toxic metals, and organic contaminants, which runoff and likely reach surface water, groundwater, and water treatment plants [155]. ## Fossil fuel feedstocks for plastic manufacture Oil, gas, and coal are chemically transformed to provide the feedstocks for plastic manufacture. These feedstocks are naphtha, a product of crude oil refinement; natural gas liquids, a product of natural gas [73156]; and syngas, a mixture of carbon monoxide and hydrogen produced from gasified coal, which is then converted to methanol [157]. The predominant components of these feedstocks are ethane, methane, and propane, saturated hydrocarbons containing single bonds, also called alkanes. In the next stage of plastic production, large chemical plants termed “crackers” (catalytic cracking plants) are used to convert alkanes into olefins (also called alkenes), unsaturated hydrocarbons containing one pair of carbon atoms linked by a double bond. Cracking involves different technologies, depending on the feedstock. Natural gas liquids and naphtha are thermally cracked, i.e., broken down into small molecules at a high temperature [158]. By contrast, catalytic processes are used to convert coal-derived methanol into olefins [157]. Cracking also produces other petrochemicals used by the plastics industry, including benzene, toluene, and xylene (as solvents and raw materials); butadiene (used to make synthetic rubber); and styrene monomer [159]. An abundance of cheap oil and gas from fracking, coupled with large stocks of underutilized coal, and declining markets for hydrocarbon fuels are driving massive investment in fossil fuel–based plastic production in the US, China, the Middle East, and Europe [73]. The US Energy Information Administration reports current total monthly dry shale gas production to be 80 billion cubic feet per day [160]. Production of Appalachian natural gas liquids, including ethane for plastic production, is projected to grow faster than any other region in the US over the next 30 years [161]. Because construction costs for ethane cracking plants are very high—US$1 billion or more per facility and US$6 billion for one cracker in Pennsylvania [162]—their build-out will lock in plastic manufacture for many decades to come [73]. Cracking facilities produce multiple airborne pollutants. *They* generate substantial quantities of CO2, thus contributing to climate change. They also release methane, carbon monoxide, nitrogen dioxide, sulfur dioxide, VOCs, and particulate pollutants (PM10 and PM2.5) [159]. Steam cracking produces hydrogen [159], but given that the feedstocks are derived from natural gas, the hydrogen is not “green” but rather is termed “blue.” [ 163164] ## Monomers Monomers are small molecules that are polymerized (see “Polymerization, Compounding, and Conversion” below) to form polymers. Ethylene and propylene are two of the main monomers used in the production of the polyolefins polyethylene (PE) and PP, which are the highest volume industrial polymers [73]. Many other monomers are used to make other plastics [8165]. Many monomers are produced in very large quantities. Thus, an estimated 6 Mt of bisphenol A (BPA), a nonolefinic monomer used to make polycarbonate (PC) and epoxy resins, was produced in 2021 alone [166]. Production volumes of monomers are expected to grow, with an anticipated compound annual growth rate of $6\%$ for BPA dominating within the Asia-Pacific region [166]. BPA’s primary use ($95\%$) is as a monomer in the production of polycarbonate plastics and epoxy resins [167]. Ethylene is used to make PE (~$32\%$ of global plastic production), but it is also an important starting material for the synthesis of other plastic monomers, such as vinyl chloride, styrene, and ethylene glycol. The second highest volume monomer for industrial plastic production is propylene, which is used to make PP ($23\%$). Other plastic monomers include vinyl chloride, which is used to make polyvinyl chloride (PVC, $16\%$); styrene, which is used to make polystyrene (PS, $7\%$); and ethylene glycol and terephthalic acid, which is used to make polyethylene terephthalate (PET, $7\%$) [73]. A comprehensive classification of the environmental and health hazards of the monomers used in production of the major classes of plastics has been developed [165]. This classification reflects the intrinsic hazardous properties of the monomers but does not consider the extent of human exposure. The classification is based on EU Classification, Labelling and Packaging criteria, and it ranks monomers on a five-point log scale that considers several properties, including carcinogenicity, reproductive toxicity, respiratory effects, specific organ toxicity, ozone depleting, explosive hazard, and oxidizing hazard. On this scale, monomers are categorized variously as Phase Out (Level V), Risk Reduction (Level IV), and Acute Toxicity (Level III). The most hazardous monomers according to this particular classification criteria include acrylamide, ethylene oxide, propylene oxide, 1,3-butadiene, and vinyl chloride [165]. Hazard ranking of 55 plastic polymers with annual production volumes exceeding 10,000 tons found that 16 were Level V (e.g., polyurethane [PUR] foam, plasticized and rigid PVC, and high impact PS) and 15 Level IV (e.g., phenol formaldehyde resin, unsaturated polyester, and PC—with phosgene). Additional polymers with high-production volumes ranked as hazardous include PVC (37 Mt), PUR (9 Mt), and the styrenic polymer acrylonitrile butadiene styrene [165]. The least hazardous included PP, polyvinyl acetate, HDPE, and LDPE [165]. ## Processing aids and additives Over 10,500 different chemicals are used to make plastic [8]. In addition to monomers ($24\%$), Wiesinger et al. describe two other broad chemical classes used to manufacture plastics—processing aids and additives [8]. Processing aids ($39\%$) include antistatic, blowing, foaming, cross-linking, and curing agents as well as catalysts, heat stabilizers, antifoaming agents, lubricants, and solvents. Additives constitute the majority of the 10,500 ($55\%$) chemicals incorporated into plastics, and they impart a wide range of functional properties, including durability (antioxidants, light stabilizers, heat stabilizers, biocides, and flame retardants), strength (glass fibers and carbon fibers), and flexibility (plasticizers). It should be noted that antioxidants protect oxidative degradation of plastic [168] and can be hazardous, while other antioxidants are used as food preservatives and are not considered as hazardous [169]. Additives also include colorants (pigments and soluble azo colorants) and fillers (mica, talc, kaolin, clay, calcium carbonate, and barium sulfate), among numerous other functions (e.g., impact modifiers, odor agents, antistatic agents, lubricants, slip agents) [8170]. It should be noted that the same additive can have more than one function and that approximately $30\%$ of the ~10,500 chemicals added to plastics are uncategorizable due to lack of information [8]. Based on EU, US, and Nordic data, approximately 4,000 of these ~10,500 chemicals are considered to be high-production volume chemicals (that is, annual production exceeds 1,000 tons). Determining how much of this production can be attributed to plastics is difficult because these substances are used in other applications and many production volumes are considered proprietary information [8]. The amounts of additives in plastic products vary widely, with plasticizers being the most predominant ($10\%$–$70\%$ wt/wt, especially in PVC), followed by flame retardants ($0.7\%$–$25\%$ wt/wt), antioxidants ($0.05\%$–$3\%$ wt/wt), and ultraviolet light (UV) stabilizers ($0.05\%$–$3\%$ wt/wt) (Table 1 in Hahladakis et al., 2018 [170]). On a weight basis, fillers represent greater than $50\%$ of all additives used, followed by plasticizers, reinforcing agents, flame retardants, and coloring agents [171]. Rating systems are used to assess the hazards associated with additives and other chemicals used to make plastic. Commonly used criteria in these ratings include persistence, bioaccumulation potential, and toxicity (PBT); carcinogenic, mutagenic, and reproductive toxicity; endocrine disruption; chronic aquatic toxicity; and specific target organ toxicity upon repeated exposure [132172173174]. Assessments under EU Registration, Evaluation, Authorization, and Restriction of Chemicals (REACH) authorization lists [175] also feature candidate lists for substances of very high concern [176], PBT chemicals [177], and endocrine disruptors [178]. Reported hazard classifications in the regulatory databases are available for ~6,400 ($61\%$) of the ~10,500 known plastic-associated chemicals, while the other 4,100 substances ($39\%$) lack classifications in the databases accessed and are therefore considered as having unknown levels of concern [8]. Of the 6,400 substances with hazard classifications, 3,950 ($37\%$) are of a low level of concern, and 2,486 ($24\%$) meet one or more of the hazard criteria under EU REACH and are identified as substances of potential concern. For those with potential concern, 1,254 ($12\%$) are also high-production volume chemicals and are therefore classified as having a high level of concern, and 1,232 ($12\%$) have a medium level of concern. An overview of hazardous chemicals in plastics is given in Figure 2.4. **Figure 2.4:** *A multitude of hazardous chemicals are used, present, and released across all stages of the plastic life cycle. PUR, polyurethane; PAN, polyacrylonitrile; PVC, polyvinyl chloride; PBT, persistent, bioaccumulative and toxic; vPvB, very persistent/very bioaccumulative; EDCs, endocrine disrupting chemicals. References: [1](Wiesinger, Wang and Hellweg, 2021); [2](Lithner, Larsson and Dave, 2011); [3](Groh et al., 2019); [4](Food Packaging Forum Foundation, 2022); [5](Hahladakis et al., 2018); [6](Lowe et al., 2021).Credit: Designed in 2022 by Will Stahl-Timmins.* ## Selected plastic additives Given the large number and diverse array of plastic additives and other chemicals used to make plastics [8165170], we include summary information in the following subsection on human health hazards for two classes of plastic additives (flame retardants and plasticizers) that were selected on the basis that they are commonly used; are included in human biomonitoring programs, such as the National Health and Examination Survey (NHANES); and among all of the many additives are the best studied. We also included three types of stabilizers (heat, antioxidant, and UV) on the basis that this is an emerging field of research. ## Flame retardants Flame retardants comprise a range of chemical groups added to plastics to prevent or slow ignition and combustion. Organohalide flame retardants incorporate chlorine, bromine, or fluorine [179]. BFRs are the most widely used of the halogenated flame retardants [180181], but other groups, such as polychlorinated biphenyls (PCBs), have also been widely used [182], although they were also used in heat exchangers in electrical capacitors and transformers [183]. BFRs have been in use since the 1970s [181184] in industrial materials, electronics, and textiles and are also used in substantial quantities in the manufacture of plastic cabinets for televisions, personal computers, and small appliances as well as in PUR cushions for chairs and sofas [184]. Polybrominated biphenyls (PBBs) were the first generation of flame retardants. Their production was discontinued in 1976 because of their toxicity to humans and animals [185]. PBB replacements (“regrettable substitutions,” see Section 4) include polybrominated diphenyl ethers (PBDEs). Because these compounds are PBT, they are mostly restricted under the Stockholm Convention and have been restricted in the US since 2004 [186]. Other halogenated flame retardants, including decabromodiphenyl ethane, hexabromocyclododecane (HBCDD), tetrabromobisphenol A, PCBs, short-chain chlorinated paraffins, the PBB hexabromobiphenyl, PBDEs (including tetra-, penta-, hexa-, hepta-, and decabromodiphenyl ethers and HBCDD are each internationally recognized and regulated as POPs under the Stockholm Convention. Medium-chain chlorinated paraffins are proposed for listing [187]. Organophosphates are another class of flame retardants and have been increasingly used in recent years [188]. Organophosphate esters (OPEs) were introduced in response to concerns about the environmental and health toxicity of BFRs and chlorinated flame retardants [188]. Five main groups of OPEs are commonly used as flame retardants: organophosphates, organophosphonates, organophosphinates, organophosphine oxide, and organophosphites [189]. OPEs may also include various degrees of halogenation [190]. Some OPEs are used as plasticizers in floor polishes, coatings, engineering thermoplastics, and epoxy resins [191]. Volatile OPEs (tributylphosphate, triethyl phosphate, tri(2-chlorethyl)phosphate) are easily released from plastic products into the air, soil, dust, sediments, surface water, and biological media [192]. Humans can be exposed to OPEs through inhalation, ingestion, or dermal contact with dust, air, or water that contains OPEs [192]. An assessment of in-home exposure of 21 mother-toddler pairs to OPFRs found levels of three of five OPE metabolites tested in the urine of both members of virtually all pairs ($100\%$, $98\%$, and $96\%$) [180]. OPE metabolite levels in the mothers’ urine were significantly correlated with levels in children’s urine, suggesting similar exposure routes or maternal-child transfer (e.g., transplacentally or via breast milk). In children, predictors of hand-mouth exposure (e.g., less frequent hand washing) were associated with elevated OPEs levels, emphasizing the particular vulnerability of children to exposure to environmental contaminants. ## Plasticizers Plasticizers are chemical substances added to polymers to aid processing and increase flexibility [193194]. Plasticizers help to make plastics softer, more pliable, and more durable [195]. Although as many as 30,000 chemicals have been identified or proposed for use as plasticizers [193], only about 100 plasticizers are commercially produced worldwide, with half of these being commercially important [193]. In 2018, the plasticizers market was estimated to be worth around US$14 million and is projected to reach US$16.7 million by 2024 [196]. Ortho-phthalate diesters comprise up to $85\%$ of the total plasticizer market, of which di(2-ethylhexyl) phthalate (DEHP) and diethyl phthalate are the most common examples [197]. As a specific example, ~$97\%$ of DEHP’s use is as a plasticizer, with the remainder (~3–$5\%$) being used as a solvent (i.e., nonplasticizer) in personal care products, such as perfumes and cosmetics [198199]. Approximately $90\%$ of all plasticizers are used to manufacture flexible PVC [193196], which is used in products such as food packaging, children’s toys, furniture, medical devices, and adhesives [197200201202]. Like many additives, plasticizers are typically not covalently bound to polymers [194]. Consequently, they can be released from polymer products through a number of pathways, including volatilization to air, leaching into liquids, abrasion of polymer products, and direct diffusion from the polymer to dust on the polymer surface, and contaminate the environment, where they may pose a risk to human health [195]. Some phthalate plasticizers are hazardous [197] and are carcinogenic [203204] neurodevelopmental toxicants [205] and endocrine disruptors [206207]. Unlike flame retardants, phthalates are metabolized and excreted and are not biomagnified in the food chain [208]. Diet is therefore not a major route of exposure that occurs instead through other routes such as direct contact with consumer products [200201202]. Given their environmental ubiquity and adverse health effects, the use of six major phthalates has been restricted in consumer products, including cosmetics and children’s clothing and toys, and medical devices in several countries in Europe as well as in Japan and the US [209]. Alternative plasticizers to ortho-phthalates have been developed [195] and currently include adipate esters (e.g., di[2-ethylhexyl] adipate [DEHA]); sebacate esters; mono- and dibenzoate esters; cyclohexane dicarboxylic acid esters (e.g., di-isononyl cyclohexane-1,2-dicarboxylate [DINCH]); glycerol esters (including acetylated glycerides); phosphate esters; terephthalate esters (e.g., di[2-ethylhexyl] terephthalate); trimellitates (e.g., trioctyl trimellitate) and biobased alternatives, such as citrate esters (e.g., acetyl tributyl citrate [ATBC]); and epoxidized vegetable oils [195209]. The extent to which these substances have been implemented varies, with some having been utilized for several decades and others only entering the market more recently [195]. While alternative plasticizers are claimed to be safer than regulated phthalate plasticizers [209], few were tested for safety or toxicity prior to their commercial introduction. Like ortho-phthalates, they have the potential to leach out from plastic products, as they are typically not covalently bound to polymers [197]. These compounds have potential to pollute the environment and threaten human health, and concerns have been raised about regrettable substitution [197210211212]. See Section 4. ## Stabilizers Stabilizers are a class of additives that protect plastics from degradation by oxidation, ozone, heat, light (including UV), and bacterial attack [213]. They are incorporated into plastics during processing as well as in final plastic product formulation, and some stabilizers have multiple roles. The EU European Chemicals Agency’s (ECHA) “Mapping Exercise—Plastic Additives Initiative,” completed in 2018, classified substances registered under REACH at above 100 metric tons per annum [214] by function and lists stabilizer chemicals under “light stabilizers” ($$n = 17$$), “heat stabilizers” ($$n = 27$$), “antioxidants” ($$n = 26$$), and “other stabilizers” ($$n = 22$$). We follow this categorization here. Biocide stabilizers were not included here because they were not a focus of the ECHA mapping exercise. For the purpose of hazard assessment, stabilizers are best categorized by chemical class. “ Harmonized hazard scores” were estimated for six groups of stabilizer chemicals [213] based on a number of sources, including classification, labeling, and packaging for [1] environmental, [2] human health hazards, or [3] endocrine disruption (based on REACH, United Nations Environment Programme [UNEP], World Health Organization [WHO]); and/or PBT properties; and/or very persistent/very bioaccumulative properties (based on EU classification) [213]. Each of the six stabilizer groups listed (tin, organophosphite, hindered phenol, benzophenone, benzotriazole, and “other”) had high hazard scores on at least one of the above three criteria [213]. Stabilizers are a class of additives that protect plastic from degradation by oxidation, light, and heat [179]. ECHA’s “Mapping Exercise—Plastic Additives Initiative,” [214] classified stabilizer chemicals as “light stabilizers” ($$n = 17$$), “heat stabilizers” ($$n = 27$$), “antioxidants” ($$n = 26$$), and “other stabilizers” ($$n = 22$$). We follow this categorization here. A second approach to categorizing stabilizers involves hazard assessment by chemical class. Groh et al. estimated “harmonized hazard scores” for six groups of stabilizer chemicals based on [1] environmental effects, [2] human health hazards, [3] endocrine disruption (based on the EU REACH, the UNEP and the WHO), [4] PBT properties, and [5] very persistent/very bioaccumulative properties [213]. Each of the six stabilizer groups listed (tin, organophosphite, hindered phenol, benzophenone, benzotriazole, and “other”) had high hazard scores on at least one of the above three criteria [213]. Many stabilizers are additionally used in consumer products other than plastic (e.g., in household and personal care products) or have other major sources of human and environmental exposure. ## UV stabilizers Benzophenones (BzPs) were initially used as preservatives in paints and varnishes and were introduced as sunscreens in the 1950s [215]. BzPs are also used in cosmetics and other personal care products [215216]. Their ability to prevent UV light from damaging scents and colors allows products such as inks, perfumes, and soaps to be stored in clear packaging (glass or plastic) that would otherwise need to be opaque. BzP UV filters are added to plastics and textiles as UV blockers to prevent photodegradation [216217]. BzP concentrations of 7,426 ng/g in plastic waste from car recycling plants (vehicle “fluff,” namely scrap consisting of, e.g., plastic parts, textiles, seals, and tire fragments) and of 4,637 ng/g in e-waste (plastic cable granulates and plastic residue) have been reported [216]. Partition experiments show that different BzPs (e.g., BzP-1, BzP-2, BzP-3) leach in varying amounts from these products [216]. BzPs have also been found in environmental samples, such as wastewater [216] and indoor dust [218]. They have also been widely detected in wastewater treatment plants, sludge, fresh water, and sediments, with BzP-3 predominating, likely reflecting its use in both commercial and industrial products [219]. Benzotriazoles are another class of high-volume chemicals used as UV stabilizers in plastics as well as in a wide range of other products, such as detergents, dry cleaning equipment, and deicing agents [220]. Although the Japanese government banned the use of UV-320 in 2007, replacement benzotriazoles, such as UV-326, UV-327, and UV-328, are still widely used [221]. Indeed, in 2022, POPs Review Committee concluded that UV-328 fulfilled the screening criteria for inclusion in the Stockholm Convention [222]. Benzotriazoles have been shown to be present in plastic bottle caps, food packaging, and shopping bags [221]. The FCCmigex database identifies at least nine benzotriazoles where migration or extraction has been demonstrated from plastic food contact materials (galaxolide and the phenolic benzotriazoles UV-P, UV-P, UV-234, UV-326, UV-327, UV-328, UV-329, and UV-360, with 1–13 entries each) [223]. Benzotriazoles are highly lipophilic, bioaccumulative, and persistent in the environment and have been detected in human blood, breast milk, and urine [221]. Other UV stabilizers include triazine UV absorbers [224]; metal chelates, such as nickel phenolates; and hindered amine free radical scavengers [225]. These are less studied, but the FCCmigex database identifies at least two triazine UV absorbers where migration or extraction has been demonstrated from plastic food contact materials (UV-1164 and UV-1577, with one entry each), as well as at least one hindered amine (CAS number 42774-15-2, a tetramethylpiperidine-based hindered amine) [223]. Eleven different triazine UV stabilizers have recently been detected in human breast milk alongside benzotriazoles and BzPs [226]. ## Antioxidants Antioxidants in plastics include hindered phenols [227], secondary aromatic amines [228], organophosphite esters [229], nonylphenols [230], and organosulfur compounds (thioethers) [231]. Hindered phenol antioxidants include butylated hydroxytoluene (BHT), which is used in a large number of products, including food, cosmetics, pharmaceuticals, and fuels as well as plastics and rubbers [232]. Exposure is primarily via diet and occupationally via inhalation in workers handling BHT. Its metabolite (BHT-acid) has been detected in $98\%$ of samples in the German Environmental Specimen Bank, with median levels being slightly higher in women than men [232]. Nonylphenols are used as antioxidants (and plasticizers) in various resins. Concern about the endocrine-disrupting properties of nonylphenols led to increasing concern for human health, particularly if used in food contact materials. In fact, migration of nonylphenols from bottles into the water they contained was shown for HDPE and PVC bottles and caps [230]. Classes of antioxidants in plastics include hindered phenols [179227], aromatic secondary amines [179228], and organosulfur compounds (thioethers) [179231]. Hindered phenol antioxidants are primary antioxidants that act as hydrogen donors and may be used alongside secondary antioxidants [179]. They include butylated hydroxytoluene (BHT), which is used in a large number of products, including food, cosmetics, pharmaceuticals, and fuels as well as plastics and rubbers [232]. Human exposure is primarily via diet and also occupationally via inhalation in workers handling BHT [1137]. Its metabolite (BHT-acid) has been detected in $98\%$ of samples in the German Environmental Specimen Bank, with median levels being slightly higher in women than men [232], and in $78\%$ of samples in a study of urine samples from China, India, Japan, Saudi Arabia, and the US [1138]. Toxicological studies of BHT have reported carcinogenicity and reproductive toxicity [1138] as well as immunosuppression when administered to rats along with other food preservatives [1139]. BHT administration in pregnant mice resulted in maternal weight loss, reduced implantation sites, and failure of the uterine lumen to close [1140]. Endometrial decidual markers were reduced along with upregulation of serum estrogen, progesterone, estrogen-receptor-a, and the progesterone receptor [1140]. There are fewer data available on other hindered phenols. However, 9 of 14 hindered phenols listed as antioxidant plastic additives in the ECHA mapping exercise have been shown to migrate from food contact plastics [223457]. Epidemiologic studies of the possible adverse health effects of BHT are lacking. Despite reports of adverse health effects in toxicological studies, and the lack of epidemiological data regarding the safety of population-wide exposure, BHT is approved as a food additive in the US up to concentrations of $0.02\%$ (w/w) of total lipid content of food [1141]. Organophosphite antioxidants are secondary antioxidants that decompose hydroperoxides released by oxidative processes to unreactive products, but they are themselves converted to OPEs in that process [1142]. They have an additional role as heat stabilizers [214]. Both organophosphite additives and their OPE products have been detected in indoor dust and have thus been recognized as an important novel source of OPEs [1142]. Tris(nonylphenyl)phosphite, an organophosphite antioxidant used in plastic food packaging and demonstrated to leach from those materials [223457], can a be hydrolyzed (in the presence of acids within packaged food or by gastric acid following ingestion) to nonylphenol [11431144], which are known endocrine disruptors [1145]. Other antioxidants used in plastics include aromatic secondary amines (primary antioxidants) [179228] and thioesters (secondary antioxidants) [179231], including bis(4-(2,4,4-trimethylpentan-2-yl)phenyl)amine, distearyl thiodipropionate, and di-octadecyl-disulfide, listed in the ECHA mapping exercise [214]. We did not find any human exposure or health data for these three chemicals. ## Heat stabilizers Heat stabilizers are essential for durable products, particularly PVC, with long service lives, as they prevent chain reactions leading to decomposition when products are heated [233]. Metal compounds that contain, for example, lead, cadmium, or tin are often used as heat stabilizers because they prevent the chain reactions and also prevent product deterioration due to sunlight exposure [233]. Organotins are widely used as heat stabilizers as well as color control in PVC, with different types of stabilizers being used for different unplasticized (e.g., pipes, fittings, sheet) or plasticized (e.g., cable covering, flooring wall covering, medical footwear, food packaging) products [233]. Organotin compounds are also widely used as antifouling agents in marine paint [234235]. They are found extensively in the environment and in humans [236]. Heavy metal stabilizers are especially used in PVC, which has particular sensitivity to heat [179]. These include organometallic compounds containing lead, cadmium, or tin [179]. Lead in the form of lead salts has a long history of use as a PVC stabilizer. Concerns about widespread leaching of lead and its toxicity have triggered a proposal under REACH to restrict PVC products containing equal to or greater than $0.1\%$ lead in the EU [1146]. Nevertheless, legacy plastic products, such as PVC piping, children’s toys, and recycled plastics, are still a major concern as sources of environmental lead and other metal contamination [114711481149]. High concentrations of lead have also continued to be detected in children’s plastic toys in a number of countries in Asia, Africa, and South America [1150]. Lead damages human health at even the lowest levels of exposure; the WHO has determined that no level of lead in blood is safe [115111521153]. Lead is toxic to multiple organs in children, most notably the brain and nervous system [1151]. It is additionally associated with hypertension in children [115411551156], and in adults, increasing evidence indicates that lead is an important risk factor for cardiovascular and renal disease [1157]. Cadmium (stearate or laurate) was previously used to stabilize PVC conferring heat stability and weatherability. The use of cadmium salts as stabilizers in plastics is now minimal and was less than $1\%$ of total consumption in 2009 [11581159]. Cadmium was also used as a pigment in plastics, particularly for brightly colored products such as toys, and in ceramics and paint [1160]. Cadmium is toxic at low levels, and systematic reviews have documented associations between cadmium exposure and increased all-cause mortality, cancer mortality, and cardiovascular disease [11611162]. Organotins remain widely used as heat stabilizers and are used also to make antifouling paint for ships [179]. They are dispersed extensively in the environment, especially the ocean, and in humans [2361163]. Dietary intake is considered the main route for human exposure [11631164]. Organotins are associated with endocrine and metabolic disturbances, including appetite regulation in toxicological studies [11651166]. Animal and in vitro human studies have revealed the association of organotins with oxidative stress and inflammatory responses [1167]. Organotins have been shown to affect fertility and reproductive function [1168], and in human infants, placental organotin concentrations are associated with increased risk of congenital cryptorchidism [1169]. As far as other heat stabilizers, nonmetallic organic stabilizers are increasingly used in place of metal stabilizers due to environmental and health concerns [179]. These are typically based on phosphites [179214]. Phosphite esters are discussed in the “Antioxidants” section above. ## Polymerization, compounding, and conversion All plastics are based on polymers—large molecules formed by the joining of monomers. During polymerization, mixtures of different monomers are often used to make different polymers. For example, flexible PUR foam is made from propylene oxide (58 wt%), ethylene oxide (7 wt%) and toluene-diisocyanate (29 wt%) monomers; epoxy resin is made from BPA ($45\%$ wt%), epichlorhydrin (37 wt%) and 4,4′-methyleneadianiline (18 wt%). Additional chemicals used in polymerization processes include initiators, catalysts, and solvents. Virgin plastics are typically produced in granular or pellet form [156], also called nurdles [237]. Once formed, polymers are compounded with additives and processing agents to produce the properties required for specific plastic products [156]. During compounding, multiple additional chemicals, including additives and processing agents (see above), are inserted into the polymers (Figure 2.5) to instill specific properties, such as color, stability, flexibility, water repellence, and UV resistance. During conversion, this mixture is converted into products using various mechanical processes, such as film extrusion, sheet extrusion, extrusion coating, blow molding, foam molding (involving expansion), injection molding, raffia, and PET conversion. Processing also involves heat and pressure to mold plastics into required shapes. Once cooled, products are trimmed, ground, drilled, and sanded and then assembled and painted. Different polymer types are used to make different categories of products. For single-use plastics, approximately $90\%$ are made from five main polymer types, namely PP, HDPE, LDPE, linear-LDPE, and PET [65238]. The remaining single-use plastics are made of polymers such as PS, PVC, and polyamide (PA) [65]. **Figure 2.5:** *Plastic is a complex chemical mixture of inter-twined polymers comprising multiple monomeric units joined by carbon-carbon bonds and multiple chemicals added to enhance production and impart properties such as flexibility, strength and durability. Straight chains are polyvinyl chloride (PVC) polymers. Additives depicted are di(2-ethylhexyl) phthalate (DEHP), di-n-butyl phthalate (DnBP), a polybrominated diphenyl ether congener (PBDE-47) and a polychlorinated biphenyl congener (PCB-153).Credit: Manuel Brunner, co-author.* There are two broad categories of plastic: thermoplastics and thermosets. Thermoplastics can be softened repeatedly by heating and then hardened by cooling, thus allowing reshaping. Thermoplastics include HDPE, LDPE, linear-LDPE, PP, PC, PVC and PS [239240]. By contrast, thermosets are plastics that change their chemical nature once heated and cannot subsequently be remelted or reshaped. Thermosetting plastics include PUR and a large number of resins, such as epoxies, melamine, acrylic, unsaturated polyesters, and vinyl esters [239]. ## Nondurable/single-use and durable plastic Plastic products can be classified into nondurable, which includes single-use plastic, and durable plastics. Nondurable and single-use plastics, both flexible and rigid, have lifetime usages of three to six months and up to three years, but they may be designed to be “disposable” and thus are typically used only once [347]. In 2019, the Global Plastic Waste Makers Index estimated that a total of 376 Mt of 14 discrete polymer types were converted into a similar volume of single-use and durable products globally [65]. ## Single-use plastics—Consumer and institutional products Single-use plastics accounted for 133 Mt, or $35\%$, of the 376 Mt of plastic produced in 2019 [65]. Of these 133 Mt, $99\%$ (130 Mt) went into packaging. The remainder (3 Mt) was used to make single-use consumer and institutional products [65], such as disposable food service ware, kitchenware, household and institutional refuse bags, and personal care items [241]. Other sources confirm that packaging is by far the major application of single-use plastic in Europe ($40\%$) [12239]. In the US, an analysis of plastic flows in 2017 showed that $27\%$ of all plastic is used for packaging ($27\%$; Table 2 in Heller et al., 2020 [47]). ## Durable plastic Durable plastics accounted for the remaining $65\%$ (243 Mt) of the polymers produced in 2019. These materials were used to make durable products in a number of consumer and industrial sectors [65], including synthetic fibers (66 Mt, $17.5\%$ of total polymers), building and construction (62 Mt, $16.4\%$ of total polymers), transport (16 Mt, $4\%$ of total polymers), electrical and electronics (16 Mt, $4\%$ of total polymers) and “other” (83 Mt, $22\%$ of total polymers) for agriculture, homewares and furniture, and large industrial containers [65]. ## Synthetic fibers A substantial proportion of plastics is used to make synthetic fibers, for example, polyester and nylon, which is used in clothing, furniture upholstery and other household goods, such as carpets and curtains [12242]. Global production of natural animal and plant fibers has grown slowly over the last 30 years, whereas synthetic fiber production has increased to almost 65 Mt per year [242]. Currently, synthetic fibers constitute almost two-thirds of all textile fibers produced globally, dominated by polyester [12]. Processes used to convert raw polymers to finished fiber-based products, such as dyeing, impregnating, coating, and plasticizing, involve multiple hazardous chemicals [243]. Azo dyes are made of aromatic amines, which are highly carcinogenic as well as genotoxic and allergenic [244]. Aromatic amines were among the first synthetic chemicals to be produced and were documented as early as 1895 to cause cancer of the bladder [245]. Disperse dyes (small polar dye molecules) used to stain synthetic fibers have been reported to be the most common causes of textile allergy [246247]. Quinoline and its derivatives are extensively used in the manufacture of dyes and have also been detected in synthetic textiles, with some being skin irritants and probable human carcinogens [248]. Benzothiazoles and benzotriazoles are found in azo dyes [220] and have been reported in clothing with printed graphics, including socks and infant bodysuits [249]. As well as being dermal sensitizers, these chemicals are genotoxic and can act as endocrine disruptors [249]. Synthetic textiles with PVC prints, which contain high levels of phthalates, are common, especially in children’s clothing. Phthalate concentrations (as sum of all phthalates) ranging from 1.4 mg/kg to 200,000 mg/kg ($20\%$ by sample weight) have been reported in non-peer-reviewed work by Greenpeace [250]. The dominant phthalates found in this investigation were DEHP, butyl benzyl phthalate (BBP), di-isononyl phthalate (DINP), and diheptyl phthalate. BPA and its analog bisphenol S (BPS) have been detected in synthetic textiles, including those marketed for infants, at concentrations of 15–366 ng/g [251]. It is unknown to what extent these chemicals are taken up by infants via, for example, mouthing. We note, however, that the draft European recommendation for tolerable daily phthalate intake is 0.4 ng/kg [252]. Similarly, phthalates in concentrations above EU limits (i.e., $0.1\%$ of mass: European Commission Regulation No. $\frac{552}{2009}$) have been reported in a range of infant products, including nylon sheets, cot mattresses, and diaper changing mats [253]. In addition to dyes, a number of metal(loid)s are found in some synthetic fabrics, including high concentrations of chromium, copper, and aluminum in polyester and high concentrations of nickel and iron in nylon. Concentrations of these metal(loid)s, such as chromium, lead, and nickel, have been found to exceed recommended limits in some clothing samples tested [248]. Some synthetic textiles are treated with formaldehyde-releasing compounds and resins to prevent creasing. Formaldehyde, a known carcinogen, has been detected in clothing items at concentrations 40 times higher than specified by international textile regulations [254]. More recently, durable-press fabrics are being produced that release less formaldehyde. Some synthetic textiles are treated with metal nanoparticles, which act as antimicrobial (silver) or UV absorption (titanium) agents. However, product disclosure is largely lacking, and the forms and concentrations in which these metals are precent are unknown. Nevertheless, dermal exposure is hypothesized to be a likely exposure route [248]. Exposure to airborne microplastics (MPs) during manufacture of synthetic textiles is an occupational risk, although natural fibers such as cotton also carry health risks [255]. ## Building and construction Building and construction is the third-largest market sector for plastics after packaging and textiles and is the least visible [12]. Industry specifications require durability over decades as well as strength and flame resistance. A wide range of thermoplastics (e.g., PC glazing) and thermosets (e.g., PUR foam insulation) meet these standards and are thus widely used in the building and construction industry. Plastics offer some specific advantages over traditional building materials because they are resistant to heat transfer and moisture diffusion, and do not undergo metallic corrosion or microbial degradation. They can also be molded into different shapes, come in a range of colors and textures, and require minimal or no painting [256]. Products include pipes for supply water and sewage removal; electrical insulation; temperature insulation; waterproof membranes in walls, ceilings, window frames, and window profiles; and flooring [256]. PVC is the most commonly used plastic in the building and construction sector and accounts for almost half ($43\%$) of the plastic used in Europe [12], with construction accounting for $75\%$ of the European PVC market [257]. Global PVC demand was 35 Mt in 2007 and has grown since. A range of heat stabilizers are used in both rigid and flexible PVC products (670,000 tons globally in 2006), including lead-based, mixed-metal stabilizers (e.g., calcium-zinc, barium-zinc) and tin-based stabilizers [257]. ( Stabilizers are discussed in detail in the “Selected plastic additives” section.) The organotin compounds used in PVC as heat stabilizers represent approximately two-thirds of global organotin consumption [258]. A Greenpeace study of PVC flooring revealed concentrations of 330–48,800 µg/kg for monobutyltin, 37,700–569,000 µg/kg for dibutyltin, and 128–17,940 µg/kg for tributyltin [259]. Phthalate plasticizers (580,000 tons globally in 2006) are also a major constituent in certain PVC building materials [257]. PVC is widely used in flooring because it is inexpensive and easy to clean, making it highly practicable in kitchens, bathrooms, and children’s playrooms [259]. Flexible PVC is commonly plasticized with phthalates, historically DEHP and more recently DINP [260]. Analysis of floor samples in 2000 showed DINP concentrations of $4.7\%$–$15.8\%$ mass/mass in all five samples tested and BBP concentrations of $1.6\%$–$5.0\%$ mass/mass in three of five samples [259]. Phthalates are not bound to the polymer matrix and are therefore leachable. Phthalates vaporize from linoleum flooring to air and have been detected in suspended air particles and sedimented dust in homes as well as in water after cleaning plastic floors [259]. Pregnant women living in homes with PVC flooring have significantly higher urinary levels of the BBP metabolite (mono-benzyl phthalate) than pregnant women living in homes made with other flooring materials [261]. Exposure to PVC flooring in early life, including during pregnancy, is related to the incidence of childhood asthma at 10 years of age [262]. As another example of leaching from PVC products, MP shedding occurs inside water pipes and acts as a vector for the release of dibutyl phthalate [263]. Sources of data on alternative building materials include the Pharos and HomeFree databases from the Healthy Building Network [264]. ## Transport, electrical, and electronic The transport, electronic, and electrical sectors have been revolutionized by the transition from metals to plastic materials reinforced with glass or carbon fibers. Such reinforced plastics (also referred to as “polymer matrix composites”) are associated with an array of functional properties, including high durability, strength, dimensional stability, thermal resistance (>300°C), and the ability to dampen mechanical vibration. They do not require anticorrosion painting. All of these properties result in their being highly versatile and widely used within these sectors [263265266]. Reinforced plastics made from both thermoplastic and thermosetting polymers are highly amenable to 3D printing, which enables their use in multiple applications, including electronics, printed circuit boards, aerospace, and biomedicine [266]. Reinforced plastics are approximately one-third lighter than metal alloys, and products from them can be molded, therefore requiring far fewer parts [265]. Because of their light weight and durability, they have significantly improved performance in every mode of road, air, and marine transport [267]. The Boeing 787 *Dreamliner is* made of $50\%$ plastics, with the remainder being metal alloys, steel, and other materials. Efficiencies are gained not only in production but also during operation. For example, a 600-fleet airline with an average capacity of 200 seats per plane saves 1.9 million liters of fuel and 4.5 million kg of CO2 per year due to the use of reinforced plastics in aircraft construction [265]. Hazards to human health during use of reinforced plastic products are as yet unknown although MP shedding from these materials is known to occur [268]. ## Agriculture Plastics are widely used within the agriculture sector. Macroplastics are commonly used in greenhouses, sheds, covering films and nets, irrigation, implements, twine, netting, storage equipment, and mulch; in addition, pesticides and fertilizers contained in MNPs are used for slow release and targeted delivery [269]. Plastics also enter soils as a result of waste mismanagement and from sewage-derived fertilizer [270]. It has been estimated that plastics constitute $0.1\%$–$0.6\%$ of soil in Swiss floodplain soils [271]. Plastics in agriculture are a source of several chemical pollutants and thus a soil contaminant. Chemicals of concern that leach into soil from plastics include phthalates and residual monomers such as BPA [269]. Metal(loid)s (cadmium, chromium, copper, zinc, nickel, lead, and arsenic) have also been detected in soils and vegetables grown in plastic shed production systems in China [272]. A study in Spain estimated plastic waste to be almost 250,000 kg over 1,500 hectares per year [273]. A recent project using satellite imaging and artificial intelligence revealed that agricultural plastic is a major component of land-based plastic waste [274]. Plastics in soils can alter soil formation, stability, and hydrology and are likely to alter primary nutrient production and cycling [275]. Furthermore, under experimental conditions, MP infill in artificial sports turf has been observed to reduce plant growth, possibly due to the leaching of additives [276]. A recent study using fluorescently labeled PS beads under controlled laboratory conditions revealed their uptake by wheat and lettuce roots, with transfer to the epidermis and xylem [276277]. Nanoplastic (NP) particles made from PS applied at high concentrations (g/kg or mg/L) have been found to trigger mild reactions in plants resulting in increased root stress. It is not yet known whether environmental MNPs are also taken up into plants from soil under natural conditions, nor whether or to what extent they may bioaccumulate through the food chain. Soil can also be polluted by chemicals (e.g., metalloids, polycyclic aromatic hydrocarbons (PAHs), VOCs, PCBs, PBDEs, polychlorinated dibenzo-p-dioxins (PCDDs), “dioxin-like” compounds) released during the production and disposal of plastics [278279280]. Analysis of farmland soils in China has revealed that flame retardants, including OPFRs and BFRs such as PBDEs, are widelydetected [281]. Plastic packaging for animal feedstock has been found to contain several bisphenol compounds, with BPA being the predominant form, and evidence of leaching of these compounds from the products [282]. Microplastics have also been detected in animal feedstock such as fish meal and soybean meal [283]. ## Single-use and durable Modern medicine has been transformed by plastic. Quantifying plastic use in medicine is challenging, but procurement data and waste audits provide a high-level overview of consumption of products used in medicine such as the following: A 2021 audit across five hospitals in Europe showed that plastics comprise $70\%$ of sanitary waste and $34\%$ of general waste [284]. Plastic waste has been recognized as a substantial footprint in the health care sector, and various procurement programs that aim to replace products with nonplastic alternatives are underway [285]. See Box 2.2 for more information about the contributions of the health care sector to plastic waste production. The COVID-19 pandemic has resulted in large increases (approximately $350\%$–$370\%$) in plastic waste. It has been reported that, in one day, care for the inhabitants of Wuhan, China (population ~11 million people) generated 200 tons of medical waste, a volume four times greater than the capacity of dedicatedincinerators [286]. Toxic chemical additives are found in a wide range of plastic-based medical supplies [287], and medical supplies are an important source of human exposure to phthalates, BPA, and PFAS (e.g., see Table 4 in [288]). Phthalates can account for $30\%$–$40\%$ of medical-use plastics by weight [287]. Phthalates are used in medications to control GI drug delivery. Parenteral nutrition enhances leaching from plastic medical products such as blood bags, endotracheal tubes, and cardiopulmonary bypass machines [289]. Leaching of DEHP from PVC-based medical equipment made more than 40 years ago has been described [290]. A 1982 study revealed DEHP metabolites in nonuremic psoriatic patients and in uremic patients undergoing hemodialysis or cardiac bypass surgery [291]. A more recent study on serum and urine samples from 35 adult intensive care unit patients also revealed exposure to DEHP, and to a lesser extent BPA, with patients on hemofiltration, extracorporeal membrane oxygenation, or both showing higher levels (100–1,000 times) than the general population [292]. A 2005–2006 audit of DEHP in medical devices in a large neonatal intensive care unit (NICU) via website searches and phone interviews found that 10 of 21 ($48\%$) devices were DEHP-free and that gaps exist with respect to alternatives to both the PVC polymer and alternative plasticizers [293]. A two-center (97 neonates) study examining PVC medical devices used for infusion therapies, parenteral nutrition, blood transfusion, and respiratory ventilation found urinary DEHP, di-(2-ethylhexyl) terephthalate, and tri-(2-ethylhexyl)trimellitate metabolites in exposed patients [294]. Larger exposures were found for DEHP compared to di-(2-ethylhexyl) terephthalate (5–10 times) and tri-(2-ethylhexyl)trimellitate (57–228 times) [294]. Both phthalates and BPA have been detected in NICU patients, and phthalates are associated with increased risk for cholestasis, necrotizing enterocolitis, and bronchopulmonary dysplasia in newborns [295296]. In very low birth weight infants, in a single-center study, serendipitous replacement of IV fluids with a DEHP-free formulation over a two-year period resulted in the near elimination of neonatal hypertension, an effect that was reversed when the original brand of DEHP-containing IV fluids was reintroduced [297]. This finding is consistent with broader epidemiological evidence of an association between phthalate exposure and increased blood pressure in children [298]. ## Box 2.2 Contributions of the Health Care Sector to Plastic Waste Production. The origin of the phrase primum non nocere (first, do no harm) is uncertain. Although likely to be hundreds rather than thousands of years old [299], it certainly predates invention of the synthetic plastics so widely used in health care today [300]. Global demand for the manufacture of plastic medical disposables, including personal protective equipment and medical devices, more than doubled between 2005 and 2020 [301]. Drivers of this increase include concerns of infection control, which have driven a transition to single-use items, and a culture of defensive medicine that has led clinicians to be overcautious. Prior to the COVID-19 pandemic, plastic medical waste accounted for $23\%$ of the total waste produced by the UK National Health Service (NHS) [284], with an average of 3.4 kg of single-use materials utilized every day when treating a patient with severe sepsis [302]. Between April 2019 and March 2020, NHS *England* generated 624,000 tons of waste, $53\%$ of which was incinerated, $29\%$ recycled, $11\%$ alternatively processed, and $7\%$ sent directly to the landfill [303]. The COVID-19 pandemic resulted in an increased demand for single-use products [286304], demonstrating the vulnerability of health care systems reliant on single-use rather than reusable equipment [305]. Present legislation related to the management of medical waste and lack of clarity on recyclability leads to the majority being incinerated, whether used or not [303306]. With climate change predicted to cause approximately 250,000 additional deaths per year from malnutrition, malaria, diarrhea, and heat stress, with estimated direct health costs of US$2billion to US$4 billion per year by 2030 [307], the health care profession needs to review its use of “planet-harming technologies and products which fuel its own workload.” Yet, when the concept of “first, do no harm” is applied to a specific patient, clinical tunnel vision can take over. Thoughts of product origin, cost, and planetary or distant community harm are forgotten, as clinicians’ main concerns become protecting the health of their patients, their colleagues, and themselves. Health care systems need to take a broader view and be cognizant of the wider consequences of decisions they make as to which products to use and their potential for reuse. There have been some limited successes in reducing demand for specific items, e.g., nitrile gloves [308]. Larger reductions in health care’s plastic footprint will, however, require a systems-level approach [285] to reverse years of culture and procurement practices, now made worse by the COVID-19 pandemic. Fortunately, the profession is becoming increasingly aware that resource extraction and the manufacture, use, and disposal of plastic medical devices cause harm to both people and the planet. This heightened awareness now needs to translate to a full assessment of options, balancing impacts of single-use items with the chemical and energy use needed to decontaminate reusable equipment. Health care systems, including the UK NHS, need to urgently reexamine the products used to provide care, and legislation that hinders progress needs to be reviewed. Up to $50\%$ of health care’s plastic waste arises from packaging [284], which offers the greatest opportunity for change. Health care services have a vested interest in reducing their impact on the ecosystem; otherwise, they will become increasingly burdened through climate-related acute and chronic ill health [307], which will increase demand for the very products that are harming the environment. In addition, there are presently many unanswered questions about the direct and indirect impacts of the MPs found in human blood, breast milk, and lungs. While macroplastic health care–associated pollution can be seen and sorted in the health care system, it is essential not to forget sources of equally important MPs. For example, $3.5\%$ (9.5 billion miles a year) of all road travel in *England is* linked to NHS activity [309], which results in significant amounts of MPs generated through tire and brake pad wear. These MPs contribute to particulate air pollution and are washed into storm drains and thereafter the ocean [310]. Reusable textiles are a source of plastic microfibers produced during washing that are then discharged into hospital wastewater [311312]. Health care can not only drive change through its considerable buying power but also through the narratives its actions generate. The NHS has 80,000 suppliers, and through communication, it can ensure that these companies understand the need for change. This health care provider-supplier collaborative approach can be replicated globally, help drive positive actions, improve health and well-being, and reduce the impact of providing health care on the planet. Considerations for health care systems, organizations, and professionals to help reduce plastic waste production include the following: ## Iatrogenic exposure An additional emerging route of iatrogenic exposure to plastics is through novel nanoparticle drug delivery systems. These systems promise a number of potential benefits in drug delivery, including transport of medications that may otherwise require complex solvents, targeted delivery or release, and sustained release [313314315]. Those synthetic polymer technologies that have made it onto the pharmaceutical marketplace have largely been based on saturated aliphatic polyesters, such as polylactic acid (PLA), polyethylene glycol, or a combination of a saturated aliphatic polyester and polyethylene glycol in block copolymer micelles [314316], expected to be fully metabolized and excreted [313314]. There are however a much more diverse range of polymers and nanoparticle technologies in the experimental and developmental phases, beyond the scope of this report, with hydrogel and dendrimer applications being particular areas of focus [313314315]. ## Disposal Disposal is the third main component of the plastic life cycle, after production and use. With continuing year-to-year increases in global plastic production, especially in the production of single-use plastics, and continuing increases in the accumulation of plastic waste, the issue of waste disposal has become increasingly prominent. Reduce, reuse, and recycle are the three traditional components of waste management programs. While these strategies have proven highly effective for paper, cardboard, glass, and aluminum, especially when they are coupled with economic incentives such as container deposit fees, they have largely failed for plastic, where recovery and recycling rates are below $10\%$ globally. The result of continuing near exponential increases in plastic production coupled with very low rates of recycling is that each year an estimated 22 Mt of plastic waste enters the environment. *Global* generation of plastic waste has more than doubled in the past two decades, from 156 Mt in 2000 to 353 Mt in 2019 [5]. For single-use plastics, 131 Mt was produced in 2019, increasing to 137 Mt in 2021, with predictions of 148 Mt in 2027 [317]. A total of more than 6 Gt of plastic waste have accumulated worldwide since 1950. Strategies currently used for disposal of plastic waste include controlled and uncontrolled landfilling, open burning, thermal conversion, and export. Of the plastic waste produced globally in 2019, almost half was disposed of in sanitary landfills, $19\%$ was incinerated, $9\%$ was recycled, and $22\%$ was discarded into uncontrolled dumpsites, burned in open pits, or leaked to the environment [5]. In the US, an analysis of plastic flows in 2017 showed that $8\%$ was recycled, $14\%$ was combusted, $76\%$ was landfilled [47]. Vast quantities of plastic waste are exported each year from high-income to low-income countries, where it accumulates in landfills, pollutes air and water, degrades vital ecosystems, befouls beaches and estuaries, and harms human health—environmental injustice on a global scale. Plastic-laden e-waste is particularly problematic. If current trends in plastic production continue unchecked, it is anticipated under a business-as-usual scenario that the dominant means of plastic waste disposal will continue to be landfilling, and by 2060, the volume of mismanaged plastic waste produced each year could triple to 155–256 Mt each year [14]. A disproportionately high fraction of this future waste is expected to be produced in Africa and Asia [318]. ## Recycling Recycling is an important component of modern waste management systems, which encourage collection of waste by assigning it an economic value, thereby reducing mismanagement and leakage and displacing primary production. Sectors like glass (EU ~$75\%$), paper (EU ~$70\%$), and aluminum (EU ~$65\%$) packaging have high recovery and recycling rates in many modern economies [676869]. Such efficient waste recovery systems are an important step toward a more circular, less wasteful, and less polluting economy. Plastic recycling, by contrast, accounts for less than $10\%$ of the new plastic produced globally each year [5]. For single-use plastic, closed-loop recycling (also referred to as “on-par” recycling, for example, bottle-to-bottle recycling) is even lower, being $1\%$ from recycled feedstocks in 2019 and $2\%$ in 2021, with predictions of $3\%$ in 2027 [317]. Current plastic recycling rates by region are US ~$5\%$, Middle East and Africa ~$5\%$, China ~$13\%$, India ~$14\%$, and Europe ~$14\%$ [319]. The amount of recycled plastic is increasing somewhat, having quadrupled over the past two decades from 6.8 Mt in 2000 to 29.1 Mt in 2019 [5], but recycling rates are still too low to displace primary production. Thus, virgin plastic production and use continue to soar, and plastic waste continues to accumulate in ever-growing amounts [314320]. Factors that contribute to the global failure to recycle more substantial quantities of plastic are the sheer volume of plastics; unfavorable economics (the high cost of collection, waste transport, and recycling vs. cheap virgin plastic); the wide variety of resins, polymers, multilayer, and composite plastic products; and plastics’ chemical heterogeneity [848]. Additional impediments are the extensive use of plastic materials of close-to-zero material value (e.g., small-format packaging such as sachets) and the decrease in plastic quality after recycling (or downcycling). A particularly intractable problem is the inclusion in plastic products of multiple toxic chemicals, such as phthalates, BPA, flame retardants, and heavy metals. The presence of these materials in recycled plastic limits its use in consumer products with high potential for human exposure, such as food packaging. ## Mechanical, physical, and chemical recycling technologies Several technologies have been developed for recycling plastics and include both mechanical and chemical technologies. Plastic waste is also burned in waste-to-energy facilities, euphemistically termed thermal recycling or pyrolysis, to produce heat or energy [321]. Plastic recycling processes may have negative environmental impacts, such as the generation of air pollution and toxic ash [322]. Exposures to these waste products are associated with health impacts in recycling workers and in residents of nearby “fenceline” communities [323]. ## Mechanical recycling Mechanical recycling is a process in which plastic waste is sorted by polymer, shredded, washed, and melted into plastic granulates. This process is suited for thermoplastic materials (e.g., PET, PE, PP, PVC) but generally cannot be applied for thermosets like epoxy resins and most PURs. Mechanical recycling can be divided into closed-loop or open-loop recycling [324]. Today, closed-loop mechanical recycling is largely limited to PET bottles (bottle-to-bottle) in waste systems where the bottles are collected in a separate waste stream. The majority of mechanical recycling is an open-loop process in which the resulting plastic granulate is of lower quality than the input material. Multiple factors contribute to the lower quality of plastic recycled via this technique. They include polymer degradation during extrusion, cross-contamination with undesired polymers, incomplete removal of odors and colorants, and unknown additive composition—notably the potential inclusion of toxic additives—in the final granulate used for product manufacture. As a result, most mechanically recycled plastics do not meet standards for such uses as food contact materials, which means the plastic is downcycled to products with lower specifications [322]. To overcome the poorer material properties of recycled plastic, mechanical recycling often involves enrichment with additives or virgin polymer. Depending on the plastic type, melting during the extrusion process can emit toxic chemicals into the workplace air, including VOCs (e.g., vinyl chloride, styrene, formaldehyde, benzene) and PAHs [325326]. Melting plastic pellets recycled from waste plastic releases additives such as phthalates and VOCs in considerably higher quantities than are released from melting virgin plastic pellets [327]. These toxic chemicals are also being released during the granulation step of mechanical recycling [328]. Workers employed in mechanical recycling operations can be directly exposed to carcinogenic metal(loid)s such as arsenic, cadmium, and chromium present in recycled plastic pellets via skin contact and through inhalation of contaminated airborne dust [329]. Mechanical recycling is described in engineering jargon as a mature technology with a technology readiness level of 8–9 (TRL 1–9: 1 = observation of a new phenomenon (science); 9 = proven application in real-life scenarios). However, mechanical recycling can be improved with advanced sorting technologies, i.e., cleaner feedstock; better pretreatment methods, like a hot washing step; and postprocessing, such as deodorization [330]. ## Physical recycling (purification) Physical recycling (often referred to as purification) is a process in which sorted plastic is treated with solvents to selectively dissolve one or multiple types of polymers [322]. Solvent treatment is followed by a series of steps intended to remove contaminants and additives and by selective precipitation of target polymers. The product is a purified polymer precipitate that can be extruded and compounded to produce close-to-virgin plastic. Physical recycling has gained interest in recent years because it has the potential to address complex plastic waste that is almost impossible to recycle mechanically. Because it conserves plastic’s material properties, physical recycling is regarded as a closed-loop process. The physical recycling technology is, however, not mature at this time (TRL 4–7) and is currently only operating at small commercial scale (e.g., CreaSolv process in Indonesia). ## Chemical recycling Chemical recycling covers a wide range of processes [5]. The common element in all of them is that the chemical structure of polymers is changed. There are two major types of chemical recycling: depolymerization and conversion. Depolymerization technology involves cleaving plastic polymers into their initial monomeric units, which can then be used in a polymerization process to yield virgin-like polymers. Depolymerization can be performed chemically (chemolysis/solvolysis), biologically (enzymolysis), or thermally (thermolysis). It is best suited for PET plastics (most common), PS, and PA plastics, but less so for polyolefins [322]. Depolymerization is regarded as a closed-loop process because additives and contaminants can be removed, and the resulting polymers are indistinguishable from their virgin counterparts. These polymers can then be used to make the same or different higher-value materials (higher than those resulting from mechanical recycling). This is called upcycling. Consequently, depolymerization is increasingly gaining traction in the circular waste economy, with multiple pilot and demonstration plants emerging (TRL 3–7) [322]. Its effectiveness at scale is unproven. Conversion is a set of processes that involves breaking down plastic waste into small molecules that can then be used as chemical feedstock in petrochemical facilities. The resulting material can contain organic contaminants, toxic chemical additives, and NIAS [331]. Conversion processes include gasification (TRL 5–8), pyrolysis (TRL 3–9), and hydrothermal conversion (TRL 4–7), based on temperature, catalyst, and reaction medium: Theoretically, all polymer types can be converted, and chemical recycling has the potential to remove contaminants, additives, and toxins from the plastic waste stream and under certain circumstances uses low-quality plastic (e.g., from downcycling) and mixed-plastic waste as feedstock. In practice, however, conversion technologies, too, have a number of disadvantages: Occupational health hazards to workers in chemical recycling include exposures to toxic solvents (e.g., chloroform, xylene, n-hexane, cyclohexane), to airborne emissions from pyrolysis and gasification (e.g., styrene, hydrogen), and to hazardous waste (e.g., chars, tars) [323332]. The pyrolysis process can result in the production and release of toxic chemicals such as VOCs, PAHs, PCBs, and PCDDs [326]. While adequate management of recycling plants and use of personal protective equipment can mitigate exposure to toxic chemicals [323], some pilot pyrolysis plants have not been equipped with all of the abatement systems intended for use in such facilities, resulting in the risk of occupational exposures of workers to noncondensable gases such as VOCs [333]. ## Recycled products The ultimate goals of recycling are to curb plastic pollution, reduce production of new virgin plastic, and contribute to a circular economy in which recycled plastic is converted into new products. These aspirational goals are evident in such documents as the EU’s Directive on Single-Use Plastic, which requires all PET bottles to be made of $25\%$ recycled plastic by 2025 [334]. The use of a large number of chemicals in plastics is, however, a persistent impediment to recycling and to achieving global plastic circularity goals. These hazardous chemicals have been shown to leach out of recycled plastic in larger quantities than from virgin plastic [335336]. Chemical analysis of products (children’s toys, fabrics, tires, food contact materials, and construction materials) made from recycled plastics has shown higher numbers of flame retardants, fragrances, solvents, biocides, pesticides, and dyes compared to products made from virgin plastics, suggesting greater exposure hazards [331]; 14 of 20 chemicals with high toxicity scores (abundance, detection frequency, and bioactivity) had a higher prevalence in recycled plastics compared to virgin plastics [331]. A further problem is that restricted/banned chemicals such as legacy POPs may be unintentionally reintroduced into the market if outdated plastic products are included in the recycling waste stream [337]. For example, PBDEs and PBBs have been detected in children’s toys made from recycled plastic in several countries [338339340]. Another problem is that chemical contaminants may be introduced into recycled plastic from multiple sources during disposal, collection, and processing [151341342343344]. For example, analysis of food containers and plastic films made from recycled PET revealed metal(loid) contamination (average concentrations: cadmium 8.8 ppm; chromium 6.8 ppm; nickel 9.4 ppm; lead 0.2 ppm; antimony 8.3 ppm) [345]. Likewise, higher concentrations of antinomy and BPA have been detected in recycled compared to virgin plastic [336]. A systematic evidence map of chemical migration from recycled PET food and drink bottles showed that, of 193 chemicals examined, over 150 chemicals migrated from the polymer into food samples [336]. These included chemicals such as antinomy, acetaldehyde, and a number of endocrine-discrupting chemicals (EDCs). Eighteen of the 150 chemicals detected were found in concentrations above EU regulatory limits, with 109 and 113 being nonauthorized substances and NIAS, respectively [151336]. ## E-waste Discarded electronic products, or e-waste, include small and large household appliances, computers, mobile phones, lighting, tools, toys, sports equipment, medical devices and batteries, circuit boards, plastic casings, cathode-ray tubes, activated glass, and lead capacitors [346]. Approximately $20\%$ of e-waste by weight is plastic [347]. Plastics commonly used in the manufacture of electronic equipment are acrylonitrile butadiene styrene, PS, PC, PVC, PE, and PP [348]. These materials are used for electrical and thermal insulation as well as for manufacture of intrinsically conducting polymers, screens, casings, cables, films, and machine parts [348]. Plastic additives in electronics vary in their amount and include antioxidants (~$1\%$), heat and light stabilizers (up to ~$5\%$), plasticizers (e.g., DEHP ~$50\%$ in films and cables), and colorants (~$1\%$) as well as mold release agents, foaming agents, mineral fillings, and coupling agents [348]. More than 200 different types of flame retardants are used and can constitute up to $15\%$ of the plastic in electronic products [348]. Flame retardants include chlorinated, brominated, phosphorus-based aluminum-trihydrate and its derived inorganic rehydrate compounds. Metals, either in plastics or from other sources, are also present [348]. All of these toxic additives are found in e-waste [348]. Because of its chemical complexity and toxicity, e-waste and e-waste recycling pose significant hazards to human health [349350]. OPEs, toxic metals, and POPs, such as PCBs, BFRs, and PFAS, are released into the environment in e-waste recycling operations—both formal and informal [351352353]. Even in a formal e-waste recycling plant in Sweden, despite industrial hygiene improvements that successfully reduced occupational exposures, workers had higher serum PBDE concentrations than unexposed controls [354]. Similarly, in Canada, e-waste recycling workers have been shown to have higher concentrations of OPE metabolites in their urine and of PBDE and lead in their blood compared to glass recycling workers [355]. PBDE concentrations were associated with changes in thyroid function, and OPE metabolite concentrations were associated with changes in sex hormones [355]. The hazards of e-waste recycling are magnified in low-income and middle-income countries (LMICs), where much recycling occurs in the informal sector, emissions controls are few, and a substantial fraction of the workers are young children and women of child-bearing age. E-waste recycling in LMICs poses threats to the health of workers and residents of nearby communities [356357358]. In India, for example, the average concentrations of PAHs, phthalates, BPA, and toxic metals in soil were higher in informal e-waste recycling sites than in other dumpsites [359]. Analyses of air, dust, and soil close to e-waste recycling facilities in southeast China revealed higher concentrations of BFRs, POPs such as PCBs, polychlorinated dioxins and dibenzofurans (PCDD/Fs), perfluoroalkyls (from fluoropolymers), and PAHs (including pyrene and benzene derivatives) compared to control areas. Metal(loid)s (lead, chromium, cadmium, mercury, zinc, nickel, lithium, barium, and beryllium) were also detected in these samples [346]. Open burning of e-waste, such as plastic-covered cables to dispose of plastics and recover copper and other metals, is an especially hazardous practice [356357358]. Open burning of plastic-coated cables generates dense clouds of black smoke that can contain PAHs, dioxins, and VOCs such as benzene. Exposures to workers, children, and pregnant women living within or near unregulated e-waste recycling sites have been associated with a range of negative health outcomes, including changes in thyroid function, altered gonadal hormone levels, adverse birth outcomes, and altered growth due to exposure during pregnancy (see systematic reviews [346360]). ## Incineration—Controlled and uncontrolled Because they are organic carbon, plastics can be incinerated. Controlled combustion with sufficient oxygen at very high temperatures (>1,000°C) mainly produces water, CO2, and trace chemicals [361]. However, this idealized type of combustion requires well-resourced infrastructure, which is mostly lacking in the LMICs where much plastic incineration occurs. The consequence of increasing export of plastic waste from higher-income countries to LMICs is uncontrolled waste disposal using landfilling and open fires [362]. A systematic review of open burning of plastic waste, mainly in the Global South, examined emissions of eight hazardous substances, including BFRs, phthalates, dioxins and related compounds, BPA, PM, and PAHs [362]. The authors concluded that large quantities of mismanaged waste, including plastic, are threatening the health of ~2 billion people in LMICs, with highest risk to an estimated 11 million waste pickers who lack safe workplaces and protective equipment. Waste incineration, including plastic waste incineration, is estimated to account for $39\%$ (approximately 334 million kg) of global atmospheric aerosol emissions [363]. These aerosolized materials eventually precipitate to accumulate in soils and sediments [362364]. Mixed waste includes general household waste, tires, and agricultural and construction materials, much of which contains, or is made of, plastics. Uncontrolled incineration of mixed waste, either from landfill or backyard burning, releases a multitude of toxic substances, including PM, PAHs, PCDDs and related compounds, BFRs (bromophenols, HBCDD, PBBs), VOCs, and semi-VOCs [363365]. Open burning of PVC plastic is particularly problematic [366]. During uncontrolled, low-temperature, open combustion, PVC acts as a chlorine donor, which leads to the substantial formation of dioxins and furans that can become airborne [366]. Estimates of the health impact of PCDDs vary in different regions. For example, PCDD release during open burning of municipal solid waste in *India is* associated with 0.1–0.2 excess cancer cases per 100,000 people, and co-incineration of waste and coal in Poland indicates excess cancer cases of 4.5–$\frac{13.2}{100}$,000 [362]. Additional toxic emissions from PVC incineration include carbon monoxide, hydrogen chloride, and PAHs. Burning of plastic waste is estimated to account for $39\%$ of global PAH emissions [362]. PAHs are potent human carcinogens and have been estimated to contribute to 8.7 cases of cancer per one million people exposed [367]. BFRs are another hazard in plastic waste combustion. Although the manufacture of many of the most highly toxic of these compounds is now prohibited by the Stockholm Convention [187368], these older BFRS can still reside in legacy plastics and thus enter the plastic waste stream [369]. Analysis of plastic waste, as well as virgin and recycled plastic, revealed a wide range of BFRs, including bromophenols, dibromophenols, hexabromocyclodecane stereoisomers, and PBDEs [369]. Acrylonitrile butadiene styrene, PS, and PE can have high concentrations of BFRs [370]. Incineration of plastic waste results in the release of gases, particulates, and ash and the formation of brominated dibenzo-p-dioxins and dibenzofurans [371372]. Analyses of hair samples from populations near e-waste recycling sites indicate that younger people (aged 15–45 years) who were more likely to be involved in recycling and waste management had higher concentrations of brominated compounds than children or older adults [371373]. PBDEs have been linked to reduced birth weight, type 2 diabetes, endometriosis, cardiovascular disease and cancers [26]. Toxic metals in combusted plastic waste include mercury, cadmium, lead, chromium, and nickel. Mercury and lead are neurotoxic, while chromium and nickel are carcinogenic [374]. Antimony is used as a synergist in the production of BFRs, and arsenic is used as a biocide [170]. Although these metal(loid)s do not tend to migrate from plastics during use, they are released into airborne and precipitated soot during plastic incineration [362]. Little is known of the fate during recycling and waste disposal of the multiple other chemicals incorporated into plastics. A study that examined open burning of plastic materials such as shopping bags, roadside trash, and landfill waste found elevated concentrations of two chemicals—1,3,5-triphenylbenzene and tris(2,4-di-tert-butylphenyl)phosphate)—that are not found in wildfire smoke [375]. A study of atmospheric aerosols from urban, rural, marine, and polar regions showed a positive correlation between 1,3,5-triphenylbenzene and BPA levels [376], a finding that suggests that open burning of plastic waste is a source of widespread atmospheric BPA [375]. In Europe, where incineration is controlled, atmospheric sources of BPA as well as plasticizers are thought to contribute to human exposure [377]. ## Waste-to-energy Waste-to-energy facilities, euphemistically called thermal recycling facilities, convert plastic waste to energy. This conversion involves the production of a wide range of hazardous chemicals, including chlorine, hydrogen chloride and phosgene, hydrogen cyanide, and ammonia as well as formic acid, formaldehyde, benzene and its derivatives, phenol, and PCDD/Fs. Depending on the facility, substantial quantities of these chemicals will be released to the atmosphere [321]. The main sources of these toxic combustion products are PVC and condensation polymers such as PUR, PA, and phenyl-formaldehyde resins as well as chemicals in plastics. Waste-to-energy conversion also generates CO2. Under controlled conditions, formal waste-to-energy plants may remove a large portion of these pollutants from their emissions. For example, in China, while the waste-to-energy incineration capacity increased by $150\%$ from 2015 to 2020, the total emissions of toxic gases decreased by $42.46\%$–$88.24\%$ due to improvements in gas cleaning treatment [378]. However, co-incineration of fossil fuel with $15\%$ of refuse-derived fuel (containing $35\%$ plastic, $30\%$ paper, $20\%$ wood, and $15\%$ textiles) in a cement plant in Spain has been shown to emit similar amounts of PM, PCDD/Fs, and metals compared to normal operations with $100\%$ fossil fuel [379]. While there are limited investigations of the health impacts of exposure to waste-to-energy emissions, a recent systematic review [380] reported that depending on the use of sorted/unsorted waste and gas cleaning technology, waste-to-energy processes may be associated with increased cancer and noncancer risks [381]. ## Incineration When plastic is burned, airborne PM is released. PM consists of “black carbon” [382] in the form of char or ash, PM10, and PM2.5 [362]. It can contain heavy metals, VOCs, PAHs, and PCDD/Fs [383384385386]. Plastic burning has been estimated to contribute to $6.8\%$ of PM2.5 in Nanjing, China [387]; $13.4\%$ of PM2.5 in Delhi, India [388]; and up to $3\%$ and $15\%$ of PM2.5 in Atlanta, Georgia, and Dhaka, Bangladesh, respectively [389]. Evidence suggests that PM solids together with PAHs may be more deleterious to health than PM alone [390]. Thus, PM2.5-bound PAH are both carcinogenic and mutagenic [77] and may be linked to immunological and developmental impairments and reproductive abnormalities [391]. ## Microplastics Data are lacking on the contribution of MPs to airborne PM, and so respirable exposures in humans cannot be estimated reliably [171]. However, in an urban setting, MPs were found in all air samples, with deposition rates ranging from 575 to 1,008 MPs/m2/day, the majority ($92\%$) of which were fibrous with 15 different petrochemical-based polymers being identified [392]. Nevertheless, airborne MPs have been identified as an emerging source of PM pollution [393394]. MPs can be released into the atmosphere from a range of sources, including compost spreading, wastewater sludge, tires, textiles, and paint [393394395], and are potentially transported by wind currents. Several MPs have been detected in the atmosphere in various forms (fibers, fragments, film) and include PE, polyester, and PUR [396]. However, little is known about the extent to which people are exposed to airborne PM MPs, and further research is required to better understand the implications for human health. Microplastics, or MPs, is a term used to denote plastic particles less than 5 mm in diameter. MPs are further classified into two categories: primary MPs and secondary MPs [5]. Primary MPs are plastic particles that have been manufactured to a small size and intentionally added to consumer products for cosmetic and biomedical purposes; they also include microfibers that are shed from synthetic materials. Primary MPs can leak into the environment from multiple sources across the plastic life cycle. Secondary MPs, on the other hand, are small plastic fragments that arise from the degradation (breakdown) of larger plastic items, particularly plastic litter, as a result of natural weathering processes after they have entered environment [5]. Nurdles are tiny plastic preproduction resin pellets from which plastic products are manufactured. Nurdles are also considered to be primary MPs due to their size and can be unintentionally released into the environment during their manufacture, land and maritime transport, loading, storage, conversion, and recycling [417418]. These resin pellets have been found in rivers, waterways, and the ocean and are often seen on beaches along coastlines around the world [419]. Up to 53 billion nurdles may be spilled annually in the UK alone [420]. Microfibers are shed to the environment from synthetic textiles during production, use, and disposal [421]. It has been reported that by the fifth washing cycle, a total of 30,000–465,000 microfibers will have been released per square meter of synthetic fabric [422]. MPs are released into wastewater, the environment, and the human body from multiple sources during use, including abrasion of tires [423], paint (e.g., marine coatings [412]; see Box 2.3), road markings [424425], and glitters [426]. Another source is microbeads that are intentionally added to rinse-off personal care products (e.g., face scrubs, toothpaste) [427]. Shed microfibers are found in the air [428], wastewater [429], rivers [430], and ocean [431432]. MPs can be generated in plastic use and during mechanical recycling of plastics, especially during cutting and shredding stages, and are released into the environment or into wastewater [433]. Additionally, MPs can enter the environment via leakage from controlled landfill sites where geomembrane liners fail. Analyses of leachates from active and closed municipal waste sites revealed PE and PP polymers as the predominant MPs, with size ranges of 100–1,000 µm [434]. Furthermore, unregulated incineration of solid waste can also result in the production and release of MPs as PM into the air and on land [418]. MPs can also be formed from degradation of already leaked land- and marine-based mismanaged macroplastic waste [418]. Airborne MPs can travel long distances via atmospheric transport to reach remote areas [435436437]. Airborne MPs and MPs deposited on land can be washed into the aquatic environment or into sewage systems via stormwater and/or surface runoff [418438]. The deposition of airborne MPs in a coastal city has been estimated at 4,885 ± 1858 MNPs/m2/day (mean ± standard deviation; range: 82–12,159 MNPs/m2/day); the highest level at an urban rooftop correlated with coastal winds, suggesting that airborne MNPs may originate from wave action [439]. MPs detected in wastewater and sewage systems include PE, PP, PS, PA, PET, and PVC [440]. These materials can come from households and industrial activities [441]. Before being released into the aquatic environment, wastewater is treated in wastewater treatment plants. Varying amounts of MPs are removed in these facilities through preliminary, primary, secondary, and tertiary treatment [442], with $88\%$ of MPs removed from wastewater by applying the first three treatment processes and $94\%$ removed with more advanced tertiary treatments, such as reverse osmosis [443]. While the removal efficiency of MPs from wastewater is high, the residual concentration of up to 54 microparticles/L treated wastewater adds up to a significant total amount of MP leakage into the aquatic environment given the very large volume of urban wastewater [443]. Additionally, sewer systems that combine sewage and stormwater through a single pipe can be overloaded during heavy rainfalls [418], and in these circumstances, untreated wastewater, with all its MP content, is discharged directly into rivers and the ocean [444]. MPs removed from wastewater are concentrated in sewage sludge. If applied to the land as a fertilizer, this can be a major source of MPs to the environment, accounting for an estimated 0.66 Mt of all plastic leakage in 2019 [5]. A recent study reported that each gram of dry solid sewage sludge contains 0.01 g of MPs [445]. Due to its high organic and nutrient content, sewage sludge is commonly used as a sustainable soil conditioner or fertilizer on agricultural lands in many countries [446447]. As a result of this circular process, MPs captured during wastewater treatment may accumulate in terrestrial environments [448] or enter aquatic environments via surface runoff or infiltration into groundwater [449]. More MPs are estimated to enter the soil from the use of wastewater sludge for agricultural purposes each year than MPs entering the ocean overall and freshwater sediments [447450]. As many as 3,500 MP particles/kg of dry soil have been reported in samples of agricultural soils that were subject to 10 years of continuous sewage sludge disposal [451]. The presence of MPs in sewage sludge poses a threat to soil health and productivity [452453454] and could cause harm to soil-dwelling biota [455]. ## Landfill Plastic waste is projected to almost triple in volume globally by 2060, with half still being landfilled at that time and less than $20\%$ recycled [14]. The amount of solid waste that enters landfills varies widely across countries [239397]; for example, Switzerland recycles $25\%$, uses $75\%$ for energy recovery, and has no landfills. Other European countries, by contrast, report $20\%$ recycling, minimal energy recovery, and $80\%$ landfill [239]. Formal landfills in Europe accumulated over 5.25 Gt of waste between 1995 and 2015, of which $5\%$–$25\%$ by weight is estimated to be plastic [398]. This contrasts with other global regions, which often lack formal landfill facilities [399]. Landfilling is associated with environmental pollution, including groundwater pollution due to the leaching of organic and inorganic substances contained in the waste, odor pollution from degradation products, and pollution of surface waters from runoffs [400]. Air pollution is an important negative impact of gas emissions from landfills, with the GHGs methane and CO2 predominating [397]. Fugitive aromatic compounds with different levels of dispersion are also emitted. These include BTEX and naphthalene. Analysis of concentration profiles emitted from the working face of municipal solid waste dumpsites in China showed wide variations. For example, toluene levels were up to 90 µg/m3 within 200 m and dropped to approximately 12 µg/m3 at 800 m; benzene levels were 12 µg/m3 within 200 m and dropped to approximately 2 µg/m3 at 800 m [401]. Potential carcinogenic risk zones were also calculated. The carcinogenic risk impact distances for benzene and ethylbenzene were 710 ± 121 m and 1,126 ± 138 m downwind of the landfill; cumulative carcinogenic risk distances were higher at 1,515 ± 205 m, and, for worst-case scenarios, the cumulative risk was estimated to persist at over 4 km from the landfill site [401]. Flame retardants (PBDEs) have also been detected in landfill [402]. ## Leakage of Plastics to the Environment Lack of end-of-life management for plastic products, particularly in LMICs, coupled with the economics of cheap virgin plastic versus expensive recycled plastic has resulted in the failure to recover most plastic-based materials and retain their economic value [65403]. As a result, there is significant “leakage” of plastics throughout the plastic life cycle out of the economy and into the environment (Figure 2.6). In 2019, an estimated 22 Mt of plastics leaked into the environment, with macroplastics accounting for $88\%$ and manufactured MPs for $12\%$ [5]. **Figure 2.6:** *Plastic life cycle: Plastic leakage. Plastic and plastic-associated chemicals leak into the environment across all stages of the plastic life cycle. Chemical-laden macroplastics constitute the bulk of plastic leakage. Plastic polymers can take many years to degrade in the environment with the rate of degradation depending on many factors such as temperature, light and mechanical action.[9] Mt, Megatons; PET, polyethylene terephthalate; PVC, polyvinyl chloride;LDPE, low-density polyethylene. References: [1](Organisation for Economic Co-operation and Development (OECD), 2022a); [2](Karlsson et al., 2018); [3](Organisation for Economic Co-operation and Development (OECD), 2021); [4](Cole and Sherrington, 2016); [5](Paruta, Pucino and Boucher, 2022); [6](Hann et al., 2018); [7](Periyasamy and Tehrani-Bagha, 2022); [8](Ryberg, Laurent and Hauschild, 2018); [9](Chamas et al., 2020).Credit: Designed in 2022 by Will Stahl-Timmins.* Vast quantities of plastic waste from developed countries are exported to LMICs [404405]. China was previously the largest destination for plastic waste, with an estimated 6.6 Mt of plastic waste imported in 2017 (~$54\%$ of total exported plastic waste) [405]. Following China’s January 2018 ban on the import of nearly all plastic waste [406], exports from developed countries shifted to Southeast Asian countries, including Thailand, Malaysia, Vietnam, and Indonesia as well as Turkey [49]. These countries often lack the infrastructure to properly manage plastic waste [5399], thus increasing the likelihood of environmental leakage [407408409]. The volume of this leakage is estimated to have been 19.4 Mt in 2017 [399], of which $58\%$ came from Asian countries, including $23\%$ from China [5]. Other sources of macroplastic leakage include littering and marine activities. Littering is the second-largest contributor, with an estimated 1.1 Mt of leaked plastic attributed to littering in 2019 [5]. Littering mainly involves incorrect disposal by consumers of plastic products, most commonly single-use plastics such as plastic packaging, straws, and cutlery. Marine activities, including fishing, can lead to copious amounts of macroplastic leakage directly into the ocean [399], largely in the form of abandoned, lost, or otherwise discarded fishing gear (ALDFG) [410]. A recent study estimated that $75\%$–$86\%$ of floating plastic mass in the North Pacific Garbage Patch can be attributed to fishing activities [318]. ## Box 2.3 Microplastics and Toxic Chemicals from Paint. It has become increasingly apparent that the total volume of marine microplastics (MPs) cannot be solely attributed to mismanaged consumer waste. Therefore, there must be other contributing sources [411]. Paint is increasingly recognized as an important source of marine MPs. Paint, including water-based acrylics, can contain up to $50\%$ plastic, including polyurethanes, polyesters, polyacrylates, polystyrene, alkyls, and epoxy resins as well as additives, adhesion promoters, thickeners, antiskinning agents, and emulsifiers [412]. Antifouling paints contain high concentrations of hazardous inorganic additives as antifouling agents as well as an array of toxic metals, such as Cu+/Cu2+, tributyl tin+ (which was banned in 2008 [413] but is still circulating in legacy materials), Pb2+, and CrO42– [412]. Paint MPs and their potentially toxic additives are shed into the environment from ships and other marine megastructures, such as rigs, as well as from road markings and the external surfaces of buildings [414]. Ships have been reported to leave “skid marks” of MPs in the ocean consisting of high-density polyethylene, low-density polyethylene, polypropylene, polyvinyl chloride, polyethylene terephthalate, and polyurethane [411]. Intentional releases of paint MPs into the environment occur during power-abrasion cleaning of ships during dry dock maintenance [415]. To reduce paint shedding in dry dock operations, closed-loop vacuum blasting is being developed commercially as a capture technology [416]. An estimated 42 billion liters of paint are applied to marine megastructures globally each year [415]. With a 20-year life span, or $5\%$ annual loss of marine paint, this equates to 2–3 Mt of paint MPs released to the ocean every year [415]. ## Nonintentionally added substances (NIAS) Nonintentionally added substances, or NIAS, is a term that was first introduced in the food industry (Box 2.4). It refers to chemicals that are not intentionally added to foods or other materials, but they enter foods or other consumer goods in manufacturing or via contact materials such as food wrappings [148]. NIAS are now known to include an enormous variety of synthetic chemicals and may outnumber intentionally added substances in food products [456]. Plastic-associated chemicals, such as many additives and NIAS, are present in plastic products as mobile and leachable components within the plastic matrix and have been shown to leach from everyday plastic products made from different polymer types [459]. Controlled aging of biopolymers such as PS, PP, PET, LDPE, and HDPE at 40°C over four weeks results in a progressive increase in the release of VOCs. Testing of MP debris collected from a beach led to the release of VOCs, including benzene, acrolein, propanal, methyl vinyl ketone, and methyl propenyl ketone, thus highlighting an additional invisible hazard of plastic pollution [461]. Both new PVC pellets and beached preproduction pellets/nurdles have been shown to leach PCBs and PAHs to seawater [462]. Additionally, beached PS pellets can degrade into styrene oligomers [463], which have been detected in sand samples from 26 countries [464]. In the same study, BPA was found in seawater samples [464]. Leaching of plastic-associated chemicals during use is specially problematic for food contact materials [170457] because it can result in human exposure via ingestion—e.g., leaching from plastic baby bottles [465] or from PET food containers and drink bottles [336]. A recent systematic evidence map identified DEHP, dibutyl phthalate, BPA, DEHA, and 2,4-di-tert-butylphenol as the five most frequently detected plastic-associated chemicals leaching from food contact materials [457]. Inhalation is another route of exposure to plastic-associated chemicals released during use. For example, semi-VOCs such as PAHs, phthalates, organophosphates, and BFRs in household products (e.g., electronic devices, furniture, carpets) have been shown to vaporize into indoor air [466467]. Human exposure to plastic-associated chemicals can also occur via the dermal route, as these chemicals leach from products in contact with skin, such as textiles [248] and personal care products [468469]. ## Box 2.4 Nonintentionally Added Substances (NIAS) in Plastic Food Packaging. NIAS in plastic products include impurities present in raw materials and/or additives used during production, oligomers and other by-products of polymer production, degradation and transformation products, contaminants from machinery, and contaminants introduced in recycling [148149150]. Breakdown products are a major source of NIAS in plastic food contact materials [149], and one breakdown product, 2,4-di-tert-butylphenol, is among the most commonly detected chemicals in migration and extraction studies on plastic and multimaterial food contact products [457]. Oligomers are also common. Cyclical silicone (polydimethylsiloxane) oligomers are among the most commonly detected migrants from plastic and multimaterial food contact materials [457], although this may reflect bias due to available methodologies for detection or selection of materials studied. A systematic evidence map of food packaging plastics detected through published migration and extraction studies has recently been developed [148223]. Migration of polyolefin oligomers into food from polypropylene food containers has recently been demonstrated during microwave heating [223458]. Multilayer packaging offers an additional level of complexity, with transformation products that can arise from the polymer and additives in plastic layers as well as from adhesives bonding those layers. In a third important group, the contaminants, examples of such contaminants frequently detected in migration and extraction studies of plastic food contact materials, include the solvent toluene and antimony (a catalyst) [223457]. Recycling can introduce additional NIAS into plastic due to contamination during use (e.g., chemicals from previously-packaged food), disposal (e.g., metals and persistent organic pollutants adsorbed from the environment), and/or recycling (e.g., monomers formed during melting) [150]. Typical recycling-related NIAS in plastic include oligomers, bisphenols, phthalates, and other additives in recycled plastics [148]. One example study of recycled polyethylene terephthalate bottles included detection of acetaldehyde, formaldehyde, 2-methyl-1,3-dioxolane, polyethylene terephthalate oligomers, toluene, xylenes, and cyclopentanone [151]. Migration of NIAS into food depends on concentration of the NIAS in the material, volatility, molecular weight, vapor pressure, hydrophilicity and lipophilicity, food contact surface area, time of contact, temperature, and nature of the food or food simulant (such as lipid content) [148149223]. In many cases, toxicity of these substances and their transformation products is unknown [456]. Transformation products may be more toxic than their progenitors [459]. Because NIAS migrating from food packaging plastics include a complex array of predicted and unpredicted substances, toxicological assessments based on classical chemical-by-chemical approaches fail to capture the potential hazards of these materials [149456]. The inescapable conclusion is that we have insufficient understanding of the chemical and toxicological profiles of common consumer plastic materials, including food packaging, to evaluate consumer safety in use and environmental safety in waste [460]. ## Environmental degradation of plastic Degradation of plastic in landfills releases polymer breakdown products [398], and different plastics produce different degradation products, which can include aldehydes and ketones from PE; hydrochloric acid from PVC; pentanes from PP; oligomers of styrene, ethyl benzene, phenol, and benzoic acid from PS; and acetaldehyde, ethylene, benzene, and biphenyl from PET with concentrations in the 0.1–7 mg/L range [398]. These pollutants can be released into air [470] or water [398471]. Metal(oid)s, plastic additives, and constitutional monomers present in plastic waste can also be released into leachate [398]. For example, BPA concentrations in leachates from municipal waste disposal sites in tropical Asia ranged from sub µg/L to mg/L [472]. For this reason, landfill leachate is considered to be heavily polluted water that requires specialized physical, biological, and chemical treatment [473]. Rainfall on landfill sites, especially on landfills without impermeable bottom liners or protective top cover layers [474], results in dissolution of organic and inorganic pollutants into leachates, which then either seep into the soil and contaminate underground water systems [400] or enter runoffs and contaminate surface waters such as rivers [475476]. High levels of PBDEs have been reported, for example, in groundwater near open dumpsites [477]. ## Abiotic pathways Plastic polymers can be degraded in the environment by abiotic pathways such as physical fragmentation or chemically through photooxidation, hydrolytic cleavage, or thermal oxidation (discussed in Section 3). Degradation processes are complex and include depolymerization, chemical modification, and changes in physical properties (e.g., strength, surface strength, integrity). In an ideal scenario, they result in complete mineralization to CO2 and H2O. For large macroplastics, the most appropriate definition of degradation is overall loss of mass, whereas surface ablation is an important mechanism for small plastic pieces, especially in the marine environment [240]. Degradation rates vary with polymer type as well as the waste material’s physical properties, including size and shape, and polarity, as well as the environmental milieu (air, terrestrial, aquatic, and landfill) in which the waste resides and the milieu’s physical parameters, such as heat, light, temperature, oxygen levels, and pressure [240]. The complexity of degradation processes, as well as the different techniques used to measure it, are reflected in widely varying estimates of plastic degradation rates, with some media reports claiming that some plastic does not degrade at all [240]. Using surface degradation rates (μm per year), average half-lives of plastic items have been estimated at between 2.3 and 2,500 years, or more. Values of greater than 2,500 years were given for studies where no degradation had been detected within the study period. Mean degradation rates for single-use PET water bottles are reported to be over 2,500 years in landfills and 2.3 years in marine environments with acceleration by UV or heat. Degradation rates for PVC pipes were 5,000 years in landfills and 530 years in marine environments, and the degradation rates for LDPE plastic bags are 4.6 years in landfills and a range of 1.4 to more than 2,500 years in marine environments [240]. ( See Section 3 for details.) ## Biotic pathways Interest in natural enzymatic polymer degradation by actinomycetes, algae, bacteria, and fungi has increased in recent years [478]. Microbial enzymes that degrade PUR, PE, PS, and nylon have been identified [478]. Their mechanisms of action include biodeterioration (surface fragmentation), biofragmentation (extracellular enzymes and free radicals), assimilation (active and passive transportation), and mineralization (end products being CO2, acetic acid, and lipids) (discussed in Section 3). Different factors can either enhance or inhibit biotic degradation. For example, the addition of bacterial nutrient sources such as starch or palmitic acid and oxidation with hydrochloric, sulfuric, or nitric acids can accelerate degradation, whereas plastic additives such as plasticizers and flame retardants can inhibit it [478]. The potential for developing engineered biotic degradation methods via various -omics approaches may be substantial [478]. However, scaling for commercial degradation is a challenge, and the extent to which biotic degradation is a viable mechanism for “cleanup” remains unknown. The impacts of this emerging technology on human health are unknown. ( See Section 3 for details.) ## Greenhouse gases Plastic is a contributor to global climate change [14479480481]. GHGs are emitted at every stage of the plastic life cycle [482483484], starting with direct and fugitive emissions from the extraction and transportation of fossil fuel feedstocks, direct process emissions from energy-intensive chemical reactions in steam crackers, indirect emissions from energy conversion in the energy sector that facilitates polymerization and conversion, and finally emissions associated with end-of-life processes [479482484485486]. Combined, the total global plastic-associated GHG emissions are higher than the total net GHG emissions of most individual countries (see Figure 2.7). **Figure 2.7:** *2020 net greenhouse gas (GHG) emissions from the world’s largest emitters. Gigatons (Gt) of carbon dioxide equivalents (CO2e), including land use, land-use change and forestry, and share of global total (%). In 2015, the annual emissions of CO2 and other greenhouse gas from plastics production was 1.96 Gt of CO2e, or 3.7% of total emissions (Cabernard et al., 2022)[1].Source & Permissions: Adapted from Rhodium Group ClimateDeck. https://rhg.com/research/preliminary-2020-global-greenhouse-gas-emissions-estimates/* By far, the largest proportion of these plastic-associated GHG emissions ($90\%$) is attributed to plastic production [5]. In 2015, the annual emissions of CO2 and other GHG from plastics production amounted to 1.96 Gt of CO2 equivalents (CO2e) [13], with an estimate of $4.5\%$ in 2019 [5]. GHG emissions from plastic are projected to more than double in volume by 2060 [14]. They will consume a significant proportion of the global carbon budget [479] and will undermine the ability of the global community to hold emissions within the targets set in international climate treaties [479483487]. In addition to increasing in absolute volume, CO2 emissions from plastic may be anticipated to account for an increasingly large proportion of global CO2 emissions in future years as emissions from combustion fossil carbon as fuel decreases. Natural gas released to the atmosphere in gas extraction and transmission principally consists of methane and is an additional contributor to plastic’s total carbon footprint. As much as $4\%$ of all natural gas produced by fracking is lost via leakage to the atmosphere from a combination of venting, flaring, and unintentional leaks [96]. These releases appear to have contributed to recent sharp increases in atmospheric methane and are responsible for a third of the total global increase in methane emissions over the past decade. Methane is a potent contributor to global warming, with a heat-trapping potential 30 times greater than that of CO2 over a 100-year span and 85 times greater over a 20-year span [96]. The expansion of shale gas and oil extraction will likely result in further increases in these releases [92]. GHG emissions resulting from the manufacture of plastics differ by polymer type. The largest emissions are associated with the production of polymer fibers used for textiles, followed by PP and LDPE [14]. Growth in plastics production in coal-based economies has resulted in doubling of the carbon footprint associated with plastic manufacture since 1995 [13]. Factors such as the efficiency, configuration, and service life of plant equipment likely influence GHG emissions from plastic production facilities [479]. End-of-life processes account for the remaining $10\%$ of plastic’s GHG emissions, though these vary according to disposal method [14]. Incineration is the disposal method responsible for the largest GHG releases and accounts for approximately $70\%$ of all end-of-life GHG emissions, followed by recycling and sanitary landfilling [5]. Recycling has the potential to reduce plastic’s overall GHG emissions through reduction of primary plastics production via substitution with secondary plastics [14]. ## Conclusion Hazards to human and planetary health occur at every stage of the plastic life cycle—production, use, and disposal. This Section of the Minderoo-Monaco Commission on Plastics and Human Health has summarized these hazards stage by stage. A key finding that emerges from this analysis is that plastics’ hazards to human and environmental health extend far beyond the visible and now well-recognized hazards of beach litter and marine microplastic. They include the hazards of the many toxic chemicals incorporated into plastics, as well as MNPs, and also the contributions of plastic production to greenhouse gas emissions and global climate change. In this Section of the report, we have built a framework and structure that we will follow in subsequent Sections, and especially in Section 4, where we will examine plastics’ impacts on human health at each stage of its life cycle. Plastic causes disease, disability, and premature death at every stage of its long and complex life cycle—from extraction of the coal, oil, and gas that are its main feedstocks; to transport, manufacture, refining, use, recycling, and combustion; and finally to reuse, recycling, and disposal into the environment. In this section of the Minderoo-Monaco Commission on Plastics and Human Health, we have summarized current information on the nature and magnitude of these hazards and also identified gaps in knowledge where additional research is needed to characterize and quantify plastics’ risks to human health. At every stage of the plastic life cycle, infants in the womb and children are the populations at highest risk. Measures taken to protect the health of children and other highly vulnerable groups, such as workers, residents of “fenceline” communities, and Indigenous populations, are ethically and morally well justified and have the added benefit of protecting the health of entire populations. We have quantified some of the externalized costs associated with plastics production and use—costs that are not borne by the industries responsible for them, but instead are imposed on governments, businesses, and persons throughout the world, without compensation. The particular externalities on which we have focused are the health-related global impacts of pollution of ambient air and workplace air due to plastics production, and the impacts of endocrine disruptors and neurotoxicants associated with plastics use in the US. The value of these externalities is summarized in Table 5.6. The total cost of the health impacts of plastics production in 2015 that we are able to quantify is $250 billion (2015 Int$), more than the GDP of New Zealand or Finland in that year [1370]. **Table 5.6** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | | --- | --- | --- | | | DEATHS, CASES, IQ POINTS LOST OR CO2 EMISSIONS | VALUE (IN 2015 PPP BILLION) | | Upstream Health Effects | Upstream Health Effects | Upstream Health Effects | | Global PM deaths in 2015 | 159,491 Deaths | 211.8 | | Global Occupational deaths in 2015 | 31,857 Deaths | 38.9 | | Sub-total | 191,348 Deaths | 250.7 | | Health Effects of Plastics Use in the USA | Health Effects of Plastics Use in the USA | Health Effects of Plastics Use in the USA | | Lost IQ Points (PBDE) | 9,818,493 Points | 145.8 | | Cases of Intellectual Disability (PBDE) | 43268 | 56.5 | | Cases of CHD (BPA) | 1540000 | 165.9 | | Cases of Stroke (BPA) | 60738 | 62.4 | | Deaths (DEHP) | 90762 | 490.0 | | Sub-total | – | 920.6 | | Social Cost of Carbon – Plastics Production | Social Cost of Carbon – Plastics Production | Social Cost of Carbon – Plastics Production | | Global CO2e emitted during plastics production | 1.96 Gt CO2e | 341.0 | The annual emissions of CO2 and other GHGs from plastics production, amounting to 1.96 Gt of CO2e in 2015, are more than the combined CO2 emissions of Brazil and Indonesia [1397]. Using the US EPA’s estimates of the SCC, we estimate that the annual cost of these emissions is $341 billion (2015 Int$). To examine the health impacts of plastics use, we examined the disease burden and economic costs associated with EDCs and neurotoxic plastic additive chemicals, PBDE, BPA, and DEHP in the US. We limited this analysis to the USA because estimates of serum and urinary concentrations of these endocrine disruptors are available for the general US population and not available in most other countries. We estimate that the total cost of the health impacts due to these endocrine disruptors in the US in a single year is over $920 billion (2015 Int$). Estimates suggest that over $90\%$ of exposure to these substances comes from plastics [1398]. What are the solutions to these externalized costs? Reducing local air pollution from plastics production, which relies heavily on the combustion of coal and oil, will require either a transition to clean, renewable energy or the use of pollution control devices to limit emissions of PM, NOx, and SO2. Reducing CO2 emissions will require moving from fossil fuel combustion to renewable sources of energy though various means, including economic incentives such as carbon trading rights, or through internalizing pollution costs through penalties and pollution taxes. If these policies were implemented, they would help reduce pollutant emissions from plastics production and from other manufacturing sectors. The incorporation of hazardous chemicals into plastic production is part of the larger problem of inadequate regulation of chemicals in consumer goods. Differences in the regulation of chemicals in the EU vs. the US are illustrated by the differential impacts of PBDEs on intellectual disability and lost IQ points in Table 5.5. The losses in IQ points and associated lifetime productivity losses are much lower in the EU, where PBDEs have been regulated for many years, than in the US, where the impact of these endocrine disruptors remains substantial. Although we cannot produce similar estimates in other countries, due to lack of data on exposure of the general population, results from the US suggest that the externalized disease burden and economic costs associated with plastic use require much closer examination. Adequately addressing SEJ with respect to plastics pollution will require honest consideration of the following: However, such change will invoke predictable resistance by those benefiting from the status quo, including powerful individuals, governments, and corporations, especially the integrated multinational fossil carbon corporations that both produce coal, oil, and gas and manufacture plastics and plastic additives. Groups with acquired privileges and power—whether individuals, organizations, or multinational corporations—will resist critical examination of current plastics production and use [461482]. Thus, governments and society will need to give extra attention to SEJ in the designing of durable and equitable strategies. Critically, action is necessary at all levels. A case study from Indonesia provides examples of how government, business, and other actors can all participate, and it provides some hope for future action—once SEJ inequities are addressed (Case Study 6.2). The Minderoo-Monaco Commission on Plastics and Human Health finds that plastics are both a boon to humanity and a stealth threat to human and planetary health. Plastics convey enormous benefits, but current linear patterns of plastic production, use, and disposal with little attention paid to sustainable design or safe materials and a near absence of recovery, reuse, and recycling are responsible for grave harms to health, widespread environmental damage, great economic costs, and deep societal injustices. These harms are rapidly worsening. While there remain gaps in knowledge about plastics’ harms and uncertainties about their full magnitude, the evidence available today demonstrates unequivocally that these impacts are great and that they will increase in severity in the absence of urgent and effective intervention at global scale. Manufacture and use of essential plastics may continue. But reckless increases in plastic production, and especially increases in the manufacture of an ever-increasing array of unnecessary single-use plastic products, need to be curbed. Global intervention against the plastic crisis is needed now, because the costs of failure to act will be immense. ## Distribution of Plastics in the Ocean For nearly as long as plastics have been made, they have escaped into nature, including the marine environment. However, the rate of escape has increased markedly since the start of widespread mass production in the 1950s. Marine life has encountered plastic debris in the ocean at least as early as the mid-1960s when global plastic production was less than $2\%$ of what it is today (See Section 2) [488]. Much of the scientific literature on plastics has reported on contamination of the marine environment—from shorelines and coastal environments to the open ocean, from the sea surface to the seafloor, from the tropics to the poles, and in association with increasing numbers of marine species. The most comprehensive datasets on the abundance and distribution of plastics in the marine environment come from beach cleanups [539] and sampling the ocean surface [540541]. Ocean Conservancy has sponsored an annual International Coastal Cleanup for more than 30 years to remove trash from beaches worldwide and document the most commonly found items while simultaneously engaging and educating local communities about the problem. Scientific measurements of plastics in the ocean itself have mostly been conducted at the sea surface by towing nets designed to sample plankton (with mesh size typically ~0.2 mm or larger). The first reports of floating plastic particles were from plankton surveys in which this anthropogenic material was incidentally collected along with the plankton of interest [1490]. Larger, highly visible floating plastic debris—macroplastic—has been quantified by visual surveys from ships or aircraft since the 1980s [542543544]. These very resource-intensive surveys are few, however. While remote sensing technologies offer promise to identify and quantify large floating plastic debris, such approaches are still under development [274545546]. Consequently, data on the number or mass abundance of floating plastics larger than centimeters in size are scarce. For decades, scientists have encountered and measured large debris composed of plastics and other materials on the seafloor, from continental shelves to the deep sea [547]. As with floating MPs, data on these materials were initially collected by scientists trawling or imaging the seafloor for other scientific purposes (e.g., [548549550]). Given the substantial challenges of accessing the deep ocean, it is perhaps surprising that there are more scientific publications documenting large debris on the seafloor than that floating at the sea surface. Much of the seafloor literature reports ALDFG [551], such as traps and nets (e.g., [552553]), or debris made of denser materials and/or consumer products that sank close to presumed coastline sources (e.g., [554555]). As public and scientific interest in MPs has grown, new methods have been employed to measure ever-smaller particles throughout the marine environment, including on beaches [556557], in deep ocean sediments [558], deeper in the ocean water column [559560561], and even in sea ice [562563]. Chemical characterization (fourier-transform infrared spectroscopy or Raman spectroscopy) is critical to confirming the identity of the smallest particles that can be isolated from environmental samples, including biota. Currently, there are no reliable methods to identify and quantify NP particles and sources in environmental samples, although there is some evidence of NP generation from laboratory weathering exposures [564] and colloidal material isolated from seawater [565]. It is likely that NP particles behave differently in the environment than larger particles of the same polymer composition [566], and much work remains to understand their abundance, distribution, and potential risks [527528529]. The scientific literature includes numerous reports of marine life, ranging from phytoplankton and zooplankton to large marine animals, interacting with plastic debris, especially through entanglement and ingestion (reviewed [505567568]; see section “Plastics in Aquatic Food Webs and Seafood” below for details). The diversity of plastics in terms of their particle size and morphology, their polymer and chemical (additive and sorbed) composition [820522], as well as a lack of clarity on relevant environmental exposure rates, makes risk assessment for individual categories of plastic debris a complex and challenging task. It is now clear that MPs contaminate every part of the marine environment globally. While variable in their regional concentrations, they are abundant and widespread in their global distribution. The level of plastic contamination in any particular location will be influenced not only by proximity to inputs or sources of plastics to that location but also by their transport in the atmosphere and ocean. This transboundary transport can render some geographies, such as remote islands [569570] and polar regions [571572], particularly vulnerable to the accumulation of plastic debris that may have originated thousands of kilometers away. The ocean may also be a source of MPs to coastal regions, through transport of sea spray from breaking waves [439532573]. Some have argued that discharges of synthetic chemicals, including plastics, onto land and into the ocean have reached a critical threshold that meets the criteria for planetary boundary threats [16574], potentially with negative impacts on global health. ## Sources Although constrained by a lack of direct measurements of fluxes of plastics (of any size) to the ocean, it is presumed that most marine plastic debris is released to the environment on land [497], where most plastics are used, and then transported by water (e.g., rivers, streams, waves, and tides on beaches) or air (wind) to the ocean [23]. The first global estimate of the input of plastics from municipal solid waste generated on land to the ocean—4.8–12.7 megatons in 2010 [24]—was computed using country-level estimates of plastic waste generation and waste management practices reported by the World Bank [575]. These data were used to model the amount of plastic waste generated by populations in coastal regions that was not properly captured and contained in waste management systems and, therefore, available to enter the environment, including the ocean. This provided a first estimate of the scale of the global problem and resulted in a strong recommendation to further refine the estimate with more robust national data on plastic municipal solid waste generation and its management and environmental measurements of plastic waste. The same basic modeling framework was subsequently adopted to estimate plastic waste input from land to inland waters that ultimately drain to the ocean [407408], with subsequent refinements added to model the likelihood of river transport [409] taking into account the complexity of flow in watersheds [576]. Additional studies have used a similar framework to estimate the amount of plastic waste entering aquatic ecosystems globally (19–23 megatons in 2016 [577]; 9–14 megatons in 2016 [57]). In addition to the broad assumptions built into these simple models, they did not always consider known activities that influence plastic waste handling and fate, such as burying, burning, illegal or unregulated dumping; unregulated or illegal discharges; informal waste collection (e.g., by waste pickers); and the international trade of waste. Further, until recently, few field data were available to anchor estimates of environmental fluxes of plastics, such as from land into rivers and rivers to the ocean. Although data are increasing, the methodologies used vary substantially according to study, the size of debris measured (large plastics vs. MPs), and the environmental matrix sampled (e.g., riverbank, water surface or sediments; river mouth or estuary). Estimating fluxes of plastic debris from beaches or shorelines to the ocean, including estuaries, is even more complex because, in contrast to rivers, fluxes may be erosional or depositional due to local bathymetry and coastline geometry, and they also vary with ocean tides. For these and other reasons, there is extremely high spatio-temporal variability in plastic flux measurements, and much work remains to constrain estimates of total plastic flux from land to ocean (Box 3.1). Many additional sources of plastics to the ocean are not captured in estimates of leaked mismanaged waste from land. A presumably very large source is ALDFG and aquaculture gear, which has yet to be robustly estimated, even to an order of magnitude [551]. Immense amounts of debris of all material types are input to the ocean due to catastrophic events such as tsunamis, hurricanes, or floods [515578579]. Plastic debris is also lost during shipping, commercial, recreational, and other maritime activities [19497499500]. MPs have additional sources and pathways to the marine environment. Sources include generation from large debris [580581582], release as a consequence of wear or abrasion of products while they are in use, such as tires [583584585586] or clothing [587], and the direct release of small pieces of plastic used in applications such as cosmetics [496]. MPs from these sources may be carried to the ocean in wastewater [588] and stormwater outflows [589], as well as by atmospheric transport [531532]. MPs can directly enter the ocean during use as industrial scrubbers or as paint from vessels or structures, for example, and plastic resin pellets, flakes, and powder (the “raw materials” of plastic products) can be lost during transport. ## Box 3.1 How Much Plastic in the Marine Environment Originates From Single-Use Products? While it is possible to confirm the polymer identity of most microplastics, it is generally not possible to identify the object from which microplastic fragments were generated. However, most of the floating plastics collected in plankton nets are polyethylene and polypropylene [590], which are high-production polymers largely used in packaging and other single-use applications [591]. Item counts from decades of beach cleaning and beach surveys have consistently been dominated by categories of single-use items, including packaging, food and beverage service ware, and cigarette butts [23]. In 2017, for the first time, Ocean Conservancy’s Top Ten List of items most frequently collected in the International Coastal Cleanup were all composed of plastics, displacing non-plastic items such as paper bags, glass bottles, and aluminium cans [592]. ## Transport The three-dimensional hydrodynamic processes that transport debris bidirectionally between shorelines and the sea are complex and largely determined by local characteristics such as the shape of the shoreline, seafloor bathymetry, and local wind and sea conditions [515]. Once in the ocean, floating plastics are transported by the surface flow resulting from the interaction of processes including large-scale current systems that form ocean gyres [593], tides [594], surface waves [595], and vertical motion associated with convection [596], ocean turbulence [597598], and frontal dynamics [599]. For larger debris that protrudes above the sea surface, surface winds also contribute to the direction and speed of transport (windage effect). The highest concentrations of floating MPs collected in plankton nets in the open ocean have been measured in subtropical ocean gyres, consistent with theoretical ocean circulation models [600601602]. Very high surface concentrations have also been measured in the Mediterranean Sea’s semi-enclosed and highly populated basin [603604605], where surface flow is unidirectional into the basin. Albeit in lower concentrations, floating MPs have been detected throughout the open ocean, including in polar regions [571572] and other regions remote from human populations. For small MPs (<100 um), atmospheric transport also carries particles over long distances where they may be deposited in the ocean far from land sources such as roads, agricultural fields, and population centers [531532]. MPs also may be transported from the ocean back to land via coastal sea spray [573]. Even in the high accumulation zones of the subtropical ocean gyres, large spatiotemporal variability in surface plastic concentration is typically observed [606607]. Variations in sampling methodology and conditions when sampling (e.g., wind speed, sea state) can explain some of this variability [608]. Still, the complex interaction of multiple physical processes such as those listed above and the diverse physical characteristics of the debris itself currently renders the prediction of “hotspots” of accumulation at small scales (tens of kilometers or less) extremely challenging. There is clear evidence of the transport of plastic debris below the sea surface, but the factors involved are less well understood for all but the densest items, which sink relatively quickly close to their point of entry. Neutrally buoyant or slowly sinking particles could be carried long distances by typically slower-moving, mainly horizontal, deep ocean currents. Even for initially buoyant items, deep sea sediments are the presumed final sink for MPs and other debris in the ocean [560609]. Laboratory studies and modelling studies have demonstrated the sinking of particles (even initially buoyant plastics) due to biotic processes such as biofilm formation [610611612], aggregation in marine snow (aggregate particles comprising organic material, detritus, fecal pellets and more, which continuously fall from upper regions of the oceans to the depths) [613], and fecal pellet formation after ingestion [614]. While MPs have been found in the deep water column [615] and in seafloor sediments [558], direct measurement of vertical MPs fluxes in marine snow has only recently been achieved [547560]. In addition to biologically mediated processes, such as biofilm growth on plastic debris and incorporation of MPs into fecal pellets, there can also be direct biological transport of plastics by birds, fish, and other marine life encountering debris by ingestion or entanglement. Biological transport can occur between ocean and land, as in the case of seabirds foraging at sea to feed their chicks on land [616], or within the ocean when animals swim or migrate at one depth or across a range of depths. For example, mesopelagic fish that feed near the sea surface at night and migrate to depth during the day could transport MPs vertically through ingestion and egestion [617618]. Not yet well understood, the relative effect of biological transport compared to physical transport may be quantitatively small but still biologically or ecologically important. ## Mass Balance: The Case of the “Missing Plastics” The concept of “missing” plastics in the ocean was first considered by Richard Thompson and co-authors in their 2004 article, “Lost at Sea: Where Is All the Plastic?” [ 2]. They reported an increase in the abundance of microscopic plastic particles in plankton samples collected since the 1960s in the northern North Atlantic, whereas previous ocean and beach samplings had not shown an increase in large plastic debris that might have been expected during a time when plastic production had greatly accelerated, and considering plastics’ persistence in the environment. They suggested that this increase in particles they dubbed “microplastics” (likely from the fragmentation of larger items) might explain the difference. Since then, the concept of “missing plastics” has evolved to describe a quantitative mismatch between estimates of plastic mass input to the ocean (with presumed negligible outputs) and the estimated standing stock of plastics in major marine compartments or reservoirs [19]. This mass balance exercise is analogous to carbon budgeting carried out in the 1990s to understand the fate of anthropogenic CO2 released into the atmosphere [619]. The goal of an environmental mass balance analysis for plastics should not be to make precise estimations, because standing stock estimates cannot capture the complexity of time-dependent processes, including the fluxes between reservoirs and the physical and chemical transformations that occur within them. Instead, the mass balance approach is most useful to identify the major knowledge gaps in understanding the flow of plastics from sources to marine sinks. This knowledge is critical to inform other goals, such as exposure and risk assessment or hotspot identification for prevention or mitigation activities. When Jambeck et al. [ 24] made the first estimate of plastic waste entering the ocean from land, they compared their estimate (4.8–12.7 megatons in a single year) to the only available standing stock estimates of plastic debris in the ocean, which reported the mass of plastics floating at the sea surface, mainly in sizes collected by plankton nets [600601602]. The estimated annual flux into the ocean was 10 to 1,000 times larger than the estimated mass of floating plastics. Further, because the input estimate represented only one of the presumed largest sources of plastic input to the marine environment, others of which include fishing and aquaculture gear and losses due to catastrophic events, it was likely an underestimate. Later environmental assessments continued to demonstrate a large quantitative mismatch with the annual input estimated by Jambeck et al. [ 24], even when paired with numerical ocean models to better estimate surface plastic concentrations in unsampled regions (Table 3.1). **Table 3.1** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | | --- | --- | --- | --- | --- | | ESTIMATED MASS | MEASUREMENT METHOD | DEBRIS SIZE | REGION(# MEASUREMENTS) | STUDY | | 1,100 tons | Plankton net(0.335 mm mesh) | Microplastic (nominal) | Western North Atlantic Ocean (6,136) | (Law et al., 2010) [600] | | 21,290 tons | Plankton net(0.335 mm mesh) | Microplastic (nominal) | Eastern North and South Pacific Ocean (2,529) | (Law et al., 2014) [607] | | 6,350–31,750 tons | Plankton nets(0.2 mm–1 mm mesh) | Microplastic (nominal) | Global (3,070) | (Cózar et al., 2014) [601] | | 66,140 tons | Plankton net(0.33 mm mesh) | 0.33–200 mm | Global (680) | (Eriksen et al., 2014) [602] | | 202,800 tons | Visual survey transects | >200 mm | Global (891) | (Eriksen et al., 2014) [602] | | 93,000–236,000 tons | Plankton nets(0.15–3 mm mesh) | Microplastic (nominal) | Global (11,854) | (van Sebille et al., 2015) [1561] | | 82,000–578,000 tons | Plankton nets & filtered continuous seawater intake(0.1–0.3 mm mesh) | Microplastic (nominal) | Global (8,209) | (Isobe et al., 2021) [541] | Although one should not expect standing stock estimates from the sea surface alone, especially from a limited range of debris size, to equal an annual input estimate, the magnitude of the mismatch revealed a major knowledge gap in understanding plastic flows in the marine environment. Proposed explanations for this mismatch include: Although the evidence demonstrates the contamination of the marine environment by plastic debris at global scales, knowledge about the sources, accumulation zones, and sinks of plastic debris remains incomplete. This primarily stems from a gross under-sampling of the marine environment to date, a challenge difficult to overcome for multiple reasons: Facing these challenges head-on, new technologies, such as remote sensing of the sea surface from aircraft or satellites and automated particle identification techniques, are actively being developed to better understand the global distribution of large and small plastic marine debris. Advances in knowledge of the baseline distribution will be necessary to inform exposure and risk assessment of this global contamination and to assess the efficacy of mitigation activities. ## Fate of Plastics in the Ocean Once in the marine environment, plastics experience abiotic and biotic processes that transform their physical and chemical properties (Figure 3.1). These processes may include photochemical degradation from UV irradiation, hydrolysis, mechanical breakdown by wave action and interaction with sediments (abrasion), and biodegradation, principally by microbes. These processes can synergize or antagonize with one another to transform and degrade plastic. For most plastics, these are surface processes; thus, geometries with greater surface area to volume ratio enhance degradation [240]. A plastic product’s structure, properties, and composition (polymer and additives) [526], its geometry (e.g., particle, fiber, sheet) [636], and the local environmental conditions it encounters (e.g., sea surface vs. seafloor) together control the extent to which any one process contributes toward its degradation [240637638]. The integration of these processes defines a plastic product’s eventual lifetime in the marine environment. **Figure 3.1:** *Fate of plastic debris in the environment. Diagram illustrating many of the possible pathways over the lifecycle of plastic litter on its journey from land to sea. Plastic debris enters the ocean through both aquatic (rivers, accidental escape at sea) and land-based sources (littering, escape from municipal waste management such as wastewater treatment plants (WWTPs)). Depending on the density of the plastic material, plastic items will remain afloat for a given part of their lifecycle or, as they become weighted down by biofouling, will begin to sink into the water column, ultimately to the ocean bottom. Changes in biofouling with depth may lead to depth oscillations (not shown) before particles end up on the seafloor (Kooi et al., 2017; Rummel et al., 2017; Royer et al., 2021). Mechanical, photochemical and biological forces break down plastic debris into microplastics and nanoplastics that subsequently become incorporated into the marine food web. Organisms such as filter feeders may further concentrate these smaller particles, given their capacity to filter large volumes of water. Microorganisms begin to attach, colonizing plastic in the water within hours, and can include potentially harmful microorganisms, such as disease-causing pathogens. PET, polyethylene terephthalate; PVC, polyvinyl chloride. Figure reproduced from (Amaral-Zettler, Zettler and Mincer, 2020) [641], with permission from Springer Nature and the Copyright Clearance Center. The reproduced figure is not part of the open access license governing the current paper.* It is apparent that, for most conventional plastics, the sum of these various processes does not result in mineralization of plastics to CO2 and water at a rate that could provide any meaningful mitigation of the increasing input of plastics to the environment [23639640]. Nevertheless, the environmental transformations of plastic are relevant for understanding their potential impacts on marine biota. ## Abiotic Processes For most plastics, abiotic processes dominate their degradation, particularly for polyolefins with carbon-carbon backbones (e.g., PE, PP, PS) [637]. For example, PS is largely resistant to microbial degradation because breaking the aromatic backbone is energetically unfavorable, but absorption of sunlight results in photochemical oxidation, producing CO2 and dissolved organic carbon [629]. Abiotic processes are categorized by their mode of action as either chemical (photochemical, hydrolytic) or mechanical (abrasion) transformations [642]. ## Photochemical Photochemical processes (i.e., sunlight) transform plastics at the molecular and microstructural levels, manifesting as macroscopic changes. Under appropriate conditions, sunlight can degrade plastics much faster than previously thought, reducing estimated lifetimes from thousands of years to tens of years [629643]. Photochemical degradation is the process by which light, principally UV radiation, initiates a series of chemical reactions in the plastics that lead to chain scission and crosslinking of the polymer matrix and the addition of new functional groups into the polymer backbone [225630]. The rapid absorption of UV radiation by plastics limits photochemical degradation to the first 50 to 100 µm into the material, making photochemical degradation a surface-level transformation [644645]. Photodegraded plastic is often identifiable in the field by its yellowed and cracked appearance [646647648]. Photochemical degradation depends on irradiation conditions (environmental dependencies) and the plastic type and microstructure (material dependencies). The amount of irradiation a plastic receives depends on its global location (latitude), local environment (on the beach, ocean surface, or position in the water column), and water clarity (e.g., the presence of other light scattering and absorbing materials) [240637]. For photochemical processes to occur, the plastic must have unsaturated chromophoric groups that absorb UV radiation [225]. Several of the major plastics (e.g., PE, PP, PVC) do not have UV absorbing groups in their backbone (unlike PS, PET); however, internal (e.g., catalyst residues, unsaturated content, carbonyls) and external impurities (e.g., trace metals, sorbents, additives) can absorb UV [225]. Generally, photochemical degradation is a three-step process: i. initiation, ii. propagation, and iii. termination [649650]. Initiation occurs when UV radiation is absorbed by the plastic and a radical is formed by breaking a C-H bond. The radical propagates in step two by reacting with oxygen to form a peroxy radical, which can become autocatalytic. The process terminates in step three when two radicals combine to yield inert products that result in chain-scission, branching, crosslinking, and incorporation of UV absorbing oxygenated and unsaturated functional groups. Specific polymer susceptibilities and photochemical reaction mechanisms have been reviewed [225630637650]. Photochemical degradation results in several interrelated physical and chemical changes that span the molecular and microstructural levels of the material, which are relevant to plastics’ interaction with biological systems [650]. At the molecular level, chain-scission reduces the molecular weight of the polymer matrix, and oxidation introduces carbonyls and hydroxyls into the polymer matrix [225637650]. A decrease in molecular weight, in turn, reduces mechanical properties [651]. Oxidation makes the surface more hydrophilic [652]. Microstructurally, most plastics are semi-crystalline materials. Amorphous regions are more susceptible to degradation processes [650653]. This leads to an increase in surface crystallinity. As a result, a mechanical mismatch between the more crystalline surface layer and the less crystalline underlying bulk polymer generates internal stress [654]. To relieve the stress, the surface of the polymer fractures, leading to increased surface roughness and thus increased surface area [655656]. The collective changes in molecular and microstructural properties contribute to increased brittleness making the material more susceptible to fragmentation [651657658]. Along with physicochemical changes, photochemical degradation results in the release of additives [637], the formation of water-soluble products [526629643659660661662], the liberation of CO2 [526629], and, importantly, the shedding of NPs [636663664665666667]. Because of these changes in plastic properties, most plastics are modified with additives to impede photodegradation. Additives can include antioxidants (e.g., phenolics and phosphites), hindered-amine light stabilizers, their combination, and benzotriazole-type UV stabilizers (BUVSs), among others [649668669]. In contrast, prooxidants and photocatalysts can be incorporated to enhance the photochemical degradation of plastics [670]. Gaining traction is the use of titanium dioxide, a common white pigment and photocatalyst, which has been shown to accelerate photochemical degradation [526542671672673674675676677]. ## Thermal Degradation and Combustion Thermal degradation and combustion of plastics are processes not often associated with marine plastics; however, recent work on ship fires and pyroplastics (see Box 3.2) has revealed that plastics altered by elevated temperature and burning enter the ocean [678679]. *In* general, thermal degradation results in the same chemical and physical property changes as photochemical degradation [637680]. The difference is at the initiation step; instead of UV light, it is thermal energy that leads to radical formation [637]. Combustion of plastics during incineration and open burning of waste results in air pollutants, ash, and charred remnants (Section 2) [362]. The remnants of burning (pyroplastics) have only recently been documented on beaches globally because of their camouflaged appearance, resembling rocks or other natural marine debris [678679681]. Owing to their recent discovery, little is known about how the transformations from burning impact plastics’ fate in the marine environment. During the recent M/V X-Press Pearl ship fire and plastic spill off the western coast of Sri Lanka, both unburnt plastic pellets and burnt plastics were released [678]. Though unfortunate, this spill provided an opportunity to study pyroplastics in the ocean. Recent evidence has shown that burnt plastics from this spill can be of diverse shape, size, and color (often darker shades), have increased brittleness, and contain soot and toxic, combustion-derived contaminants (e.g., PAHs and heavy metals) [237678682683684685686]. ## Box 3.2 New Forms of Plastic in the Anthropocene. Abiotic and biotic degradation processes in the marine environment transform plastics physically and chemically into new forms, such as plastiglomerates, plasticrusts, pyroplastics, anthropoquinas, plastitars, and petroplastics, which have properties unlike those of the material that originally entered the ocean and are unique to the marine environment [681707708709]. Plastiglomerates: multi-component composites of melted plastic, sand or sediment, volcanic rock, and organic matter, likely formed by the burning of plastic [710711712713714]. Plasticrusts: plastic encrusted on intertidal rocks with other debris often embedded within the plastic, likely formed following the wave-induced collision of plastic with rock outcrops [713715716717]. Pyroplastics: burned or melted plastic with geogenic appearance, neutral color, and increased brittleness formed from the burning of plastic [678679684713714716718]. Anthropoquinas: anthropogenic material contained within sedimentary rock including plastic formed by the deposition of sediment containing plastic [719]. Plastitars: tar encrusted on intertidal rocks with embedded plastic and other debris, similar to plasticrusts [708720]. Petroplastics: agglomerates of oil and microplastics [709721]. These new forms of plastic debris have only recently been recognized in the environment, and there is uncertainty about their prevalence, distribution, fate, and overall impact on marine systems [681]. ## Hydrolytic Degradation Hydrolytic degradation is the process by which water reacts to cleave chemical bonds in plastics, principally ester and amide bonds. Some polymers susceptible to hydrolysis include polyesters (e.g., PET, PLA, polycaprolactone, PC), PAs (e.g., nylons), and PUR [637687688689]. The rate and extent of hydrolysis largely depend on temperature, salinity, and pH (environmental dependencies) and molecular weight and crystallinity (material dependencies) [688689690]. Hydrolytic degradation results in many of the same physical and chemical changes as photochemical degradation: reduction in molecular weight, new functional groups, increase in crystallinity, reduction in mechanical properties, increased brittleness, increased surface roughness, and increased hydrophilicity [658687690691692]. Similarly, along with changes in the plastic, other products are liberated, including additives, water-soluble products, and fragmented NPs [637690]. The time required for hydrolytic degradation under marine conditions varies depending on the type of polymer [240689690]. ## Mechanical Degradation Waves and other motion act to fragment plastics by abrasion with hard particulate (e.g., sand) and surfaces (e.g., rocks). Mechanical degradation works synergistically with chemical degradation processes. As chemical degradation embrittles plastic, mechanical degradation can more readily fragment the material. ## Biotic Processes Biotic degradation processes encompass plastic’s fragmentation, assimilation, and transformation by organisms (micro and macro). It is believed that abiotic processes can prime plastics for biotic degradation by reducing the length of the plastic’s polymer molecules and introducing labile functional groups [637638677693694]. Conversely, organisms or biological macromolecules can impede abiotic processes (e.g., biofilms screening UV light [695]). Biotic processes include chemical (enzymatic, hydrolytic, oxidative) and mechanical degradation. There is an important distinction between the biodegradability and the biodegradation of plastics. Biodegradability (sometimes referred to as inherent biodegradability) indicates the potential for a plastic to be degraded by some biological means [639]. ISO and ASTM standards for biodegradability can test for this intrinsic property [696]. In contrast, biodegradation is a system property describing microbial transformation of the plastic, the rate of which is dependent on the surrounding environmental conditions [639]. Thus, biodegradability can be thought of as a prerequisite for biodegradation to occur, but it says nothing about whether it occurs in a given environment, or the rate of transformation, which are the subject of biodegradation. ## Biodegradation (Enzymatic, Hydrolytic, Oxidative) Microbes are the organisms primarily responsible for the biodegradation of plastics in the marine environment. Microbes can degrade recalcitrant materials (e.g., lignin and oil); however, the rates of degradation are highly variable [697]. Microbes can degrade plastics hydrolytically, oxidatively, and enzymatically [637638640698699700]. Together these processes lead to chain scission, oxidation, and new end groups [637638]. Like their abiotic counterparts, biodegradation processes depend on temperature and pH (environmental dependencies) and on polymer type and crystallinity (material dependencies). However, biodegradation has additional requirements such as nutrient levels and the extent to which other, more labile carbon sources are available (environmental dependencies), the molecular weight of the plastic and size of the plastic particle (material dependencies), and whether microbes in the local community can degrade the given plastic (biological dependency) [640698699]. Degradation can occur extracellularly (e.g., through the secretion of enzymes) [640698699], or, if polymer chains or plastic particles are small enough to cross cellular membranes, they can be degraded intracellularly by cellular machinery and intracellular conditions [640698699]. *In* general, biodegradation requires chemical bonds and microstructural features that are cleavable, modifiable, and accessible by enzymes; chemical bonds that are hydrolytically cleavable; or chemical bonds that are susceptible to reactive oxygen species [637638640698699700]. A recent metagenomic study examined microbial plastic-degrading enzymes in global soil and marine environments (67 locations at 3 depths across 8 oceans) [701]. Filtered environmental-hits were compared to those in the gut microbiome where plastic degrading species have not been reported. Overall, >30,000 enzyme-hits were detected. Approximately one quarter of organisms in the environmental microbiomes examined had a range of polymer-degrading enzyme-hits including enzymes capable of degrading PET, PVA, PUR, PET, and PE; phthalate-degrading enzyme-hits were also identified. Higher enzyme-hits were observed in more heavily plastic-polluted ocean locations where stratification with depth was also observed, reflecting depth-related variations in taxonomic richness. Proportions of enzyme-hits varied with location, with PUR-degrading enzymes being found in the ocean and not the soil, and twice as much PET-degrading enzymes in the soil as in the ocean [701]. Biodegradation results in many of the same physical and chemical changes as abiotic processes: reduction in molecular weight, new functional groups, increase in crystallinity, reduction in mechanical properties, increased brittleness, increased surface roughness, and increased hydrophilicity. Along with changes in the plastic, microbes may degrade organic additives, metabolize water-soluble products, and respire plastic carbon to CO2 [640698699]. It is important to emphasize that although microbial biodegradation of plastic in the ocean can occur, the rates are very low [23639640641701]. ## Mechanical Fragmentation by Marine Biota Biological fragmentation of plastic results from the biting or chewing of plastics by organisms and the excretion or regurgitation of smaller secondary particles and fragments. This process has been observed for MPs ingested and excreted by a variety of organisms, including amphipods, arctic krill, rotifers, and seabirds [568702703704]. ## Lifetimes Plastic products entering the marine environment are heterogeneous, variable, and diverse in their properties. For instance, a bottle made of PET and another made of PP may functionally both serve the same purpose, but they will have different fates in the environment. Likewise, a clear PS cup and a Styrofoam (expanded PS) cup, though both are cups and chemically PS, can differ in molecular and microstructural properties and surface area to volume ratio resulting in different fates. From the descriptions above, it is evident that different plastic degradation processes can lead to similar transformation outcomes. Differences in environmental persistence and fate arise because each plastic product has different susceptibilities to any given degradation process, based on the product’s geometry, the local environment, and the type of plastic. Collectively, degradation processes create heterogeneity, variability, and diversity in the properties of the plastic and its degradation products. This translates into a continuum of lifetimes for a given plastic product, not a single number as often quoted by infographics [240471]. Similarly, this heterogeneity can lead to the misrepresentation of plastic lifetimes without appropriate reporting of the material and environmental conditions. For instance, this is evidenced by the discrepancy between the lifetimes of PLA products in composting and marine conditions; under composting conditions, PLA readily degrades in months, while under marine conditions, it degrades over many years [689705706]. The coupling of geometry, local environment, and plastic type makes determining plastics’ lifetimes in the environment challenging and onerous. Chamas et al. [ 240] proposed the “specific surface degradation rate” (µm/year) as a measure of plastic degradation (Figure 3.2). This term is a material-environment coupled property, which was calculated for the major plastics (PE, PET, PVC, PP, PS) and some biodegradable plastics (PLA, polyhydroxyalkanoate). The specific surface degradation rate can range from 0 to 1,400 µm/year depending on the combination of plastic type and environment [240]. Applying this parameter to real products, it was estimated that a PET single-use bottle and HDPE bottle in the marine environment with accelerants (e.g., sunlight, heat, microbes, additives) could have average half-lives of 2.3 years and 26 years, respectively [240]. The surface specific degradation rate only considers gross mass loss from a product but not the fate of the degradation products, which can include water-soluble products and MNPs. **Figure 3.2:** *Degradation rates for various plastics. “Vertical columns represent different environmental conditions (L, landfill/compost/soil; M, marine; B, biological; S, sunlight) and plastic types (represented by their resin identification codes). Plastics type 7, “others”, corresponds to various nominally biodegradable plastics. The range and average value for plastic types 1–6 are shown on the right as lines and squares, respectively, as well as for biodegradable “others”. Data points representing degradation rates that were unmeasurably slow are shown on the x-axis. Gray columns represent combinations for which no data were found.” PET, polyethylene terephthalate; HDPE, high-density polyethylene; PVC, polyvinyl chloride; LDPE, low-density polyethylene; PP, polypropylene; PS, polystyrene. Figure caption and figure reprinted with permission from (Chamas et al., 2020) [240] (CC BY 4.0).* Determining rates and lifetimes of plastic degradation in the marine environment remains an active area of research frustrated by the tremendous complexity and diversity of plastic products and marine environmental conditions, as well as the relatively easy transport of lightweight plastics by wind and water. Despite having a relatively well-developed theoretical understanding of plastic degradation mechanisms, our grasp of how those mechanisms manifest together as degradation rates and lifetimes in the marine environment remains nascent. Much more would be required to fully understand the long-term fate of plastics in the ocean; it is clear, however, that rates of input to the environment considerably exceed rates of degradation, leading to environmental accumulation [639]. ## Plastics in Marine Biota Plastic-biota interactions in aquatic environments include entanglement, colonization, and ingestion. Entanglement in large plastic litter such as ALDFG is well known to impact megafauna such as marine mammals, turtles, seabirds, and some fishes; such interactions are usually lethal [505567722]. Ingestion of plastics is less visible but is widespread, and thus may be a more insidious threat to marine life. Plastic pieces and particles have been found in hundreds of marine and freshwater species from diverse environments around the world. The aquatic animal species ingesting plastic include a variety of invertebrates, fish, seabirds, turtles, and marine mammals [2567568608723724725726727728729730731732733]. Santos et al. [ 734] documented 1,565 species globally for which ingestion of plastics had been reported; most of these ($82\%$) were marine species, which may, at least in part, reflect the greater effort devoted to investigating plastics in marine systems as opposed to other environments. The probability and extent of ingestion of plastic particles is influenced by multiple factors [725729], including encounter frequency (related to particle concentrations), the size of the particles relative to the normal food types [735736737], shape, color [523738], and the presence of eco-coronas (layers of biological macromolecules adsorbed to the surface of MPs [507739] or infochemicals (biochemicals that mediate communication among organisms) from biofilms [740741742743744]. Ingestion may be indiscriminate, but some suspension feeders are able to reject plastics in favor of their natural food [735745746747]. Our current understanding regarding the ingestion of plastic particles by marine species has some limitations. One is that much of the data are for MPs that are in the GI tract, which does not represent internal concentrations in tissues [727748]. Moreover, the MPs may often be excreted without adversely affecting the animal. The location of MPs mainly in GI tract has implications for the behavior of plastics in aquatic food webs (see below). Another limitation is that most studies of plastics in biota in the field have measured only the larger sized MPs (>150 µm) [748]. Much less is known about the ingestion of NPs and small MPs [749750], which includes the sizes most likely to undergo translocation from the GI tract into tissues [751]. The dearth of information about the ingestion of these smaller plastic particles by marine species is a major knowledge gap. ## Plastics as Substrate: The Plastisphere Marine microbes (bacteria, archaea, and single-celled eukaryotes such as diatoms and dinoflagellates) form the base of the marine food web, and so their interactions with plastic particles may help drive the trophic dynamics of MPs. Any hard surface immersed in an aquatic environment will readily acquire a biofilm—an “eco-corona”. Initial formation of an “eco-corona” [507739] composed of biological macromolecules is driven by physico-chemical interactions with the surrounding water, and colonization by microorganisms typically follows within hours [612]. Over longer periods in freshwater and marine environments, succession to an assemblage of epifaunal microorganisms then follows over the following weeks to months. Plastics are no exception and readily become colonized by micro and macro biota. Microbial communities on plastic particles in the ocean have been named the Plastisphere [752]. A variety of microbial taxa have been found associated with plastics, including bacteria, archaea, diatoms, dinoflagellates, and fungi (reviewed [641753]). These taxa include pathogens (e.g., certain Vibrio species) and species of diatoms, dinoflagellates, and cyanobacteria that are associated with harmful algal blooms [605641752753754]. The colonized MPs may thus serve as vectors to transport these organisms, as well as the genes they carry (e.g., antibiotic resistance genes [755]), over long distances, with potential impacts on humans and aquaculture. In some instances, this “fouling” assemblage can overwhelm originally buoyant plastic items causing them to sink [610612756]. In addition, infochemicals such as dimethyl sulfide released from the microbial assemblage can attract planktivorous organisms and seabirds that normally use these chemical cues to indicate the location of food [740741742743744757]. The presence of an eco-corona or microbial biofilm may promote consumption and internalization of plastic particles into cells [758]. ## Bioaccumulation, Trophic Transfer, and Biomagnification of Plastic Particles The behavior of chemicals in marine food webs is governed in part by processes such as bioaccumulation (increased concentration of chemical during the lifespan of an organism), trophic transfer (movement of the chemical from one trophic level to another), and biomagnification (increased concentration with increasing trophic level). Plastic particles may potentially undergo the same processes (reviewed [748]), but there is some confusion and misunderstanding in the MP literature about these processes and the extent to which they occur for MPs and NPs [727748759]. It is important to distinguish between plastic particles and the chemicals (additives, sorbed pollutants) that they may carry. Many of the plastic-associated chemicals also have sources other than plastics and are well known to bioaccumulate and biomagnify in marine food webs. We focus here on the plastic particles themselves, for which these processes are not as well understood. Bioaccumulation of MNPs has been documented in some, but not all, laboratory and field studies involving aquatic species [748]. Bioaccumulation typically requires translocation from the GI tract into tissues. Smaller MNPs (particles <150 µm) have a greater potential for translocation [751]. There is evidence from field studies that ingested MPs can translocate into tissues [760761762], but in some cases the sizes of particles detected are larger than those seen to translocate experimentally or that can be explained by known mechanisms of transcellular and paracellular uptake [762]. Such conflicting results between laboratory and field studies (e.g., [762763764765766]) indicate that this is an area in need of further research. Although there is some evidence suggesting that NPs may persist in tissues, depuration of MPs and NPs often occurs within a few days [745746764767768]. For example, a study using radiolabeled NPs found size-dependent, rapid uptake into scallop tissues, followed by rapid depuration (Figure 3.3) [764]. A recent study suggested that less than $1\%$ of NPs move from a fish GI tract to other tissues, and the accumulation there did not persist [769770]. *In* general, the persistence of MPs and NPs in tissues—which is required for bioaccumulation and biomagnification to occur—is not yet fully understood. Insight into these processes may come from cross-disciplinary exchange with biomaterial scientists who study polymeric wear particles from orthopedic implants or nanoparticles for drug delivery. The migration and clearance of polymeric particles from orthopedic implants to other tissues in humans have been described [771772773774775776]. **Figure 3.3:** *Uptake and depuration of 24 nm polystyrene nanoparticles. (a) Tissue distributions shown by Quantitative Whole Body Autoradiography (QWBA) in Pecten maximus after 6 h uptake with (b) quantification of radioactivity levels measured in tissues (left axis; Bq g–1, S/Nnorm; right axis ng g–1), (c) Tissue distributions shown by QWBA in Pecten maximus after 8 days of depuration, with (d) quantification after 8 days of depuration (left axis Bq g–1, S/Nnorm; right axis, ng g–1). Each bar represents the mean value measured in 3–6 different sections of a given individual. No bar = radioactivity < LOD. nPS24 and nP250 are spherical polystyrene nanoparticles with sizes of approximately 24 ± 13 and 248 ± 21 nm respectively. HP, Hepatopancreas; Gi, Gills; Go, Gonad; I, Intestine; K, Kidney; M, Muscle; A, Anus. Figure caption and figure reprinted from (Al-Sid-Cheikh et al., 2018) [764] (CC BY 4.0).* There is abundant evidence for trophic transfer of plastic particles from prey to consumer [748]. However, trophic transfer does not necessarily lead to bioaccumulation or biomagnification in the consumer, at least for MPs, because much of the MP load is retained in the GI tract and rapidly excreted. In contrast to many POPs, for which biomagnification is well known, both experimental and field-based studies have indicated that MPs do not biomagnify [729748759762777778]. In some cases, there is evidence of trophic dilution—a reduction in concentrations of MPs in higher trophic levels [760778779]. An important caveat is that because we do not yet have reliable methods for measuring NPs in animals, it remains possible that their behavior could be more similar to that of POPs. ## Plastics in Seafood As noted above, a wide range of organisms are known to ingest MPs; this includes a variety of seafood species, including fish, shellfish (mollusks and crustaceans), seaweeds, echinoderms, and cephalopods. Examination of seafood species prior to and at the point sale clearly indicates the potential for human exposure to MPs through consumption of MP-containing seafood [780781782]. The discussion that follows has been informed by several recent reviews that have compiled data on concentrations of MPs in seafood and provided estimates of human exposure [762783784785786787788789790]. MP particles are typically found at greatest concentration in the GI tract and with most species this is removed before consumption; exceptions include smaller fish (e.g., anchovies, sardines) and most bivalves, which are usually eaten whole. Several studies have reported higher MP concentrations in inedible vs. edible fractions of seafood [791792793]. However, the “inedible” fractions may be further processed for use as animal feed (including fish meal), providing additional opportunities for transfer to humans later in the food supply chain [790794]. Studies examining the edible fraction of fish have often found MPs in the species examined [785795796797798799]. The environment may also influence the MP burden in animals; e.g., benthic species have been found to contain more MP than pelagic species [796800]. Most current knowledge about plastics in seafood is for the larger-sized MPs. For the smaller NPs the potential for transfer to humans is greater because of the greater potential of these particles to translocate from the GI tract to the circulatory system and onward to tissues throughout the organism, as noted earlier. Despite the prominence of seafood in discussions of human exposure to MPs and NPs, there is need for caution in the interpretation of these findings, for several reasons. For MPs, the quantities found are typically quite low (1 or 2 pieces per individual fish or shellfish) [801]. In the context of total human exposure to MPs, seafood is only one of several sources, others of which (e.g., drinking water, inhaled air) may equal or exceed the exposure from seafood (Table 3.2) [783788802]. Moreover, in addition to the MPs in the tissues of seafood species themselves, there are several points in the pathway from harvest to consumption at which MPs may be introduced [790]. For example, MPs are produced through the use of cutting boards and cookware [803804805] and the quantity of MPs settling onto the food from the atmosphere during preparation and consumption is likely to exceed that in the seafood beingconsumed [802]. **Table 3.2** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Unnamed: 5 | Unnamed: 6 | Unnamed: 7 | Unnamed: 8 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | (DANOPOULOS ET AL., 2020) [786] | (CATARINO ET AL., 2018) [802] | (VAN CAUWENBERGHE AND JANSSEN, 2014) [767] | (HANTORO ET AL., 2019) [784] | (BARBOZA ET AL., 2020) [785] | (COX ET AL., 2019) [783] | (DOMENECH AND MARCOS, 2021) [788] | (ZHANG ET AL., 2020) [1562] | | mollusks | 2,067(0–27,825) | 123–4,620 | 1,800–11,000 | 500–32,750 | | | | | | Crustaceans | 206–17,716 | | | 322–19,511 | | | | | | Shellfish | | | | | | | | 0–1.3 × 104 | | Fish | 31–8,323 | | | 25–32,375 | 112–3,078 | | | | | Total seafood | | | | | | 17448 | 22000 | | | Fruits & vegetables | | | | | | | 19 × 109 | | | Bottled water | | | | | | 15156 | 2.61 × 103–3.96 × 1010 | | | Tap water | | | | | | 3358 | | 0–4.7 × 103 | | Total water | | | | | | | | 0–2.8 × 1010 | | Salt | | | | | | 86 | 261 | 0–7.3 × 104 | | Alcohol | | | | | | 294 | 26 | | | Honey | | | | | | 73 | | | | Sugar | | | | | | 8319 | | | | Dust ingestion | | | | | | | | 100–1.9 × 104 | | Air (inhaled) | | | | | | 46501 | 2160 | | | Indoor air | | | | | | | | 1.9 × 103–1 × 105 | | Outdoor air | | | | | | | | 0–3 × 107 | Overall, the evidence available does not implicate consumption of seafood as a major pathway for transfer of MP from the environment to humans. Two caveats are 1) the potential contribution from seafood may be greater for populations for whom seafood is a higher proportion of their diet and 2) without corrective action (Section 7), the relative contribution from seafood could increase in the future. The relative importance of seafood as a vector for transfer of NPs is less clear because particles of this size are extremely difficult to quantify in environmental samples. Filling the NP knowledge void should be a researchpriority. ## Plastic as a Vector of Chemicals Plastics in marine environments may contain, or have accumulated, hundreds of chemicals classified as “additives” or “adsorbed” chemicals, respectively (see also Section 2) [843472806807]. Thus, plastic litter in marine environments is a cocktail containing chemicals added during manufacture as well as those adsorbed from polluted water, including phthalates, PBDEs, BPA, PCBs, styrenes, PAHs, and metal(loid)s such as lead and nickel [808]. Some of these chemicals can be bioaccumulated upon their ingestion or uptake by marine organisms. The mechanisms and characteristics of MP-mediated bioaccumulation are different between adsorbed chemicals and additives. A variety of anthropogenic chemicals have been found in association with plastics in the marine environment (Table 3.3). The importance of MPs as vectors of adsorbed chemicals to marine biota has been reviewed in several papers [21509809810811812813]. Mainly through equilibrium model calculations, the results suggest that the role of MPs as vectors of adsorbed chemicals is often minor compared to the accumulation of those chemicals from natural prey, especially in aquatic environments already highly contaminated by chemicals such as PCBs. **Table 3.3** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | | --- | --- | --- | | NAME OF CHEMICAL(S) | ABBREVIATIONS (WHERE APPLICABLE) | REFERENCES* | | Polychlorinated biphenyls | PCBs | (Ogata et al., 2009 [1316]; Taniguchi et al., 2016 [1563]; Camacho et al., 2019 [1564]; Yamashita et al., 2019 [819]; Arias et al., 2023 [1565]) | | Polybrominated diphenyl ethers** | PBDEs | (Taniguchi et al., 2016 [1563]; Camacho et al., 2019 [1564]; Pozo et al., 2020 [1566]; Ohgaki et al., 2021 [1567]) | | Polycyclic aromatic hydrocarbons | PAHs | (Taniguchi et al., 2016 [1563]; Yeo et al., 2017 [1568]; Camacho et al., 2019 [1564]; Arias et al., 2023 [1565]) | | Dichlorodipheyltrichloroethane and its metabolites | DDTs | (Ogata et al., 2009 [1316]; Taniguchi et al., 2016 [1563]; Camacho et al., 2019 [1564]; Pozo et al., 2020 [1566]; Arias et al., 2023 [1565]) | | Hexachlorocyclohexanes | HCHs | (Ogata et al., 2009 [1316]; Taniguchi et al., 2016 [1563]; Camacho et al., 2019 [1564]; Pozo et al., 2020 [1566]; Arias et al., 2023 [1565]) | | Cyclodiene pesticides*** | | (Taniguchi et al., 2016 [1563]; Camacho et al., 2019 [1564]) | | Chlordanes | | (Taniguchi et al., 2016 [1563]) | | Mirex | | (Taniguchi et al., 2016 [1563]; Camacho et al., 2019 [1564]) | | Hexachlorobenzene | HCB | (Taniguchi et al., 2016 [1563]; Camacho et al., 2019 [1564]; Pozo et al., 2020 [1566]) | | Pentachlorobenzene | PeCB | (Pozo et al., 2020 [1566]) | | Benzotriazole-type UV stabilizers**** | BUVSs | (Camacho et al., 2019 [1564]; Karlsson et al., 2021 [1569]) | | Organophosphorus flame retardants**** | OPFRs | (Camacho et al., 2019 [1564]) | | Triclosan**** | | H. Takada, personal communication | | Sterols | | H. Takada, personal communication | | Hopanes | | (Alidoust et al., 2021 [1570]) | Conversely, in remote areas with higher abundances of plastic, the accumulation of POPs or other sorbed chemicals from that plastic could be greater than accumulation from natural prey. Plastic from contaminated areas also may carry POPs to otherwise cleaner environments. MPs of mm-size are transported offshore by Stokes drift [814] and are sometimes found in remote islands [815], while µm-size MPs are settled and deposited in bottom sediments [816] through biofouling [610] and some other biological processes. Sorption of hydrophobic organic compounds to mm-size MP particles is not only a surface process but also involves diffusion of chemicals within the polymer matrix; for this reason, sorption/desorption takes time [817]. For example, it would take more than one year for a PCB congener with a Kow ~7 (e.g., PCB-153) to reach equilibrium [818]. This is the reason that high concentrations of PCBs are sporadically found in mm-size MP (resin pellets) on the beaches of remote islands [819]. This means that mm-size MPs can transport sorbed chemicals from anthropogenically impacted areas to remote ecosystems. The other group of chemicals found in plastics in marine environments are the additives, by-products of plastic production, and unreacted and polymer-degradation-derived monomers that are originally contained in plastic products [8]. Additives include plasticizers, antioxidants, UV stabilizers, flame retardants, and many others [8]. Many of these are EDCs and neurotoxicants [221]. ( See also Section 4.) They have a wide range of hydrophobicity. Hydrophilic additives such as BPA can be directly transferred to humans (“Direct Exposure” in Figure 3.4) or leached into seawater. However, most of the additives are hydrophobic and they are retained in plastics and MPs in seawater [820]. **Figure 3.4:** *Conceptual model of microplastic-mediated transfer of additives and POPs to marine animals and humans. Conceptual model of microplastic-mediated transfer of additives and persistent organic pollutants (POPs) persistent to marine animals and humans. Plastic additives and legacy POPs accumulate in the ocean through leaching from waste virgin and recycled plastic; similarly, microplastics, which also contain additives, accumulate as a result of fragmentation. Plastic additives and legacy POPs can be adsorbed to microplastics. Humans can be directly exposed to plastic additives and legacy POPs from use of plastic products as well as indirectly via the food chain. Human exposure to microplastics via the food chain may also occur.Credit: Shige Takada and Manuel Brunner (co-authors).* Plastic ingestion can be considered as a kind of internal exposure to additives (“Direct Accumulation” in Figure 3.4). A key question is the extent to which the additives from ingested plastics transfer to the tissue of marine organisms. Because additives are compounded in the polymer matrix, they may not readily leach out of plastics. Hydrophobic additives are retained in MPs with minimal leaching into seawater [820]. When they are ingested by marine organisms, however, oily components in digestive fluid can facilitate the leaching of the hydrophobic additives [821822]. Furthermore, as plastics become smaller through fragmentation, leaching of hydrophobic additives from plastics can be facilitated [823824]. Laboratory experiments [820] and a semi-field exposure experiment [825] demonstrated that plastic additives can be transferred from ingested plastics to the tissue of organisms. Occurrence of the process was demonstrated through the detection of BUVSs and BFRs in globally collected seabird preen gland oil samples [826]. Furthermore, the bioaccumulation of plastic additives in the tissue of marine organisms and their connection to plastic ingestion has been demonstrated for whales [827], oysters [828], seabirds [829], and lanternfish [830]. Collectively, these observations demonstrate that plastics can be an important vector of additives to some marine organisms. One impediment to interpreting these studies is the sporadic occurrence of specific additives (e.g., UV stabilizers and BFRs) in plastics [831832] and marine organisms [821829833]. To connect detection of specific additives to ingestion of plastics, many observations are therefore necessary. Another difficulty is distinguishing plastic-mediated exposure to additives from prey-mediated exposure, as prey may already be contaminated with additives via indirect bioconcentration. In such scenarios (“Indirect concentration” in Figure 3.4), additives may leach out into seawater and organisms may concentrate the leached additives from seawater. The importance of this process has been demonstrated by exposure experiments involving moderately hydrophobic additives such as hexabromocyclodecanes (log Kow ~ 6) [834] and BUVSs (log Kow ~ 7) [820]. Bioaccumulation of moderately hydrophobic additives such as nonylphenol and BUVSs has been observed in a wide range of coastal organisms [835836837838]. Highly hydrophobic additives such as decabrominated diphenyl ether (BDE-209, log Kow ~ 12) may be directly accumulated from ingested plastics into tissues of organisms (“Direct accumulation” in Figure 3.4). However, excretion of MPs would reduce retention time and thus limit the release of such additives in the digestive tract. Consistent with this, in one study the proportion of bioaccumulated BDE209 was only a few percent of the total BDE-209 in the ingested plastics [825]. However, excreted MPs could possibly be re-ingested, until ultimately becoming unavailable through burial or other processes. Various plastic additives have been detected in human adipose [839], blood [840], and urine [841842843844]. This may occur via direct transfer through daily-use plastics, i.e., leaching of additives from food and beverage containers (“Direct exposure” in Figure 3.4) as well as through indirect exposure via consumption of seafood (Figure 3.4). Overall, there are several processes through which aquatic organisms and humans may be exposed to plastic-associated chemicals. A more quantitative assessment of these processes will provide a better understanding of their relative roles in mediating the transfer of these chemicals from plastics to marine organisms and humans. ## Impacts on Marine Life There are now thousands of publications on the impacts of large and small plastic particles on marine organisms. As with all other aspects of plastic pollution, impacts on marine animal health depend on the size, shape, and chemistry of the particles [41]. The most dramatic and visible effects occur with macroplastics; for large items of debris, the evidence for impacts is extensive and well documented [505]. These involve findings of entanglement by plastics and plastics in stomach contents of birds and large animals, often reported in the media. There is more evidence of negative effects from entanglement than from ingestion. This likely reflects the direct visibility of the damage associated with entanglement by macroplastics, which are external to the organism, as opposed to effects of ingestion, which require necropsy [505]. Kuhn and van Franeker [728] reviewed more than 700 studies of entanglement and ingestion in marine megafauna, seabirds, marine mammals, turtles, fish, and invertebrates, finding 914 species affected. The frequency and abundance of the visible ingested particles varied among taxa and was greater in some than in others; however, Wilcox et al. [ 724] modelled the data records and estimate that by 2050, $99\%$ of seabird species will be affected by plastics. There is a substantial body of evidence indicating effects of plastics on marine and aquatic organisms. The majority of the observations come from laboratory experiments, but there also is evidence of harm to organisms in the natural environment. Negative effects of all sizes of plastics have been demonstrated at most levels of biological organization from the scale of macromolecules to cells, tissues, organs, organisms, and assemblages of multiple species [505506507510845]. Although effects are not always found, in a recent meta-analysis of 139 lab and field studies, Bucci et al. [ 41] found that $59\%$ of tested effects at all levels of biological organization were detected. Whether effects occur at a population level is uncertain, in part because it is extremely difficult to isolate the effects of plastics, compared to the myriad other stressors in the environment that may have population-level impacts. Nevertheless, there are studies that predict the potential for impacts at the population level, especially for species that are already threatened or endangered [846]. ## Particle Effects and Mechanisms of Micro and Nanoplastics In contrast to the effects of macroplastics, the effects of MPs and NPs are less obvious, and the mechanisms of action are not yet well understood. There is a rapidly growing number of papers on the effects attributed to MP or NP particles themselves observed in experimental exposures, although a limitation is that many of the experimental studies used concentrations of plastics greater than those reported to occur in the environment [41847848849]. Reports on impacts address diverse taxa, marine mammals to fish, crustaceans, and mollusks. Jacob et al. [ 850] reviewed studies on fish; Pisani et al. [ 851] examined studies on crustaceans; and Han et al. [ 852] did a meta-analysis of endpoints in aquatic vertebrates and invertebrates. The reported impacts include effects on survival, behavior, metabolism, and reproduction. Kogel et al. [ 514] reviewed effects observed in experimental studies with diverse taxa, considering multiple aspects of the particles as used, and noted that not all experiments are comparable. The reader is referred to these reviews for further details on the impacts. The text below considers types of health effects of the particles, and some of the mechanisms involved. Effects of associated chemicals are considered in the following subsection. Given the broad molecular similarity across the biological spectrum, fundamental mechanisms inferred from experiments and observations with one species or with cells in culture are likely to apply to other species. Thus, while the focus of this section is on effects on marine species, the mechanisms considered here are drawn from studies in mammals and human cells as well as from marine or aquatic species. Some are discussed in more detail in the section on human health effects (Section 4). MP and NP particle effects on cells, tissues and organisms depend to a large degree on the particle size, shape, and chemistry, as well as charge and concentration [41504]. There still are uncertainties regarding the size ranges of particles classified as “microplastics” or “nanoplastics,” although there are suggestions for standardizing the terminology [521737]. Commonly accepted sizes in the maximum dimension for MP particles are 1 µm to 5 mm, and for NPs as 1 to 100 nm or 1,000 nm [521]. Different size particles may differ in the physical effects related to the shape and structure of the particle, and how this can affect cells and cellular function. There have been several reviews of NP effects in vitro, in diverse human cells in culture, and in vivo, in diverse model organisms [518527528529750853854855]. Among commonly observed effects is enhanced production of reactive oxygen species (ROS) and an oxidative stress response, which can impair cellular and organ function in marine and other species [518856]. Disruption of mitochondrial function and a role in ROS generation are seen in various human and other species cells in culture [857858], The exact mechanisms of action on the mitochondria are not understood, but particles other than plastics are known to affect mitochondria. Mechanical stretching of the membranes could be involved in the effects on cells and on mitochondria [859]. Inflammation, suggesting an immune system reaction to the particles, has been observed in cells and organisms [860]. Consistent with this, changes in activation of genes involved in inflammation, including p38, mitogen-activated protein kinases, and various cytokines, have been observed [861]. Broader consequences of inflammation and oxidative stress can include GI, hepatic, reproductive, and neurotoxic effects, observed in rodents and in zebrafish [862863864]. In addition, inflammatory and other responses in gut and in liver and other tissues have been reported in invertebrates and fish [864865]. The liver increasingly appears to be a target [866867]. Metabolic effects suggest that NP may contribute to obesogenesis, possibly due to action of associated chemicals [868], including chemicals yet to be identified [869]. MPs in the GI tract may impact animal health in several ways, for example through intestinal blockage, triggering satiety in the absence of nutritional value (a maladaptive food choice referred to as an “evolutionary trap” [734]), or by interfering with the function of the animal’s microbiome [870871]. The review by Jacob [850] and several primary studies have found that MPs and NPs cause alteration of gut microbiomes in fish [872], including adults and larval zebrafish [873874], which could affect the function of the gut-brain axis. Evidence of effects on microbiomes is growing. A difference between MPs and NPs related to particle properties is that the larger MPs reportedly either pass through the digestive tract and are excreted or accumulate therein, causing obstruction and potentially a false sense of satiety. NPs and smaller MPs apparently can cross cellular membranes in the gut [770]. NP also have been found to cross cell membranes of cells in culture [875]. Particles transferring from the gut to the blood may then distribute throughout the body, whether of vertebrate or invertebrate [763]. However, there is not yet a clear understanding regarding which particle sizes are more or less likely to pass from the gut into the circulatory system. In studies in marine and aquatic species smaller NPs were found to appear in or affect internal organs. An example in zebrafish showed that 50 nm particles appeared in blood and crossed the blood brain barrier and affected dopamine metabolism, while 100 nm particles did not [876]. It is also possible that some effects could result from the disruption of membranes and membrane function by particles [877]. Thus, the MP and NP particles themselves may elicit effects that compromise growth, behavior, or reproduction and other processes. Moreover, there are suggestions that transgenerational effects of NPs may occur [878879]. Notably, the majority of the studies that address MP impacts experimentally have used virgin plastics at concentrations or doses far in excess of the concentrations in the ocean [41847]. If bioaccumulation of the NP is proven, it could signal that a latency between exposure and effect may occur, as particles and their associated chemicals accumulate in the brain, for example, over time. ## Effects and Mechanisms of Plastic-Associated Chemicals Chemical effects result from the chemical composition of the particle that may include plasticizers or others added in synthesis, or chemicals associated with the particle, adsorbed from the environment. In nature, the effects of plastics will involve both particle and chemical effects. The studies showing effects of MPs or NPs on aquatic species in the laboratory seldom have examined chemical effects at the same time. As discussed above, plastics have been referred to as a “Trojan horse,” carrying the various chemicals associated with plastics, both additives and adsorbed chemicals, into organisms that acquire plastics via dietary or respiratory pathways, whether through gills or lungs [880881882]. Smaller particles that cross membranes may carry chemicals throughout the body. Larger particles that are ingested but do not cross membranes still may desorb or release additives in the gut, which then can be accumulated in the body. As discussed above, the additives include plasticizers and stabilizers, with BPA, phthalates such as DEHP, and PBDEs as prominent examples. Adsorbed chemicals (Table 3.3) generally are POPs and other highly or moderately hydrophobic chemicals including PCBs, PCDDs, PAHs, pesticides, and others. There is a huge literature on the environmental occurrence of these compounds and their effects in aquatic and mammalian model organisms, and in animals in the ocean and Great Lakes. There is also extensive epidemiological information linking exposure to effects in humans, reviewed in Landrigan et al. [ 21] and in the Human Health section of this report (Section 4). The effects in animal models (zebrafish, medaka, rodents) exposed to plastic-associated chemicals include neurodevelopmental disorders such as behavioral and cognitive effects (PCBs, PCDDs), immune dysfunction (PCBs, PCDDs, PBDEs), reproductive impairment (BPA, other chemicals), lipid metabolism defects (phthalates), cardiovascular disease (PCBs, PCDDs, PAHs), carcinogenesis (PAHs, PCBs), and population effects (endocrine disruptors, dioxin equivalents). The mechanisms for many of these effects are known to involve cytosolic and nuclear receptors that regulate genes determining hormonal action, growth and reproduction, cell proliferation and others. Critical questions concerning effects in the ocean are whether—and where—there could be exposure to MPs or NPs at levels capable of eliciting effects like those observed experimentally [509847883], and what proportion of chemical exposure can be attributed to plastic sources. Immune system and reproductive effects have been observed in cetaceans, associated with PCBs in the tissues [884885886], developmental effects linked to population level effects have been seen in salmonids in the Great Lakes [887], and reproductive effects observed in birds have been linked to PCBs and pesticides [888]. An explicit link to plastics is seldom known, however, as there are many sources for such chemicals. Nevertheless, as we note above, in the case of additives, plastics may be a major source to the ocean, and to organisms: the bioaccumulation of plastic additives has been connected with plastic ingestion in whales, seabirds, fish, and a crustacean. Data from mesopelagic fishes are providing information regarding exposure to chemicals from plastics. Analysis of a suite of additives and PCBs in myctophid fish from the mesopelagic zone within a gyre indicated that levels of PBDEs—but not other contaminants in these fish—were associated with the prevalence of plastic particles in the water within the gyre in one study [889], but a subsequent study did not show the same result [830]. In another study, the added risk of plastics with sorbed chemicals was negligible compared to that from chemicals sorbed to natural organic matter [890]. Whether passage through the environments from rivers to the ocean inevitably confers a chemical patina that can elicit effects apart from or in addition to the particles themselves is not clear. Moreover, as noted above, plastic-associated chemicals include many that also have other sources, and the contribution of those from plastics is not quantifiable in many cases. Many mechanisms are the same, and so the health effects attributable to the chemicals should reflect the combined exposure to those chemicals from all sources and their transformation products (see Box 3.3), and how particles and chemicals together contribute to impacts. Thus, whether chemicals in the water that sorb to plastic particles present a substantial risk to fish or other consumers is not fully resolved. ## Box 3.3 Microplastics (MPs) and Toxic Chemicals From Tire-Wear. Road transport is a major source of MP leakage to the environment—an estimated 2.7 megatons globally in 2019. Specific sources include tire abrasion, brake wear and eroded road markings [5]. Tire wear particles, in particular, are recognized as an important source of MPs to aquatic environments [310583584586891]. Tires were originally made from natural rubber but are now made from a combination of natural rubber and synthetic polymers to which are added sulfur (1–$4\%$) to increase elasticity, carbon black (22–$40\%$) as a filler, oil to improve wet-grip performance, and a number of other chemicals to augment durability. During use, contact between the tire and the road causes shear and heat with the generation of MP particles [583]. The amount of tire wear-related MPs released varies between countries in relation to traffic volume. Annual estimates range from 1.25–1.80 megatons in the US, 0.76 megatons in China, 0.29 megatons in Brazil, 0.29 megatons in India, 0.24 megatons in Japan, 0.13 megatons in Germany and 0.04–0.08 megatons in the UK [583]. Tire-wear particles contribute to the particulate matter in the air, including the PM2.5 and PM10 fractions that have demonstrated human health effects [892]. Inhaled tire-wear particles caused pulmonary fibrotic injury in mice [893], and toxic effects of tire-wear products have been reported in a human lung cell line [894895]. Multiple chemicals used in tires are toxic and can be released to the environment; these include carbon black, which is possibly carcinogenic [374], toxic metal(loid)s such as cadmium, lead, nickel, and redox-active metals such as copper and zinc, and a variety of organic compounds [583896897]. No published studies have directly examined the human health hazards associated with exposure to tire-wear. In the US, unexplained deaths in Pacific Northwest Coho salmon occurred annually over decades. This acute mortality coincided with storm water runoff in high-density traffic areas as salmon were returning to their freshwater breeding grounds. It was thus termed “urban runoff mortality syndrome.” Yet the causal agent remained unidentified. Recently, a transformation product of a tire rubber additive has been identified as the chemical responsible for this syndrome [585]. N-(1,3-dimethylbutyl)-N′-phenyl-p-phenylenediamine (6PPD) is an antiozonant added to the rubber used to make tires; it is widely used and added at substantial amounts (0.4 to $2\%$)[585]. After leaching from tire wear particles, it can be transformed in the environment to 6PPD-quinone. The quinone is extremely toxic to Coho salmon, which are exposed to it when they seek to reproduce in urban creeks that receive stormwater containing roadway runoff [585]. 6PPD -quinone was detected in storm water and roadway runoff at concentrations ranging from 0.016–2.29 µg/L, concentrations up to 24 times the median lethal concentration for Coho salmon (0.095 µg/L)[585]. Follow-up studies have shown species-specific differences in susceptibility to “urban runoff mortality syndrome” with high vulnerability in Coho salmon, white-spotted char, rainbow trout, and brook trout, but lower vulnerability in chum, other salmon species, zebrafish, arctic char, white sturgeon, and three crustacean species [585898899900901902]. This case provides a compelling example of how additives in plastics can be transformed in the environment to novel toxic compounds and how such toxicity may remain unknown or unexplained for years. It also highlights the need for manufacturers to fully disclose the chemicals in their plastic products. ## Effects in the Environment A recent modeling study suggests that, at present, harmful effects of MPs in the environment are likely to be restricted to locations where their abundance is relatively high and predicts that if environmental accumulation continues at the current rate, there are likely to be widespread ecological effects in the next 50–100 years (Figure 3.5) [903904]. This future accumulation will add to what has been proposed as a global “toxicity debt”—the fragmentation of large plastics to MP and NP, which may exert greater toxicity over time due to the increase in these more toxic particles and the resulting increase in surface area and the potential release of associated chemicals [64905]. **Figure 3.5:** *Global risks of microplastic pollution based on worst case scenario (unacceptable level (PNEC) = 7.99 *103 MP m–3) displayed in a four-panel plot, in which each panel corresponded to a specific year: 1970 (A), 2010 (B), 2050 (C), and 2100 (D). Cell specific (1° by 1°) risk estimates were calculated, and a 3D visualization of the data was generated. The risk estimates were represented in 3D as elevation values. As long as the risk quotient remains lower than the value of 1 (bluish tones), policy makers consider no risk due to MPs. In case that the risk quotient exceeds the value of 1 (reddish tones), there is a risk. Figure reproduced from (Everaert et al., 2020) [904] (CC BY-NC-ND 4.0).* ## Ecosystem-Level Effects The potential for ecosystem-level effects of plastics is likely to vary among ecosystems. It has been suggested that some ecosystems may be more vulnerable [906]. While it is extremely difficult to assess changes in ecosystems such as many estuaries, coral reefs (see below), the Antarctic, and the deep ocean, the presence of plastic particles in, for example, deep sea fish [731907908] and in amphipods from the deepest parts of the ocean [909] indicate that ecosystem-level effects could occur globally. Plastics may adversely impact ocean ecosystems through effects on a range of global processes and interactions affecting the natural flux of chemicals and energy in the environment [507641]. Effects will vary across systems, e.g., comparing estuarine and open ocean systems, and the different species involved in energy transformation and transfer may respond differently to MP and NP. Factors that may influence this include polymer composition, particle size, associated chemicals, species feeding strategies, and others [910]. ## Effects on Phytoplankton and Primary Production The effects of MP and material leached from plastic on phytoplankton have been reviewed [753]. Some studies have reported alterations in photosynthesis and primary productivity (the fixation of CO2 into organic molecules). Particles and especially leachates have adverse effects, although there is limited ability to generalize because of differences in species tested, polymer identity and size, and leached material. Effects more commonly seen include ROS generation and oxidative stress, and transcriptional alterations and effects on the proteome [911912]. Both stimulation and inhibition of primary production have been observed. Some studies report increased primary productivity by organisms in the “plastisphere,” while hetero-aggregation and sinking of particles and cells may negatively affect productivity [912]. The potential link between MP effects on phytoplankton and oceanic carbon sequestration capacity are elaborated below, in Ocean Plastics and Climate. ## Effects on Zooplankton and Energy Transfer Microzooplankton comprise heterotrophic and mixotrophic organisms 20–200 µm in size, which include many protists, such as ciliates, dinoflagellates, and foraminiferans, as well as small metazoans, such as copepods and their developmental stages and some meroplanktonic larvae. We focus here on copepods as a representative and broadly important taxon. Copepods (crustaceans in the subclass Copepoda) are among the most abundant primary consumers on earth, occurring in all waters; they form the base of food chains in the ocean. While Pisani et al. [ 851] reviewed plastics’ effects in crustaceans, copepods deserve special attention given their profound role in energy transfer. They also exemplify the complexities in dissecting effects in the ocean. Some copepod species generate flow fields to bring food particles in toward the mouth and then select particles by size and chemoperception. Inert particles can be rejected or consumed. Studies with various plastic particles indicate that copepods do consume plastic particles. Studies with a common species indicated that plastic beads or fibers alone were usually rejected after “tasting,” regardless of particle size or polymer type [747]. However, when particles were present together with prey diatoms, they were ingested together [913]. This implies that plastic particles may not be ingested by some species in regions where there are low densities of prey. Behavioral studies have suggested that there is a low risk of MP ingestion by planktonic copepods, although some species and other zooplankton do ingest particles alone, without regard to the presence of actual prey [914]. The duration of plastic particles’ time in the ocean may alter their acceptability to copepods. Polystyrene beads kept for several weeks in seawater were ingested by two calanoid copepod species at higher rates than fresh beads alone. The suggestion is that the microbial content of biofilms developing on the particles could make them more palatable [742]. Studies of impacts on copepods have shown varying results. In a laboratory study with the copepod *Acartia tonsa* investigators found that polystyrene microbeads affected growth and survival suggesting a decrease in population size that cumulatively could be substantial [915]. Large effects of PS MP particles were also observed on growth rates of a doliolid, an important gelatinous zooplankton species [916]. Zooplankton fecal pellets are important in carbon flux in the ocean (see below). Effects of plastic on fecal pellet size and composition have been observed. Decreased fecal pellet size implies less energy transfer to depths [915], while at the same time fecal pellets can vector MPs to the depths [614917]. ## Effects on Coral Reefs Coral reefs are among the most important biodiversity hotspots on our planet. Corals are the keystone species of coral reefs, and scleractinian corals secrete calcium carbonate skeletons that are the basis of reef structures. These organisms have adapted to live in nutrient-poor conditions by establishing a symbiosis with a wide range of microorganisms, including dinoflagellates of the Symbiodiniaceae family. Dinoflagellates are essential to coral health as they photosynthesize and transfer most of their photosynthates to the coral host for its own needs. When corals are exposed to stress, they tend to expel their dinoflagellate symbionts (a phenomenon referred to “coral bleaching”), and thus experience starvation. If bleaching continues for too long, corals will die. Coral reefs also are among the most threatened ecosystems in the world and are affected by a range of global (e.g., warming, acidification) and local (e.g., overfishing, pollution) stressors. The combination of stressors potentially leads to greater impacts than a single factor. Currently, pollution from plastics and plastic additives such as biocides, flame-retardants, and plasticizers is a growing concern for the health of corals and other reef organisms [918919]. Indeed, macroplastics and MPs have been consistently found in the water, sediments and organisms of all coral reefs studied, although little is known about the level of NPs [919]. The potential effects of MPs on corals may involve ingestion and direct exposure and the combined action of MPs and associated chemical contaminants, contributing to coral disease, and impacts on coral-Symbiodiniaceae symbiosis, with significant bleaching associated with NPs [920]. Some corals may mistake the plastics spiked with microbial surface biofilms as their natural food source [921]. Once ingested, MPs can induce a false sense of satiety and reduce natural heterotrophic feeding, although this effect depends on the species [921922923]. In addition to active ingestion of MPs by corals, passive adhesion to the coral structure surface may also impact coral health [921922]. Laboratory studies have demonstrated that MP exposure (including active ingestion and passive surface adhesion) can influence the coral energetics, growth, and overall health, with consequences for feeding behavior, photosynthetic performance, energy expenditure, skeletal calcification, and even tissue bleaching and necrosis. Although corals are the main reef builders and have attracted most of the scientific attention, they are not the only reef species affected by plastics. For example, all members of the planktonic and benthic species described above are also present in reefs and are likely affected by plastics. A recent review [919] summarized current knowledge of the effects of plastics on reef organisms. If corals bleach and die, the whole reef ecosystem, and the services it provides to the billion humans living nearby, will disappear. Studies should thus continue assessing the actual risk of MNPs at relevant in situ concentrations and in conjunction with other anthropogenic stressors such as global warming to determine potential synergies these stressors may have. ## Ocean Plastics and Climate Plastics contribute to GHG emissions during the production of and transport of fossil fuels that are plastic feedstocks, during the processing of feedstocks in plastic production, and during degradation and incineration of plastics (see Plastic Life Cycle, Section 2) [480]. The annual volume of plastic-associated GHG emissions is estimated to have been 1.9 Gt of CO2 in 2019 [14]. Plastics in the ocean have further potential impacts on climate through direct and indirect effects on the biogeochemistry of the planet. Direct effects include GHG emissions from plastic particles, primarily PET, as they degrade in the ocean [485], although the contribution from this relative to other sources of GHGs is likely to be very small [924]. ## Effects on Global Carbon Flux As mentioned above, MP particles ingested by zooplankton, including copepods and others, can be eliminated in fecal pellets, which can alter the density of the pellets and thus the sinking rate of pellets and of marine snow that may incorporate the fecal pellets containing plastic particles [925]. The possible consequences include changing the rate of carbon export to the deep sea [614], which could impact the global carbon balance. The sinking marine snow also provides food to organisms in the midwaters and the deep ocean, and plastics in fecal pellets and marine snow can deliver plastics to those consumers [917]. Knowledge and understanding of the magnitude of plastic particles in marine snow in different parts of the ocean and the impact on the carbon export are badly needed. Biofilms on plastic particles could influence the impacts of those particles in the ocean, potentially altering the consumption by zooplankton, conveying pathogens [754], altering the degradation of the particles [695926], and increasing O2 consumption in remineralization, thereby reducing O2 availability for other processes or for efflux into the atmosphere [927]. Biofilms on the particles may affect their buoyancy and distribution [610756], including accumulation in the sea surface microlayer [928929]. The sea surface microlayer of ocean waters is composed of lipids and metabolites from microbes and is a site where heat and gases are exchanged between ocean and atmosphere [930]. Modeling and experiments suggest that plastics may be enriched in this layer, which could affect microbial growth and activity in the surface layer, and potentially affect air-sea gas exchange at the surface [929], which could modify CO2 uptake in the ocean, affecting this major sink for global CO2. MPs in the sea surface microlayer also may be ejected from the surface, with repeated settling and ejection leading to a “grasshopper” effect that can result in long-range ocean transport [931], adding to the surface microlayer and impacts in distant regions. ## Phytoplankton and O2 Generation Effects on phytoplankton that generate O2 via photosynthesis could be another mechanism through which ocean plastic affects climatic conditions [932]. Experimental studies have resulted in some, albeit conflicting, evidence for such impacts on marine phytoplankton. In a study of polystyrene leachate, the photosynthetic activity of four microalgal species was increased [933]. In contrast, materials leached from weathered PVC plastic particles were found to decrease the abundance of photosynthetic cells and to decrease photosynthetic efficiency [934]. However, in that same study populations of some heterotrophic bacteria were stimulated. Negative effects of leachates have been seen on Prochlorococcus, the most abundant photosynthesizer in the ocean [935]. As in impacts on other organisms, plastic effects on phytoplankton tend to depend on size and composition of the particles [911912936]. There are outstanding questions regarding the identity of leachate compounds that affect photosynthetic bacteria including Prochlorococcus [937], and the mechanisms of action on photosynthesizers, which may involve oxidative stress as in animals [911]. Importantly, as with effects on animals, the effects of plastics and/or leachates observed experimentally are obtained with concentrations orders of magnitude greater than the levels known or estimated in the environment [912]. The extent of such effects in the environment is thus uncertain [938]. There is a need to determine the degree to which plastics are affecting photosynthesis in the global ocean. This should include studies to determine how the effects of plastics and other stressors, including those associated with a warming ocean, might act together on primary producers [480]. ## Plastics in Freshwaters The impetus for much research on plastics, and in part for this report, (is the magnitude of the plastics problem in the ocean. However, the potential for ecological impacts of plastics is equally important in freshwaters, including lakes and rivers. As these are common sources of drinking water worldwide, the occurrence of plastics especially MP and NP) in lakes and rivers also potentially poses a more immediate exposure pathway to humans than plastics in the ocean. Searching the literature reveals that studies of MPs in freshwater systems have lagged studies in the ocean. Accordingly, there is need for attention on plastics in freshwaters. MPs are detectable in water and sediments of lakes and rivers globally [939]. The levels in more heavily contaminated freshwaters tend to be higher than those in the ocean, consistent with proximity to terrestrial sources [940]. As expected, the numbers of particles detected vary greatly, by 8 or 9 orders of magnitude, depending on location of sampling, with greater levels in waters near urban areas [941942]. In a study in China, levels were many orders of magnitude higher in water and in sediment near a “dumping river” in a heavily populated area than in a moderately urbanized environment [943]. Several studies have noted that PE, PP, PS, and PET, are common in both water and sediments [944]. Groundwater, which is a major source of drinking water for as much as $\frac{1}{3}$ of the world’s population, is of increasing concern for plastic contamination [945]. In soil ecosystems, groundwater supports nearly all terrestrial plants, from grasslands to forests. MPs and plastics’ chemicals in groundwater thus present potential ecological and human health risks. Plastics can enter groundwater from farm operations, septic discharge, waste disposal sites, road run-off (tire particles), precipitation, and tidal inflow in coastal areas, and concentrations appear to exceed those in marine waters [946]. Reviews have shown multiple types of plastic and plastic particles in groundwater [946], although in many areas, fibers are most prevalent [947]. We might also note that chemicals associated with hydrocarbon extraction, some directed to plastic feedstock, can enter groundwaters [948]. There are suggestions that better analyses are needed [512] to address the many groundwater and plastics research needs [949]. Environmental and experimental studies of plastics in freshwater biota have concentrated mostly on fish, although studies have addressed taxa from microbes and algae to birds [950]. The studies of plastics in wild freshwater fish have examined samples from Europe, Asia, and North and South America. A full appreciation of the levels and extent of contamination is hindered by the small sample sizes in many studies [951]. Most of the studies have examined gut contents for plastics [952], which reveals little about the possibility of transfer to human consumers who eat the flesh. The presence of plastics in the gut also does not convey information about possible effects on the organisms. However, McIlwraith et al. [ 762] were able to detect the presence of a considerable number of particles in flesh and liver of a large proportion of the fish they surveyed. This indicates that some particles do translocate from the gut to the edible portions of freshwater fish and thereby could be transferred to humans. Much of what has been reported above for effects in marine species can be expected to apply as well to freshwater fish. Indeed, in reviewing papers on effects of virgin MPs and NPs on fish, many of those studies included freshwater fish, model species as well as wild species [850]. The studies addressed types, shapes, exposure pathways and examined various life stages. As with marine species, however, most studies on effects in biota have been conducted with MP particle concentrations far greater than the concentrations measured in the environment, as much as 5–6 orders of magnitude higher [953954]. The presence of plastics in drinking water from lakes and rivers has attracted growing concern. The key questions have to do with the levels and nature of plastics in specific drinking water sources, whether water treatment removes particles, or whether treatment adds particles. The levels detected in drinking water range by orders of magnitude (reviewed [512]). An investigation of type, quantity, size, and shape of MPs in drinking waters found that MP levels were low in water from a “high-performance” water treatment plant [955]. Treatment plants differ in performance, yet it appears that MP concentrations generally are lower in treated than in raw waters [956]. Detection of MP and NP contamination of freshwater biota shows similar results and raises similar questions as those in marine systems. That is, there is need for improved technology, and consistency in sampling and measurement, and reporting of the nature of particles. ## Conclusions, Recommendations, and Potential Implications for the Ocean and Human Health Plastics are persistent contaminants that do not readily degrade in the environment. Large plastics and MPs have accumulated in all oceanic and freshwater environments, in lakes, rivers, polar ocean, and the deepest ocean trenches around the world. The impacts of large plastics entangling marine animals are visible, obvious, and disturbing. The impacts of MPs are less easily observed, but studies suggest that smaller MPs can adversely disrupt the physiology of plants, animals, and microbes. NP particles are less well studied in the environment, due to the inability to accurately measure them, but experimentally they have been shown to distribute in organisms to organs including the brain. Because of their small size and potential to transfer across tissues within organisms, NP exposure may pose a higher risk. In some studies, chemical additives found in plastics have been shown to affect heath and reproduction of aquatic organisms. There are evidence gaps, but this document and other recent reviews conclude that there is sufficient evidence to act in order to reduce the rate of environmental accumulation. This conclusion is based partly on the evidence of harm demonstrated in laboratory studies, and projected increases in accumulation of plastics in the environment. For example, recent consideration of the need for binding restrictions by the ECHA concluded that there was already sufficient evidence of harm to stem the release of MPs into the environment. Similarly, the recent agreement to negotiate a legally binding international UN Plastics Treaty demonstrates a clear consensus that current design, use, and disposal of plastics is problematic and needs to change. The need for action to curb the escape of plastics to the environment is based on a need to protect aquatic life, biodiversity, and ecosystems upon which humanity depends for food, livelihood, and well-being. ## Recommendations In the published record of research on MNPs in the ocean, and in human health concerns, it is commonly mentioned that there are almost no data on smaller MNPs in the environment or tissues. This deficiency is largely due to the lack of reliable and cost-effective methods adequate for assessing the presence and abundance of these particles. We strongly recommend a concerted effort and funding to support the development of methods to allow detection of smaller MNPs in environmental media (water, sediment) and in tissues critical to effects (e.g., brain, gonad) and transfer to humans (e.g., fish muscle). Information on the global distribution of plastics is currently fragmentary and depends on labor-intensive analyses. We strongly recommend international efforts to develop sensor technology to measure quantities of plastics as particles or debris, and deployment of such technology on robotic vehicles for ocean sensing of small particles and to broadly use satellite sensing to identify where plastic debris fields are changing. The information obtained may indicate where impacts may be greater, signaling need for intervention. Effects of smaller MNPs in the environment are largely unknown; inferences are based on experimental studies most using doses far exceeding the plastics’ levels in the environment. We urge the support of studies to identify the impacts of small MPs and NPs on organisms in the environment itself, and at environmentally relevant doses in model systems. ## Potential Implications for Human Health The presence of plastics in the ocean and their adverse impacts (both demonstrated and potential) are important for the health of the ocean and the earth system more generally (Figure 3.6). In addition, plastics in the ocean have potential impacts on human health and well-being, as described in detail in the following section (Section 4). The potential links between ocean health and human health include: **Figure 3.6:** *Distribution, fate, and impacts of plastics in the ocean. Plastics enter aquatic environments (marine and freshwater), undergo processes that determine their distribution and fate, and impact organisms and ecosystems in a variety of ways. Mt, Megatons; POPs, persistent organic pollutants. References: [1](Rochman et al., 2019); [2](Lau et al., 2020); [3](Borrelle et al., 2020); [4](Santos, Machovsky-Capuska and Andrades, 2021); [5](Pitt, Aluru and Hahn, 2023); [6](Everaert et al., 2020).Credit: Designed in 2022 by Will Stahl-Timmins.* Without dramatic change in production, especially of non-essential, single-use plastics, the abundance of plastics in the global environment, including the ocean, will increase. Likewise, the potential impacts of plastics on life in the ocean will increase. The prospect that increasing impacts will alter the role of the ocean in global and human health and wellbeing is inescapable. The human health effects, elaborated in the next section, and ocean health effects are inextricably linked. ## Box 4.1 Plastics’ Impacts on Children’s Health. Infants in the womb and young children are two populations at particularly high risk of plastic-related health effects at every stage of the plastic life cycle. Because of the exquisite sensitivity of early development to hazardous chemicals and children’s unique patterns of exposure, plastic-associated exposures are linked to increased risks of prematurity, stillbirth, low birth weight, birth defects of the reproductive organs, neurodevelopmental impairment, impaired lung growth, and childhood cancer. Early-life exposures to plastic-associated chemicals also increase the risk of multiple noncommunicable diseases later in life. The concept that noncommunicable diseases in adult life, such as cardiovascular disease, diabetes, cancer, and dementia, can result from adverse environmental exposures during fetal life or early infancy was elucidated in the early 2000s in the Developmental Origins of Health and Disease (DOHaD) concept [958]. This concept developed from the earlier work of David Barker and his colleagues and is based on long-term epidemiologic studies of adults who had been exposed to adverse environmental influences in utero and in early childhood [959]. The DOHaD construct provides an intellectual framework for conceptualizing the long-term impacts of plastic-associated chemicals on human health. Plastics’ disproportionate impacts on children’s health are seen in communities near coal mines, oil wells, and fracking sites. They are seen in the low-income, largely minority “fenceline” communities adjacent to plastic production facilities. They are seen among children exposed to plastic during its use in homes, schools, and playgrounds. They are seen in children who live adjacent to plastic waste disposal sites in all countries and especially in the low-income countries to which so much of the world’s plastic waste is exported. When policies for the reduction of plastics’ harms to human health are specifically designed to protect children’s health, they protect the health of all members of exposed populations. ## Health impacts of extraction of carbon feedstocks Fossil carbon derived from coal, gas, and oil is the raw material for more than $98\%$ of all plastic and the main feedstock for most of the chemicals (petrochemicals) added to plastic [912]. Extraction of coal, oil, and gas by mining, conventional drilling, and unconventional drilling (hydraulic fracturing, or “fracking”) is associated with multiple harmful impacts on human health. ## Coal mining Coal mining is a physically hazardous occupation with high rates of acute and chronic injury and injury-related death. Coal mining is also responsible for chronic health impairment in miners. Coal dust inhalation can cause coal workers’ pneumoconiosis, silicosis, chronic obstructive airway disease, emphysema, and chronic bronchitis as well as cardiovascular disease [960961962]. Miners are exposed to diesel exhaust, a known cause of cardiovascular disease and lung cancer, from trucks and drilling equipment [94]. Coal mining can also cause disease in nearby communities. Coal dust inhalation by pregnant women in communities near mines increases the risk of acute lower respiratory tract infections in their children [962]. ## Oil and gas extraction Oil and gas extraction are highly hazardous trades, and workers in these occupations are at elevated risk of disease and death caused by fire, explosion, blowouts, and physical injury. Work-related fatality rates in oil and gas development are 2.5 to 7.5 times higher than those in construction and general industry, respectively [94]. Offshore drilling operations are especially dangerous. Oil and gas field workers can be occupationally exposed to silica dust as well as to diesel exhaust from vehicles and drilling equipment [94]. ## Air and water pollution Exposures to multiple toxic chemicals in air and water compound the physical hazards of coal, oil, and gas extraction and lead to increased risks of noncommunicable disease, disability, and premature death in workers as well as in residents of “fenceline” communities [963964965]. ## Air pollution—Ozone Ground-level ozone is formed in the air surrounding gas and oil extraction sites by the photochemical reaction of aerosolized hydrocarbons with NOX [967968]. Ozone is a respiratory irritant, and exposure is especially dangerous for children, the aged, and active adults who spend time outdoors. Ozone exposure can lead to asthma and chronic obstructive pulmonary disease [98]. Atmospheric concentrations of ozone in some gas fields are reported to be as high as those in urban areas. In Utah, the Department of Environmental Quality reported ozone levels in the heavily fracked Uinta Basin at levels $85\%$ higher than US federal health standards in 2010 [969]. Gas field ozone haze can spread widely and has been measured as far as 300 km beyond drilling fields, thus presenting hazards for people in distant communities [967]. ## Air pollution—Particulate matter Airborne PM pollution is extensive in coal, oil, and gas production and arises from multiple sources, including rock and coal dust from mining operations, diesel exhaust emissions from drilling rigs and transport vehicles, hydrocarbon emissions from wells, and flaring of natural gas [78]. Since 1995, airborne PM emissions from plastic across its life cycle have increased by $70\%$, with the greatest increases in the production phase [970971]. PM air pollution causes disease and premature death in exposed workers and “fenceline” community residents. Fine particulates such as PM2.5 are the most dangerous airborne particulates because they are small enough to penetrate deep into the lungs, and in some instances, they can enter the bloodstream [972]. In adults, PM2.5 exposure increases the risk for cardiovascular disease, stroke, chronic obstructive pulmonary disease, lung cancer, and diabetes [973]. In infants and children, it increases risk for premature birth, low birth weight, stillbirth [974], impaired lung development, and asthma [975]. Prematurity and low birth weight are risk factors for cardiovascular disease, kidney disease, hypertension, and diabetes in adult life, while impaired lung growth increases risk in adult life for chronic respiratory disease [976]. Emerging evidence indicates that PM2.5 pollution is additionally associated with neurologic dysfunction in both adults and children. In adults, associations are reported between PM2.5 pollution and increased risk of dementia [977]. In children, air pollution is linked to loss of cognitive function (IQ loss), memory deficits, behavioral dysfunction, reductions in brain volume, and increased risks of attention deficit hyperactivity disorder (ADHD) and autism spectrum disorder [978]. Oxidative stress is a key initiating component in the cascade of pathophysiologic events triggered by PM2.5 exposure [979980981]. ## Air pollution—Volatile organic compounds Multiple volatile organic compounds (VOCs) are released to the atmosphere by oil and gas extraction. They include benzene, 1,3-butadiene, tetrachloroethane, methane, ethane, propane, toluene, methanol, ethanol, formaldehyde, and acetaldehyde as well as n-hexane, styrene, methanol, and 2,2,4-trimethylpentane. Both workers and “fenceline” community residents are exposed to these hazardous air pollutants. VOCs can cause eye, nose, and throat irritation [982]; headaches; loss of coordination [983]; nausea; and damage to the liver, kidneys, and central nervous system [984]. Some, such as benzene, 1,3-butadiene, and formaldehyde are known human carcinogens and can cause leukemia and lymphoma. Others are associated with increased risk of neuropathy and asthma [985986987]. ## Community health impacts of oil and gas extraction Oil and gas development are associated with multiple adverse health effects in nearby communities [992993]. Residents report foul-smelling air and burning eyes as well as coughing, migraines, dizziness, memory loss, and numbness in their hands, feet, arms, and legs. Diseases of livestock are reported, with cattle wasting and dying, and fewer calves born or surviving [991]. High rates of automobile accidents involving tanker trucks are reported in communities neighboring oil and gas extraction sites [94]. Increased cancers at various anatomic sites have been reported among adults in these communities [994] as well as in children (see Box 4.2). ## Box 4.2 Oil and Gas Extraction and Pediatric Cancer. Oil and gas are major feedstocks for plastic production. Many of the compounds used or produced in oil and gas extraction by drilling and fracking—most notably benzene, 1,3-butadiene, and formaldehyde—are known to cause leukemia and lymphoma in persons of all ages, including children [995]. Exposures to these materials during pregnancy and in early childhood are especially dangerous [996997]. Epidemiological studies conducted among children born or living near fracking sites have found elevated rates of childhood cancer, especially leukemia, and congenital heart defects [996998]. A 2017 registry-based study in Colorado found that children and young adults diagnosed with leukemia were four times more likely than controls to live in areas of extractive activity [996]. A Pennsylvania study found a nearly 2-fold increase over background in risk of acute lymphatic leukemia among children with at least one unconventional oil and gas development well within 2 km of their residence during the prenatal period and a 2.8-fold increase in risk of leukemia among children who had at least one unconventional oil and gas development well within 2 km of their residence at ages two to seven years [32]. ## Health impacts of fossil carbon transport Coal, oil, and gas destined for use as plastic feedstocks are transported by pipeline, ship, rail, and road from mines and drill sites to chemical and plastic production sites. Because many plastic feedstocks are toxic, flammable, and explosive, all of these modes of transport are associated with hazards to human health and the environment, as was seen in the February 2023 rail car disaster in East Palestine, OH, USA. ## Gas leaks Fires and explosions are the major health hazards associated with fossil carbon transport, and most involve gas. The major components of gas, which vary with site of origin, are methane ($60\%$–$90\%$), ethane (0–$20\%$), propane (0–$20\%$), and butane (0–$20\%$) [999]. All are flammable and explosive. The higher the pressure in a pipeline, the greater the danger of fire and explosion. Multiple gas pipeline explosions occur globally each year. In the US, the Pipeline and Hazardous Materials Safety Administration has collected data on more than 3,200 accidents deemed “serious” or “significant” since 1987 [1000]. ( Figure 4.2) **Figure 4.2:** *Fatalities caused by pipeline incidents.Permission: Pipeline and Hazardous Materials Safety Administration (no date) Pipeline incident 20-year trends | phmsa. Available at: https://www.phmsa.dot.gov/data-and-statistics/pipeline/pipeline-incident-20-year-trends (Accessed: 24 October 2022).Figure adapted from (Pipeline and Hazardous Materials Safety Administration (PHMSA), 2022) by Manuel Brunner (co-author).* Injuries, burns and deaths result from pipeline fires and explosions. Eighty explosions in a pipeline in Massachusetts in 2018 damaged 130 buildings, injured 23 people (including 2 firefighters), and killed 1 young man [1001]. Other notable explosions in the US have occurred in Armada Township, Michigan; Refugio, Texas; and Watford City, North Dakota. An April 2016 explosion in a high-pressure pipeline in Salem, Pennsylvania, created a crater 9 m wide, 15 m long, and 4 m deep; destroyed a house 60 m away; melted the siding off a house 300 m away; charred trees and telephone poles 1.6 km away; and hospitalized a man in his twenties with third-degree burns on over $75\%$ of his body [1002]. Compressor stations, which are sited at intervals along pipelines to push gas forward, leak gas into surrounding communities. Additionally, the diesel engines that power compressors produce incessant noise and generate both particulate and hazardous air pollutants, notably benzene, 1,3-butadiene, and formaldehyde—all known human carcinogens [1003]. Pipelines, and compressor stations are disproportionately sited in low-income, minority, and marginalized communities—environmental justice communities—where they deepen social injustice while producing no local benefit [1004]. ## Oil leaks and spills Oil spills are numerous, (Figure 4.3) and associated with an increased prevalence of respiratory problems, neurological effects, genotoxicity, endocrine disruption, immune dysfunction, and mental health disorders in both cleanup workers and community residents [119]. All of these effects were seen after the Exxon Valdez oil spill in Alaska in 1989 and following the Deepwater Horizon oil spill in the Gulf of Mexico in 2010. Improper site management following leaks is associated with elevated atmospheric concentrations of methane, benzene, mercury, xylenes, n-hexane, and toluene [1005]. **Figure 4.3:** *Global number of oil spills from tankers from 1970–2021.Permissions: no special permissions needed. Roser, M. and Ritchie, H. (2022) ‘Oil spills’, Our World in Data [Preprint]. Available at: https://ourworldindata.org/oil-spills (Accessed: 24 October 2022) and: ITOPF (2022). Oil tanker spill statistics 2021. ITOPF Ltd, London, UK. Figure adapted by Manuel Brunner (co-author).* ## Health impacts of cracking Catalytic cracking of coal, oil, and gas to create ethylene and propylene feedstocks is the first step in plastic production. Cracking occurs in massive, highly energy-intensive chemical plants. These facilities generate multiple air pollutants, including carbon monoxide, NOx, SOX, VOCs, and PM (Figure 4.4). They also release substantial quantities of CO2 and methane, thus contributing substantially to climate change [159]. They endanger the health of chemical workers and residents of nearby “fenceline” communities [159]. They are often sited in environmental justice communities. **Figure 4.4:** *Some toxic chemicals with human health impacts. The human health impacts of the chemicals listed in each phase of the plastic life cycle are described in the text.Credit: Designed in 2022 by Will Stahl-Timmins.* The principal health hazards to cracking workers result from their occupational exposures to toxic and carcinogenic petrochemicals. These include solvents (benzene, toluene, and xylene), 1,3-butadiene, and styrene [159]. Benzene and butadiene cause leukemias and lymphomas and are classed by the International Agency for Research on Cancer (IARC) as proven human carcinogens [132]. Styrene is neurotoxic and is classed by the IARC as a possible human carcinogen [1006]. Human toxicity impact potentials [1007] have been calculated across the cracking life cycle and include cancer, noncancer, and respiratory as well as environmental impact potentials of smog formation, freshwater ecotoxicity, and ozone depletion [159]. Risk of fire and explosion is high within refineries and cracking plants. From January 2000 to July 2013, 171 fires occurred in oil refineries in the US. Of these, about $8\%$ led to explosions, which can result in serious injuries and death [1008]. At least 58 workers have died at oil refineries and cracking facilities in the US since 2005 [1009]. ## Health impacts of refinement, polymerization, compounding, and conversion Many of the monomers, additives, and catalysts used to form plastic are highly toxic, and a number are carcinogenic [165]. These chemicals are released into workplace air in dust and vapor formed at virtually every stage of the production process—during handling and mixing of resins and additives, processing under heat and pressure, and finishing products after molding—and results in exposures of these chemicals to workers and to residents of nearby communities. ## Occupational health hazards Plastic production workers are occupationally exposed to toxic monomers (e.g., vinyl chloride, styrene, BPA, acrylonitrile, butadiene, ethylene, urethane); additives (e.g., phthalates, heavy metal stabilizers, flame retardants, hydrocarbons); dust, including airborne asbestos dust [96510101011]; and hazardous vapors (Figure 4.4). Compared to nonexposed workers, chemical body burdens in epoxy-resin workers are high: acrylonitrile (11-fold elevation above background), styrene (5.5-fold elevation) [238], di(2-ethylhexyl) phthalate (DEHP; 2-fold elevation), and BPA (2-fold elevation) [1012]. Elevated body burdens were even observed in workplaces where airborne concentrations of hazardous vapors were within legally permissible limits, and they remained evident after one or more days away from work. Toxic occupational exposures in plastic production workers are associated with a range of diseases: ## Community health impacts Petrochemical refining facilities, including cracking plants and plastic production facilities, are disproportionately located in low-income communities [1019]. In the US, over 530,000 people live within three miles of chemical refineries, with the majority being people of color and people living below the poverty line [1020]. People in these communities are exposed to hazardous chemical pollutants released from cracking plants into air, water, and soil. These pollutants include benzene, formaldehyde, toluene, and other hazardous air pollutants with recognized mutagenic, carcinogenic, and reproductive toxicity [9910211022]. ## Health impacts of synthetic textile manufacture Synthetic, plastic-based fibers are used in very large quantities to manufacture clothing, furniture, carpeting, tires, and myriad other consumer goods. Physical hazards of synthetic textile production include fire, noise, temperature, humidity, and unsafe machinery. Synthetic textile manufacture also results in occupational exposures to toxic chemicals and airborne dusts, including airborne plastic microfibers. Multiple diseases have been reported among workers in this industry [243]: ## Health Impacts of Plastics during Use Plastics are used to produce an enormous range of industrial and consumer goods that include construction materials, electronics, medical equipment, and children’s toys. Virtually all plastics-based products contain a wide range of chemical additives, often in very large quantities. These additives, recently estimated to number more than 10,500, include plasticizers, stabilizers, dyes, and flame retardants [8]. Depending on the product, additives can comprise $5\%$–$50\%$ by weight of manufactured plastics [213]. Most additives do not form strong chemical bonds with the polymer matrix. They can therefore leach from plastic to contaminate air, water, and soil and expose humans [170]. Systematic reviews and meta-analyses were searched for on PubMed and Epistemonoikos [10331034]. Primary research articles were searched in the Medline and Embase databases using the Ovid search platform, with additional searches on PubMed. Where available, we focus on in vivo human health research with direct quantification of chemicals in biospecimens such as blood or urine. Where human health research was absent, additional searches were made for the animal and in vitro literature. Most health hazards associated with plastics in the use phase result from exposures to toxic additives and other plastic-associated chemicals. For this reason, it is very important to consider additives in any discussion of the toxicity of plastics as well as in all negotiations on a Global Plastic Treaty. Additional health hazards of plastics in the use phase result from ingestion and inhalation of (MNPs) formed through the erosion and breakdown of plastics. An emerging body of evidence indicates that MNPs may cause direct toxic effects due to their physical accumulation in cells and tissues [1035] and that they can also cause toxicity by acting as vectors, “Trojan horses,” that transport chemical additives, adsorbed chemicals, and pathogens into tissues and cells [8821036]. Nevertheless, the contribution of MNPs to the transfer and bioaccumulation of adsorbed chemicals may be negligible compared to direct environmental transfer and bioaccumulation via prey and depends on multiple factors, such as the characteristics of the particles as well as the environmental milieu in which they are found [8111037]. ## Health impacts of plastic additives Exposures to plastic additives such as polychlorinated biphenyls (PCBs), BPA, phthalates, and BFRs are widespread, reflecting the great increases in plastic production and use in recent decades. In the US, the Centers for Disease Control and Prevention’s National Biomonitoring Program finds BPA and phthalates in the urine of nearly all people tested [10381039]. Children aged 6–11 years had the highest levels, followed by adolescents aged 12–19 years, non-Hispanic Blacks, and participants with lower-income levels. Multiple sources of human exposure to plastic additives have been documented: ## Endocrine disruptors A number of plastic additives act as EDCs—synthetic chemicals that “interfere with the synthesis, secretion, transport, metabolism, binding, or elimination of natural hormones responsible for homeostasis, reproduction, and developmental process.” [ 1041] The endocrine signaling system regulates every aspect of early human development, including body growth, organ formation, and development of the brain, reproductive system, and immune system. For this reason, EDC exposures in early life—during pregnancy and in the first two years after birth—are extremely dangerous because they can cause changes within developing cells and organs that disrupt organ formation and increase risk of disease and disability in childhood and across the life span (see Box 4.1) [1042]. EDCs in a pregnant woman’s body can be transmitted to her child through the placenta during pregnancy and through breast milk during lactation. These chemicals can also cross the developing blood-brain barrier in young children, allowing toxic levels to accumulate in brain tissue [10431044]. Children’s vulnerability to EDCs is further magnified by their age-appropriate behaviors, such as hand-to-mouth activity, crawling, and persistent contact with soil [1045], all of which increase exposure, and by the immaturity of enzymatic pathways in their liver and kidneys, which prevent them from efficiently detoxifying and eliminating EDCs [1045]. Many EDCs are lipophilic, bioaccumulate in adipose tissue, and persist in the body, thus leading to continuing exposure [1041]. At a societal level, extensive exposure to EDCs that results in widespread disease, disability, and premature death can have substantial economic impacts that are the consequence of lifelong increases in health care costs and decreases in productivity [1046]. ## Reproductive toxicity of EDCs EDCs such as BPA, PCBs, and phthalates can interfere with androgens and estrogens and thus disrupt sexual and reproductive development and function [1047]. The consequences are increased rates of premature births; increased frequency of birth defects, such as cryptorchidism and shortened anogenital distances in the reproductive organs of baby boys; increased risk for postnatal morbidity and mortality; and decreased fertility [10421047]. ## Metabolic toxicity of EDCs Some EDCs interfere with the actions of the thyroid hormone, thus disrupting energy metabolism in cells throughout the body [1041]. Others, termed obesogens, reprogram the insulin-glucagon axis that regulates appetite and energy balance [10411047], thus increasing the risk for obesity and diabetes. In experimental animals as well as in epidemiologic studies, perinatal exposures to obesogenic EDCs such as BPA induce weight gain and body fat accumulation [10431048]. ## Cardiovascular and renal toxicity of EDCs The metabolic changes initiated by obesogenic EDCs, including changes in serum lipid profiles and increased risk of hypertension, increase subsequent risk for cardiovascular disease and stroke [1049]. A major epidemiological study based on the NHANES survey in the general population of the US found that persons in the highest quartile of urinary BPA concentration had a significantly increased prevalence of myocardial infarction (odds ratio [OR] = 1.73, $95\%$ CI = 1.11–2.69) and of stroke (OR = 1.61, $95\%$ CI = 1.09–2.36) compared with persons in the lowest quartile [1050]. Additional epidemiological studies have found a positive relationship between BPA exposures and renal dysfunction [10511052]. Phthalate and PCB exposures are both associated with increased mortality from cardiovascular disease, and positive exposure-response trends are seen in these effects [1053]. ## Neurotoxicity Some plastic additives are neurotoxic in addition to being endocrine disruptors. Infants in the womb and children in the first years after birth are particularly vulnerable to brain injury caused by these chemicals because their brains are rapidly growing and developing throughout the first thousand days of life and continue to develop through childhood and adolescence and into early adulthood (see Box 4.3) [105410551056]. Injury to the developing brain caused by neurotoxic plastic additives results in neurodevelopmental disorders, behavioral change, and diminished cognitive function (reduced IQ). An unresolved question is whether brain injury caused by toxic chemicals in early life increases risk for dementia and other neurodegenerative disorders in later life. ## Box 4.3 The Vulnerability of the Developing Human Brain to Neurotoxic Chemicals. An estimated one in six children in the US has a neurodevelopmental disorder, about $17\%$ of all children [1057]. Prevalence rates of these disorders, including ADHD, autism spectrum disorder, cognitive impairment (IQ loss), dyslexia, reduced academic performance, behavioral changes, and reductions in brain volume appear to be increasing [10581059]. Toxic chemicals in the environment, including chemicals added to plastic, are important causes of neurodevelopmental disorders in children [10591060]. The brain and nervous system begin to develop in the first month of pregnancy with the formation of a thin strip of cells—the neural plate—along the dorsal side of the embryo [1061]. The cells in the neural plate are the precursors of the brain and spinal cord. Over the course of pregnancy, these cells divide and multiply at a very rapid rate, and by the time a baby is born, the brain contains approximately 100 billion neurons [1062]. As they are dividing and multiplying, neurons migrate from the positions where they are formed to their final destinations. As they migrate, the neurons form dense networks of connections with one another [1061]—an estimated 2,500 connections per neuron by the time of birth [1062]. Each connection must be precisely established, and redundant neurons and connections are pruned away through programmed cell death, or apoptosis. Almost $50\%$ of the neurons present in the brain at birth have undergone apoptosis by adolescence [1061]. All of these developments are highly interdependent, and optimal brain development requires that each step must occur in its proper sequence. A consequence of the great complexity of human brain development is exquisite vulnerability to toxic chemicals and other harmful environmental exposures [1042104310451047]. Exposures to even minute doses of toxic chemicals during critical developmental stages can have lifelong consequences for brain function and neurological health [105810601063]. These windows of vulnerability are unparalleled in adults. Multiple toxic chemicals have been found to induce adverse neurological effects in children at levels previously thought to be safe and that produce no adverse outcomes in adults [206106010641065106610671068]. Brain damage caused by exposure to toxic chemicals early in life may be chronic, irreversible, and difficult to treat [1058]. Prevention of exposure is the most effective strategy for safeguarding the developing human brain against toxic chemicals [105810691070]. Dr David Rall, former director of the US National Institute of Environmental Health Sciences, famously stated that “if thalidomide had caused a 10-point loss of IQ instead of obvious birth defects of the limbs, it would probably still be on the market [1071].” ## Neurotoxicity of PCBs Prenatal exposures to PCBs are associated with lower IQ and difficulties with motor skills, attention, and memory [10721073107410751076]. ## Neurotoxicity of phthalates Prenatal exposure to phthalates has been associated with increased risk for autism spectrum disorder [107710781079]. Certain phthalates are associated with poorer motor skills in children [10681079]. Prenatal exposures to phthalates are also associated with aggression, depression, conduct and attention problems, externalization of problems [1064], and diminished executive functioning in preschool children of both sexes [1064]. Mothers with higher exposure to phthalates during pregnancy are three times more likely than other women to have children diagnosed with ADHD [1065]. Phthalates act as anti-androgens in the developing central nervous system, disrupting the normal sexual differentiation of the brain. This finding may explain the sex-specificity of disruption in brain function in children exposed to phthalates, such as the observation that the exposure of baby boys to phthalates has been associated with less male-typical play behavior [1066]. Phthalates also interfere with estrogen synthesis and thyroid hormone production, both of which exert powerful influence on brain development. Prenatal BPA exposure is reported to influence hypothalamic morphology, concentrations of total testosterone, and sex-dependent neurological behaviors [1080]. A study to assess the impact of prenatal phthalate exposure on cognitive function conducted among 328 inner-city mothers and their children found that children whose mothers had the highest concentration of certain phthalates had significantly lower scores on perceptual reasoning (by 3.9 IQ points) and verbal comprehension (by 4.4 IQ points), respectively [1067]. A follow-up study of the children at age seven years showed similar results, where children born to women above the 75th concentration percentile for phthalates scored 6.6 and 7.6 IQ points lower than their less heavily exposed peers. These findings are consistent with those of other studies [1079] and suggest that the adverse impacts of prenatal phthalate exposure on cognitive function can persist into the early school years, with negative implications for academic performance and future earnings potential. ## Neurotoxicity of BPA and BPA substitutes Multiple and diverse adverse emotional and behavioral outcomes are associated with both in vitro and childhood BPA exposure [1081]. Maternal BPA exposure during pregnancy has been found to increase hyperactivity and aggression in two-year-old girls [1082]. In boys, prenatal BPA exposure is associated with increased aggression and emotional reactivity at ages three to five years [1083] and with increased depression and anxiety at age seven years. One study found much poorer scores across several behaviors in boys than girls, including withdrawal, depression, rule breaking, defiance, and conduct problems. These findings suggest the impacts of BPA exposure on the developing human brain may be sexually dimorphic [1084]. Studies in experimental animals confirm the findings from human studies, similarly finding neurodevelopmental and behavioral abnormalities following BPA exposure. Examples include locomotor deficits, anxiety-like behavior, and declarative memory impairments that persisted into old age following prenatal exposure [1085] and hyperactivity, anxiety-like behaviors, and elevated dopamine levels following exposure to BPA [1080]. A review found that BPA induces aggression, anxiety, cognitive deficits, and learning memory impairment [1080]. Studies to elucidate the possible mechanisms underlying these behavioral changes have found that BPA exposure during critical windows of development, such as breastfeeding and organogenesis, upregulates dopamine receptor function, while exposure at other time points does not have this effect [1080]. BPA exposure has been found to increase DNA methylation in rodents, and this change can be passed to future generations [1080]. Increasing evidence suggests that BPA substitutes can induce similar toxicity (see Box 4.4). ## Box 4.4 The Problem of Regrettable Substitution. The concept of regrettable substitution—the replacement of toxic chemicals in plastics with new materials that were never assessed for toxicity and subsequently found to be toxic—is exemplified by the flame retardant additives previously described. Without more robust, legally mandated premarket evaluation processes and without systems in place for early postmarket detection of unforeseen adverse effects, there is a risk for further regrettable substitutions as markets shift away from chemicals with established toxicity, such as BPA and phthalate plasticizers, to other bisphenol analogs and alternative plasticizers [197]. Similar to the chemicals they replace, these substitutes are able to leach out from plastic products, leading to environmental contamination and human exposure [197], and there is an urgency to have a process to identify potential hazards. We consider some selected examples here. Di-isononyl cyclohexane-1,2-dicarboxylate (DINCH). DINCH is commonly used to replace other phthalate plasticizers, such as DEHP and di-isononyl phthalate [1086], most frequently in food packaging products, toys, and medical devices [1087]. DINCH is oxidized to various metabolites in the human body, and these metabolites have been found to activate human nuclear receptors in vitro and thus interfere with hormonal activity [1086]. In vitro studies have shown that DINCH can alter lipid metabolism in steroidogenic cells and Sertoli cells and induce oxidative stress in the Sertoli cell line, indicating its potential to damage the endocrine and reproductive systems [1088]. Additional in vitro studies have found that DINCH exposure is associated with enhanced inflammatory responses in human macrophages and enhanced cellular stress [1089]. Studies conducted in zebrafish larvae have indicated that DINCH can impact transcriptional profiles, lipid metabolism, and behavior [1090]. Human epidemiological research confirms widespread exposure to DINCH [109110921093109410951096109710981099]. In the last seven years, a small number of epidemiological studies have now started to evaluate association with human health outcomes, with a particular focus on reproductive outcomes. Adverse associations have been observed between DINCH exposure and multiple outcomes, including oxidative stress in men [1093] and pregnant women [1095], sperm epigenetics [1092], in vitro fertilization outcomes [2111091], risk of preterm birth related to preconceptual and or prenatal exposure [1099], and fibroids in women [1096]. More robust evaluation is however limited by sensitivity of the techniques used to detect exposure, resulting in low detection rates in most of these studies. Di(2-ethylhexyl) adipate (DEHA). DEHA belongs to a second group of replacement plasticizers, but again, it has endocrine activity with potential for endocrine disruption, specifically with respect to metabolic and reproductive function, at least at high dose. In rat studies, high-dose intravenous injections of DEHA were shown to decrease the animals’ appetite and weight gain, increase liver weight in females, and decrease thymus weight in both males and females, indicating the compound’s disruptive effect on metabolism. In vitro studies demonstrate that DEHA exposure can alter mitochondrial activity in steroidogenic cells, disrupting the ability of these cells to synthesize sex hormones [1088]. In silico experiments show that DEHA is able to bind the ligand-binding pocket of human sex hormone–binding globulin, subsequently preventing sex hormones from binding and therefore interrupting normal endocrine function [212]. While DEHA exposure has been confirmed in human observational studies, there are not yet any observational studies to directly evaluate safety at current levels of exposure, including in terms of endocrine effects seen in animal and in vitro studies [1100]. Acetyl tributyl citrate (ATBC). ATBC is a plasticizer from a third group, the citrate esters. As with DINCH and DEHA, however, ATBC again has endocrine activity with potential for endocrine disruption, specifically with respect to reproductive function. ATBC is able to bind the ligand-binding pocket of human sex hormone–binding globulin in a similar manner to DEHA and thus has the potential to disrupt human sex hormone regulation [212]. ATBC was additionally found to have anti-estrogenic and anti-androgenic effects in the uterotrophic assay and steroidogenesis assays, respectively [1101]. Further in vitro studies using mouse cells demonstrated that exposure to ATBC reduced the viability of Leydig and fibroblast cells, with a greater effect demonstrated on Leydig cells, indicating the anti-androgenic effects of ATBC [1102]. In vivo studies have demonstrated that ATBC can slightly impair liver function in rats and interfere with the animals’ reproduction and development [1103]. There are not yet any observational studies to have directly evaluated human safety, including in terms of endocrine effects seen in vitro [1100]. Bisphenol S (BPS) and bisphenol F (BPF). BPS and BPF are chemically similar analogs of BPA, with similar application. In vivo and in vitro studies additionally demonstrated that they have endocrine-disrupting effects similar to those of BPA [1104]. Zebrafish assays studying BPA, BPS, and BPF have shown that all three chemicals can influence estrogenic activity [1105]. Similarly, studies in pigs have revealed that BPS can interfere with meiotic activity in oocytes [1106]. BPA, BPS, BPF, and bisphenol AF were all found to genetically alter steroidogenesis in H295R steroidogenic cells [1107]. Disruption of the thyroid hormone stimulating pathway as a result of BPA, BPS, and BPF exposure has also been documented in tadpoles both in vivo and in vitro, revealing the potential capacity of bisphenols as metabolic disruptors [1108]. A recent scoping review identified 21 human epidemiological studies on BPA and bisphenol analogs [1109]. BPS and BPF exposure have been associated with metabolic problems, such as increased risk of type 2 diabetes; pregnancy and reproductive issues, such as increased risk of late-term birth for girls; and skin sensitivity [1109]. ## Neurotoxicity of BFRs BFRs, including polybrominated biphenyls (PBBs), polybrominated diphenyl ethers (PBDEs), hexabromocyclododecane (HBCDD), and tetrabromobisphenol A (TBBPA), are added to consumer plastic products such as furniture, carpets, drapes, and consumer electronics to reduce flammability. A number of these compounds have been shown to be neurotoxic [181184189192197111011111112]. The toxicity of PBBs was discovered through investigation of a large contamination episode that occurred in Michigan in 1973 and resulted in the deaths of thousands of dairy cattle and other farm animals as well as widespread contamination of milk and other dairy products, beef, pork, lamb, chicken, and eggs. Human exposure was extensive [1113]. To study the human health effects of PBB exposure, a cohort of around 4,000 people—the Michigan Long-Term PBB study—was formed. It included workers at the chemical plant where the mislabeling had occurred, farm families with varying degrees of exposure, and members of the general public. The study found some evidence linking high PBB exposure to increased risk of breast cancer, lymphoma, and cancer of the digestive system [11141115]. Spontaneous abortion rates and increased weight loss were also seen in the Michigan cohort [1116]. Early-life exposure to PBBs has been shown to be associated with developmental neurotoxicity [1117]. In everyday exposure scenarios, PBB has been associated with increased risk of papillary thyroid cancer in women [1118], type 2 diabetes [11191120], and deep infiltrating endometriosis [1121]. PBDEs can leach from furniture, draperies, and electronics and settle into house dust [1122]. PBDEs in house dust can include both newer PBDEs as well as older and more highly toxic phased-out members of the class that are still present in older furniture [1911110]. PBDEs in house dust are a source of exposure for young children who crawl on the floor and exhibit age-appropriate behaviors, such as frequent touching of objects and hand-to-mouth activity [1110]. Some of the highest PBDE body burdens in the world have been reported in California children, because California’s Fire Safety Law Technical Bulletin 117 requires that furniture and baby and other household products contain very high levels of PBDEs [1111]. To investigate the developmental neurotoxicity of PBDEs, a 2013 California study undertaken within the Center for the Health Assessment of Mothers and Children of Salinas cohort analyzed the relationship between pre- and postnatal PBDE exposure and neurological functioning in children [1111]. Researchers found that both gestational and early childhood exposure to PBDEs were associated with adverse neurobehavioral development, including shortened attention span and poorer fine motor coordination at five and seven years of age, and a decrease in verbal and full-IQ scale at seven years [1111]. A 2010 longitudinal study analyzed PBDE levels in 210 cord blood specimens and assessed neurodevelopmental effects in children up to 72 months. Researchers found that children with higher cord blood concentrations of BDEs 47, 99, or 100 scored lower on mental and physical development tests such as IQ [1110]. Other epidemiological studies have found negative associations between prenatal PBDE exposure and impaired cognitive function, attention problems, anxious behavior, and increased withdrawal as well as with reductions in psychomotor development index and lower full-scale IQ performances [1123]. In these studies, PBDE congeners 47, 99, and 100 have consistently been associated with lower cognitive functioning [1124]. ## Neurotoxicity of organophosphate ester (OPE) flame retardants OPE flame retardants are structurally similar to organophosphate pesticides of known neurotoxicity, such as chlorpyrifos [192]. Human studies on the neurotoxicity of OPFRs have reported associations between early-life exposure to OPEs, decreased childhood IQ, and disrupted internalizing and externalizing behaviors [188]. In further studies, associations have been found between higher prenatal levels of OPE metabolites and poorer performance on cognitive tests, more frequent withdrawal and attention problems, and hyperactivity. Higher prenatal OPE metabolite levels were associated with poorer IQ scores at seven years of age and with more hyperactive behaviors [1125]. ## Light stabilizers Light stabilizers, also called UV stabilizers or photo stabilizers, may be added to plastics either to protect the plastic itself or—when used in plastic packaging materials—to protect goods within plastic packaging against photodegradation. Major classes of UV stabilizers include benzophenones (BzPs) and benzotriazoles. BzPs act as UV absorbers, with widespread use in sunscreens, cosmetics, and other personal care products [215216], and are added to plastics and textiles to protect them from photodegradation [216217]. BzPs can readily penetrate the skin and therefore be easily absorbed [1126]. A recent experimental study in humans showed that cotton clothing that had been exposed to elevated concentrations of BzP-3 in the air (4.4 microg/m3) for 32 days acted as an exposure route [1127]. After wearing the clothing for three hours, both the parent BzP-3 and its metabolite BzP-1 were detected in urine [1127]. BzPs are also released from food packaging materials [223457], and exposure has been linked to dietary sources such as frozen food, instant noodles, and instant coffee in a 2016 Korean study, with urinary concentrations being higher in younger cosmetics users and lean women [1128]. BzPs in personal care products have long been recognized as causing allergic contact dermatitis, with BzP-3 being the most common cause [215]. Systemic health impacts associated with exposure to BzPs in human epidemiological studies include endometriosis, renal function, oxidative stress markers in pregnant women, and disordered emotional-behavioral development in preschool children following prenatal exposure [11261129]. BzP-3 exposure is associated with decreased birth weight in girls and increased birth weight in boys; in addition, gestational age in both boys and girls is reduced [11301131]. Benzotriazoles are another class of high-volume chemicals used as UV stabilizers. They have been shown to present in plastic bottle caps, food packaging, and shopping bags [221]. Benzotriazoles are highly lipophilic, bioaccumulative, and persistent in the environment and have been detected in human blood, breast milk, and urine [221], including urine from the general population of seven countries (China, Greece, India, Japan, Korea, Vietnam, and the US) [220]. The human health effects of benzotriazoles have been investigated in only one cohort, which reported associations between three benzothiazone stabilizers (1-H-benzotriazole, tolytriazole, and xylyltriazole) and elevated estrogens and testosterone in the urine of pregnant women [1132]. In the same cohort, Chen et al. report association between benzothiazones and cord blood mitochondrial DNA copy number [1133]. Some benzothiazones appear to be endocrine disruptors that activate the aryl hydrocarbon receptor [11341135], and/or exhibit anti-androgenic, estrogenic, or anti-estrogenic activity at human sex hormone receptors [2211136]. In January 2021, the benzothiazone UV-328 was found to fulfill all persistent organic pollutant (POP) criteria under the Stockholm Convention, and in September 2022, the POPs Review Committee to the Stockholm Convention recommended that UV-328 be listed in Annex A to the Convention [222]. Other light stabilizers include carbon black (a UV absorber, with additional application as a pigment and filler in black plastics) [179]; triazine UV absorbers [224]; metal chelate UV quenchers, such as nickel chelates [179225]; and hindered amine quenchers [179225]. Eleven triazine UV stabilizers have recently been detected in human breast milk alongside benzotriazoles and BzPs [226], in addition to detection in household dust and air. Except for carbon black, which is a known human carcinogen, the health effects of these compounds have been little studied [223457]. ## Melamine Melamine-formaldehyde resins, commonly referred to as “melamine,” are thermoset aminoplast plastics produced as copolymers of melamine and formaldehyde, sometimes blended with urea-formaldehyde resins. Key consumer applications include tableware, laminate surfaces on furniture, and inside coatings of food cans [1170]. Degradation of melamine-formaldehyde resin leads to the release of melamine and formaldehyde [11711172], and in some instances cyanuric acid [1173]. The human health impacts of high-dose, acute exposure to melamine were seen in a tragic episode of adulteration of infant milk formula with melamine in China. The main clinical consequences were renal stones and associated renal damage in young children [11741175]. Subsequent research has shown that lower-level exposures to melamine in the general population through melamine-formaldehyde resins may be associated with broader health impacts [117611771178], notably more renal toxicity. In a series of Taiwanese studies, melamine exposure was associated with renal stone formation (urolithiasis) in adults [1179], increased renal damage in patients with urolithiasis [11801181], and progressive decline in renal function in patients with chronic kidney disease [1182]. Studies in the US indicate that melamine exposure levels in the general adult [11831184] and child [11841185] populations are similar or higher than those seen in the Taiwanese studies and approach the levels seen in melamine factory workers in Taiwan [1186]. *These* general population studies find evidence of association between urinary melamine and diminished renal function in adults [1183] and between urinary melamine and cyanuric acid levels and markers of renal damage in children [1185]. ## Health impacts of microplastic and nanoplastic particles (MNPs) Fragmentation of plastic through use or following disposal leads to the production and release into the environment of millions of MP particles (1–5,000 µm in diameter) [21187], which can further degrade over time into NPs; <1 µm in diameter [5281188]. Additionally, there are primary MNPs, materials that are intentionally manufactured at micro or nano size for use in domestic and biomedical products, and these materials can also leak into the environment [118911901191]. MNPs are ubiquitously present in the environment today and have been detected in air [11921193], soil [271], and water supplies [119411951196]. However, knowledge of the safety of MNPs is limited by the fact that most experimental studies use standard particles that are most often PS, spherical, and of a known size, whereas environmental MNPs to which humans are exposed are highly diverse in terms of the type of plastic and additives therein, as well as their size, shape, and adsorbed chemicals [171]. ## MNP exposure Humans are exposed to MNPs via multiple routes [78011971198], with ingestion and inhalation being the main pathways [393780]. Exposure to MNPs and their associated chemicals and pathogens can occur by direct ingestion and inhalation or via consumption of contaminated food products, such as fish [11991200] and wheat [1201] (see Table 3.2). ## MNP ingestion exposure Ingestion is a major route of human MNP exposure [788]. Humans are exposed to MNPs via consumption of contaminated food products, such as fruits and vegetables [1202], seafood [120312041205] (see Section 3), table salts [120612071208], drinking water [12091210], and other daily consumables [12081211], or consumption of drinks containing MNPs leached from plastic products, such as plastic bottles [12121213], tea bags [1214], and coffee cups [1215], during use. Additional exposure can occur via accidental ingestion of personal care products containing MNPs, such as toothpaste [120312081216]. There is a broad range of uncertainty in current estimates of human MNP exposure via ingestion [1217]. Estimates of typical exposures range from a high of 0.1–5 g/week (average exposure via ingestion only) [1218] to a low as 4.1 µg/week (median exposure in adults via both ingestion and inhalation) [530]. Key challenges to exposure estimation include combining data across studies based on mass with those based on count and heterogeneity in the polymers measured, technique used, and range of particle sizes reported upon. Once ingested, MNPs travel through the digestive tract. They can accumulate within the gut, be excreted, or be absorbed. MPs have been detected in human colonic mucosal samples [1219], and several studies have reported MNPs in human stools [12031208122012211222]. MNPs can be absorbed from the digestive tract via several pathways, depending on their size, including endocytosis by epithelial cells, transcytosis via M-cells or across tight junctions of the epithelial cell layer, and persorption via epithelial gaps formed after shedding of enterocytes from villous tips [11871223], with estimates of uptake of spherical PS MNPs being $10\%$ for 60 nm MNPs after five days of daily oral gavage in rats [1224] and $6.16\%$, $1.53\%$, and $0.46\%$ for MNPs of sizes 50, 500, and 5000 nm, respectively, 24 hours after a single oral gavage in mice [1035]. Local effects of MNP on gut permeability may also increase uptake. MNP-induced oxidative stress has been shown to cause intestinal epithelial cell apoptosis and increased gut permeability in mice, leading to increased MNP absorption and biodistribution in several organs [1035]. ## MNP inhalation exposure Inhalation of MNPs released to indoor and outdoor air during the manufacture and use of products such as synthetic textiles, synthetic rubber tires, paint, and plastic covers is a second important exposure route for systemic uptake [393780]. In evaluation of airborne particles, it is those with an aerodynamic diameter of less than 10 μm, and especially those of less than 2.5 μm, that are most likely be inhaled into lower airways rather than deposited in upper airways and ultimately swallowed [1225], although microfibers of up to 250 μm in size have been found in human lungs [1226]. Therefore, depending on their size and shape, once inhaled, MNPs can be trapped by the lungs’ lining fluid and be expelled with sputum or swallowed, or they can be deposited deeper in the lungs [780]. Upon contact with the alveolar epithelium, they can translocate into epithelial cells and macrophages via diffusion, direct cellular penetration, or active cellular uptake [780]. Humans are estimated to inhale between 26 and 170 airborne MNPs of varying sizes per day [255783]. In humans, MNPs of multiple sizes and shapes have been detected in the sputum [1227] and bronchoalveolar lavage fluid [1228] as well as lung tissue obtained from surgical resections [122612291230] and autopsies [393]. The abundance of MNPs in human lungs has been shown to increase with age [12281229]. ## MNP dermal exposure Dermal contact is another potential route of MNP exposure. Dermal exposure can occur via MNPs present in indoor dust, microfibers from textiles [11971231], and in a range of personal care and cosmetic products, such as face and body scrubs [1232123312341235], shower gels [1236], and shampoo [427]. Dermal contact is considered the least significant route of MNP exposure in terms of potential for systemic uptake [1237] because absorption via skin is believed to be minimal for MNPs greater than 10–100 nm in diameter [1238]. PS NPs (≥20 nm) have been shown to accumulate in hair follicles of pig skin [12391240] and only penetrate the outermost layer of pig skin, even with a modestly compromised skin barrier [1240]. In contrast, studies using ex vivo human skin have shown that 750–6,000 nm PS MNPs can penetrate the skin via the hair follicles [1241]. Rare events of deeper penetration of smaller PS MNPs (20–200 nm) into viable epidermis at sites of high focal particle aggregation have also been detected in ex vivo human skin tissue [12421243]. ## MNP transplacental exposure A route of exposure unique to infants is transplacental transfer of maternal MNPs during pregnancy, with a number of studies now reporting extensive detection of MNPs in human placental tissue following delivery [1244124512461247]. Transplacental transfer is supported by detection of MNPs in human meconium samples [12441248] and by ex vivo human studies [12491250]. In vivo rodent studies have similarly confirmed the transfer of 20 nm rhodamine-labeled spherical PS MNPs from the maternal to the fetal circulation [1251]. ## MNP exposure risks for children Infants are exposed to MNPs in infant feeding bottles [1252], infant formula [1253], and expressed breast milk stored in plastic containers [12531254]. Young children are exposed via age-specific behaviors, such as hand-to-mouth activity [251]. As a result, infant exposure to MNPs is likely to be greater than that of adults. Indeed, concentrations of MNPs reported in infant stools are an order of magnitude greater than in adults [1248]. The potential role of these early exposures in increasing risk for disease in later life is not yet known [125512561257]. ## MNP exposures via biomedical products and procedures Several studies have detected ultra-high molecular weight PE MNPs in periprosthetic tissues in patients with joint replacements [77112581259]. Veruva and colleagues [1260] showed strong correlations between the number of wear MNPs and macrophages and the level of inflammatory markers in periprosthetic tissue samples from patients with total disk replacement. Importantly, by inducing an inflammatory reaction [1261] mediated by macrophages [12621263], wear MNPs from prostheses can lead to periprosthetic osteolysis [12641265]. ## MNP toxicity Studies of human health effects of MNPs are limited by the lack of a standardized method for quantifying MNP exposure, as they remain in early stages of development [126612671268]. As a consequence, most health evaluation is currently based on data from animal and in vitro studies [12691270], and we include that evidence here. ## Gastrointestinal toxicity of ingested MNPs Pathways for local toxicity of ingested MNPs in the gut include direct effects on gut epithelium and local uptake and indirect effects through actions in the lumen. A recent study found significantly higher concentrations of MNPs in stools of adults with inflammatory bowel disease compared to healthy individuals, and MNP concentration positively correlated with inflammatory bowel disease symptom severity [1271]. While these data are cross-sectional and reverse causality may be possible, there is mechanistic support from in vivo animal studies [1272], with MNPs ranging in shape and polymer type and size ranging from 0.1 μm to 5 mm shown to induce intestinal inflammation in a number of animal models (zebrafish, goldfish, and mice). In addition, in vitro studies have shown that exposure to MNPs can induce oxidative stress [1273], reduce mitochondrial membrane potential [1273], and reduce viability of Caco-2 human gut cells [12731274]. Overall, MNP exposure is suspected to be a digestive hazard to humans by adversely impacting the colon and small intestine, cell proliferation and cell death, chronic inflammation, and immunosuppression [1275]. MNPs can also have indirect effects on the gut through action within the lumen, with MNPs reported to induce gut microbial dysbiosis in mice [1276]. Similarly, in vitro experiments with Caco-2 human gut cells have shown that 30–140 μm PE MNPs altered the gut microbial population, which inhibited the following microbial metabolite pathways: the pentose phosphate metabolism pathway, which has an antioxidant effect in the gut; fructose and mannose metabolism, which may overload their absorption by the gut, leading to increased risk of obesity; and synthesis of secondary bile acids, which is crucial for GI function [1273]. ## Pulmonary toxicity of inhaled MNPs and microfibers Synthetic fabrics, tires, and other fiber-based plastic products can release MP fibers to the environment through their use and disposal. MP fibers are also produced intentionally, and their manufacture can result in occupational exposures to production workers (see Section 2). Scientific evidence for adverse health effects of MP fibers is mainly related to exposures during textile manufacture [248]. Extensive studies of the toxicity of inhaled microfibers have been conducted in nylon flock workers. These studies are reviewed in the “Health impacts of synthetic textile manufacture” section. Health effects include elevated risks of a range of respiratory symptoms [127712781279], decreased lung function [1228], accumulation in pulmonary tissue [1230], and stomach and esophageal cancers [1280]. In vitro studies have shown that PS MNPs can be absorbed by human lung cells (25–100 nm) [128112821283] and can induce negative cytotoxic effects, including increased pro-inflammatory markers (25 nm–2.17 μm) [1282128312841285], oxidative stress (40 nm–2.17 μm) [1281128212841285], mitochondrial disruption (50 nm) [1282], genotoxicity (50 nm) [1281], altered cell morphology (1 μm and 10 μm) [1286], and decreased cell viability (25 nm–2.17 μm) [12821283128412851287], cell proliferation (40 nm–10 μm) [12851286], and cell metabolic activity and cohesion between cells (1 μm and 10 μm) [1286]. In vivo rodent studies have shown that exposure to 64–535 nm PS microspheres increase pro-inflammatory markers [12881289] and that MNP exposure can cause pulmonary inflammatory cell infiltration, bronchoalveolar macrophage aggregation (1–5 µm commercial synthetic polymer microspheres) [1290], impaired lung function, and increased heart weight (0.10 μm PS microsphere) [1289]. ## Dermal toxicity of MNPs In vitro human studies have shown internalization of fluorescently labeled PS MNPs (200 nm) by human keratinocytes and induction of several cytotoxic effects by PE MNPs (100–400 nm) and MNPs isolated from commercial face scrubs (30–300 nm), including oxidative stress, reduction in cell viability at high concentrations, and inhibition of cell proliferation [1291]. However, to date, there are no studies assessing human dermal exposure to MNPs and its potential health effects in vivo [1197]. ## Placental toxicity of MNPs There have been few studies on health effects of transplacental MNP exposure. However, experimental studies in rodents have shown that maternal exposure to 5 μm spherical PS MNPs can permanently alter metabolism in the F1 and F2 generations, which increases their risk of metabolic disorder [12921293]. In humans, the presence of MNPs in the placenta have been associated with several ultrastructural alterations, such as narrowing of fetal capillary and changes in mitochondrial and endoplasmic reticulum morphology in human placenta [1246]. ## Systemic toxicity of MNPS When they enter the body via any pathway, MNPs can be transported to all the organs through the circulation [12941295]. In mice, 50 nm and 500 nm spherical PS MNPs absorbed through the gut accumulated in the heart, liver, lungs, spleen, kidneys, testis, epididymis, and brain [1035]. In pregnant rats, 20 nm spherical PS MNP absorbed though the lungs accumulated in the maternal lungs, heart, and spleen and also translocated through the placenta to accumulate in the fetal heart, liver, lungs, kidneys, and brain [1251]. In humans, MNPs have been detected in several internal organs, including the para-aortic lymph nodes, liver, and spleen [7711296]. Data on human tissue-specific effects following systemic MNP exposure is derived from in vitro human studies, with experimental MNP exposure aiming to model that which should enter systemic circulation [1297]. Key recent studies have shown that spherical PS MNPs can induce cytokine and chemokine production in mast cells and peripheral blood mononuclear cells (0.46–10 μm) [1298]; induce morphologic alterations, changes in gene expression, and apoptosis in microglia (0.2 μm and 2 μm) [1299]; and accumulate in human kidney proximal tubular epithelial cells, where they induce inflammation, mitochondrial dysfunction, endoplasmic reticulum stress, and autophagy (2 μm) [1300]. PE beads of 10–45 μm can have genotoxic effects in peripheral blood lymphocytes [1301], and several types of MNP beads can destabilize lipid membranes in red blood cells by mechanical stretching (0.8 μm PS, 1 μm and 10 μm PE, and 1 μm and 8 μm polymethylmethacrylate) [859]. In experimental animals, studies of MNP toxicity have generated abundant information [1302]. A recent systematic review reported induction of inflammation, oxidative stress, and several reproductive impacts of MNP exposure in rodents due to their accumulation in the ovaries and testes [1270]. For example, spherical PS MNPs disrupted the blood-testis barrier (38.92 nm–10 μm), induced testicular atrophy (38.92 nm and 100 nm), reduced sperm count (0.5–10 μm), increased ovarian collagen and fibronectin (0.5 μm), and impaired germ cell development in rodents (0.1–10 μm) [1270]. Indeed, a recent rapid systematic review concluded that MNPs may be a hazard to the human reproductive system [1275]. In addition to reproductive impacts, a recent scoping review by da Silva Brito and colleagues [863] reported metabolic and endocrine disruption, hepatotoxicity, and neurotoxic effects related to MNP exposure in rodents. Therefore, accumulation of MNPs in internal organs can potentially lead to serious health impacts. ## Mechanisms of MNP toxicity Potential mechanisms of MNP toxicity include effects due to physical properties such as a particle size and shape; effects related to chemical properties of the polymer; effects related to leaching of additives and other chemicals from MNPs; effects related to toxic chemicals that adsorb to the surfaces of MNPs; and effects related to pathogenic microbes that can adhere to the surfaces of MNPs to form a biofilm [206417521303]. ## MNP toxicity—Physical and material toxicity Since the majority of MNPs originate from the breakdown of a wide range of larger plastics, they come in many different sizes and shapes (e.g., spheres, irregular, fibers) and are composed of many different polymers and additive chemicals [201037]. These factors determine the properties and bioavailability of individual MNPs [10371304] and therefore influence their toxicity (see systematic review [1305]). MNPs appear able to exert direct toxicity related to their size, shape, and chemical composition. This direct toxicity is a relatively recently recognized phenomenon [201303]. It is separate from the toxicity that results from exposures to monomers and additive chemicals that leach from MNPs [8611306]. Smaller [130713081309], irregular-shaped MNPs [8611310] with sharp edges [130713101311] have been shown to have the highest toxicity in in vivo animal (e.g., decreased reproductive success and body length of Daphnia magna) and in vitro human studies (e.g., increased release of pro-inflammatory cytokine and ROS and reduced cell viability). The use of control particles in experimental studies has elucidated that the effects of MNPs (e.g., reproductive toxicity in *Daphnia magna* and antioxidant activity in mussels) are likely related to the plastic material itself and are not a result of particle exposure per se [130613091312]. The toxic effects of MNPs have additionally been found to be polymer dependent. PVC MNPs showed the greatest particle toxicity in an in vivo study in *Daphnia magna* (highest reproductive toxicity) [1306] and in an in vitro study using primary human monocyte and monocyte-derived dendritic cells (greatest increase in pro-inflammatory cytokine release) [861]. Importantly, this effect of polymer has been found to be independent of leached chemicals, with an independent effect of certain leached chemicals also demonstrated [8611306]. An observation of possible relevance to understanding the direct toxicity of MNPs is that metal(-oxide) nanoparticles, similar in size to NPs, have been shown to target the central nervous system. In several animal studies, these very small particles were found able to cross the blood-brain barrier and enter the brain through olfactory nerve endings, resulting in altered neurotransmitter levels, acetylcholinesterase inhibition, oxidative stress, neuroinflammation, cell damage, and death [1295]. ## MNPs as a vector for toxic additives and monomers MNPs can cause toxicity by releasing toxic chemicals such as monomers, plasticizers, flame retardants, antioxidants, and UV stabilizers from their plastic matrix into cells, tissues, and body fluids [13131314]. This has been termed the “Trojan horse” effect [8821036] (see Section 3), and although it may not play a major role in most habitats [1037], risks to human health from this source are unknown [811]. ## MNPs as a vector for environmental toxins and pathogens MNPs can cause toxicity through their ability to adsorb toxic chemicals and pathogens from the surrounding environment and transport these materials into cells and tissues [10351223122412251228]. Toxic chemicals detected on the surface of MNPs include POPs such as PCBs [81813151316], PFAS [1317], and PAHs [5078181201131513181319]. The adsorption capacity of MNPs depends on their surface area, with smaller particles having higher active surface area, and also on the polymer type for both environmental pollutants [81813201321] and microorganisms [1322]. MNPs can also act as a vectors for microorganisms [27213231324], including human pathogens [13251326] containing antibiotic resistance genes [132713281329]. Pathogenic bacteria such as E. coli have been found on plastic pellets on bathing beaches [13301331]. A recent human study showed that the presence of pathological microbes in bronchoalveolar lavage fluid was associated with higher MP concentrations [1228]. MNPs can thus play a role in facilitating the emergence of infectious diseases [1332]. ## Health Impacts of Plastic Waste An estimated 400 Mt of plastic waste are generated globally each year, and this volume is increasing in parallel with annual increases in plastic production, especially in the production of short-lived and single-use plastics. Less than $10\%$ of plastic waste is recycled, in contrast to recycling rates of $65\%$–$70\%$ for paper and cardboard and $90\%$ for glass [1333]. Most plastic waste is discarded in landfills, incinerated, or exported to LMICs, where it threatens the health of approximately two billion people [362]. A particularly hazardous component of plastic waste is electronic waste (e-waste). ## Health impacts of landfilled plastic Plastic discarded in landfills accumulates in enormous quantities, especially in LMICs [1334]. Mismanaged plastic waste that escapes from landfills clogs waterways and disfigures beaches. Plastic waste can also catch fire, exposing nearby residents to toxic combustion products. As it breaks down, plastic waste can generate MNPs and leach toxic additives into surface water and groundwater (see Section 2). ## Health impacts of unmanaged plastic incineration Combustion of plastic in open pits generates copious amounts of particulate air pollution as well as multiple hazardous air pollutants that include polychlorinated dioxins and dibenzofurans (PCDD/Fs), PCBs, and heavy metals. In Indian cities, the burning of waste and plastic accounts for $13.5\%$ of all PM2.5 air pollution and is linked to an estimated $5.1\%$ of lung cancer cases (total 5,000 per million population) or 255 cases per million [367]. ## Health impacts of “waste-to-energy” Thermal conversion, or pyrolysis, of waste plastic via incineration in waste-to-energy facilities results in the generation of a wide range of hazardous chemicals, including chlorine, hydrogen chloride, and phosgene (mustard gas); hydrogen cyanide; and ammonia as well as formic acid, formaldehyde, benzene and its derivatives, phenol, and PCDD/Fs [321]. The main sources of these toxic combustion chemicals are PVC and condensation polymers such as PURs, PA, and phenyl-formaldehyde resins. Human exposure to toxic chemicals produced by thermal conversion of waste plastic occurs through dietary intake, inhalation of contaminated air, soil and dust ingestion [1335], and dermal contact [1336], with dietary intake accounting for $60\%$–$99\%$ of total PCDD/F intake [1336]. ## Electronic waste (e-waste) Electronic waste comprises a significant portion of discarded plastic (see Section 2). Over 50 Mt of e-waste are generated annually, most in high-income countries. E-waste consists of discarded computers, mobile phones, televisions, and appliances and contains great quantities of plastic. Planned obsolescence is a major driver. Only $17.4\%$ of discarded electronics is recycled, and $7\%$–$20\%$ is exported to LMICs [1341], where it accumulates in enormous deposits, such as those in Agbogbloshie, Ghana [1342]; Latin America; and various locations in south China [1343]. The WHO reports that more than 18 million children, some as young as five years old, are employed in e-waste recycling, where they are exposed to lead, mercury, and PCDDs under horrificconditions [1344]. Plastics commonly used in electronics manufacture and found in e-waste are acrylonitrile butadiene styrene, PS, PC, PVC, PE, and PP. In addition, more than 200 different types of flame retardants are used in electronics production and can constitute $15\%$ by weight of electronic products. These flame retardants can be chlorinated or brominated, phosphorus-based, and aluminum trihydrate and its derived inorganic trihydrates. Metals, either in plastic or from other sources, are also present [348]. Associations between e-waste exposures and compromised thyroid function have been reported, although results were inconsistent. Exposure to metals in e-waste was associated with lower forced vital capacity in eight- and 9-year-old boys. Pregnancy outcomes included consistent associations with increased spontaneous abortions, stillbirths, and premature births as well as reduced birth weight and birth length. Physical growth (i.e., height, weight, and body mass index) was also stunted. Increased lead levels in blood were associated with low scores on neonatal behavioral neurological assessment. Increased frequencies of DNA damage and micronucleated and binucleated cells were also seen in peripheralblood [346]. ## Harms to Human Health of Climate Change Caused by Plastics Plastic is a contributor to climate change [479480481]. In 2019, global greenhouse gas (GHG) emissions across the plastics life cycle were estimated to be 1.8 Gt of CO2e, approximately $3.7\%$ of current global GHG emissions [5]. The greatest fraction of these emissions arises in plastic production. GHG emissions from plastic are projected to increase to 4.3 Gt CO2e by2060 [14]. Myriad catastrophic environmental events are associated with climate change, and all have potential to significantly impact human health [307]. They include the direct effects of heat and extreme weather conditions and events; indirect ecosystem-mediated effects, such as expanded ranges of disease vectors [13451346] and effects on food systems; and effects mediated by socioeconomic pathways, such as increased poverty, sociopolitical tension and/or conflict, and population displacement [1347]. The WHO considers climate change the gravest threat to human health in the 21st century, having the potential to undo all the progress made over the past 50 years in human development, global health, and povertyreduction [307]. Studies investigating the human health impacts of climate change have increased exponentially in number [1347] and unequivocally indicate that climate change has deleterious impacts on mortality; infectious diseases; respiratory, cardiovascular, and neurological illnesses and diseases; mental health; nutrition; pregnancy and birth outcomes; skin diseases and allergies; occupational health and injuries; and health systems [134513481349]. Despite these great gains in knowledge, critical gaps remain in the evidence base, and further investigation is required [13451347]. The effects of climate change disproportionately impact populations that have contributed least to the problem [1346], and they particularly threaten society’s most vulnerable and disadvantaged groups. These include women and children, older adults, individuals with underlying health conditions, poor communities and low-income countries, ethnic minorities and Indigenous populations, and migrants and displaced persons [30713501351]. Climate change has the potential to amplify the health impacts associated with plastic across its life cycle. For example, the increasing frequency and intensity of climate-related extreme weather events and flooding have the potential to exacerbate the spread of plastic in the natural environment [480] and to accelerate the spread of MNPs and the release into the environment of plastic additives [481]. Rising sea levels have the potential to increase the amount of ocean plastic pollution [1352]. Storms have the potential to damage warehouses that store raw materials and finished plastic products, thereby causing unintentional plastic leakage into the environment. Wildfires, which are occurring with increased frequency and severity [13531354], result in the combustion of plastic materials, especially when they reach urbanized environments. This, in turn, releases copious amounts of air pollutants, including PM, ozone, carbon monoxide, NOx, and VOVs [1354] as well as PCDDs and PCDFs s, mercury, and PCBs [1355]. Increasing ocean acidification, seawater salinity, and higher temperatures combined with UV light exposure accelerate the rate at which plastics fragment and release additives into the environment [5131356135713581359]. Plastic pollution is enhancing the spread of vector-borne diseases by increasing the number of sites available for mosquito breeding [1360]. Bacteria, including pathogenic bacteria, can colonize MPs in the sea [132513611362] and may travel vast distances, expanding their geographic range [1363]. Rising sea surface temperatures have been found to increase the virulence and capacity for the marine pathogen *Vibrio parahaemolyticus* to adhere and form biofilms onplastics [1364]. ## Health costs of local air pollution Fossil fuels provide the feedstock used to make plastics and are also the main source of the energy used in plastics production. Plastics production is highly energy-intensive: $87\%$ of fossil fuels used in plastics production are combusted [13], resulting in air pollution emissions (PM2.5, NOx, and SOx) as well as in the generation of GHGs. Plastics disposal (e.g., incineration, pyrolysis, and landfilling) results in additional air pollution and GHG emissions. Multi-regional input-output models have made it possible to trace the location of plastics production facilities and associated fuel use from the extraction of fossil carbon, through their combustion in the production process, and in the incineration and landfilling of plastics. The local air pollution impacts of plastics production are greatest in the regions and countries where fossil fuel extraction and transportation, resin production, and manufacturing are concentrated, often in low-income countries. Table 5.1 presents estimates of the workforce engaged in plastics production and disposal in 2015 by geographic location and stage of production. Seventy-five percent of the global plastics workforce in 2015 were located in Asia—$41\%$ in China and $35\%$ in the rest of Asia. Only $5\%$ were located in the EU and $2\%$ in the US. The other four areas listed—the rest of the Americas, the rest of Europe, Africa, and the Middle East each accounted for no more than $6\%$ of the global plastics workforce. **Table 5.1** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Unnamed: 5 | Unnamed: 6 | Unnamed: 7 | Unnamed: 8 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | UPSTREAM CHAIN | RESIN PRODUCTION | MANUFACTURING | RECYCLING | INCINERA-TION | LAND-FILLING | TOTAL WORKFORCE | TOTAL (UNIT: %) | | EU | 4090 | 1184 | 755 | 199 | 73 | 28 | 6329 | 5.79% | | US | 1950 | 438 | 37 | 48 | 33 | 21 | 2527 | 2.31% | | China | 31300 | 11900 | 63 | 1090 | 8 | 45 | 44405 | 40.65% | | RoW Asia | 24600 | 12262 | 517 | 40 | 36 | 23 | 37478 | 34.31% | | RoW Americas | 3527 | 2626 | 263 | 7 | 7 | 130 | 6560 | 6.01% | | RoW Europe | 3308 | 1265 | 654 | 30 | 25 | 251 | 5533 | 5.07% | | RoW Africa | 4732 | 571 | 129 | 0 | 29 | 28 | 5490 | 5.03% | | RoW Middle East | 668 | 232 | 7 | 0 | 4 | 6 | 917 | 0.84% | | Total | 74175 | 30477 | 2425 | 1414 | 216 | 531 | 109238 | 100.00% | The types of fossil fuels burned to produce plastics and associated emissions reflect the location of production. The world’s largest consumers of coal—China, followed by India and Indonesia—are also major plastic producers. Thus, in 2015 coal constituted $44\%$ of fossil fuel used in plastics production globally followed by oil $40\%$, and natural gas $8\%$. The sulfur and ash content of coal, the sulfur content of oil, and the extent to which pollution control equipment is used in extraction and production processes determine the amounts of primary PM emitted. Emissions of SO2 and NOx determine secondary particulate formation. Cabernard et al. [ 13] provide estimates of exposure to ambient PM2.5 pollution and associated health effects throughout the plastics production process. Specifically, they estimate PM2.5 intake fractions [13651366] associated with the burning of different fuels used to produce (and dispose of) plastics, by country. Estimates of the number of deaths associated with PM2.5 exposures are based on concentration-response functions from the Global Burden of Disease study [97013671368] and reflect deaths due to ischemic heart disease, stroke, chronic obstructive pulmonary disease, lower respiratory infection, and lung cancer among adults 30 years of age and older. Mortality estimates associated with plastic, by country group for 2015, are presented in Table 5.2. Cabernard et al. [ 13] estimate that local air pollution from plastics production and disposal resulted in 159,000 deaths globally in 2015, of which $99\%$ are associated with plastics production. Seventy-nine percent of these deaths occurred in Asia—$31\%$ in China. Only $6\%$ occurred in the USA and the EU combined. Geographic patterns of these deaths reflect differences in ambient PM pollution levels associated with plastics production, differences in population density surrounding production facilities, and differences in baseline mortality associated with cardiorespiratory diseases of persons living near these fatalities. The overall pattern is, however, clear: deaths associated with ambient PM pollution from plastics production occur where production is taking place: primarily in low- and middle-income countries (LMICs). **Table 5.2** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | | --- | --- | --- | --- | | | ENTIRE LIFE CYCLE OF PLASTICS | DISPOSAL (DIRECT IMPACTS OF INCINERATION AND LANDFILLING) | TOTAL (UNIT: %) | | EU | 6474 | 49 | 4.06% | | US | 3540 | 80 | 2.22% | | China | 48900 | 629 | 30.66% | | RoW Asia | 76512 | 59 | 47.97% | | RoW Americas | 6290 | 11 | 3.94% | | RoW Europe | 4216 | 264 | 2.64% | | RoW Africa | 8700 | 33 | 5.45% | | RoW Middle East | 4860 | 10 | 3.05% | | Total | 159491 | 1134 | 100.00% | ## Health impacts on workers Workers involved in plastics production are exposed to a variety of hazards in addition to ambient PM2.5 pollution.1 These include toxic and carcinogenic chemicals used to produce plastics, such as benzene, formaldehyde, and vinyl chloride as well as gases and PM inside factories. These workers are also at high risk of occupational death and injury. The Global Burden of Disease study [1369] estimates occupational fatalities for various categories of hazards, by country. Those most relevant to plastics production include exposure to benzene, exposure to formaldehyde, exposures to PM and gases, and deaths due to injuries. To measure occupational deaths associated with plastics production, we assume deaths are related to the proportion of the workforce in plastics production. We multiply total worker fatalities in a category by the ratio of workers involved in making plastics to the size of the country’s workforce [1370]. Estimates of worker fatalities in plastics production in 2015 due to exposure to benzene, formaldehyde, injuries, PM, and gases are presented in Table 5.3. The largest category of fatalities are deaths due to PM and gases (18,713 globally in 2015), followed by deaths due to injuries [10,410]. Occupational exposures to benzene were responsible for an estimated 1,705 deaths, and formaldehyde exposures for 1,030 deaths. The distribution of total worker deaths by region/country generally parallels the distribution of plastics workers in Table 5.1, but there are differences that reflect differences in workplace safety across countries and differences in the proportion of the workforce employed in plastics production. **Table 5.3** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Unnamed: 5 | Unnamed: 6 | | --- | --- | --- | --- | --- | --- | --- | | | BENZENE | FORMALDEHYDE | INJURIES | PM AND GASES | TOTAL | TOTAL (UNIT: %) | | EU | 139 | 18 | 209 | 498 | 864 | 2.71% | | US | 96 | 12 | 114 | 261 | 483 | 1.52% | | China | 383 | 398 | 3939 | 10985 | 15705 | 49.30% | | RoW Asia | 526 | 378 | 4356 | 5876 | 11136 | 34.95% | | RoW Americas | 249 | 90 | 601 | 421 | 1362 | 4.27% | | RoW Europe | 77 | 19 | 400 | 380 | 876 | 2.75% | | RoW Africa | 143 | 85 | 713 | 258 | 1200 | 3.77% | | RoW Middle East | 91 | 31 | 77 | 34 | 233 | 0.73% | | Total | 1705 | 1030 | 10410 | 18713 | 31857 | 100.00% | It is important to emphasize that the data presented in Table 5.3 are almost certainly underestimates of the full burden of occupational deaths associated with plastics production. These estimates do not reflect deaths that occur after long latency periods, such as cancer deaths due to vinyl chloride monomer exposure, or cancers of long latency associated with carcinogen exposures in the extraction of fossil fuels. ## Health and other impacts of plastic-related GHG emissions In 2015, plastics production created 1.96 Gt of CO2 and other GHGs [13]—almost 2 Gt of CO2e. While representing only around $3.7\%$ of total GHG emissions in 2015, it is not unreasonable to assume that a similar proportion of the GHG effects on climate and human health could be attributed to production of plastics. The health impacts of GHG emissions, which occur through their effects on the earth’s climate now and in the future, are numerous. They include impacts of temperature changes on mortality and morbidity due to cardiorespiratory disease (fewer hospitalizations or deaths in the winter, and more in the summer); impacts of changes in temperature and precipitation on disease and death from vector-borne diseases; effects of climate on health mediated through floods, droughts and food insecurity [13711372]. Among the best studied are the impacts of temperature on mortality [137313741375]. Estimating the health impacts of current CO2 emissions requires estimating their future impacts on the climate and the effects of climate changes on health. Because of the long residence times of CO2 and other GHGs in the atmosphere, both estimates require predicting what the world will look like—in terms of population, gross domestic product (GDP), and GHG emissions—over the span of many decades. Recent estimates of the social cost of carbon (SCC) [13761377] address this complexity. They reflect the anticipated impacts of CO2 emissions on changes in temperature and the impacts of temperature changes on mortality throughout the world from the present to the year 2300. These estimates suggest that increases in deaths during the summer will outweigh reductions in deaths during the winter. The SCC measures not only the net impacts of temperature on mortality—it also includes estimates of the net impacts of temperature on agriculture, energy consumption and damages due to sea-level rise. In 2022, the US EPA [1376] released estimates of the SCC that include the present value of impacts for multiple categories of damages based on Rennert et al. [ 1377], the Climate Impact Lab [1378], and Howard and Sterner [1379]. The EPA’s estimate of the SCC corresponding to one ton of CO2 released in 2020 is $190 (2020 USD). Using 2020 United States dollars (USD), the US Environmental Protection Agency estimates that each ton of CO2 released to the environment results in damages to human health and well-being (the social costs of carbon, SCC) that cause economic losses of $190 (USD). ## Valuation of premature mortality associated with plastics production Deaths associated with plastics production can be valued using the human capital approach—the value of output lost when a person dies prematurely—or using the Value per Statistical Life (VSL)—the amount that individuals will pay for small reductions in risk of death that together sum to one statistical life. We follow the World Bank [13801381] in valuing deaths associated with air pollution—in this case, deaths from plastics production—using the VSL. We refer to these as the “welfare costs” of premature deaths. Because the deaths associated with plastics production occur throughout the world, we transfer estimates of the VSL from the OECD to individual countries using per capita Gross National Income (GNI) and an income elasticity of the VSL equal to one [13801382]. This is equivalent to using a VSL equal to 100 times per capita GNI of whichever country it is applied to. We express all damages in 2015 international dollars (Int$).2 The global welfare costs of the premature deaths associated with plastics production in 2015 were over $250 billion (2015 Int$)3 (Table 5.4). The welfare costs of ambient PM2.5 deaths (212 billion 2015 Int$) were 5.4 times greater than the welfare costs of occupational deaths (39 billion 2015 Int$), reflecting the fact that ambient PM2.5 deaths were five times as numerous as occupational deaths among workers in plastics production. Sixty-four percent of welfare costs occurred in Asia ($35\%$ in China). The EU and USA accounted for $20\%$ of these costs. **Table 5.4** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | | --- | --- | --- | --- | --- | | | VALUE OF PM DEATHS(IN 2015 PPP BILLION) | VALUE OF OCCUPATIONALDEATHS (IN 2015 PPP BILLION) | TOTAL VALUE (2015 PPP BILLION) | TOTAL (UNIT: %) | | EU | 24.42 | 3.11 | 27.54 | 10.98% | | US | 19.13 | 2.61 | 21.74 | 8.67% | | China | 66.30 | 21.29 | 87.59 | 34.94% | | RoW Asia | 64.94 | 8.03 | 72.98 | 29.11% | | RoW Americas | 8.76 | 1.60 | 10.36 | 4.13% | | RoW Europe | 8.15 | 1.32 | 9.47 | 3.78% | | RoW Africa | 8.22 | 0.47 | 8.69 | 3.47% | | RoW Middle East | 11.83 | 0.48 | 12.31 | 4.91% | | Total | 211.75 | 38.92 | 250.67 | 100.00% | The welfare cost of the 1.96 Gt of CO2 emissions from plastics production in 2015 is $341 billion (2015 Int$), using the US EPA’s estimate of the SCC. ## Valuation of health impacts of plastics use Bisphenols, phthalates, brominated compounds, and other EDCs and neurotoxic chemicals are used in very large quantities in plastics production. These chemicals can leach out of plastics and can enter the human body through ingestion, inhalation, or dermal absorption. Bisphenols such as BPA are found in food packaging. PBDE is a flame retardant used in furniture and in children’s clothing that can be ingested from contaminated dust. DEHP, a phthalate, is used in industrial food processing and may be ingested in processed food. All of these chemicals have been found through epidemiological and toxicological studies (reviewed in Section 4) to be endocrine disruptors. A growing epidemiological literature links endocrine disruptors to adverse reproductive outcomes, neurological development, morbidity, and mortality. Extrapolating these results to the general population requires estimates of population exposure to these chemicals. In the USA, the NHANES provides these estimates for some endocrine disruptors. Comparable data are available for Canada, and for some countries in the EU, but are not readily available for India or China. We therefore illustrate the impacts of three endocrine disruptors—PBDE, BPA, and DEHP—on health outcomes in Canada, the EU, and the US. ## Impacts of PBDE on intellectual development Longitudinal studies have linked PBDE levels in the blood of pregnant women to decreases in their children’s IQ [111011111123], finding that higher maternal levels of PBDE are associated with lower IQ scores in children, and negatively associated also with other neurodevelopmental and behavioral indices. Chen et al., Eskenazi et al., and Herbstman et al. [ 111011111123] all find statistically significant negative relationships between levels of PBDE congeners (BDE-47, BDE-100, and BDE-153) in mothers’ blood during pregnancy (indicating prenatal fetal exposure) and children’s IQ. In Chen et al. [ 1123], a one-unit increase in log10 BDE-47 is associated with a 4.5-point reduction in IQ at age five (CI = 0.1–8.8). The results of these longitudinal studies have been used to estimate IQ losses in the 2010 birth cohorts in the US, the EU, and Canada (see Table 5.5). Attina et al. [ 1384] apply these results to the 2010 birth cohort in the US using the distribution of BDE-47 in the female population of child-bearing age from NHANES. They estimate that PBDE exposure resulted in a loss of almost 10 million IQ points based on Chen et al. [ 1123]. This number doubles using the dose-response results in Herbstman et al. [ 1110]. Bellanger et al. [ 1385] perform a similar analysis using estimates of PBDE in cord blood for the EU. PBDE concentrations are much lower in the EU and Canada than in the US due to bans on the use of PBDE in those countries, and result therefore in much smaller IQ losses than in the USA.4 In Canada, estimated IQ losses range from 347,000 to 927,000 points [1386]. **Table 5.5** | Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Unnamed: 5 | Unnamed: 6 | Unnamed: 7 | | --- | --- | --- | --- | --- | --- | --- | --- | | GEO GROUP | SCENARIO | IQ POINTS LOST | LOST ECONOMIC PRODUCTIVITY (IN 2015 PPP BILLION $) | ATTRIBUTABLEINTELLECTUAL DISABILITY | COST OF INTELLECTUAL DISABILITY (IN 2015 PPP BILLION $) | TOTAL COSTS (IN 2015 PPP BILLION $) | TOTAL COSTS (IN % OF GDP) | | EU | PBDE (base) | 837685 | 8.30 | 3156 | 2.75 | 11 | 0.06% | | EU | PBDE (high) | 1926158 | 19.09 | 7705 | 6.72 | 26 | 0.13% | | USA | PBDE (base) | 9818493 | 145.77 | 43268 | 56.51 | 202 | 1.11% | | USA | PBDE (high) | 19035078 | 282.61 | 98769 | 129.00 | 412 | 2.26% | | Canada | PBDE (base) | 373628 | 4.19 | 1607 | 1.59 | 6 | 0.36% | | Canada | PBDE (high) | 926055 | 10.40 | 4493 | 4.44 | 15 | 0.93% | Intellectual disability is defined by IQ levels below 70 points. Estimates of numbers of cases of intellectual disability associated with PBDE appear in Table 5.5. These range in number from over 43,000 in the US (base case) to 3,200 in the EU (base case) and 1,600 in Canada (base case). We estimate the economic cost of lost IQ points in terms of foregone earnings over the lifespan of exposed children, and we estimate the cost of cases of intellectual disability using medical costs. There is a large literature linking IQ to earnings both directly and indirectly—i.e., by affecting educational attainment and labor force participation. We assume that one IQ point reduces the present value of lifetime earnings by $1.1\%$ [1383], noting that recent studies have used values between $0.9\%$ [1387] and $1.4\%$ [1388]. Our estimate of lifetime earnings (Grosse, Krueger, and Mvundura [1383] include both market earnings and non-market output.) *Using a* discount rate of $3\%$ and discounting lifetime earnings to birth yields a loss per IQ point of $14,847 2015 Int$ for the 2010 birth cohort in the US. This value is transferred to each country in the EU and to Canada based on the ratio of per capita GNI (2015 Int$) in the transfer country to per capita GNI (2015 Int$) in the US. The present value of lost productivity associated with PBDE exposure for the 2010 birth cohort (base case) is $145 billion (2015 Int$) for the US, $8.3 billion (2015 Int$) for the EU and $4.2 billion (2015 Int$) for Canada (See Table 5.5) These numbers double using higher published estimates of IQ loss. We use estimates of the present value of the medical (direct and indirect) costs of intellectual disability for the US from Honeycutt et al. [ 1389], which we transfer to other countries using the ratio of per capita GNI (2015 Int$) in the transfer country to per capita GNI (2015 Int$) in the US. Adding these costs to IQ productivity losses yields economic costs of PBDE exposure to the 2010 birth cohort of $202 billion (2015 Int$) in the US—or $1.1\%$ of GDP. Costs are significantly lower in Canada ($0.36\%$ of GDP) and the EU ($0.06\%$ of GDP), reflecting the lower levels of PBDE exposure in women in those countries. ## Impacts of BPA on cardiovascular disease Epidemiological studies over the past decade have explored associations between BPA and cardiovascular disease [1050139013911392]. Three of these studies use data from multiple waves of the NHANES survey to link urinary BPA concentrations to various cardiovascular outcomes: coronary heart disease (CHD), ischemic heart disease, congestive heart failure and stroke. Cai et al. [ 1050], using a sample of 9,139 adults from the 2003–2014 waves of NHANES, find that the natural logarithm of BPA level (ng/mL) (ln BPA) in American adults is associated with increased prevalence of any cardiovascular diseases with an OR of 1.16 (CI: 1.04–1.29). Examining specific causes of deaths, ORs are 1.19 (CI: 1.02,1.39) for CHD and 1.20 (CI: 1.05, 1.37) for stroke. Moon et al. [ 1392] use waves 2003 through 2016 of NHANES to estimate impacts of ln BPA (ng/mL) on cardiovascular outcomes using Propensity Score Matching. They estimate ORs of 1.17 (CI: 1.08–1.27) for any cardiovascular diseases; 1.13 (CI: 1.04–1.24) for ischemic heart disease; and 1.13 (CI:1.01–1.26) for stroke. We use the Propensity Score Matching results from Moon et al. [ 1392] together with the distribution of urinary BPA levels in the adult US population in the 2003–2016 waves of NHANES to calculate the fraction of CHD and stroke attributable to BPA in the US. We conservatively apply an OR of 1.13 to CHD and an OR of 1.13 to stroke. Our calculations indicate that $7.64\%$ of the prevalence of each outcome is attributable to BPA exposure. Based on the prevalence of CHD (affecting 20.1 million adults in 2020) [1393] and stroke (affecting 795,000 adults in 2020) [1394], we estimate that 1.54 million cases of CHD and 60,738 cases of stroke in the US in 2020 were attributable to the plastic additive BPA. We value the economic costs of these cases using estimates of the direct costs of illness from the American Heart Association [1395], and we value mortality attributable to BPA using the VSL used in previous analyses. Converting the medical costs of a case of CHD and a case of stroke to 2015 Int$ yields annual costs of $7.85 billion (2015 Int$) for cases of cardiovascular disease and $2.41 billion (2015 Int$) for cases of stroke associated with BPA. We value the 29,247 CHD deaths and 10,634 stroke deaths attributable to BPA exposure using a VSL of 5.4 million 2015 Int$. In the US, the total costs attributable to BPA are $165.9 billion 2015 Int$ for CHD and $62.4 billion 2015 Int$ for stroke. ## Impacts of phthalates on mortality Several recent studies in the US have investigated the relationship between phthalates and mortality [12711396]. Trasande, Bao, and Liu [1052] estimate the impact of phthalate metabolites, focusing on high-molecular weight metabolites and DEHP, on all-cause and cardiovascular mortality in the US. High-molecular weight phthalates are used in food wraps and intravenous tubing; DEHP is used in industrial food processing. Using data from the 2001–2010 waves of NHANES, the authors associate urinary phthalates with mortality data through 2015. Specifically, they estimate Cox proportional hazard models relating the natural logarithm of urinary phthalate concentrations to cardiovascular and all-cause mortality for 55- to 64-year-olds. They find that the natural logarithm of high-molecular weight phthalates is significantly related to all-cause mortality (OR = 1.14; CI: 1.06–1.23), as is the logarithm of DEHP (OR: 1.10 CI: 1.03–1.19). Applying the DEHP results to the 2013 cohort of 55- to 64-year-olds yields an estimated 90,762 deaths attributable to DEHP levels in 2013 or approximately $5.2\%$ of all deaths in that year. We value the 90,762 deaths using the same VSL as in previous analyses—5.4 million (2015 Int$). This implies a welfare cost of $490 billion (2015 Int$) for deaths attributable to DEHP exposure in the US in 2013. ## Box 6.1 Definition of Social and Environmental Justice (SEJ). Social justice strives to fairly distribute resources, opportunities, and privileges in society, regardless of an individual’s background and status [1399]. Environmental justice is defined as follows: SEJ are inherently linked because issues in the environment have direct impacts on certain people and groups, including those living in or directly dependent on particular environments. When people do not have equitable access to resources that could relieve environmental stressors, it becomes a social justice issue. Both SEJ necessitate the following: SEJ addresses both “procedural” injustice, the unequal access to information and role in decision-making, and “distributive” injustice, which is the unequal distribution of burdens on certain groups and people [1402]. ## Social and Environmental Justice Social and environmental injustice flow from a belief, common in much of modern society, that “natural and human resources exist for exploitation, commodification and control, and to fuel economic growth.” [ 1403] Such beliefs and narratives are created and perpetuated by economic, political, and social groups “to concentrate power and wealth, which necessarily requires oppression of the masses and the marginalized.” [ 1403] They are grounded in the concept that some lives are more important than others. Those beliefs take no cognizance of the view that *Earth is* a shared inheritance, a “Common Home” whose “fruits are meant to benefit everyone.” [ 46] The concepts of SEJ relate not only to the negative consequences of pollution but also to the exclusion of certain groups from participating “meaningfully in the leadership and composition of the environmental movement and related decision-making processes.” [ 1402] To reverse such injustice, SEJ focuses on the equitable distribution of both the environmental resources and burdens so that no one group of people bears a disproportionate share of the negative consequences resulting from industrial operations and/or government policies. SEJ may also consider equity and the rights of the environment [1404]. Until recently, less attention has been given to environmental justice in marine and coastal environments, where environmental harms, such as plastic and chemical pollution, are worsening and exceeding planetary boundaries through pollution and toxic waste, plastics and marine debris, climate change, fishery declines, ecosystem degradation, and biodiversity loss [14051406]. These concerns are now being framed as “ocean equity,” “ocean justice,” and “blue justice” and include the impacts of pollution and climate. The focus of this emerging area is the need to address equity and justice in ocean governance and management [14051406]. It is essential that any global agreement on plastic pollution, such as the Global Plastics Treaty [1407], address SEJ and that it include remedies to reduce the disproportionate impacts of plastics and related issues of climate change on coastal and ocean-dependent communities. These points are noted in Case Study 6.1, which focuses on the disproportionate impacts of plastic pollution and climate change on small island states. If solutions to the plastic crisis and the climate crisis are to be sustainable, they must benefit people and advance SEJ. ## Case Study 6.1 The Intersection of Climate Change and Plastics Pollution on Small Islands: An Amplifier of Social and Environmental Injustice. The intersection of plastics and climate change issues on small islands amplifies social and environmental injustice through complex linkages. Islands are on the front lines of climate change, experiencing extreme weather events, sea-level rise, brackish water inundation of freshwater sources, and other existential stressors [140814091410]. Island nations and states also experience substantial amounts of plastic washing up on their shores, including swaths of microplastics (MPs) [1410]. This pollution has negative implications for aesthetics, fisheries, and water quality. At the same time, island nations are most vulnerable to sea-level rise and will have to plan for managed retreat because of global climate change [141214131414]. Small island states contribute little to anthropogenic climate change yet are major recipients of its effects. The least developed countries contribute about $1\%$ of carbon dioxide (CO2) emissions [1415], and similarly, the small islands classified as “small island developing states” by the United Nations (a mix of least developed countries and middle-income countries) contribute less than $1\%$ of such emissions [141614171418]. In contrast, global plastics production, use, and disposal are sizable contributors to climate-driving greenhouse gas (GHG) emissions. Plastic production relies almost exclusively on fossil fuel feedstocks, and by 2050, it is forecast to account for $20\%$ of total oil consumption [73]; plastics-associated GHG emissions could reach $15\%$ of the global carbon budget that same year [7483]. Plastics products themselves emit GHG when exposed to sunlight [485], emphasizing the need for a life cycle approach to solutions [23]. Outcomes of plastics production, use, and disposal are growing threats to island communities, including to human health, at every stage of the plastics life cycle [91419]. Small Islands worldwide are disproportionately affected by plastics [14191420]. This inequitable burden results from a confluence of circumstances [1411]. For example, small islands serve as “strainers” of distal plastics concentrated and transported by ocean currents [1419]. These islands also have insufficient waste infrastructure to manage plastic waste generated locally [14211422] and experience plastic debris from all fisheries sectors (commercial, recreational/touristic, subsistence fishing, and aquaculture) [14111423]. The cultural and economic reliance on seafood in Oceania, including by marginalized or Indigenous communities, also poses a threat from MP pollution in seafood, particularly where dietary alternatives are not readily accessible [231424]. The intersection of climate change and plastic pollution is amplifying social and environmental injustice on small islands. Developing, implementing, and refining solutions to decrease social and environmental injustice on small islands can provide scalable solutions applicable to continental “islands” and ultimately our planet “island”—Earth. Such solutions are urgently needed by humanity to arrest the existential threat of climate change and address a less commonly recognized driver of GHG emissions: plastics production, use, and disposal. ## Plastic Pollution and SEJ Plastic-related pollutants intersect with SEJ across human health, environmental health, economics, and human rights [1402]. The nexus between these areas and SEJ is complex and tangled. As noted in Sections 2 and 3, plastic-related pollutants take many forms and are generated across the plastics life cycle, from fossil fuel extraction—the feedstock for virtually all plastic—production and manufacturing, to product use, and finally to intentional and unintentional dumping, littering, or unregulated disposal or discharge; managed disposal via landfills, incinerators, or chemical recycling facilities; and leakage from managed systems [2364135514251426] (see Figure 6.3). **Figure 6.3:** *Impacts of plastic to vulnerable populations. As depicted in this figure from (UNEP and Azul, 2021), vulnerable groups and populations are adversely affected by plastic pollution (which includes intentional and unintentional leakage of plastic and chemical additives to the environment), throughout the entire life cycle of plastics, beginning with extraction and production, through its market penetration and uses, to plastic waste management and disposal.Original source: (UNEP and Azul, 2021).* The adverse effects of plastic pollution and climate change are widespread but currently felt most keenly in certain geographies and among certain groups and populations least responsible for the pollution who lack the power or resources needed to address the problem [14021427]. As a result, plastic pollution has been described as a new form of “colonialism.” [ 1428] Plastic pollution creates dire inequities for people, the environment, and the countries and prevents individuals and societies from flourishing [1427]. Environmental injustices occur at local, national, and global scales, including injustices occurring between the regions of the Global North and Global South: “Global North nations continue dumping waste in both domestic and global ‘pollution havens’ where the cost of doing business is much cheaper, regulation is virtually non-existent, and residents do not hold much formal political power.” [ 1429] ## Understanding the True Costs and Impacts of Plastics Plastics are produced and manufactured inexpensively for the global market with harms and costs unaccounted for (or “externalized”), resulting in the diminished health and well-being of marginalized and vulnerable communities and the degradation of the environments and resources upon which these communities depend. The indirect social and environmental costs of plastics production and manufacturing are thus borne by poor or disempowered communities and peoples [714301431] and are not represented nor included in the costs paid by consumers. The full economic costs associated with plastic-related pollution include impaired ecosystems; reduction of their associated benefits, sometimes referred to as “ecosystem services”; and the still incompletely quantified costs of impacts on human health as noted by this Commission [133114321433]. These costs (referred to as “externalities”) are excluded from the plastic sector’s current accounting or responsibilities and shifted onto governments, taxpayers, and citizens without compensation; thus, they are disproportionately experienced by marginalized and vulnerable communities [1402143414351436]. For example, communities in low-lying island nations are experiencing substantial amounts of plastic washing up on their shores, including swaths of MPs [1411]. See Case Study 6.1 for more information. This has negative implications for aesthetics, fisheries, and water quality—all of which are sensitive to environmental change. Concurrently, island nations are particularly vulnerable to sea-level rise impacts that result from global warming and will have to plan for managed retreat from the lands where they have lived for centuries [141214131414]. Advancing SEJ will require a reckoning by the plastics sector and by consumers of these externalities (both from climate change and plastic pollution) across the plastics life cycle. Investigations of plastic-related pollution and social and environmental injustice are increasing, including the documentation of harmful emissions associated with plastics production and disposal. Growing legal and financial risks associated with the generation of plastic pollution may provide further impetus for change, particularly as public awareness and oppositiongrows [1398]. Section 4 of this Commission, addressing human health, underscores the need to reduce production of plastic at its source to protect human health and well-being and identifies the health consequences of plastic on the poor, minorities, the marginalized, and the people of the Global South. The fact that plastic pollution is also pervasive in high-income countries despite adequate and widely accessible waste management infrastructure underscores that the countries of the Global South cannot solely resolve the problems of plastic pollution through improved plastic waste management [5401437]. While increasing waste management may lessen the impacts of the problem, it alone cannot solve it. Indeed, as Sections 2 and 4 emphasize, plastic is not just a waste issue but also a health issue across the plastics life cycle, which includes exposure from everyday use. Understanding the SEJ implications of plastic-related materials and pollutants requires isolating how each stage of the plastics life cycle affects vulnerable people and groups. Here, we introduce and explore these impacts with the aim of elevating SEJ in discussions and solutions surrounding each stage of the plastics life cycle. Addressing both the urgency and complexity of these problems at all scales requires a specific set of solutions that maintains focus on the disadvantaged, marginalized, and excluded and supports the empowerment of those who suffer inequities. Specific health and environmental risks associated with each of these phases is presented in greater detail in Sections 2 (The Plastic Life Cycle and Its Hazards to Human Health) and 4 (The Impact of Plastics on Human Health). ## SEJ issues associated with extraction of raw materials Exploration and extraction of materials used to create plastics have disproportionately harmful effects on marginalized and vulnerable communities at scales from local to global [2314011435]. These range from impacts of oil extraction on Indigenous peoples reliant on vulnerable and threatened natural resources for survival [78] to environmental and health effects of pollution from the extraction and processing of petroleum products on communities sited near chemical facilities, also known as “fenceline” communities [1430] or “sacrifice zones.” [ 14381439] Specific activities associated with extraction include mining, fracking, and drilling for the oil, coal, and gas used to formulate essentially all (about $98\%$) plastics produced today [23]. For example, Section 2 details the contaminated water impacts in Nigeria in the *Niger delta* from oil pollution and gas flaring activities. The siting of these facilities and related exposures from activities conducted at these facilities affect many people and communities already at risk and are exacerbated in locations with absent or weak laws or enforcement of environmental or health protections. Many such locations are found in the Global South and on Indigenous lands, such as South Sudan and the Amazon basin. However, poor and vulnerable locations in the Global North are not immune, such as the infamous “Cancer Alley” in the US Gulf region, which accounts for about $25\%$ of US petrochemical production [1402](see Box 6.2). This harm directly affects the health of individuals and exposed groups and drives the destruction of habitats and the biodiversity needed for the survival of groups and populations. Specific drivers of these impacts include illegal or uncontrolled discharge of oil and chemicals, construction and related activities that destroy or disturb habitats and species, contamination of soil and water sources, and equipment failures and dangerous working environments for those employed in such activities. These same affected groups also commonly suffer from procedural inequities, lacking access to guidance or information while having little to no standing in decision-making; this is especially true of Indigenous groups [14021440]. ## SEJ issues associated with production and manufacturing Plastics production and manufacturing are fraught with SEJ issues. The disproportionate siting of production and manufacturing facilities as well as of pipelines and compressor stations in minority and disadvantaged communities is at the literal forefront of these issues, with repercussions that cascade across human health, economics, and human rights [23140214301439]. Poorer communities are commonly selected as sites for plastic manufacturing [14021437]. Environmental burdens resulting from plastic production and manufacturing are unjustly borne by marginalized and vulnerable communities worldwide across the Global North and South [14021435]. “ Fenceline” communities and housing (in proximity or immediately adjacent to production and manufacturing plants) are often touted as providing affordable housing and the promise of local employment, while failing to acknowledge the risks associated with exposure to documented harmful emissions [91430]. Adverse health impacts in “fenceline” communities can compound existing inequities, including in education for individuals with few to no economic options to move or exert political influence [1439]. Chemical additives, most of them petrochemicals, used in the production and manufacture of plastics are major drivers of exposure risk [731431]. Nearly all plastics are derived from fossil fuels, and when plastics production facilities and fossil fuel refineries are concentrated geographically for manufacturing efficiencies, it amplifies SEJ inequities [9]. See Box 6.2 formore information. Concentrations, or “hotspots,” of plastics production plants and refineries are of SEJ concern worldwide [1402]. The greatest sources of plastic production, and foci of many studies, is the Global North [36314341436]. In 2020, while Asia led the world in global plastics production at $52\%$, the three largest global producers were China at $32\%$, North America at $19\%$, and Europe at $15\%$ [63]. The remainder of Asia, excluding Japan, produced $17\%$, the Middle East $7\%$, and Latin America $4\%$, with the Commonwealth of Independent States and Japan each at $3\%$ of production [63]. ## Box 6.2 Plastic Pollution Hotspots. One plastic pollution hotspot is the infamous “Cancer Alley” area along the southern Mississippi River bracketed by New Orleans and Baton Rouge, Louisiana [1436]. This largely minority black region experiences the highest cancer rates in the US [14341441]. In one community, resident cancer rates are overwhelmingly higher than the national average, putatively attributed to toxic emissions from a nearby synthetic neoprene factory, among other petrochemical producers [14341442]. These disproportionate health and exposure impacts were specifically cited in a 2022 judicial decision denying a permit for a plastic production facility in the Cancer Alley parish of St. James, Louisiana [1440]. These hotspots generate exposure to both nearby residents and workers at these facilities. As an example, in the same region, which is also repeatedly impacted by hurricanes and other disasters, the National Institute of Environmental Health Science Gulf Long-Term Follow-Up Study (GuLF Study) followed a cohort of 32,608 adults involved in oil spill response and cleanup following the 2010 Deepwater Horizon oil spill disaster. The study found strong associations between exposure to natural and other hazards and mental health impacts (perceived stress, distress, depression, anxiety, and post-traumatic stress disorder [PTSD]) [1443]. ## SEJ issues associated with plastics use and market penetration Plastics are inexpensive to produce and highly accessible, with broad use and market penetration, but this convenience comes at a cost that industry neither bears nor calculates. Substantial uncounted indirect costs (“negative externalities”) associated with the everyday use of plastics, particularly associated with chemical exposures from plastic, are borne by vulnerable communities and the environment [2314311444]. Although plastic use is prevalent among lower-income communities, gross domestic product (GDP) is a key driver of plastics use globally, with plastic use rising as nations increase their economic capacity and attainment. For example, an individual in the US uses 255 kg of new plastic every year on average, while the average person in sub-Saharan Africa uses less than one-tenth that amount [5]. As Africa’s population and economy grow and transform over the coming century, its plastic consumption is expected to increase exponentially [141444]. Considering the negative health and environmental impacts associated with all stages of the plastics life cycle, the correlation of increasing GDP and plastic use warrants attention on the global stage. Examples of uses and market penetration that increase SEJ concerns surrounding plastics and related chemical pollution include plastic water bottles, plastic packaging in general, and products and plastic packaging from fast-food restaurants and discount stores. Plastic water bottles are used worldwide to provide much-needed drinking water. Globally, around two billion people lack access to safely managed drinking water at home [1445]. In some cases, local governments fail to deliver potable water to their constituents and instead rely on bottled water as the primary source of water consumption [1402]. This increases the risk of MP and plastic additive exposure in vulnerable communities because toxic additives are leachable and migrate out of products (Section 2). One study found that of 259 total bottles processed from nine different countries, $93\%$ showed some sign of MP contamination of the water [1212]. The use of plastic-bottled water in response to natural hazards, both in the Global North and South, will likely rise given the increasing frequency and intensity of storms and associated hazards(see Figure 6.4). **Figure 6.4:** *Bottled drinking water to support Hurricane Katrina personnel in New Orleans, Louisiana.Credit: MSGT Michael E. Best, USAF.* As discussed in Section 4, use of plastics in packaging can expose users to a range of harmful chemicals, including bisphenol A (BPA), phthalates, and PFAS, all of which are used in common items such as bottles and processed food containers [14441446]. These chemicals can interfere with hormonal systems, damage children’s developing brains, and increase the risk of cancer [1431]. Women have an additional risk of exposure through feminine hygiene products and household items, exacerbating existing gender-related inequalities [1402]. A growing body of evidence shows that women experience reproductive disorders (including infertility and miscarriage) when frequently exposed to such chemicals [14021446]. Consumption of fast foods is another source of disproportionate use and exposure to harmful chemicals in vulnerable populations. Fast food served at high temperatures in plastic packaging enables harmful chemicals to migrate into the food [1431]. One study found that many popular fast foods contain an abundance of ortho-phthalates; these plasticizer chemicals are established endocrine disruptors that can increase risk of diabetes, obesity, and cardiovascular disease and also impair fertility [1447]. Individuals who frequently consume high-fat, high-salt fast-food meals are especially vulnerable to such plasticizer exposures, including lower-income minority communities living in food deserts with limited healthy food options [14471448]. Market penetration of plastic products is high in vulnerable and poorer communities because of business strategies and investments. For example, in the US, “dollar stores” are more likely to be found in lower-income neighborhoods, where families seek to maximize their budget and where offerings emphasize the perceived value and convenience of plastic. However, most of the merchandise is low quality and may contain toxic chemicals, such as high levels of lead and other endocrine disruptors [1402]. Despite these challenges, the widespread use and market penetration of plastics can also provide an opening for the use of consumer purchasing power to drive market innovation. By investing in reusable products and refusing single-use plastic options, consumers can collectively advocate for a more circular economy [1402]. Unfortunately, lower-income groups have less opportunity and fewer sustainable options, and even if available, they cannot always afford the better option, undermining their ability to drive market change [1402]. As societies transition toward circularity, it will be important for sustainable living practices and necessary resources to be widely accessible across economic and social conditions. Accessibility, and other SEJ parameters, need to be explicitly incorporated in solutions, as noted below. ## SEJ issues associated with plastics waste: Leakage and management, including disposal, exports, and incineration Waste disposal, management, and leakage and plastic waste exports have a wide range of harmful impacts, particularly on vulnerable and poorer populations. Lower-income countries, such as countries in South and Southeast Asia, bear the brunt of plastic waste mismanagement, while higher-income countries are responsible for most of the plastic production and use [23244071449]. In the Global North, plastic waste may be incinerated, recycled, sent to managed landfills, or exported to countries without adequate waste management, but in the Global South, access to organized waste management systems is indeed limited. Many cities and governments rely on incineration, or plastic is dumped into uncontrolled landfills that may leak into surrounding environments [4923]. Large amounts of toxic ash by-product from incineration are another critical concern, as a large body of evidence links incineration by-products to cardiovascular diseases, cancers, stroke, and respiratory and other illnesses [9]. The piling and burning of plastic waste in dumps, as well as in towns and villages, seriously affects people living in “fenceline” communities, as particulate matter (PM), PCDDs, and PCDFs and toxic gases are released into the air as plastic burns [143814491450]. In addition, in poorer communities, plastics may be burned as fuel (heat or cooking) or as a means of disposal [1432]. The implications for human health from these practices are difficult to document, given the lack of data on exposure of the most vulnerable individuals to this range of pollutants, as described in Sections 2 and 4. Additional, and often unforeseen, consequences of plastic pollution on these communities can include flood risks from drainage systems clogged with solid waste that ends up in waterways, such as seen in Figure 6.5; increases in vector-borne diseases; and a reduction in tourism due to degraded environments with plastic and other wastes, which is exacerbated in poorer communities without adequate solid waste management systems [24407133214321451]. **Figure 6.5:** *Single use plastic waste clogging open drains in Makoko, Lagos, Nigeria.Credit: Adetoun Mustapha and Korede Out.* People who work in recycling and with recycled plastic (see health risks of recycling in Sections 2 and 4) as well as in the informal waste sector, such as waste pickers, are especially vulnerable to the impacts of plastic pollution (Box 6.3), and the informal sector is growing, influenced by climate change impacts and urbanization [1452]. In developing countries, informal waste pickers perform much of the vital role of reducing the large amount of waste (especially plastic waste) in landfills and open dumpsites, where open burning often takes place [1452] (see Figure 6.6). Informal waste-picking activities also provide important social benefits, serving as opportunities for people who have few or no marketable skills and education and no alternative sources of income to survive. However, as described in Sections 2 and 4, these workers, many of whom are young children and pregnant women, are heavily exposed to harmful chemicals and contaminants through inhalation and ingestion during waste picking [373]. **Figure 6.6:** *Female workers sort out plastic bottles for recycling in a factory in Dhaka, Bangladesh.Credit: Abir Abdullah/Climate Visuals Countdown.* Trade in waste also creates SEJ issues. Wealthy countries that produce more waste than they can recycle at home ship much of that plastic waste to low-income and middle-income countries (LMICs) for legal or illegal disposal or putative recycling [1449]. In addition, the island nation of Indonesia and small countries like Vietnam and Malaysia that accept these materials and process them for a fee do not have systems to manage all the plastic waste they import, let alone the ability to manage the river- and ocean-borne plastic pollution that reaches their shores and coastal waters from distal and local sources [23]. Leaked plastic waste and its toxic chemical additives can be transferred through the food chain and across trophic levels [23]. Exposure to MPs via seafood is likely to be greater for populations that depend on seafood for nutrition [923]. Marine plastic pollution also causes negative impacts on the fisheries sector, including reduced revenues or increased costs [1462]. Vulnerable communities who rely on fisheries as a main source of income or for subsistence may be particularly affected (Case Study 6.1). ## Box 6.3 Gender Injustice and Impacts on Women Waste Pickers. Of the estimated 20 million waste pickers worldwide, the majority are women from socially and ethnically marginalized communities [145314541455]. Unjustly, women waste pickers are often “invisible” or disrespected in their societies; they work long hours in unhealthy conditions and earn lower wages compared to men [145414561457]. There is limited attention paid to the occupational health issues and social harms experienced by women waste pickers, but a few key studies from around the world report significant impacts on women as a result of consistent exposure to toxic plastic and electronic waste (e-waste), many of which contain known endocrine disruptors [1458145914601461]. ## Embedding SEJ in Solutions to the Plastics Problem Decision makers need to prioritize and establish policies in national and international action plans to address the long-neglected and disproportionate environmental and health impacts of plastic pollution on poor, marginalized, and voiceless populations and groups [1463]. To date, the failure to do so has been due to the long latency of many pollution-related health impacts, insufficient information about pollution’s enormous economic and social costs, the vested interests of large industry, and the belief—widely held but thoroughly discredited—that pollution is an unavoidable consequence of economic development [1380]. Policies and actions to reduce plastic waste will also need to address the important social, environmental, and economic roles of high-risk groups, including workers through formal recognition of their existing roles in the waste sector and actions to reduce both worker health risks and poverty [1452]. Finally, solutions need to recognize and reduce the full cost (including “externalities” not currently considered in decision-making) associated with plastic pollution, especially to vulnerable or marginalized groups, and the environment, upon which we all ultimately depend. Solutions will need to encompass mechanisms to resolve both distributive and procedural inequity in each stage of the plastics life cycle. This includes addressing power inequities, reducing disproportionate burdens on vulnerable and disadvantaged communities (including ensuring accessibility of those solutions), and considering the role of human agency and individuals’ choices by promoting empowerment. Principles for addressing the procedural inequities and promotion of empowerment include the following: In alignment with the identified SEJ principles and numerous statements from civil society and worker groups, the Intergovernmental Negotiating Committee (INC) urged delegates to address SEJ in the Global Plastics Treaty at the first meeting (INC-1) [1464]. Additionally, the Office of the High Commissioner on Human Rights’ statement to the INC-1 about key human rights considerations for the Global Plastics Treaty informs some of the critical tasks for embedding SEJ in solutions. Recommendations included promoting the human right to a clean, healthy, and sustainable environment; safeguarding the rights of those who suffer the most from plastic production and pollution; holding businesses accountable to remediating or protecting against human rights harms; and transitioning toward a chemically safe circular economy that address all stages of the plastics life cycle [1465]. The emerging concept of ocean equity similarly asserts the need for attention to many dimensions of equity from the beginning to the end of any decision-making process, as in Figure 6.7 from Bennett [2022]: [1] recognitional equity, [2] procedural equity, [3] management equity, [4] distributional equity, [5] environmental equity, and [6] contextual equity. This framework would also be particularly applicable to the issues associated with plastic pollution even beyond coastal and ocean-dependent communities. **Figure 6.7:** *Ocean equity is comprised of several distinct dimensions. Bennett (2022) describes, and depicts in this figure, a range of equity considerations (which include many aspects of procedural and distributional equity) relevant to those working in coastal and marine conservation. These equity dimensions may also provide a useful framework for embedding social and environmental justice (SEJ) in decisions and processes relevant to plastic production and pollution at the local to global scale – even beyond coastal and marine areas.Original source: (Bennett, 2022).* ## A key inequity: Lack of funding for SEJ research A lack of long-term and sustained funding hampers our understanding and limits development of solutions to plastics-associated SEJ problems. Proportionally less funding for marine conservation, including marine pollution and marine science, goes to developing countries, particularly Africa [1466], which may contribute to limited knowledge regarding social justice–related plastic issues in areas of the Global South. Little research has been conducted on plastic pollution and impacts on human health, ecosystems, economies, and SEJ in LMICs compared to high-income countries. For example, with about 2.5 Mt of plastic waste annually, Nigeria ranks ninth globally among countries with the highest contributions to plastic pollution [24], and more than $88\%$ of the plastic waste generated in *Nigeria is* not recycled but flows to lagoons and the ocean [146714681469]. A systematic review of academic studies on plastic pollution in the environment in Nigeria conducted by Yalwaji et al. [ 2022] [1469] shows that as of May 30, 2021, there were only 26 such studies in Nigeria, compared to 62 peer-reviewed studies on the Arctic Ocean [1470]. Between 1987 and September 2020, there were only 59 studies on plastic pollution in the African aquatic environment [1471]. Due to resource constraints and the lack of research prioritization in LMICs, most LMICs’ governments have failed to allocate meaningful resources for research. Lack of research funding in LMICs has several implications for generating high-quality evidence to inform policy and practice [1472] and impedes achievement of the UN Sustainable Development Goals. Regarding marine-focused funding for conservation priorities such as pollution, grant making has historically allocated a sizable proportion of funding to global initiatives ($40\%$) and work focused on North America ($32\%$). Over the past decade, an increased proportion of marine funding ($15\%$) was allocated to Asia. Funding to Africa remains limited at less than $3\%$ of philanthropic funding [1466]. Currently, much of the literature on environmental justice issues related to plastics focuses only on local to national scales and is US-centric [1402]. Future studies should enhance focus on international case studies to better comprehend the range of environmental justice implications from global dependence on plastics [1402]. A much wider evidence base is required regarding environmental justice and equity. Transdisciplinary research that demonstrates the multifaceted interplay among the plastics life cycle, environment, and society and how the nexus of these elements produces inequities within and between countries is necessary to drive global strategies and policy interventions. Implementation research that demonstrates just and equitable outcomes with the potential to reduce burdens experienced by communities most impacted by marine plastic pollution are also necessary. Research to better understand and quantify the potential health impacts of plastic pollution to humans (see Section 5, Economic Impacts) and aquatic and terrestrial ecosystems is key for science and data-driven solutions. ## Primary SEJ solution: Reduce plastic pollution and health risks at the source and hold producers accountable While it is important to act at all stages of the plastic life cycle and at all scales, from local to global, the economic and health costs of plastic pollution on vulnerable populations require that reducing the sources of plastic pollution and health impacts (including reducing production and addressing the health risks of additives and MPs) be preeminent in the hierarchy of SEJ solutions. This includes ensuring that burdens are removed from those least responsible and ensuring that responsibility lies with those who create and profit from plastic pollution. Addressing pollution and other problems at their source will be not only efficient but also effective, and it will ensure responsibility is borne by producers, or “perpetrators.” [ 14061473] Expert reports have shown that “reducing plastics production, consumption and improving waste management to turn off the tap of plastics pollution is easier than attempting to clean up ocean plastics,” as well as being more cost-effective [1402140714371474]. Increasingly, courts are also recognizing that responsibility lies with plastic producers, manufacturers, and transporters under both statutory and common law, creating new impetus for “polluter pays” solutions [13981475]. Moreover, an approach that assigns much greater legal and financial responsibility to the plastic industry for previously externalized costs aligns well with the 2021 recommendation of the UN Special Rapporteur on Toxics, in UN General Assembly Report A/$\frac{76}{207}$, that plastics be addressed using a “human rights–based approach,” citing the specific impacts of plastic on the following vulnerable populations [1476]: The UN General Assembly’s declaration in 2022 that a “clean, healthy and sustainable environment” is a human right confirms obligations to address plastic pollution where it begins, with production [1477]. ## Key SEJ solutions: Procedural equity In addition to addressing distributive inequity, which is the disproportionate impact of harm to specific peoples and communities, there is a need to institute procedural equity in government and nongovernment sector decision-making across the plastics life cycle. Action in five key areas will be needed to advance procedural solutions to SEJ plastics pollution issues: [1] equitable and inclusive participation, [2] full economic cost estimation, [3] transparency and access to information, [4] acknowledge and address societal roles, and [5] fill data and funding gaps. ## Equitable and inclusive participation Governments and private sector actors should adopt the participatory approach to ensure transparent processes and meaningful participation of affected communities in decision-making [200]. Such an approach would include the following elements: ## Full economic cost estimation To fully inform decision-making with SEJ considerations, economic quantification methods must capture the full economic cost of an action over time and across groups (see, e.g., [200] and Section 5, Economic Impacts). A broader public understanding of the full costs associated with plastics is essential to engender actions needed to remedy social and environmental injustice associated with plastics production, use, and disposal. Many costs and benefits can be difficult to quantify—particularly indirect costs, nonmonetary costs, monetary benefits, and nonmonetary benefits—but other methods exist and should be employed for quantifying costs and benefits, such as the following: ## Transparency and access to information Lack of equal access to information and knowledge places workers and underserved groups at a significant disadvantage in decision-making on plastic pollution. Of the utmost importance is advancing the access of such individuals and groups to a range of relevant, evidence-based information, including the state-of-the-science and procedural requirements and legal rights. Some examples of actions that would address transparency and access to information follow: ## Acknowledge and address societal roles Any process should recognize and accommodate economic and societal roles in legacy to novel systems to reduce plastic pollution. Examples of such recognition and accommodation might include the following: ## Fill data and funding gaps It is imperative to invest resources to identify and fill major gaps in funding and to generate knowledge in key areas critical to equitable decision-making. This may include the following: ## Key SEJ solutions: Stage-specific and distributional equity In addition to the foci previously discussed, stage-specific actions will also be necessary to mitigate the specific SEJ issues associated with the plastics life cycle. Examples of such actions are outlined below. ## Case Study 6.2 Indonesia’s Response to Plastic Waste. Both on land and in the ocean that encircles its islands, *Indonesia is* currently experiencing a trash catastrophe (Figure 6.8). The second-largest source of the vast amount of oceanic plastic garbage is Indonesia. The nation and its citizens suffer negative economic effects because of this waste. Currently, barely $10\%$ of the 6.8 Mt of plastic trash generated in Indonesia (including imported plastic trash) each year makes it to recycling facilities. The ocean receives about 625,000 tons of plastic waste each year. Waste management is lacking or inadequate across large areas of Indonesia, resulting in informal dumping and burning of litter on a substantial scale. Where management exists, landfills are frequently located relatively close to residential areas, and contaminated effluent can leak into those areas and impede the growth of surrounding crops. **Figure 6.8:** *Trash next to a waterway in Indonesia.Credit: Credit to Richard C. Thompson, University of Plymouth.* Contaminated effluent as well as large quantities of solid plastic waste enter rivers and impact the people whose livelihoods depend on them. Due to the harm caused to marine life by plastic pollution in the waters, the fishing sector also suffers [221483]. Videos showing trash-filled beaches in popular tourist locations like Bali have gone viral, alarming the tourism sector and potentially damaging Indonesia’s economy. The possible effects of this high pollution on tourism are a source of concern. Fortunately, the problem has been acknowledged, and measures exist to deal with Indonesia’s plastictrash issue. Actions by individual and community People, organizations, and the government are mobilizing to solve and lessen Indonesia’s plastic trash crisis. The first step is to become aware of the issue. Local Indonesians have made a big contribution to organizing campaigns and raising public awareness. For example, Melati and Isabel Wijsen established the environmental nonprofit Bye Bye Plastic Bags when they were just 12 and 10 years old, respectively [1484]. Bye Bye Plastic Bags has become one of the largest environmental nonprofits in Bali and is helping to educate children on the environmental harm of plastics. Another individual, Mohamad Bijaksana Junerosano, founded the social enterprise Waste4Change [1485]. It educates the populace on sorting and sustainably managing waste. Initiatives to clean up the community have also gained popularity recently. Simple and efficient ways to engage people are beach cleanups. For a one-day beach cleaning in August 2018, that also brought attention to the garbage situation, more than 20,000 individuals organized in 76 places around Indonesia [1486]. Action by the government Both local and national levels of government have taken the most important steps to end the crisis of plastic waste in Indonesia. The island of Bali banned all single-use plastics at the end of 2018 [1487]. The capital of Jakarta also banned single-use plastic bags in its shopping centers and street markets in 2020 pursuant to Governor’s Regulation No. $\frac{142}{2019}$ [1488]. Internationally, the UK has invested over US$5 million in research to help identify solutions, including in *Indonesia via* the Pisces Partnership [1489]. Indonesia’s national government has rolled out a very ambitious plan to end the plastic waste problem. It aims to minimize marine plastic waste by $70\%$ by 2025 and be entirely rid of plastic pollution by 2040. Indonesia created five action points to make it easier to meet these overall goals [14901491]: Though reducing plastic waste in Indonesia and its oceans is a challenge, ordinary people and the government of Indonesia are taking proactive steps. Hopefully, these efforts will have a positive impact on livelihoods, the economy, and the health of people. The future looks brighter for a cleaner Indonesia. ## Main Findings This Commission has four major findings: ## Main Finding #1 Current practices for the production, use, and disposal of plastics cause great harms to human health and the global environment, and they are not sustainable. These harms arise at every stage across the plastic life cycle and are described in Sections 2 and 4 of this Commission. They include human health impacts such as developmental neurotoxicity, endocrine disruption, and carcinogenesis. In the ocean (Section 3), plastics’ harms extend far beyond the visible and well-recognized damages of beach litter, contaminated mid-ocean gyres, and physical injury to marine species and include extensive injury to marine ecosystems. Plastic production results in GHG emissions equivalent to nearly 1.96 Gt of CO2e annually that contribute to climate change. The main driver of plastics’ worsening harms is an almost exponential and still accelerating increase in global plastic production. More than half of all plastics ever produced have been manufactured since 2002. Plastics’ harms are further magnified by low rates of recovery and recycling—less than $10\%$ globally—and by the long persistence of plastic waste in the environment. The result has been the accumulation since 1950 of nearly 6 Gt of plastic waste that now pollutes every corner of the planet [3]. ## Main Finding #2 The thousands of chemicals in plastics—monomers, additives, processing agents, and NIAS—are responsible for many of plastics’ known harms to human and planetary health (Section 2). These chemicals leach out of plastics, enter the environment, cause pollution, and result in human exposure (Section 4). In the environment and in the bodies of living organisms, many plastic-associated chemicals can undergo chemical transformation to form breakdown products and metabolites, some of which are highly toxic and contribute further to plastics’ harms. Plastic manufacturers disclose little information on the identity, chemical composition, or potential toxicity of plastic chemicals at the time of entry to market and in most countries are under no legal obligation to do so. Both the complexity and the lack of transparency regarding the chemical composition of plastics has led to the current situation in which publicly funded epidemiologic research must attempt to discover possible health impacts of plastic-associated chemicals, but only after these chemicals have been released to market and resulted in potentially widespread human exposure. ## Main Finding #3 The economic costs of plastics’ harms to human health and the global environment are very high (Section 5). We estimate that in 2015 the health-related costs of plastic production exceeded $250 billion (2015 Int$) globally, and that in the US alone the health costs of disease, disability, and premature death caused by just three plastic-associated chemicals (PBDE, BPA, and DEHP) exceeded $920 billion (2015 Int$). The cost of GHG emissions from plastic cause economic harms that we value at $341 billion (2015 Int$) annually. These costs, large as they are, underestimate the full costs of plastics’ impacts on human health and the environment. All of these costs are externalized by the petrochemical and plastic manufacturing industries, and they are borne by individual citizens, taxpayers, and their governments without compensation. ## Main Finding #4 The health, environmental, and economic harms caused by plastics disproportionately affect vulnerable and at-risk populations—the poor, people of color, and Indigenous populations as well as fossil fuel extraction workers; plastic production workers; informal waste and recovery workers; persons living in communities adjacent to fossil fuel extraction, plastic production, and plastic waste facilities; and children [1494149514961497149814991500] (Sections 4 and 6). These disparate harms are seen in countries at every level of income, including high-income countries [149414971499]. They are seen globally in the export of vast quantities of plastic waste, including plastic-laden e-waste from high-income to low-income countries, where this waste accumulates in open tips and landfills, pollutes air and water, degrades vital ecosystems, befouls beaches and estuaries, damages fisheries, and harms human health, especially children’s health [14951498]. SEJ principles require reversal of these inequitable burdens to ensure that no group bears a disproportionate share of plastics’ harms and that those who benefit economically from plastics bear their fair share of its currently externalized costs. ## Recommendations for Policy Makers This Commission’s strongest recommendation is that the Intergovernmental Negotiating Committee (INC) for the Global Plastics Treaty develop and implement a strong and comprehensive, legally binding Treaty that ensures urgent action and effective interventions at an international scale across the entire life cycle of plastics to end plastic pollution, pursuant to the mandate set forth in the March 2022 resolution of the UNEA [37]. Progress toward development of this *Treaty is* already underway, and the first meeting of the INC took place in Punta del Este, Uruguay, in late 2022 [15011502]. International measures to curb plastic production and pollution are critical because the harms to human health and the environment caused by plastics, plastic-associated chemicals, and plastic waste transcend national boundaries, are planetary in their scale, and have disproportionate impacts on the health and well-being of people in some of the world’s poorest nations. A powerful and effective Treaty, consistent with fundamental principles of precaution, would build on models already elaborated in existing multilateral environmental agreements [15031504]. Experience with accelerated timelines for such other agreements as the Ottawa Land Mine Convention [15051506] and the Nuclear Weapons Prohibition Treaty [1507] suggest that the timeline proposed in the UNEA resolution, for completion of Treaty development by the end of 2024, is realistic. The Commission notes that effective implementation of the Global Plastics Treaty will require coordinated action at the global, national, regional, and local levels. This Commission encourages national, regional, state and local policymakers to be involved in the negotiations on the Treaty, including to support evaluation of the efficacy and feasibility of measures proposed for inclusion in the Treaty as negotiations proceed. National and local policymakers are uniquely well-positioned to pilot test and assess the efficacy of harm reduction strategies, and their experience can provide valuable guidance and real-world grounding to the treaty deliberations. ## 1.1 Global Cap on Plastic Production Given the great and growing magnitude of the harms caused by plastics to human and planetary health [15081509] (Sections 2, 3, 4, and 6), the substantial and still undercounted economic costs resulting from those harms (Section 5), and the enormous increases in plastic production projected for coming decades (Section 2), this *Commission is* of the considered opinion that a cap on plastic production is well-justified, much needed, timely—and importantly—the most effective harm-reduction strategy. The great power of a global cap on plastic production is that it will reduce the volume of plastics and plastic waste at its root source. It will slow the current massive global buildout of plastic production infrastructure. It will help put the world on track to end plastic pollution by 2040, a target put forth by the High Ambition Coalition to End Plastic Pollution [1510]. Like the phase-out of ozone-depleting chlorofluorocarbons under the Montreal Protocol [1511] and the removal of lead from gasoline [1512], a global cap on plastic production will have far-reaching benefits for planetary and human health. As was the case with both of those interventions, a cap on plastic production can be expected to have salutary public policy impacts by encouraging industry to develop new technologies and substitutes for existing uses. Complementary, “downstream” control strategies such as enhanced plastic recovery, recycling, and reuse need also to be encouraged. Experience indicates, however, that these approaches are inherently less effective than “upstream” prevention and that without curbing continuing unsustainable production of plastics, they may be expected to continue to lag behind for the foreseeable future (Section 2) [57577]. This Commission recognizes that for a global cap on plastic production to be effective, the Treaty will need to include a roadmap that stipulates targets and timetables, and the Treaty includes national contributions that are binding. Targets and timetables for LMICs will likely need to be less stringent than those for high-income countries, to facilitate a just transition. ## 1.2 Bans on Unnecessary, Avoidable and Problematic Plastic Items As is described in Section 2, manufacture of single-use plastic products accounts for 35–$40\%$ of current plastic production, and this fraction is growing rapidly. Inclusion in the Global Plastic Treaty of a provision restricting manufacture and use of unnecessary single-use plastics will help curb current unsustainable increases in plastic production and slow the accumulation of plastic waste. The Montreal Protocol and the Stockholm Convention both provide precedents on how such a ban or restriction could be structured under the Global Plastic Treaty. Many countries, states, and cities have already successfully imposed bans on some single-use plastic items [1513151415151516]. Additional strategies for limiting use of single-use plastics at the national and local levels are suggested in the UNEP report, Single-Use Plastics: A Roadmap for Sustainability [1432]. This Commission suggests that manufactured MPs such as microbeads in cosmetics (see Box 7.1) be considered as a target for banning under the Global Plastics Treaty. ## Box 7.1 Manufactured microplastic particles—“microbeads”. Manufactured microplastic (MP) particles. Manufactured MP particles, often called “microbeads” are now intentionally added to many personal care products and cosmetics, including sunscreen, shampoo, makeup, and deodorants as well as in other commercial and consumer products [1233]. These products contribute directly to environmental contamination and human exposure yet their perceived benefit to society is trivial [1517]. To counter these materials’ potential hazards to human health and the environment, several countries including New Zealand, Canada, UK, and the Republic of Korea and have banned plastic microbeads in cosmetics and personal care products [1518]. In the US, the Microbead-Free Waters Act of 2015 prohibits the manufacture, packaging, and distribution of rinse-off cosmetics containing plastic microbeads [1519]. In 2019, the European Chemicals Agency proposed a sweeping restriction on the use of MPs in all types of EU market products. A year later, the agency’s Committee for Risk Assessment recommended an additional ban on all MPs utilized in infill for artificial turf fields [1520]. The final EU rule on intentionally added MPs in products is scheduled to be released soon [1518]. The Nordic Council of Ministers recommend minimizing MP releases at every stage of the plastic life cycle [1521]. ## 2.1 Inclusion of Chemicals in Scope Thousands of chemicals, including monomers, additives, and NIAS, are incorporated into plastics during manufacture and are integral components of plastic products, macroplastic waste, and MP particles. Many of these chemicals are responsible for a very great part of the harms to human health and the environment caused by plastic. Over 2,400 plastics chemicals have hazard ratings that are of high concern. Most of the rest have never been assessed for their potential impacts on human and ecosystem health. Plastics chemicals include neurotoxicants, human carcinogens and endocrine disruptors. As is documented in Section 4, plastics chemicals are especially dangerous for infants in the womb, young children, and pregnant women. ## 2.2 Establishment of Health Protective Standards for plastic-associated chemicals The incorporation of hazardous chemicals into plastic production is part of the larger problem of inadequate regulation of chemicals in consumer goods. Health-protective standards for plastic-associated chemicals and their associated reporting obligations should address five major areas: sustainable design; mutual acceptance of data; transparency on the chemical composition and toxicity of plastic-associated chemicals; and systems for human biomonitoring and post-market surveillance [1523]. ## 2.3 Reducing the Complexity of Plastic Products Currently, many different chemicals are used to perform similar functions in plastic production. This, together with a lack of coordination among manufacturers, has resulted in an enormous proliferation of different types of plastic. This complexity poses great challenges to downstream control efforts, including recovery and recycling. These problems are compounded by the lack of any system to trace the identities and levels of the chemicals present in specificplastic products. To address this issue, the UNEA resolution specifically identifies the need to address the design of products and materials This Commission notes that several suggestions on how to achieve product streamlining have recently been put forth [1510]. They include developing design standards for plastics to simplify their chemical composition and increase compatibility with end-of-life management [43], for example, through removing toxic chemicals from plastics and replacing them with more environmentally friendly materials, redesigning clothing to reduce fiber shedding and redesigning tires to reduce PM and microfiber release. ## 3.1 Mandatory Extended Producer Responsibility (EPR) Frameworks The Treaty could prescribe minimum global standards for EPR frameworks to be implemented at the national level by treaty parties. EPR legislation makes plastic producers and the manufacturers of plastic products legally and financially responsible for the safety and end-of-life management of all the materials they produce and sell. By making plastic producers responsible for costs that until now they have externalized and shifted onto governments, taxpayers, and the general public, the goal of EPR is to incentivize change within industry. By analogy, liability protocols have been adopted for releases of hazardous substances [1527], wastes [1528], and genetically modified organisms [1529] under other major multilateral conventions. At a minimum, EPR programs require producers of plastic products to either take back these products at the end of their useful life (with take back encouraged in some instances through deposit fees) or to cover costs of waste management and clean-up (see examples in Boxes 7.2, 7.3, 7.4) [1530]. For example, Canada has had a national EPR policy in place since 2009 that has been implemented in at least five provinces [1531]. Container deposit laws in multiple countries provide additional examples of successful EPR strategies (Box 7.4). A striking example of the benefit that could have been delivered by EPR is seen in the case of plastic microbeads in cosmetics. The patent for the use of plastic microbeads in cosmetics was filed decades ago. If EPR been in place at that time it would have likely prevented the hundreds of thousands of tons of totally avoidable environmental contamination that have resulted from use of these materials. EPR programs can be coordinated with national and state strategies to reduce virgin plastic production. The ultimate goal of this suite of interventions is to make the plastic supply chain more circular and less linear, thus reducing need for virgin plastic production and slowing the accumulation of plastic waste. ## Box 7.2 Extended producer responsibility (EPR) and e-waste. EPR and E-Waste. Electric-powered and electronic products contain large quantities of many types of plastic. Substantial opportunities exist to reduce both plastic production and the generation of e-waste by requiring that all electric-powered and electronic products be repairable, that their components be reusable, and that manufacturers take their products back at the end of their useful life for reuse, remanufacturing, recycling, or safe disposal. Such requirements will make manufacturers of electric-powered and electronic products responsible for the costs of e-waste handling that they currently externalize and shift on to state and local governments and vulnerable populations. It will also make them responsible for the large volumes of e-waste that they currently send to landfills in high-income countries and into e-waste dumpsites in LMICs. Product design standards that require easy disassembly of all electric and electronic products and the repair and reuse of components are key to achieving this goal. They should include standardization of the plastics used in electrical appliances and electronic goods. The European Commission has already proposed design requirements mandating most of these features [1532]. “Right to Repair” laws are a further strategy for reducing e-waste. These laws prohibit manufacturers from placing limitations on access to repair materials such as parts, tools, diagnostics, and programming such as firmware. In July 2017, the European Parliament approved a recommendation that Member States should pass laws giving consumers the right to repair their electronics. Likewise, the British government introduced a “Right to Repair” law that went into effect on July 1, 2021. ## Box 7.3 Extended producer responsibility (EPR) for abandoned, lost, and discarded fishing gear (ALDFG). EPR and ALDFG. Abandoned, lost, or discarded fishing gear constitutes a significant portion of ocean plastics [551]. Entanglement in ALDFG is well known to endanger marine mammals, turtles, seabirds, and some fishes and poses a threat to ecosystems [505567722]. ALDFG may also serve as an ongoing source of marine leakage of harmful chemicals and MPs [15331534]. Several EPR provisions pertaining to fishing gear are already in place in the EU, where Directive $\frac{2019}{904}$ requires producers of fishing gear containing plastic to cover the costs of 1) the collection of waste gear and 2) awareness raising measures to prevent and reduce the abandonment of gear at sea. EU Member States are required to monitor and assess compliance [334]. The measure was built on earlier directives, such as that of 2008 obliging national minimum annual collection rates of gear containing plastic for recycling [1535] and of 2009 requiring that gear be tagged with an external identification number, and that the master of the vessel attempt to retrieve lost gear as soon as possible, reporting losses to authorities of the flag state within 24 hours [1536]. An international program of EPR can incorporate and expand on these provisions, for instance by implementing a universal system of equipment registration, and by marking registered gear with acoustic transponder tags, which can be used to ensure that lost gear is tracked and, when possible, retrieved. Such tags are relatively inexpensive and are already in use in fisheries in Southwest England [1537]. Commercial fishers who use fish aggregating devices employ similar devices, and these have been successfully utilized in demonstration projects by bodies such as the European Climate Infrastructure and Environment Executive Agency to locate and retrieve lost nets [15381539]. Technologies such as these, in combination with deposit schemes or regulations that impose financial penalties for discarded or lost equipment, can introduce stronger systems of accountability. Drawing on the EU’s minimum annual collection rates, international requirements can be instituted requiring manufacturers to offer take-back credits, and to ensure that new gear be made using a percentage of recycled materials. A certification or labelling scheme can be initiated to identify products made from recycled fishing gear, thus conferring a higher value [1537]. ## Box 7.4 The effectiveness of bottle deposit laws. Container Deposit Laws. Deposit laws in multiple countries for bottles, cans and other containers provide an example of a highly successful EPR strategy implemented at national level. Under EC Directive $\frac{2019}{904}$, EU Member States are encouraged to establish bottle bills that incentivize the collection and reuse of containers, including plastic bottles [334]. Such bills are already in place in many countries worldwide [1540]. In 2022, the 11 Canadian provinces with deposit laws had an average return rate of $74\%$, the 13 EU countries averaged $90\%$, and the 13 jurisdictions in Oceania averaged $69\%$. All of these rates are far higher than the global average plastic recycling rate of about $9\%$. Across the US, polyethylene terephthalate (PET) plastic beverage containers without a deposit were recycled at a rate of only $17\%$, in contrast to a nationwide average recycling rate of $57\%$ for PET bottles with a deposit [1541]. In 2019, Oregon and Michigan—the two US states with 10¢ deposits—had overall redemption rates of $86\%$ and $89\%$ respectively. In contrast, states such as Massachusetts and Connecticut, which have 5¢ deposits, had redemption rates of $43\%$ and $44\%$. In the European Union, the five best performing Member States with deposit schemes for PET bottles (Germany, Denmark, Finland, the Netherlands, and Estonia) reached an average collection rate for PET of $94\%$ in 2014 [1532]. ## 3.2. Explore Listing Plastic Polymers as Persistent Organic Pollutants (POPs) under the Stockholm Convention Plastics meet many of the cardinal criteria for listing as POPs: Based on the foregoing considerations, this Commission urges the Parties to the Stockholm Convention to consider listing of some plastic polymers as POPs and urges the INC expressly to call for exploration of this action. This Commission makes two further observations in regard to the Stockholm Convention: ## 4.1 Consideration of Vulnerable and At-Risk Populations in Developing Globally Binding Controls The UNEA resolution on the Global Plastics Treaty specifically identifies human health in its preamble. Building on this foundation, the operational provisions of the Global Plastics Treaty should be crafted to assure that protection of human health is a central goal of the instrument. The UNEA remit specifically instructs the INC to adopt a full life cycle approach to addressing the health impacts of the global plastics problem. The groups at greatest risk of harms to health caused by plastic across the life cycle are infants, children, pregnant women, workers, Indigenous populations, and persons living in “fenceline” communities adjacent to plastic industries (Sections 4 and 6). Protection of the health of these vulnerable and at-risk groups is ethically well justified, and measures crafted to protect their health will safeguard the health of entire populations. Consistent with the directive of the UNEA resolution, it is essential that affected communities, civil society organizations, Indigenous populations, environmental justice organizations, the scientific community, faith-based organizations, and LMIC representatives such as waste-pickers contribute to the treaty negotiations. People living near plastic production and waste facilities in both high-income countries and LMICs have a particularly important voice that needs to be heard and factored into decision-making. Their participation will help ensure that the Global Plastics Treaty includes measures specifically designed to address the disproportionate harms that plastics impose on these populations. Liberal standards for admission of accredited non-governmental observers to negotiations under UN auspices for major conventions on climate, biodiversity, international trade in wastes, POPs, stratospheric ozone protection, and others is now the global good practice standard. ## 4.2 Strengthening Restrictions on Transnational Export of Plastic Waste The provisions of the Global Plastics Treaty will need to build upon and reinforce the work of the Basel Convention on the Control of Transboundary Movements of Hazardous Wastes and Their Disposal. Having entered into force in May 1992, the Basel Convention [1546] was designed to reduce the movement of hazardous waste between nations, and specifically to discourage wealthier countries from exporting unsafe waste to lower-income nations. In 2019, this Convention was updated by the Plastic Waste Amendment, which modified three annexes to include certain plastic wastes [1528]. This amendment requires prior written consent of the importing country for most categories of plastic waste, including e-waste. A companion amendment [1547] has the effect of banning the export of contaminated and highly mixed plastic waste from the OECD countries, the EU, and Lichtenstein to LMICs. Despite the strong and well-crafted provisions of the Basel Convention and its Plastic Waste Amendment, massive amounts of plastic waste continue to flow into the world’s least developed countries where it degrades environments, harms human health, and deepens social injustices. To address this continuing crisis, this Commission recommends that the Global Plastics Treaty include a provision calling for collaboration with the Basel Convention to strengthen enforcement of Basel Convention’s regulations and increase awareness among national leaders of their power to refuse unwanted shipments of plastic waste under the provisions of the Basel Convention. Additionally, recognizing that The Basel *Convention is* primarily limited in scope to international trade in wastes, this Commission notes that the Global Plastics Treaty needs to penetrate to the domestic level, establishing globally agreed-upon standards for production, use, and disposal of plastics within countries and not only on shipments that enter into international trade. The London Convention and Protocol [1549] can and should be mobilized further to address the problem of the dumping of plastics into the marine environment. The Rotterdam Convention [1548] creates a similar prior-informed-consent procedure for chemicals and pesticides in international trade, potentially including many plastics or their precursors and starting materials. The parties to the Rotterdam Convention should consider a plastics amendment analogous to that under the Basel Convention. Even without such an instrument, the Rotterdam Convention can be mobilized to address an important additional component of the global plastics problem. ## 5. Creation of a Permanent Science Policy Advisory Body The governing UNEA resolution directs the INC to consider “the possibility of a mechanism to provide policy-relevant scientific and socioeconomic information and assessment related to plastics pollution.” Currently, in parallel with negotiations on the plastics treaty, the global community is working on establishing an intergovernmental, independent science-policy panel on chemicals, waste and pollution prevention, with the ambition to establish it by 2024. This body will need to interface closely with the Plastics Treaty and appropriate frameworks for that interface should be established. Negotiators may consider whether the science-policy functions for the Plastics Treaty could be realized through this new science-policy panel on chemicals, waste, and pollution prevention, or would be better served by a dedicated science-policy body established under the Treaty (or some combination of both). Such a body should be transdisciplinary in scope and include expertise in the natural, economic, and social sciences as well as regional expertise and Indigenous knowledge. It will be essential that this body is shielded from special interests. The overall priority of such a dedicated Permanent Science Policy Advisory Body could be to guide Treaty Parties in evaluating which solutions are most effective in reducing plastic production and consumption, curbing the generation of plastic waste, enhancing plastic waste recovery and recycling, and ensuring proper disposal of plastic waste. This Body could also assess trade-offs among proposed solutions, evaluate unintended consequences to interventions, and evaluate safer alternatives to current plastics (see Box 7.5). Specific functions of this Body could be to: ## Box 7.5 Independent evaluation of evidence on the efficacy of solutions is critical for the Global Plastics Treaty. The Need for Independent Evaluation of Evidence on the Efficacy of Solutions. It will be essential that solutions proposed to the plastics crisis be subject to careful review and due diligence to avoid “regrettable substitutions.” Some examples of inadequately vetted solutions that have been found, on review, to aggravate global plastics crisis include: *It is* critical that we learn from these mistakes and take a far more precautionary approach, e.g., based on EPR. This transition needs to start now; it needs to be evidence-based; and it needs to be enshrined in all approaches intended to address the issue of plastic production [60]. It is also critical to adopt an evidence-based approach to identify which aspects of the plastic crisis are best addressed by actions at an international versus the national level. A case study is seen in the release of MP fibers from textiles. Three main intervention points exist to reduce microfiber shedding: Options 1 and 2 are more attractive in the Global North because they can be implemented at a national scale. For example, new legislation in France will mandate filters on washing machines. By contrast, Options 1 and 2 will not be very effective in the Global South, where many populations do not have the benefit of washing machines or advanced wastewater treatment. Option 3, improved fabric design, which could be mandated by international legislation under the Global Plastics Treaty, appears to be effective in reducing microfiber shedding in all countries at every level of income, because $50\%$ of all microfiber emissions occur while garments are being worn rather than while they are being washed [1555]. Improved fabric design therefore has far greater potential to address the issue of microfiber shedding than any “downstream” solution [1555]. ## 6. Additional Recommendations for Harm Reduction This *Commission is* very clear in our view that the most effective strategies for slowing the accumulation of plastic waste and reducing the harms associated with plastics across its life cycle are “upstream” solutions that address the root causes of the plastic crisis. They include: These “upstream” solutions need to be supplemented by complementary, “downstream” strategies such as effective recovery, recycling, and reuse. The Commission makes the following comments regarding some existing and emerging, expensive, and unproven downstream waste management methods from the perspective of minimizing harm to human health and the environment: This Commission notes the following risks associated with these technologies: This Commission advises countries to approach these emerging, expensive, and unproven technologies with great caution and to undertake thorough, independent assessments of their environmental impacts prior to any adoption. Any implementation of these technologies should proceed step-wise and be closely scrutinized. A final consideration in regard to chemical and thermal conversion is that investments into these costly yet unproven technologies will divert funding away from proven effective strategies. Investments made in chemical and thermal recycling thus have potential to derail efforts to address the root causes of the global plastics crisis. ## Recommendations for Research While much actionable information is already available on plastics’ hazards to human and planetary health, gaps in knowledge remain and additional research is needed to better safeguard health. Governments will likely support much of this research using public funds. An additional source of support could be a carbon tax levied against the industries that produce the coal, oil, and gas feedstocks used in plastic manufacture. Such a tax would complement EPR fees paid by the manufacturers of plastic products to prevent the accumulation of plastic waste. Like EPR, it would have the effect of making the producers of fossil carbon responsible for the harms to human health and the global environment caused by the materials they produce and sell. ## 1. Operational Research This research will need to determine which solutions are most effective and cost-effective in the context of particular countries and assess the risks, benefits, and trade-offs of proposed solutions. It will need additionally to evaluate any potential unintended consequences of interventions. This Commission recommends that there be close coordination between research on these topics at the national and international level and the Permanent Science Policy Advisory Body that this Commission suggests be established under the treaty. ## 1. Education Because plastics’ manufacture, use, disposal, and pollution are globally pervasive and growing, it is imperative that today’s youth be educated about the plastic world that they will inherit and have to address as the consumers and leaders of tomorrow. Youth need to have as keen an appreciation of the dangers of chemical and plastic pollution as they do of climate change. Educators can instill in students the importance of a healthy body and a healthy environment and help students grow into well-informed consumers and members of society who have the power to make change with their individual voices, choices, and actions. ## 2. Outreach to Impacted Communities Although plastics are materials upon which modern society is heavily reliant, substantial outreach is required to inform the general populace about the dangers of pervasive use of unnecessary plastic products and the negative impacts these products have on human and environmental health from production through use, disposal, and pollution of the environment. It is critical that accurate and timely information be shared, and that concrete, actionable alternatives are co-created according to the specific needs and circumstances of each local community. ## Recommendations for the Medical, Public Health, and Scientific Communities While the oceanographic and marine biology communities have been aware of plastic’s negative impacts on the environment since the 1970s [1490], the medical and public health communities have until now not been widely aware of plastics’ impacts on human health [2841560]. This is not surprising, given that the science in this area is still emerging and the majority of reports on plastics’ hazards have appeared in environmental and oceanographic journals and in government agency publications not widely read by physicians, nurses, and public health professionals. Moreover, plastic and chemical pollution have been worsening quietly, while the world’s attention has been focused on climate change. Today, however, as knowledge of plastics’ many harms to human and planetary health increases and becomes more widely available, physicians, nurses, and public health professionals have an opportunity to lead the global effort to reduce these hazards and to protect the health of their patients. Health professionals can educate themselves and their patients about plastics and their hazards. They can take a leadership role in reducing plastic use and plastic waste generation in hospitals and health care facilities. The health care sector generates substantial volumes of plastic waste, and depending on the medical facility, plastics comprise between 20–$65\%$ of all waste generated [284]. While some plastics are essential in health care, many are not, and careful review of inventories and supply chains has potential to greatly reduce plastic use. Examples include going back to cotton rather than plastic sheets and using sterilizable, reusable surgical instruments instead of single-use plastic devices. Public health agencies can increase support for toxicological and epidemiological research into the health hazards of plastics and plastic additives. They can launch and support large-scale, multi-year human biomonitoring programs and observational studies. They can increase their educational offerings on chemical and plastic pollution. Medical societies and public health organizations are uniquely well positioned to educate elected officials about plastics’ harms to human health and the environment and to advocate governments at every level to reduce plastic production, use, and disposal to protect the health of their patients. They can advocate for such goals as reductions in plastic production; moratoria on fracking and on the construction of cracker plants, pipelines, and compressor stations; restrictions on single use plastics; and passage of EPR legislation. By pointing out to elected officials the increasingly well-documented links between plastics and harms to human health, and by noting that actions to control plastic production will also help control climate change, prevent pollution, and save tax dollars, doctors, nurses, public health professionals, and scientists are in a powerful position to oppose the forces that call for endless, unchecked increases in plastic production. These trusted advocates are uniquely well positioned to catalyze enduring action to safeguard human health, protect the planet, and advance the common good. ## Additional File The additional file for this article can be found as follows: ## Funding Information We acknowledge the generous support of the Minderoo Foundation, a modern philanthropic organization, the Centre Scientifique de Monaco, and the Prince Albert II of Monaco Foundation. The Minderoo-Monaco *Commission is* an independent and objective scientific review of the peer-reviewed literature that itself has undergone peer review. Neither Minderoo Foundation, nor its benefactors, had any influence over the conduct, the findings, or the recommendations of the Commission. ## Competing Interests In addition to the adjunct positions at the Nigerian Institute of Medical Research and Lead City University in Nigeria, AM works for Shell Nigeria Exploration & Production Company but did not receive any support from the company for her research and for this study, and the company is not in any way involved in this study. MB, DC, AG, YM, BJS, CS, and SD are employed by the Minderoo Foundation, an independent not-for-profit philanthropic organization. The contributions of the following authors were supported by the Minderoo Foundation: MC, MH, RH, AL, AM, MP, YP, MS, JJS, HT & RCT. JJS’s work was also supported by the Woods Hole Center for Oceans and Human Health (NIH grant P01ES028938 and National Science Foundation grant OCE-1840381). MEH’s work was also supported by the Minderoo Foundation as well as by grants from Woods Hole Sea Grant (Award No. NA18OAR4170104, project R/P–89) and the March Marine Initiative, a program of March Limited, Bermuda. BDJ was supported by the Postdoctoral Scholar Program at the Woods Hole Oceanographic Institution (WHOI), with funding provided by the Weston Howland Jr. Postdoctoral Scholarship. JAP was supported by the US National Science Foundation (NSF) Graduate Research Fellowship Program as well as by a grant from Woods Hole Sea Grant (Award No. NA180AR4170104, project R/P-89). ZW gratefully acknowledges funding by the European Union under the Horizon 2020 Research and Innovation Programme (grant agreement number 101036756). No other authors have conflicts of interest. ## Author Contributions This project was conceived by PJL, SD, HR, and CS. All authors approved the manuscript. PJL took the lead in writing and reviewing the introduction. JM, AKY, SD, CS, HR, JJS, MEH, and LEF contributed to the writing and reviewing the introduction. RT took the lead in writing and reviewing the “Our Plastic Age” textbox. LEF and JE contributed to reviewing of this textbox. SD took the lead in writing and reviewing Section 2. ZW, DC, JE, HT, AG, BJS, MB, CS, MW, BH, JJS, KLL, MH, MS, MR and PJL contributed to the writing and reviewing this Section. MEH and JJS took the lead in writing and reviewing Section 3. RT, RH, MLP, HT, KLL, CFP, BDJ, and JAP contributed to the writing and reviewing of Section 3; SD contributed to reviewing. PJL and CS took the lead in writing and reviewing Section 4. JJS, AL, SD, MEH, HR, PF, CB, EW, HI, CG, BJS, AG, MW, WS, LEF, AM, and AKY contributed to writing and reviewing this Section. RH took the lead in writing and reviewing “The Contribution of the Healthcare Sector to Plastic Waste Production” textbox. SD, PJL, MEH, and LEF contributed to the reviewing the textbox. MC and YP took the lead in writing and reviewing Section 5. PK, RF, and PJL contributed to reviewing. MS and AM took the lead in writing and reviewing Section 6. EMC, MJD, LEF, BH and AV contributed to the writing and reviewing Section 6 and PL and SD contributed to reviewing this Section. MW, HR, and PJL took the lead in writing and reviewing Section 7. JE, MH, SD, CS, PJL, KM, JN, DW, RT, KLL, SP, JJS, MEH, and TC and ZW contributed to reviewing this Section. YM conceptualized the infographic design and contributed to the review of all sections. ## References 1. Carpenter EJ, Anderson SJ, Harvey GR, Miklas HP, Peck BB. **Polystyrene spherules in coastal waters**. *Science* (1972.0) **178** 749-750. DOI: 10.1126/science.178.4062.749 2. Thompson RC, Olsen Y, Mitchell RP. **Lost at sea: where is all the plastic?**. *Science* (2004.0) **304** 838. DOI: 10.1126/science.1094559 3. Geyer R, Jambeck JR, Law KL. **Production, use, and fate of all plastics ever made**. *Sci Adv* (2017.0) **3**. DOI: 10.1126/sciadv.1700782 4. Ritchie H, Roser M. **Plastic Pollution. Our world in data** 5. 5Organisation for Economic Co-operation and Development (OECD). Global Plastics Outlook: Economic Drivers, Environmental Impacts and Policy Options. Organisation for Economic Co-operation and Development; 2022. Accessed December 15, 2022. DOI: 10.1787/de747aef-en. *Global Plastics Outlook: Economic Drivers, Environmental Impacts and Policy Options* (2022.0). DOI: 10.1787/de747aef-en 6. 6Ellen MacArthur Foundation. The New Plastics Economy: Rethinking the Future of Plastics & Catalysing Action. Ellen MacArthur Foundation; 2017: 68. Accessed October 7, 2022. https://emf.thirdlight.com/link/cap0qk3wwwk0-l3727v/@/#id=1.. *The New Plastics Economy: Rethinking the Future of Plastics & Catalysing Action* (2017.0) 68 7. 7World Economic Forum. The New Plastics Economy: Rethinking the Future of Plastics. World Economic Forum (WEF). 2016: 36. Accessed December 23, 2022. https://www3.weforum.org/docs/WEF_The_New_Plastics_Economy.pdf.. *The New Plastics Economy: Rethinking the Future of Plastics* (2016.0) **36** 8. Wiesinger H, Wang Z, Hellweg S. **Deep dive into plastic monomers, additives, and processing aids**. *Environ Sci Technol* (2021.0) **55** 9339-9351. DOI: 10.1021/acs.est.1c00976 9. Azoulay D, Villa P, Arellano Y. *Plastic & Health: The Hidden Costs of a Plastic Planet* (2019.0) 10. **Plastics: a carbon copy of the climate crisis** 11. Bauer F, Fontenit G. **Plastic dinosaurs – digging deep into the accelerating carbon lock-in of plastics**. *Energy Policy* (2021.0) **156** 112418. DOI: 10.1016/j.enpol.2021.112418 12. 12European Environment Agency (EEA). Plastics, the Circular Economy and Europe’s Environment – A Priority for Action. European Environment Agency; 2021. DOI: 10.2800/5847. *Plastics, the Circular Economy and Europe’s Environment – A Priority for Action* (2021.0). DOI: 10.2800/5847 13. Cabernard L, Pfister S, Oberschelp C, Hellweg S. **Growing environmental footprint of plastics driven by coal combustion**. *Nat Sustain* (2022.0) **5** 139-148. DOI: 10.1038/s41893-021-00807-2 14. 14Organisation for Economic Co-operation and Development (OECD). Global Plastics Outlook: Policy Scenarios to 2060. Organisation for Economic Co-operation and Development; 2022. DOI: 10.1787/aa1edf33-en. *Global Plastics Outlook: Policy Scenarios to 2060* (2022.0). DOI: 10.1787/aa1edf33-en 15. Rockström J, Steffen W, Noone K. **Planetary boundaries: exploring the safe operating space for humanity**. *Ecol Soc* (2009.0) **14** 16. Persson L, Carney Almroth BM, Collins CD. **Outside the safe operating space of the planetary boundary for novel entities**. *Environ Sci Technol* (2022.0) **56** 1510-1521. DOI: 10.1021/acs.est.1c04158 17. Barnes DKA, Galgani F, Thompson RC, Barlaz M. **Accumulation and fragmentation of plastic debris in global environments**. *Philos Trans R Soc B Biol Sci* (2009.0) **364** 1985-1998. DOI: 10.1098/rstb.2008.0205 18. Kershaw P. J.. **Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection Sources, Fate and Effects of Microplastics in the Marine Environment: A Global Assessment**. *International Maritime Organization* (2015.0) 19. Law KL. **Plastics in the marine environment**. *Annu Rev Mar Sci* (2017.0) **9** 205-229. DOI: 10.1146/annurev-marine-010816-060409 20. Rochman CM, Brookson C, Bikker J. **Rethinking microplastics as a diverse contaminant suite**. *Environ Toxicol Chem* (2019.0) **38** 703-711. DOI: 10.1002/etc.4371 21. Landrigan PJ, Stegeman JJ, Fleming LE. **Human health and ocean pollution**. *Ann Glob Health* (2020.0) **86** 151. DOI: 10.5334/aogh.2831 22. 22United Nations Environment Programme (UNEP). From Pollution to Solution: A Global Assessment of Marine Litter and Plastic Pollution. United Nations Environment Programme (UNEP); 2021: 148. Accessed October 7, 2022. http://www.unep.org/resources/pollution-solution-global-assessment-marine-litter-and-plastic-pollution.. *From Pollution to Solution: A Global Assessment of Marine Litter and Plastic Pollution* **2021** 148 23. 23National Academies of Sciences, Engineering and Medicine. Reckoning with the U.S. Role in Global Ocean Plastic Waste. The National Academies Press; 2022. DOI: 10.17226/26132. *Reckoning with the U.S. Role in Global Ocean Plastic Waste* (2022.0). DOI: 10.17226/26132 24. Jambeck JR, Geyer R, Wilcox C. **Plastic waste inputs from land into the ocean**. *Science* (2015.0) **347** 768-771. DOI: 10.1126/science.1260352 25. Landrigan PJ, Frumkin H, Lundberg BE. **The false promise of natural gas**. *N Engl J Med* (2020.0) **382** 104-107. DOI: 10.1056/NEJMp1913663 26. Symeonides C, Brunner M, Mulders Y. **Buy-now-pay-later: hazards to human and planetary health from plastics production, use and waste**. *J Paediatr Child Health* (2021.0) **57** 1795-1804. DOI: 10.1111/jpc.15777 27. Creech JL, Johnson MN. **Angiosarcoma of liver in the manufacture of polyvinyl chloride**. *J Occup Med Off Publ Ind Med Assoc* (1974.0) **16** 150-151 28. Rinsky RA. **Benzene and leukemia: an epidemiologic risk assessment**. *Environ Health Perspect* (1989.0) **82** 189-191. DOI: 10.1289/ehp.8982189 29. Casey JA, Savitz DA, Rasmussen SG. **Unconventional natural gas development and birth outcomes in Pennsylvania, USA**. *Epidemiology* 1. DOI: 10.1097/EDE.0000000000000387 30. O’Callaghan-Gordo C, Orta-Martínez M, Kogevinas M. **Health effects of non-occupational exposure to oil extraction**. *Environ Health* (2016.0) **15** 56. DOI: 10.1186/s12940-016-0140-1 31. Saunders PJ, McCoy D, Goldstein R, Saunders AT, Munroe A. **A review of the public health impacts of unconventional natural gas development**. *Environ Geochem Health* (2018.0) **40** 1-57. DOI: 10.1007/s10653-016-9898-x 32. Clark CJ, Johnson NP, Soriano M. **Unconventional oil and gas development exposure and risk of childhood acute lymphoblastic leukemia: a case-control study in Pennsylvania, 2009–2017**. *Environ Health Perspect* (2022.0) **130** 087001. DOI: 10.1289/EHP11092 33. **Don’t pollute my future! The impact of the environment on children’s health** 34. Harremoës P, Gee D, MacGarvin M. *Late Lessons from Early Warnings: The Precautionary Principle 1896–2000* (2001.0) 1-211 35. Landrigan PJ, Raps H, Symeonides C. **Announcing the Minderoo – Monaco Commission on Plastics and Human Health**. *Ann Glob Health* (2022.0) **88** 73. DOI: 10.5334/aogh.3916 36. Wang Z, Altenburger R, Backhaus T. **We need a global science-policy body on chemicals and waste**. *Science* (2021.0) **371** 774-776. DOI: 10.1126/science.abe9090 37. 37United Nations Environment Programme (UNEP). End Plastic Pollution – Towards an International Legally Binding Instrument; 2022(UNEP/EA.5/Res.14): 1–4. Accessed October 7, 2022. https://wedocs.unep.org/xmlui/bitstream/handle/20.500.11822/39764/END%20PLASTIC%20POLLUTION%20-%20TOWARDS%20AN%20INTERNATIONAL%20LEGALLY%20BINDING%20INSTRUMENT%20-%20English.pdf?sequence=1&isAllowed=y.. *End Plastic Pollution – Towards an International Legally Binding Instrument* (2022.0) 1-4 38. 38Institute of Medicine (US) Committee for a Planning Study on Ongoing Study of Costs of Environmental-Related Health Effects. Costs of Environment-Related Health Effects: A Plan for Continuing Study. National Academies Press (US); 1981. Accessed November 9, 2022. http://www.ncbi.nlm.nih.gov/books/NBK219029/.. *Costs of Environment-Related Health Effects: A Plan for Continuing Study* (1981.0) 39. 39United Nations Environment Programme (UNEP). Faith for Earth – an ethical approach to global challenges. UNEP. Published August 22, 2018. Accessed November 9, 2022. http://www.unep.org/news-and-stories/story/faith-earth-ethical-approach-global-challenges. 40. Landrigan PJ, Vicini A. *Ethical Challenges in Global Public Health: Climate Change, Pollution, and the Health of the Poor* (2021.0) 41. Bucci K, Tulio M, Rochman CM. **What is known and unknown about the effects of plastic pollution: a meta-analysis and systematic review**. *Ecol Appl* (2020.0) **30**. DOI: 10.1002/eap.2044 42. 42United Nations Environment Programme (UNEP). Resolution 5/14: End Plastic Pollution: Towards an International Legally Binding Instrument. Vol UNEP/EA. 5/Res.14. 2022; 4. Accessed November 22, 2022. https://www.unep.org/news-and-stories/press-release/historic-day-campaign-beat-plastic-pollution-nations-commit-develop.. *Resolution 5/14: End Plastic Pollution: Towards an International Legally Binding Instrument* (2022.0) **4** 43. Wang Z, Praetorius A. **Integrating a chemicals perspective into the global plastic treaty**. *Environ Sci Technol Lett* (2022.0) **9** 1000-1006. DOI: 10.1021/acs.estlett.2c00763 44. Hakuzimana J. **Break free from plastics: environmental perspectives and evidence from rwanda**. *Environ Ecosyst Sci* (2021.0) **5** 27-36. DOI: 10.26480/ees.01.2021.27.36 45. **SB-54 Solid waste: reporting, packaging, and plastic food service ware. California Legislative Information** 46. **Encyclical Letter Laudato Si’ of the Holy Father Francis on Care for Our Common Home. Published online May 24, 2015** 47. Heller MC, Mazor MH, Keoleian GA. **Plastics in the US: toward a material flow characterization of production, markets and end of life**. *Environ Res Lett* (2020.0) **15** 094034. DOI: 10.1088/1748-9326/ab9e1e 48. Wang C, Liu Y, Chen WQ, Zhu B, Qu S, Xu M. **Critical review of global plastics stock and flow data**. *J Ind Ecol* (2021.0) **25** 1300-1317. DOI: 10.1111/jiec.13125 49. Wang C, Zhao L, Lim MK, Chen WQ, Sutherland JW. **Structure of the global plastic waste trade network and the impact of China’s import ban**. *Resour Conserv Recycl* (2020.0) **153** 104591. DOI: 10.1016/j.resconrec.2019.104591 50. Vince J, Hardesty BD. **Plastic pollution challenges in marine and coastal environments: from local to global governance**. *Restor Ecol* (2017.0) **25** 123-128. DOI: 10.1111/rec.12388 51. Hatzonikolakis Y, Giakoumi S, Raitsos DE. **Quantifying transboundary plastic pollution in marine protected areas across the Mediterranean Sea**. *Front Mar Sci* (2022.0) **8** 762235. DOI: 10.3389/fmars.2021.762235 52. 52World Health Organization (WHO). Global Health Risks: Mortality and Burden of Disease Attributable to Selected Major Risks. WHO Press; 2009. Accessed February 10, 2023. https://apps.who.int/iris/handle/10665/44203.. *Global Health Risks: Mortality and Burden of Disease Attributable to Selected Major Risks* (2009.0) 53. Thompson RC, Swan SH, Moore CJ, vom Saal FS. **Our plastic age**. *Philos Trans R Soc B Biol Sci* (2009.0) **364** 1973-1976. DOI: 10.1098/rstb.2009.0054 54. Yearsley VE, Couzens EG. *Plastics* (1945.0) 55. Andrady AL, Neal MA. **Applications and societal benefits of plastics**. *Philos Trans R Soc Lond B Biol Sci* (2009.0) **364** 1977-1984. DOI: 10.1098/rstb.2008.0304 56. Thompson RC, Pahl S. *Plastics and the Environment* (2018.0) 177-187. DOI: 10.1039/9781788013314-00177 57. Lau WWY, Shiran Y, Bailey RM. **Evaluating scenarios toward zero plastic pollution**. *Science* (2020.0) **369** 1455-1461. DOI: 10.1126/science.aba9475 58. Napper IE, Thompson RC. **Plastic debris in the marine environment: history and future challenges**. *Glob Chall* (2020.0) **4** 1900081. DOI: 10.1002/gch2.201900081 59. Zalasiewicz J, Williams M, Steffen W, Crutzen P. **The new world of the Anthropocene**. *Environ Sci Technol* (2010.0) **44** 2228-2231. DOI: 10.1021/es903118j 60. Thompson RC, Pahl S, Sembiring E. **Plastics treaty – research must inform action**. *Nature* (2022.0) **608** 472-472. DOI: 10.1038/d41586-022-02201-0 61. 61United Nations Environment Assembly of the United Nations Environment Programme (UNEP). Fifth Meeting of the Conference of the Parties to the Basel Convention. United Nations Environment Programme (UNEP). 1999; 1–111. Accessed January 24, 2023. http://www.basel.int/TheConvention/ConferenceoftheParties/Meetings/COP5/tabid/6150/Default.aspx.. (1999.0) 1-111 62. Thompson RC, Moore CJ, vom Saal FS, Swan SH. **Plastics, the environment and human health: current consensus and future trends**. *Philos Trans R Soc B Biol Sci* (2009.0) **364** 2153-2166. DOI: 10.1098/rstb.2009.0053 63. 63Plastics Europe. Plastics – the Facts 2020: An Analysis of European Plastics Production, Demand and Waste Data. Plastics Europe and EPRO. 2020; 64. Accessed December 28, 2022. https://plasticseurope.org/knowledge-hub/plastics-the-facts-2020/.. *Plastics – the Facts 2020: An Analysis of European Plastics Production, Demand and Waste Data* (2020.0) **64** 64. MacLeod M, Arp HPH, Tekman MB, Jahnke A. **The global threat from plastic pollution**. *Science* (2021.0) **373** 61-65. DOI: 10.1126/science.abg5433 65. Charles D, Kimman L, Saran N. *The Plastic Waste Makers Index* (2021.0) 66. 66Beyond Plastics. The New Coal: Plastics & Climate Change. Beyond Plastics at Bennington College. 2021; 27. Accessed October 9, 2022. https://static1.squarespace.com/static/5eda91260bbb7e7a4bf528d8/t/616ef29221985319611a64e0/1634661022294/REPORT_The_New-Coal_Plastics_and_Climate-Change_10-21-2021.pdf.. *The New Coal: Plastics & Climate Change* (2021.0) **27** 67. 67Metal Packaging Europe. Aluminium beverage can recycling remains at a high 76% in 2019. Metal Packaging Europe. Published December 16, 2021. Accessed December 11, 2022. https://www.metalpackagingeurope.org/article/aluminium-beverage-can-recycling-remains-high-76-2019. 68. 68European Paper Recycling Council (EPCR). European Declaration on Paper Recycling 2021–2030. European Paper Recycling Council. 2022; 1–12. Accessed December 11, 2022. https://www.cepi.org/wp-content/uploads/2022/06/EPRC-22-010.pdf.. *European Declaration on Paper Recycling 2021–2030* (2022.0) 1-12 69. Tiseo I 70. Stubbins A, Law KL, Muñoz SE, Bianchi TS, Zhu L. **Plastics in the Earth system**. *Science* (2021.0) **373** 51-55. DOI: 10.1126/science.abb0354 71. Steffen W, Richardson K, Rockström J. **Planetary boundaries: guiding human development on a changing planet**. *Science* (2015.0) **347** 1259855. DOI: 10.1126/science.1259855 72. Bar-On YM, Phillips R, Milo R. **The biomass distribution on Earth**. *Proc Natl Acad Sci* (2018.0) **115** 6506-6511. DOI: 10.1073/pnas.1711842115 73. 73Center for International Environmental Law (CIEL). Fueling Plastics: Fossils, Plastics, & Petrochemical Feedstocks. Center for International Environmental Law (CIEL). 2017: 5. Accessed October 9, 2022. https://www.ciel.org/wp-content/uploads/2017/09/Fueling-Plastics-Fossils-Plastics-Petrochemical-Feedstocks.pdf.. *Fueling Plastics: Fossils, Plastics, & Petrochemical Feedstocks* **2017** 5 74. **Energy Information Administration (EIA). U.S. refinery net production. Petroleum & other liquids** 75. **U.S. crude oil production will increase to new records in 2023 and 2024**. *Today in Energy* 76. 76British Petroleum (BP p.l.c.). BP Energy Outlook 2019. 2019; 73. Accessed February 8, 2023. https://www.bp.com/content/dam/bp/business-sites/en/global/corporate/pdfs/energy-economics/energy-outlook/bp-energy-outlook-2019.pdf.. *BP Energy Outlook 2019* (2019.0) **73** 77. Chen J, Liu G, Kang Y. **Coal utilization in China: environmental impacts and human health**. *Environ Geochem Health* (2014.0) **36** 735-753. DOI: 10.1007/s10653-013-9592-1 78. O’Rourke D, Connolly S. **Just oil? The distribution of environmental and social impacts of oil production and consumption**. *Annu Rev Environ Resour* (2003.0) **28** 587-617. DOI: 10.1146/annurev.energy.28.050302.105617 79. Fink JK. *Oil Field Chemicals* (2003.0). DOI: 10.1016/B978-0-7506-7703-5.X5025-8 80. Solvay SA 81. **Barium sulfate. American Chemical Society: Chemistry for Life** 82. **Environmental contaminants – Environment Norway** 83. Raimi MO, Sawyerr OH, Ezekwe CI, Salako G. **Many oil wells, one evil: comprehensive assessment of toxic metals concentration, seasonal variation and human health risk in drinking water quality in areas surrounding crude oil exploration facilities in Rivers state, Nigeria**. DOI: 10.1101/2021.11.06.21266005 84. Eklund RL, Knapp LC, Sandifer PA, Colwell RC. **Oil spills and human health: contributions of the Gulf of Mexico Research Initiative**. *GeoHealth* (2019.0) **3** 391-406. DOI: 10.1029/2019GH000217 85. Omara M, Sullivan MR, Li X, Subramanian R, Robinson AL, Presto AA. **Methane emissions from conventional and unconventional natural gas production sites in the Marcellus shale basin**. *Environ Sci Technol* (2016.0) **50** 2099-2107. DOI: 10.1021/acs.est.5b05503 86. Zhang P, Dougherty H, Su D, Trackemas J, Tulu B. **Influence of longwall mining on the stability of gas wells in chain pillars**. *Int J Min Sci Technol* (2020.0) **30** 3-9. DOI: 10.1016/j.ijmst.2019.12.005 87. Kassotis CD, Tillitt DE, Davis JW, Hormann AM, Nagel SC. **Estrogen and androgen receptor activities of hydraulic fracturing chemicals and surface and ground water in a drilling-dense region**. *Endocrinology* (2014.0) **155** 897-907. DOI: 10.1210/en.2013-1697 88. Kassotis CD, Vu DC, Vo PH. **Endocrine-disrupting activities and organic contaminants associated with oil and gas operations in Wyoming groundwater**. *Arch Environ Contam Toxicol* (2018.0) **75** 247-258. DOI: 10.1007/s00244-018-0521-2 89. Norris JQ, Turcotte DL, Moores EM, Brodsky EE, Rundle JB. **Fracking in tight shales: what is it, what does it accomplish, and what are its consequences?**. *Annu Rev Earth Planet Sci* (2016.0) **44** 321-351. DOI: 10.1146/annurev-earth-060115-012537 90. Elliott EG, Ettinger AS, Leaderer BP, Bracken MB, Deziel NC. **A systematic evaluation of chemicals in hydraulic-fracturing fluids and wastewater for reproductive and developmental toxicity**. *J Expo Sci Environ Epidemiol* (2017.0) **27** 90-99. DOI: 10.1038/jes.2015.81 91. Barati R, Liang JT. **A review of fracturing fluid systems used for hydraulic fracturing of oil and gas wells**. *J Appl Polym Sci* (2014.0) **131**. DOI: 10.1002/app.40735 92. Moss J, Coram A, Blashki G. *Is Fracking Good for Your Health? An Analysis of the Impacts of Unconventional Gas on Health and Climate* (2013.0) 93. Colborn T, Kwiatkowski C, Schultz K, Bachran M. **Natural gas operations from a public health perspective**. *Hum Ecol Risk Assess Int J* (2011.0) **17** 1039-1056. DOI: 10.1080/10807039.2011.605662 94. Adgate JL, Goldstein BD, McKenzie LM. **Potential public health hazards, exposures and health effects from unconventional natural gas development**. *Environ Sci Technol* (2014.0) **48** 8307-8320. DOI: 10.1021/es404621d 95. Kovats S, Depledge M, Haines A. **The health implications of fracking**. *The Lancet* (2014.0) **383** 757-758. DOI: 10.1016/S0140-6736(13)62700-2 96. Howarth RW. **Methane emissions and climatic warming risk from hydraulic fracturing and shale gas development: implications for policy**. *Energy Emiss Control Technol* (2015.0) **3** 45-54. DOI: 10.2147/EECT.S61539 97. Srebotnjak T, Rotkin-Ellman M. *Fracking Fumes: Air Pollution from Fracking Threatens Public Health and Communities* (2014.0) 1-12 98. Colborn T, Schultz K, Herrick L, Kwiatkowski C. **An exploratory study of air quality near natural gas operations**. *Hum Ecol Risk Assess Int J* (2014.0) **20** 86-105. DOI: 10.1080/10807039.2012.749447 99. McKenzie LM, Witter RZ, Newman LS, Adgate JL. **Human health risk assessment of air emissions from development of unconventional natural gas resources**. *Sci Total Environ* (2012.0) **424** 79-87. DOI: 10.1016/j.scitotenv.2012.02.018 100. **Compliance with eight-hour ozone standard** 101. Rappazzo KM, Nichols JL, Rice RB, Luben TJ. **Ozone exposure during early pregnancy and preterm birth: a systematic review and meta-analysis**. *Environ Res* (2021.0) **198** 111317. DOI: 10.1016/j.envres.2021.111317 102. Zhang Y, Ma Y, Feng F. **Respiratory mortality associated with ozone in China: a systematic review and meta-analysis**. *Environ Pollut* (2021.0) **280** 116957. DOI: 10.1016/j.envpol.2021.116957 103. Sun HZ, Yu P, Lan C. **Cohort-based long-term ozone exposure-associated mortality risks with adjusted metrics: a systematic review and meta-analysis**. *The Innovation* (2022.0) **3** 100246. DOI: 10.1016/j.xinn.2022.100246 104. Orellano P, Reynoso J, Quaranta N, Bardach A, Ciapponi A. **Short-term exposure to particulate matter (PM10 and PM2.5), nitrogen dioxide (NO2), and ozone (O3) and all-cause and cause-specific mortality: systematic review and meta-analysis**. *Environ Int* (2020.0) **142** 105876. DOI: 10.1016/j.envint.2020.105876 105. Kondash A, Vengosh A. **Water footprint of hydraulic fracturing**. *Environ Sci Technol Lett* (2015.0) **2** 276-280. DOI: 10.1021/acs.estlett.5b00211 106. Luek JL, Gonsior M. **Organic compounds in hydraulic fracturing fluids and wastewaters: A review**. *Water Res* (2017.0) **123** 536-548. DOI: 10.1016/j.watres.2017.07.012 107. Tao Z, Liu C, He Q, Chang H, Ma J. **Detection and treatment of organic matters in hydraulic fracturing wastewater from shale gas extraction: a critical review**. *Sci Total Environ* (2022.0) **824** 153887. DOI: 10.1016/j.scitotenv.2022.153887 108. Kargbo DM, Wilhelm RG, Campbell DJ. **Natural gas plays in the marcellus shale: challenges and potential opportunities**. *Environ Sci Technol* (2010.0) **44** 5679-5684. DOI: 10.1021/es903811p 109. Schmidt CW. **Blind rush? Shale gas boom proceeds amid human health questions**. *Environ Health Perspect* (2011.0) **119** A348-A353. DOI: 10.1289/ehp.119-a348 110. Sun Y, Wang D, Tsang DCW, Wang L, Ok YS, Feng Y. **A critical review of risks, characteristics, and treatment strategies for potentially toxic elements in wastewater from shale gas extraction**. *Environ Int* (2019.0) **125** 452-469. DOI: 10.1016/j.envint.2019.02.019 111. Denchak M 112. Ankathi S, Lu Z, Zaimes GG, Hawkins T, Gan Y, Wang M. **Greenhouse gas emissions from the global transportation of crude oil: current status and mitigation potential**. *J Ind Ecol* 1-12. DOI: 10.1111/jiec.13262 113. Vidmar P, Perkovič M. **Safety assessment of crude oil tankers**. *Saf Sci* (2018.0) **105** 178-191. DOI: 10.1016/j.ssci.2018.02.009 114. 114United Nation Conference on Trade and Development (UNCTAD). Review of Maritime Transport 2019. United Nations Publications; 2019. https://unctad.org/system/files/official-document/rmt2019_en.pdf.. *Review of Maritime Transport 2019* (2019.0) 115. Wang J, Zhou Y, Zhuang L, Shi L, Zhang S. **Study on the critical factors and hot spots of crude oil tanker accidents**. *Ocean Coast Manag* (2022.0) **217** 106010. DOI: 10.1016/j.ocecoaman.2021.106010 116. Burclaff N. **Library of Congress Research Guides: Oil and gas industry: modes of transportation** 117. Little DI, Sheppard SRJ, Hulme D. **A perspective on oil spills: what we should have learned about global warming**. *Ocean Coast Manag* (2021.0) **202** 105509. DOI: 10.1016/j.ocecoaman.2020.105509 118. Brekke C, Solberg AHS. **Oil spill detection by satellite remote sensing**. *Remote Sens Environ* (2005.0) **95** 1-13. DOI: 10.1016/j.rse.2004.11.015 119. Laffon B, Pásaro E, Valdiglesias V. **Effects of exposure to oil spills on human health: updated review**. *J Toxicol Environ Health Part B* (2016.0) **19** 105-128. DOI: 10.1080/10937404.2016.1168730 120. Kollo M, Laanearu J, Tabri K. **Hydraulic modelling of oil spill through submerged orifices in damaged ship hulls**. *Ocean Eng* (2017.0) **130** 385-397. DOI: 10.1016/j.oceaneng.2016.11.032 121. Simecek-Beatty D, Lehr WJ. **Extended oil spill spreading with Langmuir circulation**. *Mar Pollut Bull* (2017.0) **122** 226-235. DOI: 10.1016/j.marpolbul.2017.06.047 122. Chen J, Zhang W, Wan Z, Li S, Huang T, Fei Y. **Oil spills from global tankers: status review and future governance**. *J Clean Prod* (2019.0) **227** 20-32. DOI: 10.1016/j.jclepro.2019.04.020 123. Aguilera F, Méndez J, Pásaro E, Laffon B. **Review on the effects of exposure to spilled oils on human health**. *J Appl Toxicol* (2010.0) **30** 291-301. DOI: 10.1002/jat.1521 124. Nwaoha TC, Yang Z, Wang J, Bonsall S. **Adoption of new advanced computational techniques to hazards ranking in LNG carrier operations**. *Ocean Eng* (2013.0) **72** 31-44. DOI: 10.1016/j.oceaneng.2013.06.010 125. Betteridge S. **Modelling large LNG pool fires on water**. *J Loss Prev Process Ind* (2018.0) **56** 46-56. DOI: 10.1016/j.jlp.2018.08.008 126. **Safety data sheet: liquefied natural gas (LNG)**. 11 127. Bush Trevino M, Pierce MA. *Encyclopedia of Energy* (2013.0) 128. Liu H 129. 129American Petroleum Institute. State of pipelines. American Petroleum Institute. Published 2021. Accessed September 30, 2022. https://energyinfrastructure.org/pipeline/state-of-pipelines. 130. Furchtgott-Roth D, Green KP. *Intermodal Safety in the Transport of Oil* (2013.0) **31** 131. 131FracTracker Alliance. An introduction to oil and gas pipelines. FracTracker Alliance. Published June 14, 2016. Accessed September 30, 2022. https://www.fractracker.org/2016/06/introduction-oil-gas-pipelines/. 132. **List of classifications – IARC monographs on the identification of carcinogenic hazards to humans** 133. Nethery RC, Mealli F, Dominici F. **Estimating population average causal effects in the presence of non-overlap: the effect of natural gas compressor station exposure on cancer mortality**. *Ann Appl Stat* (2019.0) **13** 1242-1267. DOI: 10.1214/18-AOAS1231 134. Peekema RM. **Causes of natural gas pipeline explosive ruptures**. *J Pipeline Syst Eng Pract* (2013.0) **4** 74-80. DOI: 10.1061/(ASCE)PS.1949-1204.0000116 135. Brunet R. **Gastown’s explosive potential: a lack of anti-corrosion protection increases the risk of deadly gas leaks**. *Rep Newsmag Br Columbia Ed* (1997.0) **8** 17 136. Willey RJ. **Insights to the Columbia gas explosions Lawrence and North Andover, MA September 13, 2018**. *Process Saf Prog* (2021.0) **40**. DOI: 10.1002/prs.12195 137. Carlson LC, Rogers TT, Kamara TB. **Petroleum pipeline explosions in sub-Saharan Africa: a comprehensive systematic review of the academic and lay literature**. *Burns* (2014.0) **41** 497-501. DOI: 10.1016/j.burns.2014.08.013 138. Frittelli J, Andrews A, Parfomak PW, Pirog R, Ramseur JL, Ratner M. *U.S. Rail Transportation of Crude Oil: Background and Issues for Congress* (2014.0) **25** 139. 139Great Lakes Commission. Issue Brief 3: Crude Oil Transport: Risks and Impacts. 2015; 1–13. Accessed September 30, 2022. https://www.glc.org/wp-content/uploads/Oil-Transportation-IssueBrief3-2015.pdf.. *Issue Brief 3: Crude Oil Transport: Risks and Impacts* (2015.0) 1-13 140. Clay K, Jha A, Muller N, Walsh R. **External costs of transporting petroleum products: evidence from shipments of crude oil from North Dakota by pipelines and rail**. *Energy J* (2019.0) **40** 55-72. DOI: 10.5547/01956574.40.1.kcla 141. Liu X. **Risk analysis of transporting crude oil by rail: methodology and decision support system**. *Transp Res Rec* (2016.0) **2547** 57-65. DOI: 10.3141/2547-09 142. Généreux M, Petit G, Maltais D. **The public health response during and after the Lac-Mégantic train derailment tragedy: a case study**. *Disaster Health* (2014.0) **2** 113-120. DOI: 10.1080/21665044.2014.1103123 143. Iranitalab A, Khattak A, Thompson E. **Statistical modeling of types and consequences of rail-based crude oil release incidents in the United States**. *Reliab Eng Syst Saf* (2019.0) **185** 232-239. DOI: 10.1016/j.ress.2018.12.009 144. Schneller AJ, Smemo KA, Mangan E, Munisteri C, Hobbs C, MacKay C. **Crude oil transportation by rail in Saratoga county, New York: Public perceptions of technological risk, state responses, and policy**. *Risk Hazards Crisis Public Policy* (2020.0) **11** 377-410. DOI: 10.1002/rhc3.12200 145. van der Beek AJ. **World at work: truck drivers**. *Occup Environ Med* (2012.0) **69** 291-295. DOI: 10.1136/oemed-2011-100342 146. Gangadhari RK, Murty S, Khanzode V. **Analysis of accidents involving petroleum tankers and their consequences in India**. *Process Saf Prog* (2021.0) **40**. DOI: 10.1002/prs.12154 147. 147India Brand Equity Foundation. Plastic manufacturers and exporters in India. India Brand Equity Foundation. Published July 2022. Accessed November 16, 2022. https://www.ibef.org/exports/plastic-industry-india. 148. Geueke B. **Dossier: Non-intentionally added substances (NIAS)**. DOI: 10.5281/zenodo.1265331 149. Kato LS, Conte-Junior CA. **Safety of plastic food packaging: the challenges about non-intentionally added substances (NIAS) discovery, identification and risk assessment**. *Polymers* (2021.0) **13** 2077. DOI: 10.3390/polym13132077 150. Nerin C, Alfaro P, Aznar M, Domeño C. **The challenge of identifying non-intentionally added substances from food packaging materials: a review**. *Anal Chim Acta* (2013.0) **775** 14-24. DOI: 10.1016/j.aca.2013.02.028 151. Horodytska O, Cabanes A, Fullana A. **Non-intentionally added substances (NIAS) in recycled plastics**. *Chemosphere* (2020.0) **251** 126373. DOI: 10.1016/j.chemosphere.2020.126373 152. 152United Nations Environment Programme (UNEP). Global Chemicals Outlook II – From Legacies to Innovative Solutions: Implementing the 2030 Agenda for Sustainable Development. United Nations Environment Programme (UNEP). 2019; 700. Accessed October 7, 2022. http://www.unep.org/resources/report/global-chemicals-outlook-ii-legacies-innovative-solutions.. *Global Chemicals Outlook II – From Legacies to Innovative Solutions: Implementing the 2030 Agenda for Sustainable Development* (2019.0) **700** 153. 153European Environment Agency (EEA). Chemicals for a Sustainable Future: Report of the EEA Scientific Committee Seminar. European Environment Agency; 2018. Accessed October 7, 2022. https://www.eea.europa.eu/about-us/governance/scientific-committee/reports/chemicals-for-a-sustainable-future.. *Chemicals for a Sustainable Future: Report of the EEA Scientific Committee Seminar* (2018.0) 154. Levi PG, Cullen JM. **Mapping global flows of chemicals: from fossil fuel feedstocks to chemical products**. *Environ Sci Technol* (2018.0) **52** 1725-1734. DOI: 10.1021/acs.est.7b04573 155. Tasker TL, Burgos WD, Piotrowski P. **Environmental and human health impacts of spreading oil and gas wastewater on roads**. *Environ Sci Technol* (2018.0) **52** 7081-7091. DOI: 10.1021/acs.est.8b00716 156. Nielsen TD, Bauer F. *Plastics and Sustainable Investments- An Information Brief for Investors* (2019.0) 157. Cui K 158. 158Encyclopaedia Britannica. Olefin. Brittanica. Published May 22, 2020. Accessed December 11, 2022. https://www.britannica.com/science/olefin. 159. Young B, Hawkins TR, Chiquelin C, Sun P, Gracida-Alvarez UR, Elgowainy A. **Environmental life cycle assessment of olefins and by-product hydrogen from steam cracking of natural gas liquids, naphtha, and gas oil**. *J Clean Prod* (2022.0) **359** 131884. DOI: 10.1016/j.jclepro.2022.131884 160. 160U.S. Energy Information Administration (EIA). Natural gas weekly update. Natural Gas. Published December 15, 2022. Accessed December 19, 2022. https://www.eia.gov/naturalgas/weekly/#itn-tabs-1. 161. de Place E 162. Heyl E 163. Dawood F, Anda M, Shafiullah GM. **Hydrogen production for energy: an overview**. *Int J Hydrog Energy* (2020.0) **45** 3847-3869. DOI: 10.1016/j.ijhydene.2019.12.059 164. Newborough M, Cooley G. **Developments in the global hydrogen market: the spectrum of hydrogen colours**. *Fuel Cells Bull* (2020.0). DOI: 10.1016/S1464-2859(20)30546-0 165. Lithner D, Larsson Å, Dave G. **Environmental and health hazard ranking and assessment of plastic polymers based on chemical composition**. *Sci Total Environ* (2011.0) **409** 3309-3324. DOI: 10.1016/j.scitotenv.2011.04.038 166. **Bisphenol-A (BPA) market – growth, trends, COVID-19 impact, and forecast (2022–2027)** 167. 167World Health Organization, Food and Agriculture Organization of the United Nations. Joint FAO/WHO Expert Meeting to Review Toxicological and Health Aspects of Bisphenol A: Final Report, Including Report of Stakeholder Meeting on Bisphenol A, 1–5 November 2010, Ottawa, Canada. 2011; 1–60. Accessed February 8, 2023. https://apps.who.int/iris/handle/10665/44624.. (2011.0) 1-60 168. Al-Malaika S, Pritchard G. *Plastic Additives: An A-Z Reference* (1998.0) 55-72. DOI: 10.1007/978-94-011-5862-6 169. Shahidi F. **Antioxidants in food and food antioxidants**. *Food Nahr* (2000.0) **44** 158-163. DOI: 10.1002/1521-3803(20000501)44:3<158::AID-FOOD158>3.0.CO;2-L 170. Hahladakis JN, Velis CA, Weber R, Iacovidou E, Purnell P. **An overview of chemical additives present in plastics: migration, release, fate and environmental impact during their use, disposal and recycling**. *J Hazard Mater* (2018.0) **344** 179-199. DOI: 10.1016/j.jhazmat.2017.10.014 171. 171World Health Organization (WHO). Dietary and Inhalation Exposure to Nano- and Microplastic Particles and Potential Implications for Human Health. World Health Organization; 2022. Accessed October 7, 2022. https://www.who.int/publications-detail-redirect/9789240054608.. *Dietary and Inhalation Exposure to Nano- and Microplastic Particles and Potential Implications for Human Health* (2022.0) 172. **Hazardous Chemical Information System (HCIS)** 173. 173European Chemicals Agency (ECHA). Classification & Labelling (C&L) Inventory. Information on Chemicals. Published 2022. Accessed November 2, 2022. https://echa.europa.eu/information-on-chemicals/cl-inventory-database. 174. 174Japanese National Institute of Technology and Evaluation (NITE). GHS classification results. Chemical Management. Published 2022. Accessed November 2, 2022. https://www.nite.go.jp/chem/english/ghs/ghs_download.html. 175. **Authorisation list** 176. 176European Chemicals Agency (ECHA). Candidate list of substances of very high concern for authorisation. Information on Chemicals. Published 2022. Accessed November 2, 2022. https://echa.europa.eu/candidate-list-table. 177. **PBT assessment list. Information on Chemicals** 178. 178European Chemicals Agency (ECHA). Endocrine disruptor assessment list. Information on Chemicals. Published October 20, 2022. Accessed November 2, 2022. https://echa.europa.eu/de/ed-assessment. 179. Ambrogi V, Carfagna C, Cerruti P, Marturano V, Jasso-Gastinel CF, Kenny JM. *Modification of Polymer Properties* (2017.0) 87-108. DOI: 10.1016/B978-0-323-44353-1.00004-X 180. de Wit CA. **An overview of brominated flame retardants in the environment**. *Chemosphere* (2002.0) **46** 583-624. DOI: 10.1016/S0045-6535(01)00225-9 181. Hendriks HS, Westerink RHS. **Neurotoxicity and risk assessment of brominated and alternative flame retardants**. *Neurotoxicol Teratol* (2015.0) **52** 248-269. DOI: 10.1016/j.ntt.2015.09.002 182. **Flame retardants. National Institute of Environmental Health Sciences: Health and Education: Environmental Health Topics** 183. Kimbrough RD, Jensen AA. *Halogenated Biphenyls, Terphenyls, Naphthalenes, Dibenzodioxins and Related Products* (2012.0) 184. Feiteiro J, Mariana M, Cairrão E. **Health toxicity effects of brominated flame retardants: from environmental to human exposure**. *Environ Pollut* (2021.0) **285** 117475. DOI: 10.1016/j.envpol.2021.117475 185. **Polybrominated biphenyls (PBBs) overview. ATSDR TOXZINE – PBBs** 186. **Technical fact sheet – polybrominated diphenyl ethers (PBDEs)** 187. **Stockholm Convention on persistent organic pollutants (POPS)** 188. Zhao J yi, Zhan Z xiang, Lu M juan, Tao F biao, Wu D, Gao H. **A systematic scoping review of epidemiological studies on the association between organophosphate flame retardants and neurotoxicity**. *Ecotoxicol Environ Saf* (2022.0) **243** 113973. DOI: 10.1016/j.ecoenv.2022.113973 189. AbouDonia M, Salama MBADM, Elgamal M, Elkholi I, Wang Q. **Organophosphorus flame retardants (OPFR): neurotoxicity**. *J Environ Health Sci JEHS* (2016.0) **2**. DOI: 10.15436/2378-6841.16.022 190. van der Veen I, de Boer J. **Phosphorus flame retardants: properties, production, environmental occurrence, toxicity and analysis**. *Chemosphere* (2012.0) **88** 1119-1153. DOI: 10.1016/j.chemosphere.2012.03.067 191. Blum A, Behl M, Birnbaum L. **Organophosphate ester flame retardants: are they a regrettable substitution for polybrominated diphenyl ethers?**. *Environ Sci Technol Lett* (2019.0) **6** 638-649. DOI: 10.1021/acs.estlett.9b00582 192. Yang J, Zhao Y, Li M, Du M, Li X, Li Y. **A review of a class of emerging contaminants: the classification, distribution, intensity of consumption, synthesis routes, environmental effects and expectation of pollution abatement to organophosphate flame retardants (OPFRs)**. *Int J Mol Sci* (2019.0) **20** 2874. DOI: 10.3390/ijms20122874 193. Godwin AD, Kutz M. *Applied Plastics Engineering Handbook (Second Edition)* (2017.0) 533-553. DOI: 10.1016/B978-0-323-39040-8.00025-0 194. Jamarani R, Erythropel HC, Nicell JA, Leask RL, Marić M. **How green is your plasticizer?**. *Polymers* (2018.0) **10** 834. DOI: 10.3390/polym10080834 195. Bui TT, Giovanoulis G, Cousins AP, Magnér J, Cousins IT, de Wit CA. **Human exposure, hazard and risk of alternative plasticizers to phthalate esters**. *Sci Total Environ* (2016.0) **541** 451-467. DOI: 10.1016/j.scitotenv.2015.09.036 196. **Plasticizers market – industry forecast, 2024** 197. Qadeer A, Kirsten KL, Ajmal Z, Jiang X, Zhao X. **Alternative plasticizers as emerging global environmental and health threat: another regrettable substitution?**. *Environ Sci Technol* (2022.0) **56** 1482-1488. DOI: 10.1021/acs.est.1c08365 198. 198Agency for Toxic Substances and Disease Registry (ATSDR), Di(2-Ethylhexyl)Phthalate (DEHP). Toxicological Profile for Di(2-Ethylhexyl)Phthalate (DEHP). 2022; 1–547. https://www.atsdr.cdc.gov/ToxProfiles/tp9.pdf.. *Toxicological Profile for Di(2-Ethylhexyl)Phthalate (DEHP)* (2022.0) 1-547 199. 199United States Consumer Product Safety Commission (US CPSC). Memorandum to Michael Babich from Kent Carlson Regarding the Toxicity Review of Di(2-Ethylhexyl) Phthalate (DEHP) Toxicity Review of Di(2-Ethylhexyl) Phthalate (DEHP). 2010; 1–329. Accessed February 8, 2023. https://www.cpsc.gov/s3fs-public/ToxicityReviewOfDEHP.pdf.. *Memorandum to Michael Babich from Kent Carlson Regarding the Toxicity Review of Di(2-Ethylhexyl) Phthalate (DEHP) Toxicity Review of Di(2-Ethylhexyl) Phthalate (DEHP)* (2010.0) 1-329 200. Murphy J. *The Additives for Plastics Handbook* (2001.0) 201. Wilkes CE, Summers JW, Daniels CA, Berard MT. *PVC Handbook* (2005.0) 202. Sampson J, de Korte D. **DEHP-plasticised PVC: relevance to blood services**. *Transfus Med* (2011.0) **21** 73-83. DOI: 10.1111/j.1365-3148.2010.01056.x 203. Wang YC, Chen HS, Long CY. **Possible mechanism of phthalates-induced tumorigenesis**. *Kaohsiung J Med Sci* (2012.0) **28** S22-S27. DOI: 10.1016/j.kjms.2012.05.006 204. Mughees M, Chugh H, Wajid S. **Mechanism of phthalate esters in the progression and development of breast cancer**. *Drug Chem Toxicol* (2022.0) **45** 1021-1025. DOI: 10.1080/01480545.2020.1802480 205. Wang Y, Qian H. **Phthalates and their impacts on human health**. *Healthcare* (2021.0) **9** 603. DOI: 10.3390/healthcare9050603 206. Giuliani A, Zuccarini M, Cichelli A, Khan H, Reale M. **Critical review on the presence of phthalates in food and evidence of their biological impact**. *Int J Environ Res Public Health* (2020.0) **17** 5655. DOI: 10.3390/ijerph17165655 207. Hlisníková H, Petrovičová I, Kolena B, Šidlovská M, Sirotkin A. **Effects and mechanisms of phthalates’ action on reproductive processes and reproductive health: a literature review**. *Int J Environ Res Public Health* (2020.0) **17** 6811. DOI: 10.3390/ijerph17186811 208. Mackintosh CE, Maldonado J, Hongwu J. **Distribution of phthalate esters in a marine aquatic food web: comparison to polychlorinated biphenyls**. *Environ Sci Technol* (2004.0) **38** 2011-2020. DOI: 10.1021/es034745r 209. Kim S, Kim Y, Moon HB. **Contamination and historical trends of legacy and emerging plasticizers in sediment from highly industrialized bays of Korea**. *Sci Total Environ* (2021.0) **765** 142751. DOI: 10.1016/j.scitotenv.2020.142751 210. Chen XH, Ma L, Hu YX. **Transcriptome profiling and pathway analysis of hepatotoxicity induced by tris (2-ethylhexyl) trimellitate (TOTM) in mice**. *Environ Toxicol Pharmacol* (2016.0) **41** 62-71. DOI: 10.1016/j.etap.2015.11.007 211. Mínguez-Alarcón L, Souter I, Chiu YH. **Urinary concentrations of cyclohexane-1,2-dicarboxylic acid monohydroxy isononyl ester, a metabolite of the non-phthalate plasticizer di(isononyl)cyclohexane-1,2-dicarboxylate (DINCH), and markers of ovarian response among women attending a fertility center**. *Environ Res* (2016.0) **151** 595-600. DOI: 10.1016/j.envres.2016.08.012 212. Sheikh IA, Beg MA. **Structural characterization of potential endocrine disrupting activity of alternate plasticizers di-(2-ethylhexyl) adipate (DEHA), acetyl tributyl citrate (ATBC) and 2,2,4-trimethyl 1,3-pentanediol diisobutyrate (TPIB) with human sex hormone-binding globulin**. *Reprod Toxicol* (2019.0) **83** 46-53. DOI: 10.1016/j.reprotox.2018.11.003 213. Groh KJ, Backhaus T, Carney-Almroth B. **Overview of known plastic packaging-associated chemicals and their hazards**. *Sci Total Environ* (2019.0) **651** 3253-3268. DOI: 10.1016/j.scitotenv.2018.10.015 214. 214European Chemicals Agency (ECHA). Mapping exercise – plastic additives initiative. European Chemicals Agency. Published 2020. https://echa.europa.eu/de/mapping-exercise-plastic-additives-initiative. 215. Heurung AR, Raju SI, Warshaw EM. **Benzophenones**. *Dermatitis®* (2014.0) **25** 3-10. DOI: 10.1097/DER.0000000000000025 216. Runde K, Castro G, Vike-Jonas K, González SV, Asimakopoulos AG, Arp HPH, Lee B.. **Occurrence and sorption behaviour of bisphenols and benzophenone UV-filters in e-waste plastic and vehicle fluff**. *J Hazard Mater* (2022.0) **426** 9. DOI: 10.1016/j.jhazmat.2021.127814 217. Wejnerowska G, Narloch I. **Determination of benzophenones in water and cosmetics samples: a comparison of solid-phase extraction and microextraction by packed sorbent methods**. *Molecules* (2021.0) **26** 6896. DOI: 10.3390/molecules26226896 218. Wang L, Asimakopoulos AG, Moon HB, Nakata H, Kannan K. **Benzotriazole, benzothiazole, and benzophenone compounds in indoor dust from the United States and East Asian countries**. *Environ Sci Technol* (2013.0) **47** 4752-4759. DOI: 10.1021/es305000d 219. Mao F, He Y, Gin KYH. **Occurrence and fate of benzophenone-type UV filters in aquatic environments: a review**. *Environ Sci Water Res Technol* (2019.0) **5** 209-223. DOI: 10.1039/C8EW00539G 220. Asimakopoulos AG, Wang L, Thomaidis NS, Kannan K. **Benzotriazoles and benzothiazoles in human urine from several countries: a perspective on occurrence, biotransformation, and human exposure**. *Environ Int* (2013.0) **59** 274-281. DOI: 10.1016/j.envint.2013.06.007 221. Sakuragi Y, Takada H, Sato H. **An analytical survey of benzotriazole UV stabilizers in plastic products and their endocrine-disrupting potential via human estrogen and androgen receptors**. *Sci Total Env* (2021.0) **800** 149374. DOI: 10.1016/j.scitotenv.2021.149374 222. **Eighteenth meeting of the Persistent Organic Pollutants Review Committee (POPRC.18)** 223. **FCCmigex Database** 224. Porfyris AD, Luyt AS, Gasmi S. **Enhancing the UV/heat stability of LLDPE irrigation pipes via different stabilizer formulations**. *SPE Polym* (2021.0) **2** 336-350. DOI: 10.1002/pls2.10055 225. Yousif E, Haddad R. **Photodegradation and photostabilization of polymers, especially polystyrene: review**. *SpringerPlus* (2013.0) **2** 398. DOI: 10.1186/2193-1801-2-398 226. Sun Y, Xie H, Du B. **Widening the lens on UV filters: newfound ubiquity of triazine UV filters in breast milk from South China and implications for augmented “cocktail” exposure in infants**. *Environ Sci Technol Lett* (2022.0) **9** 844-850. DOI: 10.1021/acs.estlett.2c00535 227. Li J, Zhou C, Xu S, Shen L. **Investigation of hindered phenol antioxidant effects on the aging performance of cross-linked LDPE in the presence of copper**. *Sci Rep* (2020.0) **10** 10189. DOI: 10.1038/s41598-020-67131-1 228. Carloni P, Greci L, Mar’in A, Stipa P. **Aromatic secondary amines as antioxidants for polyolefins: part 1—9,10-dihydroacridine (acridan) derivatives**. *Polym Degrad Stab* (1994.0) **44** 201-209. DOI: 10.1016/0141-3910(94)90165-1 229. Schwetlick K, Scott G. *Mechanisms of Polymer Degradation and Stabilisation* (1990.0) 23-60. DOI: 10.1007/978-94-011-3838-3_2 230. Loyo-Rosales JE, Rosales-Rivera GC, Lynch AM, Rice CP, Torrents A. **Migration of nonylphenol from plastic containers to water and a milk surrogate**. *J Agric Food Chem* (2004.0) **52** 2016-2020. DOI: 10.1021/jf0345696 231. Kulich DM, Reid Shelton J. **Organosulfur antioxidants: mechanisms of action**. *Polym Degrad Stab* (1991.0) **33** 397-410. DOI: 10.1016/0141-3910(91)90085-6 232. Schmidtkunz C, Küpper K, Weber T, Leng G, Kolossa-Gehring M. **A biomonitoring study assessing the exposure of young German adults to butylated hydroxytoluene (BHT)**. *Int J Hyg Environ Health* (2020.0) **228** 113541. DOI: 10.1016/j.ijheh.2020.113541 233. 233South East Europe PVC Forum. Stabilisers. PVC Forum South East Europe. Published 2010. Accessed December 11, 2022. http://www.seepvcforum.com/en/content/17-stabilisers. 234. Whalen MM, Loganathan BG, Kannan K. **Immunotoxicity of environmentally relevant concentrations of butyltins on human natural killer cells in vitro**. *Environ Res* (1999.0) **81** 108-116. DOI: 10.1006/enrs.1999.3968 235. Ruff A, Lewis M, Whalen M. **Organotin and organochlorine toxicants activate key translational regulatory proteins in human immune cells**. *Arch Toxicol*. DOI: 10.1007/s00204-022-03413-z 236. Okoro HK, Fatoki OS, Adekola FA, Ximba BJ, Snyman RG, Opeolu B, Whitacre DM. *Reviews of Environmental Contamination and Toxicology Volume 213* (2011.0) **213** 27-54. DOI: 10.1007/978-1-4419-9860-6_2 237. Sewwandi M, Hettithanthri O, Egodage SM, Amarathunga AAD, Vithanage M. **Unprecedented marine microplastic contamination from the X-Press Pearl container vessel disaster**. *Sci Total Environ* (2022.0) **828** 154374. DOI: 10.1016/j.scitotenv.2022.154374 238. Dematteo R, Keith MM, Brophy JT. **Chemical exposures of women workers in the plastics industry with particular reference to breast cancer and reproductive hazards**. *NEW Solut J Environ Occup Health Policy* (2013.0) **22** 427-448. DOI: 10.2190/NS.22.4.d 239. 239PlasticsEurope. Plastics- the facts 2016. An analysis of of European plastics production, demand and waste data. Presented at: 2016; Brussels, Belgium. https://plasticseurope.org/wp-content/uploads/2021/10/2016-Plastic-the-facts.pdf. 240. Chamas A, Moon H, Zheng J. **Degradation rates of plastics in the environment**. *ACS Sustain Chem Eng* (2020.0) **8** 3494-3511. DOI: 10.1021/acssuschemeng.9b06635 241. 241American Chemistry Council. The Resin Review – 2019. American Chemistry Council. 2019; 1–80. https://store.americanchemistry.com/products/the-resin-review-2019-electronic-version.. *The Resin Review – 2019* (2019.0) 1-80 242. Sandin G, Roos S, Johansson M. *Environmental Impact of Textile Fibers – What We Know and What We Don’t Know: Fiber Bible Part 2* (2019.0) 243. Brigden K, Labunska I, House E, Santillo D, Johnston P. *Hazardous Chemicals in Branded Textile Products on Sale in 27 Countries During 2012* (2012.0) **45** 244. 244International Agency for Research on Cancer (IARC). Some Aromatic Amines, Organic Dyes, and Related Exposures. Vol 99. IARC Press; 2010. Accessed December 11, 2022. https://monographs.iarc.who.int/wp-content/uploads/2018/06/mono99.pdf.. *Some Aromatic Amines, Organic Dyes, and Related Exposures* (2010.0) **99** 245. Freudenthal RI, Stephens E, Anderson DP. **Determining the potential of aromatic amines to induce cancer of the urinary bladder**. *Int J Toxicol* (1999.0) **18** 353-359. DOI: 10.1080/109158199225260 246. Ryberg K, Goossens A, Isaksson M. **Is contact allergy to disperse dyes and related substances associated with textile dermatitis?**. *Br J Dermatol* (2009.0) **160** 107-115. DOI: 10.1111/j.1365-2133.2008.08953.x 247. Coman G, Blattner CM, Blickenstaff NR, Andersen R, Maibach HI. **Textile allergic contact dermatitis: current status**. *Rev Environ Health* (2014.0) **29** 163-168. DOI: 10.1515/reveh-2014-0061 248. Rovira J, Domingo JL. **Human health risks due to exposure to inorganic and organic chemicals from textiles: a review**. *Environ Res* (2019.0) **168** 62-69. DOI: 10.1016/j.envres.2018.09.027 249. Liu W, Xue J, Kannan K. **Occurrence of and exposure to benzothiazoles and benzotriazoles from textiles and infant clothing**. *Sci Total Environ* (2017.0) **592** 91-96. DOI: 10.1016/j.scitotenv.2017.03.090 250. Pedersen H, Hartmann J. *Toxic Childrenswear by Disney – A Worldwide Investigation of Hazardous Chemicals in Disney Clothes* (2004.0) 251. Xue J, Liu W, Kannan K. **Bisphenols, benzophenones, and bisphenol A diglycidyl ethers in textiles and infant clothing**. *Environ Sci Technol* (2017.0) **51** 5279-5286. DOI: 10.1021/acs.est.7b00701 252. **Bisphenol A** 253. Negev M, Berman T, Reicher S, Sadeh M, Ardi R, Shammai Y. **Concentrations of trace metals, phthalates, bisphenol A and flame-retardants in toys and other children’s products in Israel**. *Chemosphere* (2018.0) **192** 217-224. DOI: 10.1016/j.chemosphere.2017.10.132 254. Novick RM, Nelson ML, McKinley MA, Anderson GL, Keenan JJ. **The effect of clothing care activities on textile formaldehyde content**. *J Toxicol Environ Health A* (2013.0) **76** 883-893. DOI: 10.1080/15287394.2013.821439 255. Prata JC. **Airborne microplastics: consequences to human health?**. *Environ Pollut* (2018.0) **234** 115-126. DOI: 10.1016/j.envpol.2017.11.043 256. Agarwal S, Gupta RK, Kutz M. *Applied Plastics Engineering Handbook (Second Edition)* (2017.0) 635-649. DOI: 10.1016/B978-0-323-39040-8.00030-4 257. Markarian J. **PVC additives – what lies ahead?**. *Plast Addit Compd* (2007.0) **9** 22-25. DOI: 10.1016/S1464-391X(07)70153-8 258. Sadiki AD, Williams DT. **A study on organotin levels in Canadian drinking water distributed through PVC pipesa**. *Chemosphere* (1999.0) **38** 1541-1548. DOI: 10.1016/S0045-6535(98)00374-9 259. Allsopp M, Santillo D, Johnston P. *Hazardous Chemicals in PVC Flooring* (2000.0) 1-10 260. Stringer R, Labunska I, Santillo D, Johnston P, Siddorn J, Stephenson A. **Concentrations of phthalate esters and identification of other additives in PVC children’s toys**. *Environ Sci Pollut Res* (2000.0) **7** 27-36. DOI: 10.1065/espr199910.007 261. Shu H, Jönsson BAG, Gennings C, Lindh CH, Nånberg E, Bornehag CG. **PVC flooring at home and uptake of phthalates in pregnant women**. *Indoor Air* (2019.0) **29** 43-54. DOI: 10.1111/ina.12508 262. Shu H, Jönsson BA, Larsson M, Nånberg E, Bornehag CG. **PVC flooring at home and development of asthma among young children in Sweden, a 10-year follow-up**. *Indoor Air* (2014.0) **24** 227-235. DOI: 10.1111/ina.12074 263. Ye X, Wang P, Wu Y, Zhou Y, Sheng Y, Lao K. **Microplastic acts as a vector for contaminants: the release behavior of dibutyl phthalate from polyvinyl chloride pipe fragments in water phase**. *Environ Sci Pollut Res* (2020.0) **27** 42082-42091. DOI: 10.1007/s11356-020-10136-0 264. 264Healthy Building Network. Healthy Building Network homepage. Healthy Building Network. Published 2022. Accessed December 16, 2022. https://healthybuilding.net/. 265. Koniuszewska AG, Kaczmar JW. **Application of polymer based composite materials in transportation**. *Prog Rubber Plast Recycl Technol* (2016.0) **32** 1-24. DOI: 10.1177/147776061603200101 266. Wang X, Jiang M, Zhou Z, Gou J, Hui D. **3D printing of polymer matrix composites: a review and prospective**. *Compos Part B Eng* (2017.0) **110** 442-458. DOI: 10.1016/j.compositesb.2016.11.034 267. Mathijsen D. **What does the future hold for composites in transportation markets?**. *Reinf Plast* (2017.0) **61** 41-46. DOI: 10.1016/j.repl.2016.12.061 268. Salthammer T. **Microplastics and their additives in the indoor environment**. *Angew Chem Int Ed* (2022.0) **61**. DOI: 10.1002/anie.202205713 269. 269Food and Agriculture Organization of the United Nations (FAO), United Nations Environment Programme (UNEP). Global Assessment of Soil Pollution – Summary for Policy Makers. Food and Agriculture Organization of the United Nations (FAO). 2021; 84. Accessed October 7, 2022. https://www.fao.org/documents/card/fr/c/CB4827EN/.. *Global Assessment of Soil Pollution – Summary for Policy Makers* (2021.0) **84** 270. Guo JJ, Huang XP, Xiang L. **Source, migration and toxicology of microplastics in soil**. *Environ Int* (2020.0) **137** 105263. DOI: 10.1016/j.envint.2019.105263 271. Scheurer M, Bigalke M. **Microplastics in Swiss floodplain soils**. *Environ Sci Technol* (2018.0) **52** 3591-3598. DOI: 10.1021/acs.est.7b06003 272. Meng M, Yang L, Wei B, Cao Z, Yu J, Liao X, Barbosa Fernando. **Plastic shed production systems: the migration of heavy metals from soil to vegetables and human health risk assessment**. *Ecotoxicol Environ Saf* (2021.0) **215** 112106. DOI: 10.1016/j.ecoenv.2021.112106 273. Vox G, Loisi RV, Blanco I, Mugnozza GS, Schettini E. **Mapping of agriculture plastic waste**. *Agric Agric Sci Procedia* (2016.0) **8** 583-591. DOI: 10.1016/j.aaspro.2016.02.080 274. **Global Plastic Watch** 275. Rillig MC, Lehmann A. **Microplastic in terrestrial ecosystems**. *Science* (2020.0) **368** 1430-1431. DOI: 10.1126/science.abb5979 276. van Kleunen M, Brumer A, Gutbrod L, Zhang Z. **A microplastic used as infill material in artificial sport turfs reduces plant growth**. *Plants People Planet* (2020.0) **2** 157-166. DOI: 10.1002/ppp3.10071 277. Li L, Luo Y, Li R. **Effective uptake of submicrometre plastics by crop plants via a crack-entry mode**. *Nat Sustain* (2020.0) **3** 929-937. DOI: 10.1038/s41893-020-0567-9 278. Zhang JY, Qiu LM, He J, Liao Y, Luo YM. **Occurrence and congeners specific of polychlorinated biphenyls in agricultural soils from Southern Jiangsu, China**. *J Environ Sci* (2007.0) **19** 338-342. DOI: 10.1016/S1001-0742(07)60055-2 279. Zhu ZC, Chen SJ, Zheng J. **Occurrence of brominated flame retardants (BFRs), organochlorine pesticides (OCPs), and polychlorinated biphenyls (PCBs) in agricultural soils in a BFR-manufacturing region of North China**. *Sci Total Environ* (2014.0) **481** 47-54. DOI: 10.1016/j.scitotenv.2014.02.023 280. Gorovtsov AV, Sazykin IS, Sazykina MA. **The influence of heavy metals, polyaromatic hydrocarbons, and polychlorinated biphenyls pollution on the development of antibiotic resistance in soils**. *Environ Sci Pollut Res* (2018.0) **25** 9283-9292. DOI: 10.1007/s11356-018-1465-9 281. Liao R, Jiang J, Li Y. **Distribution and leaching behavior of organophosphorus and brominated flame retardants in soil in Chengdu**. *Environ Sci Process Impacts* (2020.0) **22** 1295-1305. DOI: 10.1039/D0EM00106F 282. Wang R, Huang Y, Dong S, Wang P, Su X. **The occurrence of bisphenol compounds in animal feed plastic packaging and migration into feed**. *Chemosphere* (2021.0) **265** 129022. DOI: 10.1016/j.chemosphere.2020.129022 283. Walkinshaw C, Tolhurst TJ, Lindeque PK, Thompson R, Cole M. **Detection and characterisation of microplastics and microfibres in fishmeal and soybean meal**. *Mar Pollut Bull* (2022.0) **185** 114189. DOI: 10.1016/j.marpolbul.2022.114189 284. 284Health Care Without Harm. Measuring and Reducing Plastics in the Healthcare Sector. Health Care Without Harm; 2021. Accessed October 7, 2022. https://noharm-europe.org/sites/default/files/documents-files/6886/2021-09-23-measuring-and-reducing-plastics-in-the-healthcare-sector.pdf.. *Measuring and Reducing Plastics in the Healthcare Sector* (2021.0) 285. 285Health Care Without Harm. Plastics and Health: An Urgent Environmental, Climate, and Health Issue; 2022: 1–14. Accessed December 28, 2022. https://noharm-europe.org/sites/default/files/documents-files/7272/HCWH%20-%20Plastics%20and%20health.pdf.. *Plastics and Health: An Urgent Environmental, Climate, and Health Issue* **2022** 1-14 286. Patrício Silva AL, Prata JC, Walker TR. **Increased plastic pollution due to COVID-19 pandemic: challenges and recommendations**. *Chem Eng J* (2021.0) **405** 126683. DOI: 10.1016/j.cej.2020.126683 287. Genco M, Anderson-Shaw L, Sargis RM. **Unwitting accomplices: endocrine disruptors confounding clinical care**. *J Clin Endocrinol Metab* (2020.0) **105**. DOI: 10.1210/clinem/dgaa358 288. Gosetti F, Bolfi B, Robotti E. **Study of endocrine disrupting compound release from different medical devices through an on-line SPE UHPLC-MS/MS method**. *Anal Chim Acta* (2018.0) **1042** 141-154. DOI: 10.1016/j.aca.2018.07.028 289. Morton WJ, Muller CT, Goodwin N, Wilkes AR, Hall JE. **Investigation of phthalate release from tracheal tubes**. *Anaesthesia* (2013.0) **68** 377-381. DOI: 10.1111/anae.12083 290. Baker RWR. **Diethylhexyl phthalate as a factor in blood transfusion and haemodialysis** 291. Draviam EJ, Pearson KH, Kerkay J. **Human Metabolism of Bis(2-ethylhexyl)phthalate**. *Anal Lett* (1982.0) **15** 1729-1750. DOI: 10.1080/00032718208065150 292. Huygh J, Clotman K, Malarvannan G. **Considerable exposure to the endocrine disrupting chemicals phthalates and bisphenol-A in intensive care unit (ICU) patients**. *Environ Int* (2015.0) **81** 64-72. DOI: 10.1016/j.envint.2015.04.008 293. Van Vliet EDS, Reitano EM, Chhabra JS, Bergen GP, Whyatt RM. **A review of alternatives to di (2-ethylhexyl) phthalate-containing medical devices in the neonatal intensive care unit**. *J Perinatol* (2011.0) **31** 551-560. DOI: 10.1038/jp.2010.208 294. Bernard L, Masse M, Boeuf B. **Medical devices used in NICU: the main source of plasticisers’ exposure of newborns**. *Sci Total Environ* (2023.0) **858** 159994. DOI: 10.1016/j.scitotenv.2022.159994 295. von Rettberg H, Hannman T, Subotic U. **Use of di(2-ethylhexyl)phthalate–containing infusion systems increases the risk for cholestasis**. *Pediatrics* (2009.0) **124** 710-716. DOI: 10.1542/peds.2008-1765 296. Fischer CJ, Bickle Graz M, Muehlethaler V, Palmero D, Tolsa JF. **Phthalates in the NICU: is it safe?**. *J Paediatr Child Health* (2013.0) **49** e413-E419. DOI: 10.1111/jpc.12244 297. Jenkins R, Farnbach K, Iragorri S. **Elimination of intravenous di-2-ethylhexyl phthalate exposure abrogates most neonatal hypertension in premature infants with bronchopulmonary dysplasia**. *Toxics* (2021.0) **9** 75. DOI: 10.3390/toxics9040075 298. Golestanzadeh M, Riahi R, Kelishadi R. **Association of exposure to phthalates with cardiometabolic risk factors in children and adolescents: a systematic review and meta-analysis**. *Environ Sci Pollut Res* (2019.0) **26** 35670-35686. DOI: 10.1007/s11356-019-06589-7 299. Smith CM. **Origin and uses of primum non nocere—above all, do no harm!**. *J Clin Pharmacol* (2005.0) **45** 371-377. DOI: 10.1177/0091270004273680 300. **The age of plastic: from Parkesine to pollution** 301. 301Statista. Raw materials demand for medical disposables globally by type 2025. Statista. Accessed November 29, 2022. https://www.statista.com/statistics/689249/demand-of-raw-materials-for-medical-disposables-product-type-worldwide/. 302. McGain F, Burnham JP, Lau R, Aye L, Kollef MH, McAlister S. **The carbon footprint of treating patients with septic shock in the intensive care unit**. *Crit Care Resusc J Australas Acad Crit Care Med* (2018.0) **20** 304-312 303. **Estates Returns Information Collection summary page and dataset for ERIC 2019/20** 304. Hopkinson P, Smith R, Fleming L, Morrissey K, Hu X. **Single use plastic in healthcare must not become the new normal – The BMJ** 305. Burki T. **Global shortage of personal protective equipment**. *Lancet Infect Dis* (2020.0) **20** 785-786. DOI: 10.1016/S1473-3099(20)30501-6 306. 306Department of Health (UK). Environment and Sustainability: Health Technical Memorandum 07–01: Safe Management of Healthcare Waste. Department of Health; 2006. https://data.parliament.uk/DepositedPapers/Files/DEP2008-2426/DEP2008-2426.pdf.. *Environment and Sustainability: Health Technical Memorandum 07–01: Safe Management of Healthcare Waste* (2006.0) 307. **Climate change and health** 308. **NHS England » ‘The gloves are off’ campaign** 309. **Greener NHS » Delivering a net zero NHS** 310. Baensch-Baltruschat B, Kocher B, Stock F, Reifferscheid G. **Tyre and road wear particles (TRWP) – a review of generation, properties, emissions, human health risk, ecotoxicity, and fate in the environment**. *Sci Total Environ* (2020.0) **733** 137823. DOI: 10.1016/j.scitotenv.2020.137823 311. Brodin M, Norin H, Hanning AC, Persson C, Okcabol S. *Microplastics from Industrial Laundries: A Study of Laundry Effluents* **2018** 1-35 312. De Falco F, Di Pace E, Cocca M, Avella M. **The contribution of washing processes of synthetic clothes to microplastic pollution**. *Sci Rep* (2019.0) **9** 6633. DOI: 10.1038/s41598-019-43023-x 313. Wolfram J, Zhu M, Yang Y. **Safety of nanoparticles in medicine**. *Curr Drug Targets* (2015.0) **16** 1671-1681. DOI: 10.2174/1389450115666140804124808 314. Lombardo D, Kiselev MA, Caccamo MT. **Smart nanoparticles for drug delivery application: development of versatile nanocarrier platforms in biotechnology and nanomedicine**. *J Nanomater* (2019.0) **2019**. DOI: 10.1155/2019/3702518 315. Battistella C, Liang Y, Gianneschi NC. **Innovations in disease state responsive soft materials for targeting extracellular stimuli associated with cancer, cardiovascular disease, diabetes, and beyond**. *Adv Mater* (2021.0) **33** 2007504. DOI: 10.1002/adma.202007504 316. Bobo D, Robinson KJ, Islam J, Thurecht KJ, Corrie SR. **Nanoparticle-based medicines: a review of FDA-approved materials and clinical trials to date**. *Pharm Res* (2016.0) **33** 2373-2387. DOI: 10.1007/s11095-016-1958-5 317. Charles D, Kimman L. *The Plastic Waste Makers Index 2023* (2023.0) 318. Lebreton L, Royer SJ, Peytavin A, Strietman WJ, Smeding-Zuurendonk I, Egger M. **Industrialised fishing nations largely contribute to floating plastic pollution in the North Pacific subtropical gyre**. *Sci Rep* (2022.0) **12** 12666. DOI: 10.1038/s41598-022-16529-0 319. **Plastic waste in 2019** 320. Geyer R, Kuczenski B, Zink T, Henderson A. **Common misconceptions about recycling**. *J Ind Ecol* (2016.0) **20** 1010-1017. DOI: 10.1111/jiec.12355 321. Kijo-Kleczkowska A, Gnatowski A. **Recycling of plastic waste, with particular emphasis on thermal methods—review**. *Energies* (2022.0) **15** 2114. DOI: 10.3390/en15062114 322. King S, Hutchinson S, Boxall N. *Advanced Recycling Technologies to Address Australia’s Plastic Waste* (2021.0) 1-74. DOI: 10.25919/7stt-ke60 323. Cook E, Velis CA, Cottom JW. **Scaling up resource recovery of plastics in the emergent circular economy to prevent plastic pollution: assessment of risks to health and safety in the Global South**. *Waste Manag Res* (2022.0) **40** 1680-1707. DOI: 10.1177/0734242X221105415 324. Schyns ZOG, Shaver MP. **Mechanical recycling of packaging plastics: a review**. *Macromol Rapid Commun* (2021.0) **42** 2000415. DOI: 10.1002/marc.202000415 325. Tsai CJ, Chen ML, Chang KF, Chang FK, Mao IF. **The pollution characteristics of odor, volatile organochlorinated compounds and polycyclic aromatic hydrocarbons emitted from plastic waste recycling plants**. *Chemosphere* (2009.0) **74** 1104-1110. DOI: 10.1016/j.chemosphere.2008.10.041 326. He Z, Li G, Chen J, Huang Y, An T, Zhang C. **Pollution characteristics and health risk assessment of volatile organic compounds emitted from different plastic solid waste recycling workshops**. *Environ Int* (2015.0) **77** 85-94. DOI: 10.1016/j.envint.2015.01.004 327. Yamashita K, Yamamoto N, Mizukoshi A, Noguchi M, Ni Y, Yanagisawa Y. **Compositions of volatile organic compounds emitted from melted virgin and waste plastic pellets**. *J Air Waste Manag Assoc* (2009.0) **59** 273-278. DOI: 10.3155/1047-3289.59.3.273 328. Huang DY, Zhou SG, Hong W, Feng WF, Tao L. **Pollution characteristics of volatile organic compounds, polycyclic aromatic hydrocarbons and phthalate esters emitted from plastic wastes recycling granulation plants in Xingtan Town, South China**. *Atmos Environ* (2013.0) **71** 327-334. DOI: 10.1016/j.atmosenv.2013.02.011 329. Huang G, Xie J, Li T, Zhang P. **Worker health risk of heavy metals in pellets of recycled plastic: a skin exposure model**. *Int Arch Occup Environ Health* (2021.0) **94** 1581-1589. DOI: 10.1007/s00420-021-01727-6 330. Roosen M, Harinck L, Ügdüler S. **Deodorization of post-consumer plastic waste fractions: a comparison of different washing media**. *Sci Total Environ* (2022.0) **812** 152467. DOI: 10.1016/j.scitotenv.2021.152467 331. Lowe CN, Phillips KA, Favela KA. **Chemical characterization of recycled consumer products using suspect screening analysis**. *Environ Sci Technol* (2021.0) **55** 11375-11387. DOI: 10.1021/acs.est.1c01907 332. Maqsood T, Dai J, Zhang Y, Guang M, Li B. **Pyrolysis of plastic species: a review of resources and products**. *J Anal Appl Pyrolysis* (2021.0) **159** 105295. DOI: 10.1016/j.jaap.2021.105295 333. Paladino O, Moranda A. **Human health risk assessment of a pilot-plant for catalytic pyrolysis of mixed waste plastics for fuel production**. *J Hazard Mater* (2021.0) **405** 124222. DOI: 10.1016/j.jhazmat.2020.124222 334. 334European Union. Directive (EU) 2019/904 of the European Parliament and of the Council of 5 June 2019: On the Reduction of the Impact of Certain Plastic Products on the Environment; 2019. Accessed October 7, 2022. https://eur-lex.europa.eu/eli/dir/2019/904/oj.. *Directive (EU) 2019/904 of the European Parliament and of the Council of 5 June 2019: On the Reduction of the Impact of Certain Plastic Products on the Environment* (2019.0) 335. Brosché S, Strakova J, Bell L, Karlsson T. *Widespread Chemical Contamination of Recycled Plastic Pellets Globally* (2021.0) 336. Gerassimidou S, Lanska P, Hahladakis JN. **Unpacking the complexity of the PET drink bottles value chain: a chemicals perspective**. *J Hazard Mater* (2022.0) **430** 128410. DOI: 10.1016/j.jhazmat.2022.128410 337. Conesa JA, Nuñez SS, Ortuño N, Moltó J. **PAH and POP presence in plastic waste and recyclates: state of the art**. *Energies* (2021.0) **14** 3451. DOI: 10.3390/en14123451 338. Chen SJ, Ma YJ, Wang J, Chen D, Luo XJ, Mai BX. **Brominated flame retardants in children’s toys: concentration, composition, and children’s exposure and risk assessment**. *Environ Sci Technol* (2009.0) **43** 4200-4206. DOI: 10.1021/es9004834 339. Ionas AC, Dirtu AC, Anthonissen T, Neels H, Covaci A. **Downsides of the recycling process: harmful organic chemicals in children’s toys**. *Environ Int* (2014.0) **65** 54-62. DOI: 10.1016/j.envint.2013.12.019 340. Kajiwara N, Matsukami H, Malarvannan G, Chakraborty P, Covaci A, Takigami H. **Recycling plastics containing decabromodiphenyl ether into new consumer products including children’s toys purchased in Japan and seventeen other countries**. *Chemosphere* (2022.0) **289** 133179. DOI: 10.1016/j.chemosphere.2021.133179 341. Puype F, Samsonek J, Knoop J, Egelkraut-Holtus M, Ortlieb M. **Evidence of waste electrical and electronic equipment (WEEE) relevant substances in polymeric food-contact articles sold on the European market**. *Food Addit Contam Part A* (2015.0) **32** 410-426. DOI: 10.1080/19440049.2015.1009499 342. Eriksen MK, Pivnenko K, Olsson ME, Astrup TF. **Contamination in plastic recycling: influence of metals on the quality of reprocessed plastic**. *Waste Manag* (2018.0) **79** 595-606. DOI: 10.1016/j.wasman.2018.08.007 343. Dreolin N, Aznar M, Moret S, Nerin C. **Development and validation of a LC–MS/MS method for the analysis of bisphenol a in polyethylene terephthalate**. *Food Chem* (2019.0) **274** 246-253. DOI: 10.1016/j.foodchem.2018.08.109 344. Núñez SS, Moltó J, Conesa JA, Fullana A. **Heavy metals, PAHs and POPs in recycled polyethylene samples of agricultural, post-commercial, post-industrial and post-consumer origin**. *Waste Manag* (2022.0) **144** 113-121. DOI: 10.1016/j.wasman.2022.03.016 345. Whitt M, Vorst K, Brown W, Baker S, Gorman L. **Survey of heavy metal contamination in recycled polyethylene terephthalate used for food packaging**. *J Plast Film Sheeting* (2013.0) **29** 163-173. DOI: 10.1177/8756087912467028 346. Grant K, Goldizen FC, Sly PD. **Health consequences of exposure to e-waste: a systematic review**. *Lancet Glob Health* (2013.0) **1** e350-e361. DOI: 10.1016/S2214-109X(13)70101-3 347. Sahajwalla V, Gaikwad V. **The present and future of e-waste plastics recycling**. *Curr Opin Green Sustain Chem* (2018.0) **13** 102-107. DOI: 10.1016/j.cogsc.2018.06.006 348. Jia C, Das P, Kim I, Yoon YJ, Tay CY, Lee JM. **Applications, treatments, and reuse of plastics from electrical and electronic equipment**. *J Ind Eng Chem* (2022.0) **110** 84-99. DOI: 10.1016/j.jiec.2022.03.026 349. Bakhiyi B, Gravel S, Ceballos D, Flynn MA, Zayed J. **Has the question of e-waste opened a Pandora’s box? An overview of unpredictable issues and challenges**. *Environ Int* (2018.0) **110** 173-192. DOI: 10.1016/j.envint.2017.10.021 350. Okeme JO, Arrandale VH. **Electronic waste recycling: occupational exposures and work-related health effects**. *Curr Environ Health Rep* (2019.0) **6** 256-268. DOI: 10.1007/s40572-019-00255-3 351. Zennegg M, Schluep M, Streicher-Porte M, Lienemann P, Haag R, Gerecke AC. **Formation of PBDD/F from PBDE in electronic waste in recycling processes and under simulated extruding conditions**. *Chemosphere* (2014.0) **116** 34-39. DOI: 10.1016/j.chemosphere.2013.12.101 352. Gravel S, Lavoué J, Bakhiyi B. **Halogenated flame retardants and organophosphate esters in the air of electronic waste recycling facilities: evidence of high concentrations and multiple exposures**. *Environ Int* (2019.0) **128** 244-253. DOI: 10.1016/j.envint.2019.04.027 353. Ahirwar R, Tripathi AK. **E-waste management: a review of recycling process, environmental and occupational health hazards, and potential solutions**. *Environ Nanotechnol Monit Manag* (2021.0) **15** 100409. DOI: 10.1016/j.enmm.2020.100409 354. Thuresson K, Bergman Å, Rothenbacher K. **Polybrominated diphenyl ether exposure to electronics recycling workers – a follow-up study**. *Chemosphere* (2006.0) **64** 1855-1861. DOI: 10.1016/j.chemosphere.2006.01.055 355. Gravel S, Lavoué J, Bakhiyi B. **Multi-exposures to suspected endocrine disruptors in electronic waste recycling workers: associations with thyroid and reproductive hormones**. *Int J Hyg Environ Health* (2020.0) **225** 113445. DOI: 10.1016/j.ijheh.2019.113445 356. Sepúlveda A, Schluep M, Renaud FG. **A review of the environmental fate and effects of hazardous substances released from electrical and electronic equipments during recycling: examples from China and India**. *Environ Impact Assess Rev* (2010.0) **30** 28-41. DOI: 10.1016/j.eiar.2009.04.001 357. Chi X, Streicher-Porte M, Wang MYL, Reuter MA. **Informal electronic waste recycling: a sector review with special focus on China**. *Waste Manag* (2011.0) **31** 731-742. DOI: 10.1016/j.wasman.2010.11.006 358. Tue NM, Takahashi S, Subramanian A, Sakai S, Tanabe S. **Environmental contamination and human exposure to dioxin-related compounds in e-waste recycling sites of developing countries**. *Environ Sci Process Impacts* (2013.0) **15** 1326. DOI: 10.1039/c3em00086a 359. Chakraborty P, Sampath S, Mukhopadhyay M, Selvaraj S, Bharat GK, Nizzetto L. **Baseline investigation on plasticizers, bisphenol A, polycyclic aromatic hydrocarbons and heavy metals in the surface soil of the informal electronic waste recycling workshops and nearby open dumpsites in Indian metropolitan cities**. *Environ Pollut* (2019.0) **248** 1036-1045. DOI: 10.1016/j.envpol.2018.11.010 360. Parvez SM, Jahan F, Brune MN. **Health consequences of exposure to e-waste: an updated systematic review**. *Lancet Planet Health* (2021.0) **5** e905-e920. DOI: 10.1016/S2542-5196(21)00263-1 361. Darivakis GS, Howard JB, Peters WA. **Release rates of condensables and total volatiles from rapid devolatilization of polyethylene and polystyrene**. *Combust Sci Technol* (1990.0) **74** 267-281. DOI: 10.1080/00102209008951692 362. Velis CA, Cook E. **Mismanagement of plastic waste through open burning with emphasis on the global south: a systematic review of risks to occupational and public health**. *Environ Sci Technol* (2021.0) **55** 7186-7207. DOI: 10.1021/acs.est.0c08536 363. Wiedinmyer C, Yokelson RJ, Gullett BK. **Global emissions of trace gases, particulate matter, and hazardous air pollutants from open burning of domestic waste**. *Environ Sci Technol* (2014.0) **48** 9523-9530. DOI: 10.1021/es502250z 364. Peng C, Chen W, Liao X. **Polycyclic aromatic hydrocarbons in urban soils of Beijing: status, sources, distribution and potential risk**. *Environ Pollut* (2011.0) **159** 802-808. DOI: 10.1016/j.envpol.2010.11.003 365. Lemieux PM, Lutes CC, Santoianni DA. **Emissions of organic air toxics from open burning: a comprehensive review**. *Prog Energy Combust Sci* (2004.0) **30** 1-32. DOI: 10.1016/j.pecs.2003.08.001 366. Zhang M, Buekens A, Jiang X, Li X. **Dioxins and polyvinylchloride in combustion and fires**. *Waste Manag Res* (2015.0) **33** 630-643. DOI: 10.1177/0734242X15590651 367. Shivani R, Gadi SK, Sharma TK. **Seasonal variation, source apportionment and source attributed health risk of fine carbonaceous aerosols over National Capital Region, India**. *Chemosphere* (2019.0) **237** 124500. DOI: 10.1016/j.chemosphere.2019.124500 368. **Listing of POPs in the Stockholm Convention** 369. Pivnenko K, Granby K, Eriksson E, Astrup TF. **Recycling of plastic waste: screening for brominated flame retardants (BFRs)**. *Waste Manag* (2017.0) **69** 101-109. DOI: 10.1016/j.wasman.2017.08.038 370. Ni HG, Lu SY, Mo T, Zeng H. **Brominated flame retardant emissions from the open burning of five plastic wastes and implications for environmental exposure in China**. *Environ Pollut* (2016.0) **214** 70-76. DOI: 10.1016/j.envpol.2016.03.049 371. Tang Z, Huang Q, Cheng J. **Polybrominated diphenyl ethers in soils, sediments, and human hair in a plastic waste recycling area: a neglected heavily polluted area**. *Environ Sci Technol* (2014.0) **48** 1508-1516. DOI: 10.1021/es404905u 372. Tang Z, Huang Q, Yang Y. **Polybrominated diphenyl ethers (PBDEs) and heavy metals in road dusts from a plastic waste recycling area in north China: implications for human health**. *Environ Sci Pollut Res* (2016.0) **23** 625-637. DOI: 10.1007/s11356-015-5296-7 373. Tang Z, Zhang L, Huang Q. **Contamination and risk of heavy metals in soils and sediments from a typical plastic waste recycling area in North China**. *Ecotoxicol Environ Saf* (2015.0) **122** 343-351. DOI: 10.1016/j.ecoenv.2015.08.006 374. **Agents classified by the IARC Monographs, Volumes 1–132. IARC Monographs on the Identification of Carcinogenic Hazards to Humans** 375. Simoneit BRT, Medeiros PM, Didyk BM. **Combustion products of plastics as indicators for refuse burning in the atmosphere**. *Environ Sci Technol* (2005.0) **39** 6961-6970. DOI: 10.1021/es050767x 376. Fu P, Kawamura K. **Ubiquity of bisphenol A in the atmosphere**. *Environ Pollut* (2010.0) **158** 3138-3143. DOI: 10.1016/j.envpol.2010.06.040 377. Guerranti C, Martellini T, Fortunati A, Scodellini R, Cincinelli A. **Environmental pollution from plasticiser compounds: do we know enough about atmospheric levels and their contribution to human exposure in Europe?**. *Curr Opin Environ Sci Health* (2019.0) **8** 1-5. DOI: 10.1016/j.coesh.2018.10.004 378. Ma W, Cui J, Abdoulaye B. **Air pollutant emission inventory of waste-to-energy plants in China and prediction by the artificial neural network approach**. *Environ Sci Technol*. DOI: 10.1021/acs.est.2c01087 379. Rovira J, Mari M, Nadal M, Schuhmacher M, Domingo JL. **Partial replacement of fossil fuel in a cement plant: risk assessment for the population living in the neighborhood**. *Sci Total Environ* (2010.0) **408** 5372-5380. DOI: 10.1016/j.scitotenv.2010.07.060 380. Cole-Hunter T, Johnston FH, Marks GB. **The health impacts of waste-to-energy emissions: a systematic review of the literature**. *Environ Res Lett* (2020.0) **15** 123006. DOI: 10.1088/1748-9326/abae9f 381. Li H, Nitivattananon V, Li P. **Municipal solid waste management health risk assessment from air emissions for China by applying life cycle analysis**. *Waste Manag Res* (2015.0) **33** 401-409. DOI: 10.1177/0734242X15580191 382. Long CM, Nascarella MA, Valberg PA. **Carbon black vs. black carbon and other airborne materials containing elemental carbon: physical and chemical distinctions**. *Environ Pollut* (2013.0) **181** 271-286. DOI: 10.1016/j.envpol.2013.06.009 383. Shemwell BE, Levendis YA. **Particulates generated from combustion of polymers (plastics)**. *J Air Waste Manag Assoc* (2000.0) **50** 94-102. DOI: 10.1080/10473289.2000.10463994 384. Valavanidis A, Iliopoulos N, Gotsis G, Fiotakis K. **Persistent free radicals, heavy metals and PAHs generated in particulate soot emissions and residue ash from controlled combustion of common types of plastic**. *J Hazard Mater* (2008.0) **156** 277-284. DOI: 10.1016/j.jhazmat.2007.12.019 385. Barabad MLM, Jung W, Versoza ME, Lee Y il, Choi K, Park D. **Characteristics of particulate matter and volatile organic compound emissions from the combustion of waste vinyl**. *Int J Environ Res Public Health* (2018.0) **15** 1390. DOI: 10.3390/ijerph15071390 386. Wu D, Li Q, Shang X. **Commodity plastic burning as a source of inhaled toxic aerosols**. *J Hazard Mater* (2021.0) **416** 125820. DOI: 10.1016/j.jhazmat.2021.125820 387. Haque MM, Kawamura K, Deshmukh DK. **Characterization of organic aerosols from a Chinese megacity during winter: predominance of fossil fuel combustion**. *Atmospheric Chem Phys* (2019.0) **19** 5147-5164. DOI: 10.5194/acp-19-5147-2019 388. Gadi R, Shivani SK, Sharma TK. **Source apportionment and health risk assessment of organic constituents in fine ambient aerosols (PM2.5): a complete year study over national capital region of India**. *Chemosphere* (2019.0) **221** 583-596. DOI: 10.1016/j.chemosphere.2019.01.067 389. Islam MdR, Welker J, Salam A, Stone EA. **Plastic burning impacts on atmospheric fine particulate matter at urban and rural sites in the USA and Bangladesh**. *ACS Environ Au* (2022.0) **2** 409-417. DOI: 10.1021/acsenvironau.1c00054 390. Pehnec G, Jakovljević I. **Carcinogenic potency of airborne polycyclic aromatic hydrocarbons in relation to the particle fraction size**. *Int J Environ Res Public Health* (2018.0) **15** 2485. DOI: 10.3390/ijerph15112485 391. Etchie TO, Sivanesan S, Etchie AT. **The burden of disease attributable to ambient PM2.5-bound PAHs exposure in Nagpur, India**. *Chemosphere* (2018.0) **204** 277-289. DOI: 10.1016/j.chemosphere.2018.04.054 392. Wright SL, Ulke J, Font A, Chan KLA, Kelly FJ. **Atmospheric microplastic deposition in an urban environment and an evaluation of transport**. *Environ Int* (2020.0) **136** 105411. DOI: 10.1016/j.envint.2019.105411 393. Amato-Lourenço LF, dos Santos Galvão L, de Weger LA, Hiemstra PS, Vijver MG, Mauad T. **An emerging class of air pollutants: potential effects of microplastics to respiratory human health?**. *Sci Total Environ* (2020.0) **749** 141676. DOI: 10.1016/j.scitotenv.2020.141676 394. Sridharan S, Kumar M, Singh L, Bolan NS, Saha M. **Microplastics as an emerging source of particulate air pollution: a critical review**. *J Hazard Mater* (2021.0) **418** 126245. DOI: 10.1016/j.jhazmat.2021.126245 395. Chen G, Feng Q, Wang J. **Mini-review of microplastics in the atmosphere and their risks to humans**. *Sci Total Environ* (2020.0) **703** 135504. DOI: 10.1016/j.scitotenv.2019.135504 396. Enyoh CE, Verla AW, Verla EN, Ibe FC, Amaobi CE. **Airborne microplastics: a review study on method for analysis, occurrence, movement and risks**. *Environ Monit Assess* (2019.0) **191** 668. DOI: 10.1007/s10661-019-7842-0 397. Vaverková MD. **Landfill impacts on the environment – review**. *Geosciences* (2019.0) **9** 431. DOI: 10.3390/geosciences9100431 398. Canopoli L, Fidalgo B, Coulon F, Wagland ST. **Physico-chemical properties of excavated plastic from landfill mining and current recycling routes**. *Waste Manag* (2018.0) **76** 55-67. DOI: 10.1016/j.wasman.2018.03.043 399. Ryberg MW, Laurent A, Hauschild MZ. *Mapping of Global Plastics Value Chain and Hotspots: With a Particular Focus on Marine Environment* (2018.0) 400. Siddiqua A, Hahladakis JN, Al-Attiya WAKA. **An overview of the environmental pollution and health effects associated with waste landfilling and open dumping**. *Environ Sci Pollut Res* (2022.0) **29** 58514-58536. DOI: 10.1007/s11356-022-21578-z 401. Kim YJ, Osako M, Sakai S ichi. **Leaching characteristics of polybrominated diphenyl ethers (PBDEs) from flame-retardant plastics**. *Chemosphere* (2006.0) **65** 506-513. DOI: 10.1016/j.chemosphere.2006.01.019 402. Liu Y, Liu Y, Li H. **Health risk impacts analysis of fugitive aromatic compounds emissions from the working face of a municipal solid waste landfill in China**. *Environ Int* (2016.0) **97** 15-27. DOI: 10.1016/j.envint.2016.10.010 403. Forrest A, Giacovazzi L, Dunlop S. **Eliminating plastic pollution: how a voluntary contribution from industry will drive the circular plastics economy**. *Front Mar Sci* (2019.0) **6** 627. DOI: 10.3389/fmars.2019.00627 404. Brooks AL, Wang S, Jambeck JR. **The Chinese import ban and its impact on global plastic waste trade**. *Sci Adv* (2018.0) **4**. DOI: 10.1126/sciadv.aat0131 405. Brown A, Laubinger F, Börkey P. *Monitoring Trade in Plastic Waste and Scrap* (2022.0). DOI: 10.1787/8f3e9c56-en 406. Clarke H. **China’s tsunami of waste [plastic waste]**. *Eng Technol* (2019.0) **14** 48-51. DOI: 10.1049/et.2019.0308 407. Lebreton LCM, van der Zwet J, Damsteeg JW, Slat B, Andrady A, Reisser J. **River plastic emissions to the world’s oceans**. *Nat Commun* (2017.0) **8** 15611. DOI: 10.1038/ncomms15611 408. Schmidt C, Krauth T, Wagner S. **Export of plastic debris by rivers into the sea**. *Environ Sci Technol* (2017.0) **51** 12246-12253. DOI: 10.1021/acs.est.7b02368 409. Meijer LJJ, van Emmerik T, van der Ent R, Schmidt C, Lebreton L. **More than 1000 rivers account for 80% of global riverine plastic emissions into the ocean**. *Sci Adv* (2021.0) **7**. DOI: 10.1126/sciadv.aaz5803 410. Macfadyen G, Huntington T, Cappell R. *Abandoned, Lost or Otherwise Discarded Fishing Gear* (2009.0) 411. Dibke C, Fischer M, Scholz-Böttcher BM. **Microplastic mass concentrations and distribution in German bight waters by pyrolysis–gas chromatography–mass spectrometry/thermochemolysis reveal potential impact of marine coatings: do ships leave skid marks?**. *Environ Sci Technol* (2021.0) **55** 2285-2295. DOI: 10.1021/acs.est.0c04522 412. Turner A. **Paint particles in the marine environment: an overlooked component of microplastics**. *Water Res X* (2021.0) **12** 100110. DOI: 10.1016/j.wroa.2021.100110 413. Gipperth L. **The legal design of the international and European Union ban on tributyltin antifouling paint: direct and indirect effects**. *J Environ Manage* (2009.0) **90** S86-S95. DOI: 10.1016/j.jenvman.2008.08.013 414. Gaylarde CC, Neto JAB, da Fonseca EM. **Paint fragments as polluting microplastics: a brief review**. *Mar Pollut Bull* (2021.0) **162** 111847. DOI: 10.1016/j.marpolbul.2020.111847 415. McAdams D, Angelskar T 416. Pinovo AS. **Clean blasting. Pinovo – dust-free abrasive vacuum blasting for industrial use** 417. Karlsson TM, Arneborg L, Broström G, Almroth BC, Gipperth L, Hassellöv M. **The unaccountability case of plastic pellet pollution**. *Mar Pollut Bull* (2018.0) **129** 52-60. DOI: 10.1016/j.marpolbul.2018.01.041 418. 418Organisation for Economic Co-operation and Development (OECD). Policies to Reduce Microplastics Pollution in Water: Focus on Textiles and Tyres. OECD Publishing; 2021. DOI: 10.1787/7ec7e5ef-en. *Policies to Reduce Microplastics Pollution in Water: Focus on Textiles and Tyres* (2021.0). DOI: 10.1787/7ec7e5ef-en 419. **The Great Nurdle Hunt: tackling worldwide pollution** 420. Cole G, Sherrington C. *Study to Quantify Pellet Emissions in the UK* (2016.0) 421. Carney Almroth BM, Åström L, Roslund S, Petersson H, Johansson M, Persson NK. **Quantifying shedding of synthetic fibers from textiles: a source of microplastics released into the environment**. *Environ Sci Pollut Res* (2018.0) **25** 1191-1199. DOI: 10.1007/s11356-017-0528-7 422. Belzagui F, Crespi M, Álvarez A, Gutiérrez-Bouzán C, Vilaseca M. **Microplastics’ emissions: microfibers’ detachment from textile garments**. *Environ Pollut* (2019.0) **248** 1028-1035. DOI: 10.1016/j.envpol.2019.02.059 423. Goehler LO, Moruzzi RB, Tomazini da Conceição F. **Relevance of tyre wear particles to the total content of microplastics transported by runoff in a high-imperviousness and intense vehicle traffic urban area**. *Environ Pollut* (2022.0) **314** 120200. DOI: 10.1016/j.envpol.2022.120200 424. Ryberg MW, Hauschild MZ, Wang F, Averous-Monnery S, Laurent A. **Global environmental losses of plastics across their value chains**. *Resour Conserv Recycl* (2019.0) **151** 104459. DOI: 10.1016/j.resconrec.2019.104459 425. Burghardt TE, Pashkevich A, Babić D, Mosböck H, Babić D, Żakowska L. **Microplastics and road markings: the role of glass beads and loss estimation**. *Transp Res Part Transp Environ* (2022.0) **102** 103123. DOI: 10.1016/j.trd.2021.103123 426. Yurtsever M. **Tiny, shiny, and colorful microplastics: are regular glitters a significant source of microplastics?**. *Mar Pollut Bull* (2019.0) **146** 678-682. DOI: 10.1016/j.marpolbul.2019.07.009 427. Nawalage NSK, Bellanthudawa BKA. **Synthetic polymers in personal care and cosmetics products (PCCPs) as a source of microplastic (MP) pollution**. *Mar Pollut Bull* (2022.0) **182** 113927. DOI: 10.1016/j.marpolbul.2022.113927 428. Dris R, Gasperi J, Mirande C. **A first overview of textile fibers, including microplastics, in indoor and outdoor environments**. *Environ Pollut* (2017.0) **221** 453-458. DOI: 10.1016/j.envpol.2016.12.013 429. Jiang L, Chen M, Huang Y. **Effects of different treatment processes in four municipal wastewater treatment plants on the transport and fate of microplastics**. *Sci Total Environ* (2022.0) **831** 154946. DOI: 10.1016/j.scitotenv.2022.154946 430. Miller RZ, Watts AJR, Winslow BO, Galloway TS, Barrows APW. **Mountains to the sea: river study of plastic and non-plastic microfiber pollution in the northeast USA**. *Mar Pollut Bull* (2017.0) **124** 245-251. DOI: 10.1016/j.marpolbul.2017.07.028 431. Liu J, Yang Y, Ding J, Zhu B, Gao W. **Microfibers: a preliminary discussion on their definition and sources**. *Environ Sci Pollut Res* (2019.0) **26** 29497-29501. DOI: 10.1007/s11356-019-06265-w 432. Liu J, Liang J, Ding J. **Microfiber pollution: an ongoing major environmental issue related to the sustainable development of textile and clothing industry**. *Environ Dev Sustain* (2021.0) **23** 11240-11256. DOI: 10.1007/s10668-020-01173-3 433. Suzuki G, Uchida N, Tuyen LH. **Mechanical recycling of plastic waste as a point source of microplastic pollution**. *Environ Pollut* (2022.0) **303** 119114. DOI: 10.1016/j.envpol.2022.119114 434. He P, Chen L, Shao L, Zhang H, Lü F. **Municipal solid waste (MSW) landfill: a source of microplastics? Evidence of microplastics in landfill leachate**. *Water Res* (2019.0) **159** 38-45. DOI: 10.1016/j.watres.2019.04.060 435. Allen S, Allen D, Phoenix VR. **Atmospheric transport and deposition of microplastics in a remote mountain catchment**. *Nat Geosci* (2019.0) **12** 339-344. DOI: 10.1038/s41561-019-0335-5 436. Evangeliou N, Grythe H, Klimont Z. **Atmospheric transport is a major pathway of microplastics to remote regions**. *Nat Commun* (2020.0) **11** 3381. DOI: 10.1038/s41467-020-17201-9 437. Trainic M, Flores JM, Pinkas I. **Airborne microplastic particles detected in the remote marine atmosphere**. *Commun Earth Environ* (2020.0) **1** 1-9. DOI: 10.1038/s43247-020-00061-y 438. Shruti VC, Pérez-Guevara F, Elizalde-Martínez I, Kutralam-Muniasamy G. **Current trends and analytical methods for evaluation of microplastics in stormwater**. *Trends Environ Anal Chem* (2021.0) **30**. DOI: 10.1016/j.teac.2021.e00123 439. Fan W, Salmond JA, Dirks KN, Cabedo Sanz P, Miskelly GM, Rindelaub JD. **Evidence and mass quantification of atmospheric microplastics in a coastal New Zealand city**. *Environ Sci Technol* (2022.0) **56** 17556-17568. DOI: 10.1021/acs.est.2c05850 440. Menéndez-Manjón A, Martínez-Díez R, Sol D. **Long-term occurrence and fate of microplastics in WWTPs: a case study in southwest Europe**. *Appl Sci* (2022.0) **12** 2133. DOI: 10.3390/app12042133 441. Long Z, Wang W, Yu X, Lin Z, Chen J. **Heterogeneity and contribution of microplastics from industrial and domestic sources in a wastewater treatment plant in Xiamen, China**. *Front Environ Sci* (2021.0) **9**. DOI: 10.3389/fenvs.2021.770634 442. Sun J, Dai X, Wang Q, van Loosdrecht MCM, Ni BJ. **Microplastics in wastewater treatment plants: detection, occurrence and removal**. *Water Res* (2019.0) **152** 21-37. DOI: 10.1016/j.watres.2018.12.050 443. Iyare PU, Ouki SK, Bond T. **Microplastics removal in wastewater treatment plants: a critical review**. *Environ Sci Water Res Technol* (2020.0) **6** 2664-2675. DOI: 10.1039/D0EW00397B 444. Woodward J, Li J, Rothwell J, Hurley R. **Acute riverine microplastic contamination due to avoidable releases of untreated wastewater**. *Nat Sustain* (2021.0) **4** 793-802. DOI: 10.1038/s41893-021-00718-2 445. Lofty J, Muhawenimana V, Wilson CAME, Ouro P. **Microplastics removal from a primary settler tank in a wastewater treatment plant and estimations of contamination onto European agricultural land via sewage sludge recycling**. *Environ Pollut* (2022.0) **304** 119198. DOI: 10.1016/j.envpol.2022.119198 446. Rolsky C, Kelkar V, Driver E, Halden RU. **Municipal sewage sludge as a source of microplastics in the environment**. *Curr Opin Environ Sci Health* (2020.0) **14** 16-22. DOI: 10.1016/j.coesh.2019.12.001 447. Hatinoğlu MD, Sanin FD. **Sewage sludge as a source of microplastics in the environment: a review of occurrence and fate during sludge treatment**. *J Environ Manage* (2021.0) **295** 113028. DOI: 10.1016/j.jenvman.2021.113028 448. Weber CJ, Santowski A, Chifflard P. **Investigating the dispersal of macro- and microplastics on agricultural fields 30 years after sewage sludge application**. *Sci Rep* (2022.0) **12** 6401. DOI: 10.1038/s41598-022-10294-w 449. Nizzetto L, Futter M, Langaas S. **Are Agricultural Soils Dumps for Microplastics of Urban Origin?**. *Environ Sci Technol* (2016.0) **50** 10777-10779. DOI: 10.1021/acs.est.6b04140 450. Zhang B, Yang X, Chen L, Chao J, Teng J, Wang Q. **Microplastics in soils: a review of possible sources, analytical methods and ecological impacts**. *J Chem Technol Biotechnol* (2020.0) **95** 2052-2068. DOI: 10.1002/jctb.6334 451. Corradini F, Meza P, Eguiluz R, Casado F, Huerta-Lwanga E, Geissen V. **Evidence of microplastic accumulation in agricultural soils from sewage sludge disposal**. *Sci Total Environ* (2019.0) **671** 411-420. DOI: 10.1016/j.scitotenv.2019.03.368 452. Zhou J, Wen Y, Marshall MR. **Microplastics as an emerging threat to plant and soil health in agroecosystems**. *Sci Total Environ* (2021.0) **787** 147444. DOI: 10.1016/j.scitotenv.2021.147444 453. Chia RW, Lee JY, Jang J, Kim H, Kwon KD. **Soil health and microplastics: a review of the impacts of microplastic contamination on soil properties**. *J Soils Sediments* (2022.0) **22** 2690-2705. DOI: 10.1007/s11368-022-03254-4 454. Gao H, Liu Q, Yan C. **Macro-and/or microplastics as an emerging threat effect crop growth and soil health**. *Resour Conserv Recycl* (2022.0) **186** 106549. DOI: 10.1016/j.resconrec.2022.106549 455. Zhang L, Xie Y, Liu J, Zhong S, Qian Y, Gao P. **An overlooked entry pathway of microplastics into agricultural soils from application of sludge-based fertilizers**. *Environ Sci Technol* (2020.0) **54** 4248-4255. DOI: 10.1021/acs.est.9b07905 456. Grob K, Biedermann M, Scherbaum E, Roth M, Rieger K. **Food Contamination with organic materials in perspective: packaging materials as the largest and least controlled source? A view focusing on the European situation**. *Crit Rev Food Sci Nutr* (2006.0) **46** 529-535. DOI: 10.1080/10408390500295490 457. Geueke B, Groh KJ, Maffini MV. **Systematic evidence on migrating and extractable food contact chemicals: most chemicals detected in food contact materials are not listed for use**. *Crit Rev Food Sci Nutr* (2022.0) **0** 1-11. DOI: 10.1080/10408398.2022.2067828 458. Conchione C, Lucci P, Moret S. **Migration of polypropylene oligomers into ready-to-eat vegetable soups**. *Foods* (2020.0) **9** 1365. DOI: 10.3390/foods9101365 459. Zimmermann L, Bartosova Z, Braun K, Oehlmann J, Völker C, Wagner M. **Plastic products leach chemicals that induce in vitro toxicity under realistic use conditions**. *Environ Sci Technol* (2021.0) **55** 11814-11823. DOI: 10.1021/acs.est.1c01103 460. Grob K. **The European system for the control of the safety of food-contact materials needs restructuring: a review and outlook for discussion**. *Food Addit Contam Part A* (2017.0) **34** 1643-1659. DOI: 10.1080/19440049.2017.1332431 461. Lomonaco T, Manco E, Corti A. **Release of harmful volatile organic compounds (VOCs) from photo-degraded plastic debris: a neglected source of environmental pollution**. *J Hazard Mater* (2020.0) **394** 122596. DOI: 10.1016/j.jhazmat.2020.122596 462. Rendell-Bhatti F, Paganos P, Pouch A. **Developmental toxicity of plastic leachates on the sea urchin Paracentrotus lividus**. *Environ Pollut* (2021.0) **269** 115744. DOI: 10.1016/j.envpol.2020.115744 463. Kwon BG, Koizumi K, Chung SY, Kodera Y, Kim JO, Saido K. **Global styrene oligomers monitoring as new chemical contamination from polystyrene plastic marine pollution**. *J Hazard Mater* (2015.0) **300** 359-367. DOI: 10.1016/j.jhazmat.2015.07.039 464. Kwon BG, Chung SY, Saido K. **Sandy beaches as hotspots of bisphenol A**. *Environ Res* (2020.0) **191** 110175. DOI: 10.1016/j.envres.2020.110175 465. Simoneau C, Van den Eede L, Valzacchi S. **Identification and quantification of the migration of chemicals from plastic baby bottles used as substitutes for polycarbonate**. *Food Addit Contam Part A* (2012.0) **29** 469-480. DOI: 10.1080/19440049.2011.644588 466. He C, Wang X, Thai P. **Organophosphate and brominated flame retardants in Australian indoor environments: levels, sources, and preliminary assessment of human exposure**. *Environ Pollut* (2018.0) **235** 670-679. DOI: 10.1016/j.envpol.2017.12.017 467. Lucattini L, Poma G, Covaci A, de Boer J, Lamoree MH, Leonards PEG. **A review of semi-volatile organic compounds (SVOCs) in the indoor environment: occurrence in consumer products, indoor air and dust**. *Chemosphere* (2018.0) **201** 466-482. DOI: 10.1016/j.chemosphere.2018.02.161 468. Lu S, Yu Y, Ren L, Zhang X, Liu G, Yu Y. **Estimation of intake and uptake of bisphenols and triclosan from personal care products by dermal contact**. *Sci Total Environ* (2018.0) **621** 1389-1396. DOI: 10.1016/j.scitotenv.2017.10.088 469. Pagoni A, Arvaniti OS, Kalantzi OI. **Exposure to phthalates from personal care products: urinary levels and predictors of exposure**. *Environ Res* (2022.0) **212** 113194. DOI: 10.1016/j.envres.2022.113194 470. Martí V, Jubany I, Pérez C, Rubio X, De Pablo J, Giménez J. **Human Health Risk Assessment of a landfill based on volatile organic compounds emission, immission and soil gas concentration measurements**. *Appl Geochem* (2014.0) **49** 218-224. DOI: 10.1016/j.apgeochem.2014.06.018 471. Ward CP, Reddy CM. **We need better data about the environmental persistence of plastic goods**. *Proc Natl Acad Sci* (2020.0) **117** 14618-14621. DOI: 10.1073/pnas.2008009117 472. Teuten EL, Saquing JM, Knappe DR. **Transport and release of chemicals from plastics to the environment and to wildlife**. *Philos Trans R Soc B Biol Sci* (2009.0) **364** 2027-2045. DOI: 10.1098/rstb.2008.0284 473. Luo H, Zeng Y, Cheng Y, He D, Pan X. **Recent advances in municipal landfill leachate: a review focusing on its characteristics, treatment, and toxicity assessment**. *Sci Total Environ* (2020.0) **703** 135468. DOI: 10.1016/j.scitotenv.2019.135468 474. Zhang B, Li G, Cheng P, Yeh TCJ, Hong M. **Landfill risk assessment on groundwater based on vulnerability and pollution index**. *Water Resour Manag* (2016.0) **30** 1465-1480. DOI: 10.1007/s11269-016-1233-x 475. Melnyk A, Kuklińska K, Wolska L, Namieśnik J. **Chemical pollution and toxicity of water samples from stream receiving leachate from controlled municipal solid waste (MSW) landfill**. *Environ Res* (2014.0) **135** 253-261. DOI: 10.1016/j.envres.2014.09.010 476. Wijesekara SSRMDHR, Mayakaduwa SS, Siriwardana AR. **Fate and transport of pollutants through a municipal solid waste landfill leachate in Sri Lanka**. *Environ Earth Sci* (2014.0) **72** 1707-1719. DOI: 10.1007/s12665-014-3075-2 477. Olujimi OO, Adegbite KI, Sojinu SO. **Occurrence, compositional profiles and health risks assessment of polybrominated diphenyl ethers (PBDEs) in groundwater in and around dumpsites in Lagos State, Nigeria**. *Environ Forensics* (2022.0) **0** 1-10. DOI: 10.1080/15275922.2022.2081887 478. Amobonye A, Bhagwat P, Singh S, Pillai S. **Plastic biodegradation: frontline microbes and their enzymes**. *Sci Total Environ* (2021.0) **759** 143536. DOI: 10.1016/j.scitotenv.2020.143536 479. Shen M, Huang W, Chen M, Song B, Zeng G, Zhang Y. **(Micro)plastic crisis: un-ignorable contribution to global greenhouse gas emissions and climate change**. *J Clean Prod* (2020.0) **254** 120138. DOI: 10.1016/j.jclepro.2020.120138 480. Ford HV, Jones NH, Davies AJ. **The fundamental links between climate change and marine plastic pollution**. *Sci Total Environ* (2022.0) **806** 150392. DOI: 10.1016/j.scitotenv.2021.150392 481. Lavers JL, Bond AL, Rolsky C. **Far from a distraction: plastic pollution and the planetary emergency**. *Biol Conserv* (2022.0) **272** 109655. DOI: 10.1016/j.biocon.2022.109655 482. Hamilton LA, Feit S. *Plastic & Climate: The Hidden Costs of a Plastic Planet* (2019.0) 483. Zheng J, Suh S. **Strategies to reduce the global carbon footprint of plastics**. *Nat Clim Change* (2019.0) **9** 374-378. DOI: 10.1038/s41558-019-0459-z 484. Bauer F, Nielsen TD, Nilsson LJ. **Plastics and climate change – breaking carbon lock-ins through three mitigation pathways**. *One Earth* (2022.0) **5** 361-376. DOI: 10.1016/j.oneear.2022.03.007 485. Royer SJ, Ferrón S, Wilson ST, Karl DM. **Production of methane and ethylene from plastic in the environment**. *PLoS ONE* (2018.0) **13**. DOI: 10.1371/journal.pone.0200574 486. Nicholson SR, Rorrer NA, Carpenter AC, Beckham GT. **Manufacturing energy and greenhouse gas emissions associated with plastics consumption**. *Joule* (2021.0) **5** 673-686. DOI: 10.1016/j.joule.2020.12.027 487. **Decision -/CP.26, Glasgow Climate Pact. United Nations Framework Convention on Climate Change**. (2021.0) 488. Kenyon KW, Kridler E. **Laysan albatrosses swallow indigestible matter**. *The Auk* (1969.0) **86** 339-343. DOI: 10.2307/4083505 489. Buchanan JB. **Pollution by synthetic fibres**. *Mar Pollut Bull* (1971.0) **2** 23. DOI: 10.1016/0025-326X(71)90136-6 490. Carpenter EJ, Smith KL. **Plastics on the Sargasso sea surface**. *Science* (1972.0) **175** 1240-1241. DOI: 10.1126/science.175.4027.1240 491. Venrick EL, Backman TW, Bartram WC, Platt CJ, Thornhill MS, Yates RE. **Man-made objects on the surface of the central North Pacific Ocean**. *Nature* (1973.0) **241** 271. DOI: 10.1038/241271a0 492. Colton JBJ, Burns BR, Knapp FD. **Plastic particles in surface waters of the northwestern atlantic**. *Science* (1974.0) **185** 491-497. DOI: 10.1126/science.185.4150.491 493. Wong CS, Green DR, Cretney WJ. **Quantitative tar and plastic waste distributions in the Pacific Ocean**. *Nature* (1974.0) **247** 30-32. DOI: 10.1038/247030a0 494. Venkataramanan R, Burckart GJ, Ptachcinski RJ. **Leaching of diethylhexyl phthalate from polyvinyl chloride bags into intravenous cyclosporine solution**. *Am J Hosp Pharm* (1986.0) **43** 2800-2802. DOI: 10.1093/ajhp/43.11.2800 495. Moore C. **Trashed: across the Pacific Ocean, plastics, plastics, everywhere**. *Nat Hist* (2003.0) **2003** 6 496. Sun Q, Ren SY, Ni HG. **Incidence of microplastics in personal care products: an appreciable part of plastic pollution**. *Sci Total Environ* (2020.0) **742** 140218. DOI: 10.1016/j.scitotenv.2020.140218 497. 497GESAMP (Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Sources, Fate and Effects of Microplastics in the Marine Environment: Part Two of a Global Assessment (Kershaw, P.J., and Rochman, C.M., Eds). (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Rep. Stud. GESAMP No. 93, 220p; 2016.. *Sources, Fate and Effects of Microplastics in the Marine Environment: Part Two of a Global Assessment (Kershaw, P.J., and Rochman, C.M., Eds). (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Rep. Stud. GESAMP No. 93, 220p* (2016.0) 498. 498GESAMP (Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Guidelines or the Monitoring and Assessment of Plastic Litter and Microplastics in the Ocean (Kershaw P.J., Turra A. and Galgani F. Editors), (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Rep. Stud. GESAMP No. 99, 130p; 2019.. *Guidelines or the Monitoring and Assessment of Plastic Litter and Microplastics in the Ocean (Kershaw P.J., Turra A. and Galgani F. Editors), (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Rep. Stud. GESAMP No. 99, 130p* (2019.0) 499. 499GESAMP (Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Proceedings of the GESAMP International Workshop on Assessing the Risks Associated with Plastics and Microplastics in the Marine Environment (Kershaw, P.J., Carney Almroth, B., Villarrubia-Gómez, P., Koelmans, A.A., and Gouin, T., Eds.). (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Reports to GESAMP No. 103, 68p; 2020.. *Proceedings of the GESAMP International Workshop on Assessing the Risks Associated with Plastics and Microplastics in the Marine Environment (Kershaw, P.J., Carney Almroth, B., Villarrubia-Gómez, P., Koelmans, A.A., and Gouin, T., Eds.). (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Reports to GESAMP No. 103, 68p* (2020.0) 500. 500GESAMP (Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Sea-Based Sources of Marine Litter, (Gilardi, K., Ed.) (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Rep. Stud. GESAMP No. 108, 109p; 2021.. *Sea-Based Sources of Marine Litter, (Gilardi, K., Ed.) (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on the Scientific Aspects of Marine Environmental Protection). Rep. Stud. GESAMP No. 108, 109p* (2021.0) 501. Andrady AL. *Plastics and the Ocean: Origin, Characterization, Fate, and Impacts* (2022.0). DOI: 10.1002/9781119768432 502. Andrady AL. **Microplastics in the marine environment**. *Mar Pollut Bull* (2011.0) **62** 1596-1605. DOI: 10.1016/j.marpolbul.2011.05.030 503. Cole M, Lindeque P, Halsband C, Galloway TS. **Microplastics as contaminants in the marine environment: a review**. *Mar Pollut Bull* (2011.0) **62** 2588-2597. DOI: 10.1016/j.marpolbul.2011.09.025 504. Wright SL, Thompson RC, Galloway TS. **The physical impacts of microplastics on marine organisms: a review**. *Environ Pollut* (2013.0) **178** 483-492. DOI: 10.1016/j.envpol.2013.02.031 505. Gall SC, Thompson RC. **The impact of debris on marine life**. *Mar Pollut Bull* (2015.0) **92** 170-179. DOI: 10.1016/j.marpolbul.2014.12.041 506. Rochman CM, Browne MA, Underwood AJ, van Franeker JA, Thompson RC, Amaral-Zettler LA. **The ecological impacts of marine debris: unraveling the demonstrated evidence from what is perceived**. *Ecology* (2016.0) **97** 302-312. DOI: 10.1890/14-2070.1 507. Galloway TS, Cole M, Lewis C. **Interactions of microplastic debris throughout the marine ecosystem**. *Nat Ecol Evol* (2017.0) **1** 116. DOI: 10.1038/s41559-017-0116 508. Worm B, Lotze HK, Jubinville I, Wilcox C, Jambeck J. **Plastic as a persistent marine pollutant**. *Annu Rev Environ Resour* (2017.0) **42** 1-26. DOI: 10.1146/annurev-environ-102016-060700 509. Burns EE, Boxall ABA. **Microplastics in the aquatic environment: evidence for or against adverse impacts and major knowledge gaps**. *Environ Toxicol Chem* (2018.0) **37** 2776-2796. DOI: 10.1002/etc.4268 510. 510Science Advice for Policy by European Academies (SAPEA). A Scientific Perspective on Microplastics in Nature and Society. SAPEA. 2018; 176. DOI: 10.26356/microplastics. *A Scientific Perspective on Microplastics in Nature and Society* (2018.0) **176**. DOI: 10.26356/microplastics 511. Besseling E, Redondo-Hasselerharm P, Foekema EM, Koelmans AA. **Quantifying ecological risks of aquatic micro- and nanoplastic**. *Crit Rev Environ Sci Technol* (2019.0) **49** 32-80. DOI: 10.1080/10643389.2018.1531688 512. Koelmans AA, Mohamed Nor NH, Hermsen E, Kooi M, Mintenig SM, De France J. **Microplastics in freshwaters and drinking water: critical review and assessment of data quality**. *Water Res* (2019.0) **155** 410-422. DOI: 10.1016/j.watres.2019.02.054 513. Hale RC, Seeley ME, La Guardia MJ, Mai L, Zeng EY. **A global perspective on microplastics**. *J Geophys Res Oceans* (2020.0) **125**. DOI: 10.1029/2018JC014719 514. Kögel T, Bjorøy Ø, Toto B, Bienfait AM, Sanden M. **Micro- and nanoplastic toxicity on aquatic life: determining factors**. *Sci Total Environ* (2020.0) **709** 136050. DOI: 10.1016/j.scitotenv.2019.136050 515. van Sebille E, Aliani S, Law KL. **The physical oceanography of the transport of floating marine debris**. *Environ Res Lett* (2020.0) **15** 023003. DOI: 10.1088/1748-9326/ab6d7d 516. Huang W, Song B, Liang J. **Microplastics and associated contaminants in the aquatic environment: a review on their ecotoxicological effects, trophic transfer, and potential impacts to human health**. *J Hazard Mater* (2021.0) **405** 124187. DOI: 10.1016/j.jhazmat.2020.124187 517. Khan FR, Catarino AI, Clark NJ. **The ecotoxicological consequences of microplastics and co-contaminants in aquatic organisms: a mini-review**. *Emerg Top Life Sci* (2022.0) **6** 339-348. DOI: 10.1042/ETLS20220014 518. Trevisan R, Ranasinghe P, Jayasundara N, Di Giulio RT. **Nanoplastics in aquatic environments: impacts on aquatic species and interactions with environmental factors and pollutants**. *Toxics* (2022.0) **10**. DOI: 10.3390/toxics10060326 519. **COMMISSION REGULATION (EU) …/… of XXX amending Annex XVII to Regulation (EC) No 1907/2006 of the European Parliament and of the Council concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) as regards synthetic polymer microparticles** 520. Subramanian M. **Plastics tsunami: can a landmark treaty stop waste from choking the oceans?**. *Nature* (2022.0) **611** 650-653. DOI: 10.1038/d41586-022-03793-3 521. Hartmann NB, Hüffer T, Thompson RC. **Are we speaking the same language? Recommendations for a definition and categorization framework for plastic debris**. *Environ Sci Technol* (2019.0) **53** 1039-1047. DOI: 10.1021/acs.est.8b05297 522. Kooi M, Koelmans AA. **Simplifying microplastic via continuous probability distributions for size, shape, and density**. *Environ Sci Technol Lett* (2019.0) **6** 551-557. DOI: 10.1021/acs.estlett.9b00379 523. Martí E, Martin C, Galli M, Echevarría F, Duarte CM, Cózar A. **The colors of the ocean plastics**. *Env Sci Technol* (2020.0) **54** 6594-6601. DOI: 10.1021/acs.est.9b06400 524. Kooi M, Primpke S, Mintenig SM, Lorenz C, Gerdts G, Koelmans AA. **Characterizing the multidimensionality of microplastics across environmental compartments**. *Water Res* (2021.0) **202** 117429. DOI: 10.1016/j.watres.2021.117429 525. Bank MS, Mitrano DM, Rillig MC, Sze Ki Lin C, Ok YS. **Embrace complexity to understand microplastic pollution**. *Nat Rev Earth Environ* (2022.0) **3** 736-737. DOI: 10.1038/s43017-022-00365-x 526. Walsh AN, Reddy CM, Niles SF, McKenna AM, Hansel CM, Ward CP. **Plastic formulation is an emerging control of its photochemical fate in the ocean**. *Environ Sci Technol* (2021.0) **55** 12383-12392. DOI: 10.1021/acs.est.1c02272 527. Bouwmeester H, Hollman PC, Peters RJ. **Potential health impact of environmentally released micro- and nanoplastics in the human food production chain: experiences from nanotoxicology**. *Env Sci Technol* (2015.0) **49** 8932-8947. DOI: 10.1021/acs.est.5b01090 528. Gigault J, El Hadri H, Nguyen B. **Nanoplastics are neither microplastics nor engineered nanoparticles**. *Nat Nanotechnol* (2021.0) **16** 501-507. DOI: 10.1038/s41565-021-00886-4 529. Mitrano DM, Wick P, Nowack B. **Placing nanoplastics in the context of global plastic pollution**. *Nat Nanotechnol* (2021.0) **16** 491-500. DOI: 10.1038/s41565-021-00888-2 530. Mohamed Nor NH, Kooi M, Diepens NJ, Koelmans AA. **Lifetime accumulation of microplastic in children and adults**. *Environ Sci Technol* (2021.0) **55** 5084-5096. DOI: 10.1021/acs.est.0c07384 531. Brahney J, Mahowald N, Prank M. **Constraining the atmospheric limb of the plastic cycle**. *Proc Natl Acad Sci USA* (2021.0) **118**. DOI: 10.1073/pnas.2020719118 532. Allen D, Allen S, Abbasi S. **Microplastics and nanoplastics in the marine-atmosphere environment**. *Nat Rev Earth Environ* (2022.0) **3** 393-405. DOI: 10.1038/s43017-022-00292-x 533. Dissanayake PD, Kim S, Sarkar B. **Effects of microplastics on the terrestrial environment: a critical review**. *Environ Res* (2022.0) **209** 112734. DOI: 10.1016/j.envres.2022.112734 534. Zhang Y, Cai C, Gu Y, Shi Y, Gao X. **Microplastics in plant-soil ecosystems: a meta-analysis**. *Environ Pollut* (2022.0) **308** 119718. DOI: 10.1016/j.envpol.2022.119718 535. 535National Academies of Sciences, Engineering, and Medicine (NASEM). Oil in the Sea IV: Inputs, Fates, and Effects. The National Academies Press. 2022; 516. DOI: 10.17226/26410. *Oil in the Sea IV: Inputs, Fates, and Effects* (2022.0) **516**. DOI: 10.17226/26410 536. Peng C, Zhao X, Liu G. **Noise in the Sea and Its Impacts on Marine Organisms**. *Int J Environ Res Public Health* (2015.0) **12** 12304-12323. DOI: 10.3390/ijerph121012304 537. Kunc HP, McLaughlin KE, Schmidt R. **Aquatic noise pollution: implications for individuals, populations, and ecosystems**. *Proc Biol Sci* (2016.0) **283**. DOI: 10.1098/rspb.2016.0839 538. Southall BL, Scholik-Schlomer AR, Hatch L. **Underwater noise from large commercial ships—international collaboration for noise reduction**. *Encyclopedia of Maritime and Offshore Engineering* **2017** 1-9. DOI: 10.1002/9781118476406.emoe056 539. **TIDES: Trash Information and Data for Education Studies** 540. Law KL, Starr N, Siegler TR, Jambeck JR, Mallos NJ, Leonard GH. **The United States’ contribution of plastic waste to land and ocean**. *Sci Adv* (2020.0) **6**. DOI: 10.1126/sciadv.abd0288 541. Isobe A, Azuma T, Cordova MR. **A multilevel dataset of microplastic abundance in the world’s upper ocean and the Laurentian Great Lakes**. *Microplastics Nanoplastics* (2021.0) **1** 16. DOI: 10.1186/s43591-021-00013-z 542. Day RH, Shaw DG, Ignell SE. **The quantitative distribution and characteristics of marine debris in the North Pacific Ocean, 1984–88**. (1990.0) 185-211 543. Pichel WG, Veenstra TS, Churnside JH. **GhostNet marine debris survey in the Gulf of Alaska – Satellite guidance and aircraft observations**. *Mar Pollut Bull* (2012.0) **65** 28-41. DOI: 10.1016/j.marpolbul.2011.10.009 544. Lebreton L, Slat B, Ferrari F. **Evidence that the Great Pacific Garbage Patch is rapidly accumulating plastic**. *Sci Rep* (2018.0) **8** 4666. DOI: 10.1038/s41598-018-22939-w 545. Martínez-Vicente V, Clark JR, Corradi P. **Measuring marine plastic debris from space: initial assessment of observation requirements**. *Remote Sens* (2019.0) **11** 2443. DOI: 10.3390/rs11202443 546. Martinez-Vicente V. **The need for a dedicated marine plastic litter satellite mission**. *Nat Rev Earth Environ* (2022.0) **3** 728-729. DOI: 10.1038/s43017-022-00360-2 547. Canals M, Pham CK, Bergmann M. **The quest for seafloor macrolitter: a critical review of background knowledge, current methods and future prospects**. *Environ Res Lett* (2021.0) **16** 023001. DOI: 10.1088/1748-9326/abc6d4 548. Holmstrom A. **Plastic films on the bottom of the Skagerack**. *Nature* (1975.0) **255** 622-623. DOI: 10.1038/255622a0 549. Galgani F, Jaunet S, His E. **Distribution and abundance of debris on the continental shelf of the north-western Mediterranean Sea**. *Science* (1995.0) **31** 713-717. DOI: 10.1016/0025-326X(95)00055-R 550. Bergmann M, Klages M. **Increase of litter at the Arctic deep-sea observatory Hausgarten**. *Mar Pollut Bull* (2012.0) **64** 2734-2741. DOI: 10.1016/j.marpolbul.2012.09.018 551. Richardson K, Hardesty BD, Vince J, Wilcox C. **Global estimates of fishing gear lost to the ocean each year**. *Sci Adv* (2022.0) **8**. DOI: 10.1126/sciadv.abq0135 552. Goodman AJ, McIntyre J, Smith A, Fulton L, Walker TR, Brown CJ. **Retrieval of abandoned, lost, and discarded fishing gear in Southwest Nova Scotia, Canada: preliminary environmental and economic impacts to the commercial lobster industry**. *Mar Pollut Bull* (2021.0) **171** 112766. DOI: 10.1016/j.marpolbul.2021.112766 553. Teng G, Shan X, Jin X, Yang T. **Marine litter on the seafloors of the Bohai Sea, Yellow Sea and northern East China Sea**. *Mar Pollut Bull* (2021.0) **169** 112516. DOI: 10.1016/j.marpolbul.2021.112516 554. Scotti G, Esposito V, D’Alessandro M. **Seafloor litter along the Italian coastal zone: an integrated approach to identify sources of marine litter**. *Waste Manag* (2021.0) **124** 203-212. DOI: 10.1016/j.wasman.2021.01.034 555. Cau A, Franceschini S, Moccia D. **Scattered accumulation hotspots of macro-litter on the seafloor: insights for mitigation actions**. *Environ Pollut* (2022.0) **292** 118338. DOI: 10.1016/j.envpol.2021.118338 556. Browne MA, Galloway TS, Thompson RC. **Spatial patterns of plastic debris along estuarine shorelines**. *Env Sci Technol* (2010.0) **44** 3404-3409. DOI: 10.1021/es903784e 557. Claessens M, De Meester S, Van Landuyt L, De Clerck K, Janssen CR. **Occurrence and distribution of microplastics in marine sediments along the Belgian coast**. *Mar Pollut Bull* (2011.0) **62** 2199-2204. DOI: 10.1016/j.marpolbul.2011.06.030 558. Woodall LC, Sanchez-Vidal A, Canals M. **The deep sea is a major sink for microplastic debris**. *R Soc Open Sci* (2014.0) **1** 140317. DOI: 10.1098/rsos.140317 559. Kanhai LDK, Gårdfeldt K, Lyashevska O, Hassellöv M, Thompson RC, O’Connor I. **Microplastics in sub-surface waters of the Arctic Central Basin**. *Mar Pollut Bull* (2018.0) **130** 8-18. DOI: 10.1016/j.marpolbul.2018.03.011 560. Choy CA, Robison BH, Gagne TO. **The vertical distribution and biological transport of marine microplastics across the epipelagic and mesopelagic water column**. *Sci Rep* (2019.0) **9** 7843. DOI: 10.1038/s41598-019-44117-2 561. Pabortsava K, Lampitt RS. **High concentrations of plastic hidden beneath the surface of the Atlantic Ocean**. *Nat Commun* (2020.0) **11** 4073. DOI: 10.1038/s41467-020-17932-9 562. Obbard RW, Sadri S, Wong YQ, Khitun AA, Baker I, Thompson RC. **Global warming releases microplastic legacy frozen in Arctic Sea ice**. *Earths Future* (2014.0) **2** 315-320. DOI: 10.1002/2014EF000240 563. Peeken I, Primpke S, Beyer B. **Arctic sea ice is an important temporal sink and means of transport for microplastic**. *Nat Commun* (2018.0) **9** 1505. DOI: 10.1038/s41467-018-03825-5 564. Lambert S, Wagner M. **Characterisation of nanoplastics during the degradation of polystyrene**. *Chemosphere* (2016.0) **145** 265-268. DOI: 10.1016/j.chemosphere.2015.11.078 565. Ter Halle A, Jeanneau L, Martignac M. **Nanoplastic in the North Atlantic Subtropical Gyre**. *Env Sci Technol* (2017.0) **51** 13689-13697. DOI: 10.1021/acs.est.7b03667 566. Alimi OS, Farner Budarz J, Hernandez LM, Tufenkji N. **Microplastics and nanoplastics in aquatic environments: aggregation, deposition, and enhanced contaminant transport**. *Environ Sci Technol* (2018.0) **52** 1704-1724. DOI: 10.1021/acs.est.7b05559 567. Laist DW. **Overview of the biological effects of lost and discarded plastic debris in the marine environment**. *Mar Pollut Bull* (1987.0) **18** 319-326. DOI: 10.1016/S0025-326X(87)80019-X 568. Kühn S, Bravo Rebolledo EL, van Franeker JA, Bergmann M, Gutow L, Klages M. *Marine Anthropogenic Litter* (2015.0) 75-116. DOI: 10.1007/978-3-319-16510-3_4 569. Lavers JL, Bond AL. **Exceptional and rapid accumulation of anthropogenic debris on one of the world’s most remote and pristine islands**. *Proc Natl Acad Sci* (2017.0) **114** 6052-6055. DOI: 10.1073/pnas.1619818114 570. Barnes DKA, Morley SA, Bell J. **Marine plastics threaten giant Atlantic marine protected areas**. *Curr Biol* (2018.0) **28** R1137-R1138. DOI: 10.1016/j.cub.2018.08.064 571. Cózar A, Martí E, Duarte CM. **The Arctic Ocean as a dead end for floating plastics in the North Atlantic branch of the Thermohaline Circulation**. *Sci Adv* (2017.0) **3**. DOI: 10.1126/sciadv.1600582 572. Lacerda AL d. F, Rodrigues L dos S, van Sebille E. **Plastics in sea surface waters around the Antarctic Peninsula**. *Sci Rep* (2019.0) **9** 3977. DOI: 10.1038/s41598-019-40311-4 573. Allen S, Allen D, Moss K, Le Roux G, Phoenix VR, Sonke JE. **Examination of the ocean as a source for atmospheric microplastics**. *PLOS ONE* (2020.0) **15**. DOI: 10.1371/journal.pone.0232746 574. Arp HPH, Kühnel D, Rummel C. **Weathering plastics as a planetary boundary threat: exposure, fate, and hazards**. *Env Sci Technol* (2021.0) **55** 7246-7255. DOI: 10.1021/acs.est.1c01512 575. Hoornweg D, Bhada-Tata P. *What a Waste: A Global Review of Solid Waste Management* (2012.0) 576. Hoellein TJ, Rochman CM. **The “plastic cycle”: a watershed-scale model of plastic pools and fluxes**. *Front Ecol Environ* (2021.0) **19** 176-183. DOI: 10.1002/fee.2294 577. Borrelle SB, Ringma J, Law KL. **Predicted growth in plastic waste exceeds efforts to mitigate plastic pollution**. *Science* (2020.0) **369** 1515-1518. DOI: 10.1126/science.aba3656 578. Carlton JT, Chapman JW, Geller JB. **Tsunami-driven rafting: transoceanic species dispersal and implications for marine biogeography**. *Science* (2017.0) **357** 1402-1406. DOI: 10.1126/science.aao1498 579. Maximenko N, Hafner J, Kamachi M, MacFadyen A. **Numerical simulations of debris drift from the Great Japan Tsunami of 2011 and their verification with observational reports**. *Mar Pollut Bull* (2018.0) **132** 5-25. DOI: 10.1016/j.marpolbul.2018.03.056 580. Ryan PG, Moore CJ, van Franeker JA, Moloney CL. **Monitoring the abundance of plastic debris in the marine environment**. *Philos Trans R Soc Lond B Biol Sci* (2009.0) **364** 1999-2012. DOI: 10.1098/rstb.2008.0207 581. Boucher J, Friot D. *Primary Microplastics in the Oceans: A Global Evaluation of Sources* (2017.0). DOI: 10.2305/IUCN.CH.2017.01.en 582. 582OSPAR Commission. CEMP Guidelines for Marine Monitoring and Assessment of Beach Litter. OSPAR Commission. 2020; 29. DOI: 10.25607/OBP-1728. *CEMP Guidelines for Marine Monitoring and Assessment of Beach Litter* (2020.0) **29**. DOI: 10.25607/OBP-1728 583. Kole PJ, Löhr AJ, Van Belleghem FGAJ, Ragas AMJ. **Wear and tear of tyres: a stealthy source of microplastics in the environment**. *Int J Environ Res Public Health* (2017.0) **14** 1265. DOI: 10.3390/ijerph14101265 584. Wagner S, Hüffer T, Klöckner P, Wehrhahn M, Hofmann T, Reemtsma T. **Tire wear particles in the aquatic environment – a review on generation, analysis, occurrence, fate and effects**. *Water Res* (2018.0) **139** 83-100. DOI: 10.1016/j.watres.2018.03.051 585. Tian Z, Zhao H, Peter KT. **A ubiquitous tire rubber–derived chemical induces acute mortality in coho salmon**. *Science* (2021.0) **371** 185-189. DOI: 10.1126/science.abd6951 586. Werbowski LM, Gilbreath AN, Munno K. **Urban stormwater runoff: a major pathway for anthropogenic particles, black rubbery fragments, and other types of microplastics to urban receiving waters**. *ACS EST Water* (2021.0) **1** 1420-1428. DOI: 10.1021/acsestwater.1c00017 587. McIlwraith HK, Lin J, Erdle LM, Mallos N, Diamond ML, Rochman CM. **Capturing microfibers – marketed technologies reduce microfiber emissions from washing machines**. *Mar Pollut Bull* (2019.0) **139** 40-45. DOI: 10.1016/j.marpolbul.2018.12.012 588. Carr SA, Liu J, Tesoro AG. **Transport and fate of microplastic particles in wastewater treatment plants**. *Water Res* (2016.0) **91** 174-182. DOI: 10.1016/j.watres.2016.01.002 589. Sutton R, Franz A, Gilbreath A. *Understanding Microplastic Levels, Pathways, and Transport in the San Francisco Bay Region* (2019.0) 1-402 590. Moret-Ferguson S, Law KL, Proskurowski G, Murphy EK, Peacock EE, Reddy CM. **The size, mass, and composition of plastic debris in the western North Atlantic Ocean**. *Mar Pollut Bull* (2010.0) **60** 1873-1878. DOI: 10.1016/j.marpolbul.2010.07.020 591. Law KL, Narayan R. **Reducing environmental plastic pollution by designing polymer materials for managed end-of-life**. *Nat Rev Mater* (2022.0) **7** 104-116. DOI: 10.1038/s41578-021-00382-0 592. 592Ocean Conservancy. Building a Clean Swell: 2018 Report; 2018. Accessed December 14, 2022. https://oceanconservancy.org/wp-content/uploads/2018/07/Building-A-Clean-Swell.pdf.. *Building a Clean Swell: 2018 Report* (2018.0) 593. Onink V, Wichmann D, Delandmeter P, Sebille E. **The role of Ekman currents, geostrophy, and Stokes drift in the accumulation of floating microplastic**. *J Geophys Res Oceans* (2019.0) **124** 1474-1490. DOI: 10.1029/2018JC014547 594. Sterl MF, Delandmeter P, Sebille E. **Influence of barotropic tidal currents on transport and accumulation of floating microplastics in the global open ocean**. *J Geophys Res Oceans* (2020.0) **125**. DOI: 10.1029/2019JC015583 595. Cunningham HJ, Higgins C, van den Bremer TS. **The role of the unsteady surface wave-driven Ekman–Stokes flow in the accumulation of floating marine litter**. *J Geophys Res Oceans* (2022.0) **127**. DOI: 10.1029/2021JC018106 596. Khatmullina L, Chubarenko I. **Thin synthetic fibers sinking in still and convectively mixing water: laboratory experiments and projection to oceanic environment**. *Environ Pollut* (2021.0) **288** 117714. DOI: 10.1016/j.envpol.2021.117714 597. Brunner K, Kukulka T, Proskurowski G, Law KL. **Passive buoyant tracers in the ocean surface boundary layer: 2. Observations and simulations of microplastic marine debris**. *J Geophys Res Oceans* (2015.0) **120** 7559-7573. DOI: 10.1002/2015JC010840 598. Kukulka T. **Horizontal transport of buoyant material by turbulent jets in the upper ocean**. *J Phys Oceanogr* (2020.0) **50** 827-843. DOI: 10.1175/JPO-D-19-0276.1 599. Tarry DR, Essink S, Pascual A. **Frontal convergence and vertical velocity measured by drifters in the Alboran Sea**. *J Geophys Res Oceans* (2021.0) **126**. DOI: 10.1029/2020JC016614 600. Law KL, Morét-Ferguson S, Maximenko N a. **Plastic accumulation in the North Atlantic subtropical gyre**. *Science* (2010.0) **329** 1185-1188. DOI: 10.1126/science.1192321 601. Cózar A, Echevarría F, González-Gordillo JI. **Plastic debris in the open ocean**. *Proc Natl Acad Sci UA* (2014.0) **111** 10239-10244. DOI: 10.1073/pnas.1314705111 602. Eriksen M, Lebreton LCM, Carson HS. **Plastic pollution in the world’s oceans: more than 5 trillion plastic pieces weighing over 250,000 tons afloat at sea**. *PLoS One* (2014.0) **9**. DOI: 10.1371/journal.pone.0111913 603. Cózar A, Sanz-Martín M, Martí E. **Plastic accumulation in the Mediterranean Sea**. *PloS ONE* (2015.0) **10**. DOI: 10.1371/journal.pone.0121762 604. Suaria G, Avio CG, Mineo A. **The Mediterranean plastic soup: synthetic polymers in Mediterranean surface waters**. *Sci Rep* (2016.0) **6** 37551. DOI: 10.1038/srep37551 605. Pedrotti ML, Lombard F, Baudena A. **An integrative assessment of the plastic debris load in the Mediterranean Sea**. *Sci Total Env* (2022.0) **838** 155958. DOI: 10.1016/j.scitotenv.2022.155958 606. Goldstein MC, Titmus AJ, Ford M. **Scales of spatial heterogeneity of plastic marine debris in the northeast Pacific Ocean**. *PLoS ONE* (2013.0) **8**. DOI: 10.1371/journal.pone.0080020 607. Law KL, Moret-Ferguson SE, Goodwin DS. **Distribution of surface plastic debris in the eastern Pacific Ocean from an 11-year data set**. *Env Sci Technol* (2014.0) **48** 4732-4738. DOI: 10.1021/es4053076 608. Wilcox C, Hardesty BD, Law KL. **Abundance of floating plastic particles is increasing in the western North Atlantic Ocean**. *Environ Sci Technol* (2020.0) **54** 790-796. DOI: 10.1021/acs.est.9b04812 609. Kane IA, Clare MA, Miramontes E. **Seafloor microplastic hotspots controlled by deep-sea circulation**. *Science* (2020.0) **368** 1140-1145. DOI: 10.1126/science.aba5899 610. Kooi M, Nes EHV, Scheffer M, Koelmans AA. **Ups and downs in the ocean: effects of biofouling on vertical transport of microplastics**. *Env Sci Technol* (2017.0) **51** 7963-7971. DOI: 10.1021/acs.est.6b04702 611. Tibbetts J, Krause S, Lynch I, Sambrook Smith G. **Abundance, distribution, and drivers of microplastic contamination in urban river environments**. *Water* (2018.0) **10** 1597. DOI: 10.3390/w10111597 612. Rummel CD, Jahnke A, Gorokhova E, Kühnel D, Schmitt-Jansen M. **Impacts of biofilm formation on the fate and potential effects of microplastic in the aquatic environment**. *Environ Sci Technol Lett* (2017.0) **4** 258-267. DOI: 10.1021/acs.estlett.7b00164 613. Long M, Moriceau B, Gallinari M. **Interactions between microplastics and phytoplankton aggregates: impact on their respective fates**. *Mar Chem* (2015.0) **175** 39-46. DOI: 10.1016/j.marchem.2015.04.003 614. Cole M, Lindeque PK, Fileman E. **Microplastics alter the properties and sinking rates of zooplankton faecal pellets**. *Environ Sci Technol* (2016.0) **50** 3239-3246. DOI: 10.1021/acs.est.5b05905 615. Bergmann M, Wirzberger V, Krumpen T. **High quantities of microplastic in arctic deep-sea sediments from the Hausgarten observatory**. *Env Sci Technol* (2017.0) **51** 11000-11010. DOI: 10.1021/acs.est.7b03331 616. Fry DM, Fefer SI, Sileo L. **Ingestion of plastic debris by Laysan albatrosses and wedge-tailed shearwaters in the Hawaiian islands**. *Mar Pollut Bull* (1987.0) **18** 339-343. DOI: 10.1016/S0025-326X(87)80022-X 617. Boerger CM, Lattin GL, Moore SL, Moore CJ. **Plastic ingestion by planktivorous fishes in the North Pacific Central Gyre**. *Mar Pollut Bull* (2010.0) **60** 2275-2278. DOI: 10.1016/j.marpolbul.2010.08.007 618. Lusher AL, O’Donnell C, Officer R, O’Connor I. **Microplastic interactions with North Atlantic mesopelagic fish**. *ICES J Mar Sci J Cons* (2016.0) **73** 1214-1225. DOI: 10.1093/icesjms/fsv241 619. Keeling CD, Bacastow RB, Carter AF. **A three-dimensional model of atmospheric CO2 transport based on observed winds: 1. Analysis of observational data**. *Asp Clim Var Pac West Am* (1989.0) **55** 165-236. DOI: 10.1029/GM055p0165 620. Browne MA, Crump P, Niven SJ. **Accumulation of microplastic on shorelines woldwide: sources and sinks**. *Env Sci Technol* (2011.0) **45** 9175-9179. DOI: 10.1021/es201811s 621. Lebreton L, Egger M, Slat B. **A global mass budget for positively buoyant macroplastic debris in the ocean**. *Sci Rep* (2019.0) **9** 12922. DOI: 10.1038/s41598-019-49413-5 622. Barrett J, Chase Z, Zhang J. **Microplastic pollution in deep-sea sediments from the Great Australian Bight**. *Front Mar Sci* (2020.0) **7** 576170. DOI: 10.3389/fmars.2020.576170 623. Zhao S, Zettler ER, Bos RP, Lin P, Amaral-Zettler LA, Mincer TJ. **Large quantities of small microplastics permeate the surface ocean to abyssal depths in the South Atlantic Gyre**. *Glob Chang Biol* (2022.0) **28** 2991-3006. DOI: 10.1111/gcb.16089 624. Kako S, Isobe A, Magome S. **Sequential monitoring of beach litter using webcams**. *Mar Pollut Bull* (2010.0) **60** 775-779. DOI: 10.1016/j.marpolbul.2010.03.009 625. Kako S, Isobe A, Kataoka T. **Sequential webcam monitoring and modeling of marine debris abundance**. *Mar Pollut Bull* (2018.0) **132** 33-43. DOI: 10.1016/j.marpolbul.2018.04.075 626. Weiss L, Ludwig W, Heussner S. **The missing ocean plastic sink: gone with the rivers**. *Science* (2021.0) **373** 107-111. DOI: 10.1126/science.abe0290 627. Koelmans AA, Kooi M, Law KL, van Sebille E. **All is not lost: deriving a top-down mass budget of plastic at sea**. *Environ Res Lett* (2017.0) **12** 114028. DOI: 10.1088/1748-9326/aa9500 628. Zhu X, Nguyen B, You JB, Karakolis E, Sinton D, Rochman C. **Identification of microfibers in the environment using multiple lines of evidence**. *Env Sci Technol* (2019.0) **53** 11877-11887. DOI: 10.1021/acs.est.9b05262 629. Ward CP, Armstrong CJ, Walsh AN, Jackson JH, Reddy CM. **Sunlight converts polystyrene to carbon dioxide and dissolved organic carbon**. *Environ Sci Technol Lett* (2019.0) **6** 669-674. DOI: 10.1021/acs.estlett.9b00532 630. Andrady AL. **Weathering and fragmentation of plastic debris in the ocean environment**. *Mar Pollut Bull* (2022.0) **180** 113761. DOI: 10.1016/j.marpolbul.2022.113761 631. Kukulka T, Proskurowski G, Morét-Ferguson S, Meyer DW, Law KL. **The effect of wind mixing on the vertical distribution of buoyant plastic debris**. *Geophys Res Lett* (2012.0) **39** L07601. DOI: 10.1029/2012GL051116 632. Poulain M, Mercier MJ, Brach L. **Small microplastics as a main contributor to plastic mass balance in the North Atlantic Subtropical Gyre**. *Env Sci Technol* (2019.0) **53** 1157-1164. DOI: 10.1021/acs.est.8b05458 633. Di Benedetto M, Donohue J, Tremblay K, Edson E, Law KL. **Microplastics segregation by rise velocity at the ocean surface**. *Environ Res Lett*. DOI: 10.1088/1748-9326/acb505 634. Royer SJ, Wiggin K, Kogler M, Deheyn DD. **Degradation of synthetic and wood-based cellulose fabrics in the marine environment: comparative assessment of field, aquarium, and bioreactor experiments**. *Sci Total Environ* (2021.0) **791** 148060. DOI: 10.1016/j.scitotenv.2021.148060 635. Stark M. **Plausibility checks are needed in microplastic research to prevent misinterpretations**. *Environ Sci Technol* (2022.0) **56** 17495-17497. DOI: 10.1021/acs.est.2c05989 636. Ter Halle A, Ladirat L, Gendre X. **Understanding the fragmentation pattern of marine plastic debris**. *Environ Sci Technol* (2016.0) **50** 5668-5675. DOI: 10.1021/acs.est.6b00594 637. Gewert B, Plassmann MM, MacLeod M. **Pathways for degradation of plastic polymers floating in the marine environment**. *Environ Sci Process Impacts* (2015.0) **17** 1513-1521. DOI: 10.1039/C5EM00207A 638. Wayman C, Niemann H. **The fate of plastic in the ocean environment – a minireview**. *Environ Sci Process Impacts* (2021.0) **23** 198-212. DOI: 10.1039/D0EM00446D 639. 639Science Advice for Policy by European Academies (SAPEA). Biodegradability of Plastics in the Open Environment. SAPEA. 2020; 1–231. DOI: 10.26356/biodegradabilityplastics. *Biodegradability of Plastics in the Open Environment* (2020.0) 1-231. DOI: 10.26356/biodegradabilityplastics 640. Oberbeckmann S, Labrenz M. **Marine microbial assemblages on microplastics: diversity, adaptation, and role in degradation**. *Annu Rev Mar Sci* (2020.0) **12** 209-232. DOI: 10.1146/annurev-marine-010419-010633 641. Amaral-Zettler LA, Zettler ER, Mincer TJ. **Ecology of the plastisphere**. *Nat Rev Microbiol* (2020.0) **18** 139-151. DOI: 10.1038/s41579-019-0308-0 642. Ward CP, Armstrong CJ, Walsh AN, Jackson JH, Reddy CM. **Sunlight converts polystyrene to carbon dioxide and dissolved organic carbon**. *Environ Sci Technol Lett* (2019.0) **6** 669-674. DOI: 10.1021/acs.estlett.9b00532 643. Zhu L, Zhao S, Bittar TB, Stubbins A, Li D. **Photochemical dissolution of buoyant microplastics to dissolved organic carbon: rates and microbial impacts**. *J Hazard Mater* (2020.0) **383** 121065. DOI: 10.1016/j.jhazmat.2019.121065 644. Gulmine JV, Janissek PR, Heise HM, Akcelrud L. **Degradation profile of polyethylene after artificial accelerated weathering**. *Polym Degrad Stab* (2003.0) **79** 385-397. DOI: 10.1016/S0141-3910(02)00338-5 645. Andrady AL, Lavender Law K, Donohue J, Koongolla B. **Accelerated degradation of low-density polyethylene in air and in sea water**. *Sci Total Environ* (2022.0) **811** 151368. DOI: 10.1016/j.scitotenv.2021.151368 646. Klemchuk PP, Horng PL. **Transformation products of hindered phenolic antioxidants and colour development in polyolefins**. *Polym Degrad Stab* (1991.0) **34** 333-346. DOI: 10.1016/0141-3910(91)90127-D 647. Fisner M, Majer A, Taniguchi S, Bícego M, Turra A, Gorman D. **Colour spectrum and resin-type determine the concentration and composition of Polycyclic Aromatic Hydrocarbons (PAHs) in plastic pellets**. *Mar Pollut Bull* (2017.0) **122** 323-330. DOI: 10.1016/j.marpolbul.2017.06.072 648. Fanini L, Bozzeda F. **Dynamics of plastic resin pellets deposition on a microtidal sandy beach: informative variables and potential integration into sandy beach studies**. *Ecol Indic* (2018.0) **89** 309-316. DOI: 10.1016/j.ecolind.2018.02.027 649. Scott G, Scott G. *Atmospheric Oxidation and Antioxidants* (1993.0) 385-429. DOI: 10.1016/B978-0-444-89616-2.50012-4 650. Rabek JF. *Polymer Photodegradation: Mechanisms and Experimental Methods* (2012.0) 651. Nunes RW, Martin JR, Johnson JF. **Influence of molecular weight and molecular weight distribution on mechanical properties of polymers**. *Polym Eng Sci* (1982.0) **22** 205-228. DOI: 10.1002/pen.760220402 652. Cacuro TA, Freitas ASM, Waldman WR. **Demonstration of polymer photodegradation using a simple apparatus**. *J Chem Educ* (2018.0) **95** 2222-2226. DOI: 10.1021/acs.jchemed.7b00836 653. Rull F, Rodriguez J, Alia JM, Arroyo F, Edwards H. **Evolution of crystallinity in photodegraded polyethylene films studied by ft-raman spectroscopy**. *Macromol Symp* (1995.0) **94** 189-200. DOI: 10.1002/masy.19950940117 654. Sarkar AK, Rubin AE, Zucker I. **Engineered polystyrene-based microplastics of high environmental relevance**. *Environ Sci Technol* (2021.0) **55** 10491-10501. DOI: 10.1021/acs.est.1c02196 655. Kaczmarek H. **Changes to polymer morphology caused by u.v. irradiation: 1. Surface damage**. *Polymer* (1996.0) **37** 189-194. DOI: 10.1016/0032-3861(96)81086-X 656. Rabello MS, White JR. **The role of physical structure and morphology in the photodegradation behaviour of polypropylene**. *Polym Degrad Stab* (1997.0) **56** 55-73. DOI: 10.1016/S0141-3910(96)00202-9 657. Vincent PI. **The tough-brittle transition in thermoplastics**. *Polymer* (1960.0) **1** 425-444. DOI: 10.1016/0032-3861(60)90059-8 658. Perego G, Cella GD, Bastioli C. **Effect of molecular weight and crystallinity on poly(lactic acid) mechanical properties**. *J Appl Polym Sci* (1996.0) **59** 37-43. DOI: 10.1002/(SICI)1097-4628(19960103)59:1<37::AID-APP6>3.0.CO;2-N 659. Eyheraguibel B, Traikia M, Fontanella S. **Characterization of oxidized oligomers from polyethylene films by mass spectrometry and NMR spectroscopy before and after biodegradation by a Rhodococcus rhodochrous strain**. *Chemosphere* (2017.0) **184** 366-374. DOI: 10.1016/j.chemosphere.2017.05.137 660. Gewert B, Plassmann M, Sandblom O, MacLeod M. **Identification of chain scission products released to water by plastic exposed to ultraviolet light**. *Environ Sci Technol Lett* (2018.0) **5** 272-276. DOI: 10.1021/acs.estlett.8b00119 661. Lee YK, Romera-Castillo C, Hong S, Hur J. **Characteristics of microplastic polymer-derived dissolved organic matter and its potential as a disinfection byproduct precursor**. *Water Res* (2020.0) **175** 115678. DOI: 10.1016/j.watres.2020.115678 662. Lee YK, Murphy KR, Hur J. **Fluorescence signatures of dissolved organic matter leached from microplastics: polymers and additives**. *Environ Sci Technol* (2020.0) **54** 11905-11914. DOI: 10.1021/acs.est.0c00942 663. Song YK, Hong SH, Eo S, Han GM, Shim WJ. **Rapid production of micro- and nanoplastics by fragmentation of expanded polystyrene exposed to sunlight**. *Environ Sci Technol* (2020.0) **54** 11191-11200. DOI: 10.1021/acs.est.0c02288 664. Meides N, Menzel T, Poetzschner B. **Reconstructing the environmental degradation of polystyrene by accelerated weathering**. *Environ Sci Technol* (2021.0) **55** 7930-7938. DOI: 10.1021/acs.est.0c07718 665. Sorasan C, Edo C, González-Pleiter M. **Generation of nanoplastics during the photoageing of low-density polyethylene**. *Environ Pollut* (2021.0) **289** 117919. DOI: 10.1016/j.envpol.2021.117919 666. Liu Z, Zhu Y, Lv S. **Quantifying the dynamics of polystyrene microplastics UV-aging process**. *Environ Sci Technol Lett* (2022.0) **9** 50-56. DOI: 10.1021/acs.estlett.1c00888 667. Pfohl P, Wagner M, Meyer L. **Environmental degradation of microplastics: how to measure fragmentation rates to secondary micro- and nanoplastic fragments and dissociation into dissolved organics**. *Environ Sci Technol* (2022.0) **56** 11323-11334. DOI: 10.1021/acs.est.2c01228 668. Al-malaika S, Scott G. *Atmospheric Oxidation and Antioxidants* (1993.0) 161-224. DOI: 10.1016/B978-0-444-89615-5.50009-3 669. Scott G, Scott G. *Atmospheric Oxidation and Antioxidants* (1993.0) 121-160. DOI: 10.1016/B978-0-444-89615-5.50008-1 670. Scott G, Scott G. *Atmospheric Oxidation and Antioxidants* (1993.0) 83-119. DOI: 10.1016/B978-0-444-89615-5.50007-X 671. Gesenhues U. **Influence of titanium dioxide pigments on the photodegradation of poly(vinyl chloride)**. *Polym Degrad Stab* (2000.0) **68** 185-196. DOI: 10.1016/S0141-3910(99)00184-6 672. Jang J, Lee HS, Lyoo WS. **Effect of UV irradiation on cellulase degradation of cellulose acetate containing TiO2**. *Fibers Polym* (2007.0) **8** 19-24. DOI: 10.1007/BF02908155 673. Roy PK, Hakkarainen M, Varma IK, Albertsson AC. **Degradable polyethylene: fantasy or reality**. *Environ Sci Technol* (2011.0) **45** 4217-4227. DOI: 10.1021/es104042f 674. Thomas RT, Sandhyarani N. **Enhancement in the photocatalytic degradation of low density polyethylene–TiO2 nanocomposite films under solar irradiation**. *RSC Adv* (2013.0) **3** 14080-14087. DOI: 10.1039/c3ra42226g 675. Tu-morn M, Pairoh N, Sutapun W, Trongsatitkul T. **Effects of titanium dioxide nanoparticle on enhancing degradation of polylactic acid/low density polyethylene blend films**. *Mater Today Proc* (2019.0) **17** 2048-2061. DOI: 10.1016/j.matpr.2019.06.253 676. Nabi I, Bacha AUR, Ahmad F, Zhang L. **Application of titanium dioxide for the photocatalytic degradation of macro- and micro-plastics: a review**. *J Environ Chem Eng* (2021.0) **9** 105964. DOI: 10.1016/j.jece.2021.105964 677. Walsh AN, Mazzotta MG, Nelson TF, Reddy CM, Ward CP. **Synergy between sunlight, titanium dioxide, and microbes enhances cellulose diacetate degradation in the ocean**. *Environ Sci Technol* (2022.0) **56** 13810-13819. DOI: 10.1021/acs.est.2c04348 678. de Vos A, Aluwihare L, Youngs S. **The M/V X-Press Pearl nurdle spill: contamination of burnt plastic and unburnt nurdles along Sri Lanka’s beaches**. *ACS Environ Au* (2022.0) **2** 128-135. DOI: 10.1021/acsenvironau.1c00031 679. Turner A, Huang Wei. **Cadmium pigments in consumer products and their health risks**. *Sci Total Environ* (2019.0) **657** 1409-1418. DOI: 10.1016/j.scitotenv.2018.12.096 680. Witkowski A, Stec AA, Hull TR, Hurley MJ, Gottuk D, Hall JR. *SFPE Handbook of Fire Protection Engineering* (2016.0) 167-254. DOI: 10.1007/978-1-4939-2565-0_7 681. De-la-Torre GE, Dioses-Salinas DC, Pizarro-Ortega CI, Santillán L. **New plastic formations in the Anthropocene**. *Sci Total Environ* (2021.0) **754** 142216. DOI: 10.1016/j.scitotenv.2020.142216 682. Rubesinghe C, Brosché S, Withanage H, Pathragoda D, Karlsson T. *X-Press Pearl, a ‘New Kind of Oil Spill’ Consisting of a Toxic Mix of Plastics and Invisible Chemicals* (2022.0) 683. Partow H, Lacroix C, Le Floch S, Alcaro L. *X-Press Pearl Maritime Disaster Sri Lanka* (2022.0) 684. James BD, de Vos A, Aluwihare LI. **Divergent forms of pyroplastic: lessons learned from the M/V**. *ACS Environ Au* (2022.0) **2** 467-479. DOI: 10.1021/acsenvironau.2c00020 685. Perera ULHP, Subasinghe HCS, Ratnayake AS, Weerasingha WADB, Wijewardhana TDU. **Maritime pollution in the Indian Ocean after the MV**. *Mar Pollut Bull* (2022.0) **185** 114301. DOI: 10.1016/j.marpolbul.2022.114301 686. Sewwandi M, Amarathunga AAD, Wijesekara H, Mahatantila K, Vithanage M. **Contamination and distribution of buried microplastics in Sarakkuwa beach ensuing the MV**. *Mar Pollut Bull* (2022.0) **184** 114074. DOI: 10.1016/j.marpolbul.2022.114074 687. Tamada JA, Langer R. **Erosion kinetics of hydrolytically degradable polymers**. *Proc Natl Acad Sci USA* (1993.0) **90** 552-556. DOI: 10.1073/pnas.90.2.552 688. Uhrich KE, Cannizzaro SM, Langer RS, Shakesheff KM. **Polymeric systems for controlled drug release**. *Chem Rev* (1999.0) **99** 3181-3198. DOI: 10.1021/cr940351u 689. Wang GX, Huang D, Ji JH, Völker C, Wurm FR. **Seawater-degradable polymers—fighting the marine plastic pollution**. *Adv Sci* (2021.0) **8** 2001121. DOI: 10.1002/advs.202001121 690. Sang T, Wallis CJ, Hill G, Britovsek GJP. **Polyethylene terephthalate degradation under natural and accelerated weathering conditions**. *Eur Polym J* (2020.0) **136** 109873. DOI: 10.1016/j.eurpolymj.2020.109873 691. Allen NS, Edge M, Mohammadian M, Jones K. **Hydrolytic degradation of poly(ethylene terephthalate): importance of chain scission versus crystallinity**. *Eur Polym J* (1991.0) **27** 1373-1378. DOI: 10.1016/0014-3057(91)90237-I 692. Sammon C, Yarwood J, Everall N. **An FT–IR study of the effect of hydrolytic degradation on the structure of thin PET films**. *Polym Degrad Stab* (2000.0) **67** 149-158. DOI: 10.1016/S0141-3910(99)00104-4 693. Scott G. **Environmental stability of polymers**. *Polymers and the Environment* (1999.0) 38-67. DOI: 10.1039/9781847551726-00038 694. Scott G. **Antioxidant control of polymer biodegradation**. *Macromol Symp* (1999.0) **144** 113-125. DOI: 10.1002/masy.19991440111 695. Nelson TF, Reddy CM, Ward CP. **Product formulation controls the impact of biofouling on consumer plastic photochemical fate in the ocean**. *Env Sci Technol* (2021.0) **55** 8898-8907. DOI: 10.1021/acs.est.1c02079 696. Degli Innocenti F, Breton T. **Intrinsic biodegradability of plastics and ecological risk in the case of leakage**. *ACS Sustain Chem Eng* (2020.0) **8** 9239-9249. DOI: 10.1021/acssuschemeng.0c01230 697. Atlas RM. **Microbial degradation of petroleum hydrocarbons: an environmental perspective**. *Microbiol Rev* (1981.0) **45** 30. DOI: 10.1128/mr.45.1.180-209.1981 698. Mohanan N, Montazer Z, Sharma PK, Levin DB. **Microbial and enzymatic degradation of synthetic plastics**. *Front Microbiol* (2020.0) **11** 580709. DOI: 10.3389/fmicb.2020.580709 699. Vaksmaa A, Hernando-Morales V, Zeghal E, Niemann H, Joshi SJ, Deshmukh A, Sarma H. *Biotechnology for Sustainable Environment* (2021.0) 111-154. DOI: 10.1007/978-981-16-1955-7_5 700. Zadjelovic V, Erni-Cassola G, Obrador-Viel T. **A mechanistic understanding of polyethylene biodegradation by the marine bacterium Alcanivorax**. *J Hazard Mater* (2022.0) **436** 129278. DOI: 10.1016/j.jhazmat.2022.129278 701. Zrimec J, Kokina M, Jonasson S, Zorrilla F, Zelezniak A. **Plastic-degrading potential across the global microbiome correlates with recent pollution trends**. *mBio* (2021.0) **12** e02155-21. DOI: 10.1128/mBio.02155-21 702. Dawson AL, Kawaguchi S, King CK. **Turning microplastics into nanoplastics through digestive fragmentation by Antarctic krill**. *Nat Commun* (2018.0) **9** 1001. DOI: 10.1038/s41467-018-03465-9 703. Hodgson DJ, Bréchon AL, Thompson RC. **Ingestion and fragmentation of plastic carrier bags by the amphipod Orchestia gammarellus: effects of plastic type and fouling load**. *Mar Pollut Bull* (2018.0) **127** 154-159. DOI: 10.1016/j.marpolbul.2017.11.057 704. Zhao J, Xing B. **Fragmentation of microplastics by marine and freshwater rotifers** 705. Karamanlioglu M, Robson GD. **The influence of biotic and abiotic factors on the rate of degradation of poly(lactic) acid (PLA) coupons buried in compost and soil**. *Polym Degrad Stab* (2013.0) **98** 2063-2071. DOI: 10.1016/j.polymdegradstab.2013.07.004 706. Deroiné M, Le Duigou A, Corre YM. **Accelerated ageing of polylactide in aqueous environments: comparative study between distilled water and seawater**. *Polym Degrad Stab* (2014.0) **108** 319-329. DOI: 10.1016/j.polymdegradstab.2014.01.020 707. Haram LE, Carlton JT, Ruiz GM, Maximenko NA. **A plasticene lexicon**. *Mar Pollut Bull* (2020.0) **150** 110714. DOI: 10.1016/j.marpolbul.2019.110714 708. Domínguez-Hernández C, Villanova-Solano C, Sevillano-González M. **Plastitar: a new threat for coastal environments**. *Sci Total Environ* (2022.0) **839** 156261. DOI: 10.1016/j.scitotenv.2022.156261 709. Yang M, Zhang B, Xin X, Lee K, Chen B. **Microplastic and oil pollution in oceans: interactions and environmental impacts**. *Sci Total Environ* (2022.0) **838** 156142. DOI: 10.1016/j.scitotenv.2022.156142 710. Corcoran PL, Moore CJ, Jazvac K. **An anthropogenic marker horizon in the future rock record**. *GSA Today* (2014.0) **24** 4-8. DOI: 10.1130/GSAT-G198A.1 711. Corcoran PL, Jazvac K, Ballent A, Dellasala DA, Goldstein MI. *Encyclopedia of the Anthropocene* (2018.0) 163-170. DOI: 10.1016/B978-0-12-809665-9.10000-X 712. Corcoran PL, Jazvac K. **The consequence that is plastiglomerate**. *Nat Rev Earth Environ* (2020.0) **1** 6-7. DOI: 10.1038/s43017-019-0010-9 713. De-la-Torre GE, Pizarro-Ortega CI, Dioses-Salinas DC. **First record of plastiglomerates, pyroplastics, and plasticrusts in South America**. *Sci Total Environ* (2022.0) **833** 155179. DOI: 10.1016/j.scitotenv.2022.155179 714. Ellrich JA, Ehlers SM. **Field observations in pebble beach habitats link plastiglomerate to pyroplastic via pebble clasts**. *Mar Pollut Bull* (2022.0) **174** 113187. DOI: 10.1016/j.marpolbul.2021.113187 715. Gestoso I, Cacabelos E, Ramalhosa P, Canning-Clode J. **Plasticrusts: a new potential threat in the Anthropocene’s rocky shores**. *Sci Total Environ* (2019.0) **687** 413-415. DOI: 10.1016/j.scitotenv.2019.06.123 716. Ehlers SM, Ellrich JA. **First record of ‘plasticrusts’ and ‘pyroplastic’ from the Mediterranean Sea**. *Mar Pollut Bull* (2020.0) **151** 110845. DOI: 10.1016/j.marpolbul.2019.110845 717. Ehlers SM, Ellrich JA, Gestoso I. **Plasticrusts derive from maritime ropes scouring across raspy rocks**. *Mar Pollut Bull* (2021.0) **172** 112841. DOI: 10.1016/j.marpolbul.2021.112841 718. Furukuma S. **A study of ‘new plastic formations’ found in the Seto Inland Sea, Japan**. *Int J Sci Res Publ IJSRP* (2021.0) **11** 185-188. DOI: 10.29322/IJSRP.11.06.2021.p11427 719. Fernandino G, Elliff CI, Francischini H, Dentzien-Dias P. **Anthropoquinas: first description of plastics and other man-made materials in recently formed coastal sedimentary rocks in the southern hemisphere**. *Mar Pollut Bull* (2020.0) **154** 111044. DOI: 10.1016/j.marpolbul.2020.111044 720. Shiber JG. **Plastic pellets and tar on Spain’s Mediterranean beaches**. *Mar Pollut Bull* (1987.0) **18** 84-86. DOI: 10.1016/0025-326X(87)90573-X 721. Yang M, Zhang B, Xin X. **Microplastic-oil-dispersant agglomerates in the marine environment: formation mechanism and impact on oil dispersion**. *J Hazard Mater* (2022.0) **426** 127825. DOI: 10.1016/j.jhazmat.2021.127825 722. Gregory MR. **Environmental implications of plastic debris in marine settings—entanglement, ingestion, smothering, hangers-on, hitch-hiking and alien invasions**. *Philos Trans R Soc Lond B Biol Sci* (2009.0) **364** 2013-2025. DOI: 10.1098/rstb.2008.0265 723. van Franeker JA, Law KL. **Seabirds, gyres and global trends in plastic pollution**. *Env Pollut* (2015.0) **203** 89-96. DOI: 10.1016/j.envpol.2015.02.034 724. Wilcox C, Van Sebille E, Hardesty BD. **Threat of plastic pollution to seabirds is global, pervasive, and increasing**. *Proc Natl Acad Sci USA* (2015.0) **112** 11899-11904. DOI: 10.1073/pnas.1502108112 725. Botterell ZLR, Beaumont N, Dorrington T, Steinke M, Thompson RC, Lindeque PK. **Bioavailability and effects of microplastics on marine zooplankton: a review**. *Env Pollut* (2019.0) **245** 98-110. DOI: 10.1016/j.envpol.2018.10.065 726. Gove JM, Whitney JL, McManus MA. **Prey-size plastics are invading larval fish nurseries**. *Proc Natl Acad Sci UA* (2019.0) **116** 24143-24149. DOI: 10.1073/pnas.1907496116 727. Gouin T. **Toward an improved understanding of the ingestion and trophic transfer of microplastic particles: critical review and implications for future research**. *Environ Toxicol Chem* (2020.0) **39** 1119-1137. DOI: 10.1002/etc.4718 728. Kühn S, van Franeker JA. **Quantitative overview of marine debris ingested by marine megafauna**. *Mar Pollut Bull* (2020.0) **151** 110858. DOI: 10.1016/j.marpolbul.2019.110858 729. Walkinshaw C, Lindeque PK, Thompson R, Tolhurst T, Cole M. **Microplastics and seafood: lower trophic organisms at highest risk of contamination**. *Ecotoxicol Environ Saf* (2020.0) **190** 110066. DOI: 10.1016/j.ecoenv.2019.110066 730. López-Martínez S, Morales-Caselles C, Kadar J, Rivas ML. **Overview of global status of plastic presence in marine vertebrates**. *Glob Change Biol* (2021.0) **27** 728-737. DOI: 10.1111/gcb.15416 731. Savoca MS, McInturf AG, Hazen EL. **Plastic ingestion by marine fish is widespread and increasing**. *Glob Change Biol* (2021.0) **27** 2188-2199. DOI: 10.1111/gcb.15533 732. Kahane-Rapport SR, Czapanskiy MF, Fahlbusch JA. **Field measurements reveal exposure risk to microplastic ingestion by filter-feeding megafauna**. *Nat Commun* (2022.0) **13** 6327. DOI: 10.1038/s41467-022-33334-5 733. Lim KP, Lim PE, Yusoff S, Sun C, Ding J, Loh KH. **A meta-analysis of the characterisations of plastic ingested by fish globally**. *Toxics* (2022.0) **10** 186. DOI: 10.3390/toxics10040186 734. Santos RG, Machovsky-Capuska GE, Andrades R. **Plastic ingestion as an evolutionary trap: toward a holistic understanding**. *Science* (2021.0) **373** 56-60. DOI: 10.1126/science.abh0945 735. Zhao S, Ward JE, Danley M, Mincer TJ. **Field-based evidence for microplastic in marine aggregates and mussels: implications for trophic transfer**. *Environ Sci Technol* (2018.0) **52** 11038-11048. DOI: 10.1021/acs.est.8b03467 736. Jâms IB, Windsor FM, Poudevigne-Durance T, Ormerod SJ, Durance I. **Estimating the size distribution of plastics ingested by animals**. *Nat Commun* (2020.0) **11** 1594. DOI: 10.1038/s41467-020-15406-6 737. Bermúdez JR, Swarzenski PW. **A microplastic size classification scheme aligned with universal plankton survey methods**. *MethodsX* (2021.0) **8** 101516. DOI: 10.1016/j.mex.2021.101516 738. Santos RG, Andrades R, Fardim LM, Martins AS. **Marine debris ingestion and Thayer’s law – The importance of plastic color**. *Environ Pollut* (2016.0) **214** 585-588. DOI: 10.1016/j.envpol.2016.04.024 739. Lynch I, Dawson KA, Lead JR, Valsami-Jones E, Lead JR, Valsami-Jones E. *Frontiers of Nanoscience* (2014.0) **7** 127-156. DOI: 10.1016/B978-0-08-099408-6.00004-9 740. Savoca MS, Wohlfeil ME, Ebeler SE, Nevitt GA. **Marine plastic debris emits a keystone infochemical for olfactory foraging seabirds**. *Sci Adv* (2016.0) **2**. DOI: 10.1126/sciadv.1600395 741. Savoca MS, Tyson CW, McGill M, Slager CJ. **Odours from marine plastic debris induce food search behaviours in a forage fish**. *Proc R Soc B Biol Sci* (2017.0) **284** 20171000. DOI: 10.1098/rspb.2017.1000 742. Vroom RJE, Koelmans AA, Besseling E, Halsband C. **Aging of microplastics promotes their ingestion by marine zooplankton**. *Environ Pollut* (2017.0) **231** 987-996. DOI: 10.1016/j.envpol.2017.08.088 743. Botterell ZLR, Beaumont N, Cole M. **Bioavailability of microplastics to marine zooplankton: effect of shape and infochemicals**. *Environ Sci Technol* (2020.0) **54** 12024-12033. DOI: 10.1021/acs.est.0c02715 744. Polhill L, de Bruijn R, Amaral-Zettler L, Praetorius A, van Wezel A. **Daphnia magna’s favorite snack: biofouled plastics**. *Environ Toxicol Chem* (2022.0) **41** 1977-1981. DOI: 10.1002/etc.5393 745. Ward JE, Zhao S, Holohan BA. **Selective ingestion and egestion of plastic particles by the blue mussel (Mytilus edulis) and eastern oyster (Crassostrea virginica): implications for using bivalves as bioindicators of microplastic pollution**. *Environ Sci Technol* (2019.0) **53** 8776-8784. DOI: 10.1021/acs.est.9b02073 746. Mladinich K, Holohan BA, Shumway SE, Brown K, Ward JE. **Determining the properties that govern selective ingestion and egestion of microplastics by the blue mussel (Mytilus edulis) and eastern oyster (Crassostrea virginica)**. *Environ Sci Technol* (2022.0) **56** 15770-15779. DOI: 10.1021/acs.est.2c06402 747. Xu J, Rodríguez-Torres R, Rist S. **Unpalatable plastic: efficient taste discrimination of microplastics in planktonic copepods**. *Environ Sci Technol* (2022.0) **56** 6455-6465. DOI: 10.1021/acs.est.2c00322 748. Pitt JA, Aluru N, Hahn ME, Shumway SE, Ward JE. *Plastics in the Sea: Occurrence and Impacts* (2023.0) 749. Ribeiro F, Okoffo ED, O’Brien JW. **Out of sight but not out of mind: size fractionation of plastics bioaccumulated by field deployed oysters**. *J Hazard Mater Lett* (2021.0) **2** 100021. DOI: 10.1016/j.hazl.2021.100021 750. Kokilathasan N, Dittrich M. **Nanoplastics: detection and impacts in aquatic environments – a review**. *Sci Total Environ* (2022.0) **849** 157852. DOI: 10.1016/j.scitotenv.2022.157852 751. Lusher AL, Hollman PCH, Mendoza-Hill JJ. **Microplastics in fisheries and aquaculture: status of knowledge on their occurrence and implications for aquatic organisms and food safety**. *FAO Fish Aquac Tech Pap No 615 Rome Italy* 752. Zettler ER, Mincer TJ, Amaral-Zettler LA. **Life in the “plastisphere”: microbial communities on plastic marine debris**. *Env Sci Technol* (2013.0) **47** 7137-7146. DOI: 10.1021/es401288x 753. Amaneesh C, Anna Balan S, Silpa PS. **Gross negligence: impacts of microplastics and plastic leachates on phytoplankton community and ecosystem dynamics**. *Environ Sci Technol* (2022.0) **57** 5-24. DOI: 10.1021/acs.est.2c05817 754. Bowley J, Baker-Austin C, Porter A, Hartnell R, Lewis C. **Oceanic hitchhikers – assessing pathogen risks from marine microplastic**. *Trends Microbiol* (2021.0) **29** 107-116. DOI: 10.1016/j.tim.2020.06.011 755. Arias-Andres M, Klümper U, Rojas-Jimenez K, Grossart HP. **Microplastic pollution increases gene exchange in aquatic ecosystems**. *Environ Pollut* (2018.0) **237** 253-261. DOI: 10.1016/j.envpol.2018.02.058 756. Amaral-Zettler LA, Zettler ER, Mincer TJ, Klaassen MA, Gallager SM. **Biofouling impacts on polyethylene density and sinking in coastal waters: a macro/micro tipping point?**. *Water Res* (2021.0) **201** 117289. DOI: 10.1016/j.watres.2021.117289 757. Dell’Ariccia G, Phillips RA, van Franeker JA. **Comment on “Marine plastic debris emits a keystone infochemical for olfactory foraging seabirds” by Savoca et al**. *Sci Adv* (2017.0) **3**. DOI: 10.1126/sciadv.1700526 758. Ramsperger AFRM, Narayana VKB, Gross W. **Environmental exposure enhances the internalization of microplastic particles into cells**. *Sci Adv* (2020.0) **6**. DOI: 10.1126/sciadv.abd1211 759. Provencher JF, Ammendolia J, Rochman CM, Mallory ML. **Assessing plastic debris in aquatic food webs: what we know and don’t know about uptake and trophic transfer**. *Environ Rev* (2019.0) **27** 304-317. DOI: 10.1139/er-2018-0079 760. Akhbarizadeh R, Moore F, Keshavarzi B. **Investigating microplastics bioaccumulation and biomagnification in seafood from the Persian Gulf: a threat to human health?**. *Food Addit Contam Part A* (2019.0) **36** 1696-1708. DOI: 10.1080/19440049.2019.1649473 761. Ribeiro F, Okoffo ED, O’Brien JW. **Quantitative analysis of selected plastics in high-commercial-value Australian seafood by pyrolysis gas chromatography mass spectrometry**. *Environ Sci Technol* (2020.0) **54** 9408-9417. DOI: 10.1021/acs.est.0c02337 762. McIlwraith HK, Kim J, Helm P, Bhavsar SP, Metzger JS, Rochman CM. **Evidence of microplastic translocation in wild-caught fish and implications for microplastic accumulation dynamics in food webs**. *Environ Sci Technol* (2021.0) **55** 12372-12382. DOI: 10.1021/acs.est.1c02922 763. Browne MA, Dissanayake A, Galloway TS, Lowe DM, Thompson RC. **Ingested microscopic plastic translocates to the circulatory system of the mussel, Mytilus edulis (L)**. *Env Sci Technol* (2008.0) **42** 5026-5031. DOI: 10.1021/es800249a 764. Al-Sid-Cheikh M, Rowland SJ, Stevenson K, Rouleau C, Henry TB, Thompson RC. **Uptake, whole-body distribution, and depuration of nanoplastics by the scallop pecten maximus at environmentally realistic concentrations**. *Environ Sci Technol* (2018.0) **52** 14480-14486. DOI: 10.1021/acs.est.8b05266 765. De Sales-Ribeiro C, Brito-Casillas Y, Fernandez A, Caballero MJ. **An end to the controversy over the microscopic detection and effects of pristine microplastics in fish organs**. *Sci Rep* (2020.0) **10** 12434. DOI: 10.1038/s41598-020-69062-3 766. Kim J, Poirier DG, Helm PA, Bayoumi M, Rochman CM. **No evidence of spherical microplastics (10–300 µm) translocation in adult rainbow trout (Oncorhynchus mykiss) after a two-week dietary exposure**. *PloS One* (2020.0) **15**. DOI: 10.1371/journal.pone.0239128 767. Van Cauwenberghe L, Janssen CR. **Microplastics in bivalves cultured for human consumption**. *Environ Pollut* (2014.0) **193** 65-70. DOI: 10.1016/j.envpol.2014.06.010 768. Ribeiro F, Mitrano DM, Hacker C. **Short depuration of oysters intended for human consumption is effective at reducing exposure to nanoplastics**. *Environ Sci Technol* (2022.0) **56** 16716-16725. DOI: 10.1021/acs.est.2c02269 769. Clark NJ, Khan FR, Crowther C, Mitrano DM, Thompson RC. **Uptake, distribution and elimination of palladium-doped polystyrene nanoplastics in rainbow trout (Oncorhynchus mykiss) following dietary exposure**. *Sci Total Environ* (2022.0) **854** 158765. DOI: 10.1016/j.scitotenv.2022.158765 770. Clark NJ, Khan FR, Mitrano DM, Boyle D, Thompson RC. **Demonstrating the translocation of nanoplastics across the fish intestine using palladium-doped polystyrene in a salmon gut-sac**. *Environ Int* (2022.0) **159** 106994. DOI: 10.1016/j.envint.2021.106994 771. Urban RM, Jacobs JJ, Tomlinson MJ, Gavrilovic J, Black J, Peoc’h M. **Dissemination of wear particles to the liver, spleen, and abdominal lymph nodes of patients with hip or knee replacement**. *J Bone Jt Surg-Am Vol* (2000.0) **82** 457-476. DOI: 10.2106/00004623-200004000-00002 772. Alexis F, Pridgen E, Molnar LK, Farokhzad OC. **Factors affecting the clearance and biodistribution of polymeric nanoparticles**. *Mol Pharm* (2008.0) **5** 505-515. DOI: 10.1021/mp800051m 773. Richards L, Brown C, Stone MH, Fisher J, Ingham E, Tipper JL. **Identification of nanometre-sized ultra-high molecular weight polyethylene wear particles in samples retrieved in vivo**. *J Bone Joint Surg Br* (2008.0) **90** 1106-1113. DOI: 10.1302/0301-620X.90B8.20737 774. Brown TD, Habibi N, Wu D, Lahann J, Mitragotri S. **Effect of nanoparticle composition, size, shape, and stiffness on penetration across the blood–brain barrier**. *ACS Biomater Sci Eng* (2020.0) **6** 4916-4928. DOI: 10.1021/acsbiomaterials.0c00743 775. Aliyandi A, Reker-Smit C, Bron R, Zuhorn IS, Salvati A. **Correlating corona composition and cell uptake to identify proteins affecting nanoparticle entry into endothelial cells**. *ACS Biomater Sci Eng* (2021.0) **7** 5573-5584. DOI: 10.1021/acsbiomaterials.1c00804 776. Pednekar DD, Liguori MA, Marques CNH. **From static to dynamic: a review on the role of mucus heterogeneity in particle and microbial transport**. *ACS Biomater Sci Eng* (2022.0) **8** 2825-2848. DOI: 10.1021/acsbiomaterials.2c00182 777. Covernton GA, Davies HL, Cox KD. **A Bayesian analysis of the factors determining microplastics ingestion in fishes**. *J Hazard Mater* (2021.0) **413** 125405. DOI: 10.1016/j.jhazmat.2021.125405 778. Covernton GA, Cox KD, Fleming WL. **Large size (>100-µm) microplastics are not biomagnifying in coastal marine food webs of British Columbia, Canada**. *Ecol Appl* (2022.0) **32**. DOI: 10.1002/eap.2654 779. He S, Chi HY, Li C. **Distribution, bioaccumulation, and trophic transfer of palladium-doped nanoplastics in a constructed freshwater ecosystem**. *Environ Sci Nano* (2022.0) **9** 1353-1363. DOI: 10.1039/D1EN00940K 780. Wright SL, Kelly FJ. **Plastic and human health: a micro issue?**. *Environ Sci Technol* (2017.0) **51** 6634-6647. DOI: 10.1021/acs.est.7b00423 781. Smith M, Love DC, Rochman CM, Neff RA. **Microplastics in seafood and the implications for human health**. *Curr Environ Health Rep* (2018.0) **5** 375-386. DOI: 10.1007/s40572-018-0206-z 782. Zarus GM, Muianga C, Hunter CM, Pappas RS. **A review of data for quantifying human exposures to micro and nanoplastics and potential health risks**. *Sci Total Environ* (2021.0) **756** 144010. DOI: 10.1016/j.scitotenv.2020.144010 783. Cox KD, Covernton GA, Davies HL, Dower JF, Juanes F, Dudas SE. **Human consumption of microplastics**. *Env Sci Technol* (2019.0) **53** 7068-7074. DOI: 10.1021/acs.est.9b01517 784. Hantoro I, Löhr AJ, Van Belleghem FGAJ, Widianarko B, Ragas AMJ. **Microplastics in coastal areas and seafood: implications for food safety**. *Food Addit Contam Part Chem Anal Control Expo Risk Assess* (2019.0) **36** 674-711. DOI: 10.1080/19440049.2019.1585581 785. Barboza LGA, Lopes C, Oliveira P. **Microplastics in wild fish from North East Atlantic Ocean and its potential for causing neurotoxic effects, lipid oxidative damage, and human health risks associated with ingestion exposure**. *Sci Total Environ* (2020.0) **717** 134625. DOI: 10.1016/j.scitotenv.2019.134625 786. Danopoulos E, Jenner LC, Twiddy M, Rotchell JM. **Microplastic contamination of seafood intended for human consumption: a systematic review and meta-analysis**. *Env Health Perspect* (2020.0) **128** 126002. DOI: 10.1289/EHP7171 787. Dawson AL, Santana MFM, Miller ME, Kroon FJ. **Relevance and reliability of evidence for microplastic contamination in seafood: a critical review using Australian consumption patterns as a case study**. *Environ Pollut* (2021.0) **276** 116684. DOI: 10.1016/j.envpol.2021.116684 788. Domenech J, Marcos R. **Pathways of human exposure to microplastics, and estimation of the total burden**. *Curr Opin Food Sci* (2021.0) **39** 144-151. DOI: 10.1016/j.cofs.2021.01.004 789. Mortensen NP, Fennell TR, Johnson LM. **Unintended human ingestion of nanoplastics and small microplastics through drinking water, beverages, and food sources**. *NanoImpact* (2021.0) **21** 100302. DOI: 10.1016/j.impact.2021.100302 790. Lusher AL, Covernton GA, Andrady AL. *Plastics and the Ocean: Origin, Characterization, Fate, and Impacts* (2022.0) 367-388. DOI: 10.1002/9781119768432.ch13 791. Su L, Deng H, Li B. **The occurrence of microplastic in specific organs in commercially caught fishes from coast and estuary area of east China**. *J Hazard Mater* (2019.0) **365** 716-724. DOI: 10.1016/j.jhazmat.2018.11.024 792. Hosseinpour A, Chamani A, Mirzaei R, Mohebbi-Nozar SL. **Occurrence, abundance and characteristics of microplastics in some commercial fish of northern coasts of the Persian Gulf**. *Mar Pollut Bull* (2021.0) **171** 112693. DOI: 10.1016/j.marpolbul.2021.112693 793. Makhdoumi P, Hossini H, Pirsaheb M. **A review of microplastic pollution in commercial fish for human consumption**. *Rev Environ Health*. DOI: 10.1515/reveh-2021-0103 794. Thiele CJ, Hudson MD, Russell AE, Saluveer M, Sidaoui-Haddad G. **Microplastics in fish and fishmeal: an emerging environmental challenge?**. *Sci Rep* (2021.0) **11** 2045. DOI: 10.1038/s41598-021-81499-8 795. Abbasi S, Soltani N, Keshavarzi B, Moore F, Turner A, Hassanaghaei M. **Microplastics in different tissues of fish and prawn from the Musa Estuary, Persian Gulf**. *Chemosphere* (2018.0) **205** 80-87. DOI: 10.1016/j.chemosphere.2018.04.076 796. Akhbarizadeh R, Moore F, Keshavarzi B. **Investigating a probable relationship between microplastics and potentially toxic elements in fish muscles from northeast of Persian Gulf**. *Environ Pollut* (2018.0) **232** 154-163. DOI: 10.1016/j.envpol.2017.09.028 797. Atamanalp M, Köktürk M, Uçar A. **Microplastics in tissues (brain, gill, muscle and gastrointestinal) of Mullus barbatus and Alosa immaculata**. *Arch Environ Contam Toxicol* (2021.0) **81** 460-469. DOI: 10.1007/s00244-021-00885-5 798. Makhdoumi P, Hossini H, Nazmara Z, Mansouri K, Pirsaheb M. **Occurrence and exposure analysis of microplastic in the gut and muscle tissue of riverine fish in Kermanshah province of Iran**. *Mar Pollut Bull* (2021.0) **173** 112915. DOI: 10.1016/j.marpolbul.2021.112915 799. Selvam S, Manisha A, Roy PD. **Microplastics and trace metals in fish species of the Gulf of Mannar (Indian Ocean) and evaluation of human health**. *Environ Pollut* (2021.0) **291** 118089. DOI: 10.1016/j.envpol.2021.118089 800. Li Z, Chao M, He X. **Microplastic bioaccumulation in estuary-caught fishery resource**. *Environ Pollut* (2022.0) **306** 119392. DOI: 10.1016/j.envpol.2022.119392 801. Carbery M, O’Connor W, Palanisami T. **Trophic transfer of microplastics and mixed contaminants in the marine food web and implications for human health**. *Environ Int* (2018.0) **115** 400-409. DOI: 10.1016/j.envint.2018.03.007 802. Catarino AI, Macchia V, Sanderson WG, Thompson RC, Henry TB. **Low levels of microplastics (MP) in wild mussels indicate that MP ingestion by humans is minimal compared to exposure via household fibres fallout during a meal**. *Env Pollut* (2018.0) **237** 675-684. DOI: 10.1016/j.envpol.2018.02.069 803. Habib RZ, Kindi RA, Salem FA. **Microplastic contamination of chicken meat and fish through plastic cutting boards**. *Int J Environ Res Public Health* (2022.0) **19**. DOI: 10.3390/ijerph192013442 804. Habib RZ, Poulose V, Alsaidi R. **Plastic cutting boards as a source of microplastics in meat**. *Food Addit Contam Part Chem Anal Control Expo Risk Assess* (2022.0) **39** 609-619. DOI: 10.1080/19440049.2021.2017002 805. Luo Y, Gibson CT, Chuah C, Tang Y, Naidu R, Fang C. **Raman imaging for the identification of Teflon microplastics and nanoplastics released from non-stick cookware**. *Sci Total Environ* (2022.0) **851** 158293. DOI: 10.1016/j.scitotenv.2022.158293 806. Hirai H, Takada H, Ogata Y. **Organic micropollutants in marine plastics debris from the open ocean and remote and urban beaches**. *Mar Pollut Bull* (2011.0) **62** 1683-1692. DOI: 10.1016/j.marpolbul.2011.06.004 807. Mato Y, Isobe T, Takada H, Kanehiro H, Ohtake C, Kaminuma T. **Plastic resin pellets as a transport medium for toxic chemicals in the marine environment**. *Environ Sci Technol* (2001.0) **35** 318-324. DOI: 10.1021/es0010498 808. Rochman CM, Bergmann M, Gutow L, Klages M. *Marine Anthropogenic Litter* (2015.0) 117-140. DOI: 10.1007/978-3-319-16510-3_5 809. Gouin T, Roche N, Lohmann R, Hodges G. **A thermodynamic approach for assessing the environmental exposure of chemicals absorbed to microplastic**. *Env Sci Technol* (2011.0) **45** 1466-1472. DOI: 10.1021/es1032025 810. Bakir A, O’Connor IA, Rowland SJ, Hendriks AJ, Thompson RC. **Relative importance of microplastics as a pathway for the transfer of hydrophobic organic chemicals to marine life**. *Env Pollut* (2016.0) **219** 56-65. DOI: 10.1016/j.envpol.2016.09.046 811. Koelmans AA, Bakir A, Burton GA, Janssen CR. **Microplastic as a vector for chemicals in the aquatic environment: critical review and model-supported reinterpretation of empirical studies**. *Env Sci Technol* (2016.0) **50** 3315-3326. DOI: 10.1021/acs.est.5b06069 812. Ziccardi LM, Edgington A, Hentz K, Kulacki KJ, Kane Driscoll S. **Microplastics as vectors for bioaccumulation of hydrophobic organic chemicals in the marine environment: a state-of-the-science review**. *Env Toxicol Chem* (2016.0) **35** 1667-1676. DOI: 10.1002/etc.3461 813. Hartmann NB, Rist S, Bodin J. **Microplastics as vectors for environmental contaminants: exploring sorption, desorption, and transfer to biota**. *Integr Environ Assess Manag* (2017.0) **13** 488-493. DOI: 10.1002/ieam.1904 814. Isobe A, Kubo K, Tamura Y, Kako S, Nakashima E, Fujii N. **Selective transport of microplastics and mesoplastics by drifting in coastal waters**. *Mar Pollut Bull* (2014.0) **89** 324-330. DOI: 10.1016/j.marpolbul.2014.09.041 815. Heskett M, Takada H, Yamashita R. **Measurement of persistent organic pollutants (POPs) in plastic resin pellets from remote islands: toward establishment of background concentrations for International Pellet Watch**. *Mar Pollut Bull* (2012.0) **64** 445-448. DOI: 10.1016/j.marpolbul.2011.11.004 816. Takada H, Koro M, Kwan C, Nakajima T. **Chapter 10. Marine plastic pollution: chemical aspects and possible solutions**. *Overcoming Environmental Risks to Achieve Sustainable Development Goals, Current Topics in Environmental Health and Preventive Medicine* (2022.0). DOI: 10.1007/978-981-16-6249-2_10 817. Karapanagioti HK, Werner D, Takada H, Karapanagioti HK. *Hazardous Chemicals Associated with Plastics in the Marine Environment* (2019.0) 205-219. DOI: 10.1007/698_2018_256 818. Rochman CM, Hoh E, Hentschel BT, Kaye S. **Long-term field measurement of sorption of organic contaminants to five types of plastic pellets: implications for plastic marine debris**. *Environ Sci Technol* (2013.0) **47** 1646-1654. DOI: 10.1021/es303700s 819. Yamashita R, Tanaka K, Yeo BG, Takada H, Karapanagioti HK. *Hazardous Chemicals Associated with Plastics in the Marine Environment* (2019.0) 163-183. DOI: 10.1007/698_2018_299 820. Hasegawa T, Mizukawa K, Yeo BG, Sekioka T, Takada H, Nakaoka M. **The significance of trophic transfer of microplastics in the accumulation of plastic additives in fish: an experimental study using brominated flame retardants and UV stabilizers**. *Mar Pollut Bull* (2022.0) **185** 114343. DOI: 10.1016/j.marpolbul.2022.114343 821. Tanaka K, Takada H, Yamashita R, Mizukawa K, Fukuwaka MA, Watanuki Y. **Facilitated leaching of additive-derived PBDEs from plastic by seabirds’ stomach oil and accumulation in tissues**. *Env Sci Technol* (2015.0) **49** 11799-11807. DOI: 10.1021/acs.est.5b01376 822. Kühn S, Booth AM, Sørensen L, van Oyen A, van Franeker JA. **Transfer of additive chemicals from marine plastic debris to the stomach oil of northern fulmars**. *Front Environ Sci* (2020.0) **8**. DOI: 10.3389/fenvs.2020.00138 823. Sun B, Hu Y, Cheng H, Tao S. **Releases of brominated flame retardants (BFRs) from microplastics in aqueous medium: kinetics and molecular-size dependence of diffusion**. *Water Res* (2019.0) **151** 215-225. DOI: 10.1016/j.watres.2018.12.017 824. Turner A. **PBDEs in the marine environment: sources, pathways and the role of microplastics**. *Environ Pollut* (2022.0) **301** 118943. DOI: 10.1016/j.envpol.2022.118943 825. Tanaka K, Watanuki Y, Takada H. **In vivo accumulation of plastic-derived chemicals into seabird tissues**. *Curr Biol* (2020.0) **30** 723-728. DOI: 10.1016/j.cub.2019.12.037 826. Yamashita R, Hiki N, Kashiwada F. **Plastic additives and legacy persistent organic pollutants in the preen gland oil of seabirds sampled across the globe**. *Environ Monit Contam Res* (2021.0) **1** 97-112. DOI: 10.5985/emcr.20210009 827. Fossi MC, Coppola D, Baini M. **Large filter feeding marine organisms as indicators of microplastic in the pelagic environment: the case studies of the Mediterranean basking shark (Cetorhinus maximus) and fin whale (Balaenoptera physalus)**. *Mar Environ Res* (2014.0) **100** 17-24. DOI: 10.1016/j.marenvres.2014.02.002 828. Jang M, Shim WJ, Han GM, Rani M, Song YK, Hong SH. **Styrofoam debris as a source of hazardous additives for marine organisms**. *Environ Sci Technol* (2016.0) **50** 4951-4960. DOI: 10.1021/acs.est.5b05485 829. Tanaka K, Takada H, Yamashita R, Mizukawa K, Fukuwaka MA, Watanuki Y. **Accumulation of plastic-derived chemicals in tissues of seabirds ingesting marine plastics**. *Mar Pollut Bull* (2013.0) **69** 219-222. DOI: 10.1016/j.marpolbul.2012.12.010 830. Gassel M, Rochman CM. **The complex issue of chemicals and microplastic pollution: a case study in North Pacific lanternfish**. *Environ Pollut* (2019.0) **248** 1000-1009. DOI: 10.1016/j.envpol.2019.03.002 831. Tanaka K, van Franeker JA, Deguchi T, Takada H. **Piece-by-piece analysis of additives and manufacturing byproducts in plastics ingested by seabirds: implication for risk of exposure to seabirds**. *Mar Pollut Bull* (2019.0) **145** 36-41. DOI: 10.1016/j.marpolbul.2019.05.028 832. Tanaka K, Takada H, Ikenaka Y, Nakayama SMM, Ishizuka M. **Occurrence and concentrations of chemical additives in plastic fragments on a beach on the island of Kauai, Hawaii**. *Mar Pollut Bull* (2020.0) **150** 110732. DOI: 10.1016/j.marpolbul.2019.110732 833. Herzke D, Anker-Nilssen T, Nost TH. **Negligible impact of ingested microplastics on tissue concentrations of persistent organic pollutants in northern fulmars off coastal Norway**. *Env Sci Technol* (2016.0) **50** 1924-1933. DOI: 10.1021/acs.est.5b04663 834. Jang M, Shim WJ, Han GM, Cho Y, Moon Y, Hong SH. **Relative importance of aqueous leachate versus particle ingestion as uptake routes for microplastic additives (hexabromocyclododecane) to mussels**. *Environ Pollut* (2021.0) **270** 116272. DOI: 10.1016/j.envpol.2020.116272 835. Nakata H, Murata S, Filatreau J. **Occurrence and concentrations of benzotriazole UV stabilizers in marine organisms and sediments from the Ariake Sea, Japan**. *Environ Sci Technol* (2009.0) **43** 6920-6926. DOI: 10.1021/es900939j 836. Nakata H, Shinohara RI, Nakazawa Y. **Asia-Pacific mussel watch for emerging pollutants: distribution of synthetic musks and benzotriazole UV stabilizers in Asian and US coastal waters**. *Mar Pollut Bull* (2012.0) **64** 2211-2218. DOI: 10.1016/j.marpolbul.2012.07.049 837. Diao P, Chen Q, Wang R. **Phenolic endocrine-disrupting compounds in the Pearl River Estuary: Occurrence, bioaccumulation and risk assessment**. *Sci Total Environ* (2017.0) **584–585** 1100-1107. DOI: 10.1016/j.scitotenv.2017.01.169 838. Peng X, Fan Y, Jin J, Xiong S, Liu J, Tang C. **Bioaccumulation and biomagnification of ultraviolet absorbents in marine wildlife of the Pearl River Estuarine, South China Sea**. *Environ Pollut* (2017.0) **225** 55-65. DOI: 10.1016/j.envpol.2017.03.035 839. Shen R, Yang K, Cheng X. **Accumulation of polystyrene microplastics induces liver fibrosis by activating cGAS/STING pathway**. *Environ Pollut* (2022.0) **300** 118986. DOI: 10.1016/j.envpol.2022.118986 840. Cobellis L, Colacurci N, Trabucco E, Carpentiero C, Grumetto L. **Measurement of bisphenol A and bisphenol B levels in human blood sera from healthy and endometriotic women**. *Biomed Chromatogr BMC* (2009.0) **23** 1186-1190. DOI: 10.1002/bmc.1241 841. Liao C, Liu F, Alomirah H. **Bisphenol S in urine from the United States and seven Asian countries: occurrence and human exposures**. *Environ Sci Technol* (2012.0) **46** 6860-6866. DOI: 10.1021/es301334j 842. Wu LH, Zhang XM, Wang F. **Occurrence of bisphenol S in the environment and implications for human exposure: a short review**. *Sci Total Environ* (2018.0) **615** 87-98. DOI: 10.1016/j.scitotenv.2017.09.194 843. Dong J, Ma Y, Leng K. **Associations of urinary di-(2-ethylhexyl) phthalate metabolites with the residential characteristics of pregnant women**. *Sci Total Environ* (2020.0) **707** 135671. DOI: 10.1016/j.scitotenv.2019.135671 844. Choi JY, Lee J, Huh DA, Moon KW. **Urinary bisphenol concentrations and its association with metabolic disorders in the US and Korean populations**. *Environ Pollut* (2022.0) **295** 118679. DOI: 10.1016/j.envpol.2021.118679 845. Browne MA, Underwood AJ, Chapman MG, Williams R, Thompson RC, van Franeker JA. **Linking effects of anthropogenic debris to ecological impacts**. *Proc Biol Sci* (2015.0) **282** 20142929. DOI: 10.1098/rspb.2014.2929 846. Wilcox C, Puckridge M, Schuyler QA, Townsend K, Hardesty BD. **A quantitative analysis linking sea turtle mortality and plastic debris ingestion**. *Sci Rep* (2018.0) **8** 12536. DOI: 10.1038/s41598-018-30038-z 847. Lenz R, Enders K, Nielsen TG. **Microplastic exposure studies should be environmentally realistic**. *Proc Natl Acad Sci USA* (2016.0) **113** e4121-4122. DOI: 10.1073/pnas.1606615113 848. Burton GA. **Stressor exposures determine risk: so, why do fellow scientists continue to focus on superficial microplastics risk?**. *Env Sci Technol* (2017.0) **51** 13515-13516. DOI: 10.1021/acs.est.7b05463 849. Prata JC, da Costa JP, Lopes I, Andrady AL, Duarte AC, Rocha-Santos T. **A One Health perspective of the impacts of microplastics on animal, human and environmental health**. *Sci Total Environ* (2021.0) **777** 146094. DOI: 10.1016/j.scitotenv.2021.146094 850. Jacob H, Besson M, Swarzenski PW, Lecchini D, Metian M. **Effects of virgin micro- and nanoplastics on fish: trends, meta-analysis, and perspectives**. *Environ Sci Technol* (2020.0) **54** 4733-4745. DOI: 10.1021/acs.est.9b05995 851. Pisani XG, Lompré JS, Pires A, Greco LL. **Plastics in scene: a review of the effect of plastics in aquatic crustaceans**. *Environ Res* (2022.0) **212** 113484. DOI: 10.1016/j.envres.2022.113484 852. Han Y, Lian F, Xiao Z. **Potential toxicity of nanoplastics to fish and aquatic invertebrates: current understanding, mechanistic interpretation, and meta-analysis**. *J Hazard Mater* (2022.0) **427** 127870. DOI: 10.1016/j.jhazmat.2021.127870 853. Lehner R, Weder C, Petri-Fink A, Rothen-Rutishauser B. **Emergence of nanoplastic in the environment and possible impact on human health**. *Env Sci Technol* (2019.0) **53** 1748-1765. DOI: 10.1021/acs.est.8b05512 854. Stapleton PA. **Toxicological considerations of nano-sized plastics**. *AIMS Env Sci* (2019.0) **6** 367-378. DOI: 10.3934/environsci.2019.5.367 855. Schröter L, Ventura N. **Nanoplastic toxicity: insights and challenges from experimental model systems**. *Small Weinh Bergstr Ger* (2022.0) **18**. DOI: 10.1002/smll.202201680 856. Hu M, Palić D. **Micro- and nano-plastics activation of oxidative and inflammatory adverse outcome pathways**. *Redox Biol* (2020.0) **37** 101620. DOI: 10.1016/j.redox.2020.101620 857. Lee SE, Yi Y, Moon S, Yoon H, Park YS. **Impact of micro- and nanoplastics on mitochondria**. *Metabolites* (2022.0) **12**. DOI: 10.3390/metabo12100897 858. Lin S, Zhang H, Wang C. **Metabolomics reveal nanoplastic-induced mitochondrial damage in human liver and lung cells**. *Environ Sci Technol* (2022.0) **56** 12483-12493. DOI: 10.1021/acs.est.2c03980 859. Fleury JB, Baulin VA. **Microplastics destabilize lipid membranes by mechanical stretching**. *Proc Natl Acad Sci USA* (2021.0) **118**. DOI: 10.1073/pnas.2104610118 860. Vlacil AK, Bänfer S, Jacob R. **Polystyrene microplastic particles induce endothelial activation**. *PloS One* (2021.0) **16**. DOI: 10.1371/journal.pone.0260181 861. Weber A, Schwiebs A, Solhaug H. **Nanoplastics affect the inflammatory cytokine release by primary human monocytes and dendritic cells**. *Environ Int* (2022.0) **163** 107173. DOI: 10.1016/j.envint.2022.107173 862. Sarasamma S, Audira G, Siregar P. **Nanoplastics cause neurobehavioral impairments, reproductive and oxidative damages, and biomarker responses in zebrafish: throwing up alarms of widespread health risk of exposure**. *Int J Mol Sci* (2020.0) **21** 1410. DOI: 10.3390/ijms21041410 863. da Silva Brito WA, Mutter F, Wende K, Cecchini AL, Schmidt A, Bekeschus S. **Consequences of nano and microplastic exposure in rodent models: the known and unknown**. *Part Fibre Toxicol* (2022.0) **19** 28. DOI: 10.1186/s12989-022-00473-y 864. Pulvirenti E, Ferrante M, Barbera N. **Effects of nano- and microplastics on the inflammatory process: in vitro and in vivo studies systematic review**. *Front Biosci Landmark* (2022.0) **27** 287. DOI: 10.31083/j.fbl2710287 865. Hodkovicova N, Hollerova A, Svobodova Z, Faldyna M, Faggio C. **Effects of plastic particles on aquatic invertebrates and fish – a review**. *Environ Toxicol Pharmacol* (2022.0) **96** 104013. DOI: 10.1016/j.etap.2022.104013 866. Zhao L, Shi W, Hu F, Song X, Cheng Z, Zhou J. **Prolonged oral ingestion of microplastics induced inflammation in the liver tissues of C57BL/6J mice through polarization of macrophages and increased infiltration of natural killer cells**. *Ecotoxicol Environ Saf* (2021.0) **227** 112882. DOI: 10.1016/j.ecoenv.2021.112882 867. Yin J, Ju Y, Qian H. **Nanoplastics and microplastics may be damaging our livers**. *Toxics* (2022.0) **10** 586. DOI: 10.3390/toxics10100586 868. Kannan K, Vimalkumar K. **A review of human exposure to microplastics and insights into microplastics as obesogens**. *Front Endocrinol* (2021.0) **12** 724989. DOI: 10.3389/fendo.2021.724989 869. Völker J, Ashcroft F, Vedøy Å, Zimmermann L, Wagner M. **Adipogenic activity of chemicals used in plastic consumer products**. *Environ Sci Technol* (2022.0) **56** 2487-2496. DOI: 10.1021/acs.est.1c06316 870. Santos AL, Rodrigues CC, Oliveira M, Rocha TL. **Microbiome: a forgotten target of environmental micro(nano)plastics?**. *Sci Total Environ* (2022.0) **822** 153628. DOI: 10.1016/j.scitotenv.2022.153628 871. Li W, Chen X, Li M, Cai Z, Gong H, Yan M. **Microplastics as an aquatic pollutant affect gut microbiota within aquatic animals**. *J Hazard Mater* (2022.0) **423** 127094. DOI: 10.1016/j.jhazmat.2021.127094 872. Kang HM, Byeon E, Jeong H, Kim MS, Chen Q, Lee JS. **Different effects of nano- and microplastics on oxidative status and gut microbiota in the marine medaka Oryzias melastigma**. *J Hazard Mater* (2021.0) **405** 124207. DOI: 10.1016/j.jhazmat.2020.124207 873. Zhao Y, Qin Z, Huang Z, Bao Z, Luo T, Jin Y. **Effects of polyethylene microplastics on the microbiome and metabolism in larval zebrafish**. *Environ Pollut* (2021.0) **282** 117039. DOI: 10.1016/j.envpol.2021.117039 874. Medriano CA, Bae S. **Acute exposure to microplastics induces metabolic disturbances and gut dysbiosis in adult zebrafish (Danio rerio)**. *Ecotoxicol Environ Saf* (2022.0) **245** 114125. DOI: 10.1016/j.ecoenv.2022.114125 875. Yang M, Wang WX. **Recognition and movement of polystyrene nanoplastics in fish cells**. *Environ Pollut* (2023.0) **316** 120627. DOI: 10.1016/j.envpol.2022.120627 876. Hwang KS, Son Y, Kim SS. **Size-dependent effects of polystyrene nanoparticles (PS-NPs) on behaviors and endogenous neurochemicals in zebrafish larvae**. *Int J Mol Sci* (2022.0) **23**. DOI: 10.3390/ijms231810682 877. Hollóczki O, Gehrke S. **Can nanoplastics alter cell membranes?**. *Chemphyschem Eur J Chem Phys Phys Chem* (2020.0) **21** 9-12. DOI: 10.1002/cphc.201900481 878. Pitt JA, Trevisan R, Massarsky A, Kozal JS, Levin ED, Di Giulio RT. **Maternal transfer of nanoplastics to offspring in zebrafish (Danio rerio): a case study with nanopolystyrene**. *Sci Total Environ* (2018.0) **643** 324-334. DOI: 10.1016/j.scitotenv.2018.06.186 879. Zhao X, Liu Z, Ren X, Duan X. **Parental transfer of nanopolystyrene-enhanced tris(1,3-dichloro-2-propyl) phosphate induces transgenerational thyroid disruption in zebrafish**. *Aquat Toxicol* (2021.0) **236** 105871. DOI: 10.1016/j.aquatox.2021.105871 880. Trevisan R, Uzochukwu D, Di Giulio RT. **PAH sorption to nanoplastics and the Trojan horse effect as drivers of mitochondrial toxicity and PAH localization in zebrafish**. *Front Environ Sci* (2020.0) **8** 78. DOI: 10.3389/fenvs.2020.00078 881. Katsumiti A, Losada-Carrillo MP, Barros M, Cajaraville MP. **Polystyrene nanoplastics and microplastics can act as Trojan horse carriers of benzo(a)pyrene to mussel hemocytes in vitro**. *Sci Rep* (2021.0) **11** 22396. DOI: 10.1038/s41598-021-01938-4 882. Hu L, Zhao Y, Xu H. **Trojan horse in the intestine: a review on the biotoxicity of microplastics combined environmental contaminants**. *J Hazard Mater* (2022.0) **439** 129652. DOI: 10.1016/j.jhazmat.2022.129652 883. Annenkov VV, Danilovtseva EN, Zelinskiy SN, Pal’shin VA. **Submicro- and nanoplastics: how much can be expected in water bodies?**. *Environ Pollut* (2021.0) **278** 116910. DOI: 10.1016/j.envpol.2021.116910 884. Schwacke LH, Zolman ES, Balmer BC. **Anaemia, hypothyroidism and immune suppression associated with polychlorinated biphenyl exposure in bottlenose dolphins (Tursiops truncatus)**. *Proc Biol Sci* (2012.0) **279** 48-57. DOI: 10.1098/rspb.2011.0665 885. Desforges JP, Hall A, McConnell B. **Predicting global killer whale population collapse from PCB pollution**. *Science* (2018.0) **361** 1373-1376. DOI: 10.1126/science.aat1953 886. Hall AJ, McConnell BJ, Schwacke LH, Ylitalo GM, Williams R, Rowles TK. **Predicting the effects of polychlorinated biphenyls on cetacean populations through impacts on immunity and calf survival**. *Environ Pollut* (2018.0) **233** 407-418. DOI: 10.1016/j.envpol.2017.10.074 887. Cook PM, Robbins JA, Endicott DD. **Effects of aryl hydrocarbon receptor-mediated early life stage toxicity on lake trout populations in Lake Ontario during the 20th century**. *Environ Sci Technol* (2003.0) **37** 3864-3877. DOI: 10.1021/es034045m 888. Tillitt DE, Giesy JP. **Ecotoxicology of organochlorine chemicals in birds of the great lakes**. *Environ Toxicol Chem* (2013.0) **32** 490-492. DOI: 10.1002/etc.2109 889. Rochman CM, Lewison RL, Eriksen M, Allen H, Cook AM, Teh SJ. **Polybrominated diphenyl ethers (PBDEs) in fish tissue may be an indicator of plastic contamination in marine habitats**. *Sci Total Env* (2014.0) **476–477** 622-633. DOI: 10.1016/j.scitotenv.2014.01.058 890. Hanslik L, Huppertsberg S, Kämmer N, Knepper TP, Braunbeck T. **Rethinking the relevance of microplastics as vector for anthropogenic contaminants: adsorption of toxicants to microplastics during exposure in a highly polluted stream – analytical quantification and assessment of toxic effects in zebrafish (Danio rerio)**. *Sci Total Environ* (2022.0) **816** 151640. DOI: 10.1016/j.scitotenv.2021.151640 891. Luo Z, Zhou X, Su Y. **Environmental occurrence, fate, impact, and potential solution of tire microplastics: Similarities and differences with tire wear particles**. *Sci Total Environ* (2021.0) **795** 148902. DOI: 10.1016/j.scitotenv.2021.148902 892. Wright S, Borm PJA. **Applying existing particle paradigms to inhaled microplastic particles**. *Front Public Health* (2022.0) **10** 868822. DOI: 10.3389/fpubh.2022.868822 893. Li Y, Shi T, Li X. **Inhaled tire-wear microplastic particles induced pulmonary fibrotic injury via epithelial cytoskeleton rearrangement**. *Environ Int* (2022.0) **164** 107257. DOI: 10.1016/j.envint.2022.107257 894. Gualtieri M, Mantecca P, Cetta F, Camatini M. **Organic compounds in tire particle induce reactive oxygen species and heat-shock proteins in the human alveolar cell line A549**. *Environ Int* (2008.0) **34** 437-442. DOI: 10.1016/j.envint.2007.09.010 895. Gualtieri M, Mantecca P, Cetta F, Camatini M. **Organic compounds in tire particle induce reactive oxygen species and heat-shock proteins in the human alveolar cell line A549**. *Environ Int* (2008.0) **34** 437-442. DOI: 10.1016/j.envint.2007.09.010 896. Halle LL, Palmqvist A, Kampmann K, Jensen A, Hansen T, Khan FR. **Tire wear particle and leachate exposures from a pristine and road-worn tire to Hyalella azteca: comparison of chemical content and biological effects**. *Aquat Toxicol* (2021.0) **232** 105769. DOI: 10.1016/j.aquatox.2021.105769 897. Chibwe L, Parrott JL, Shires K. **A deep dive into the complex chemical mixture and toxicity of tire wear particle leachate in fathead minnow**. *Environ Toxicol Chem* (2022.0) **41** 1144-1153. DOI: 10.1002/etc.5140 898. Hiki K, Asahina K, Kato K. **Acute toxicity of a tire rubber-derived chemical, 6PPD quinone, to freshwater fish and crustacean species**. *Environ Sci Technol Lett* (2021.0) **8** 779-784. DOI: 10.1021/acs.estlett.1c00453 899. McIntyre JK, Prat J, Cameron J. **Treading water: tire wear particle leachate recreates an urban runoff mortality syndrome in coho but not chum salmon**. *Environ Sci Technol* (2021.0) **55** 11767-11774. DOI: 10.1021/acs.est.1c03569 900. Brinkmann M, Montgomery D, Selinger S. **Acute toxicity of the tire rubber-derived chemical 6PPD-quinone to four fishes of commercial, cultural, and ecological importance**. *Environ Sci Technol Lett* (2022.0) **9** 333-338. DOI: 10.1021/acs.estlett.2c00050 901. French BF, Baldwin DH, Cameron J. **Urban roadway runoff is lethal to juvenile coho, steelhead, and chinook salmonids, but not congeneric sockeye**. *Environ Sci Technol Lett* (2022.0) **9** 733-738. DOI: 10.1021/acs.estlett.2c00467 902. Hiki K, Yamamoto H. **The tire-derived chemical 6PPD-quinone is lethally toxic to the white-spotted char Salvelinus leucomaenis pluvius but not to two other salmonid species**. *Environ Sci Technol Lett* (2022.0) **9** 1050-1055. DOI: 10.1021/acs.estlett.2c00683 903. Everaert G, Van Cauwenberghe L, De Rijcke M. **Risk assessment of microplastics in the ocean: modelling approach and first conclusions**. *Environ Pollut* (2018.0) **242** 1930-1938. DOI: 10.1016/j.envpol.2018.07.069 904. Everaert G, De Rijcke M, Lonneville B. **Risks of floating microplastic in the global ocean**. *Env Pollut* (2020.0) **267** 115499. DOI: 10.1016/j.envpol.2020.115499 905. Rillig MC, Kim SW, Kim TY, Waldman WR. **The global plastic toxicity debt**. *Environ Sci Technol* (2021.0) **55** 2717-2719. DOI: 10.1021/acs.est.0c07781 906. Horton AA, Barnes DKA. **Microplastic pollution in a rapidly changing world: implications for remote and vulnerable marine ecosystems**. *Sci Total Environ* (2020.0) **738** 140349. DOI: 10.1016/j.scitotenv.2020.140349 907. Pereira JM, Rodríguez Y, Blasco-Monleon S, Porter A, Lewis C, Pham CK. **Microplastic in the stomachs of open-ocean and deep-sea fishes of the North-East Atlantic**. *Environ Pollut* (2020.0) **265** 115060. DOI: 10.1016/j.envpol.2020.115060 908. Justino AKS, Ferreira GVB, Schmidt N. **The role of mesopelagic fishes as microplastics vectors across the deep-sea layers from the Southwestern Tropical Atlantic**. *Environ Pollut* (2022.0) **300** 118988. DOI: 10.1016/j.envpol.2022.118988 909. Jamieson AJ, Brooks LSR, Reid WDK, Piertney SB, Narayanaswamy BE, Linley TD. **Microplastics and synthetic particles ingested by deep-sea amphipods in six of the deepest marine ecosystems on Earth**. *R Soc Open Sci* (2019.0) **6** 180667. DOI: 10.1098/rsos.180667 910. Bai Z, Wang N, Wang M. **Effects of microplastics on marine copepods**. *Ecotoxicol Environ Saf* (2021.0) **217** 112243. DOI: 10.1016/j.ecoenv.2021.112243 911. Zhang B, Tang X, Liu Q, Li L, Zhao Y, Zhao Y. **Different effecting mechanisms of two sized polystyrene microplastics on microalgal oxidative stress and photosynthetic responses**. *Ecotoxicol Environ Saf* (2022.0) **244** 114072. DOI: 10.1016/j.ecoenv.2022.114072 912. Dedman CJ, Christie-Oleza JA, Fernández-Juárez V, Echeveste P. **Cell size matters: nano- and micro-plastics preferentially drive declines of large marine phytoplankton due to co-aggregation**. *J Hazard Mater* (2022.0) **424** 127488. DOI: 10.1016/j.jhazmat.2021.127488 913. Köster M, Paffenhöfer GA. **Feeding of marine zooplankton on microplastic fibers**. *Arch Environ Contam Toxicol* (2022.0) **83** 129-141. DOI: 10.1007/s00244-022-00948-1 914. Rodríguez Torres R, Almeda R, Xu J. **The behavior of planktonic copepods minimizes the entry of microplastics in marine food webs**. *Environ Sci Technol*. DOI: 10.1021/acs.est.2c04660 915. Shore EA, deMayo JA, Pespeni MH. **Microplastics reduce net population growth and fecal pellet sinking rates for the marine copepod, Acartia tonsa**. *Environ Pollut* (2021.0) **284** 117379. DOI: 10.1016/j.envpol.2021.117379 916. Paffenhöfer GA, Köster M. **The effects of microplastics on Dolioletta gegenbauri (Tunicata, Thaliacea)**. *Arch Environ Contam Toxicol* (2020.0) **78** 94-105. DOI: 10.1007/s00244-019-00676-z 917. Kvale K, Prowe AEF, Chien CT, Landolfi A, Oschlies A. **The global biological microplastic particle sink**. *Sci Rep* (2020.0) **10** 16670. DOI: 10.1038/s41598-020-72898-4 918. Sendra M, Pereiro P, Yeste MP, Mercado L, Figueras A, Novoa B. **Size matters: zebrafish (Danio rerio) as a model to study toxicity of nanoplastics from cells to the whole organism**. *Environ Pollut* (2021.0) **268** 115769. DOI: 10.1016/j.envpol.2020.115769 919. Pantos O. **Microplastics: impacts on corals and other reef organisms**. *Emerg Top Life Sci* (2022.0) **6** 81-93. DOI: 10.1042/ETLS20210236 920. Marangoni LFB, Beraud E, Ferrier-Pagès C. **Polystyrene nanoplastics impair the photosynthetic capacities of Symbiodiniaceae and promote coral bleaching**. *Sci Total Env* (2022.0) **815** 152136. DOI: 10.1016/j.scitotenv.2021.152136 921. Corona E, Martin C, Marasco R, Duarte CM. **Passive and active removal of marine microplastics by a mushroom coral (Danafungia scruposa)**. *Front Mar Sci* (2020.0) **7** 128. DOI: 10.3389/fmars.2020.00128 922. Martin C, Corona E, Mahadik GA, Duarte CM. **Adhesion to coral surface as a potential sink for marine microplastics**. *Environ Pollut* (2019.0) **255** 113281. DOI: 10.1016/j.envpol.2019.113281 923. Rotjan RD, Sharp KH, Gauthier AE. **Patterns, dynamics and consequences of microplastic ingestion by the temperate coral, Astrangia poculata**. *Proc Biol Sci* (2019.0) **286** 20190726. DOI: 10.1098/rspb.2019.0726 924. Kvale K. **Implications of plastic pollution on global marine carbon cycling and climate**. *Emerg Top Life Sci* (2022.0) **6** 359-369. DOI: 10.1042/ETLS20220013 925. Porter A, Lyons BP, Galloway TS, Lewis C. **Role of marine snows in microplastic fate and bioavailability**. *Environ Sci Technol* (2018.0) **52** 7111-7119. DOI: 10.1021/acs.est.8b01000 926. Jacquin J, Cheng J, Odobel C. **Microbial Ecotoxicology of Marine Plastic Debris: A Review on Colonization and Biodegradation by the “Plastisphere”**. *Front Microbiol* (2019.0) **10** 865. DOI: 10.3389/fmicb.2019.00865 927. Kvale K, Prowe AEF, Chien CT, Landolfi A, Oschlies A. **Zooplankton grazing of microplastic can accelerate global loss of ocean oxygen**. *Nat Commun* (2021.0) **12** 2358. DOI: 10.1038/s41467-021-22554-w 928. Song YK, Hong SH, Jang M. **Large accumulation of micro-sized synthetic polymer particles in the sea surface microlayer**. *Environ Sci Technol* (2014.0) **48** 9014-9021. DOI: 10.1021/es501757s 929. Galgani L, Tzempelikou E, Kalantzi I. **Marine plastics alter the organic matter composition of the air-sea boundary layer, with influences on CO(2) exchange: a large-scale analysis method to explore future ocean scenarios**. *Sci Total Environ* (2023.0) **857** 159624. DOI: 10.1016/j.scitotenv.2022.159624 930. Engel A, Bange HW, Cunliffe M. **The ocean’s vital skin: toward an integrated understanding of the sea surface microlayer**. *Front Mar Sci* (2017.0) **4** 165. DOI: 10.3389/fmars.2017.00165 931. Ferrero L, Scibetta L, Markuszewski P. **Airborne and marine microplastics from an oceanographic survey at the Baltic Sea: an emerging role of air-sea interaction?**. *Sci Total Environ* (2022.0) **824** 153709. DOI: 10.1016/j.scitotenv.2022.153709 932. Larue C, Sarret G, Castillo-Michel H, Pradas Del Real AE. **A critical review on the impacts of nanoplastics and microplastics on aquatic and terrestrial photosynthetic organisms**. *Small Weinh Bergstr Ger* (2021.0) **17**. DOI: 10.1002/smll.202005834 933. Chae Y, Hong SH, An YJ. **Photosynthesis enhancement in four marine microalgal species exposed to expanded polystyrene leachate**. *Ecotoxicol Environ Saf* (2020.0) **189** 109936. DOI: 10.1016/j.ecoenv.2019.109936 934. Focardi A, Moore LR, Raina JB, Seymour JR, Paulsen IT, Tetu SG. **Plastic leachates impair picophytoplankton and dramatically reshape the marine microbiome**. *Microbiome* (2022.0) **10** 179. DOI: 10.1186/s40168-022-01369-x 935. Tetu SG, Sarker I, Schrameyer V. **Plastic leachates impair growth and oxygen production in Prochlorococcus, the ocean’s most abundant photosynthetic bacteria**. *Commun Biol* (2019.0) **2** 184. DOI: 10.1038/s42003-019-0410-x 936. Sjollema SB, Redondo-Hasselerharm P, Leslie HA, Kraak MHS, Vethaak AD. **Do plastic particles affect microalgal photosynthesis and growth?**. *Aquat Toxicol* (2016.0) **170** 259-261. DOI: 10.1016/j.aquatox.2015.12.002 937. Tetu SG, Sarker I, Moore LR. **How will marine plastic pollution affect bacterial primary producers?**. *Commun Biol* (2020.0) **3** 55. DOI: 10.1038/s42003-020-0789-4 938. Troost TA, Desclaux T, Leslie HA, van Der Meulen MD, Vethaak AD. **Do microplastics affect marine ecosystem productivity?**. *Mar Pollut Bull* (2018.0) **135** 17-29. DOI: 10.1016/j.marpolbul.2018.05.067 939. Dusaucy J, Gateuille D, Perrette Y, Naffrechoux E. **Microplastic pollution of worldwide lakes**. *Environ Pollut* (2021.0) **284** 117075. DOI: 10.1016/j.envpol.2021.117075 940. Horton AA, Walton A, Spurgeon DJ, Lahive E, Svendsen C. **Microplastics in freshwater and terrestrial environments: evaluating the current understanding to identify the knowledge gaps and future research priorities**. *Sci Total Environ* (2017.0) **586** 127-141. DOI: 10.1016/j.scitotenv.2017.01.190 941. Koutnik VS, Leonard J, Alkidim S. **Distribution of microplastics in soil and freshwater environments: global analysis and framework for transport modeling**. *Environ Pollut* (2021.0) **274** 116552. DOI: 10.1016/j.envpol.2021.116552 942. Talbot R, Chang H. **Microplastics in freshwater: a global review of factors affecting spatial and temporal variations**. *Environ Pollut* (2022.0) **292** 118393. DOI: 10.1016/j.envpol.2021.118393 943. Phuong NN, Duong TT, Le TPQ. **Microplastics in Asian freshwater ecosystems: current knowledge and perspectives**. *Sci Total Environ* (2022.0) **808** 151989. DOI: 10.1016/j.scitotenv.2021.151989 944. Li C, Busquets R, Campos LC. **Assessment of microplastics in freshwater systems: a review**. *Sci Total Environ* (2020.0) **707** 135578. DOI: 10.1016/j.scitotenv.2019.135578 945. Karunanidhi D, Subramani T, Srinivasamoorthy K, Yang Q. **Environmental chemistry, toxicity and health risk assessment of groundwater: environmental persistence and management strategies**. *Environ Res* (2022.0) **214** 113884. DOI: 10.1016/j.envres.2022.113884 946. Lee M, Kim H, Ryu HS. **Review on invasion of microplastic in our ecosystem and implications**. *Sci Prog* (2022.0) **105** 368504221140766. DOI: 10.1177/00368504221140766 947. Panno SV, Kelly WR, Scott J. **Microplastic contamination in karst groundwater systems**. *Groundwater* (2019.0) **57** 189-196. DOI: 10.1111/gwat.12862 948. Soriano MAJr, Deziel NC, Saiers JE. **Regional scale assessment of shallow groundwater vulnerability to contamination from unconventional hydrocarbon extraction**. *Environ Sci Technol* (2022.0) **56** 12126-12136. DOI: 10.1021/acs.est.2c00470 949. Belkhiri AH, Carre F, Quiot F. **State of knowledge and future research needs on microplastics in groundwater**. *J Water Health* (2022.0) **20** 1479-1496. DOI: 10.2166/wh.2022.048 950. Kukkola A, Krause S, Lynch I, Sambrook Smith GH, Nel H. **Nano and microplastic interactions with freshwater biota – current knowledge, challenges and future solutions**. *Environ Int* (2021.0) **152** 106504. DOI: 10.1016/j.envint.2021.106504 951. Collard F, Gasperi J, Gabrielsen GW, Tassin B. **Plastic particle ingestion by wild freshwater fish: a critical review**. *Environ Sci Technol* (2019.0) **53** 12974-12988. DOI: 10.1021/acs.est.9b03083 952. Cera A, Scalici M. **Freshwater wild biota exposure to microplastics: a global perspective**. *Ecol Evol* (2021.0) **11** 9904-9916. DOI: 10.1002/ece3.7844 953. O’Connor JD, Mahon AM, Ramsperger AFRM. **Microplastics in freshwater biota: a critical review of isolation, characterization, and assessment methods**. *Glob Chall* (2020.0) **4** 1800118. DOI: 10.1002/gch2.201800118 954. Ockenden A, Tremblay LA, Dikareva N, Simon KS. **Towards more ecologically relevant investigations of the impacts of microplastic pollution in freshwater ecosystems**. *Sci Total Environ* (2021.0) **792** 148507. DOI: 10.1016/j.scitotenv.2021.148507 955. Kirstein IV, Hensel F, Gomiero A. **Drinking plastics? Quantification and qualification of microplastics in drinking water distribution systems by µFTIR and Py-GCMS**. *Water Res* (2021.0) **188** 116519. DOI: 10.1016/j.watres.2020.116519 956. Pivokonsky M, Cermakova L, Novotna K, Peer P, Cajthaml T, Janda V. **Occurrence of microplastics in raw and treated drinking water**. *Sci Total Environ* (2018.0) **643** 1644-1651. DOI: 10.1016/j.scitotenv.2018.08.102 957. Wyles KJ, Pahl S, Thomas K, Thompson RC. **Factors that can undermine the psychological benefits of coastal environments: exploring the effect of tidal state, presence, and type of litter**. *Environ Behav* (2016.0) **48** 1095-1126. DOI: 10.1177/0013916515592177 958. Gillman MW. **Developmental origins of health and disease**. *N Engl J Med* (2005.0) **353** 1848-1850. DOI: 10.1056/NEJMe058187 959. Gillman MW. **Developmental origins of health and disease**. *N Engl J Med* (2005.0) **353** 1848-1850. DOI: 10.1056/NEJMe058187 960. Liu T, Liu S. **The impacts of coal dust on miners’ health: a review**. *Environ Res* (2020.0) **190** 109849. DOI: 10.1016/j.envres.2020.109849 961. Fan L, Liu S. **Respirable nano-particulate generations and their pathogenesis in mining workplaces: a review**. *Int J Coal Sci Technol* (2021.0) **8** 179-198. DOI: 10.1007/s40789-021-00412-w 962. Gasparotto J, Da Boit Martinello K. **Coal as an energy source and its impacts on human health**. *Energy Geosci* (2021.0) **2** 113-120. DOI: 10.1016/j.engeos.2020.07.003 963. Harrison RJ, Retzer K, Kosnett MJ. **Sudden deaths among oil and gas extraction workers resulting from oxygen deficiency and inhalation of hydrocarbon gases and vapors — United States, January 2010–March 2015**. *MMWR Morb Mortal Wkly Rep* (2016.0) **65** 6-9. DOI: 10.15585/mmwr.mm6501a2 964. Benson C, Dimopoulos C, Argyropoulos CD, Varianou Mikellidou C, Boustras G. **Assessing the common occupational health hazards and their health risks among oil and gas workers**. *Saf Sci* (2021.0) **140** 105284. DOI: 10.1016/j.ssci.2021.105284 965. Onyije FM, Hosseini B, Togawa K, Schüz J, Olsson A. **Cancer incidence and mortality among petroleum industry workers and residents living in oil producing communities: a systematic review and meta-analysis**. *Int J Environ Res Public Health* (2021.0) **18** 4343. DOI: 10.3390/ijerph18084343 966. **OSH answers fact sheets: sulfur dioxide** 967. Olaguer EP. **The potential near-source ozone impacts of upstream oil and gas industry emissions**. *J Air Waste Manag Assoc* (2012.0) **62** 966-977. DOI: 10.1080/10962247.2012.688923 968. Malley CS, Henze DK, Kuylenstierna JCI. **Updated global estimates of respiratory mortality in adults ≥30 years of age attributable to long-term ozone exposure**. *Environ Health Perspect* (2017.0) **125** 087021. DOI: 10.1289/EHP1390 969. 969Robin Cooley. Petition for Reconsideration of 2008 National Standards for Ozone and Petition for New Final Nonattainment Designation for the Uinta Basin, Utah. EarthJustice; 2012.. *Petition for Reconsideration of 2008 National Standards for Ozone and Petition for New Final Nonattainment Designation for the Uinta Basin, Utah* (2012.0) 970. Cabernard L, Pfister S, Hellweg S. **A new method for analyzing sustainability performance of global supply chains and its application to material resources**. *Sci Total Environ* (2019.0) **684** 164-177. DOI: 10.1016/j.scitotenv.2019.04.434 971. Stadler K, Wood R, Bulavskaya T. **EXIOBASE 3: Developing a time series of detailed environmentally extended multi-regional input-output tables**. *J Ind Ecol* (2018.0) **22** 502-515. DOI: 10.1111/jiec.12715 972. Wang C, Tu Y, Yu Z, Lu R. **PM2.5 and cardiovascular diseases in the elderly: an overview**. *Int J Environ Res Public Health* (2015.0) **12** 8187-8197. DOI: 10.3390/ijerph120708187 973. Murray CJL, Aravkin AY, Zheng P. **Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *The Lancet* (2020.0) **396** 1223-1249. DOI: 10.1016/S0140-6736(20)30752-2 974. Bekkar B, Pacheco S, Basu R, DeNicola N. **Association of air pollution and heat exposure with preterm birth, low birth weight, and stillbirth in the US: a systematic review**. *JAMA Netw Open* (2020.0) **3**. DOI: 10.1001/jamanetworkopen.2020.8243 975. Patel MM, Miller RL. **Air pollution and childhood asthma: recent advances and future directions**. *Curr Opin Pediatr* (2009.0) **21** 235-242. DOI: 10.1097/MOP.0b013e3283267726 976. Gauderman WJ, Urman R, Avol E. **Association of improved air quality with lung development in children**. *N Engl J Med* (2015.0) **372** 905-913. DOI: 10.1056/NEJMoa1414123 977. Dimakakou E, Johnston H, Streftaris G, Cherrie J. **Exposure to environmental and occupational particulate air pollution as a potential contributor to neurodegeneration and diabetes: a systematic review of epidemiological research**. *Int J Environ Res Public Health* (2018.0) **15** 1704. DOI: 10.3390/ijerph15081704 978. Volk HE, Perera F, Braun JM. **Prenatal air pollution exposure and neurodevelopment: a review and blueprint for a harmonized approach within ECHO**. *Environ Res* (2021.0) **196** 110320. DOI: 10.1016/j.envres.2020.110320 979. Rehman K, Fatima F, Waheed I, Akash MSH. **Prevalence of exposure of heavy metals and their impact on health consequences**. *J Cell Biochem* (2018.0) **119** 157-184. DOI: 10.1002/jcb.26234 980. Maciejczyk P, Chen LC, Thurston G. **The role of fossil fuel combustion metals in PM2.5 air pollution health associations**. *Atmosphere* (2021.0) **12** 34. DOI: 10.3390/atmos12091086 981. Rajagopalan S, Landrigan PJ, Longo DL. **Pollution and the heart**. *N Engl J Med* (2021.0) **385** 1881-1892. DOI: 10.1056/NEJMra2030281 982. Mølhave L, Bach B, Pedersen OF. **Human reactions to low concentrations of volatile organic compounds**. *Environ Int* (1986.0) **12** 167-175. DOI: 10.1016/0160-4120(86)90027-9 983. Ratner MH, Jabre JF, Ewing WM, Abou-Donia M, Oliver LC. **Amyotrophic lateral sclerosis—a case report and mechanistic review of the association with toluene and other volatile organic compounds**. *Am J Ind Med* (2018.0) **61** 251-260. DOI: 10.1002/ajim.22791 984. Tsai WT. **An overview of health hazards of volatile organic compounds regulated as indoor air pollutants**. *Rev Environ Health* (2019.0) **34** 81-89. DOI: 10.1515/reveh-2018-0046 985. Jørgensen NK, Cohr KH. **n-Hexane and its toxicologic effects – a review**. *Scand J Work Environ Health* (1981.0) **7** 157-168. DOI: 10.5271/sjweh.3107 986. Willis MD, Jusko TA, Halterman JS, Hill EL. **Unconventional natural gas development and pediatric asthma hospitalizations in Pennsylvania**. *Environ Res* (2018.0) **166** 402-408. DOI: 10.1016/j.envres.2018.06.022 987. Garcia-Gonzales DA, Shonkoff SBC, Hays J, Jerrett M. **Hazardous air pollutants associated with upstream oil and natural gas development: a critical synthesis of current peer-reviewed literature**. *Annu Rev Public Health* (2019.0) **40** 283-304. DOI: 10.1146/annurev-publhealth-040218-043715 988. Wilbur S, Bosch S. *Interaction Profile For: Benzene, Toluene, Ethylbenzene, and Xylenes (BTEX)* (2004.0) 1-154 989. Xu X, Zhang X, Carrillo G, Zhong Y, Kan H, Zhang B. **A systematic assessment of carcinogenicity of chemicals in hydraulic-fracturing fluids and flowback water**. *Environ Pollut* (2019.0) **251** 128-136. DOI: 10.1016/j.envpol.2019.04.016 990. Shonkoff SBC, Hays J, Finkel ML. **Environmental public health dimensions of shale and tight gas development**. *Environ Health Perspect* (2014.0) **122** 787-795. DOI: 10.1289/ehp.1307866 991. Bamberger M, Oswald RE. **Impacts of gas drilling on human and animal health**. *NEW Solut J Environ Occup Health Policy* (2012.0) **22** 51-77. DOI: 10.1371/journal.pone.0154164 992. Hays J, Shonkoff SBC, Carpenter DO. **Toward an understanding of the environmental and public health impacts of unconventional natural gas development: a categorical assessment of the peer-reviewed scientific literature, 2009–2015**. *PLOS ONE* (2016.0) **11**. PMID: 27096432 993. 993Concerned Health Professionals of New York and Physicians for Social Responsibility. Compendium of Scientific, Medical, and Media Findings Demonstrating Risks and Harms of Fracking and Associated Gas and Oil Infrastructure, Eighth Edition, April 28, 2022; 2022. Accessed November 3, 2022. https://concernedhealthny.org/compendium/.. *Compendium of Scientific, Medical, and Media Findings Demonstrating Risks and Harms of Fracking and Associated Gas and Oil Infrastructure, Eighth Edition, April 28, 2022* (2022.0) 994. Finkel ML. **Shale gas development and cancer incidence in southwest Pennsylvania**. *Public Health* (2016.0) **141** 198-206. DOI: 10.1016/j.puhe.2016.09.008 995. Heck JE, Park AS, Qiu J, Cockburn M, Ritz B. **Risk of leukemia in relation to exposure to ambient air toxics in pregnancy and early childhood**. *Int J Hyg Environ Health* (2014.0) **217** 662-668. DOI: 10.1016/j.ijheh.2013.12.003 996. McKenzie LM, Allshouse WB, Byers TE, Bedrick EJ, Serdar B, Adgate JL, Meliker J. **Childhood hematologic cancer and residential proximity to oil and gas development**. *PLOS ONE* (2017.0) **12**. DOI: 10.1371/journal.pone.0170423 997. Elliott EG, Ma X, Leaderer BP. **A community-based evaluation of proximity to unconventional oil and gas wells, drinking water contaminants, and health symptoms in Ohio**. *Environ Res* (2018.0) **167** 550-557. DOI: 10.1016/j.envres.2018.08.022 998. McKenzie LM, Allshouse W, Daniels S. **Congenital heart defects and intensity of oil and gas well site activities in early pregnancy**. *Environ Int* (2019.0) **132** 104949. DOI: 10.1016/j.envint.2019.104949 999. **Natural gas composition** 1000. **Pipeline incident 20 year trends** 1001. 1001National Transportation Safety Board. Overpressurization of Natural Gas Distribution System, Explosions, and Fires in Merrimack Valley, Massachusetts September 13, 2018; 2019. https://www.ntsb.gov/investigations/AccidentReports/Reports/PAR1902.pdf.. *Overpressurization of Natural Gas Distribution System, Explosions, and Fires in Merrimack Valley, Massachusetts September 13, 2018* (2019.0) 1002. Baker A. *Flammable, High-Pressure Industry in a Populated Coastal Flood Zone* (2019.0) 1003. Elliott EG, Trinh P, Ma X, Leaderer BP, Ward MH, Deziel NC. **Unconventional oil and gas development and risk of childhood leukemia: assessing the evidence**. *Sci Total Environ* (2017.0) **576** 138-147. DOI: 10.1016/j.scitotenv.2016.10.072 1004. Clough E. **Environmental justice and fracking: a review**. *Curr Opin Environ Sci Health* (2018.0) **3** 14-18. DOI: 10.1016/j.coesh.2018.02.005 1005. Nordgaard CL, Jaeger JM, Goldman JSW, Shonkoff SBC, Michanowicz DR. **Hazardous air pollutants in transmission pipeline natural gas: an analytic assessment**. *Environ Res Lett* (2022.0) **17** 104032. DOI: 10.1088/1748-9326/ac9295 1006. **Known and probable human carcinogens** 1007. Hertwich EG, Mateles SF, Pease WS, McKone TE. **Human toxicity potentials for life-cycle assessment and toxics release inventory risk screening**. *Environ Toxicol Chem* (2001.0) **20** 928-939. DOI: 10.1002/etc.5620200431 1008. 1008Bluefield Process Safety. External plant fires: what’s the likelihood? Bluefield Process Safety. Published September 7, 2017. Accessed January 12, 2023. https://www.bluefieldsafety.com/2017/09/external-plant-fires-whats-the-likelihood/. 1009. **Industries at a glance: petroleum and coal products manufacturing: NAICS 324** 1010. Lilis R, Daum S, Anderson H, Sirota M, Andrews G, Selikoff IJ. **Asbestos disease in maintenance workers of the chemical industry**. *Ann N Y Acad Sci* (1979.0) **330** 127-135. DOI: 10.1111/j.1749-6632.1979.tb18713.x 1011. Mehlman MA. **Dangerous and cancer-causing properties of products and chemicals in the oil-refining and petrochemical industries part IX: asbestos exposure and analysis of exposures**. *Ann NY Acad Sci* (1991.0) **643** 368-389. DOI: 10.1111/j.1749-6632.1991.tb24482.x 1012. Hanaoka T, Kawamura N, Hara K, Tsugane S. **Urinary bisphenol A and plasma hormone concentrations in male workers exposed to bisphenol A diglycidyl ether and mixed organic solvents**. *Occup Environ Med* (2002.0) **59** 625-628. DOI: 10.1136/oem.59.9.625 1013. Mundt KA, Dell LD, Crawford L, Gallagher AE. **Quantitative estimated exposure to vinyl chloride and risk of angiosarcoma of the liver and hepatocellular cancer in the US industry-wide vinyl chloride cohort: mortality update through 2013**. *Occup Environ Med* (2017.0) **74** 709-716. DOI: 10.1136/oemed-2016-104051 1014. Osinubi OY, Gochfeld M, Kipen HM. **Health effects of asbestos and nonasbestos fibers**. *Environ Health Perspect* (2000.0) **108** 665-674. DOI: 10.1289/ehp.00108s4665 1015. Reid A, de Klerk N, Musk AW. **Does exposure to asbestos cause ovarian cancer? A systematic literature review and meta-analysis**. *Cancer Epidemiol Biomarkers Prev* (2011.0) **20** 1287-1295. DOI: 10.1158/1055-9965.EPI-10-1302 1016. Rice JM, Boffetta P. **1,3-Butadiene, isoprene and chloroprene: reviews by the IARC monographs programme, outstanding issues, and research priorities in epidemiology**. *Chem Biol Interact* (2001.0) **135–136** 11-26. DOI: 10.1016/S0009-2797(01)00175-2 1017. Nielsen GD, Larsen ST, Wolkoff P. **Re-evaluation of the WHO (2010) formaldehyde indoor air quality guideline for cancer risk assessment**. *Arch Toxicol* (2017.0) **91** 35-61. DOI: 10.1007/s00204-016-1733-8 1018. Baranski B. **Effects of the workplace on fertility and related reproductive outcomes**. *Environ Health Perspect* (1993.0) **101** 81-90. DOI: 10.1289/ehp.93101s281 1019. Johnston J, Cushing L. **Chemical exposures, health, and environmental justice in communities living on the fenceline of industry**. *Curr Environ Health Rep* (2020.0) **7** 48-57. DOI: 10.1007/s40572-020-00263-8 1020. Kunstman B, Schaeffer E, Shaykevich A. **Environmental justice and refinery pollution** 1021. Hassani S, Namvar M, Ghoreishvandi M. **Menstrual disturbances and hormonal changes in women workers exposed to a mixture of organic solvents in a pharmaceutical company**. *Med J Islam Repub Iran* (2014.0) **28** 156. PMID: 25695014 1022. Marquès M, Domingo JL, Nadal M, Schuhmacher M. **Health risks for the population living near petrochemical industrial complexes. 2. Adverse health outcomes other than cancer**. *Sci Total Environ* (2020.0) **730** 139122. DOI: 10.1016/j.scitotenv.2020.139122 1023. Jephcote C, Brown D, Verbeek T, Mah A. **A systematic review and meta-analysis of haematological malignancies in residents living near petrochemical facilities**. *Environ Health* (2020.0) **19** 53. DOI: 10.1186/s12940-020-00582-1 1024. Lin CK, Hsu YT, Christiani DC, Hung HY, Lin RT. **Risks and burden of lung cancer incidence for residential petrochemical industrial complexes: a meta-analysis and application**. *Environ Int* (2018.0) **121** 404-414. DOI: 10.1016/j.envint.2018.09.018 1025. Rovira E, Cuadras A, Aguilar X. **Asthma, respiratory symptoms and lung function in children living near a petrochemical site**. *Environ Res* (2014.0) **133** 156-163. DOI: 10.1016/j.envres.2014.05.022 1026. Chang WW, Boonhat H, Lin RT. **Incidence of respiratory symptoms for residents living near a petrochemical industrial complex: a meta-analysis**. *Int J Environ Res Public Health* (2020.0) **17** 2474. DOI: 10.3390/ijerph17072474 1027. Huang C chun, Pan S chun, Chin W shan. **Maternal proximity to petrochemical industrial parks and risk of premature rupture of membranes**. *Environ Res* (2021.0) **194** 110688. DOI: 10.1016/j.envres.2020.110688 1028. Hill EL. **Shale gas development and infant health: evidence from Pennsylvania**. *J Health Econ* (2018.0) **61** 134-150. DOI: 10.1016/j.jhealeco.2018.07.004 1029. Dietrich H, Dietrich B. **Ludwig Rehn (1849–1930) – pioneering findings on the aetiology of bladder tumours**. *World J Urol* (2001.0) **19** 151-153. DOI: 10.1007/s003450100211 1030. 1030International Agency for Research on Cancer (IARC). Styrene, Styrene-7,8-Oxide, and Quinoline. International Agency for Research on Cancer; 2019. Accessed November 24, 2022. http://www.ncbi.nlm.nih.gov/books/NBK551039/.. *Styrene, Styrene-7,8-Oxide, and Quinoline* (2019.0) 1031. Kern DG, Kern E, Crausman RS, Clapp RW. **A retrospective cohort study of lung cancer incidence in nylon flock workers, 1998–2008**. *Int J Occup Environ Health* (2011.0) **17** 345-351. DOI: 10.1179/oeh.2011.17.4.345 1032. McCallum RI. **Occupational exposure to antimony compounds**. *J Environ Monit* (2005.0) **7** 1245. DOI: 10.1039/b509118g 1033. **Epistemonikos: database of the best evidence-based health care** 1034. Bravo-Soto G, Schulze C, Morel-Marambio M. **How comprehensive is the Epistemonikos database for identifying systematic reviews in health: a methodological study. Abstracts Accepted for the 26th Cochrane Colloquium, Santiago, Chile**. (2020.0) **1** 395. DOI: 10.1002/14651858.CD201901 1035. Liang B, Zhong Y, Huang Y. **Underestimated health risks: polystyrene micro- and nanoplastics jointly induce intestinal barrier dysfunction by ROS-mediated epithelial cell apoptosis**. *Part Fibre Toxicol* (2021.0) **18** 20. DOI: 10.1186/s12989-021-00414-1 1036. Zhang M, Xu L. **Transport of micro- and nanoplastics in the environment: Trojan-Horse effect for organic contaminants**. *Crit Rev Environ Sci Technol* 1-37. DOI: 10.1080/10643389.2020.1845531 1037. Koelmans AA, Redondo-Hasselerharm PE, Nor NHM, de Ruijter VN, Mintenig SM, Kooi M. **Risk assessment of microplastic particles**. *Nat Rev Mater* (2022.0) **7** 138-152. DOI: 10.1038/s41578-021-00411-y 1038. **National biomonitoring program** 1039. Ye X, Pirkle JL. *Personal Care and Consumer Product Chemicals and Metabolites* 1040. Fowler JF, Skinner SM, Belsito DV. **Allergic contact dermatitis from formaldehyde resins in permanent press clothing: an underdiagnosed cause of generalized dermatitis**. *J Am Acad Dermatol* (1992.0) **27** 962-968. DOI: 10.1016/0190-9622(92)70295-Q 1041. Yilmaz B, Terekeci H, Sandal S, Kelestimur F. **Endocrine disrupting chemicals: exposure, effects on human health, mechanism of action, models for testing and strategies for prevention**. *Rev Endocr Metab Disord* (2020.0) **21** 127-147. DOI: 10.1007/s11154-019-09521-z 1042. Street ME, Bernasconi S. **Endocrine-disrupting chemicals in human fetal growth**. *Int J Mol Sci* (2020.0) **21** 1430. DOI: 10.3390/ijms21041430 1043. Dutta S, Haggerty DK, Rappolee DA, Ruden DM. **Phthalate exposure and long-term epigenomic consequences: a review**. *Front Genet* (2020.0) **11** 405. DOI: 10.3389/fgene.2020.00405 1044. Denuzière A, Ghersi-Egea JF. **Cerebral concentration and toxicity of endocrine disrupting chemicals: the implication of blood-brain interfaces**. *NeuroToxicology* (2022.0) **91** 100-118. DOI: 10.1016/j.neuro.2022.04.004 1045. Etzel RA. **The special vulnerability of children**. *Int J Hyg Environ Health* (2020.0) **227** 113516. DOI: 10.1016/j.ijheh.2020.113516 1046. Kassotis CD, Vandenberg LN, Demeneix BA, Porta M, Slama R, Trasande L. **Endocrine-disrupting chemicals: economic, regulatory, and policy implications**. *Lancet Diabetes Endocrinol* (2020.0) **8** 719-730. DOI: 10.1016/S2213-8587(20)30128-5 1047. Gore AC, Chappell VA, Fenton SE. **Executive summary to EDC-2: The Endocrine Society’s second scientific statement on endocrine-disrupting chemicals**. *Endocr Rev* (2015.0) **36** 593-602. DOI: 10.1210/er.2015-1093 1048. Shih MK, Tain YL, Chen YW. **Resveratrol butyrate esters inhibit obesity caused by perinatal exposure to bisphenol A in female offspring rats**. *Molecules* (2021.0) **26** 4010. DOI: 10.3390/molecules26134010 1049. Oladipe TT, Uadia PO. **Occupational exposure to bisphenol A (BPA) is a risk factor for coronary heart disease**. *J Health Sci* (2022.0) **12** 1-6 1050. Cai S, Rao X, Ye J. **Relationship between urinary bisphenol a levels and cardiovascular diseases in the U.S. adult population, 2003–2014**. *Ecotoxicol Env Saf* (2020.0) **192** 110300. DOI: 10.1016/j.ecoenv.2020.110300 1051. Moreno-Gómez-Toledano R, Arenas MI, Vélez-Vélez E. **Bisphenol A exposure and kidney diseases: systematic review, meta-analysis, and NHANES 03–16 study**. *Biomolecules* (2021.0) **11** 1046. DOI: 10.3390/biom11071046 1052. Trasande L, Liu B, Bao W. **Phthalates and attributable mortality: a population-based longitudinal cohort study and cost analysis**. *Environ Pollut* (2022.0) **292** 118021. DOI: 10.1016/j.envpol.2021.118021 1053. Fu X, Xu J, Zhang R, Yu J. **The association between environmental endocrine disruptors and cardiovascular diseases: a systematic review and meta-analysis**. *Environ Res* (2020.0) **187** 109464-109464. DOI: 10.1016/j.envres.2020.109464 1054. Casey BJ, Giedd JN, Thomas KM. **Structural and functional brain development and its relation to cognitive development**. *Biol Psychol* (2000.0) **54** 241-257. DOI: 10.1016/S0301-0511(00)00058-2 1055. Lebel C, Deoni S. **The development of brain white matter microstructure**. *NeuroImage* (2018.0) **182** 207-218. DOI: 10.1016/j.neuroimage.2017.12.097 1056. Kwon D, Pfefferbaum A, Sullivan EV, Pohl KM. **Regional growth trajectories of cortical myelination in adolescents and young adults: longitudinal validation and functional correlates**. *Brain Imaging Behav* (2020.0) **14** 242-266. DOI: 10.1007/s11682-018-9980-3 1057. Cogswell ME, Coil E, Tian LH. **Health needs and use of services among children with developmental disabilities — United States, 2014–2018**. (2022.0) **71** 6. DOI: 10.15585/mmwr.mm7112a3 1058. Grandjean P, Landrigan PJ. **Neurobehavioural effects of developmental toxicity**. *Lancet Neurol* (2014.0) **13** 330-338. DOI: 10.1016/S1474-4422(13)70278-3 1059. Tran NQV, Miyake K. **Neurodevelopmental disorders and environmental toxicants: epigenetics as an underlying mechanism**. *Int J Genomics* (2017.0) **2017** 7526592. DOI: 10.1155/2017/7526592 1060. Rauh VA, Margolis A. **Research review: environmental exposures, neurodevelopment and child mental health – new paradigms for the study of brain and behavioral effects**. *J Child Psychol Psychiatry* (2016.0) **57** 775-793. DOI: 10.1111/jcpp.12537 1061. Tau GZ, Peterson BS. **Normal development of brain circuits**. *Neuropsychopharmacology* (2010.0) **35** 147-168. DOI: 10.1038/npp.2009.115 1062. Graham J 1063. Li N, Papandonatos GD, Calafat AM. **Identifying periods of susceptibility to the impact of phthalates on children’s cognitive abilities**. *Environ Res* (2019.0) **172** 604-614. DOI: 10.1016/j.envres.2019.03.009 1064. Engel SM, Miodovnik A, Canfield RL. **Prenatal phthalate exposure is associated with childhood behavior and executive functioning**. *Environ Health Perspect* (2010.0) **118** 565-571. DOI: 10.1289/ehp.0901470 1065. Engel SM, Villanger GD, Nethery RC. **Prenatal phthalates, maternal thyroid function, and risk of attention-deficit hyperactivity disorder in the Norwegian mother and child cohort**. *Environ Health Perspect* (2022.0) **126** 057004. DOI: 10.1289/EHP2358 1066. Swan SH, Liu F, Hines M. **Prenatal phthalate exposure and reduced masculine play in boys**. *Int J Androl* (2010.0) **33** 259-269. DOI: 10.1111/j.1365-2605.2009.01019.x 1067. Factor-Litvak P, Insel B, Calafat AM. **Persistent associations between maternal prenatal exposure to phthalates on child IQ at age 7 years**. *PLOS ONE* (2014.0) **9**. DOI: 10.1371/journal.pone.0114003 1068. Balalian AA, Whyatt RM, Liu X. **Prenatal and childhood exposure to phthalates and motor skills at age 11 years**. *Environ Res* (2019.0) **171** 416-427. DOI: 10.1016/j.envres.2019.01.046 1069. Landrigan PJ, Fuller R, Fisher S. **Pollution and children’s health**. *Sci Total Environ* (2019.0) **650** 2389-2394. DOI: 10.1016/j.scitotenv.2018.09.375 1070. Lam J, Lanphear BP, Bellinger D. **Developmental PBDE exposure and IQ/ADHD in childhood: a systematic review and meta-analysis**. *Environ Health Perspect* (2022.0) **125** 086001. DOI: 10.1289/EHP1632 1071. Weiss B. **Food additives and environmental chemicals as sources of childhood behavior disorders**. *J Am Acad Child Psychiatry US* (1982.0) **21**. DOI: 10.1016/S0002-7138(09)60913-4 1072. Ethier AA, Muckle G, Jacobson SW, Ayotte P, Jacobson JL, Saint-Amour D. **Assessing new dimensions of attentional functions in children prenatally exposed to environmental contaminants using an adapted Posner paradigm**. *Neurotoxicol Teratol* (2015.0) **51** 27-34. DOI: 10.1016/j.ntt.2015.07.005 1073. Jacobson JL, Jacobson SW, Humphrey HE. **Effects of exposure to PCBs and related compounds on growth and activity in children**. *Neurotoxicol Teratol* (1990.0) **12** 319-326. DOI: 10.1016/0892-0362(90)90050-M 1074. Berghuis SA, Soechitram SD, Hitzert MM, Sauer PJJ, Bos AF. **Prenatal exposure to polychlorinated biphenyls and their hydroxylated metabolites is associated with motor development of three-month-old infants**. *NeuroToxicology* (2013.0) **38** 124-130. DOI: 10.1016/j.neuro.2013.07.003 1075. Vreugdenhil HJI, Lanting CI, Mulder PGH, Boersma ER, Weisglas-Kuperus N. **Effects of prenatal PCB and dioxin background exposure on cognitive and motor abilities in Dutch children at school age**. *J Pediatr* (2002.0) **140** 48-56. DOI: 10.1067/mpd.2002.119625 1076. Patandin S, Lanting CI, Mulder PG, Boersma ER, Sauer PJ, Weisglas-Kuperus N. **Effects of environmental exposure to polychlorinated biphenyls and dioxins on cognitive abilities in Dutch children at 42 months of age**. *J Pediatr* (1999.0) **134** 33-41. DOI: 10.1016/S0022-3476(99)70369-0 1077. Ejaredar M, Nyanza EC, Ten Eycke K, Dewey D. **Phthalate exposure and childrens neurodevelopment: a systematic review**. *Environ Res* (2015.0) **142** 51-60. DOI: 10.1016/j.envres.2015.06.014 1078. Ponsonby AL, Symeonides C, Saffery R. **Prenatal phthalate exposure, oxidative stress-related genetic vulnerability and early life neurodevelopment: a birth cohort study**. *NeuroToxicology* (2020.0) **80** 20-28. DOI: 10.1016/j.neuro.2020.05.006 1079. Radke EG, Braun JM, Nachman RM, Cooper GS. **Phthalate exposure and neurodevelopment: a systematic review and meta-analysis of human epidemiological evidence**. *Environ Int* (2020.0) **137** 105408. DOI: 10.1016/j.envint.2019.105408 1080. Inadera H. **Neurological effects of bisphenol A and its analogues**. *Int J Med Sci* (2015.0) **12** 926-936. DOI: 10.7150/ijms.13267 1081. Ejaredar M, Lee Y, Roberts DJ, Sauve R, Dewey D. **Bisphenol A exposure and children’s behavior: a systematic review**. *J Expo Sci Environ Epidemiol* (2017.0) **27** 175-183. DOI: 10.1038/jes.2016.8 1082. Braun JM, Yolton K, Dietrich KN. **Prenatal bisphenol A exposure and early childhood behavior**. *Environ Health Perspect* (2009.0) **117** 1945-1952. DOI: 10.1289/ehp.0900979 1083. Perera F, Vishnevetsky J, Herbstman JB. **Prenatal bisphenol a exposure and child behavior in an inner-city cohort**. *Environ Health Perspect* (2012.0) **120** 1190-1194. DOI: 10.1289/ehp.1104492 1084. Evans SF, Kobrosly RW, Barrett ES. **Prenatal bisphenol A exposure and maternally reported behavior in boys and girls**. *Neurotoxicology* (2014.0) **45** 91-99. DOI: 10.1016/j.neuro.2014.10.003 1085. Wang Z, Alderman MH, Asgari C, Taylor HS. **Fetal bisphenol-A induced changes in murine behavior and brain gene expression persisted in adult-aged offspring**. *Endocrinology* (2020.0) **161** bqaa164. DOI: 10.1210/endocr/bqaa164 1086. Engel A, Buhrke T, Kasper S. **The urinary metabolites of DINCH® have an impact on the activities of the human nuclear receptors ERα, ERβ, AR, PPARα and PPARγ**. *Toxicol Lett* (2018.0) **287** 83-91. DOI: 10.1016/j.toxlet.2018.02.006 1087. Kasper-Sonnenberg M, Koch HM, Apel P. **Time trend of exposure to the phthalate plasticizer substitute DINCH in Germany from 1999 to 2017: biomonitoring data on young adults from the Environmental Specimen Bank (ESB)**. *Int J Hyg Environ Health* (2019.0) **222** 1084-1092. DOI: 10.1016/j.ijheh.2019.07.011 1088. Rajkumar A, Luu T, Beal MA, Barton-Maclaren TS, Hales BF, Robaire B. **Phthalates and alternative plasticizers differentially affect phenotypic parameters in gonadal somatic and germ cell lines**. *Biol Reprod* (2022.0) **106** 613-627. DOI: 10.1093/biolre/ioab216 1089. Schaffert A, Arnold J, Karkossa I, Blüher M, von Bergen M, Schubert K. **The emerging plasticizer alternative DINCH and its metabolite MINCH induce oxidative stress and enhance inflammatory responses in human THP-1 macrophages**. *Cells* (2021.0) **10** 2367. DOI: 10.3390/cells10092367 1090. Saad N, Bereketoglu C, Pradhan A. **Di(isononyl) cyclohexane-1,2-dicarboxylate (DINCH) alters transcriptional profiles, lipid metabolism and behavior in zebrafish larvae**. *Heliyon* (2021.0) **7**. DOI: 10.1016/j.heliyon.2021.e07951 1091. Wu H, Ashcraft L, Whitcomb BW. **Parental contributions to early embryo development: influences of urinary phthalate and phthalate alternatives among couples undergoing IVF treatment**. *Hum Reprod*. DOI: 10.1093/humrep/dew301 1092. Wu H, Estill MS, Shershebnev A. **Preconception urinary phthalate concentrations and sperm DNA methylation profiles among men undergoing IVF treatment: a cross-sectional study**. *Hum Reprod* (2017.0) **32** 2159-2169. DOI: 10.1093/humrep/dex283 1093. Wu H, Olmsted A, Cantonwine DE. **Urinary phthalate and phthalate alternative metabolites and isoprostane among couples undergoing fertility treatment**. *Environ Res* (2017.0) **153** 1-7. DOI: 10.1016/j.envres.2016.11.003 1094. Martinez RM, Hauser R, Liang L. **Urinary concentrations of phenols and phthalate metabolites reflect extracellular vesicle microRNA expression in follicular fluid**. *Environ Int* (2019.0) **123** 20-28. DOI: 10.1016/j.envint.2018.11.043 1095. van ′t Erve TJ, Rosen EM, Barrett ES. **Phthalates and phthalate alternatives have diverse associations with oxidative stress and inflammation in pregnant women**. *Environ Sci Technol* (2019.0) **53** 3258-3267. DOI: 10.1021/acs.est.8b05729 1096. Lee G, Kim S, Bastiaensen M. **Exposure to organophosphate esters, phthalates, and alternative plasticizers in association with uterine fibroids**. *Environ Res* (2020.0) **189** 109874. DOI: 10.1016/j.envres.2020.109874 1097. Schwedler G, Conrad A, Rucic E. **Hexamoll® DINCH and DPHP metabolites in urine of children and adolescents in Germany. Human biomonitoring results of the German Environmental Survey GerES V, 2014–2017**. *Int J Hyg Environ Health* (2020.0) **229** 113397. DOI: 10.1016/j.ijheh.2019.09.004 1098. Woodward MJ, Obsekov V, Jacobson MH, Kahn LG, Trasande L. **Phthalates and sex steroid hormones among men From NHANES, 2013–2016**. *J Clin Endocrinol Metab* (2020.0) **105** e1225-e1234. DOI: 10.1210/clinem/dgaa039 1099. Zhang Y, Mustieles V, Yland J. **Association of parental preconception exposure to phthalates and phthalate substitutes with preterm birth**. *JAMA Netw Open* (2020.0) **3**. DOI: 10.1001/jamanetworkopen.2020.2159 1100. Seewoo BJ, Goodes LM, Mofflin L. **A systematic evidence map of research on human exposure to plastic and in vivo health impacts** 1101. Park J, Park C, Gye MC, Lee Y. **Assessment of endocrine-disrupting activities of alternative chemicals for bis(2-ethylhexyl)phthalate**. *Environ Res* (2019.0) **172** 10-17. DOI: 10.1016/j.envres.2019.02.001 1102. Sung CR, Kang HG, Hong JY, Kwack SJ. **Citrate ester substitutes for di-2-ethylhexyl phthalate: in vivo reproductive and in vitro cytotoxicity assessments**. *J Toxicol Environ Health A* (2020.0) **83** 589-595. DOI: 10.1080/15287394.2020.1798832 1103. Hirata-Koizumi M, Takahashi M, Matsumoto M, Kawamura T, Ono A, Hirose A. **[Toxicity effects of phthalate substitute plasticizers used in toys]**. *Kokuritsu Iyakuhin Shokuhin Eisei Kenkyusho Hokoku* (2012.0) 31-42. PMID: 23243985 1104. Rochester JR, Bolden AL. **Bisphenol S and F: a systematic review and comparison of the hormonal activity of bisphenol A substitutes**. *Environ Health Perspect* (2015.0) **123** 643-650. DOI: 10.1289/ehp.1408989 1105. Le Fol V, Aït-Aïssa S, Sonavane M. **In vitro and in vivo estrogenic activity of BPA, BPF and BPS in zebrafish-specific assays**. *Ecotoxicol Environ Saf* (2017.0) **142** 150-156. DOI: 10.1016/j.ecoenv.2017.04.009 1106. Žalmanová T, Hošková K, Nevoral J. **Bisphenol S negatively affects the meotic maturation of pig oocytes**. *Sci Rep* (2017.0) **7** 485. DOI: 10.1038/s41598-017-00570-5 1107. Feng Y, Jiao Z, Shi J, Li M, Guo Q, Shao B. **Effects of bisphenol analogues on steroidogenic gene expression and hormone synthesis in H295R cells**. *Chemosphere* (2016.0) **147** 9-19. DOI: 10.1016/j.chemosphere.2015.12.081 1108. Zhang YF, Ren XM, Li YY. **Bisphenol A alternatives bisphenol S and bisphenol F interfere with thyroid hormone signaling pathway in vitro and in vivo**. *Environ Pollut* (2018.0) **237** 1072-1079. DOI: 10.1016/j.envpol.2017.11.027 1109. Pelch K, Wignall JA, Goldstone AE. **A scoping review of the health and toxicological activity of bisphenol A (BPA) structural analogues and functional alternatives**. *Toxicology* (2019.0) **424** 152235. DOI: 10.1016/j.tox.2019.06.006 1110. Herbstman JB, Sjödin A, Kurzon M. **Prenatal exposure to PBDEs and neurodevelopment**. *Environ Health Perspect* (2010.0) **118** 712-719. DOI: 10.1289/ehp.0901340 1111. Eskenazi B, Chevrier J, Rauch SA. **In utero and childhood polybrominated diphenyl ether (pbde) exposures and neurodevelopment in the CHAMACOS study**. *Environ Health Perspect* (2013.0) **121** 257-262. DOI: 10.1289/ehp.1205597 1112. Braun JM, Yolton K, Stacy SL. **Prenatal environmental chemical exposures and longitudinal patterns of child neurobehavior**. *NeuroToxicology* (2017.0) **62** 192-199. DOI: 10.1016/j.neuro.2017.07.027 1113. Chang CJ, Terrell ML, Marcus M. **Serum concentrations of polybrominated biphenyls (PBBs), polychlorinated biphenyls (PCBs) and polybrominated diphenyl ethers (PBDEs) in the Michigan PBB Registry 40 years after the PBB contamination incident**. *Environ Int* (2020.0) **137** 105526. DOI: 10.1016/j.envint.2020.105526 1114. Henderson AK, Rosen D, Miller GL. **Breast cancer among women exposed to polybrominated biphenyls**. *Epidemiol Camb Mass* (1995.0) **6** 544-546. DOI: 10.1097/00001648-199509000-00014 1115. Hoque A, Sigurdson AJ, Burau KD, Humphrey HEB, Hess KR, Sweeney AM. **Cancer among a Michigan cohort exposed to polybrominated biphenyls in 1973**. *Epidemiology* (1998.0) **9** 373-378. DOI: 10.1097/00001648-199807000-00005 1116. Small CM, Cheslack-Postava K, Terrell M. **Risk of spontaneous abortion among women exposed to polybrominated biphenyls**. *Environ Res* (2007.0) **105** 247-255. DOI: 10.1016/j.envres.2006.11.010 1117. Henck J, Mattsson J, Rezabek D, Carlson C, Rech R. **Developmental neurotoxicity of polybrominated biphenyls**. *Neurotoxicol Teratol* (1994.0) **16** 391-399. DOI: 10.1016/0892-0362(94)90028-0 1118. Deziel NC, Alfonso-Garrido J, Warren JL, Huang H, Sjodin A, Zhang Y. **Exposure to polybrominated diphenyl ethers and a polybrominated biphenyl and risk of thyroid cancer in women: single and multi-pollutant approaches**. *Cancer Epidemiol Biomarkers Prev* (2019.0) **28** 1755-1764. DOI: 10.1158/1055-9965.EPI-19-0526 1119. Lee DH, Steffes MW, Sjödin A, Jones RS, Needham LL, Jacobs DR. **Low dose of some persistent organic pollutants predicts Type 2 diabetes: a nested case-control study**. *Environ Health Perspect* (2010.0) **118** 1235-1242. DOI: 10.1289/ehp.0901480 1120. Lim JS, Lee DH, Jacobs DR. **Association of brominated flame retardants with diabetes and metabolic syndrome in the U.S. population, 2003–2004**. *Diabetes Care* (2008.0) **31** 1802-1807. DOI: 10.2337/dc08-0850 1121. Ploteau S, Cano-Sancho G, Volteau C. **Associations between internal exposure levels of persistent organic pollutants in adipose tissue and deep infiltrating endometriosis with or without concurrent ovarian endometrioma**. *Environ Int* (2017.0) **108** 195-203. DOI: 10.1016/j.envint.2017.08.019 1122. Sjödin A, Päpke O, McGahee E. **Concentration of polybrominated diphenyl ethers (PBDEs) in household dust from various countries**. *Chemosphere* (2008.0) **73** S131-S136. DOI: 10.1016/j.chemosphere.2007.08.075 1123. Chen A, Yolton K, Rauch SA. **Prenatal polybrominated diphenyl ether exposures and neurodevelopment in U.S. children through 5 years of age: the HOME study**. *Environ Health Perspect* (2014.0) **122** 856-862. DOI: 10.1289/ehp.1307562 1124. Roze E, Meijer L, Bakker A, Van BKNJA, Sauer PJJ, Bos AF. **Prenatal exposure to organohalogens, including brominated flame retardants, influences motor, cognitive, and behavioral performance at school age**. *Environ Health Perspect* (2009.0) **117** 1953-1958. DOI: 10.1289/ehp.0901015 1125. Doherty BT, Hammel SC, Daniels JL, Stapleton HM, Hoffman K. **Organophosphate esters: are these flame retardants & plasticizers affecting children’s health?**. *Curr Environ Health Rep* (2019.0) **6** 201-213. DOI: 10.1007/s40572-019-00258-0 1126. Mao JF, Li W, Ong CN, He Y, Jong MC, Gin KYH. **Assessment of human exposure to benzophenone-type UV filters: a review**. *Environ Int* (2022.0) **167** 107405. DOI: 10.1016/j.envint.2022.107405 1127. Morrison GC, Bekö G, Weschler CJ. **Dermal uptake of benzophenone-3 from clothing**. *Environ Sci Technol* (2017.0) **51** 11371-11379. DOI: 10.1021/acs.est.7b02623 1128. Bokyung K, Bareum K, Sol J, Pan-Gyi K, Kyunghee J, Covaci Adrian. **Major benzophenone concentrations and influence of food consumption among the general population in Korea, and the association with oxidative stress biomarker**. *Sci Total Environ* (2016.0) **565**. DOI: 10.1016/j.scitotenv.2016.05.009 1129. Kang H, Lee JP, Choi K. **Exposure to phthalates and environmental phenols in association with chronic kidney disease (CKD) among the general US population participating in multi-cycle NHANES (2005–2016)**. *Sci Total Environ* (2021.0) **791** 148343. DOI: 10.1016/j.scitotenv.2021.148343 1130. Ghazipura M, McGowan R, Arslan A, Hossain T. **Exposure to benzophenone-3 and reproductive toxicity: a systematic review of human and animal studies**. *Reprod Toxicol* (2017.0) **73** 175-183. DOI: 10.1016/j.reprotox.2017.08.015 1131. Teiri H, Samaei MR, Dehghani M. **The association of prenatal exposure to benzophenones with gestational age and offspring size at birth**. *Environ Sci Pollut Res Int* (2022.0) **29** 24682-24695. DOI: 10.1007/s11356-021-17634-9 1132. Li Y, Zhou Y, Cai Z. **Associations of benzotriazoles and benzothiazoles with estrogens and androgens among pregnant women: a cohort study with repeated measurements**. *Sci Total Environ* (2022.0) **838** 155998. DOI: 10.1016/j.scitotenv.2022.155998 1133. Chen X, Zhou Y, Hu C. **Prenatal exposure to benzotriazoles and benzothiazoles and cord blood mitochondrial DNA copy number: a prospective investigation**. *Environ Int* (2020.0) **143** 105920. DOI: 10.1016/j.envint.2020.105920 1134. Nagayoshi H, Kakimoto K, Takagi S, Konishi Y, Kajimura K, Matsuda T. **Benzotriazole ultraviolet stabilizers show potent activities as human aryl hydrocarbon receptor ligands**. *Environ Sci Technol* (2015.0) **49** 578-587. DOI: 10.1021/es503926w 1135. Kubota A, Terasaki M, Sakuragi Y. **Effects of benzotriazole UV stabilizers, UV-PS and UV-P, on the differentiation of splenic regulatory T cells via aryl hydrocarbon receptor**. *Ecotoxicol Environ Saf* (2022.0) **238** 113549. DOI: 10.1016/j.ecoenv.2022.113549 1136. Fent K, Chew G, Li J, Gomez E. **Benzotriazole UV-stabilizers and benzotriazole: antiandrogenic activity in vitro and activation of aryl hydrocarbon receptor pathway in zebrafish eleuthero-embryos**. *Sci Total Environ* (2014.0) **482–483** 125-136. DOI: 10.1016/j.scitotenv.2014.02.109 1137. **Butylhydroxytoluol (BHT) [MAK value documentation in German language, 2012]**. *The MAK-Collection for Occupational Health and Safety: Annual Thresholds and Classifications for the Workplace* (2012.0) 1-4. DOI: 10.1002/3527600418.mb12837d0052 1138. Wang W, Kannan K. **Quantitative identification of and exposure to synthetic phenolic antioxidants, including butylated hydroxytoluene, in urine**. *Environ Int* (2019.0) **128** 24-29. DOI: 10.1016/j.envint.2019.04.028 1139. Abd-Elhakim YM, Hashem MMM, Abo-El-Sooud K. **Involvement of tumor necrosis factor-α, interferon gamma-γ, and interleukins 1β, 6, and 10 in immunosuppression due to long-term exposure to five common food preservatives in rats**. *Gene* (2020.0) **742** 144590. DOI: 10.1016/j.gene.2020.144590 1140. Sun Z, Gao R, Chen X. **Exposure to butylated hydroxytoluene compromises endometrial decidualization during early pregnancy**. *Environ Sci Pollut Res Int* (2021.0) **28** 42024-42036. DOI: 10.1007/s11356-021-13720-0 1141. **Food & Drug Administration (FDA)**. *Code Fed Regul* (2022.0) **3** 1142. Liu R, Mabury SA. **Organophosphite antioxidants in indoor dust represent an indirect source of organophosphate esters**. *Environ Sci Technol* (2019.0) **53** 1805-1811. DOI: 10.1021/acs.est.8b05545 1143. Howe SR, Surana P, Jakupca MR, Borodinsky L. **Potential dietary exposure to p-nonylphenol from food-contact use of tris(nonylphenyl)phosphite (TNPP)**. *Food Addit Contam* (2001.0) **18** 1021-1039. DOI: 10.1080/02652030110050320 1144. Mottier P, Frank N, Dubois M. **LC-MS/MS analytical procedure to quantify tris(nonylphenyl)phosphite, as a source of the endocrine disruptors 4-nonylphenols, in food packaging materials**. *Food Addit Contam Part Chem Anal Control Expo Risk Assess* (2014.0) **31** 962-972. DOI: 10.1080/19440049.2014.896481 1145. 1145European Chemicals Agency (ECHA). Support Document for Identification of 4-Nonylphenol, Branched and Linear, Ethoxylated as Substances of Very High Concern Because, Due to Their Degradation to Substances of Very High Concern (4-Nonylphenol, Branched and Linear) with Endocrine Disrupting Properties, They Cause Probable Serious Effects to the Environment Which Give Rise to an Equivalent Level of Concern to Those of Cmr2 and Pbt/Vpvb3 Substances. 2013; 1–54. Accessed January 5, 2023. https://echa.europa.eu/documents/10162/9af34d5f-cd2f-4e63-859c-529bb39da7ae.. *Support Document for Identification of 4-Nonylphenol, Branched and Linear, Ethoxylated as Substances of Very High Concern Because, Due to Their Degradation to Substances of Very High Concern (4-Nonylphenol, Branched and Linear) with Endocrine Disrupting Properties, They Cause Probable Serious Effects to the Environment Which Give Rise to an Equivalent Level of Concern to Those of Cmr2 and Pbt/Vpvb3 Substances* (2013.0) 1-54 1146. **ECHA proposes a restriction on lead compounds in PVC articles** 1147. Turner A. **Concentrations and migratabilities of hazardous elements in second-hand children’s plastic toys**. *Environ Sci Technol* (2018.0) **52** 3110-3116. DOI: 10.1021/acs.est.7b04685 1148. Wagner S, Schlummer M. **Legacy additives in a circular economy of plastics: current dilemma, policy analysis, and emerging countermeasures**. *Resour Conserv Recycl* (2020.0) **158** 104800. DOI: 10.1016/j.resconrec.2020.104800 1149. Turner A, Filella M. **Hazardous metal additives in plastics and their environmental impacts**. *Environ Int* (2021.0) **156** 106622-106622. DOI: 10.1016/j.envint.2021.106622 1150. Njati SY, Maguta MM. **Lead-based paints and children’s PVC toys are potential sources of domestic lead poisoning – a review**. *Environ Pollut* (2019.0) **249** 1091-1105. DOI: 10.1016/j.envpol.2019.03.062 1151. Lanphear BP, Hornung R, Khoury J. **Low-level environmental lead exposure and children’s intellectual function: an international pooled analysis**. *Environ Health Perspect* (2005.0) **113** 894-899. DOI: 10.1289/ehp.7688 1152. Hanna-Attisha M, LaChance J, Sadler RC, Champney Schnepp A. **Elevated blood lead levels in children associated with the flint drinking water crisis: a spatial analysis of risk and public health response**. *Am J Public Health* (2016.0) **106** 283-290. DOI: 10.2105/AJPH.2015.303003 1153. Landrigan PJ, Bellinger D. **It’s time to end lead poisoning in the United States**. *JAMA Pediatr* (2021.0) **175** 1216-1217. DOI: 10.1001/jamapediatrics.2021.3525 1154. Farzan SF, Howe CG, Chen Y. **Prenatal lead exposure and elevated blood pressure in children**. *Environ Int* (2018.0) **121** 1289-1296. DOI: 10.1016/j.envint.2018.10.049 1155. Lu X, Xu X, Zhang Y, Zhang Y, Wang C, Huo X. **Elevated inflammatory Lp-PLA2 and IL-6 link e-waste Pb toxicity to cardiovascular risk factors in preschool children**. *Environ Pollut* (2018.0) **234** 601-609. DOI: 10.1016/j.envpol.2017.11.094 1156. Sanders AP, Saland JM, Wright RO, Satlin L. **Perinatal and childhood exposure to environmental chemicals and blood pressure in children: a review of literature 2007–2017**. *Pediatr Res* (2018.0) **84** 165-180. DOI: 10.1038/s41390-018-0055-3 1157. Sanders AP, Mazzella MJ, Malin AJ. **Combined exposure to lead, cadmium, mercury, and arsenic and kidney health in adolescents age 12–19 in NHANES 2009–2014**. *Environ Int* (2019.0) **131** 104993. DOI: 10.1016/j.envint.2019.104993 1158. Morrow H. *Kirk-Othmer Encyclopedia of Chemical Technology* (2010.0) 1-36. DOI: 10.1002/0471238961.0301041303011818.a01.pub3 1159. **Workplace safety and health topics: cadmium** 1160. Turner A, Wei H. **Cadmium pigments in consumer products and their health risks**. *Sci Total Environ* (2019.0) **657** 1409-1418. DOI: 10.1016/j.scitotenv.2018.12.096 1161. Tellez-Plaza M, Jones MR, Dominguez-Lucas A, Guallar E, Navas-Acien A. **Cadmium exposure and clinical cardiovascular disease: a systematic review**. *Curr Atheroscler Rep* (2013.0) **15** 356. DOI: 10.1007/s11883-013-0356-2 1162. Larsson SC, Wolk A. **Urinary cadmium and mortality from all causes, cancer and cardiovascular disease in the general population: systematic review and meta-analysis of cohort studies**. *Int J Epidemiol* (2016.0) **45** 782-791. DOI: 10.1093/ije/dyv086 1163. Antizar-Ladislao B. **Environmental levels, toxicity and human exposure to tributyltin (TBT)-contaminated marine environment. A review**. *Environ Int* (2008.0) **34** 292-308. DOI: 10.1016/j.envint.2007.09.005 1164. Sadighara P, Jahanbakhsh M, Nazari Z, Mostashari P. **The organotin contaminants in food: sources and methods for detection: a systematic review and meta-analysis**. *Food Chem X* (2021.0) **12** 100154. DOI: 10.1016/j.fochx.2021.100154 1165. Marques VB, Faria RA, Dos Santos L. **Overview of the pathophysiological implications of organotins on the endocrine system**. *Front Endocrinol* (2018.0) **9**. DOI: 10.3389/fendo.2018.00101 1166. Tinkov AA, Ajsuvakova OP, Skalnaya MG. **Organotins in obesity and associated metabolic disturbances**. *J Inorg Biochem* (2019.0) **191** 49-59. DOI: 10.1016/j.jinorgbio.2018.11.002 1167. Graceli JB. **Editorial: Organotins as a complete physiologic and endocrine disruptor: role of disease development**. *Front Endocrinol* (2019.0) **10**. DOI: 10.3389/fendo.2019.00799 1168. de Araújo JFP, Podratz PL, Merlo E. **Organotin exposure and vertebrate reproduction: a review**. *Front Endocrinol* (2018.0) **9**. DOI: 10.3389/fendo.2018.00064 1169. Rantakokko P, Main KM, Wohlfart-Veje C. **Association of placenta organotin concentrations with congenital cryptorchidism and reproductive hormone levels in 280 newborn boys from Denmark and Finland**. *Hum Reprod* (2013.0) **28** 1647-1660. DOI: 10.1093/humrep/det040 1170. Bradley EL, Castle L, Day JS, Leak J. **Migration of melamine from can coatings cross-linked with melamine-based resins, into food simulants and foods**. *Food Addit Contam Part A* (2011.0) **28** 243-250. DOI: 10.1080/19440049.2010.536167 1171. Bradley EL, Boughtflower V, Smith TL, Speck DR, Castle L. **Survey of the migration of melamine and formaldehyde from melamine food contact articles available on the UK market**. *Food Addit Contam* (2005.0) **22** 597-606. DOI: 10.1080/02652030500135243 1172. Lund KH, Petersen JH. **Migration of formaldehyde and melamine monomers from kitchen- and tableware made of melamine plastic**. *Food Addit Contam* (2006.0) **23** 948-955. DOI: 10.1080/02652030500415660 1173. Takazawa M, Suzuki S, Kannan K. **Leaching of melamine and cyanuric acid from melamine-based tableware at different temperatures and water-based simulants**. *Environ Chem Ecotoxicol* (2020.0) **2** 91-96. DOI: 10.1016/j.enceco.2020.07.002 1174. Hau AK ching, Kwan TH, Li PK tao. **Melamine Toxicity and the Kidney**. *J Am Soc Nephrol* (2009.0) **20** 245-250. DOI: 10.1681/ASN.2008101065 1175. Lam CW, Lan L, Che X. **Diagnosis and spectrum of melamine-related renal disease: plausible mechanism of stone formation in humans**. *Clin Chim Acta* (2009.0) **402** 150-155. DOI: 10.1016/j.cca.2008.12.035 1176. An L, Sun W. **A brief review of neurotoxicity induced by melamine**. *Neurotox Res* (2017.0) **32** 301-309. DOI: 10.1007/s12640-017-9731-z 1177. Bolden AL, Rochester JR, Kwiatkowski CF. **Melamine, beyond the kidney: a ubiquitous endocrine disruptor and neurotoxicant?**. *Toxicol Lett* (2017.0) **280** 181-189. DOI: 10.1016/j.toxlet.2017.07.893 1178. Yalçin S, Güneş B, Yalçin S. **Presence of melamine in human milk and the evaluation of the effects on mother–infant pairs in a cohort study**. *Hum Exp Toxicol* (2020.0) **39** 624-633. DOI: 10.1177/0960327119898748 1179. Liu CC, Wu CF, Chen BH. **Low exposure to melamine increases the risk of urolithiasis in adults**. *Kidney Int* (2011.0) **80** 746-752. DOI: 10.1038/ki.2011.154 1180. Liu CC, Hsieh TJ, Wu CF. **Urinary melamine excretion and increased markers of renal tubular injury in patients with calcium urolithiasis: a cross-sectional study**. *Environ Pollut* (2017.0) **231** 1284-1290. DOI: 10.1016/j.envpol.2017.08.091 1181. Wang YH, Wu CF, Liu CC. **A probabilistic approach for benchmark dose of melamine exposure for a marker of early renal dysfunction in patients with calcium urolithiasis**. *Ecotoxicol Environ Saf* (2020.0) **200** 110741. DOI: 10.1016/j.ecoenv.2020.110741 1182. Tsai YC, Wu CF, Liu CC. **Urinary melamine levels and progression of CKD**. *Clin J Am Soc Nephrol* (2019.0) **14** 1133-1141. DOI: 10.2215/CJN.01740219 1183. Guo J, Wu C, Zhang J. **Associations of melamine and cyanuric acid exposure with markers of kidney function in adults: results from NHANES 2003–2004**. *Environ Int* (2020.0) **141** 105815. DOI: 10.1016/j.envint.2020.105815 1184. Melough M, Foster D, Fretts AM, Sathyanarayana S. **Dietary sources of melamine exposure among US children and adults in the National Health and Nutrition Examination Survey 2003–2004**. *Nutrients* (2020.0) **12** 3844. DOI: 10.3390/nu12123844 1185. Sathyanarayana S, Flynn JT, Messito MJ. **Melamine and cyanuric acid exposure and kidney injury in US children**. *Environ Res* (2019.0) **171** 18-23. DOI: 10.1016/j.envres.2018.10.038 1186. Liu CC, Hsieh TJ, Wu CF. **Interrelationship of environmental melamine exposure, biomarkers of oxidative stress and early kidney injury**. *J Hazard Mater* (2020.0) **396** 122726. DOI: 10.1016/j.jhazmat.2020.122726 1187. Rahman A, Sarkar A, Yadav OP, Achari G, Slobodnik J. **Potential human health risks due to environmental exposure to nano- and microplastics and knowledge gaps: a scoping review**. *Sci Total Environ* (2021.0) **757** 143872-143872. DOI: 10.1016/j.scitotenv.2020.143872 1188. Gigault J, Halle A ter, Baudrimont M. **Current opinion: what is a nanoplastic?**. *Environ Pollut 1987* (2018.0) **235** 1030-1034. DOI: 10.1016/j.envpol.2018.01.024 1189. Koelmans AA, Besseling E, Shim WJ. **Chapter 12. Nanoplastics in the aquatic environment. Critical review**. *Marine Anthropogenic Litter* (2015.0) 325-340. DOI: 10.1007/978-3-319-16510-3_12 1190. Karbalaei S, Hanachi P, Walker TR, Cole M. **Occurrence, sources, human health impacts and mitigation of microplastic pollution**. *Environ Sci Pollut Res Int* (2018.0) **25** 36046-36063. DOI: 10.1007/s11356-018-3508-7 1191. El Hadri H, Gigault J, Maxit B, Grassl B, Reynaud S. **Nanoplastic from mechanically degraded primary and secondary microplastics for environmental assessments**. *NanoImpact* (2020.0) **17** 100206. DOI: 10.1016/j.impact.2019.100206 1192. Dris R, Gasperi J, Saad M, Mirande C, Tassin B. **Synthetic fibers in atmospheric fallout: a source of microplastics in the environment?**. *Mar Pollut Bull* (2016.0) **104** 290-293. DOI: 10.1016/j.marpolbul.2016.01.006 1193. Liao Z, Ji X, Ma Y. **Airborne microplastics in indoor and outdoor environments of a coastal city in Eastern China**. *J Hazard Mater* (2021.0) **417** 126007. DOI: 10.1016/j.jhazmat.2021.126007 1194. Mintenig SM, Loder MGJ, Primpke S, Gerdts G. **Low numbers of microplastics detected in drinking water from ground water sources**. *Sci Total Env* (2019.0) **648** 631-635. DOI: 10.1016/j.scitotenv.2018.08.178 1195. Lu HC, Ziajahromi S, Neale PA, Leusch FDL. **A systematic review of freshwater microplastics in water and sediments: recommendations for harmonisation to enhance future study comparisons**. *Sci Total Environ* (2021.0) **781** 146693. DOI: 10.1016/j.scitotenv.2021.146693 1196. Kumar P, Inamura Y, Bao PN, Abeynayaka A, Dasgupta R, Abeynayaka HDL. **Microplastics in freshwater environment in Asia: a systematic scientific review**. *Water* (2022.0) **14** 1737. DOI: 10.3390/w14111737 1197. Khalid Ageel H, Harrad S, Abou-Elwafa Abdallah M. **Occurrence, human exposure, and risk of microplastics in the indoor environment**. *Environ Sci Process Impacts* (2022.0) **24** 17-31. DOI: 10.1039/D1EM00301A 1198. Ramsperger AFRM, Bergamaschi E, Panizzolo M. **Nano- and microplastics: a comprehensive review on their exposure routes, translocation, and fate in humans**. *NanoImpact* (2022.0) **29** 100441. DOI: 10.1016/j.impact.2022.100441 1199. Rochman CM, Tahir A, Williams SL. **Anthropogenic debris in seafood: plastic debris and fibers from textiles in fish and bivalves sold for human consumption**. *Sci Rep* (2015.0) **5**. DOI: 10.1038/srep14340 1200. Wardrop P, Shimeta J, Nugegoda D. **Chemical pollutants sorbed to ingested microbeads from personal care products accumulate in fish**. *Environ Sci Technol* (2016.0) **50** 4037-4044. DOI: 10.1021/acs.est.5b06280 1201. Abbasi S, Moore F, Keshavarzi B. **PET-microplastics as a vector for heavy metals in a simulated plant rhizosphere zone**. *Sci Total Environ* (2020.0) **744** 140984. DOI: 10.1016/j.scitotenv.2020.140984 1202. Oliveri Conti G, Ferrante M, Banni M. **Micro- and nano-plastics in edible fruit and vegetables. The first diet risks assessment for the general population**. *Environ Res* (2020.0) **187** 109677. DOI: 10.1016/j.envres.2020.109677 1203. Luqman A, Nugrahapraja H, Wahyuono RA. **Microplastic contamination in human stools, foods, and drinking water associated with Indonesian coastal population**. *Environments* (2021.0) **8** 138. DOI: 10.3390/environments8120138 1204. Müller C. **Not as bad as it seems? A literature review on the case of microplastic uptake in fish**. *Front Mar Sci* (2021.0) **8**. DOI: 10.3389/fmars.2021.672768 1205. Blackburn K, Green D. **The potential effects of microplastics on human health: what is known and what is unknown**. *Ambio* (2022.0) **51** 518-530. DOI: 10.1007/s13280-021-01589-9 1206. Kosuth M, Mason SA, Wattenberg EV, Zhou Z. **Anthropogenic contamination of tap water, beer, and sea salt**. *PLOS ONE* (2018.0) **13**. DOI: 10.1371/journal.pone.0194970 1207. Danopoulos E, Jenner L, Twiddy M, Rotchell JM. **Microplastic contamination of salt intended for human consumption: a systematic review and meta-analysis**. *SN Appl Sci* (2020.0) **2** 1950. DOI: 10.1007/s42452-020-03749-0 1208. Wibowo AT, Nugrahapraja H, Wahyuono RA. **Microplastic contamination in the human gastrointestinal tract and daily consumables associated with an Indonesian farming community**. *Sustainability* (2021.0) **13** 12840. DOI: 10.3390/su132212840 1209. 1209World Health Organization (WHO). Microplastics in Drinking-Water. 2019; 1–101. https://apps.who.int/iris/handle/10665/326499.. *Microplastics in Drinking-Water* (2019.0) 1-101 1210. Danopoulos E, Twiddy M, Rotchell JM. **Microplastic contamination of drinking water: a systematic review**. *PLoS One* (2020.0) **15**. DOI: 10.1371/journal.pone.0236838 1211. Toussaint B, Raffael B, Angers-Loustau A. **Review of micro- and nanoplastic contamination in the food chain**. *Food Addit Contam Part A* (2019.0) **36** 639-673. DOI: 10.1080/19440049.2019.1583381 1212. Mason SA, Welch VG, Neratko J. **Synthetic polymer contamination in bottled water**. *Front Chem* (2018.0) **6**. DOI: 10.3389/fchem.2018.00407 1213. Oßmann BE, Sarau G, Holtmannspötter H, Pischetsrieder M, Christiansen SH, Dicke W. **Small-sized microplastics and pigmented particles in bottled mineral water**. *Water Res* (2018.0) **141** 307-316. DOI: 10.1016/j.watres.2018.05.027 1214. Hernandez LM, Xu EG, Larsson HCE, Tahara R, Maisuria VB, Tufenkji N. **Plastic teabags release billions of microparticles and nanoparticles into tea**. *Environ Sci Technol* (2019.0) **53** 12300-12310. DOI: 10.1021/acs.est.9b02540 1215. Zangmeister CD, Radney JG, Benkstein KD, Kalanyan B. **Common single-use consumer plastic products release trillions of sub-100 nm nanoparticles per liter into water during normal use**. *Environ Sci Technol* (2022.0) **56** 5448-5455. DOI: 10.1021/acs.est.1c06768 1216. Praveena SM, Shaifuddin SNM, Akizuki S. **Exploration of microplastics from personal care and cosmetic products and its estimated emissions to marine environment: an evidence from Malaysia**. *Mar Pollut Bull* (2018.0) **136** 135-140. DOI: 10.1016/j.marpolbul.2018.09.012 1217. Pletz M. **Ingested microplastics: do humans eat one credit card per week?**. *J Hazard Mater Lett* (2022.0) **3** 100071. DOI: 10.1016/j.hazl.2022.100071 1218. Senathirajah K, Attwood S, Bhagwat G, Carbery M, Wilson S, Palanisami T. **Estimation of the mass of microplastics ingested – a pivotal first step towards human health risk assessment**. *J Hazard Mater* (2021.0) **404** 124004. DOI: 10.1016/j.jhazmat.2020.124004 1219. Ibrahim YS, Tuan Anuar S, Azmi AA. **Detection of microplastics in human colectomy specimens**. *JGH Open* (2021.0) **5** 116-121. DOI: 10.1002/jgh3.12457 1220. Schwabl P, Koppel S, Konigshofer P. **Detection of various microplastics in human stool: a prospective case series**. *Ann Intern Med* (2019.0) **171** 453-457. DOI: 10.7326/M19-0618 1221. Pérez-Guevara F, Kutralam-Muniasamy G, Shruti VC. **Critical review on microplastics in fecal matter: research progress, analytical methods and future outlook**. *Sci Total Environ* (2021.0) **778** 146395. DOI: 10.1016/j.scitotenv.2021.146395 1222. Zhang N, Li YB, He HR, Zhang JF, Ma GS. **You are what you eat: microplastics in the feces of young men living in Beijing**. *Sci Total Environ* (2021.0) **767** 144345. DOI: 10.1016/j.scitotenv.2020.144345 1223. Powell JJ, Faria N, Thomas-McKay E, Pele LC. **Origin and fate of dietary nanoparticles and microparticles in the gastrointestinal tract**. *J Autoimmun* (2010.0) **34** J226-J233. DOI: 10.1016/j.jaut.2009.11.006 1224. Hillery A, Jani P, Florence A. **Comparative, quantitative study of lymphoid and non-lymphoid uptake of 60 nm polystyrene particles**. *J Drug Target* (1994.0) **2** 151-156. DOI: 10.3109/10611869409015904 1225. **Ambient (outdoor) air pollution** 1226. Pauly JL, Stegmeier SJ, Allaart HA. **Inhaled cellulosic and plastic fibers found in human lung tissue**. *Cancer Epidemiol Biomarkers Prev* (1998.0) **7** 419-428. PMID: 9610792 1227. Huang S, Huang X, Bi R. **Detection and analysis of microplastics in human sputum**. *Environ Sci Technol* (2022.0) **56** 2476-2486. DOI: 10.1021/acs.est.1c03859 1228. Baeza-Martínez C, Olmos S, González-Pleiter M. **First evidence of microplastics isolated in European citizens’ lower airway**. *J Hazard Mater* (2022.0) **438** 129439. DOI: 10.1016/j.jhazmat.2022.129439 1229. Chen Q, Gao J, Yu H. **An emerging role of microplastics in the etiology of lung ground glass nodules**. *Environ Sci Eur* (2022.0) **34** 25. DOI: 10.1186/s12302-022-00605-3 1230. Jenner LC, Rotchell JM, Bennett RT, Cowen M, Tentzeris V, Sadofsky LR. **Detection of microplastics in human lung tissue using µFTIR spectroscopy**. *Sci Total Environ* (2022.0) **831** 154907. DOI: 10.1016/j.scitotenv.2022.154907 1231. Harrad S, Drage DS, Sharkey M, Berresheim H. **Perfluoroalkyl substances and brominated flame retardants in landfill-related air, soil, and groundwater from Ireland**. *Sci Total Environ* (2020.0) **705** 135834. DOI: 10.1016/j.scitotenv.2019.135834 1232. Fendall LS, Sewell MA. **Contributing to marine pollution by washing your face: microplastics in facial cleansers**. *Mar Pollut Bull* (2009.0) **58** 1225-1228. DOI: 10.1016/j.marpolbul.2009.04.025 1233. Napper IE, Bakir A, Rowland SJ, Thompson RC. **Characterisation, quantity and sorptive properties of microplastics extracted from cosmetics**. *Mar Pollut Bull* (2015.0) **99** 178-185. DOI: 10.1016/j.marpolbul.2015.07.029 1234. Hernandez LM, Yousefi N, Tufenkji N. **Are there nanoplastics in your personal care products?**. *Environ Sci Technol Lett* (2017.0) **4** 280-285. DOI: 10.1021/acs.estlett.7b00187 1235. Habib RZ, Salim Abdoon MM, Al Meqbaali RM. **Analysis of microbeads in cosmetic products in the United Arab Emirates**. *Environ Pollut* (2020.0) **258** 113831. DOI: 10.1016/j.envpol.2019.113831 1236. Lei K, Qiao F, Liu Q. **Microplastics releasing from personal care and cosmetic products in China**. *Mar Pollut Bull* (2017.0) **123** 122-126. DOI: 10.1016/j.marpolbul.2017.09.016 1237. Sangkham S, Faikhaw O, Munkong N. **A review on microplastics and nanoplastics in the environment: their occurrence, exposure routes, toxic studies, and potential effects on human health**. *Mar Pollut Bull* (2022.0) **181** 113832. DOI: 10.1016/j.marpolbul.2022.113832 1238. Schneider M, Stracke F, Hansen S, Schaefer UF. **Nanoparticles and their interactions with the dermal barrier**. *Dermatoendocrinol* (2009.0) **1** 197-206. DOI: 10.4161/derm.1.4.9501 1239. Alvarez-Román R, Naik A, Kalia YN, Guy RH, Fessi H. **Skin penetration and distribution of polymeric nanoparticles**. *J Controlled Release* (2004.0) **99** 53-62. DOI: 10.1016/j.jconrel.2004.06.015 1240. Campbell CSJ, Contreras-Rojas LR, Delgado-Charro MB, Guy RH. **Objective assessment of nanoparticle disposition in mammalian skin after topical exposure**. *J Controlled Release* (2012.0) **162** 201-207. DOI: 10.1016/j.jconrel.2012.06.024 1241. Toll R, Jacobi U, Richter H, Lademann J, Schaefer H, Blume-Peytavi U. **Penetration profile of microspheres in follicular targeting of terminal hair follicles**. *J Invest Dermatol* (2004.0) **123** 168-176. DOI: 10.1111/j.0022-202X.2004.22717.x 1242. Vogt A, Combadiere B, Hadam S. **40 nm, but not 750 or 1,500 nm, nanoparticles enter epidermal CD1a+ cells after transcutaneous application on human skin**. *J Invest Dermatol* (2006.0) **126** 1316-1322. DOI: 10.1038/sj.jid.5700226 1243. Döge N, Hadam S, Volz P. **Identification of polystyrene nanoparticle penetration across intact skin barrier as rare event at sites of focal particle aggregations**. *J Biophotonics* (2018.0) **11**. DOI: 10.1002/jbio.201700169 1244. Braun T, Ehrlich L, Henrich W. **Detection of microplastic in human placenta and meconium in a clinical setting**. *Pharmaceutics* (2021.0) **13** 921. DOI: 10.3390/pharmaceutics13070921 1245. Ragusa A, Svelato A, Santacroce C. **Plasticenta: first evidence of microplastics in human placenta**. *Environ Int* (2021.0) **146** 106274. DOI: 10.1016/j.envint.2020.106274 1246. Ragusa A, Matta M, Cristiano L. **Deeply in plasticenta: presence of microplastics in the intracellular compartment of human placentas**. *Int J Environ Res Public Health* (2022.0) **19** 11593. DOI: 10.3390/ijerph191811593 1247. Zhu L, Zhu J, Zuo R, Xu Q, Qian Y, An L. **Identification of microplastics in human placenta using laser direct infrared spectroscopy**. *Sci Total Environ* (2023.0) **856** 159060. DOI: 10.1016/j.scitotenv.2022.159060 1248. Zhang J, Wang L, Trasande L, Kannan K. **Occurrence of polyethylene terephthalate and polycarbonate microplastics in infant and adult feces**. *Environ Sci Technol Lett* (2021.0) **8** 989-994. DOI: 10.1021/acs.estlett.1c00559 1249. Wick P, Malek A, Manser P. **Barrier capacity of human placenta for nanosized materials**. *Environ Health Perspect* (2010.0) **118** 432-436. DOI: 10.1289/ehp.0901200 1250. Grafmueller S, Manser P, Diener L. **Bidirectional transfer study of polystyrene nanoparticles across the placental barrier in an ex vivo human placental perfusion model**. *Environ Health Perspect* (2015.0) **123** 1280-1286. DOI: 10.1289/ehp.1409271 1251. Fournier SB, D’Errico JN, Adler DS. **Nanopolystyrene translocation and fetal deposition after acute lung exposure during late-stage pregnancy**. *Part Fibre Toxicol* (2020.0) **17** 55. DOI: 10.1186/s12989-020-00385-9 1252. Li D, Shi Y, Yang L. **Microplastic release from the degradation of polypropylene feeding bottles during infant formula preparation**. *Nat Food* (2020.0) **1** 746-754. DOI: 10.1038/s43016-020-00171-y 1253. Liu S, Guo J, Liu X. **Detection of various microplastics in placentas, meconium, infant feces, breastmilk and infant formula: a pilot prospective study**. *Sci Total Environ* (2023.0) **854** 158699. DOI: 10.1016/j.scitotenv.2022.158699 1254. Ragusa A, Notarstefano V, Svelato A. **Raman microspectroscopy detection and characterisation of microplastics in human breastmilk**. *Polymers* (2022.0) **14** 2700. DOI: 10.3390/polym14132700 1255. Barker DJP. **Developmental origins of adult health and disease**. *J Epidemiol Community Health* (2004.0) **58** 114. DOI: 10.1136/jech.58.2.114 1256. Davis GK, Wund MA, Kliman RM. *Encyclopedia of Evolutionary Biology* (2016.0) 430-440. DOI: 10.1016/B978-0-12-800049-6.00135-9 1257. Sripada K, Wierzbicka A, Abass K. **A children’s health perspective on nano- and microplastics**. *Environ Health Perspect* (2022.0) **130** 015001. DOI: 10.1289/EHP9086 1258. Campbell P, Ma S, Yeom B, McKellop H, Schmalzried TP, Amstutz HC. **Isolation of predominantly submicron-sized UHMWPE wear particles from periprosthetic tissues**. *J Biomed Mater Res* (1995.0) **29** 127-131. DOI: 10.1002/jbm.820290118 1259. Devane PA, Bourne RB, Rorabeck CH, MacDonald S, Robinson EJ. **Measurement of polyethylene wear in metal-backed acetabular cups. II. Clinical application**. *Clin Orthop* (1995.0) **319** 317-326. DOI: 10.1097/00003086-199510000-00034 1260. Veruva SY, Lanman TH, Isaza JE, Freeman TA, Kurtz SM, Steinbeck MJ. **Periprosthetic UHMWPE wear debris induces inflammation, vascularization, and innervation after total disc replacement in the lumbar spine**. *Clin Orthop Relat Res* (2017.0) **475** 1369-1381. DOI: 10.1007/s11999-016-4996-8 1261. Kaufman AM, Alabre CI, Rubash HE, Shanbhag AS. **Human macrophage response to UHMWPE, TiAlV, CoCr, and alumina particles: analysis of multiple cytokines using protein arrays**. *J Biomed Mater Res A* (2008.0) **84A** 464-474. DOI: 10.1002/jbm.a.31467 1262. Nich C, Takakubo Y, Pajarinen J. **Macrophages – key cells in the response to wear debris from joint replacements**. *J Biomed Mater Res A* (2013.0) **101** 3033-3045. DOI: 10.1002/jbm.a.34599 1263. Nich C, Goodman SB. **The role of macrophages in the biological reaction to wear debris from joint replacements**. *J Long Term Eff Med Implants* (2014.0) **24** 259-265. DOI: 10.1615/JLongTermEffMedImplants.2014010562 1264. Schmalzried TP, Jasty M, Harris WH. **Periprosthetic bone loss in total hip arthroplasty. Polyethylene wear debris and the concept of the effective joint space**. *J Bone Joint Surg Am* (1992.0) **74** 849-863. DOI: 10.2106/00004623-199274060-00006 1265. Orishimo KF, Claus AM, Sychterz CJ, Engh CA. **Relationship between polyethylene wear and osteolysis in hips with a second-generation porous-coated cementless cup after seven years of follow-up**. *JBJS* (2003.0) **85** 1095-1099. DOI: 10.2106/00004623-200306000-00018 1266. **What are the impacts of micro- and nanoplastics on the human body?** 1267. 1267Horizon Europe. Accessed November 24, 2022. http://ec.europa.eu/horizon-europe. 1268. Leslie HA, van Velzen MJM, Brandsma SH, Vethaak AD, Garcia-Vallejo JJ, Lamoree MH. **Discovery and quantification of plastic particle pollution in human blood**. *Environ Int* (2022.0) **163** 107199. DOI: 10.1016/j.envint.2022.107199 1269. Danopoulos E, Twiddy M, West R, Rotchell JM. **A rapid review and meta-regression analyses of the toxicological impacts of microplastic exposure in human cells**. *J Hazard Mater* (2022.0) **427** 127861. DOI: 10.1016/j.jhazmat.2021.127861 1270. Marcelino RC, Cardoso RM, Domingues ELBC, Gonçalves RV, Lima GDA, Novaes RD. **The emerging risk of microplastics and nanoplastics on the microstructure and function of reproductive organs in mammals: a systematic review of preclinical evidence**. *Life Sci* (2022.0) **295** 120404. DOI: 10.1016/j.lfs.2022.120404 1271. Yan S, Hu C, Wang Y. **Association of phthalate exposure with all-cause and cause-specific mortality among people with hypertension: the U.S. National Health and Nutrition Examination Survey, 2003–2014**. *Chemosphere* (2022.0) **303** 135190. DOI: 10.1016/j.chemosphere.2022.135190 1272. Hirt N, Body-Malapel M. **Immunotoxicity and intestinal effects of nano- and microplastics: a review of the literature**. *Part Fibre Toxicol* (2020.0) **17** 57. DOI: 10.1186/s12989-020-00387-7 1273. Huang W, Yin H, Yang Y, Jin L, Lu G, Dang Z. **Influence of the co-exposure of microplastics and tetrabromobisphenol A on human gut: simulation in vitro with human cell Caco-2 and gut microbiota**. *Sci Total Environ* (2021.0) **778** 146264. DOI: 10.1016/j.scitotenv.2021.146264 1274. Thubagere A, Reinhard BM. **Nanoparticle-induced apoptosis propagates through hydrogen-peroxide-mediated bystander killing: insights from a human intestinal epithelium in vitro model**. *ACS Nano* (2010.0) **4** 3611-3622. DOI: 10.1021/nn100389a 1275. 1275California State Policy Evidence Consortium (CalSPEC). Microplastics Occurrence, Health Effects, and Mitigation Policies: An Evidence Review for the California State Legislature.; 2023.. *Microplastics Occurrence, Health Effects, and Mitigation Policies: An Evidence Review for the California State Legislature* (2023.0) 1276. Li B, Ding Y, Cheng X. **Polyethylene microplastics affect the distribution of gut microbiota and inflammation development in mice**. *Chemosphere* (2020.0) **244** 125492. DOI: 10.1016/j.chemosphere.2019.125492 1277. Eschenbacher WL, Kreiss K, Lougheed MD, Pransky GS, Day B, Castellan RM. **Nylon flock–associated interstitial lung disease**. *Am J Respir Crit Care Med* (1999.0) **159** 2003-2008. DOI: 10.1164/ajrccm.159.6.9808002 1278. Atis S, Tutluoglu B, Levent E. **The respiratory effects of occupational polypropylene flock exposure**. *Eur Respir J* (2005.0) **25** 110-117. DOI: 10.1183/09031936.04.00138403 1279. Turcotte SE, Chee A, Walsh R. **Flock worker’s lung disease**. *Chest* (2013.0) **143** 1642-1648. DOI: 10.1378/chest.12-0920 1280. Gallagher LG, Li W, Ray RM. **Occupational exposures and risk of stomach and esophageal cancers: update of a cohort of female textile workers in Shanghai, China: risk of cancer in textile workers**. *Am J Ind Med* (2015.0) **58** 267-275. DOI: 10.1002/ajim.22412 1281. Paget V, Dekali S, Kortulewski T. **Specific uptake and genotoxicity induced by polystyrene nanobeads with distinct surface chemistry on human lung epithelial cells and macrophages**. *PLoS ONE* (2015.0) **10**. DOI: 10.1371/journal.pone.0123297 1282. Ruenraroengsak P, Tetley TD. **Differential bioreactivity of neutral, cationic and anionic polystyrene nanoparticles with cells from the human alveolar compartment: robust response of alveolar type 1 epithelial cells**. *Part Fibre Toxicol* (2015.0) **12** 19. DOI: 10.1186/s12989-015-0091-7 1283. Xu M, Halimu G, Zhang Q. **Internalization and toxicity: a preliminary study of effects of nanoplastic particles on human lung epithelial cell**. *Sci Total Environ* (2019.0) **694** 133794. DOI: 10.1016/j.scitotenv.2019.133794 1284. Dong CD, Chen CW, Chen YC, Chen HH, Lee JS, Lin CH. **Polystyrene microplastic particles: in vitro pulmonary toxicity assessment**. *J Hazard Mater* (2020.0) **385** 121575. DOI: 10.1016/j.jhazmat.2019.121575 1285. Yang S, Cheng Y, Chen Z. **In vitro evaluation of nanoplastics using human lung epithelial cells, microarray analysis and co-culture model**. *Ecotoxicol Environ Saf* (2021.0) **226** 112837. DOI: 10.1016/j.ecoenv.2021.112837 1286. Goodman KE, Hare JT, Khamis ZI, Hua T, Sang QXA. **Exposure of human lung cells to polystyrene microplastics significantly retards cell proliferation and triggers morphological changes**. *Chem Res Toxicol* (2021.0) **34** 1069-1081. DOI: 10.1021/acs.chemrestox.0c00486 1287. Ruenraroengsak P, Novak P, Berhanu D. **Respiratory epithelial cytotoxicity and membrane damage (holes) caused by amine-modified nanoparticles**. *Nanotoxicology* (2012.0) **6** 94-108. DOI: 10.3109/17435390.2011.558643 1288. Brown DM, Wilson MR, MacNee W, Stone V, Donaldson K. **Size-dependent proinflammatory effects of ultrafine polystyrene particles: a role for surface area and oxidative stress in the enhanced activity of ultrafines**. *Toxicol Appl Pharmacol* (2001.0) **175** 191-199. DOI: 10.1006/taap.2001.9240 1289. Lim D, Jeong J, Song KS, Sung JH, Oh SM, Choi J. **Inhalation toxicity of polystyrene micro(nano)plastics using modified OECD TG 412**. *Chemosphere* (2021.0) **262** 128330. DOI: 10.1016/j.chemosphere.2020.128330 1290. Lu K, Lai KP, Stoeger T. **Detrimental effects of microplastic exposure on normal and asthmatic pulmonary physiology**. *J Hazard Mater* (2021.0) **416** 126069. DOI: 10.1016/j.jhazmat.2021.126069 1291. Gopinath PM, Twayana KS, Ravanan P. **Prospects on the nano-plastic particles internalization and induction of cellular response in human keratinocytes**. *Part Fibre Toxicol* (2021.0) **18** 35. DOI: 10.1186/s12989-021-00428-9 1292. Luo T, Wang C, Pan Z, Jin C, Fu Z, Jin Y. **Maternal polystyrene microplastic exposure during gestation and lactation altered metabolic homeostasis in the dams and their F1 and F2 offspring**. *Environ Sci Technol* (2019.0) **53** 10978-10992. DOI: 10.1021/acs.est.9b03191 1293. Luo T, Zhang Y, Wang C. **Maternal exposure to different sizes of polystyrene microplastics during gestation causes metabolic disorders in their offspring**. *Environ Pollut* (2019.0) **255** 113122. DOI: 10.1016/j.envpol.2019.113122 1294. Ribeiro F, O’Brien JW, Galloway T, Thomas KV. **Accumulation and fate of nano- and micro-plastics and associated contaminants in organisms**. *TrAC Trends Anal Chem* (2019.0) **111** 139-147. DOI: 10.1016/j.trac.2018.12.010 1295. Prüst M, Meijer J, Westerink RHS. **The plastic brain: neurotoxicity of micro- and nanoplastics**. *Part Fibre Toxicol* (2020.0) **17** 24. DOI: 10.1186/s12989-020-00358-y 1296. Horvatits T, Tamminga M, Liu B. **Microplastics detected in cirrhotic liver tissue**. *eBioMedicine* (2022.0) **82** 104147. DOI: 10.1016/j.ebiom.2022.104147 1297. Gautam R, Jo J, Acharya M. **Evaluation of potential toxicity of polyethylene microplastics on human derived cell lines**. *Sci Total Environ* (2022.0) **838** 156089. DOI: 10.1016/j.scitotenv.2022.156089 1298. Hwang J, Choi D, Han S, Jung SY, Choi J, Hong J. **Potential toxicity of polystyrene microplastic particles**. *Sci Rep* (2020.0) **10** 7391. DOI: 10.1038/s41598-020-64464-9 1299. Kwon W, Kim D, Kim HY. **Microglial phagocytosis of polystyrene microplastics results in immune alteration and apoptosis in vitro and in vivo**. *Sci Total Environ* (2022.0) **807** 150817. DOI: 10.1016/j.scitotenv.2021.150817 1300. Wang YL, Lee YH, Hsu YH. **The kidney-related effects of polystyrene microplastics on human kidney proximal tubular epithelial cells HK-2 and male C57BL/6 mice**. *Environ Health Perspect* (2021.0) **129** 057003. DOI: 10.1289/EHP7612 1301. Çobanoğlu H, Belivermiş M, Sıkdokur E, Kılıç Ö, Çayır A. **Genotoxic and cytotoxic effects of polyethylene microplastics on human peripheral blood lymphocytes**. *Chemosphere* (2021.0) **272** 129805. DOI: 10.1016/j.chemosphere.2021.129805 1302. Yong C, Valiyaveettil S, Tang B. **Toxicity of microplastics and nanoplastics in mammalian systems**. *Int J Environ Res Public Health* (2020.0) **17** 1509. DOI: 10.3390/ijerph17051509 1303. Bucci K, Rochman CM. **Microplastics: a multidimensional contaminant requires a multidimensional framework for assessing risk**. *Microplastics Nanoplastics* (2022.0) **2**. DOI: 10.1186/s43591-022-00028-0 1304. Yee MSL, Hii LW, Looi CK. **Impact of microplastics and nanoplastics on human health**. *Nanomaterials* (2021.0) **11** 496. DOI: 10.3390/nano11020496 1305. Cho YM, Choi KH. **The current status of studies of human exposure assessment of microplastics and their health effects: a rapid systematic review**. *Environ Anal Health Toxicol* (2021.0) **36**. DOI: 10.5620/eaht.2021004 1306. Zimmermann L, Göttlich S, Oehlmann J, Wagner M, Völker C. **What are the drivers of microplastic toxicity? Comparing the toxicity of plastic chemicals and particles to Daphnia magna**. *Environ Pollut* (2020.0) **267** 115392. DOI: 10.1016/j.envpol.2020.115392 1307. Hwang J, Choi D, Han S, Choi J, Hong J. **An assessment of the toxicity of polypropylene microplastics in human derived cells**. *Sci Total Environ* (2019.0) **684** 657-669. DOI: 10.1016/j.scitotenv.2019.05.071 1308. An D, Na J, Song J, Jung J. **Size-dependent chronic toxicity of fragmented polyethylene microplastics to Daphnia magna**. *Chemosphere* (2021.0) **271** 129591. DOI: 10.1016/j.chemosphere.2021.129591 1309. Schwarzer M, Brehm J, Vollmer M. **Shape, size, and polymer dependent effects of microplastics on Daphnia magna**. *J Hazard Mater* (2022.0) **426** 128136. DOI: 10.1016/j.jhazmat.2021.128136 1310. Choi D, Hwang J, Bang J. **In vitro toxicity from a physical perspective of polyethylene microplastics based on statistical curvature change analysis**. *Sci Total Environ* (2021.0) **752** 142242. DOI: 10.1016/j.scitotenv.2020.142242 1311. Kelpsiene E, Torstensson O, Ekvall MT, Hansson LA, Cedervall T. **Long-term exposure to nanoplastics reduces life-time in Daphnia magna**. *Sci Rep* (2020.0) **10** 1-7. DOI: 10.1038/s41598-020-63028-1 1312. Brehm J, Wilde MV, Reiche L. **In-depth characterization revealed polymer type and chemical content specific effects of microplastic on Dreissena bugensis**. *J Hazard Mater* (2022.0) **437** 129351. DOI: 10.1016/j.jhazmat.2022.129351 1313. Zimmermann L, Dierkes G, Ternes TA, Völker C, Wagner M. **Benchmarking the in vitro toxicity and chemical composition of plastic consumer products**. *Environ Sci Technol* (2019.0) **53** 11467-11477. DOI: 10.1021/acs.est.9b02293 1314. Okoye CO, Addey CI, Oderinde O. **Toxic chemicals and persistent organic pollutants associated with micro- and nanoplastics pollution**. *Chem Eng J Adv* (2022.0) **11** 100310. DOI: 10.1016/j.ceja.2022.100310 1315. Rios LM, Moore C, Jones PR. **Persistent organic pollutants carried by synthetic polymers in the ocean environment**. *Mar Pollut Bull* (2007.0) **54** 1230-1237. DOI: 10.1016/j.marpolbul.2007.03.022 1316. Ogata Y, Takada H, Mizukawa K. **International pellet watch: global monitoring of persistent organic pollutants (POPs) in coastal waters. 1. Initial phase data on PCBs, DDTs, and HCHs**. *Mar Pollut Bull* (2009.0) **58** 1437-1446. DOI: 10.1016/j.marpolbul.2009.06.014 1317. Llorca M, Farré M, Karapanagioti HK, Barceló D. **Levels and fate of perfluoroalkyl substances in beached plastic pellets and sediments collected from Greece**. *Mar Pollut Bull* (2014.0) **87** 286-291. DOI: 10.1016/j.marpolbul.2014.07.036 1318. Brennecke D, Duarte B, Paiva F, Caçador I, Canning-Clode J. **Microplastics as vector for heavy metal contamination from the marine environment**. *Estuar Coast Shelf Sci* (2016.0) **178** 189-195. DOI: 10.1016/j.ecss.2015.12.003 1319. Khalid N, Aqeel M, Noman A, Khan SM, Akhter N. **Interactions and effects of microplastics with heavy metals in aquatic and terrestrial environments**. *Environ Pollut* (2021.0) **290** 118104. DOI: 10.1016/j.envpol.2021.118104 1320. Llorca M, Schirinzi G, Martínez M, Barceló D, Farré M. **Adsorption of perfluoroalkyl substances on microplastics under environmental conditions**. *Environ Pollut* (2018.0) **235** 680-691. DOI: 10.1016/j.envpol.2017.12.075 1321. Maity S, Biswas C, Banerjee S. **Interaction of plastic particles with heavy metals and the resulting toxicological impacts: a review**. *Environ Sci Pollut Res* (2021.0) **28** 60291-60307. DOI: 10.1007/s11356-021-16448-z 1322. Frère L, Maignien L, Chalopin M. **Microplastic bacterial communities in the Bay of Brest: Influence of polymer type and size**. *Environ Pollut* (2018.0) **242** 614-625. DOI: 10.1016/j.envpol.2018.07.023 1323. Miao L, Wang P, Hou J. **Distinct community structure and microbial functions of biofilms colonizing microplastics**. *Sci Total Environ* (2019.0) **650** 2395-2402. DOI: 10.1016/j.scitotenv.2018.09.378 1324. Tavelli R, Callens M, Grootaert C, Abdallah MF, Rajkovic A. **Foodborne pathogens in the plastisphere: can microplastics in the food chain threaten microbial food safety?**. *Trends Food Sci Technol* (2022.0) **129** 1-10. DOI: 10.1016/j.tifs.2022.08.021 1325. Kirstein IV, Kirmizi S, Wichels A. **Dangerous hitchhikers? Evidence for potentially pathogenic Vibrio spp. on microplastic particles**. *Mar Environ Res* (2016.0) **120** 1-8. DOI: 10.1016/j.marenvres.2016.07.004 1326. Wu X, Pan J, Li M, Li Y, Bartlam M, Wang Y. **Selective enrichment of bacterial pathogens by microplastic biofilm**. *Water Res* (2019.0) **165** 114979. DOI: 10.1016/j.watres.2019.114979 1327. Rasool FN, Saavedra MA, Pamba S. **Isolation and characterization of human pathogenic multidrug resistant bacteria associated with plastic litter collected in Zanzibar**. *J Hazard Mater* (2021.0) **405** 124591. DOI: 10.1016/j.jhazmat.2020.124591 1328. Shi J, Wu D, Su Y, Xie B. **Selective enrichment of antibiotic resistance genes and pathogens on polystyrene microplastics in landfill leachate**. *Sci Total Environ* (2021.0) **765** 142775. DOI: 10.1016/j.scitotenv.2020.142775 1329. Baysal A, Saygin H. **Co-occurence of antibiotics and micro(nano)plastics: a systematic review between 2016–2021**. *J Environ Sci Health Part A* (2022.0) **57** 519-539. DOI: 10.1080/10934529.2022.2082222 1330. Rodrigues A, Oliver DM, McCarron A, Quilliam RS. **Colonisation of plastic pellets (nurdles) by E. coli at public bathing beaches**. *Mar Pollut Bull* (2019.0) **139** 376-380. DOI: 10.1016/j.marpolbul.2019.01.011 1331. Keswani A, Oliver DM, Gutierrez T, Quilliam RS. **Microbial hitchhikers on marine plastic debris: human exposure risks at bathing waters and beach environments**. *Mar Environ Res* (2016.0) **118** 10-19. DOI: 10.1016/j.marenvres.2016.04.006 1332. Loiseau C, Sorci G. **Can microplastics facilitate the emergence of infectious diseases?**. *Sci Total Environ* (2022.0) **823** 153694. DOI: 10.1016/j.scitotenv.2022.153694 1333. **Our planet is choking on plastic** 1334. Chen DMC, Bodirsky BL, Krueger T, Mishra A, Popp A. **The world’s growing municipal solid waste: trends and impacts**. *Environ Res Lett* (2020.0) **15** 074021. DOI: 10.1088/1748-9326/ab8659 1335. Ma J, Addink R, Yun S, Cheng J, Wang W, Kannan K. **Polybrominated dibenzo- p -dioxins/dibenzofurans and polybrominated diphenyl ethers in soil, vegetation, workshop-floor dust, and electronic shredder residue from an electronic waste recycling facility and in soils from a chemical industrial complex in eastern China**. *Environ Sci Technol* (2009.0) **43** 7350-7356. DOI: 10.1021/es901713u 1336. Chan JKY, Wong MH. **A review of environmental fate, body burdens, and human health risk assessment of PCDD/Fs at two typical electronic waste recycling sites in China**. *Sci Total Environ* (2013.0) **463–464** 1111-1123. DOI: 10.1016/j.scitotenv.2012.07.098 1337. Lin S, Ali MU, Zheng C, Cai Z, Wong MH. **Toxic chemicals from uncontrolled e-waste recycling: exposure, body burden, health impact**. *J Hazard Mater* (2022.0) **426** 127792. DOI: 10.1016/j.jhazmat.2021.127792 1338. Han GG, Ding G, Lou X. **Correlations of PCBs, DIOXIN, and PBDE with TSH in children’s blood in areas of computer e-waste recycling**. *Biomed Environ Sci* (2011.0) **24** 112-116. DOI: 10.3967/0895-3988.2011.02.004 1339. Yorifuji T, Noguchi M, Tsuda T. **Does open-air exposure to volatile organic compounds near a plastic recycling factory cause health effects?**. *J Occup Health* (2012.0) **54** 79-87. DOI: 10.1539/joh.11-0202-OA 1340. Sarigiannis DA. **Assessing the impact of hazardous waste on children’s health: the exposome paradigm**. *Environ Res* (2017.0) **158** 531-541. DOI: 10.1016/j.envres.2017.06.031 1341. Forti V, Baldé CP, Kuehr R, Bel G. *The Global E-Waste Monitor 2020: Quantities, Flows and the Circular Economy Potential* (2020.0) 1-120 1342. Srigboh RK, Basu N, Stephens J. **Multiple elemental exposures amongst workers at the Agbogbloshie electronic waste (e-waste) site in Ghana**. *Chemosphere* (2016.0) **164** 68-74. DOI: 10.1016/j.chemosphere.2016.08.089 1343. Han W, Gao G, Geng J, Li Y, Wang Y. **Ecological and health risks assessment and spatial distribution of residual heavy metals in the soil of an e-waste circular economy park in Tianjin, China**. *Chemosphere* (2018.0) **197** 325-335. DOI: 10.1016/j.chemosphere.2018.01.043 1344. 1344World Health Organization (WHO). Children and Digital Dumpsites: E-Waste Exposure and Child Health. World Health Organization; 2021. Accessed October 28, 2022. https://apps.who.int/iris/handle/10665/341718.. *Children and Digital Dumpsites: E-Waste Exposure and Child Health* (2021.0) 1345. Rocque RJ, Beaudoin C, Ndjaboue R. **Health effects of climate change: an overview of systematic reviews**. *BMJ Open* (2021.0) **11**. DOI: 10.1136/bmjopen-2020-046333 1346. Watts N, Amann M, Arnell N. **The 2020 report of The Lancet Countdown on health and climate change: responding to converging crises**. *The Lancet* (2021.0) **397** 129-170. DOI: 10.1016/S0140-6736(20)32290-X 1347. Berrang-Ford L, Sietsma AJ, Callaghan M. **Systematic mapping of global research on climate and health: a machine learning review**. *Lancet Planet Health* (2021.0) **5** e514-e525. DOI: 10.1016/S2542-5196(21)00179-0 1348. Cianconi P, Betrò S, Janiri L. **The impact of climate change on mental health: a systematic descriptive review**. *Front Psychiatry* (2020.0) **11**. DOI: 10.3389/fpsyt.2020.00074 1349. Louis S, Carlson AK, Suresh A. **Impacts of climate change and air pollution on neurologic health, disease, and practice: a scoping review**. *Neurology*. DOI: 10.1212/WNL.0000000000201630 1350. 1350HEAL Network, CRE-STRIDE. Climate Change and Aboriginal and Torres Strait Islander Health. Lowitja Institute. 2021; 1–72. DOI: 10.48455/bthg-aj15. *Climate Change and Aboriginal and Torres Strait Islander Health* (2021.0) 1-72. DOI: 10.48455/bthg-aj15 1351. **Climate Change: United Nations For Indigenous Peoples** 1352. Chowdhury GW, Koldewey HJ, Niloy MNH, Sarker S. **The ecological impact of plastic pollution in a changing climate**. *Emerg Top Life Sci* (2022.0) **6** 389-402. DOI: 10.1042/ETLS20220016 1353. Jones MW, Smith A, Betts R, Canadell JG, Prentice CI, Le Quere C. *Climate Change Increases the Risk of Wildfires* (2020.0) 1-2 1354. Xu R, Yu P, Abramson MJ. **Wildfires, global climate change, and human health**. *N Engl J Med* (2020.0) **383** 2173-2181. DOI: 10.1056/NEJMsr2028985 1355. Verma R, Vinoda KS, Papireddy M, Gowda ANS. **Toxic pollutants from plastic waste – a review**. *Procedia Environ Sci* (2016.0) **35** 701-708. DOI: 10.1016/j.proenv.2016.07.069 1356. Rani M, Shim WJ, Jang M, Han GM, Hong SH. **Releasing of hexabromocyclododecanes from expanded polystyrenes in seawater – field and laboratory experiments**. *Chemosphere* (2017.0) **185** 798-805. DOI: 10.1016/j.chemosphere.2017.07.042 1357. Masry M, Rossignol S, Gardette JL, Therias S, Bussière PO, Wong-Wah-Chung P. **Characteristics, fate, and impact of marine plastic debris exposed to sunlight: a review**. *Mar Pollut Bull* (2021.0) **171** 112701. DOI: 10.1016/j.marpolbul.2021.112701 1358. Balbela C, Soroldoni S, Kessler F, Fernandes A, Pinho GLL. **Could a future ocean acidification scenario influence the photodegradation of microplastics?**. DOI: 10.2139/ssrn.4199146 1359. Born MP, Brüll C. **From model to nature – a review on the transferability of marine (micro-) plastic fragmentation studies**. *Sci Total Environ* (2022.0) **811** 151389. DOI: 10.1016/j.scitotenv.2021.151389 1360. 1360Technical University of Mombasa and Stanford University. Mitigating Plastic Litter Will Significantly Reduce Mosquito-Borne Disease. Technical University of Mombasa and Standford University. 2021; 1–2. Accessed December 7, 2022. https://globalhealth.stanford.edu/wp-content/uploads/2021/06/SGH-Brief-v9-1.pdf.. *Mitigating Plastic Litter Will Significantly Reduce Mosquito-Borne Disease* (2021.0) 1-2 1361. Harrison JP, Schratzberger M, Sapp M, Osborn AM. **Rapid bacterial colonization of low-density polyethylene microplastics in coastal sediment microcosms**. *BMC Microbiol* (2014.0) **14** 232. DOI: 10.1186/s12866-014-0232-4 1362. Laverty AL, Primpke S, Lorenz C, Gerdts G, Dobbs FC. **Bacterial biofilms colonizing plastics in estuarine waters, with an emphasis on Vibrio spp. and their antibacterial resistance**. *PLOS ONE* (2020.0) **15**. DOI: 10.1371/journal.pone.0237704 1363. Rech S, Thiel M, Borrell Pichs YJ, García-Vazquez E. **Travelling light: fouling biota on macroplastics arriving on beaches of remote Rapa Nui (Easter Island) in the South Pacific Subtropical Gyre**. *Mar Pollut Bull* (2018.0) **137** 119-128. DOI: 10.1016/j.marpolbul.2018.10.015 1364. Billaud M, Seneca F, Tambutté E, Czerucka D. **An increase of seawater temperature upregulates the expression of Vibrio parahaemolyticus virulence factors implicated in adhesion and biofilm formation**. *Front Microbiol* (2022.0) **13**. DOI: 10.3389/fmicb.2022.840628 1365. Fantke P, Jolliet O, Apte JS. **Characterizing aggregated exposure to primary particulate matter: recommended intake fractions for indoor and outdoor sources**. *Env Sci Technol* (2017.0) **51** 9089-9100. DOI: 10.1021/acs.est.7b02589 1366. Fantke P, McKone TE, Tainio M. **Global effect factors for exposure to fine particulate matter**. *Env Sci Technol* (2019.0) **53** 6855-6868. DOI: 10.1021/acs.est.9b01800 1367. Frischknecht R, Jolliet O. **Global guidance for life cycle impact assessment indicators**. *Publ UNEPSETAC Life Cycle Initiat Paris DTI2081PA ISBN* 978-992 1368. Cabernard L, Pfister S. **A highly resolved MRIO database for analyzing environmental footprints and green economy progress**. *Sci Total Env* (2021.0) **755** 142587. DOI: 10.1016/j.scitotenv.2020.142587 1369. **Global and regional burden of disease and injury in 2016 arising from occupational exposures: a systematic analysis for the Global Burden of Disease Study 2016**. *Occup Env Med* (2020.0) **77** 133-141. DOI: 10.1136/oemed-2019-106008 1370. **World Development Indicators** 1371. Haines A, Patz JA. **Health effects of climate change**. *JAMA* (2004.0) **291** 99-103. DOI: 10.1001/jama.291.1.99 1372. Tang L, Ii R, Tokimatsu K, Itsubo N. **Development of human health damage factors related to CO2 emissions by considering future socioeconomic scenarios**. *Int J Life Cycle Assess* (2018.0) **23** 2288-2299. DOI: 10.1007/s11367-015-0965-9 1373. Gasparrini A, Guo Y, Sera F. **Projections of temperature-related excess mortality under climate change scenarios**. *Lancet Planet Health* (2017.0) **1** e360-e367. DOI: 10.1016/S2542-5196(17)30156-0 1374. Carleton T, Jina A, Delgado M. **Valuing the global mortality consequences of climate change accounting for adaptation costs and benefits**. *Q J Econ* (2022.0) **137** 2037-2105. DOI: 10.1093/qje/qjac020 1375. Cromar KR, Anenberg SC, Balmes JR. **Global health impacts for economic models of climate change: a systematic review and meta-analysis**. *Ann Am Thorac Soc* (2022.0) **19** 1203-1212. DOI: 10.1513/AnnalsATS.202110-1193OC 1376. **Supplementary material for the regulatory impact analysis for the supplemental proposed rulemaking, “Standards of Performance for New, Reconstructed, and Modified Sources and Emissions Guidelines for Existing Sources: Oil and Natural Gas Sector Climate Review.”** 1377. Rennert K, Errickson F, Prest BC. **Comprehensive evidence implies a higher social cost of CO2**. *Nature* (2022.0) **610** 687-692. DOI: 10.1038/s41586-022-05224-9 1378. **Data-driven spatial climate impact model user manual, version 092022- EPA** 1379. Howard P, Schwartz J. **Think global: international reciprocity as justification for a global social cost of carbon**. *Colum J Envtl L* (2016.0) **42** 203. DOI: 10.2139/ssrn.2822513 1380. Landrigan PJ, Fuller R, Acosta NJR. **The Lancet Commission on pollution and health**. *The Lancet* (2018.0) **391** 462-512. DOI: 10.1016/S0140-6736(17)32345-0 1381. 1381World Bank. The Global Health Cost of PM2.5 Air Pollution: A Case for Action Beyond 2021. World Bank Publications; 2022. https://play.google.com/store/books/details?id=w15jEAAAQBAJ.. *The Global Health Cost of PM2.5 Air Pollution: A Case for Action Beyond 2021* (2022.0) 1382. Robinson LA, Hammitt JK, Jamison DT, Walker DG. **Conducting benefit-cost analysis in low- and middle-income countries: introduction to the special issue**. *J Benefit Cost Anal* (2019.0) **10** 1-14. DOI: 10.1017/bca.2019.4 1383. Grosse SD, Krueger KV, Mvundura M. **Economic productivity by age and sex: 2007 estimates for the United States**. *Med Care* (2009.0) **47** S94-103. DOI: 10.1097/MLR.0b013e31819c9571 1384. Attina TM, Hauser R, Sathyanarayana S. **Exposure to endocrine-disrupting chemicals in the USA: a population-based disease burden and cost analysis**. *Lancet Diabetes Endocrinol* (2016.0) **4** 996-1003. DOI: 10.1016/S2213-8587(16)30275-3 1385. Bellanger M, Demeneix B, Grandjean P, Zoeller RT, Trasande L. **Neurobehavioral deficits, diseases, and associated costs of exposure to endocrine-disrupting chemicals in the European Union**. *J Clin Endocrinol Metab* (2015.0) **100** 1256-1266. DOI: 10.1210/jc.2014-4323 1386. Malits J, Naidu M, Trasande L. **Exposure to endocrine disrupting chemicals in Canada: population-based estimates of disease burden and economic costs**. *Toxics* (2022.0) **10**. DOI: 10.3390/toxics10030146 1387. Rice GE, Hammitt JK, Evans JS. **A probabilistic characterization of the health benefits of reducing methyl mercury intake in the United States**. *Env Sci Technol* (2010.0) **44** 5216-5224. DOI: 10.1021/es903359u 1388. Grosse SD, Zhou Y. **Monetary valuation of children’s cognitive outcomes in economic evaluations from a societal perspective: a review**. *Children* (2021.0) **8**. DOI: 10.3390/children8050352 1389. Honeycutt A, Dunlap L, Chen H, Homsi G. **Economic costs associated with mental retardation, cerebral palsy, hearing loss, and vision impairment-United States, 2003**. *MMWR Morb Mortal Wkly Rep* (2004.0) **53** 57-57. PMID: 14749614 1390. Melzer D, Rice NE, Lewis C, Henley WE, Galloway TS. **Association of urinary bisphenol A concentration with heart disease: evidence from NHANES 2003/06**. DOI: 10.1371/journal.pone.0008673 1391. Melzer D, Osborne NJ, Henley WE. **Urinary bisphenol A concentration and risk of future coronary artery disease in apparently healthy men and women**. *Circulation* (2012.0) **125** 1482-1490. DOI: 10.1161/CIRCULATIONAHA.111.069153 1392. Moon S, Yu SH, Lee CB, Park YJ, Yoo HJ, Kim DS. **Effects of bisphenol A on cardiovascular disease: an epidemiological study using National Health and Nutrition Examination Survey 2003–2016 and meta-analysis**. *Sci Total Env* (2021.0) **763** 142941. DOI: 10.1016/j.scitotenv.2020.142941 1393. **Heart disease facts** 1394. **Stroke facts** 1395. Tsao CW, Aday AW, Almarzooq ZI. **Heart disease and stroke statistics—2022 update: a report from the American Heart Association**. *Circulation* (2022.0) **145** e153-e639. DOI: 10.1161/CIR.0000000000001052 1396. Zeng G, Zhang Q, Wang X, Wu KH. **Low-level plasticizer exposure and all-cause and cardiovascular disease mortality in the general population**. *Env Health* (2022.0) **21** 32. DOI: 10.1186/s12940-022-00841-3 1397. **Historical GHG Emissions** 1398. Merkl A, Charles D. *The Price of Plastic Pollution: Social Costs and Corporate Liabilities* **2022** 48 1399. Rawls J. *A Theory of Justice* (1999.0) 1400. 1400U.S. Environmental Protection Agency (EPA). EPA Annual Environmental Justice Progress Report FY 2020; 2020: 52. Accessed December 23, 2022. https://www.epa.gov/sites/default/files/2021-01/documents/2020_ej_report-final-web-v4.pdf.. *EPA Annual Environmental Justice Progress Report FY 2020* **2020** 52 1401. Bryant B. **Pollution prevention and participatory research as a methodology for environmental justice**. *Va Environ Law J* (1995.0) **14** 589-613 1402. 1402United Nations Environment Programme (UNEP). Neglected: Environmental Justice Impacts of Marine Litter and Plastic Pollution. UN Environment Programme. 2021; 1–66. Accessed August 26, 2022. http://www.unep.org/resources/report/neglected-environmental-justice-impacts-marine-litter-and-plastic-pollution.. *Neglected: Environmental Justice Impacts of Marine Litter and Plastic Pollution* (2021.0) 1-66 1403. Solomonian L, Di Ruggiero E. **The critical intersection of environmental and social justice: a commentary**. *Glob Health* (2021.0) **17** 30. DOI: 10.1186/s12992-021-00686-4 1404. Strang V. *Conservation: Integrating Social and Ecological Justice* (2019.0) 105-119. DOI: 10.1007/978-3-030-13905-6_8 1405. Bennett NJ. **Mainstreaming equity and justice in the ocean**. *Front Mar Sci* (2022.0) **9** 873572. DOI: 10.3389/fmars.2022.873572 1406. Bennett NJ, Alava JJ, Ferguson CE. *Environmental Justice in the Ocean* (2022.0) **39** 1407. 1407United Nations Environment Programme (UNEP). Historic day in the campaign to beat plastic pollution: nations commit to develop a legally binding agreement. UN Environment Programme. Published March 2, 2022. Accessed December 23, 2022. http://www.unep.org/news-and-stories/press-release/historic-day-campaign-beat-plastic-pollution-nations-commit-develop. 1408. Keener VW, Marra JJ, Finucane ML, Spooner D, Smith MH. *Climate Change and Pacific Islands: Indicators and Impacts: Report for the 2012 Pacific Islands Regional Climate Assessment (PIRCA)* (2012.0) 1409. 1409Hawai’i Climate Change Mitigation and Adaptation Commission. Sea Level Rise Vulnerability and Adaptation Report. Tetra Tech, Inc. and the State of Hawai’i Department of Land and Natural Resources, Office of Conservation and Coastal Lands. 2017; 302. Accessed December 27, 2022. https://climateadaptation.hawaii.gov/wp-content/uploads/2017/12/SLR-Report_Dec2017.pdf.. *Sea Level Rise Vulnerability and Adaptation Report* (2017.0) **302** 1410. Storlazzi CD, Gingerich SB, van Dongeren A. **Most atolls will be uninhabitable by the mid-21st century because of sea-level rise exacerbating wave-driven flooding**. *Sci Adv* (2018.0) **4**. DOI: 10.1126/sciadv.aap9741 1411. Farrelly T, Borrelle S, Fuller S. *Plastic Pollution Prevention in Pacific Island Countries: Gap Analysis of Current Legislation, Policies and Plans* (2020.0) **33** 1412. 1412Hawai‘i State Coastal Zone Management Program. Assessing the Feasibility and Implications of Managed Retreat Strategies for Vulnerable Coastal Areas in Hawai’i: Final Report. Hawai‘i State Office of Planning. 2019; 84. Accessed December 23, 2022. https://files.hawaii.gov/dbedt/op/czm/ormp/assessing_the_feasibility_and_implications_of_managed_retreat_strategies_for_vulnerable_coastal_areas_in_hawaii.pdf.. *Assessing the Feasibility and Implications of Managed Retreat Strategies for Vulnerable Coastal Areas in Hawai’i: Final Report* (2019.0) **84** 1413. Carey J. **Managed retreat increasingly seen as necessary in response to climate change’s fury**. *Proc Natl Acad Sci* (2020.0) **117** 13182-13185. DOI: 10.1073/pnas.2008198117 1414. Goering L, Clarke E. **Analysis: as seas rise, coastal communities face hard choices over “managed retreat.”**. *Reuters* 1415. 1415United Nation Conference on Trade and Development (UNCTAD). Smallest footprints, largest impacts: least-developed countries need a just sustainable transition. United Nations Conference on Trade and Development. Published October 1, 2021. Accessed August 26, 2022. https://unctad.org/topic/least-developed-countries/chart-october-2021. 1416. Betzold C. **“Borrowing” power to influence international negotiations: AOSIS in the climate change regime, 1990–1997**. *Sage J* (2010.0) **30**. DOI: 10.1111/j.1467-9256.2010.01377.x 1417. Thomas A, Baptiste A, Martyr-Koller R, Pringle P, Rhiney K. **Climate change and small island developing states**. *Annu Rev Environ Resour* (2020.0) **45** 1-27. DOI: 10.1146/annurev-environ-012320-083355 1418. Mead L 1419. Lachmann F, Almroth BC, Baumann H. *Marine Plastic Litter on Small Island Developing States (SIDS): Impacts and Measures* **2017** 77. DOI: 10.13140/RG.2.2.16168.37121 1420. Filho WL, Havea PH, Balogun AL. **Plastic debris on Pacific Islands: ecological and health implications**. *Sci Total Environ* (2019.0) **670** 181-187. DOI: 10.1016/j.scitotenv.2019.03.181 1421. Mohee R, Mauthoor S, Bundhoo ZMA, Somaroo G, Soobhany N, Gunasee S. **Current status of solid waste management in small island developing states: a review**. *Waste Manag* (2015.0) **43** 539-549. DOI: 10.1016/j.wasman.2015.06.012 1422. Fuldauer LI, Ives MC, Adshead D, Thacker S, Hall JW. **Participatory planning of the future of waste management in small island developing states to deliver on the Sustainable Development Goals**. *J Clean Prod* (2019.0) **223** 147-162. DOI: 10.1016/j.jclepro.2019.02.269 1423. Burt AJ, Raguain J, Sanchez C. **The costs of removing the unsanctioned import of marine plastic litter to small island states**. *Sci Rep* (2020.0) **10** 14458. DOI: 10.1038/s41598-020-71444-6 1424. Forrest AK, Hindell M. **Ingestion of plastic by fish destined for human consumption in remote South Pacific Islands**. *Aust J Marit Ocean Aff* (2018.0) **10** 81-97. DOI: 10.1080/18366503.2018.1460945 1425. Davies T. **Toxic space and time: slow violence, necropolitics, and petrochemical pollution**. *Ann Am Assoc Geogr* (2018.0) **108** 1537-1553. DOI: 10.1080/24694452.2018.1470924 1426. 1426Global Alliance for Incinerator Alternatives (GAIA). Discarded: Communities on the Frontlines of the Global Plastic Crisis. Global Alliance for Incinerator Alternatives (GAIA). 2019; 24. Accessed December 28, 2022. https://www.no-burn.org/wp-content/uploads/2022/02/Report-July-12-2019-Spreads-no-marks-1.pdf.. *Discarded: Communities on the Frontlines of the Global Plastic Crisis* (2019.0) **24** 1427. Conlon K. **Adaptive injustice: responsibility to act in the plastics economy**. *Resour Conserv Recycl* (2020.0) **153**. DOI: 10.1016/j.resconrec.2019.104563 1428. Liboiron M. *Pollution Is Colonialism* (2021.0). DOI: 10.1515/9781478021445 1429. Pellow DN, Brulle RJ. *Power, Justice, and the Environment: A Critical Appraisal of the Environmental Justice Movement* (2005.0) 1430. White R. *Life at the Fenceline: Understanding Cumulative Health Hazards in Environmental Justice Communities* (2018.0) **60** 1431. Coffee D, Diaz R. *The Perils of Plastic: Tracking Environmental, Climate and Human Impacts of Plastic Production, Use and Disposal* (2022.0) **30** 1432. 1432United Nations Environment Programme (UNEP). Single-Use Plastics, a Roadmap for Sustainability. Rev. 2. United Nations Environment Programme; 2018. Accessed December 28, 2022. https://www.unep.org/resources/report/single-use-plastics-roadmap-sustainability.. *Single-Use Plastics, a Roadmap for Sustainability* (2018.0) 1433. Beaumont NJ, Aanesen M, Austen MC. **Global ecological, social and economic impacts of marine plastic**. *Mar Pollut Bull* (2019.0) **142** 189-195. DOI: 10.1016/j.marpolbul.2019.03.022 1434. Mizutani J. **In the backyard of segregated neighborhoods: an environmental justice case study of Louisiana**. *Georget Envl Law Rev* (2018.0) **31** 363-390 1435. Bai Y, Givens JE. **Ecologically unequal exchange of plastic waste? A longitudinal analysis of international trade in plastic waste**. *J World-Syst Res* (2021.0) **27** 265-287. DOI: 10.5195/jwsr.2021.1026 1436. Castellón I. **Cancer alley and the fight against environmental racism**. *Villanova Environ Law J* (2021.0) **32** 15-43 1437. Owens KA, Conlon K. **Mopping up or turning off the tap? Environmental injustice and the ethics of plastic pollution**. *Front Mar Sci* (2021.0) **8**. DOI: 10.3389/fmars.2021.713385 1438. Bullard R, Mohai P, Saha R, Wright B. **Toxic wastes and race at twenty: why race still matters after all of these years**. *Environ Law* (2008.0) **38** 371-411 1439. Lerner S. *Sacrifice Zones: The Front Lines of Toxic Chemical Exposure in the United States* (2010.0). DOI: 10.7551/mitpress/8157.001.0001 1440. White TM. *Rise St. James, et. al v. Louisiana Department of Environmental Quality* 1441. Terrell KA, Julien GS. **Air pollution is linked to higher cancer rates among black or impoverished communities in Louisiana**. *Environ Res Lett* (2022.0) **17**. DOI: 10.1088/1748-9326/ac4360 1442. 1442U.S. Environmental Protection Agency (EPA). 2014 NATA: Assessment Results. US EPA. Published 2014. Accessed December 23, 2022. https://www.epa.gov/national-air-toxics-assessment/2014-nata-assessment-results. 1443. Hu MD, Lawrence KG, Gall M. **Natural hazards and mental health among US Gulf Coast residents**. *J Expo Sci Environ Epidemiol* (2021.0) **31** 842-851. DOI: 10.1038/s41370-021-00301-z 1444. Jambeck J, Moss E, Dubey B. *Leveraging Multi-Target Strategies to Address Plastic Pollution in the Context of an Already Stressed Ocean* (2020.0) **68** 1445. 1445World Health Organization, United Nations Children’s Fund. Progress on Household Drinking Water, Sanitation and Hygiene, 2000–2020: Five Years into the SDGs; 2021. Accessed December 24, 2022. https://data.unicef.org/resources/progress-on-household-drinking-water-sanitation-and-hygiene-2000-2020/.. *Progress on Household Drinking Water, Sanitation and Hygiene, 2000–2020: Five Years into the SDGs* (2021.0) 1446. Rashtian J, Chavkin DE, Merhi Z. **Water and soil pollution as determinant of water and food quality/contamination and its impact on female fertility**. *Reprod Biol Endocrinol* (2019.0) **17**. DOI: 10.1186/s12958-018-0448-5 1447. Edwards L, McCray NL, VanNoy BN. **Phthalate and novel plasticizer concentrations in food items from U.S. fast food chains: a preliminary analysis**. *J Expo Sci Environ Epidemiol* (2022.0) **32** 366-373. DOI: 10.1038/s41370-021-00392-8 1448. Dutko P, Ploeg MV, Farrigan T. *Characteristics and Influential Factors of Food Deserts* (2012.0) **36** 1449. Petrlik J, Bell L, Beeler B, Møller M, Jopkova M, Brabcova K. *Plastic Waste Poisoning Food and Threatening Communities in Africa, Asia, Central & Eastern Europe and Latin America* (2021.0) **50** 1450. Baptista AI, Perovich A. *U.S. Municipal Solid Waste Incinerators: An Industry in Decline* (2019.0) **87** 1451. Godfrey L. **Waste plastic, the challenge facing developing countries—ban it, change it, collect it?**. *Recycling* (2019.0) **4** 3. DOI: 10.3390/recycling4010003 1452. Morais J, Corder G, Golev A, Lawson L, Ali S. **Global review of human waste-picking and its contribution to poverty alleviation and a circular economy**. *Environ Res Lett* (2022.0) **17**. DOI: 10.1088/1748-9326/ac6b49 1453. 1453UNEP-IETC, GRID-Arendal. Gender and Waste Nexus: Experiences from Bhutan, Mongolia and Nepal; 2019. Accessed December 28, 2022. http://www.unep.org/resources/report/gender-and-waste-nexus-experiences-bhutan-mongolia-and-nepal.. *Gender and Waste Nexus: Experiences from Bhutan, Mongolia and Nepal* (2019.0) 1454. Buch R, Marseille A, Williams M, Aggarwal R, Sharma A. **From waste pickers to producers: an inclusive circular economy solution through development of cooperatives in waste management**. *Sustainability* (2021.0) **13** 9. DOI: 10.3390/su13168925 1455. Marques CP, Zolnikov TR, Noronha JM de, Angulo-Tuesta A, Bashashi M, Cruvinel VRN. **Social vulnerabilities of female waste pickers in Brasília, Brazil**. *Arch Environ Occup Health* (2021.0) **76** 173-180. DOI: 10.1080/19338244.2020.1787315 1456. 1456Women in Informal Employment Globalizing and Organizing (WIEGO). Waste Pickers: The Right to Be Recognized as Workers. 2013; 4. Accessed December 27, 2022. https://www.wiego.org/sites/default/files/resources/files/WIEGO-Waste-Pickers-Position-Paper.pdf.. *Waste Pickers: The Right to Be Recognized as Workers* (2013.0) **4** 1457. Allen C. *An Inclusive Recovery: The Social, Environmental and Economic Benefits of Partnering with Informal Recyclers* (2021.0). DOI: 10.46556/VPKH5682 1458. Dias S, Ogando AC. *From Theory to Action: Gender and Waste Recycling: A Toolkit for Teachers, Researchers and Practitioners* (2015.0) **32** 1459. 1459GA Circular. The Role of Gender in Waste Management: Gender Perspectives on Waste in India, Indonesia, the Philippines and Vietnam. 2019; 64. https://oceanconservancy.org/wp-content/uploads/2019/06/The-Role-of-Gender-in-Waste-Management.pdf.. *The Role of Gender in Waste Management: Gender Perspectives on Waste in India, Indonesia, the Philippines and Vietnam* (2019.0) **64** 1460. 1460Global Plastic Action Partnership (NPAP). A Roadmap for Radical Reduction of Plastic Pollution in Ghana. Ghana Plastic Action Partnership. 2021; 52. Accessed December 27, 2022. https://www.wacaprogram.org/sites/waca/files/knowdoc/NPAP_Ghana_Action_Roadmap.pdf.. *A Roadmap for Radical Reduction of Plastic Pollution in Ghana* (2021.0) **52** 1461. Obiria M, Donovan LS. **‘The smoke enters your body’: a toxic trash site in Kenya is making women sick**. *Vice* 1462. 1462European Parliament. Report on the Impact of Fisheries of Marine Litter. 2021; 26. https://www.europarl.europa.eu/doceo/document/A-9-2021-0030_EN.pdf.. *Report on the Impact of Fisheries of Marine Litter* (2021.0) **26** 1463. Morrissey K. **Aligning ocean plastic pollution and human health a co-benefits approach**. *J Ocean Coast Econ* (2019.0) **6** 17. DOI: 10.15351/2373-8456.1090 1464. 1464United Nations Environment Programme (UNEP). Intergovernmental negotiating committee to develop an international legally binding instrument on plastic pollution, including in the marine environment. UNEP – UN Environment Programme. Published 2022. Accessed December 13, 2022. http://www.unep.org/events/conference/inter-governmental-negotiating-committee-meeting-inc-1. 1465. **Key human rights considerations for the negotiations to develop an international legally binding instrument on plastic pollution** 1466. 1466Our Shared Seas. A Decade of Funding: 2010–2020 Landscape Review; 2021. Accessed December 28, 2022. https://oursharedseas.com/funding/.. *A Decade of Funding: 2010–2020 Landscape Review* (2021.0) 1467. Babayemi JO, Ogundiran MB, Weber R, Osibanjo O. **Initial inventory of plastics imports in Nigeria as a basis for more sustainable management policies**. *J Health Pollut* (2018.0) **8** 15. DOI: 10.5696/22156-9614-8.18.1 1468. Chukwuone NA, Amaechina EC, Ifelunini IA. **Determinants of household’s waste disposal practices and willingness to participate in reducing the flow of plastics into the ocean: evidence from coastal city of Lagos Nigeria**. *PLoS ONE* (2022.0) **17** 23. DOI: 10.1371/journal.pone.0267739 1469. Yalwaji B, John-Nwagwu HO, Sogbanmu TO. **Plastic pollution in the environment in Nigeria: a rapid systematic review of the sources, distribution, research gaps and policy needs**. *Sci Afr* (2022.0) **16** 12. DOI: 10.1016/j.sciaf.2022.e01220 1470. Bergmann M, Collard F, Fabres J. **Plastic pollution in the Arctic**. *Nat Rev Earth Environ* (2022.0) **3** 323-337. DOI: 10.1038/s43017-022-00279-8 1471. Akindele EO, Alimba CG. **Plastic pollution threat in Africa: current status and implications for aquatic ecosystem health**. *Environ Sci Pollut Res* (2021.0) **28** 7636-7651. DOI: 10.1007/s11356-020-11736-6 1472. Lusambili AM, Shumba CS. **Inequity is an iniquity: speaking up for aspiring African researchers**. *Health Hum Rights J* 1473. Boyd E, Chaffin BC, Dorkenoo K. **Loss and damage from climate change: a new climate justice agenda**. *One Earth* (2021.0) **4** 1365-1370. DOI: 10.1016/j.oneear.2021.09.015 1474. 1474The Pew Charitable Trusts, SYSTEMIQ. Breaking the Plastic Wave: A Comprehensive Assessment of Pathways Towards Stopping Ocean Plastic Pollution. 2020; 78. Accessed November 23, 2022. https://www.pewtrusts.org/-/media/assets/2020/07/breakingtheplasticwave_report.pdf.. *Breaking the Plastic Wave: A Comprehensive Assessment of Pathways Towards Stopping Ocean Plastic Pollution* (2020.0) **78** 1475. 1475ClientEarth. Plastics on Trial: A Briefing Series on Evolving Liability Risks Related to Plastics; 2022. Accessed November 23, 2022. https://www.clientearth.org/latest/documents/plastics-on-trial-2-hazardous-chemicals/.. *Plastics on Trial: A Briefing Series on Evolving Liability Risks Related to Plastics* (2022.0) 1476. 1476International Institute for Sustainable Development (IISD). Special Rapporteur Highlights Right to Science, Impact of Plastics on Human Rights. IISD SDG Knowledge Hub. Published September 16, 2021. Accessed November 23, 2022. https://sdg.iisd.org:443/news/special-rapporteur-highlights-right-to-science-impact-of-plastics-on-human-rights/. 1477. 1477United Nations General Assembly. Resolution 76/300: The Human Right to a Clean, Healthy and Sustainable Environment. Vol A/RES/76/300; 2022: 3. Accessed November 23, 2022. https://digitallibrary.un.org/record/3983329?ln=en.. *Resolution 76/300: The Human Right to a Clean, Healthy and Sustainable Environment* **2022** 3 1478. Bojke L, Schmitt L, Lomas J, Richardson G, Weatherly H. **Economic evaluation of environmental interventions: reflections on methodological challenges and developments**. *Int J Environ Res Public Health* (2018.0) **15** 9. DOI: 10.3390/ijerph15112459 1479. Seppelt R, Fath B, Burkhard B. **Form follows function? Proposing a blueprint for ecosystem service assessments based on reviews and case studies**. *Ecol Indic* (2012.0) **21** 145-154. DOI: 10.1016/j.ecolind.2011.09.003 1480. Gonzalez L. *Oil and Gas: Operations: Location Restrictions: Notice of Intention: Health Protection Zone: Sensitive Receptors* (2022.0) 1481. 1481GlobalRec. Kenya and South Africa announce just transition initiative for waste-pickers & other workers in the plastic value chain. International Alliance of Waste Pickers. Published December 2, 2022. Accessed December 13, 2022. https://globalrec.org/2022/12/02/kenya-south-africa-just-transition_wastepickers_plastic/. 1482. Castillo DP. **Integral ecology as a liberationist concept**. *Theol Stud* (2016.0) **77** 353-376. DOI: 10.1177/0040563916635781 1483. Jing Z, Sutikno S. **Legal issues on Indonesian marine plastic debris pollution**. *Indones Law Rev* (2020.0) **10**. DOI: 10.15742/ilrev.v10n1.592 1484. Gilchrist K. **She got plastic bags banned on Bali by 18. Now she wants to mobilize other young activists**. *CNBC* 1485. 1485The World Bank. Meet the innovators battling plastic waste in Indonesia: Mohamad Bijaksana Junerosano. The World Bank. Published May 30, 2019. Accessed November 23, 2022. https://www.worldbank.org/en/news/feature/2019/05/31/meet-the-innovators-battling-plastic-waste-in-indonesia-mohamad-bijaksana-junerosano. 1486. **Indonesia, a top plastic polluter, mobilizes 20,000 citizens to clean up the mess**. *Mongabay Environmental News* 1487. Goulopoulos S. **Bali officially bans all single-use plastic to curb ocean pollution**. *New York Post* 1488. Bhwana PG. **Jakarta officially bans the use of disposable plastic bags** 1489. 1489Piscespartnership.org. PISCES Partnership. PISCES. Accessed February 3, 2023. https://www.piscespartnership.org/. 1490. Siniwi RM. **Indonesia to reduce plastic waste 70% by 2025**. *JakartaGlobe* 1491. 1491Ministry of Environment and Forestry. National Plastic Waste Reduction Strategic Actions for Indonesia. 2020; 46. Accessed December 28, 2022. http://www.unep.org/ietc/resources/policy-and-strategy/national-plastic-waste-reduction-strategic-actions-indonesia.. *National Plastic Waste Reduction Strategic Actions for Indonesia* (2020.0) **46** 1492. **Benefits and costs of the Clean Air Act 1990–2020, the second prospective study** 1493. Grosse SD, Matte TD, Schwartz J, Jackson RJ. **Economic gains resulting from the reduction in children’s exposure to lead in the United States**. *Environ Health Perspect* (2002.0) **110** 563-569. DOI: 10.1289/ehp.02110563 1494. Bullard RD. *Dumping In Dixie: Race, Class, And Environmental Quality* (1990.0). DOI: 10.4324/9780429495274 1495. Bullard RD. *The Quest for Environmental Justice: Human Rights and the Politics of Pollution* (2005.0) **19** 1496. Claudio L. **Standing on principle: the global push for environmental justice**. *Environ Health Perspect* (2007.0) **115** A500-503. DOI: 10.1289/ehp.115-a500 1497. Morrone M, Liotta PH, Mouat DA, Kepner WG, Lancaster JM. *Environmental Change and Human Security: Recognizing and Acting on Hazard Impacts* (2008.0) 299-323. DOI: 10.1007/978-1-4020-8551-2_14 1498. Landrigan PJ, Rauh VA, Galvez MP. **Environmental justice and the health of children**. *Mt Sinai J Med J Transl Pers Med* (2010.0) **77** 178-187. DOI: 10.1002/msj.20173 1499. McDermott K. **Plastic pollution and the global throwaway culture: environmental injustices of single-use plastic**. *ENV 434 Environ Justice* 1500. **Frequently Asked Questions (FAQs)** 1501. **Summary of the first meeting of the Intergovernmental Negotiating Committee to Develop an International Legally Binding Instrument on Plastic Pollution: 28 November–2 December 2022**. *Earth Negot Bull* (2022.0) **36** 1-9 1502. 1502United Nations Environment Assembly of the United Nations Environment Programme (UNEP). Draft Report of the Intergovernmental Negotiating Committee to Develop an International Legally Binding Instrument on Plastic Pollution, Including in the Marine Environment, on the Work of Its First Session. United Nations Environment Programme (UNEP). 2022; 1–6. Accessed January 29, 2023. https://wedocs.unep.org/bitstream/handle/20.500.11822/41346/DraftReportINCWorkE.pdf.. *Draft Report of the Intergovernmental Negotiating Committee to Develop an International Legally Binding Instrument on Plastic Pollution, Including in the Marine Environment, on the Work of Its First Session* (2022.0) 1-6 1503. Wirth DA, Sachs NM, Rajamani L, Peel J. *Oxford Handbook of International Environmental Law* (2021.0) 574-592. DOI: 10.1093/law/9780198849155.003.0033 1504. Sachs NM, Wirth DA, Abelkop A, Abelkop A, Berkamp L, Bergeson LL, Auerbach B. (2023.0) 1505. **Convention on the Prohibition of the Use, Stockpiling, Production and Transfer of Anti-Personnel Mines and on their Destruction** 1506. Maslen S, Herby P. **An international ban on anti-personnel mines: history and negotiation of the “Ottawa treaty.”**. *Int Rev Red Cross* (1998.0) **325**. DOI: 10.1017/S0020860400091579 1507. **Treaty on the Prohibition of Nuclear Weapons** 1508. Simon N, Raubenheimer K, Urho N. **A binding global agreement to address the life cycle of plastics**. *Science* (2021.0) **373** 43-47. DOI: 10.1126/science.abi9010 1509. Bergmann M, Almroth BC, Brander SM. **A global plastic treaty must cap production**. *Science* (2022.0) **376** 469-470. DOI: 10.1126/science.abq0082 1510. 1510High Ambition Coalition to End Plastic Pollution (HAC). High Ambition Coallition to End Plastic Pollution homepage. High Ambition Coalition to End Plastic Pollution. Published January 5, 2023. Accessed January 5, 2023. https://hactoendplasticpollution.org/. 1511. **The Montreal Protocol on Substances that Deplete the Ozone Layer | Ozone Secretariat** 1512. **Era of leaded petrol over, eliminating a major threat to human and planetary health**. *UN Environment* 1513. Jakovcevic A, Steg L, Mazzeo N. **Charges for plastic bags: motivational and behavioral effects**. *J Environ Psychol* (2014.0) **40** 372-380. DOI: 10.1016/j.jenvp.2014.09.004 1514. Karasik R, Vegh T, Diana Z. *20 Years of Government Responses to the Global Plastic Pollution Problem: The Plastics Policy Inventory* (2020.0) **311** 1515. Gesley J 1516. **Sovereign Ordinance No. 8.633 of April 29, 2021 amending and codifying the regulations relating to waste** 1517. Anderson AG, Grose J, Pahl S, Thompson RC, Wyles KJ. **Microplastics in personal care products: exploring perceptions of environmentalists, beauticians and students**. *Mar Pollut Bull* (2016.0) **113** 454-460. DOI: 10.1016/j.marpolbul.2016.10.048 1518. **Microplastics** 1519. Pallone F. **H.R.1321 –**. (2015.0) 1520. Zuccaro P, Thompson DC, de Boer J. **Artificial turf and crumb rubber infill: an international policy review concerning the current state of regulations**. *Environ Chall* (2022.0) **9** 100620. DOI: 10.1016/j.envc.2022.100620 1521. Rognerud I, Steindal EH, Kumar N, Tartiu VE. *International Sustainability Criteria for Plastic Products in a Global Agreement on Plastic Pollution* (2021.0). DOI: 10.6027/temanord2022-509 1522. Wang Z, Adu-Kumi S, Diamond ML. **Enhancing scientific support for the Stockholm Convention’s implementation: an analysis of policy needs for scientific evidence**. *Environ Sci Technol* (2022.0) **56** 2936-2949. DOI: 10.1021/acs.est.1c06120 1523. Raubenheimer K, Uhro N. **Global governance of plastics and associated chemicals** 1524. Landrigan PJ, Goldman LR. **Children’s vulnerability to toxic chemicals: a challenge and opportunity to strengthen health and environmental policy**. *Health Aff (Millwood)* (2011.0) **30** 842-850. DOI: 10.1377/hlthaff.2011.0151 1525. 1525Organisation for Economic Co-operation and Development (OECD). OECD’s Environment, Health and Safety Programme. In: International Regulatory Co-Operation in Competition Law and Chemical Safety. Organisation for Economic Co-operation and Development (OECD). 2022; 1–50. DOI: 10.1787/23f53a28-en. *International Regulatory Co-Operation in Competition Law and Chemical Safety* (2022.0) 1-50. DOI: 10.1787/23f53a28-en 1526. **Food & Drug Administration (FDA). FDA: user fees explained**. *FDA* 1527. **Protocol on Civil Liability and Compensation for Damage Caused by the Transboundary Effects of Industrial Accidents on Transboundary Waters to the 1992 Convention on the Protection and Use of Transboundary Watercourses and International Lakes and to the 1992 Convention on the Transboundary Effects of Industrial Accidents** 1528. 1528United Nations Environment Assembly of the United Nations Environment Programme (UNEP). Basel Convention on the Control of Transboundary Movements of Hazardous Wastes and Their Disposal – Protocol on Liability and Compensation for Damage Resulting from Transboundary Movements of Hazardous Wastes and Their Disposal: Text and Annexes (Revised in 2019). United Nations Environment Programme (UNEP): Secretariat of the Basel Convention; 2020. Accessed January 29, 2023. http://www.basel.int/TheConvention/Overview/TextoftheConvention/tabid/1275/Default.aspx.. *Basel Convention on the Control of Transboundary Movements of Hazardous Wastes and Their Disposal – Protocol on Liability and Compensation for Damage Resulting from Transboundary Movements of Hazardous Wastes and Their Disposal: Text and Annexes (Revised in 2019)* (2020.0) 1529. 1529United Nations Environment Assembly of the United Nations Environment Programme (UNEP). Nagoya – Kuala Lumpur Supplementary Protocol on Liability and Redress to the Cartagena Protocol on Biosafety. United Nations Environment Programme (UNEP): Secretariat of the Convention on Biological Diversity. 2018; 1–40. Accessed January 24, 2023. https://bch.cbd.int/protocol/NKL_text.shtml.. *Nagoya – Kuala Lumpur Supplementary Protocol on Liability and Redress to the Cartagena Protocol on Biosafety* (2018.0) 1-40 1530. **Extended Producer Responsibility Project** 1531. Diggle A, Walker TR. **Implementation of harmonized Extended Producer Responsibility strategies to incentivize recovery of single-use plastic packaging waste in Canada**. *Waste Manag* (2020.0) **110** 20-23. DOI: 10.1016/j.wasman.2020.05.013 1532. 1532European Union. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions: A European Strategy for Plastics in a Circular Economy; 2018. Accessed January 12, 2023. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=COM%3A2018%3A28%3AFIN.. *Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions: A European Strategy for Plastics in a Circular Economy* (2018.0) 1533. Stolte A, Lamp J, Dederer G, Cocca M, Di Pace E, Errico ME. **Derelict fishing gear – removing a source of microplastics from the marine environment**. (2020.0) 72-81. DOI: 10.1007/978-3-030-45909-3_14 1534. Wright LS, Napper IE, Thompson RC. **Potential microplastic release from beached fishing gear in Great Britain’s region of highest fishing litter density**. *Mar Pollut Bull* (2021.0) **173** 113115. DOI: 10.1016/j.marpolbul.2021.113115 1535. 1535European Union. Directive 2008/56/EC of the European Parliament and of the Council of 17 June 2008: Establishing a Framework for Community Action in the Field of Marine Environmental Policy (Marine Strategy Framework Directive) (Text with EEA Relevance). Vol 164/19.; 2008. Accessed January 12, 2023. http://data.europa.eu/eli/dir/2008/56/oj/eng.. *Directive 2008/56/EC of the European Parliament and of the Council of 17 June 2008: Establishing a Framework for Community Action in the Field of Marine Environmental Policy (Marine Strategy Framework Directive) (Text with EEA Relevance)* (2008.0) **164/19** 1536. 1536European Union. Council Regulation (EC) No 1224/2009 of 20 November 2009: Establishing a Community Control System for Ensuring Compliance with the Rules of the Common Fisheries Policy, Amending Regulations (EC) No 847/96, (EC) No 2371/2002, (EC) No 811/2004, (EC) No 768/2005, (EC) No 2115/2005, (EC) No 2166/2005, (EC) No 388/2006, (EC) No 509/2007, (EC) No 676/2007, (EC) No 1098/2007, (EC) No 1300/2008, (EC) No 1342/2008 and Repealing Regulations (EEC) No 2847/93, (EC) No 1627/94 and (EC) No 1966/2006.; 2009. Accessed January 12, 2023. https://eur-lex.europa.eu/eli/reg/2009/1224/oj/eng.. *Council Regulation (EC) No 1224/2009 of 20 November 2009: Establishing a Community Control System for Ensuring Compliance with the Rules of the Common Fisheries Policy, Amending Regulations (EC) No 847/96, (EC) No 2371/2002, (EC) No 811/2004, (EC) No 768/2005, (EC) No 2115/2005, (EC) No 2166/2005, (EC) No 388/2006, (EC) No 509/2007, (EC) No 676/2007, (EC) No 1098/2007, (EC) No 1300/2008, (EC) No 1342/2008 and Repealing Regulations (EEC) No 2847/93, (EC) No 1627/94 and (EC) No 1966/2006* (2009.0) 1537. 1537Global Ghost Gear Initiative (GGGI). Best Practice Framework for the Management of Fishing Gear: June 2021 Update. Poseidon Aquatic Resources Management Ltd. 2021; 94. Accessed January 12, 2023. https://static1.squarespace.com/static/5b987b8689c172e29293593f/t/6160715a8230495ecf5af265/1633710447232/GGGI+Best+Practice+Framework+for+the+Management+of+Fishing+Gear+%28C-BPF%29+2021+Update+-+FINAL.pdf.. *Best Practice Framework for the Management of Fishing Gear: June 2021 Update* (2021.0) **94** 1538. **NetTag – Press Office – Newcastle University** 1539. **NetTag** 1540. Warrings R, Fellner J. **Management of aluminium packaging waste in selected European countries**. *Waste Manag Res* (2019.0) **37** 357-364. DOI: 10.1177/0734242X19832368 1541. Millette S. *Global Deposit Book 2022: An Overview of Deposit Return Systems for Single-Use Beverage Containers* (2022.0) 1-172 1542. Cousins IT, Ng CA, Wang Z, Scheringer M. **Why is high persistence alone a major cause of concern?**. *Environ Sci Process Impacts* (2019.0) **21** 781-792. DOI: 10.1039/C8EM00515J 1543. Scheringer M, Johansson JH, Salter ME, Sha B, Cousins IT. **Stories of global chemical pollution: will we ever understand environmental persistence?**. *Environ Sci Technol*. DOI: 10.1021/acs.est.2c06611 1544. Stephenson ME. **An approach to the identification of organic compounds hazardous to the environment and human health**. *Ecotoxicol Environ Saf* (1977.0) **1** 39-48. DOI: 10.1016/0147-6513(77)90016-1 1545. 1545United Nations Environment Assembly of the United Nations Environment Programme (UNEP). Minamata Convention on Mercury: Text and Annexes. United Nations Environment Programme (UNEP). 2019; 1–72. Accessed January 18, 2023. https://www.mercuryconvention.org/sites/default/files/2021-06/Minamata-Convention-booklet-Sep2019-EN.pdf.. *Minamata Convention on Mercury: Text and Annexes* (2019.0) 1-72 1546. 1546United Nations Environment Assembly of the United Nations Environment Programme (UNEP). Basel Convention on the Control of Transboundary Movements of Hazardous Wastes and Their Disposal. Entered in Force May 5, 1992. United Nations Environment Programme (UNEP): Secretariat of the Basel Convention; 1989. Accessed January 29, 2023. http://www.basel.int/Portals/4/download.aspx?d=UNEP-CHW-IMPL-CONVTEXT.English.pdf.. *Basel Convention on the Control of Transboundary Movements of Hazardous Wastes and Their Disposal. Entered in Force May 5, 1992* (1989.0) 1547. 1547United Nations Environment Programme (UNEP). Decisions Adopted by the Third Meeting of the Conference of the Parties to the Basel Convention: Decision III/1 Amendment to the Basel Convention. United Nations Environment Programme (UNEP). 1995; 1–48. Accessed January 24, 2023. https://treaties.un.org/doc/source/docs/UNEP_CHW.3_35-E.pdf.. (1995.0) 1-48 1548. 1548United Nations Environment Assembly of the United Nations Environment Programme (UNEP). Rotterdam Convention on the Prior Informed Consent Procedure for Certain Hazardous Chemicals and Pesticides in International Trade: Text and Annexes (Revised in 2019). United Nations Environment Programme (UNEP); 2020. Accessed January 11, 2023. http://www.pic.int/TheConvention/Overview/TextoftheConvention/tabid/1048/.. *Rotterdam Convention on the Prior Informed Consent Procedure for Certain Hazardous Chemicals and Pesticides in International Trade: Text and Annexes (Revised in 2019)* (2020.0) 1549. 1549United Nations International Martime Organization. Convention on the Prevention of Marine Pollution by Dumping of Wastes and Other Matters (London Protocol, 1972) and 1996 Amendments. United Nations International Maritime Organization. 2006; 1–25. Accessed January 11, 2023. https://wwwcdn.imo.org/localresources/en/OurWork/Environment/Documents/PROTOCOLAmended2006.pdf.. *Convention on the Prevention of Marine Pollution by Dumping of Wastes and Other Matters (London Protocol, 1972) and 1996 Amendments* (2006.0) 1-25 1550. Napper IE, Thompson RC. **Environmental deterioration of biodegradable, oxo-biodegradable, compostable, and conventional plastic carrier bags in the sea, soil, and open-air over a 3-year period**. *Environ Sci Technol* (2019.0) **53** 4775-4783. DOI: 10.1021/acs.est.8b06984 1551. Napper IE, Barrett AC, Thompson RC. **The efficiency of devices intended to reduce microfibre release during clothes washing**. *Sci Total Environ* (2020.0) **738** 140412. DOI: 10.1016/j.scitotenv.2020.140412 1552. Parker-Jurd FNF, Smith NS, Gibson L, Nuojua S, Thompson RC. **Evaluating the performance of the ‘Seabin’ – A fixed point mechanical litter removal device for sheltered waters**. *Mar Pollut Bull* (2022.0) **184** 114199. DOI: 10.1016/j.marpolbul.2022.114199 1553. Murphy F, Ewins C, Carbonnier F, Quinn B. **Wastewater treatment works (WwTW) as a source of microplastics in the aquatic environment**. *Environ Sci Technol* (2016.0) **50** 5800-5808. DOI: 10.1021/acs.est.5b05416 1554. Napper IE, Thompson RC. **Release of synthetic microplastic plastic fibres from domestic washing machines: effects of fabric type and washing conditions**. *Mar Pollut Bull* (2016.0) **112** 39-45. DOI: 10.1016/j.marpolbul.2016.09.025 1555. De Falco F, Cocca M, Avella M, Thompson RC. **Microfiber release to water, via laundering, and to air, via everyday use: a comparison between polyester clothing with differing textile parameters**. *Environ Sci Technol* (2020.0) **54** 3288-3296. DOI: 10.1021/acs.est.9b06892 1556. Shaykevich A. **No plastics panacea: chemical recycling causes pollution, promotes waste** 1557. Yang H, Ma M, Thompson JR, Flower RJ. **Waste management, informal recycling, environmental pollution and public health**. *J Epidemiol Community Health* (2018.0) **72** 237-243. DOI: 10.1136/jech-2016-208597 1558. McGlone C. **Is chemical recycling greenwashing?** 1559. **Trash incineration more polluting than coal** 1560. Marie C, Lémery D, Vendittelli F, Sauvant-Rochat MP. **Phthalate exposure in pregnant women: risk perception and preventive advice of perinatal health professionals**. *Matern Child Health J* (2019.0) **23** 335-345. DOI: 10.1007/s10995-018-2668-x 1561. van Sebille E, Wilcox C, Lebreton L. **A global inventory of small floating plastic debris**. *Environ Res Lett* (2015.0) **10** 124006. DOI: 10.1088/1748-9326/10/12/124006 1562. Zhang Q, Xu EG, Li J. **A Review of Microplastics in Table Salt, Drinking Water, and Air: Direct Human Exposure**. *Environ Sci Technol* (2020.0) **54** 3740-3751. DOI: 10.1021/acs.est.9b04535 1563. Taniguchi S, Colabuono FI, Dias PS. **Spatial variability in persistent organic pollutants and polycyclic aromatic hydrocarbons found in beach-stranded pellets along the coast of the state of Sao Paulo, southeastern Brazil**. *Mar Pollut Bull* (2016.0) **106** 87-94. DOI: 10.1016/j.marpolbul.2016.03.024 1564. Camacho M, Herrera A, Gómez M. **Organic pollutants in marine plastic debris from Canary Islands beaches**. *Sci Total Environ* (2019.0) **662** 22-31. DOI: 10.1016/j.scitotenv.2018.12.422 1565. Arias AH, Alvarez G, Pozo K. **Beached microplastics at the Bahia Blanca Estuary (Argentina): Plastic pellets as potential vectors of environmental pollution by POPs**. *Mar Pollut Bull* (2023.0) **187** 114520. DOI: 10.1016/j.marpolbul.2022.114520 1566. Pozo K, Urbina W, Gómez V. **Persistent organic pollutants sorbed in plastic resin pellet—“Nurdles” from coastal areas of Central Chile**. *Mar Pollut Bull* (2020.0) **151** 110786. DOI: 10.1016/j.marpolbul.2019.110786 1567. Ohgaki T, Takada H, Yoshida R. **International pellet watch: Global monitoring of polybrominated diphenyl ethers (PBDEs) in plastic resin pellets**. *Environ Monit Contam Res* (2021.0) **1** 75-90. DOI: 10.5985/emcr.20210002 1568. Yeo BG, Takada H, Hosoda J. **Polycyclic Aromatic Hydrocarbons (PAHs) and Hopanes in Plastic Resin Pellets as Markers of Oil Pollution via International Pellet Watch Monitoring**. *Arch Env Contam Toxicol* (2017.0) **73** 196-206. DOI: 10.1007/s00244-017-0423-8 1569. Karlsson T, Brosché S, Alidoust M, Takada H. **Plastic Pellets Found on Beaches All over the World Contain Toxic Chemicals**. *International Pollutants Elimination Network (IPEN)* (2021.0) 1570. Alidoust M, Yeo GB, Mizukawa K, Takada H. **Monitoring of polycyclic aromatic hydrocarbons, hopanes, and polychlorinated biphenyls in the Persian Gulf in plastic resin pellets**. *Mar Pollut Bull* (2021.0) **165** 112052. DOI: 10.1016/j.marpolbul.2021.112052
--- title: Impact of administration route on nanocarrier biodistribution in a murine colitis model authors: - Catherine C. Applegate - Hongping Deng - Brittany L. Kleszynski - Tzu-Wen L. Cross - Christian J. Konopka - L. Wawrzyniec Dobrucki - Erik R. Nelson - Matthew A. Wallig - Andrew M. Smith - Kelly S. Swanson journal: Journal of experimental nanoscience year: 2022 pmcid: PMC10038121 doi: 10.1080/17458080.2022.2134563 license: CC BY 4.0 --- # Impact of administration route on nanocarrier biodistribution in a murine colitis model ## Body The incidence of inflammatory bowel disease (IBD), including Crohn’s disease and ulcerative colitis, is increasing worldwide, with the United States showing the highest prevalence rate [1]. IBD is characterized by chronic, idiopathic, and relapsing inflammation of the mucosal lining of the gastrointestinal (GI) tract. If left untreated, this chronic inflammation leads to GI distress and tissue damage, lowering patient quality of life and potentially resulting in more serious complications such as strictures, abscesses, and fistulas as well as an increased risk for colon cancer [2]. While the etiology of IBD remains unknown, a wealth of accumulating evidence suggests that aberrant infiltration and activation of proinflammatory immune cells is the driving force of disease pathogenesis [3, 4]. The diagnosis of IBD typically requires invasive colonoscopy and biopsy for histopathological confirmation of intestinal inflammation, with magnetic resonance imaging (MRI) and X-ray fluoroscopy or computed tomography (CT) used to evaluate the extent of the disease and to monitor disease activity [5]. Although highly sensitive and high in resolution, MRI and CT lack the specificity of positron emission tomography (PET), which can be used to quantitatively assess inflammation throughout the whole gut, anatomically localized by correlation with CT imaging [6]. Radiolabeled fluorodeoxyglucose (18F-FDG) can be used to identify inflammatory lesions by PET/CT imaging; however, non-specific uptake of FDG in the healthy bowel can lead to false positive results, and the probe is more a direct measure of metabolism than inflammation [7]. PET detection of radiolabeled autologous leukocytes that home to the site of intestinal inflammation offers improved specificity compared with FDG, but this method requires specialized handling of patient blood for leukocyte isolation, radiolabeling, and reinjection [8]. Nanoparticles have been widely investigated as contrast agents for imaging-based evaluation of disease extent [9–11] and for delivering colon-targeted therapies for IBD [12–16]. However, few studies have examined the potential to detect intestinal inflammation using radiolabeled nanoparticles for high-specificity PET imaging [17]. Furthermore, it is not clear which administration route would be best suited for delivery of inflammation-targeting nanomaterials in IBD to maximize the contrast between inflamed and healthy intestines. Nanoparticles natively distribute to healthy tissues [18], and uptake specificity in inflamed tissues depends on the route of administration [19]. In the intestines, the luminal surface is inflamed in IBD and is directly accessible via oral and enema administrations, while the visceral membrane is accessible through intraperitoneal (IP) administration. Systemic administration (e.g. intravenous) can transport nanomaterials to intestinal tissues; however, uptake in inflamed tissue derives from vascular permeability rather than specific interstitial cell populations, and the vast majority of material distributes to organs of the mononuclear phagocytotic system (liver and spleen) [17, 20]. The goal of this study was to evaluate the impact of administration route on the distribution of nanocarriers to inflamed intestinal tissue in a rodent model of colitis. We focused on a polysaccharide (dextran) nanocarrier due to its specific uptake by macrophages, cells that are prevalent in inflamed intestines. We previously used the same material conjugated to a copper (Cu) radiolabel to target inflamed visceral adipose depots of obese mice and found that >$90\%$ of the 500 kDa nanocarrier was taken up by adipose tissue (AT) macrophages [21]. In addition to their macrophage targeting ability, polysaccharides such as dextran are advantageous to use as nanocarriers due to their low cost, biocompatibility, ease of chemical conjugation, and long record of clinical safety for use both as a blood volume expander and to reduce platelet aggregation [22]. In this study, we investigated the differences between healthy and inflamed tissue biodistributions of a dextran-based nanocarrier conjugate across oral, enema, and IP routes of administration. Through PET/CT imaging and quantitative radioisotopic analysis, we found that IP delivery resulted in the highest nanocarrier retention across all measured tissues of both healthy animals and animals with dextran sodium sulfate (DSS)-induced colitis. Few intestinal tissues were statistically distinguishable between the healthy and DSS colitis animal groups for any of the administration routes due to a high degree of inter-subject variability in absolute distribution to intestinal tissues. This variability was reduced by ratiometric normalization to liver or total intestinal uptake to reveal significant group differences, especially for oral and IP administration routes. IP delivery resulted in higher levels of nanocarrier accumulation within liver tissues of DSS colitis animals, which may indicate disease-specific hepatic processes that can be probed independently by these contrast agents. These findings can inform the design of larger scale studies aimed to evaluate longitudinal progression or therapeutic response in IBD using PET/CT. ## Abstract The incidence of inflammatory bowel disease (IBD) is increasing worldwide. Although current diagnostic and disease monitoring tests for IBD sensitively detect gut inflammation, they lack the molecular and cellular specificity of positron emission tomography (PET). In this proof-of-concept study, we use a radiolabeled macrophage-targeted nanocarrier probe (64Cu-NOTA-D500) administered by oral, enema, and intraperitoneal routes to evaluate the delivery route dependence of biodistribution across healthy and diseased tissues in a murine model of dextran sodium sulfate (DSS)-induced colitis. High inter-subject variability of probe uptake in intestinal tissue was reduced by normalization to uptake in liver or total intestines. Differences in normalized uptake between healthy and DSS colitis animal intestines were highest for oral and IP routes. Differences in absolute liver uptake reflected a possible secondary diagnostic metric of IBD pathology. These results should inform the preclinical development of inflammation-targeted contrast agents for IBD and related gut disorders to improve diagnostic accuracy. ## Synthesis of 64Cu-NOTA-D500 Dextran conjugates of the chelator 1,4,7-triazacyclononane-1,4,7-triacetic acid were synthesized from aminated 500 kDa dextran (D500-NH2) as reported previously [21]. D500-NH2 (10 mg) was dissolved in 5 mL anhydrous dimethylsulfoxide (DMSO) and anhydrous triethylamine (10 μL) was added. A solution of p-SCN-Bn-NOTA (0.68 mg) in 1 mL anhydrous DMSO was added, and the mixture was stirred at room temperature for 16 hr. The product was purified using an Amicon filter (MWCO 30 kDa), and the solid product was collected after lyophilization. To label with 64Cu, a solution of NOTA-D500 in sodium acetate buffer (pH = 5.5, 0.1 M) was mixed with a solution of 64CuCl2 (Washington University, St. Louis, MO, USA). The mixture was incubated for 30 min at 37 °C before nonradioactive CuCl2 (equivalent molar amount to NOTA in the dextran conjugate) was added to saturate unreacted NOTA. The mixture was incubated for 20 min and ethylenediaminetetraacetic acid (EDTA, 2 equivalents to total Cu) was added to scavenge nonspecifically bound and/or free Cu. After incubation for 10 min, free Cu, EDTA, and EDTA-Cu chelates were removed using an Amicon filter (MWCO 30 kDa), and buffer was exchanged to phosphate buffered saline (PBS). Radiochemical purity (RCP) was determined by thin-layer chromatography, and samples with RCP > $95\%$ were used for imaging experiments. ## Animals and diets All animal procedures were approved by the University of Illinois Institutional Animal Care and Use Committee prior to experimentation (protocol #17087). Six-week old male C57Bl/6J mice (Jackson Laboratory, Bar Harbor, ME, USA) were purchased and housed individually in standard shoebox cages in a temperature- and humidity-controlled room, with a 12 hr light/dark cycle. Upon arrival, mice were fed a low-fat diet ($10\%$ kcal from fat; Research Diets Inc. #D12450J, New Brunswick, NJ, USA) ad libitum with free access to fresh water and allowed to acclimate to the facility for 2 wk prior to DSS treatment regimen. ## DSS treatment Previous studies showed that providing $3\%$ DSS water for 5 d is sufficient to induce colitis in C57Bl/6J mice, with similar pathology to UC in humans [23, 24]. After acclimation, mice were randomly assigned to receive distilled water (control; Con) or water containing DSS (DSS) for 5 d (d 1–5) to induce acute colitis. For animals treated with DSS, regular drinking water was replaced with $3\%$ (wt:vol) DSS (36,000–50,000 kDa molecular weight; MP Biochemicals, Santa Ana, CA, USA) in sterile water. Freshly made DSS-water was provided on d 1, 3, and 5, and fecal scores were assessed daily to monitor colitis. Fecal scores were assigned based on stool consistency and the presence of occult blood, as previously described [25], and recorded daily using fresh (within 15 min of defecation) fecal samples. A hemoccult kit (Beckman Coulter, Brea, CA, USA) was used to determine the presence of occult blood in stool. Following DSS treatment, all mice were provided with fresh, untreated water for the remaining duration of the study. ## Biodistribution studies To determine the optimal nanocarrier administration route, mice were administered 64Cu-NOTA-D500 (~100 μCi) by oral, IP, or enema routes, 2 days after completing DSS treatment (d 8). Con and DSS mice were both used ($$n = 3$$/group). For oral administration, 100 μL of nanocarrier in PBS was administered by oral gavage to unanesthetized mice. For IP administration, unanesthetized mice were injected with 100 μL of nanocarrier in PBS on the left side. For enema administration, mice were anesthetized by isoflurane and infused with 100 μL of nanocarrier in PBS. Mice were euthanized by CO2 asphyxiation and cervical dislocation 24 hr following nanocarrier treatment. Blood was collected by cardiac puncture, and tissues were collected including adipose depots (subcutaneous, gonadal, perirenal, mesenteric), liver, spleen, stomach, and intestines (sectioned by duodenum, jejunum, ileum, proximal colon, mid colon, distal colon, and cecum). Radioactivity in tissues was measured using a Wizard2 Automatic γ-counter (PerkinElmer, Waltham, MA, USA) as previously described [21]. One mouse from each group was randomly selected to undergo serial live PET/CT imaging at 4 and 24 hr following conjugate delivery using an Inveon PET/CT system (Siemens Healthcare, Malvern, PA, USA) as previously described [21]. CT contrast of the intestines was provided by enema-delivered iohexol (GE Healthcare, Chicago, IL, USA). ## Histopathology Mid-colon tissues were fixed in $10\%$ neutral-buffered formalin for 24 hr and then transferred to $70\%$ ethanol until embedding in paraffin (Tissue Tek VIP, Sakura Finetek USA, Inc., Torrance, CA, USA). Tissues were sectioned into 7 μm-thick slices using a micro-tome (Microm HM 310, MICROM Laborgeräte GmbH, Berlin, Germany) and mounted onto glass slides (Surgipath® R X-Tra® R Microscope Slides, Leica Biosystems, Buffalo Grove, IL, USA). Mounted sections were stained with Harris’ hematoxylin and eosin (H&E), and sections were blindly evaluated for the presence of colitis by a board-certified veterinary pathologist at the University of Illinois (MAW). ## Statistical analysis Biodistribution data were analyzed between treatment groups within each mode of delivery (Wilcoxon Rank Sum) and between modes of delivery (Kruskal-Wallis with post hoc analysis by the Dwass, Steel, Critchlow-Fligner method) using SAS (v.9.4, SAS Institute, Cary, NC, USA). Statistical significance was set as $p \leq 0.05$ and a trend was set as $p \leq 0.10.$ ## Dextran nanocarrier radiochelate: 64Cu-NOTA-D500 The synthesis scheme for 64Cu-NOTA-D500 is shown in Figure 1A. Polysaccharides such as dextran are selectively internalized by macrophages through binding with plasma membrane C-type lectins and class A scavenger receptors [26, 27]. Our previous work demonstrated efficient uptake of dextran polysaccharides of high molecular weight (500 kDa) by macrophages both in vitro and in vivo, with higher specificity for uptake by M1 macrophages present within inflamed AT in a murine model of obesity [21]. We therefore synthesized contrast agents using 500 kDa dextran derived from aminated commercial biopolymers, linked to NOTA through stable thiourea bonds. The dextran amines were saturated by excess p-SCN-NOTA (2.5:1 to amines) with base catalysis and purified. Conjugation was confirmed via proton nuclear magnetic resonance (1H NMR) spectroscopy (Figure S1), showing the appearance of peaks at 7.0–7.3 ppm that were consistent with NOTA. The number of NOTA groups per dextran was near 92 based on ICP-MS elemental analysis of Cu after reaction with excess CuCl2 and purification by filtration. To prepare the radiochelate, 64CuCl2 was added to NOTA-D500 in acetate buffer (pH 5.0) before saturation of NOTA with excess cold Cu, followed by repeated purification by filtration. ## Dependence of absolute 64Cu-NOTA-D500 biodistribution on administration route Mice were treated with $3\%$ DSS in water for 5 d to induce a colitis-like phenotype [23, 24] or treated with plain water (Con). Animals were allowed to recover for 2 d and were randomized within treatment groups to receive a single dose of radiolabeled probe (64Cu-NOTA-D500) by either oral gavage, enema, or IP injection ($$n = 3$$/group; Figure 1B). Colitis was monitored by presence of occult fecal blood, and extensive colitis was observed in both the colon and cecum of all mice, as evaluated by a board-certified veterinary pathologist (MAW) post mortem. Colitis in both animal models and humans is associated with a disruption in the mucus layer of the GI tract at the site of inflammation, leading to mucosal collapse [28–30], as verified in our histologic observations (Figure 2). A healthy mucus barrier may trap nanoparticles [31], as polymeric particles can bind to proteins present within the mucus [32]. Alternatively, loss of mucosa in colitis may be associated with more rapid probe diffusion, uptake, and/or clearance. Twenty-four hours after 64Cu-NOTA-D500 administration, post mortem gamma well counting (GWC) was used to evaluate tissue uptake of the probe in excised tissues, including intestines, stomach, liver, spleen, and AT depots. Animals administered with oral gavage showed detectable probe levels in all tissues of the GI tract as well as the liver (Figure 3A). Complete data are provided in Table 1. Retention in the stomach and total intestines was higher on average for Con mice compared with DSS mice, although absolute values were highly variable, so these differences were not statistically significant ($27.2\%$ vs. $3.9\%$ and $20.5\%$ vs. $10.3\%$ of the injected dose per gram [I.D./g] tissue, respectively). This effect may be due to decreased gut transit time with IBD or degradation by microbiota-derived dextranases during 64Cu-NOTA-D500 passage through the digestive system [33]. Overall, retention in individual tissues of the intestine, where inflammation is observed with colitis, was low (<$2\%$) and not statistically different between Con and DSS mice. Enema administration led to significantly less uptake by the small intestines (duodenum, jejunum, ileum) compared with oral administration (Figure 3B). Higher mean retention was measured throughout intestinal tissues of DSS mice ($6.4\%$ I.D./g) compared with those of Con mice ($4.4\%$ I.D./g), which was opposite to that of oral gavage. Intestinal uptake differences between the DSS and Con mice were only statistically significant in the distal colon ($1.6\%$ for DSS mice vs. $0.2\%$ I.D./g for Con mice), which is the primary site of DSS-induced colitis [23]. IP administration of 64Cu-NOTA-D500 led to significantly greater absolute retention throughout the body 24 hr post-injection in comparison to oral and enema delivery modes (Figure 3C). As apparent from the PET/CT images, a large proportion of the I.D. localized to the liver and AT in both Con and DSS animals. Interestingly, IP injection resulted in significantly higher probe uptake in livers of DSS-treated animals ($116.0\%$ I.D./g) compared with livers of Con animals ($65.0\%$ I.D./g), while uptake in the spleen followed the opposite pattern. The liver and spleen contain the highest concentrations of macrophages and are thus the sites of dextran biodistribution after systemic administration [34]. The liver is strongly affected during IBD progression, with up to $50\%$ of patients also experiencing some form of hepatobiliary disease [35]. As such, increased liver uptake of the probe in DSS mice may reflect an increased population of liver macrophages in the inflamed colitis state, redirecting 64Cu-NOTA-D500 from the spleen to the liver. Absolute retention by the intestines was significantly higher following IP administration compared with oral gavage or enema. Overall intestinal distribution trended higher for Con animals ($48.7\%$ I.D./g) compared with DSS-treated animals ($32.6\%$ I.D./g). These differences were statistically significant within the cecum ($7.6\%$ vs. $3.1\%$ I.D./g) but were only trending within the proximal colon ($7.1\%$ vs. $3.8\%$ I.D./g). It was not determined if intestinal uptake was due to direct uptake across the visceral mesothelium of the intestines or as a result of hepatobiliary excretion [36], although an intact mesothelium in both inflamed and non-inflamed colons was evident from histopathology (Figure 2), so the latter is more likely. Therefore, while both IP injection and oral gavage led to superior absolute probe uptake and retention in intestines relative to enema administration, only enema delivery led to greater disease-specific probe retention at the site of colitis-associated inflammation (colon) in DSS animals compared with healthy Con animals. ## Relative tissue uptake: liver normalization High variability of biodistribution is an expected outcome for the three administration routes explored here for which the probe solution is initially dispersed to the surface or lumen of the GI tract, which is itself variable, heterogenous in contents, and undergoing peristalsis [37]. A useful approach to addressing this type of variability is to calibrate the target tissue uptake based on probe dose in a secondary tissue for which the probe must transport across similar biological barriers to reach that tissue. Because nanoparticle accumulation in the liver is predominant after most administration routes [38], we compared intestinal tissue uptake relative to liver uptake as a means to evaluate organ-level specificity of 64Cu-NOTA-D500 for GI tract inflammation. The liver is the major xenobiotic metabolizer and was a site of significant 64Cu-NOTA-D500 uptake following all administration routes. After adjusting for distribution to the liver, we observed that oral gavage led to significantly lower probe retention in total intestines and most intestinal tissue regions of DSS mice compared with tissues of Con mice, which was similar to non-significant trends in absolute probe biodistribution. Data and results are shown in Table 2. Unlike oral administration, little difference in liver-normalized intestine tissues was measured between DSS and Con animals after enema administration, with the exception of the cecum. IP injection was associated with lower liver-normalized probe retention across most tissues in DSS animals compared with Con animals, including within select AT depots (subcutaneous, right gonadal, mesenteric) and individual intestinal tissues (all except mid colon). This outcome reflects the significantly higher uptake of the probe in livers of DSS mice relative to Con mice. Across all three modes of administration, total intestinal uptake of probe relative to the liver was lowest following IP injection. Low intestinal uptake after IP administration likely reflects a lack of direct uptake by the intestines after administration, instead requiring hepatobiliary transport to reach the target tissue. Histologic observations (Figure 2) as well as histopathological reports in other DSS colitis murine models have shown that, although commonly observed in humans, macrophage transepithelial migration and transmural inflammation are only occasionally observed in DSS colitis models [23, 39]. As a result, IP injection would direct the probe toward alternative tissue targets, specifically to the liver, with greater uptake for colitis-affected animals compared with healthy animals. The global impact of liver-based normalization on inter-subject variability is shown in Figure 4A, comparing the relative standard deviation (RSD) of each intestinal tissue with and without normalization. For oral gavage, RSD reduced with liver-based normalization in most tissues, presumably because the quantity of probe entering the intestines distributes proportionally to the hepatic portal system and to further regions of the intestines. In contrast, liver-based normalization after enema administration resulted in mixed outcomes in terms of variability of intestine uptake: RSD for small intestinal tissues and the cecum were strongly reduced; however, RSD for large intestine tissues increased. In addition, a trend of increasing RSD was observed with posterior intestinal position, suggesting that liver normalization is most impactful for the most anterior small intestinal tissues where hepatic uptake occurs. For IP delivery, absolute RSD was smaller overall and uptake normalization by liver had less impact on RSD, although normalization was necessary to observe differences between Con and DSS animals. ## Relative tissue uptake: total intestine normalization Uptake of 64Cu-NOTA-D500 in individual intestinal tissues was evaluated relative to total intestinal retention for each of the administration routes and animal groups (Table 2). As shown in Figure 4B, normalization by total intestines led to a reduction in inter-individual variability for most intestinal tissues for both oral and enema administration routes. With oral administration, DSS animals exhibited significantly greater intestine-normalized probe uptake in tissues of both the small intestine (duodenum, jejunum) and large intestine (proximal and distal colon) compared with Con animals. However, the opposite association was observed for the cecum across all routes of administration. As discussed above, lower probe retention in the cecum of DSS animals may be associated with more efficient clearing from the GI tract due to less intact mucosal lining in some areas of the intestines. In addition, more rapid colonic motility is observed in colitis [40], reducing the time the conjugate spends in the cecum of DSS animals. Although results across modes of administration were not drastically different, both IP and oral administration led to significantly greater relative probe uptake in the duodenum and jejunum compared with enema delivery, while IP injection led to significantly greater uptake by the distal colon compared with oral gavage. As in the case of liver normalization, the RSD was much smaller for IP delivery compared with other administration routes; however, normalization to the total intestines did little to improve group-based differences between Con and DSS animals. Overall, the distribution of probe relative to the total intestines shows that there may be successful targeting of the dextran conjugate within specific intestinal tissues in both colitis-affected and healthy animals when accounting for the total dose that reaches and is retained in the intestines. ## In vivo PET/CT imaging of 64Cu-NOTA-D500 biodistribution for three administration routes Whole-body PET/CT imaging of 64Cu-NOTA-D500 was performed in mice longitudinally at 4 hr and 24 hr following conjugate administration as a comparison to values observed by ex vivo GWC. Tissue biodistributions were measured by CT image segmentation of tissues, using iodine-based contrast for intestinal tissues (Figure 5A). Oral gavage led to a distribution pattern that was primarily confined to the GI tract after 4 and 24 hr, consistent with our ex vivo GWC studies and similar to previous studies assessing bioavailability of dextran-based oral prodrugs [41]. PET quantification of radioactivity 4 hr after oral administration showed significantly greater probe localization in colon tissues of DSS mice compared with Con mice (Figure 5B), especially relative to total intestines, which was also consistent with ex vivo GWC studies. At the 24-hr timepoint, the majority of the probe had cleared, and differences were no longer apparent. Good agreement was observed between PET and ex vivo radioisotopic quantification by GWC at the 24 hr time point (Figure 5C), suggesting that oral gavage is an effective route of administration for accurate measurement of intestinal tissue distribution by PET/CT. PET/CT images displaying probe uptake by tissues following enema administration (Figure 5A) showed predominant localization to the colon, small intestines, and stomach in both Con and DSS mice 4 and 24 hr post-administration (Figure 5B). Significantly higher total probe distribution was measured in the healthy mouse, and as in the oral gavage case, most of the probe had cleared at the 24 hr time point. PET radioisotopic quantification showed poor agreement with ex vivo GWC results at the 24 hr time point (Figure 5C) likely due to the significant impact of intestinal contents following enema administration. IP administration resulted primarily in uptake by the liver and spleen, with less distribution to the colon in both Con and DSS mice at the 4- and 24-hr timepoints (Figure 5A). Consistent with ex vivo studies above, PET contrast was significantly higher in livers of DSS mice. Unlike for the case of oral gavage and enema, the vast majority of radioactivity was retained within the body after 24 hr and PET/CT quantification was consistent at 4 and 24 hr post-injection (Figure 5B), providing a large timeframe in which to image. Furthermore, PET radioisotopic quantification closely followed ex vivo GWC results across both Con and DSS animals (Figure 5C), suggesting that IP injection may provide the most accurate means for quantifying contrast agent localization. ## Discussion With the increasing prevalence of IBD worldwide and the deleterious consequences associated with a lack of treatment, it is important to develop diagnostic tools that accurately identify and characterize areas of inflammation associated with IBD. Colonoscopy and endoscopy can visualize limited sections of the intestines, but these methods are invasive and lack repeatability [42]. Although MRI and CT imaging provide a more global assessment of intestinal and extraintestinal inflammation, these methods rely on anatomical contrast alone or non-specific accumulation of contrast agents in areas of active inflammation [43]. Radiolabeled nanomaterials can directly target inflammatory cells and can be visualized by PET to quantitatively assess inflammation severity across independent lesions, improving the specificity of the combined CT imaging modality. Here, we showed that oral gavage of a radiolabeled dextran conjugate (64Cu-NOTA-D500) led to greater uptake by DSS colitis-affected tissues compared with healthy tissues when individual intestinal regions were measured relative to uptake by the total intestines. In vivo PET image quantification results followed ex vivo GWC results, indicating that oral administration may be preferential for contrast agent visualization of regional gut inflammation. The opposite distribution pattern was observed when intestinal tissues were normalized by the liver, with DSS tissue accumulating less probe than Con tissues. Because the same trend was observed after IP injection, for which DSS animal livers retained nearly two times the quantity of probe compared with Con animals, we attribute this result to altered hepatic function in the case of DSS. Analysis of relative uptake in specific intestinal regions, total intestines, and liver may together be a useful approach for assessing pathological changes across these organ systems in vivo. We also showed that enema administration of 64Cu-NOTA-D500 led to significantly greater absolute contrast accumulation within the distal colon of animals affected with DSS colitis compared with healthy control animals. As the primary site of inflammation induced by DSS colitis [23], the distal colon is an appropriate region of interest for evaluating differential probe localization. However, in vivo PET/CT measurements were inconsistent with ex vivo GWC results, and differences were no longer significant after tissue normalization to total intestines or liver. The inter-subject variability was highest for enema administrations, suggesting that this administration route is inferior to the others, likely due to the intrinsic variability of intestinal contents and possible heterogeneous mechanical effects related to this administration route [37]. When evaluating the biodistribution of 64Cu-NOTA-D500 relative to the liver and the intestines, we unexpectedly observed greater probe accumulation within the cecum of Con mice vs. DSS mice across all modes of administration. Con mice should exhibit a more intact mucosa and reduced colonic motility overall throughout the intestines, but there may be less inflammation of this intestinal tissue region in the DSS model compared to other regions [23]. It is also important to consider that reduced intraintestinal pH is commonly observed in IBD patients [44]. Dextran uptake by macrophage lectin receptors such as the mannose receptor is pH-dependent, with binding affinity of both receptors attenuated at pH levels below neutral, physiological values [45–47]. Moreover, pH reductions lead to a closed conformation of the mannose receptor [48, 49], additionally inhibiting dextran uptake by macrophages in a more acidic environment [50]. We also demonstrated that IP delivery of 64Cu-NOTA-D500 resulted in higher uptake in most measured tissues in both healthy and DSS colitis mice when compared with oral or enema delivery. While the highest total I.D. of the probe was found in off-target tissues (e.g. visceral AT, liver), absolute retention of the probe by intestinal tissues was much higher 24 hr following IP administration compared with other delivery routes. Notably, in vivo measurements remained relatively steady between 4 and 24 hr post administration, and PET measurements were consistent with ex vivo GWC results at the 24-hr timepoint. Moreover, inter-subject variability was much less for intestinal tissues after IP injection compared with either oral or enema delivery routes. However, lower liver-normalized contrast within intestinal tissues of DSS animals was observed compared with healthy Con animals, which likely reflects altered liver physiology as opposed to intestinal inflammation. Together with the clinical challenges associated with the use of contrast agents with long retention durations, these results suggest that IP injection is likely to be less effective overall than oral administration. Histopathological evaluation of colonic tissues indicated that animals with the greatest intestinal damage presented with inflammatory cell infiltration into the submucosa following mucosal collapse. However, there was no evidence of transmural inflammation, as shown in Figure 2. Similar histopathological reports in other DSS colitis murine models have also shown that macrophage transepithelial migration and transmural inflammation are only occasionally observed in DSS colitis models, whereas IBD in humans is commonly associated with transmural inflammation [23, 39]. As such, 64Cu-NOTA-D500 contact with the affected luminal tissues of the colon by administration within the peritoneal cavity may be limited by the shallow depth of inflammation with insufficient permeability of the outer muscularis and serosa layers. Previous studies in rats showed a likely molecular weight limit of 45 kDa for permeation into healthy intestinal tissues through the viscera [51]. In models of transmural inflammation with macrophage transepithelial migration, not only may peritoneal macrophages that have taken up the dextran conjugate migrate into the lumen of the colon in response to the inflammatory process, but the outermost tissue layers may be sufficiently damaged to allow dextran direct entry into inflamed tissues. Without this transmural inflammation and damage to the serosal side of the intestines, dextran would be unable to target intestinal inflammation following IP injection. This biological difference is an important point to consider when interpreting results in animal models, as IP administration may lead to greater intestinal uptake in models of transmural inflammation. Interestingly, like other reports [52, 53], we observed an opposing balance of radiolabel accumulation between the liver and the spleen after IP injection. Within DSS animals, 64Cu-NOTA-D500 retention was higher in the liver and lower in the spleen, whereas the opposite effect was seen in Con animals. While the reason for this opposing balance is unknown, differences in conjugate retention in the livers and spleens of Con and DSS mice may be dependent on differences in the biology of the diseased state. With the greatest concentrations of macrophages, the liver and spleen are the primary sites of nanodrug biodistribution [34]; however, higher dextran conjugate uptake by livers in DSS colitis animals compared with liver uptake in Con animals is consistent with greater liver inflammation observed to occur in IBD, in part by immune cell migration directly from the intestines to the liver [54]. This contrast modality could potentially make possible the identification of additional organs affected by inflammation and could be an important diagnostic target for some patients with IBD. Currently, most available research into nanoparticles for IBD are function-based, examining disease outcome as a marker for nanoparticle delivery [32, 55], and few studies are disease-based, examining the biodistribution of nanoparticles in control vs. diseased tissues [9, 56]. This study sought to examine the route-dependent, disease-based biodistribution of nanoparticles in a murine IBD model, ultimately providing evidence for researchers to apply in considering experimental design, internal normalization parameters, and applying the Reduction principle for animal numbers [57]. We showed that oral administration of a dextran conjugate may lead to differential uptake by colitis-affected tissues when normalizing based on total intestine or liver uptake. Both oral and enema delivery methods led to high inter-individual variability, but internal normalization significantly improved group differences for oral delivery, but not for enema delivery. In contrast, IP delivery led to much lower variability as well as greater absolute uptake of the probe, although this method should be tested in models of transmural inflammation and with lower molecular weight probes. ## Conclusions Oral and enema delivery of 64Cu-NOTA-D500 leads to higher uptake in colitis-affected tissues depending on tissue contrast (liver vs. intestines). Normalized intestinal uptake after oral delivery is the preferred method; however, IP delivery could be preferential for quantification standpoint if transmural inflammation is present, as is often seen in humans. Oral and IP delivery may additionally diagnose liver changes, affecting about $50\%$ of all individuals afflicted with IBD. Results here suggest that oral delivery with appropriate normalization may be the best route of administration for quantitative imaging in biologically variable IBD models, which should guide research design towards applying the Reduction principle of the 3 R’s of animal research (Replace, Reduce, Refine). ## Data availability statement The data that support the findings of this study are available from the corresponding author, KSS and AMS, upon reasonable request. ## References 1. Alatab S, Sepanlou SG, Ikuta K. **The global, regional, and national burden of inflammatory bowel disease in 195 countries and territories, 1990–2017: a systematic analysis for the global burden of disease study 2017**. *Lancet Gastroenterol Hepatol* (2020) **1** 17-30 2. Khor B, Gardet A, Xavier RJ. **Genetics and pathogenesis of inflammatory bowel disease**. *Nature* (2011) **7351** 307-317 3. Singh UP, Singh NP, Murphy EA. **Chemokine and cytokine levels in inflammatory bowel disease patients**. *Cytokine* (2016) **77** 44-49. PMID: 26520877 4. Műzes G, Molnár B, Tulassay Z. **Changes of the cytokine profile in inflammatory bowel diseases**. *World J Gastroenterol* (2012) **41** 5848-5861 5. Schiavone C, Romano M. **Diagnosis and management of Crohn’s disease**. *J Ultrasound* (2015) **1** 1-2 6. Cherry SR. **Fundamentals of positron emission tomography and applications in preclinical drug development**. *J Clin Pharmacol* (2001) **41** 482-491. PMID: 11361044 7. Kung BT, Seraj SM, Zadeh MZ. **An update on the role of (18)F-Fdg-Pet/Ct in major infectious and inflammatory diseases**. *Am J Nucl Med Mol Imaging* (2019) **6** 255-273 8. Basu S, Zhuang H, Torigian DA. **Functional imaging of inflammatory diseases using nuclear medicine techniques**. *Semin Nucl Med* (2009) **2** 124-145 9. Luo Y, Gao C, Chen W. **Molecular magnetic resonance imaging with contrast agents for assessment of inflammatory bowel disease: a systematic review**. *Contrast Media Mol Imaging* (2020) **2020** 10. Yu Y, Wu Y, Liu J. **Ultrasmall dopamine-coated nanogolds: preparation, characteristics, and CT imaging**. *J Exp Nanosci* (2016) **11** S1-s11. PMID: 27366201 11. Sharma R, Kwon S. **New applications of nanoparticles in cardiovascular imaging**. *J Exp Nanosci* (2007) **1–2** 115-126 12. Fedorak RN, Haeberlin B, Empey LR. **Colonic delivery of dexamethasone from a prodrug accelerates healing of colitis in rats without adrenal suppression**. *Gastroenterology* (1995) **6** 1688-1699 13. Huang Z, Gan J, Jia L. **An orally administrated nucleotide-delivery vehicle targeting colonic macrophages for the treatment of inflammatory bowel disease**. *Biomaterials* (2015) **48** 26-36. PMID: 25701029 14. McLeod AD, Friend DR, Tozer TN. **Synthesis and chemical stability of glucocorticoid-dextran esters: potential prodrugs for colon-specific delivery**. *Int J Pharm* (1993) **92** 105-114 15. Pang YN, Zhang Y, Zhang ZR. **Synthesis of an enzyme-dependent prodrug and evaluation of its potential for colon targeting**. *World J Gastroenterol* (2002) **5** 913-917 16. Varshosaz J, Emami J, Ahmadi F. **Preparation of budesonide-dextran conjugates using gluta-rate spacer as a colon-targeted drug delivery system: in vitro/in vivo evaluation in induced ulcerative colitis**. *J Drug Target* (2011) **2** 140-153 17. Wu Y, Briley-Saebo K, Xie J. **Inflammatory bowel disease: MR- and SPECT/CT-based macrophage imaging for monitoring and evaluating disease activity in experimental mouse model–pilot study**. *Radiology* (2014) **2** 400-407 18. Sharma A, Cornejo C, Mihalic J. **Physical characterization and in vivo organ distribution of coated iron oxide nanoparticles**. *Sci Rep* (2018) **8** 4916. PMID: 29559734 19. Shannahan J. **The biocorona: a challenge for the biomedical application of nanoparticles**. *Nanotechnol Rev* (2017) **6** 345-353. PMID: 29607287 20. Zhang-Hong H, Yiang L, Liang Z. **Preparation of a novel liver-targeting nanoparticle of norcantharidin derivative and evaluation of its anti-tumour activity**. *J Exp Nanosci* (2011) **2** 183-199 21. Ma L, Liu TW, Wallig MA. **Efficient targeting of adipose tissue macrophages in obesity with polysaccharide nanocarriers**. *ACS Nano* (2016) **7** 6952-6962 22. Zinderman CE, Landow L, Wise RP. **Anaphylactoid reactions to dextran 40 and 70: Reports to the United States food and drug administration, 1969 to 2004**. *J Vascular Surg* (2006) **5** 1004-1009 23. Chassaing B, Aitken JD, Malleshappa M. **Dextran sulfate sodium (DSS)-induced colitis in mice**. *Curr Protoc Immunol* (2014) **15** 11-15 24. Perse M, Cerar A. **Dextran sodium sulphate colitis mouse model: traps and tricks**. *J Biomed Biotechnol* (2012) **2012** 25. Kim JJ, Shajib MS, Manocha MM. **Investigating intestinal inflammation in DSS-induced model of IBD**. *J Vis Exp* (2012) **60** 3678 26. Pustylnikov S, Sagar D, Jain P. **Targeting the C-type lectins-mediated host-pathogen interactions with dextran**. *J Pharm Pharma Sci* (2014) **3** 371-392 27. Shimaoka T, Kume N, Minami M. **Molecular cloning of a novel scavenger receptor for oxidized low density lipoprotein, SR-PSOX, on macrophages**. *J Biol Chem* (2000) **52** 40663-40666 28. Fu J, Wei B, Wen T. **Loss of intestinal core 1-derived O-glycans causes spontaneous colitis in mice**. *J Clin Invest* (2011) **4** 1657-1666 29. Heazlewood CK, Cook MC, Eri R. **Aberrant mucin assembly in mice causes endoplasmic reticulum stress and spontaneous inflammation resembling ulcerative colitis**. *PLoS Med* (2008) **3** e54 30. Swidsinski A, Loening-Baucke V, Theissig F. **Comparative study of the intestinal mucus barrier in normal and inflamed colon**. *Gut* (2007) **3** 343-350 31. Olmsted SS, Padgett JL, Yudin AI. **Diffusion of macromolecules and virus-like particles in human cervical mucus**. *Biophys J* (2001) **4** 1930-1937 32. Lamprecht A, Ubrich N, Yamamoto H. **Biodegradable nanoparticles for targeted drug delivery in treatment of inflammatory bowel disease**. *J Pharmacol Exp Therap* (2001) **2** 775-781 33. Hovgaard L, Brøndsted H. **Dextran hydrogels for colon-specific drug delivery**. *J Controll Release* (1995) **1** 159-166 34. Weissleder R, Nahrendorf M, Pittet MJ. **Imaging macrophages with nanoparticles**. *Nat Mater* (2014) **13** 125-138. PMID: 24452356 35. Vavricka SR, Schoepfer A, Scharl M. **Extraintestinal manifestations of inflammatory bowel disease**. *Inflamm Bowel Dis* (2015) **8** 1982-1992 36. Kumar R, Roy I, Ohulchanskky TY. **In vivo biodistribution and clearance studies using multi-modal organically modified silica nanoparticles**. *ACS Nano* (2010) **2** 699-708 37. Martinez MN, Amidon GL. **A mechanistic approach to understanding the factors affecting drug absorption: a review of fundamentals**. *J Clin Pharmacol* (2002) **6** 620-643 38. Haute DV, Berlin JM. **Challenges in realizing selectivity for nanoparticle biodistribution and clearance: lessons from gold nanoparticles**. *Ther Deliv* (2017) **8** 763-774. PMID: 28825391 39. Melgar S, Karlsson A, Michaelsson E. **Acute colitis induced by dextran sulfate sodium progresses to chronicity in C57bl/6 but not in balb/C mice: correlation between symptoms and inflammation**. *Am J Physiol Gastrointest Liver Physiol* (2005) **6** G1328-1338 40. Coulie B, Camilleri M, Bharucha AE. **Colonic motility in chronic ulcerative proctosigmoiditis and the effects of nicotine on colonic motility in patients and healthy subjects**. *Aliment Pharmacol Ther* (2001) **5** 653-663 41. Larsen C, Jensen BH, Olesen HP. **Bioavailability of ketoprofen from orally administered ketoprofen-dextran ester prodrugs in the pig**. *Acta Pharma Nor* (1991) **2** 71-76 42. Osterman MT. **Mucosal healing in inflammatory bowel disease**. *J Clin Gastroenterol* (2013) **47** 212-221. PMID: 23340060 43. Kilcoyne A, Kaplan JL, Gee MS. **Inflammatory bowel disease imaging: current practice and future directions**. *World J Gastroenterol* (2016) **3** 917-932 44. Nugent SG, Kumar D, Rampton DS. **Intestinal luminal pH in inflammatory bowel disease: possible determinants and implications for therapy with aminosalicylates and other drugs**. *Gut* (2001) **4** 571-577 45. Ko HS, Kim IS, Lee K. **Characterization of binding and phagocytosis of oxidatively damaged erythrocyte to macrophage**. *Korean J Intern Med* (2002) **4** 220-226 46. Mullin NP, Hall KT, Taylor ME. **Characterization of ligand binding to a carbohydrate-recognition domain of the macrophage mannose receptor**. *J Biol Chem* (1994) **45** 28405-28413 47. Simpson DZ, Hitchen PG, Elmhirst EL. **Multiple interactions between pituitary hormones and the mannose receptor**. *Biochem J* (1999) **343** 403-411. PMID: 10510307 48. Gazi U, Martinez-Pomares L. **Influence of the mannose receptor in host immune responses**. *Immunobiology* (2009) **7** 554-561 49. Hu Z, Shi X, Yu B. **Structural insights into the pH-dependent conformational change and collagen recognition of the human mannose receptor**. *Structure* (2018) **1** 60-71.e63 50. Goswami S, Bajpai J, Bajpai AK. **Calcium alginate nanocarriers as possible vehicles for oral delivery of insulin**. *J Exp Nanosci* (2014) **4** 337-356 51. Nishida K, Kuma A, Fumoto S. **Absorption characteristics of model compounds from the small intestinal serosal surface and a comparison with other organ surfaces**. *J Pharm Pharmacol* (2005) **8** 1073-1077 52. Zhang Y, Liu AT, Cornejo YR. **A systematic comparison of in vitro cell uptake and in vivo biodistribution for three classes of gold nanoparticles with saturated peg coatings**. *PLoS One* (2020) **7** e0234916 53. Elci SG, Jiang Y, Yan B. **Surface charge controls the suborgan biodistributions of gold nanoparticles**. *ACS Nano* (2016) **5** 5536-5542 54. Mathies F, Steffens N, Kleinschmidt D. **Colitis promotes a pathological condition of the liver in the absence of Foxp3(+) regulatory T cells**. *J Immunol* (2018) **12** 3558-3568 55. Lee A, De Mei C, Fereira M. **Dexamethasone-loaded polymeric nanoconstructs for monitoring and treating inflammatory bowel disease**. *Theranostics* (2017) **7** 3653-3666. PMID: 29109767 56. Wu Y, Briley K, Tao X. **Nanoparticle-based imaging of inflammatory bowel disease**. *Wiley Interdiscipl Rev Nanomed Nanobiotechnol* (2016) **2** 300-315 57. Flecknell P. **Replacement, reduction and refinement**. *Altex* (2002) **2** 73-78
--- title: The genome of Tripterygium wilfordii and characterization of the celastrol biosynthesis pathway authors: - Tianlin Pei - Mengxiao Yan - Yu Kong - Hang Fan - Jie Liu - Mengying Cui - Yumin Fang - Binjie Ge - Jun Yang - Qing Zhao journal: GigaByte year: 2021 pmcid: PMC10038137 doi: 10.46471/gigabyte.14 license: CC BY 4.0 --- # The genome of Tripterygium wilfordii and characterization of the celastrol biosynthesis pathway ## Abstract Tripterygium wilfordii is a vine from the Celastraceae family that is used in traditional Chinese medicine (TCM). The active ingredient, celastrol, is a friedelane-type pentacyclic triterpenoid with putative roles as an antitumor, immunosuppressive, and anti-obesity agent. Here, we report a reference genome assembly of T. wilfordii with high-quality annotation using a hybrid sequencing strategy. The total genome size obtained is 340.12 Mb, with a contig N50 value of 3.09 Mb. We successfully anchored $91.02\%$ of sequences into 23 pseudochromosomes using high-throughput chromosome conformation capture (Hi–C) technology. The super-scaffold N50 value was 13.03 Mb. We also annotated 31,593 structural genes, with a repeat percentage of $44.31\%$. These data demonstrate that T. wilfordii diverged from Malpighiales species approximately 102.4 million years ago. By integrating genome, transcriptome and metabolite analyses, as well as in vivo and in vitro enzyme assays of two cytochrome P450 (CYP450) genes, TwCYP712K1 and TwCYP712K2, it is possible to investigate the second biosynthesis step of celastrol and demonstrate that this was derived from a common ancestor. These data provide insights and resources for further investigation of pathways related to celastrol, and valuable information to aid the conservation of resources, as well as understand the evolution of Celastrales. ## Introduction Tripterygium wilfordii Hook. f. (NCBI: txid458696) is a perennial twining shrub belonging to the Celastraceae family. It is known in China as ‘Lei gong teng’ (meaning: Thunder God Vine). It is indigenous to Southeast China, the Korean Peninsula, and Japan, and has been cultivated worldwide as a medicinal plant [1, 2] (Figure 1). The extract of T. wilfordii bark has been used as a pesticide in China since ancient times, and was first recorded in the Illustrated Catalogues of Plants published in 1848 [3]. The potential medicinal activity of T. wilfordii has been studied since the 1960s, with its root being used to alleviate the symptoms of leprosy patients in Gutian County, Fujian Province, China [4]. This application ignited the interest of researchers in various fields. T. wilfordii was then reported to be effective in the treatment of autoimmune diseases, such as rheumatoid arthritis and systemic psoriasis [5, 6]. In recent decades, many studies have examined the potential anticancer, antidiabetic and anti-inflammatory effects of extracts of T. wilfordii [7–9]. Investigations into the pharmacological activities of T. wilfordii have mainly focused on the various compounds accumulating in its root, such as alkaloids, diterpenoids and triterpenoids [10, 11]. Celastrol is a friedelane-type triterpenoid that is mainly found in the root bark of T. wilfordii [12]. In Chinese medicine, it has been used for the treatment of inflammatory and autoimmune diseases [13], tumors [14], and as a possible treatment for Alzheimer’s disease [15]. Celastrol is also a leptin sensitizer and may be useful in the treatment of obesity [16, 17]. Despite the commercial importance of natural products found in T. wilfordii and the growing demand for these products, traditional methods of production are becoming unsustainable owing to the slow growth rate of the vines and low accumulation of celastrol [18]. There is therefore a need for novel production methods, such as synthetic biological methods. Genome sequencing will provide a reference for mining the genes involved in the pathways of these bioactive compounds. Celastrol is a pentacyclic triterpenoid synthesized from 2,3-oxidosqualene, the common biosynthetic precursor of triterpenoids derived from the cytosolic mevalonate (MVA) and plastid 2-C-methyl-D-erythritol-4-phosphate (MEP) pathways [19, 20]. Two oxidosqualene cyclases (OSCs), namely, TwOSC1 and TwOSC3, were identified as key enzymes in the cyclization of 2,3-oxidosqualene to form friedelin, the first step in celastrol formation [21]. The next step in this pathway is thought to be hydroxylation of the C-29 position of friedelin to produce 29-hydroxy-friedelin-3-one. This is then converted, via carboxylation, to polpunonic acid, which in turn undergoes a series of oxidation reactions and rearrangements to produce celastrol [21]. Here, we report the reference genome assembly of T. wilfordii using a combined sequencing strategy. After integrating genome, transcriptome and metabolite analyses, several novel cytochrome P450 (CYPs) proteins related to celastrol biosynthesis were identified. TwCYP712K1 and TwCYP712K2 were then functionally characterized using in vivo yeast and in vitro enzyme assays. These data represent a strategy to reveal the evolution of Celastrales and the key genes involved in celastrol biosynthesis. **Figure 1.:** *Picture of Tripterygium wilfordii. With thanks to Dr. Bin Chen from the Shanghai Chenshan Herbarium for providing the image.* ## Methods A protocol collection including methods for DNA-extraction, Hi–C assembly and optical mapping is available via protocols.io (Figure 2) [22]. **Figure 2.:** *Protocol collection for the genome analysis of Tripterygium wilfordii. https://www.protocols.io/widgets/doi?uri=dx.doi.org/10.17504/protocols.io.bspfndjn* ## Plant materials Tripterygium wilfordii plants were collected from the experimental fields of Shanghai Chenshan Botanical Garden (31° 04 ′ 30.00 ′′ N, 121° 10 ′ 58.93 ′′ E) and cultured in a greenhouse by cutting propagation. All materials used for genome sequencing originated from a single plant (grown in the greenhouse of our laboratory, voucher TW1). For RNA sequencing (RNA-seq), tissues from roots (R), stems (S), young leaves (YL), mature leaves (L), flower buds (FB) and flowers (F) were harvested with three independent biological replicates. ## DNA sequencing Total DNA was isolated from leaves using the modified cetyltrimethylammonium bromide (CTAB) method [23]. DNA purity was checked by electrophoretic analysis on a $1\%$ agarose gel and using a NanoPhotometer spectrophotometer (IMPLEN, CA, USA). The DNA concentration was determined using a Qubit 2.0 fluorometer (Life Technologies, CA, USA). For Illumina sequencing, qualified DNA was fragmented using a Covaris device (MA, USA). Fragmented DNA was end-repaired; poly(A) tail and adaptor addition was performed using the Next Ultra DNA Library Prep Kit (NEB, MA, USA), then the appropriate samples were selected by electrophoretic analysis. The size-selected product was PCR-amplified, and the final product was purified and validated using AMPure XP beads (Beckman Coulter, CA, USA) and an Agilent Bioanalyzer 2100. Using the HiSeq 2500 platform, 150-bp paired-were sequenced. Clean data were obtained by removing adaptor reads, unidentified nucleotides (N) and low-quality reads from the raw reads, and Q20, Q30 and GC content of the clean data were calculated for quality assessment (Table 1). We estimated the genome size by performing k-mer frequency analysis. The k-mer frequencies (k-mer size = 17) were obtained using Jellyfish v2.2.7 [24] with jellyfish count -G 2 -s 5G -m 17 and jellyfish stats as the default parameters. For long-read sequencing, qualified DNA was sheared into fragments in a g-TUBE (Covaris, MA, USA) by centrifugation, and quantity and quality were controlled by an Agilent Bioanalyzer 2100. To construct a sequencing library, the fragmented DNA was end-repaired and poly(A) tail and adaptor addition was performed using Next Ultra II End Repair/dA-Tailing Module, Next FFPE DNA Repair Mix and Next Quick Ligation Module (NEB, MA, USA), respectively, according to the manufacturer’s instructions. The final product was validated using an Agilent Bioanalyzer 2100. Finally, the qualified DNA library was sequenced using Oxford Nanopore Technology (ONT) on the PromethION platform. **Table 1** | Raw paired reads | Raw Base (bp) | Effective Rate (%) | Error Rate (%) | Q20 (%) | Q30 (%) | GC Content (%) | | --- | --- | --- | --- | --- | --- | --- | | 84395810.0 | 25318743000.0 | 99.69 | 0.05 | 95.44 | 88.88 | 38.22 | ## Genome assembly De novo genome assembly was carried out using NextDenovo v2.3.0 [25]. The correct_option parameters used were: read_cutoff = 1k, seed_cutoff = 28087, pa_correction = 20, seed_cutfiles = 100, sort_options = -m 15g -t 8 -k 40, minimap2_options_raw = -x ava-ont -t 8. The assemble_option parameters used were: random_round = 20, minimap2_options_cns = -x ava-ont -t 8 -k17 -w17, nextgraph_options = -a 1. Racon v1.3.1 [26] and Pilon v1.22 (Pilon, RRID:SCR_014731) [27] were used for error correction with ONT data and Illumina data, respectively. Error correction was performed three times with default parameters. The completeness of the genome assembly was assessed using Benchmarking Universal Single-Copy Orthologs (BUSCO) v3.0.2 (BUSCO, RRID:SCR_015008) [28] with the parameters: -m genome -c 15 -sp arabidopsis. Assembly accuracy was evaluated using Burrows–Wheeler Aligner (BWA) software (version: 0.7.8-r455) [29] to align Illumina reads back to the genome. Variant calling was performed using SAMtools (version: 0.1.19-44428cd, SAMTOOLS, RRID:SCR_002105) [30, 31] with parameters: -m 2 -F 0.002 -d 110 -u -f. To assess the genome assembly quality, transcriptome data were assembled using Trinity (Trinity, RRID:SCR_013048) [32], then mapped back to the scaffolds using BLAT (BLAT, RRID:SCR_011919) [33]. For Bionano sequencing, genomic DNA (molecules > 300 kb) of leaves from living T. wilfordii plants was extracted using the Plant DNA Isolation Kit (Bionano Genomics, CA, USA). Using the NLRS DNA Labeling Kit (Bionano Genomics, CA, USA), DNA molecules were digested with Nt. BspQI endonucleases (determined after evaluation by electronic digestion), and fluorescently labeled. Labeled DNA molecules were electrophoretically stretched into linearization by Saphyr Chip (Bionano Genomics, CA, USA), passed through the NanoChannels [34], and then captured on the Saphyr platform with a high-resolution camera. Raw image data were first converted to digital representations of the motif-specific label pattern, then analyzed using Bionano Solve v3.1 [35] and its in-house scripts. Bionano data were compared with the draft genome (Nanopore version) with the parameters: -U, -d, -T, 3, -j, 3, -N, 20, -I, 3, and scaffolds were generated by connecting contigs with the parameters: -f, -B, 1, -N, 1. ## Sequence anchoring Hi–C library preparation and sequencing were based on a protocol described previously, with some modifications [36–38]. Leaves from living T. wilfordii plants were treated with $1\%$ formaldehyde solution to fix chromatin. Approximately 2 g of fixed tissue was homogenized with liquid nitrogen, resuspended in nucleus isolation buffer and filtered with a 40-nm cell strainer. Extracted chromatin was cut with the HindIII restriction enzyme (NEB, MA, USA), end-filled, then labeled with biotin. After ligation with T4 DNA ligase (NEB, CA, USA) and reversal of crosslinking by proteinase K, DNA was purified, cleaved into 350-bp fragments and end-repaired. DNA fragments labeled with biotin were separated using Dynabeads M-280 Streptavidin (Thermo Fisher, MA, USA), purified, and end-repaired. A-tails were added and adaptors were ligated, and the sequences were amplified by PCR to generate Hi–C libraries. Finally, the qualified libraries were sequenced on an Illumina platform. Clean data were obtained by removing adaptor reads, unidentified nucleotides (N) and low-quality reads from the raw reads, and Q20, Q30 and GC content of the clean data were calculated for quality assessment (Table 2). Clean data were first mapped to the draft genome using BWA software (version: 0.7.8-r455) [29]. After removal of PCR duplicates and unmapped reads using SAMtools v1.9 [39], based on the numbers of interacting read pairs, contigs were clustered and ordered into chromosome groups using LACHESIS (version 201701) [40] with the parameters: RE_SITE_SEQ = GATC, CLUSTER_N = 23, CLUSTER_CONTIGS_WITH_CENS = −1, CLUSTER_MIN_RE_SITES = 388, CLUSTER_MAX_LINK_DENSITY = 3, CLUSTER_NONINFORMATIVE_RATIO = 0, CLUSTER_DRAW_HEATMAP = 1, and CLUSTER_DRAW_DOTPLOT = 1. **Table 2** | Raw bases (bp) | Clean bases (bp) | Effective Rate(%) | Error rate | Q20 (%) | Q30 (%) | GC content (%) | | --- | --- | --- | --- | --- | --- | --- | | 78004410900.0 | 77678361000.0 | 99.58 | 0.04 | 96.66 | 91.13 | 39.83 | ## Transcriptome sequencing Total RNA was extracted from collected tissues using the RNAprep Pure Plant Kit (TIANGEN, Beijing, China). Qualified RNA from each sample was used to generate sequencing libraries using the NEBNext Ultra RNA Library Prep Kit for Illumina sequencing (NEB, MA, USA), following the manufacturer’s instructions. mRNA was purified from total RNA using poly(T) oligo-attached magnetic beads, then cleaved into short fragments that were used as templates for cDNA synthesis. After purification, repair, adenylation and adaptor ligation of the 3 ′ end, 150 to 200-bp cDNA fragments were separated for PCR amplification. Finally, libraries were quality controlled using an Agilent 2100 Bioanalyzer and qPCR, then sequenced on the Illumina HiSeq 2500 platform. Raw reads with adapter, poly(N) and low-quality reads were removed to generate clean data, and Q20, Q30 and GC content of the clean data were calculated for quality assessment (Table 3). RNA-seq data was mapped back to the genome assembly of T. wilfordii using HISAT v2.0.4 with default parameters [41]. The read numbers of each gene were counted using HTSeq v0.6.1 with the parameter: –m union [42]. Fragments per kilobase of transcript per million fragments mapped (FPKM) of each gene was calculated based on the length of the gene and number of read counts mapped to this gene [43]. For full-length transcriptome sequencing by PacBio, the best quality RNA samples of each tissue were mixed together to build an isoform sequencing library using the Clontech SMARTer PCR cDNA Synthesis Kit and the BluePippin Size Selection System protocol, as described by Pacific Biosciences (PN 100-092-800-03). Samples were then sequenced on the PacBio Sequel platform. Sequence data were processed using SMRTlink 7.0 software [44] with the parameters: –minLength 50, –maxLength 15000, –minPasses 1. Error correction was achieved using the Illumina RNA-seq data with LoRDEC v0.7, with the parameters: -k 23, -s 3 [45]. Redundancy in the corrected consensus reads was removed by CD-HIT v4.6.8 [46], with the parameters: -c 0.95, -T 6, -G 0, -aL 0.00, -aS 0.99, -AS 30 to obtain the final transcripts for the subsequent analysis. **Table 3** | Sample name | Raw reads (bp) | Clean reads (bp) | Clean bases | Error rate (%) | Q20 (%) | Q30 (%) | GC content (%) | | --- | --- | --- | --- | --- | --- | --- | --- | | R1 | 68286484 | 67508000 | 10.13G | 0.03 | 98.0 | 94.11 | 45.48 | | R2 | 58613186 | 57798058 | 8.67G | 0.03 | 98.0 | 94.09 | 46.14 | | R3 | 63088548 | 61991218 | 9.3G | 0.02 | 98.05 | 94.22 | 45.28 | | YL1 | 61340056 | 60217856 | 9.03G | 0.03 | 98.02 | 94.11 | 45.98 | | YL2 | 41099134 | 40542280 | 6.08G | 0.02 | 98.28 | 94.94 | 45.82 | | YL3 | 69124902 | 68072478 | 10.21G | 0.02 | 98.11 | 94.35 | 45.6 | | L1 | 66626224 | 65459234 | 9.82G | 0.02 | 98.12 | 94.34 | 45.39 | | L2 | 58487674 | 57306282 | 8.6G | 0.02 | 98.3 | 94.76 | 45.35 | | L3 | 70363762 | 69393172 | 10.41G | 0.02 | 98.14 | 94.41 | 45.88 | | S1 | 55910376 | 55176316 | 8.28G | 0.02 | 98.1 | 94.27 | 44.96 | | S2 | 67911592 | 66929018 | 10.04G | 0.03 | 98.02 | 94.11 | 44.93 | | S3 | 59810518 | 58957160 | 8.84G | 0.03 | 97.97 | 93.94 | 44.95 | | FB1 | 64749004 | 63837752 | 9.58G | 0.03 | 98.01 | 94.06 | 45.3 | | FB2 | 53079594 | 52523514 | 7.88G | 0.03 | 98.01 | 94.14 | 45.77 | | FB3 | 53084804 | 52578262 | 7.89G | 0.02 | 98.08 | 94.3 | 45.76 | | F1 | 63007696 | 62341814 | 9.35G | 0.02 | 98.14 | 94.38 | 44.91 | | F2 | 60886358 | 60100754 | 9.02G | 0.02 | 98.13 | 94.36 | 45.68 | | F3 | 59719778 | 59080108 | 8.86G | 0.03 | 97.97 | 93.99 | 44.95 | ## Genome annotation Homolog alignment and de novo prediction were applied for repeat annotation. For homolog alignment, the Repbase database employing RepeatMasker software v4.0.7 (RepeatMasker, RRID:SCR_012954) and its in-house scripts (RepeatProteinMask v4.0.7) was used with default parameters to extract repeat sequences [47]. For de novo prediction, LTR_FINDER v1.0.7 (LTR_Finder, RRID:SCR_015247) [48], RepeatScout v1.0.5 (RepeatScout, RRID:SCR_014653) [49], and RepeatModeler v1.0.3 (RepeatModeler, RRID:SCR_015027) [50] were used with default parameters to build a de novo repetitive element database for repeat identification. Tandem repeats were also extracted by de novo prediction using TRF v4.0.9 [51]. A combined strategy based on homology, gene prediction, RNA-seq and PacBio data was used to annotate gene structure. For homolog prediction, sequences of proteins from six species, including Arabidopsis thaliana, Vitis vinifera, Medicago truncatula, Cucumis sativus, Ricinus communis, and Glycyrrhiza uralensis, were downloaded from Ensembl/National Center for Biotechnology Information (NCBI)/DNA Database of Japan (DDBJ). Protein sequences were aligned to the genome using TblastN v2.2.26 [52] (E-value ≤ 1 × 10−5), then the matching proteins were aligned to homologous genome sequences for accurate spliced alignments with GeneWise v2.4.1 (GeneWise, RRID:SCR_015054) [53]. De novo gene structure identification was based on Augustus v3.2.3 (Augustus, RRID:SCR_008417) [54], GlimmerHMM v3.04 (GlimmerHMM, RRID:SCR_002654) [55], and SNAP [2013-11-29] [56]. Based on the above prediction results, RNA-seq reads from different tissues, and PacBio reads, were aligned to the genome using HISAT v2.0.4 (HiSat2, RRID:SCR_015530) [41] and TopHat v2.0.12 (TopHat, RRID:SCR_013035) [57] with default parameters to identify exon regions and splice positions. Alignment results were then used as input for Stringtie v1.3.3 (StringTie, RRID:SCR_016323) [58] with default parameters for genome-based transcript assembly. Alignment results were then integrated into a nonredundant gene set using EVidenceModeler v1.1.1 and further corrected with Program to Assemble Spliced Alignment (PASA) to predict untranslated regions and alternative splicing to generate the final gene set [59]. According to the final gene set, gene function was predicted by aligning the protein sequences to Swiss-Prot [60] and the Non-Redundant Protein Sequence Database (NR) (version 20190709) [61]. The motifs and domains were annotated using InterProScan70 v5.31 by searching against the Protein Families Database (Pfam) [62], Kyoto Encyclopedia of Genes and Genomes (KEGG, version 20190601) [63], and Integrative Protein Signature Database (InterPro) v32.0 [64] using Blastp (E-value ≤ 1 × 10−5). Gene Ontology (GO) IDs for each gene were assigned according to the corresponding InterPro entry. Noncoding RNA was annotated using tRNAscan-SE v1.4 (for tRNA) [65] or INFERNAL v1.1.2 with default parameters (for miRNA and snRNA) [66]. rRNA was predicted by BLAST using rRNA sequences from A. thaliana and O. sativa as references, which are highly conserved among plants. ## Comparative genome analyses Gene family clustering of 12 species, including T. wilfordii, A. thaliana, Citrus sinensis, V. vinifera, Glycine max, M. truncatula, G. uralensis, C. sativus, Populus trichocarpa, R. communis, Oryza sativa, and Amborella trichopoda, was inferred through all-against-all protein sequence similarity searches using OthoMCL v1.4 [67], with the parameters: -mode 3 and -inflation 1.5. Proteins containing fewer than 50 amino acids were removed, and only the longest predicted transcript per locus was retained. Single-copy orthologous genes were retrieved from the 12 species and aligned using MUSCLE v3.8.31 with default parameters [68]. All alignments were combined to produce a super-alignment matrix, which was used to construct a maximum likelihood (ML) phylogenetic tree using RAxML v8.2.12 [69] with the parameters: cds: -m GTRGAMMA -p 12345 -x 12345 -#100 -f ad -T 20, pep: -m PROTGAMMAAUTO -p 12345 -x 12345 -#100 -f ad -T 20. Divergence times between species were calculated using the MCMCtree v4.9 program implemented for phylogenetic analysis by maximum likelihood (PAML) with the default parameters [70]. The following calibration points were applied: M. truncatula–G. uralensis (15–91 million years ago, Mya), G. max–M. truncatula (46–109 Mya), G. max–C. sativus (95–135 Mya), A. thaliana–C. sinensis (96–104 Mya), P. trichocarpa–R. communis (70–86 Mya), A. thaliana–P. trichocarpa (98–117 Mya), C. sativus–R. communis (101–131 Mya), V. vinifera–A. thaliana (107–135 Mya), V. vinifera–O. sativa (115–308 Mya), and O. sativa–A. trichopoda (173–199 Mya). These calibrations were extracted from TimeTree [71]. Expansion and contraction of gene families were analyzed by using CAFÉ v4.2 [72] with the parameters: -p 0.05 -t 4 -r 10000. To avoid false positives, results were filtered and the enrichment results screened with a family-wide P-value < 0.05 and Viterbi P-values < 0.05. ## Genome-wide identification of CYP genes The hidden Markov model (HMM) profile of Pfam PF06200 [62] was used to extract full-length CYP candidates from the T. wilfordii genome by the HMM algorithm (HMMER) [71], filtering by a length between 400 and 600 amino acids [74]. ## Phylogenetic analyses Multiple sequence alignments and phylogenetic tree construction were performed using MEGA X [75], with either the neighbor-joining or ML method with a bootstrap test ($$n = 1000$$ replications). ## Co-expression analysis Gene expression pattern analysis was performed using Short Time-series Expression Miner (STEM) software [76] on the OmicShare tools platform [77]. The parameters were set as follows: the maximum unit change in model profiles between time points was 1; the maximum output profile number was 20 (similar profiles were merged); the minimum ratio of fold change of differentially expressed genes (DEGs) was no less than 2.0, and the P-value was <0.05. ## Gene cloning The complete open reading frames (ORFs) of the putative CYP genes were amplified using the primers listed in Table 4, with cDNA from T. wilfordii root used as the template. According to the manufacturer’s instructions, fragments were cloned into the entry vector pDONR207 and yeast expression vector pYesdest52 using the Gateway BP Clonase II Enzyme Kit and LR Clonase II Enzyme Kit (Invitrogen, MA, USA), respectively. **Table 4** | Primer names | Squence (5 ′ to 3 ′ )∗ | | --- | --- | | TwCYP712K1-F | GGGGACAAGTTTGTACAAAAAAGCAGGCTTCATGGCCACCATCACTGACATC | | TwCYP712K1-R | GGGGACCACTTTGTACAAGAAAGCTGGGTTTTAACCGGCAAATGGATTGAA | | TwCYP712K2-F | GGGGACAAGTTTGTACAAAAAAGCAGGCTTCATGACAACAATCACTGATGTGAA | | TwCYP712K2-R | GGGGACCACTTTGTACAAGAAAGCTGGGTTTTAAGAAGAAAATGGATTGAACC | | TwCYP712K3-F | GGGGACAAGTTTGTACAAAAAAGCAGGCTTCATGGCCACCACTACCATCATT | | TwCYP712K3-F | GGGGACCACTTTGTACAAGAAAGCTGGGTTTTAGCAAGAAAAGGGATGGAATC | ## Standard compounds Friedelin, 29-hydroxy-friedelan-3-one, and celastrol were purchased from Yuanye-Biotech (Shanghai, China), and polpunonic acid and wilforic acid A were purchased from Weikeqi-Biotech (Sichuan, China). Friedelin was dissolved in dimethyl sulfoxide (DMSO)/isopropanol (v/$v = 1$:2) following 30 min of ultrasonication in a water bath, while 29-hydroxy-friedelan-3-one, celastrol, polpunonic acid and wilforic acid A were dissolved in methanol. ## Metabolite analysis Plant tissue was ground into powder in liquid nitrogen then freeze dried. Fifty milligrams of sample was suspended in 2 mL of $80\%$ (v/v) methanol, set overnight at room temperature, then extracted in an ultrasonic water bath for 60 min. After centrifugation at 12,000g for 2 min, the supernatant was filtered through a 0.2-μm Millipore filter before liquid chromatography–mass spectrometry (LC–MS) analysis. Levels of celastrol and wilforic acid A were analyzed using an Agilent 1260LC-6400 QQQ (triple quadrupole mass spectrometer). Chromatographic separation was carried out on an Agilent Eclipse XDB-C18 analytical column (4.6 × 250 mm, 5 μm) with a guard column. The flow rate of the mobile phase consisting of $0.1\%$ (v/v) formic acid in water (A) and acetonitrile (B) was set to 0.8 mL/min. The gradient program was as follows: 0–12 min, 10–$60\%$ B; 12–17 min, $70\%$ B; 17–25 min, $95\%$ B; 25–28 min, $95\%$ B; 28–29 min, $5\%$ B; 29–35 min, $5\%$ B. The detection wavelength of celastrol was 425 nm, and UV spectra from 190–500 nm were also recorded. The injection volume was 10 μl and the column temperature was 35 °C. The liquid chromatography (LC) effluent was introduced into the electrospray ionization (ESI) source by a split-flow valve with a ratio of 3:1. All mass spectra were acquired in negative ion mode, and the parameters were as follows: drying gas 4 L/min; drying gas temperature 300 °C; nebulizer (high-purity nitrogen) pressure 15 psi; capillary voltage 4.0 kV; fragmentor voltage 135 V; and cell accelerator voltage 7 V. For full-scan mass spectrometry (MS) analysis, the spectra were recorded in the m/z range of 100–750. Levels of 29-hydroxy-friedelan-3-one and polpunonic acid were analyzed using Thermo Q Exactive Plus. Chromatographic separation was carried out on a Thermo Syncronis C18 column (2.1 × 100 mm, 1.7 μm). The flow rate of the mobile phase, consisting of $0.1\%$ (v/v) formic acid in water (A) and acetonitrile (B), was set to 0.4 mL/min. The gradient program was as follows: 0–12 min, 10–$60\%$ B; 12–17 min, $70\%$ B; 17–25 min, $95\%$ B; 25–28 min, $95\%$ B; 28–29 min, $5\%$ B. Mass spectra were acquired in both positive and negative ion modes with a heated ESI source, and the parameters were as follows: aus. gas flow 10 L/min; aus. gas heater 350 °C; sheath gas flow 40 L/min; spray voltage 3.5 kV; capillary temperature 320 °C. For full-scan MS/data-dependent (ddMS2) analysis, spectra were recorded in the m/z range of 50–750 at a resolution of 17,500 with automatic gain control (AGC) targets of 1 × 106 and 2 × 105, respectively. Levels of metabolites in different tissues were measured by comparing the area of the individual peaks with standard curves obtained from standard compounds. ## Enzyme assays of yeast in vivo Yeast in vivo assays were performed following a previously described protocol with some modifications [78]. The yeast expression vector constructs or empty vector were transformed into the yeast *Saccharomyces cerevisiae* WAT11 [79, 80] using the Yeast Transformation II Kit (ZYMO, CA, USA), and screened on synthetic-dropout (SD) medium lacking uracil (SD-Ura) with 20 g/L glucose. After growing at 28 °C for 48–72 h, transformant colonies were initially grown in 20 ml of SD-Ura liquid medium with 20 g/L glucose at 28 °C for approximately 24 h until the OD600 reached 2–3. Yeast cells were harvested by centrifugation at 4000g and resuspended in 20 mL of SD-Ura liquid medium supplemented with 20 g/L galactose to induce target proteins, while friedelin or 29-hydroxy-friedelane-3-one was applied to the cultures at a final concentration of 25 mM. After 48 h of fermentation (supplemented with 2 mL galactose after 24 h), yeast cells were harvested by centrifugation and extracted with 2 mL of $70\%$ methanol in an ultrasonic water bath for 2 h. The supernatants were filtered with a 0.2-μm Millipore filter and analyzed by LC–MS. ## Enzyme assays in vitro The protocol for enzyme assays in vitro was performed as described previously with some modifications [81]. Yeast transformation and target protein induction were performed as described above, except for 24 h of fermentation after galactose supplementation. Yeast cells were harvested by centrifugation and suspended in a 10-mL mixture of 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 0.6 M sorbitol and ddH2O. High pressure cell disruption equipment (Constant Systems, Northants, UK) was used to crush the yeast cells. After centrifugation, approximately 10 mL of supernatant was collected, and CaCl2 was applied at a final concentration of 18 mM. Microsomal proteins were then collected by centrifugation and suspended in storage buffer containing 50 mM Tris-HCl (pH 7.5), 1 mM EDTA and $20\%$ (v/v) glycerol with a final concentration of 10–15 mg/mL determined by the Bradford method [82]. The catalytic activity of putative CYP was assayed in a 100-μl reaction volume, which contained 100 mM sodium phosphate buffer (pH 7.9), 0.5 mM reduced glutathione, 2.5 μg of extracted protein and 100 μM substrate (friedelin or 29-hydroxy-friedelan-3-one). The reaction was initiated by adding NADPH at 1 mM and incubating for 12 h at 28 °C. Methanol was then added at a final concentration of $70\%$ to quench the reaction. The reaction mixture was filtered with a 0.2-μm Millipore filter and analyzed by LC–MS. Microsomal proteins extracted from yeast harboring the empty vector were used as a negative control. ## Syntenic analyses The genomes of T. wilfordii, O. sativa japonica and V. vinifera were compared using MCScan Toolkit v1.1 [83] implemented in Python. The genomes of O. sativa v32 and V. vinifera v32 were downloaded from Ensembl Plants [84]. *Syntenic* gene pairs were identified using an all-vs-all BLAST search using LAST [85], filtered to remove pairs with scores below 0.7, and clustered into syntenic blocks in MCScan. Microsynteny plots were constructed using MCScan. ## Results ## Genome sequencing, assembly, and annotation We obtained 77.86 Gb of Nanopore reads, amounting to 207.16× coverage of the 375.84-Mb genome, a size estimated by k-mer distribution analysis (Figure 3 and Table 5). The draft genome was assembled to obtain primary contigs, with a total size of 340.12 Mb and contig N50 of 3.09 Mb (Table 6). The GC content of the genome was $37.19\%$, with $0.00\%$ N (Table 7). Variant calling showed a heterozygosity rate of $0.25\%$ (Table 8). BUSCO analysis showed $95.2\%$ complete single-copy genes (Table 9). Short reads obtained from Illumina sequencing in the genome survey were aligned to the genome (Table 5), which exhibited a high consistency with a $95.31\%$ mapping rate and $93.99\%$ coverage (Table 10). In addition, $87.85\%$ of expressed sequence tags (ESTs) could be identified in the assembly, indicating high coverage of the genome (Table 11). For assembly improvement, 60.80 Gb (161.77× of estimated genome size) of reads from Bionano sequencing were obtained and integrated with the draft genome to construct scaffolds, which updated the genome of T. wilfordii from a total contig length of 340.12 Mb and a contig N50 of 3.09 Mb to a total scaffold length of 342.59 Mb and a scaffold N50 of 5.43 Mb (Tables 12 and 5). Furthermore, we anchored $91.02\%$ of the original 342.61-Mb assembly into 23 groups using Hi–C technology (Tables 13 and 14). All the super-scaffold was able to be placed in one of 23 groups (Figure 4). The super-scaffold N50 reached 13.03 Mb, with the longest super-scaffold being 17.75 Mb in size (Tables 13 and 14). The number of groups, hereafter referred to as pseudochromosomes, corresponded well to the number of chromosomes reported previously [86]. For genome annotation and gene expression profile analyses, roots, stems, young leaves, mature leaves, flower buds and flowers of T. wilfordii plants were collected prior to RNA-seq using the Illumina platform. Furthermore, RNA samples from different tissues were mixed, then sequenced using the PacBio platform to obtain full-length transcriptome sequences (Table 5). A combined strategy involving de novo prediction, homology prediction, RNA-seq and PacBio read alignment was used to construct the gene structure for the T. wilfordii genome. The final set of annotated genes amounted to 31,593 genes, with an average length of 3180 bp and an average coding sequence length of 1182 bp (Table 15). A total of 27,301 genes ($86.41\%$) were supported by RNA-seq data and 23,229 genes ($73.53\%$) were supported by all the methods used; these genes were annotated with high confidence. Gene function annotation was performed by BLAST analysis of the protein sequences of predicted genes against public databases, including NR, Swiss-Prot, KEGG, GO, Pfam and InterPro. A total of 30,535 ($96.70\%$) gene products could be functionally predicted, and 22,491 sequences could be annotated by at least one of the databases (NR, SwissProt, InterPro and KEGG) (Table 16). Repeat sequence annotation showed that the T. wilfordii genome contained $44.31\%$ repetitive sequences. Among these sequences, tandem repeats (small satellites and microsatellites) and interspersed repeats accounted for $0.95\%$ and $43.36\%$, respectively. Long terminal repeats (LTRs) of retroelements were the most abundant interspersed repeats, occupying $36.74\%$ of the genome, including $13.70\%$ Gypsy LTRs and $9.84\%$ Copia LTRs, followed by DNA transposable elements at $1.68\%$ (Table 17). Noncoding RNA annotation revealed that the T. wilfordii genome possessed 355 microRNAs (miRNAs), 797 transfer RNAs (tRNAs), 827 ribosomal RNAs (rRNAs), and 982 small nuclear RNAs (snRNAs) (Table 18). Integrated distributions of the genes, repeats, noncoding RNA densities, and all detected segmental duplications are shown in Figure 5. **Figure 4.:** *Interaction heat-map of chromosomal fragments based on Hi–C analysis. LG1–LG23 indicate Lachesis Groups 1-23. X and Y axes indicate the order positions of scaffolds on corresponding pseudochromosomes. The bar represents interaction strength between sequence segments.* **Figure 5.:** *Landscape of the *Tripterygium wilfordii* genome assembly. The circles (outer to inner) represent: pseudochromosomes, gene density, miRNAs, repeats, rRNAs, snRNAs, tRNAs, and duplicated gene links within the genome. The scale shows chromosomes in a 500-kb window; gene density in a 100-kb window (0–$100\%$, which means that the percentage of gene density indicated by the color gradient starts from 0 and goes to $100\%$ of 100 kb of DNA); miRNA density in a 100-kb window (0–$1\%$); repeat density in a 100-kb window (0–$100\%$); rRNA density in a 100-kb window (0–$18\%$); snRNA density in a 100-kb window (0–$1.3\%$); tRNA density in a 100-kb window (0–$1.8\%$); and detected gene duplication links (570).* TABLE_PLACEHOLDER:Table 9 TABLE_PLACEHOLDER:Table 10 TABLE_PLACEHOLDER:Table 11 TABLE_PLACEHOLDER:Table 12 TABLE_PLACEHOLDER:Table 13 TABLE_PLACEHOLDER:Table 14 TABLE_PLACEHOLDER:Table 15 TABLE_PLACEHOLDER:Table 16 TABLE_PLACEHOLDER:Table 17 TABLE_PLACEHOLDER:Table 18 ## Comparative genomic analysis To identify evolutionary characteristics and gene families, the T. wilfordii genome was compared with 11 published genomes of nine eudicot species (A. thaliana, C. sinensis, V. vinifera, G. max, M. truncatula, G. uralensis, C. sativus, P. trichocarpa and R. communis) and a monocot species (O. sativa). In addition, Amborella trichopoda, one of the basal groups of angiosperms, was selected as an outgroup. Based on gene family clustering analysis, 29,189 gene families were identified, of which 7296 were shared by all 12 species, and 485 of these shared families were single-copy gene families (Figure 6). Gene family numbers were compared between T. wilfordii and four fabid species. As shown in Figure 7A, 10,722 gene families were shared by G. max, C. sativus, G. uralensis, and M. truncatula, and 1086 gene families were specific to T. wilfordii. Compared with the most recent common ancestor (MRCA) of the 12 plant species, 15 gene families with 152 genes were expanded, including CYPs (see GigaDB for table [87]), and 42 gene families with 54 genes were contracted in T. wilfordii (Figure 7B). KEGG analysis showed that the expanded genes were enriched in pathways related to ‘ubiquinone and other terpenoid-quinone biosynthesis’ and ‘steroid hormone biosynthesis’, suggesting that gene family expansion contributed to specialized metabolite biosynthesis in T. wilfordii. A phylogenetic tree was constructed based on the super-alignment matrix of 485 single-copy orthologous genes from the 12 species. The branching order showed that A. thaliana (Brassicales) and C. sinensis (Sapindales) were sister to P. trichocarpa, R. communis (Malpighiales) and T. wilfordii (Celastrales), which diverged approximately 109.1 Mya, followed by divergence of T. wilfordii and species from Malpighiales approximately 102.4 Mya (Figures 5B and 8). These results were consistent with a previously proposed phylogenetic order, in which Celastrales and Malpighiales were found to be sister to each other [88]. **Figure 6.:** *The distribution of genes in different species.* **Figure 7.:** *Comparative genomic analysis. (A) Venn diagram of common and unique gene families in Tripterygium wilfordii with those in other species. (B) Phylogenetic analysis, divergence time estimation, and gene family expansions and contractions. Divergence times (Mya) are indicated by blue numbers, and numbers in brackets represent confidence intervals. Gene family expansions and contractions are indicated by green and red numbers, respectively.* **Figure 8.:** *Phylogenetic tree of Tripterygium wilfordii and other selected species. The branch length represents the evolution rate, and the value on the branch represents the value of bootstrap support.* ## Genome-wide identification and analysis of CYP candidates involved in celastrol biosynthesis **Figure 9.:** *The procedure of candidate CYP450 gene identification.* **Figure 10.:** *Phylogenetic tree of Tripterygium wilfordii CYPs. Diverse functions of CYPs are annotated by CYPs from Arabidopsis thaliana. A phylogenetic tree was built using the neighbor-joining method with a bootstrap test (n = 1000 replications). Numbers on the branches represent bootstrap support values.* The candidate gene identification procedure is illustrated in Figure 9. Based on HMMER analysis, 213 full-length ORFs of CYP genes were extracted from the T. wilfordii genome; these were annotated phylogenetic analysis with CYPs from A. thaliana. Thirty-five CYPs related to triterpenoid oxidases were identified, belonging to different subfamilies, including CYP716, CYP72, CYP71, CYP93, CYP705A and CYP81, which were previously reported to be functionally associated with diverse triterpenoid structural modifications (Figure 10) [89]. On the other hand, the expression patterns of the 213 identified CYPs were identified with TwOSC1 and TwOSC3, which are the two committed enzymes involved in the biosynthesis of the precursor of celastrol in T. wilfordii [21]. Based on RNA-seq data for various tissues, 20 profiles of gene coexpression were obtained, of which only profiles #3 and #13 showed significance (P-value < 0.05) (Figure 11). Profile #3 contained TwOSC3, and 45 CYPs showed similar expression patterns, while profile #13 included TwOSC1, and 51 CYPs had coexpression trends (Figure 12). This suggests that these CYPs are potentially involved in the biosynthesis of celastrol. To narrow down the candidate genes, the 35 CYPs identified by phylogenetic analysis were compared, and the genes showed patterns of coexpression with TwOSC1 and TwOSC3. As shown in Figure 13A, nine and seven triterpenoid biosynthesis-related CYPs showed patterns of coexpression with TwOSC1 and TwOSC3, respectively. However, no CYPs were common between the TwOSC1 group and TwOSC3 group. Based on tissue expression profiles, the 16 CYPs were clustered separately into two clades with TwOSC1 and TwOSC3, in which TwOSC3 exhibited root-specific expression, while TwOSC1 was highly expressed in leaves and other aerial parts (Figure 13B). Gene-to-gene and gene-to-metabolite Pearson’s correlation coefficients (r) were calculated using the tissue expression profiles of the 16 outstanding CYPs mentioned above, as well as three other known genes related to celastrol biosynthesis (TwHMGR1, TwFPS1 and TwDXR) [90–92], and the known intermediate product and celastrol concentrations [21]. As shown in Figure 13C, seven CYPs positively correlated with celastrol biosynthesis-related genes, with high Pearson’s r and significant P-values. In addition, these CYPs highly correlated with the levels of 29-hydroxy-friedelan-3-one, polpunonic acid, wilforic acid A and celastrol, which all specifically accumulated in the roots of T. wilfordii (Figure 14). Phylogenetic analysis placed these seven CYPs and CYPs from other species into three clades representing different functions in the structural modifications of triterpenoids (Figure 13D). Two CYPs (tw18g03520.1 and tw01g08960.1) were clustered with CYP81Q58 from C. sativus, which catalyzes hydroxylation of the C-25 position in cucurbitacins [93]; four CYPs (tw12g11350.1, tw18g06010.1, tw18g06210.1 and tw23g06640.1) were clustered with CYP712K4 from Monteverdia ilicifolia, which catalyzes the oxidation of the C-29 position using friedelin as a substrate [94]; and tw03g08450.1 was clustered with CYP716C11 from Centella asiatica, which hydroxylates the C-2 position of oleanolic acid and ursolic acid [95]. Since we were interested in identifying a C-29 position oxidase that could catalyze the conversion of friedelin to polpunonic acid, 3 CYPs were finally chosen as candidates for functional validation: tw18g06010.1, tw18g06210.1 and tw23g06640.1 (hereafter referred to as TwCYP712K1, TwCYP712K2 and TwCYP712K3) according to the closer relationship with CYP712K4 from M. ilicifolia, which is related to C-29 hydroxylation [94]. **Figure 11.:** *Maps of expression trends of CYPs with TwOSC1 and TwOSC3. Profiles ordered based on the P-value significance of number of genes assigned versus expected. Numbers on the top left corner represent the profiles number; numbers on the left bottom represent the P-value; numbers on the top right corner represent the total number of genes. Colored maps represent the significant enrichment with P-value < 0.05.* **Figure 12.:** *Co-expression of potential CYP genes with TwOSC1 and TwOSC3. The upper map indicates the similar expression patterns of CYPs with TwOSC3 and the lower map indicates the similar expression patterns of CYPs with TwOSC1. R, root; YL, young leaf; L, leaf; S, stem; FB, flower bud; F, flower; numbers 1–3 represent three biological replicates, respectively.* **Figure 13.:** *Identification of polpunonic acid producing CYPs via integration analysis. (A) Venn diagram of CYPs identified by phylogenetic analysis versus co-expression patterns. (B) Tissue-specific expression profiles and clustering of CYPs with TwOSC1 and TwOSC3. The gradient bar represents the expression levels from high (red) to low (blue) with log2 normalization, and the gray color represents the empty value. R, root; YL, young leaf; L, leaf; S, stem; FB, flower bud; F, flower. The numbers 1–3 represent three biological replicates. (C) Matrix of Pearson’s correlation coefficient and corresponding P-value of compounds, biosynthesis-related genes and CYP candidates. The lower triangle matrix represents Pearson’s correlation coefficient, and the upper triangle matrix presents P-values (green indicates P-values < 0.01, and gray indicates nonsignificant correlation). The gradient bar represents the correlation coefficient of positive or negative correlation from high to low. The black boxes enclose the highly correlated genes or compounds. (D) Phylogenetic analysis of putative CYPs in celastrol biosynthesis and CYPs known to catalyze structural modifications on triterpenoid scaffolds A phylogenetic tree was built using the maximum-likelihood method with a bootstrap test (n = 1000 replications). Allene oxide synthases AtCYP74A1 was set as an outgroup.* **Figure 14.:** *The accumulation of celastrol and intermediate products in different tissues of T. wilfordii. Bars are means ± SD from three independent biological replicates, n.d=not detected in our experimental condition; the value of not detected compounds was set to 0.* ## Heterologous expression and characterization of putative CYPs The full-length ORFs of TwCYP712K1, TwCYP712K2 and TwCYP712K3 were successfully clones and separately expressed in yeast fed with friedelin or 29-hydroxy-friedelan-3-one. However, no new peaks could be detected from the yeast strains expressing the enzymes and supplemented with friedelin compared with the empty vector (EV) control (see GigaDB [87]). This probably because the hydrophobic substrate could not be transported into the yeast cells. When fed with 29-hydroxy-friedelan-3-one, both TwCYP712K1 and TwCYP712K2 converted the substrate to a new compound possessing the same mass charge ratio (m/z) as the polpunonic acid standard, while TwCYP712K3 showed no such activity in this assay (Figure 15A and C). To further explore the enzyme activities, microsomes were extracted from yeast cells and the proteins incubated with friedelin or 29-hydroxy-friedelan-3-one for 12 h. As shown in Figure 15B, TwCYP712K1 and TwCYP712K2 converted 29-hydroxy-friedelan-3-one to a new compound in vitro, consistent with previous yeast in vivo assays. In addition, no new peak was detected from the enzyme reactions supplemented with friedelin compared with the EV control (see GigaDB [87]). **Figure 15.:** *Characterization of candidate CYPs for polpunonic acid biosynthesis. (A) Liquid chromatography–mass spectrometry (LC–MS) analyses of yeast samples fed with 29-hydroxy-friedelan-3-one as a substrate in vivo. Top, polpunonic acid standard; EV, empty vector; TwCYP712K1-TwCYP712K3, yeast expressing the corresponding proteins. New peaks with the same retention time as polpunonic acid are highlighted. (B) LC–MS analyses of microsome samples incubated with 29-hydroxy-friedelan-3-one as a substrate in vitro. Top, polpunonic acid standard; EV, empty vector; TwCYP712K1-TwCYP712K3, corresponding microsomes extracted from yeast cells. New peaks with the same retention time as polpunonic acid are highlighted. (C) Accurate masses of polpunonic acid. (D) Putative oxidation of 29-hydroxy-friedelan-3-one catalyzed by TwCYP712K1 and TwCYP712K2. The dashed arrow indicates multiple catalyzed steps that were unidentified.* ## Evolutionary analyses of TwCYP712K1 and TwCYP712K2 Genome analysis showed that TwCYP712K1 (tw18g06010.1) and TwCYP712K2 (tw18g06210.1) were located in pseudochromosome 18 within an approximately 200-kb region. This led us to examine the evolutionary relationship between specific CYPs. We hypothesized that a gene duplication event must have occurred during evolution. However, amino acid alignment indicated only $70.57\%$ identity between TwCYP712K1 and TwCYP712K2 (Figure 16), suggesting that these two genes became specialized a long time ago. Syntenic analysis showed that TwCYP712K1 and TwCYP712K2 had corresponding collinear genes in V. vinifera but not in O. sativa, indicating that these two genes appeared after the species differentiation of Poaceae and Vitales (<178.6 Mya), and they came from the common ancestor but divided after the species differentiation of Vitales (<130.6 Mya) (Figures 7 and 17). **Figure 16.:** *Alignment of TwCYP712K1 and TwCYP712K2 sequences. TwCYP712K1 and TwCYP712K2 exhibited 70.57% identity. The consensus sequences were highlighted by deep blue color.* **Figure 17.:** *Syntenic analysis of TwCYP712K1 and TwCYP712K2 genes. Focused CYP genes are colored purple.* ## Discussion In this study, we provided a high-quality reference genome of T. wilfordii with a 340.12 Mb genome assembly ($90.5\%$ of the 375.84 Mb estimated genome size) and 3.09 Mb contig N50, and successfully anchored $91.02\%$ of the sequences into 23 pseudochromosomes (Table 12). The quality of our genome is close to that of the recently published T. wilfordii genome (348.38 Mb total contigs and 4.36 Mb contig N50), which was sequenced and assembled using Illumina, PacBio and Hi–C sequencing [86]. They also identified a key CYP gene that can catalyze the oxidation of a methyl group to the acid moiety of dehydroabietic acid in triptolide biosynthesis, another clinically used specialized metabolite in T. wilfordii. Based on this genomic data, 35 CYP genes related to triterpenoid structure modification were identified according to phylogenetic analysis (Figure 8); 16 of these were co-expressed with TwOSC1 or TwOSC3 according to tissue-specific transcript profiles (Figures 11 and 12). *These* genes could be divided into two groups: the TwOSC1 group was highly expressed in leaves or other aerial parts, and the TwOSC3 group was specifically expressed in roots (Figure 13B), suggesting a sub-functionalization of TwOSC1 and TwOSC3 at the expression level in mediating the biosynthesis of friedelane-type triterpenoids. Correlation coefficient testing revealed that the expression levels of six CYPs significantly correlated with the expression patterns of genes involved in celastrol biosynthesis and the accumulation patterns of celastrol and its biosynthetic intermediates (Figure 13C). A more subdivided phylogenetic tree showed that three putative CYPs were clustered close to CYP712K4, which was cloned from M. ilicifolia (Figure 13D), another plant belonging to the Celastraceae family. CYP712K4 encodes an enzyme that catalyzes the oxidation of the C-29 position of friedelin to produce polpunonic acid [94]. Both in vivo and in vitro assays revealed that TwCYP712K1 and TwCYP712K2 could use 29-hydroxyfriedelan-3-one as a substrate to produce a new compound as the only product, indicating the oxidation of 29-hydroxyfriedelan-3-one at the C-29 position catalyzed by CYPs (Figure 15D). However, the peak of product from reactions of TwCYP712K1 and TwCYP712K3 were deviated with the peak of polpunonic acid standard. To demonstrate that these were same compound, future experiments will need to add another reaction containing 29-hydroxyfriedelan-3-one, buffer, enzyme and polpunonic acid standard (small amount, mixed into the reaction before LC-MS) in our follow on research. Comparative genome analysis showed that TwCYP712K1 and TwCYP712K2 derived from a common ancestor (Figure 17). Although they catalyzed the same reaction and were located close to each other on the same chromosome, the identity of the amino acid sequence ($70.57\%$) was not high (Figure 16). This suggests that TwCYP712K1 and TwCYP712K2 did not come from recent gene duplication, but separated during the evolution of the Celastraceae family. Interestingly, important catalytic activity for polpunonic acid biosynthesis in T. wilfordii was conserved in both these enzymes. As more genomes of the Celastraceae family are released, further evolutionary details of TwCYP712K1 and TwCYP712K2 can be investigated. There are many reports of genes encoding certain natural product pathways being grouped together in gene clusters to catalyze the biosynthesis of plant specialized metabolism, including triterpenoids [93, 96, 97]. However, neither the CYPs we identified, nor the signature enzymes TwOSC1 (tw21g04301.1) and TwOSC3 (tw20g03871.1), were clustered together. ## Potential for reuse We reported the reference genome assembly of T. wilfordii and provided a useful strategy for screening the genes involved in plant specialized metabolism. For further exploration, the genome can be used for comparative genomic analyses; for example, to resolve the controversial phylogenetic relationships within the COM (Celastrales, Oxalidales and Malpighiales) clade [88]. Additionally, full-length transcriptome and tissue-specific RNA-seq data can be used to mine all the biosynthetic pathway genes of celastrol, as well as the biosynthetic pathways of the diterpenoid and alkaloid active ingredients. ## Ethics approval and consent to participate Not applicable. ## Consent for publication Not applicable. ## Availability of data and materials The datasets generated and analyzed during the current study are available in GenBank of the National Center for Biotechnology Information (NCBI), under the BioProject number PRJNA640746. Gene and protein sequences of TwCYP712K1 (MT633088) and TwCYP712K2 (MT633089) are deposited in GenBank. Raw mass spectrometry data are available in MetaboLights under the study number MTBLS1080. Additional datasets are available in the GigaScience GigaDB repository [87]. ## Competing interests The authors declare that they have no competing interests. ## Funding This work was supported by the National Key R&D Program of China (grant numbers 2018YFC1706202, 2019YFD1000703, 2018YFD1000701, and 2020YFA0907901), the National Natural Science Foundation of China (grant numbers 31870282, 31700268), Youth Innovation Promotion Association of the Chinese Academy of Sciences, and the Chenshan Special Fund for Shanghai Landscaping Administration Bureau Program (grant numbers G182401, G182402, G192419, G192413, G192414 and G202402). Q.Z. is also support by the Shanghai Youth Talent Support Program and SA-SIBS Scholarship Program. ## Authors’ contributions T.L.P. and Q.Z initiated the program, coordinated the project, and wrote the manuscript. B.J.G prepared the plant materials. Y.K., J.L., M.Y.C., Y.M.F., and H.F prepared and analyzed the samples. T.L.P., M.X.Y., J.Y. and Q.Z. analyzed the genome sequence. M.X.Y. and J.Y. revised the manuscript. All authors read and approved the final manuscript. ## References 1. Helmstaedter A. *Pharmazie* (2013.0) **68** 643-646. PMID: 23923652 2. Tian Z, Zhang BH, Yang LP. **Ancient and modern evolution on clinical application of**. *Clin. Med. J.* (2019.0) **17** 19-23 3. Wu QJ. *Illustrated Catalogue of Plants* (1957.0) 4. Li RL. **Introduction to application of**. *Jiangsu J. Tradit. Chin. Med.* (1990.0) **11** 46-48 5. Bao J, Dai SM. **A Chinese herb**. *Rheumatol. Int.* (2011.0) **31** 1123-1129. PMID: 21365177 6. Han R, Rostami-Yazdi M, Gerdes S, Mrowietz U. **Triptolide in the treatment of psoriasis and other immune-mediated inflammatory diseases**. *Br. J. Clin. Pharmacol.* (2012.0) **74** 424-436. PMID: 22348323 7. Ding XL, Zhou XR, Jiang B, Zhao Q, Zhou GX. **Triptolide suppresses proliferation, hypoxia-inducible factor-1α and c-Myc expression in pancreatic cancer cells**. *Mol. Med. Rep.* (2015.0) **12** 4508-4513. PMID: 26094625 8. Yuan K. **Celastrol alleviates arthritis by modulating the inflammatory activities of neutrophils**. *J. Trad. Chin. Med. Sci.* (2017.0) **4** 50-58 9. Gao Q. **Treatment of db/db diabetic mice with triptolide: a novel therapy for diabetic nephropathy**. *Nephrol. Dialysis Transplant.* (2010.0) **25** 3539-3547 10. Lv HW. **The genus**. *Fitoterapia* (2019.0) **137**. PMID: 31163199 11. Duan HQ. **Triterpenoids from**. *Phytochemistry* (2000.0) **53** 805-810. PMID: 10783986 12. Lange BM, Fischedick JT, Lange MF, Srividya N, Šamec D, Poirier BC. **Integrative approaches for the identification and localization of specialized metabolites in**. *Plant Physiol.* (2017.0) **173** 456. PMID: 27864443 13. Ma J. **Anti-inflammatory and immunosuppressive compounds from**. *Phytochemistry* (2007.0) **68** 1172-1178. PMID: 17399748 14. Pang XF. **Celastrol suppresses angiogenesis-mediated tumor growth through inhibition of akt/mammalian target of rapamycin pathway**. *Cancer Res.* (2010.0) **70** 1951-1959. PMID: 20160026 15. Allison A, Cacabelos R, Lombardi V, Alvarez X, Vigo C. **Celastrol, a potent antioxidant and anti-inflammatory drug, as a possible treatment for Alzheimer’s disease**. *Prog. Neuropsychopharmacol. Biol. Psychiatry* (2001.0) **25** 1341-1357. PMID: 11513350 16. Liu JL, Lee JM, Salazar HMA, Mazitschek R, Ozcan U.. **Treatment of obesity with celastrol**. *Cell* (2015.0) **161** 999-1011. PMID: 26000480 17. Ma XR. **Celastrol protects against obesity and metabolic dysfunction through activation of a HSF1-PGC1α transcriptional axis**. *Cell Metab.* (2015.0) **22** 695-708. PMID: 26344102 18. Lin GM. **Research progress on**. *Chinese Agri. Sci. Bull.* (2009.0) **25** 90-93 19. Thimmappa R, Geisler K, Louveau T, O’Maille P, Osbourn A. **Triterpene biosynthesis in plants**. *Annu. Rev. Plant Biol.* (2014.0) **65** 225-257. PMID: 24498976 20. Miettinen K. **The TriForC database: a comprehensive up-to-date resource of plant triterpene biosynthesis**. *Nucleic Acids Res.* (2017.0) **46** D586-D594 21. Zhou JW. **Friedelane-type triterpene cyclase in celastrol biosynthesis from**. *New Phytol.* (2019.0) **223** 722-735. PMID: 30895623 22. Pei TL. **Protocols for “The genome analysis of Tripterygium wilfordii reveals TwCYP712K1 and TwCYP712K2 responsible for oxidation of friedelin in celastrol biosynthesis pathway”**. (2021.0). DOI: 10.17504/protocols.io.bspfndjn 23. Allen GC, Flores-Vergara MA, Krasynanski S, Kumar S, Thompson WF. **A modified protocol for rapid DNA isolation from plant tissues using cetyltrimethylammonium bromide**. *Nat. Protocols* (2006.0) **1** 2320-2325. PMID: 17406474 24. Marçais G, Kingsford C. **A fast, lock-free approach for efficient parallel counting of occurrences of k-mers**. *Bioinformatics* (2011.0) **27** 764-770. PMID: 21217122 25. 25.Nextomics. NextDenovo. https://github.com/Nextomics/NextDenovo. 26. Vaser R, Sović I, Nagarajan N, Šikić M. **Fast and accurate de novo genome assembly from long uncorrected reads**. *Genome Res* (2017.0) **27** 737-746. PMID: 28100585 27. Walker BJ. **Pilon: an integrated tool for comprehensive microbial variant detection and genome assembly improvement**. *PLoS One* (2014.0) **9**. PMID: 25409509 28. Simão FA, Waterhouse RM, Ioannidis P, Kriventseva EV, Zdobnov EM. **BUSCO: assessing genome assembly and annotation completeness with single-copy orthologs**. *Bioinformatics* (2015.0) **31** 3210-3212. PMID: 26059717 29. Li H, Durbin R. **Fast and accurate long-read alignment with Burrows–Wheeler transform**. *Bioinformatics* (2010.0) **26** 589-595. PMID: 20080505 30. Li H. **A statistical framework for SNP calling, mutation discovery, association mapping and population genetical parameter estimation from sequencing data**. *Bioinformatics* (2011.0) **27** 2987-2993. PMID: 21903627 31. Danecek P. **Twelve years of SAMtools and BCFtools**. (2021.0). DOI: 10.1093/gigascience/giab008 32. Grabherr MG. **Full-length transcriptome assembly from RNA-Seq data without a reference genome**. *Nat. Biotechnol.* (2011.0) **29** 644-652. PMID: 21572440 33. Kent WJ. **BLAT—The BLAST-Like Alignment Tool**. *Genome Res.* (2002.0) **12** 656-664. PMID: 11932250 34. Etherington GJ. **Bionano genome mapping from animal tissue**. (2020.0). DOI: 10.17504/protocols.io.bd7ei9je 35. 35.Bionano Genomics. https://bionanogenomics.com/support/software-downloads/. 36. Dekker J, Rippe K, Dekker M, Kleckner N. **Capturing chromosome conformation**. *Science* (2002.0) **295** 1306. PMID: 11847345 37. Belton JM, McCord RP, Gibcus JH, Naumova N, Zhan Y, Dekker J. **Hi–C: A comprehensive technique to capture the conformation of genomes**. *Methods* (2012.0) **58** 268-276. PMID: 22652625 38. Liu X. **The pipeline of Hi–C assembly**. (2018.0). DOI: 10.17504/protocols.io.qradv2e 39. Li H. **The Sequence Alignment/Map format and SAMtools**. *Bioinformatics* (2009.0) **25** 2078-2079. PMID: 19505943 40. Burton JN, Adey A, Patwardhan RP, Qiu R, Kitzman JO, Shendure J. **Chromosome-scale scaffolding of de novo genome assemblies based on chromatin interactions**. *Nat. Biotechnol.* (2013.0) **31** 1119-1125. PMID: 24185095 41. Kim D, Langmead B, Salzberg SL. **HISAT: a fast spliced aligner with low memory requirements**. *Nat. Methods* (2015.0) **12** 357-360. PMID: 25751142 42. Anders S, Pyl P, Huber W. **HTSeq-A Python framework to work with high-throughput sequencing data**. *Bioinformatics* (2014.0) **31** 166-169. PMID: 25260700 43. Trapnell C. **Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation**. *Nat. Biotechnol.* (2010.0) **28** 511-515. PMID: 20436464 44. 44.PacBio software downloads. https://www.pacb.com/support/software-downloads/. 45. Salmela L, Rivals E. **LoRDEC: accurate and efficient long read error correction**. *Bioinformatics* (2014.0) **30** 3506-3514. PMID: 25165095 46. Li WZ, Godzik A. **CD-HIT: a fast program for clustering and comparing large sets of protein or nucleotide sequences**. *Bioinformatics* (2006.0) **22** 1658-1659. PMID: 16731699 47. Jurka J, Kapitonov VV, Pavlicek A, Klonowski P, Kohany O, Walichiewicz J. **Repbase Update, a database of eukaryotic repetitive elements**. *Cytogenet. Genome Res.* (2005.0) **110** 462-467. PMID: 16093699 48. Xu Z, Wang H. **LTR_FINDER: an efficient tool for the prediction of full-length LTR retrotransposons**. *Nucleic Acids Res.* (2007.0) **35** W265-W268. PMID: 17485477 49. Price AL, Jones NC, Pevzner PA. **De novo identification of repeat families in large genomes**. *Bioinformatics* (2005.0) **21** 351-358 50. Smit AFA, Hubley R. **RepeatModeler (Version 2.0.1)**. (2019.0) 51. Benson G.. **Tandem repeats finder: a program to analyze DNA sequences**. *Nucleic Acids Res.* (1999.0) **27** 573-580. PMID: 9862982 52. 52.BLAST. tblastn. (version 2.2.26) http://blast.ncbi.nlm.nih.gov/Blast.cgi. 53. Birney E, Clamp M, Durbin R. **GeneWise and Genomewise**. *Genome Res.* (2004.0) **14** 988-995. PMID: 15123596 54. Stanke M, Keller O, Gunduz I, Hayes A, Waack S, Morgenstern B. **AUGUSTUS: ab initio prediction of alternative transcripts**. *Nucleic Acids Res.* (2006.0) **34** W435-W439. PMID: 16845043 55. Majoros W, Pertea M, Salzberg S. **TigrScan and GlimmerHMM: Two open source ab initio eukaryotic gene-finders**. *Bioinformatics* (2004.0) **20** 2878-2879. PMID: 15145805 56. Korf I. **Gene finding in novel genomes**. *BMC Bioinformatics* (2004.0) **5** 59. PMID: 15144565 57. Trapnell C, Pachter L, Salzberg SL. **TopHat: discovering splice junctions with RNA-Seq**. *Bioinformatics* (2009.0) **25** 1105-1111. PMID: 19289445 58. Pertea M, Pertea GM, Antonescu CM, Chang T-C, Mendell JT, Salzberg SL. **StringTie enables improved reconstruction of a transcriptome from RNA-seq reads**. *Nat. Biotechnol.* (2015.0) **33** 290-295. PMID: 25690850 59. Haas BJ. **Automated eukaryotic gene structure annotation using EVidenceModeler and the Program to Assemble Spliced Alignments**. *Genome Biol.* (2008.0) **9** R7. PMID: 18190707 60. 60.UniProtKB/Swiss-Prot. https://web.expasy.org/docs/swiss-prot˙guideline.html. 61. 61.Protein database. http://www.ncbi.nlm.nih.gov/protein. 62. **Protein families database (version 33.1)**. (2020.0) 63. 63.KEGG: Kyoto Encyclopedia of Genes and Genomes. 2021; KEGG http://www.genome.jp/kegg/.. (2021.0) 64. Blum M. **The InterPro protein families and domains database: 20 years on**. *Nucleic Acids Res.* (2020.0) **49** D344-D354 65. Lowe TM, Chan PP. **tRNAscan-SE On-line: integrating search and context for analysis of transfer RNA genes**. *Nucleic Acids Res.* (2016.0) **44** 54-57 66. Nawrocki EP, Eddy SR. **Infernal 1.1: 100-fold faster RNA homology searches**. *Bioinformatics* (2013.0) **29** 2933-2935. PMID: 24008419 67. Li L, Stoeckert CJ, Roos DS. **OrthoMCL: identification of ortholog groups for eukaryotic genomes**. *Genome Res.* (2003.0) **13** 2178-2189. PMID: 12952885 68. Edgar RC. **MUSCLE: multiple sequence alignment with high accuracy and high throughput**. *Nucleic Acids Res.* (2004.0) **32** 1792-1797. PMID: 15034147 69. Stamatakis A.. **RAxML-VI-HPC: Maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models**. *Bioinformatics* (2006.0) **22** 2688-2690. PMID: 16928733 70. Yang ZH. **PAML 4: Phylogenetic analysis by maximum likelihood**. *Mol. Biol. Evol.* (2007.0) **24** 1586-1591. PMID: 17483113 71. Kumar S, Stecher G, Suleski M, Hedges SB. **TimeTree: a resource for timelines, timetrees, and divergence times**. *Mol. Biol. Evol.* (2017.0) **34** 1812-1819. PMID: 28387841 72. Bie T, Cristianini N, Demuth J, Hahn M. **CAFE: A computational tool for the study of gene family evolution**. *Bioinformatics* (2006.0) **22** 1269-1271. PMID: 16543274 73. Eddy SR. **Profile hidden Markov models**. *Bioinformatics* (1998.0) **14** 755. PMID: 9918945 74. Christ B. **Repeated evolution of cytochrome P450-mediated spiroketal steroid biosynthesis in plants**. *Nat. Commun.* (2019.0) **10** 3206. PMID: 31324795 75. Kumar S, Stecher G, Li M, Knyaz C, Tamura K. **MEGA X: molecular evolutionary genetics analysis across computing platforms**. *Mol. Biol. Evol.* (2018.0) **35** 1547-1549. PMID: 29722887 76. Ernst J, Bar-Joseph Z. **STEM: a tool for the analysis of short time series gene expression data**. *BMC Bioinformatics* (2006.0) **7** 191. PMID: 16597342 77. 77.OmicShare Tools. http://www.omicshare.com/tools. 78. Zhao Q. **Two CYP82D enzymes function as flavone hydroxylases in the biosynthesis of root-specific 4**. *Mol. Plant.* (2018.0) **11** 135-148. PMID: 28842248 79. Truan G, Cullin C, Reisdorf P, Urban P, Pompon D. **Enhanced in vivo monooxygenase activities of mammalian P450s in engineered yeast cells producing high levels of NADPH-P450 reductase and human cytochrome b5**. *Gene* (1993.0) **125** 49-55. PMID: 8449412 80. Pompon D, Louerat B, Bronine A, Urban P. **Yeast expression of animal and plant P450s in optimized redox environments**. *Methods Enzymol.* (1996.0) **272** 51-64. PMID: 8791762 81. Zhao Q. *Sci. Adv.* (2016.0) **2** 82. Bradford MM. **A rapid and sensitive method for the quantitation of microgram quantities of protein using the principle of protein dye binding**. *Anal. Biochem.* (1976.0) **6** 3177-3188 83. Wang YP. **MCScanX: A toolkit for detection and evolutionary analysis of gene synteny and collinearity**. *Nucleic Acids Res.* (2012.0) **40** 49 84. Bolser DM, Staines DM, Perry E, Kersey PJ. **Ensembl Plants: integrating tools for visualizing, mining, and analyzing plant genomic data**. *Methods Mol. Biol.* (2017.0) **1533** 1-31. PMID: 27987162 85. Frith MC, Hamada M, Horton P. **Parameters for accurate genome alignment**. *BMC Bioinformatics* (2010.0) **11** 80. PMID: 20144198 86. Tu L. **Genome of**. *Nat. Commun.* (2020.0) **11** 971. PMID: 32080175 87. Pei T, Yan M, Kong Y, Fan H, Liu J, Cui M, Fang Y, Ge B, Yang J, Zhao Q. **Supporting data for “The genome of**. (2021.0). DOI: 10.5524/100864 88. Zhao L. **Phylogenomic analyses of large-scale nuclear genes provide new insights into the evolutionary relationships within the rosids**. *Mol. Phylogenet. Evol.* (2016.0) **105** 166-176. PMID: 27369455 89. Ghosh S.. **Triterpene structural diversification by plant cytochrome P450 enzymes**. *Front. Plant Sci.* (2017.0) **8** 1886. PMID: 29170672 90. Wang JD. **The effects of TwHMGR overexpression on the biosynthesis of triptolide and celastrol in**. *Acta Pharm. Sin.* (2018.0) **53** 1225-1232 91. Zhang YF. **Overexpression and RNA interference of TwDXR regulate the accumulation of terpenoid active ingredients in**. *Biotechnol. Lett.* (2018.0) **40** 419-425. PMID: 29270714 92. Zhao YJ, Zhang YF, Su P, Yang J, Huang LQ, Gao W. **Genetic transformation system for woody plant**. *Front. Plant Sci.* (2018.0) **8** 2221. PMID: 29375599 93. Shang Y. **Biosynthesis, regulation, and domestication of bitterness in cucumber**. *Science* (2014.0) **346** 1084. PMID: 25430763 94. Bicalho KU. **CYP712K4 catalyzes the C-29 oxidation of friedelin in the Maytenus ilicifolia quinone methide triterpenoid biosynthesis pathway**. *Plant Cell Physiol.* (2019.0) **60** 2510-2522. PMID: 31350564 95. Miettinen K. **The ancient CYP716 family is a major contributor to the diversification of eudicot triterpenoid biosynthesis**. *Nat. Commun.* (2017.0) **8**. PMID: 28165039 96. Nutzmann HW, Huang A, Osbourn A. **Plant metabolic clusters - from genetics to genomics**. *New Phytol.* (2016.0) **211** 771-789. PMID: 27112429 97. Mugford ST. **Modularity of plant metabolic gene clusters: A trio of linked genes that are collectively required for acylation of triterpenes in oat**. *Plant Cell.* (2013.0) **25** 1078. PMID: 23532069
--- title: Estradiol regulates voltage-gated potassium currents in corticotropin-releasing hormone neurons authors: - Emmet M. Power - Dharshini Ganeshan - Karl J. Iremonger journal: The Journal of Experimental Biology year: 2023 pmcid: PMC10038157 doi: 10.1242/jeb.245222 license: CC BY 4.0 --- # Estradiol regulates voltage-gated potassium currents in corticotropin-releasing hormone neurons ## ABSTRACT Corticotropin-releasing hormone (CRH) neurons are the primary neural population controlling the hypothalamic–pituitary–adrenal (HPA) axis and the secretion of adrenal stress hormones. Previous work has demonstrated that stress hormone secretion can be regulated by circulating levels of estradiol. However, the effect of estradiol on CRH neuron excitability is less clear. Here, we show that chronic estradiol replacement following ovariectomy increases two types of potassium channel currents in CRH neurons: fast inactivating voltage-gated A-type K+ channel currents (IA) and non-inactivating M-type K+ channel currents (IM). Despite the increase in K+ currents following estradiol replacement, there was no overall change in CRH neuron spiking excitability assessed with either frequency–current curves or current ramps. Together, these data reveal a complex picture whereby ovariectomy and estradiol replacement differentially modulate distinct aspects of CRH neuron and HPA axis function. ## Abstract Summary: Chronic estradiol replacement in ovariectomized mice influences voltage-gated potassium channel function. ## INTRODUCTION Corticotropin-releasing hormone (CRH) neurons in the paraventricular nucleus (PVN) of the hypothalamus are neuroendocrine neurons that control activity of the hypothalamic–pituitary–adrenal (HPA) axis (Herman and Cullinan, 1997; Ulrich-Lai and Herman, 2009; Kim et al., 2019a,b; Füzesi et al., 2016; Sterley et al., 2018; Daviu et al., 2020). These neurons are activated in response to stress (Kim et al., 2019b), which leads to CRH secretion from the median eminence into the portal circulation. This triggers secretion of adrenocorticotropic hormone (ACTH) from the anterior pituitary, which subsequently stimulates corticosteroid synthesis and release from the adrenal cortex. Activity of the HPA axis is sexually dimorphic. In rodents, females have higher levels of circulating corticosterone as well as stress-evoked corticosterone release (Seale et al., 2004). There are also marked changes in activity of the HPA axis across the female reproductive cycle, with both basal and stress-evoked levels of corticosterone being highest on proestrus (Atkinson and Waddell, 1997; Viau and Meaney, 1991). Because estradiol is a primary sex hormone in females and is at its highest levels on proestrus, this has led to the theory that estradiol could be responsible for both sex and estrous cycle differences in HPA axis activity (Nilsson et al., 2015). Consistent with this idea, basal corticosterone secretion in female rats is reduced following ovariectomy (Babb et al., 2013; Seale et al., 2004; Young et al., 2001). Circulating corticosterone levels can also be elevated in ovariectomized (Ovx) rats with subsequent estradiol replacement (Figueiredo et al., 2007; Kitay, 1966; Lo et al., 2000). Despite this, other data in rats show that estradiol suppresses stress-evoked ACTH release (Babb et al., 2013; Young et al., 2001) as well as stress-evoked cFos labelling in CRH neurons (Dayas et al., 2000; Figueiredo et al., 2007; Gerrits et al., 2005). To add to this complex picture, studies investigating the effect of estradiol replacement in Ovx mice are conflicting. Some studies report that estradiol replacement in Ovx mice can increase corticosterone levels (Kreisman et al., 2020), whereas others report no effect (Aoki et al., 2010; Speert et al., 2002; Wada et al., 2018) or even reduced corticosterone (Daodee et al., 2022; Eid et al., 2020; Tantipongpiradet et al., 2019; Ghobadi et al., 2016; Tang et al., 2005). Overall, the impact of estradiol on HPA axis function and CRH neuron activity is complex and may differ between species. We have recently shown that K+ channel function and CRH neuron excitability are regulated over the estrous cycle in mice (Power and Iremonger, 2021). During the proestrus phase of the estrous cycle, coinciding with a peak in estradiol levels, CRH neurons exhibit smaller K+ channel currents and higher levels of excitability measured by electrophysiological recordings. This corresponds with previous publications showing basal and stress-evoked corticosterone levels being highest during proestrus (Atkinson and Waddell, 1997; Viau and Meaney, 1991). However, it is currently unclear whether these changes in excitability are driven by estradiol alone. Previous work has shown that estradiol can regulate K+ currents and excitability in other central neurons (Arroyo et al., 2011; Vastagh et al., 2019; DeFazio and Moenter, 2002). Therefore, in the present study, we aimed to determine the effect of Ovx and high or low levels of estradiol replacement on K+ channel function and CRH neural excitability. Using patch-clamp recordings from CRH neurons, we show that chronic elevations in estradiol levels in Ovx female mice leads to increased levels of two K+ channel currents: fast inactivating voltage-gated A-type K+ channel currents (IA) and non-inactivating M-type K+ channel currents (IM). However, chronic estradiol elevations did not significantly change intrinsic excitability compared with Ovx animals. This suggests that the magnitude of changes in K+ channel currents were not sufficient to impact spiking excitability. Despite this, we speculate that enhanced K+ channel function may affect how these neurons integrate and process stress-relevant synaptic inputs. ## Animals All electrophysiological experiments were carried out in adult female (2–6 months old) Crh-IRES-Cre;Ai14 (tdTomato) mice. These mice were generated by crossing the Crh-IRES-Cre (B6(CG)-Crhtm1(cre)Zjh/J) (Taniguchi et al., 2011) strain with the Ai14 (B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomto)Hze/J) strain, both originally obtained from The Jackson Laboratory (stock numbers 012704 and 007914, respectively). These mice have been previously shown to faithfully label CRH neurons in the PVN (Chen et al., 2015; Wamsteeker-Cusulin et al., 2013; Jamieson et al., 2017). Serum corticosterone and tissue samples were taken from a mixture of C57/Bl6J (The Jackson Laboratory) and Crh-IRES-Cre mice (2–4 months). Animals were subjected to a 12 h:12 h light:dark cycle (07:00–19:00 h lights on) with food and water available ad libitum. All protocols and procedures were approved by the University of Otago Animal Ethics Committee and carried out in accordance with the New Zealand Animal Welfare Act. ## Ovariectomy and hormone replacement Adult female mice (>3 months) were bilaterally ovariectomized under isoflurane general anesthetic. Simultaneously mice received a 10 mm long silastic capsule (inner diameter: 1.57 mm; outer diameter: 2.41 mm) containing 17β-estradiol (estradiol) subcutaneously implanted between the shoulder blades and neck. The dose of estradiol (E8875, Sigma-Aldrich) was based on previous publications and estimated to give levels similar to estrus/diestrus for the Ovx–low estradiol (OvxLowE) group and proestrus (or higher) for the Ovx–high estradiol (OvxHighE) group (Desroziers et al., 2017; Hellier et al., 2018; Porteous et al., 2021). OvxLowE mice received an implant with 4 µg estradiol dissolved in absolute ethyl alcohol and mixed with silastic gel. OvxHighE mice received an implant containing crystalline estradiol mixed 1:1 with cholesterol. One group of mice were Ovx and received an implant containing only cholesterol. All mice were left for 2–3 weeks before being used for tissue collection or electrophysiology. ## Blood, tissue collection and ELISA All mice were habituated to handling for at least 4 days prior to tissue collection. Mice were euthanized (between 09:00 and 11:00 h) and trunk blood was collected in tubes. All blood samples were kept on ice before being centrifuged. Uterus and adrenal glands were dissected out and weighed immediately following decapitation. Uterus masses were also taken from a subset of animals used for electrophysiology; the protocol for dissection and weighing remained the same. Adrenal gland mass is the combined mass of both left and right adrenals for each animal. Thymus glands were dissected out and stored in $4\%$ PFA before being weighed. Serum corticosterone was measured using an ELISA (Arbor Assays, catalogue no. K014,RRID AB_2877626) according to the manufacturer's instructions. ## Slice preparation Mice were killed by cervical dislocation between 09:00 and 11:00 h, their brain was quickly removed and placed in ice-cold oxygenated ($95\%$ O2, $5\%$ CO2) slicing solution containing (in mmol l−1): 87 NaCl, 2.5 KCl, 25 NaHCO3, 1.25 NaH2PO4, 0.5 CaCl2, 6 MgCl2, 25 d-glucose and 75 sucrose, pH 7.2–7.4. A vibratome (VT1200S, Lecia Microsystems) was used to cut 200-µm-thick coronal slices of the PVN, which were then incubated in oxygenated artificial cerebrospinal fluid (aCSF) containing (in mmol l−1): 126 NaCl, 2.5 KCl, 26 NaHCO3, 1.25 NaH2PO4, 2.5 CaCl2, 1.5 MgCl2 and 10 d-glucose at 30°C for at least 1 h before recording. For recording, slices were transferred to a recording chamber and continuously perfused with 30°C aCSF at 1.5 ml min−1. CRH neurons within the PVN were visualized using a 40× objective and epifluorescence to excite tdTomato. ## Whole-cell electrophysiology recordings Electrophysiological recordings were collected with a Multiclamp 700B amplifier (Molecular Devices), filtered at 2 kHz, and digitized using the Digidata 1440a (Molecular Devices). Data were analysed with Clampfit 10.7 (Molecular Devices). For whole-cell recordings, borosilicate glass pipettes (tip resistance: 2–5 MΩ) were filled with an internal solution containing (in mmol l−1): 120 K-gluconate, 15 KCl, 0.5 Na2EGTA, 2 Mg2ATP, 0.4 Na2GTP, 10 HEPES, 5 Na2-phosphocreatine and $0.25\%$ neurobiotin (adjusted to pH 7.2 with KOH; adjusted to ≈290 mOsm with sucrose). All current clamp experiments were performed in the presence of 10 μmol l−1 cyanquixaline (6-cyano-7-nitroquinoxaline-2,3-dione) (CNQX) and picrotoxin (50 μmol l−1). Each cell was held at approximately −60 mV. The liquid junction potential was calculated to be approximately −14.1 mV and was not compensated for. Cells were not recorded from if input resistance was below 0.7 GΩ or access resistance was above 30 MΩ and both input and access resistance were monitored throughout to ensure stable recording. We used a current step protocol to determine spike output and first spike latency (FSL). The step protocol consisted of 300 ms square steps from 0 to +50 pA in 5 pA increments. Spikes were detected using a threshold search in Clampfit and were analysed for rise time, decay time, amplitude and half width. FSL was calculated from the time of the depolarizing step initiation to the action potential (AP) threshold for the first spike evoked at steps equal or greater than 10 pA. AP threshold was defined as the voltage at which the AP first derivative crossed 10 mV ms−1. The same analysis criteria were used to identify FSL and AP threshold for a 1 s, +40 pA s−1 ramp protocol. For all voltage clamp recordings, neurons were clamped at −60 mV. Input resistance, access resistance and capacitance were monitored periodically throughout recordings. IA current recordings were performed in the presence of CNQX (10 μmol l−1), picrotoxin (50 μmol l−1), tetrodotoxin (TTX; 0.5 μmol l−1), XE991 (40 μmol l−1) and nifedipine (100 μmol l−1). To evoke IA currents, neurons were hyperpolarized from −60 to −110 mV for 500 ms before a family of depolarizing steps were delivered in 10 mV steps from −100 to +30 mV. Peak IA amplitude for each voltage step was measured and normalized to capacitance to give the current densities (pA pF−1). A protocol was used to measure the IM relaxation current, similar to that used in previous studies (Hu et al., 2016; Roepke et al., 2011). These recordings were performed in the presence of CNQX (10 μmol l−1), picrotoxin (50 μmol l−1) and TTX (0.5 μmol l−1). This protocol consisted of a pre-pulse to −20 mV for 300 ms followed by 500 ms steps from −30 to −75 mV. The IM relaxation current was measured as the amplitude difference between the initial current and the sustained current at the end of the voltage step. ## Analysis Statistical analysis was performed using GraphPad Prism 8. All reported values are means±s.e.m. Comparisons between groups were carried out using either one- or two-way ANOVA where appropriate, with Tukey’s post hoc multiple comparison tests. All n-values represent neuron number, all groups had N>3 animals. $P \leq 0.05$ was considered statistically significant. P-values reported on figures are for post hoc multiple comparison tests. ## Chronic estradiol effects on uterine mass, adrenal mass and corticosterone In order to manipulate estradiol levels, female mice were ovariectomized and either received no treatment (Ovx) or received a low (OvxLowE) or high (OvxHighE) dose estradiol implant. Two to three weeks later, animals were euthanized and blood and tissue were collected. The uterus is highly sensitive to estradiol, shows enlargement in response to estradiol elevations and has been previously used as a bioassay for estrogen levels (Owens and Ashby, 2002; Serova et al., 2010). Estradiol treatment induced a significant increase in uterine mass (one-way ANOVA, F2,32=51.91, $P \leq 0.0001$; Fig. 1A), consistent with previous studies. Previous work has shown that estradiol treatment can also elevate basal corticosterone levels in rats (Figueiredo et al., 2007; Kitay, 1963; Lo et al., 2000), but have no effect (Aoki et al., 2010; Speert et al., 2002; Wada et al., 2018) or reduce corticosterone levels in mice (Daodee et al., 2022; Eid et al., 2020; Tantipongpiradet et al., 2019). Here, we found that in Ovx mice, high or low dose estradiol implants did not significantly change morning corticosterone levels (one-way ANOVA, F2,17=0.55, $$P \leq 0.58$$; Fig. 1B). Likewise, combined adrenal mass was also not different across the estradiol treatment groups (one-way ANOVA, F2,18=0.64, $$P \leq 0.54$$; Fig. 1C). Interestingly, thymus mass, which can be influenced by both corticosterone (Karatsoreos et al., 2010) and estradiol (Clarke and Kendall, 1989; Utsuyama and Hirokawa, 1989; Zoller and Kersh, 2006), was significantly different between the groups (one-way ANOVA, F2,17=30.52, $P \leq 0.0001$; Fig. 1D). Post hoc Tukey's multiple comparisons showed a significant difference between OvxHighE and both Ovx and OvxLowE ($P \leq 0.0001$ and $$P \leq 0.0004$$, respectively), and a significant difference between Ovx and OvxLowE ($$P \leq 0.023$$). **Fig. 1.:** *Consequences of chronic estradiol treatment. (A) Uterus mass for each group measured immediately after the animals were euthanized: Ovx (green), OvxLowE (orange) and OvxHighE (grey). Asterisks indicate significance identified by Tukey's multiple comparisons. Both low and high chronic estradiol implants increase uterus mass compared with Ovx animals. (B) There was no significant difference in circulating serum CORT levels taken from trunk blood samples. (C) Comparison of combined (mass of both left and right adrenal) adrenal mass between the groups. There was no significant difference in adrenal mass between the groups. (D) There was a significant difference in thymus mass between the three groups. Black asterisks indicate significant result of one-way ANOVA, coloured asterisks indicate post hoc Tukey's multiple comparisons test. Green asterisks indicate significance with the Ovx group, orange asterisks indicate significant difference with the OvxLowE group. N>6 mice for all groups. P-values: *P≤0.05, ***P≤0.001, ****P≤0.0001.* ## Estradiol regulates IA potassium channel currents in CRH neurons Neuronal intrinsic excitability is dictated in part by voltage-gated ion channel density and function. We have previously shown that IA, a transient K+ current, is regulated over the estrous cycle in CRH neurons (Power and Iremonger, 2021). To investigate the link between estradiol levels and IA currents, we used a voltage clamp protocol on CRH neurons from Ovx, OvxLowE or OvxHighE manipulated mice. Electrophysiological recordings were performed 2–3 weeks post ovariectomy. A two-way repeated-measures (RM) ANOVA revealed that there was a significant effect of estradiol treatment on IA current density (F2,18=5.32, $$P \leq 0.015$$; Fig. 2A,B), a significant effect of voltage step (F14,252=134.1, $P \leq 0.0001$) and a significant interaction (F28,252=4.68, $P \leq 0.0001$). Post hoc tests showed that current densities at multiple voltage steps were smallest in Ovx animals compared with OvxLowE ($P \leq 0.05$) and OvxHighE animals ($P \leq 0.05$). Peak amplitude of the current at the maximum voltage step (+30 mV) was also significantly different between groups (one-way ANOVA, F2,17=4.55, $$P \leq 0.026$$; Fig. 2C). Post hoc comparison revealed a significant difference between Ovx and OvxHighE ($$P \leq 0.021$$) but not with OvxLowE ($$P \leq 0.18$$). These findings show that chronic estradiol manipulations lead to changes in IA K+ currents in CRH neurons. **Fig. 2.:** *IA currents are influenced by chronic estradiol treatment. (A) Evoked IA currents from individual corticotropin releasing hormone (CRH) cells in each group: Ovx (green), OvxLowE (orange) and OvxHighE (grey). (B) IA current densities plotted for each 10 mV voltage step from −100 to +30 mV. Cells from Ovx animals had significantly smaller IA current densities compared with cells from OvxLowE and OvxHighE. (C) Peak amplitude IA currents (not normalized to capacitance) evoked by a +30 mV step. Peak IA currents in Ovx animals were significantly smaller than in OvxLowE and OvxHighE. Results of one- and two-way ANOVAs are reported in Table 2. Asterisks denote significance by Tukey's multiple comparisons test: orange, OvxLowE versus Ovx; grey, OvxHighE versus Ovx. There were no significant differences between the OvxLowE and OvxHighE groups. N=3–5 mice for all groups. P-values: *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001.* ## Estradiol regulates IM potassium channel currents in CRH neurons In addition to IA, M-type (IM) potassium currents were also investigated. IM currents are slowly activating, non-inactivating voltage-gated currents. They can contribute to intrinsic excitability via regulation of resting membrane potential and are ubiquitously found in neurons (Gutman et al., 2005). A voltage clamp protocol was used to measure the relaxation of the IM current (see Materials and Methods). Comparison of the three groups using a two-way RM ANOVA revealed a significant effect of chronic estradiol treatment on IM­ current densities (F2,19=5.37, $$P \leq 0.014$$; Fig. 3A,B). Post hoc multiple comparisons revealed that the OvxHighE group had a significantly higher IM current density compared with both Ovx ($$P \leq 0.04$$) and OvxLowE ($$P \leq 0.04$$) at the highest voltage step (–30 mV). A one-way ANOVA comparing peak IM­ current amplitude in the three groups was also significant (F2,19=5.01, $$P \leq 0.018$$; Fig. 3C), with multiple comparisons revealing significant differences between OvxHighE and OvxLowE ($$P \leq 0.026$$) but not between OvxHighE and Ovx ($$P \leq 0.075$$). These results show that estradiol levels also regulate IM K+ channel currents in CRH neurons. **Fig. 3.:** *IM currents are influenced by high chronic estradiol levels but not low. (A) Evoked IM currents from individual CRH cells in each group: Ovx (green), OvxLowE (orange) and OvxHighE (grey). (B) IM current densities plotted for each 5 mV voltage step from −75 to −30 mV. Cells from OvxHighE animals had significantly larger IM current densities compared with cells from OvxLowE and Ovx. (C) Peak amplitude IM currents (not normalized to capacitance) evoked by a −30 mV step. Peak IM currents in OvxHighE animals were significantly larger than in OvxLowE, but not compared with cells from Ovx animals. Results of one- and two-way ANOVAs are reported in Table 2. Asterisks denote significance by Tukey's multiple comparisons test: orange, OvxLowE versus OvxHighE; green, Ovx versus OvxHighE. There were no significant differences between the OvxLowE and Ovx groups. N=3–5 mice for all groups. P-values: *P≤0.05.* ## Chronic estradiol manipulations do not alter CRH neuron intrinsic excitability As both IA and IM currents are altered by artificially induced estradiol concentrations, we next investigated whether CRH neuron intrinsic excitability was also influenced. Given the elevated K+ currents in CRH neurons from OvxLowE and OvxHighE mice, we expected lower intrinsic excitability levels from these neurons compared with those from Ovx animals. Neurons were held around −60 mV in current clamp before injecting a family of current steps from 0 to +50 pA in 5 pA increments (Fig. 4A,B). This protocol was used to generate a frequency–current curve (F–I curve) and was performed on CRH neurons from Ovx, OvxLowE and OvxHighE animals. CRH neuron firing frequency was not different between the groups (two-way RM ANOVA, F2,32=0.386, $$P \leq 0.61$$; Fig. 4A,B, Table 1), nor was peak firing frequency (one-way ANOVA, F2,27=3.102, $$P \leq 0.68$$; Fig. 4B inset), or the slope of the F–I curves (one-way ANOVA, F2,32=0.19, $$P \leq 0.83$$; Table 1). FSL, measured from the 10 pA current step onwards, was not affected by chronic estradiol treatment (two-way RM ANOVA, F2,31=0.059, $$P \leq 0.94$$; Fig. 4C, Table 1). The total number of APs fired over all current steps was also similar between groups (one-way ANOVA, F2,30=0.17, $$P \leq 0.84$$; Fig. 4D). Analysis of AP parameters showed no significant differences in amplitude, rise time, half width or decay time between the three groups (Fig. 4E–H, see Table 1 for mean values and Table 2 for statistics). There were also no significant differences in capacitance or input resistance between groups (one-way ANOVA, F2,85=1.93, $$P \leq 0.15$$, and F2,85=1.6, $$P \leq 0.21$$, respectively; Tables 1 and 2). **Fig. 4.:** *Chronic estradiol does not change CRH neuron intrinsic excitability or action potential (AP) parameters. (A) Representative responses of CRH neurons to 0, 10, 30 and 50 pA current steps of the frequncy–current (F–I) curve in each group: Ovx (green), OvxLowE (orange) and OvxHighE (grey). (B) *Summary data* for the F–I curve. There was no significant difference between the three groups across the F–I curve or at peak firing frequency (50 pA step, inset). Results of two-way ANOVA are reported in Table 2. (C) First spike latency (FSL) for each current step. There was no significant difference between the groups. Results of two-way ANOVA are reported in Table 2. (D) Total number of APs fired over all current steps for each group. There was no significant difference between the groups. All AP parameters were measured from the first AP fired from each cell. There was no significant difference between the groups in AP amplitude (E), rise time (F), half width (G) or decay time (H). Results of one-way ANOVA are reported in Table 2.* TABLE_PLACEHOLDER:Table 1. TABLE_PLACEHOLDER:Table 2. In addition to an F–I curve, CRH neuron excitability was also tested using a current clamp ramp protocol consisting of a 40 pA ramp delivered over 1 s (Fig. 5A). This protocol gives a more accurate measurement of latency to first spike and AP threshold compared with measurements from F–I curves. Neither the number of APs fired (Fig. 5B), AP threshold (Fig. 5C) or FSL (Fig. 5D) were significantly different between the groups (one-way ANOVA, F2,29=1.89, $$P \leq 0.17$$, F2,31=0.64, $$P \leq 0.54$$, and F2,31=0.22, $$P \leq 0.80$$, respectively). Despite chronic estradiol treatment causing changes in K+ channel function, these results show that chronic estradiol manipulation had no impact on CRH neuron intrinsic excitability. **Fig. 5.:** *Chronic estradiol does not influence CRH neuron intrinsic excitability measured by a current ramp. (A) Example traces showing CRH neuron spiking response to a 1 s, 40 pA, current ramp protocol (bottom right) in each group: Ovx (green), OvxLowE (orange) and OvxHighE (grey). (B) There was no significant difference in the total number of APs fired during the ramp protocol between the three groups. (C) Neither high nor low estradiol concentrations altered AP threshold compared with Ovx. AP threshold was defined as the voltage at which the AP first derivative crossed 10 mV ms−1. (D) There was no change in AP latency between the groups. Results of one-way ANOVAs are reported in Table 2.* ## IA­ currents correlate with CRH neuron excitability We have previously shown that IA currents are regulated over the female estrous cycle and control CRH neuron intrinsic excitability (Power and Iremonger, 2021). We took the data from this previous work along with data from the present study and performed Pearson's correlation tests for IA current density versus various parameters of excitability measured from F–I curves and ramp protocol (Fig. 6). We included data from the following groups: intact estrus, intact proestrus, intact diestrus, Ovx, OvxLowE and OvxHighE. **Fig. 6.:** *IA current densities correlate strongly with measures of CRH neuron excitability. (A) Correlation analysis between the number of APs fired at the highest (50 pA) current step given during the F–I curve and peak IA current densities. Red is proestrus, blue is estrus, purple is diestrus, green is Ovx, orange is OvxLowE and grey is OvxHighE. (B) IA current densities versus total number of APs fired over the entire F–I curve for each group. (C) The slope of the F–I curve was correlated with IA current densities. (D) FSL of current ramp protocol correlated with IA current densities. r, R2 and P-values for each comparison are listed on individual graphs.* IA current densities were found to be negatively correlated with the number of APs fired at the 50 pA current step ($r = 0.95$, $$P \leq 0.002$$; Fig. 6A), the total number of APs fired ($r = 0.89$, $$P \leq 0.016$$; Fig. 6B) and the slope of the F–I curve ($r = 0.87$, $$P \leq 0.023$$; Fig. 6C). IA current densities were also positively correlated with current ramp FSL, although this was not significant ($r = 0.63$, $$P \leq 0.18$$; Fig. 6D). These data show that changes in IA current density in CRH neurons are correlated with several parameters of intrinsic excitability. ## DISCUSSION Circulating levels of estradiol have been previously shown to regulate the HPA axis (Patchev et al., 1995; Roy et al., 1999; Figueiredo et al., 2007); however, the impact of estradiol on CRH neuron excitability has been less clear. In the present study, we found that compared with Ovx mice, replacement with either low or high doses of estradiol increased IA current density in CRH neurons. For IM currents, only high estradiol concentrations led to an increase in current density. Despite these changes in K+ currents following estradiol manipulations, there were no significant changes in intrinsic excitability parameters. However, when we combined data from the present study with that of previous work looking at excitability in CRH neurons from intact, cycling females, we found significant correlations between IA current density and several measures of excitability. These findings differ compared with previous studies investigating estradiol effects on K+ currents in hypothalamic neurons. Hu et al. [ 2016] demonstrated that acute bath application of estradiol (100 nmol l−1, 10 min) onto CRH neurons from Ovx mice could supress IM­ currents. This effect could be replicated with a membrane-associated estrogen receptor (ER) agonist, suggesting that the fast suppression of M currents by estradiol was mediated via a non-genomic signalling mechanism (Hu et al., 2016). In other neural populations, chronic estradiol replacement in Ovx animals has also been shown to suppress K+ currents. In rostral ventrolateral medulla projecting preautonomic PVN neurons, estradiol treatment in Ovx rats was sufficient to reduce IA current density (Lee et al., 2013). In GnRH neurons, estradiol treatment in Ovx mice also decreased both IA and IK currents (DeFazio and Moenter, 2002). Previously, we have shown that I­A currents in CRH neurons are smallest during the proestrus phase and largest during the estrus stage of the mouse estrous cycle (Power and Iremonger, 2021). However, hormone profiles in intact animals will be different compared with those in Ovx animals with estradiol replacement, and this may underlie the differing findings. What signalling pathways could be responsible for the effects of estradiol on K+ channel function in CRH neurons in the present study? Estradiol acts through two main receptors, ERα and ERβ. In Ovx rats, these receptors have opposing effects on stress induced glucocorticoid secretion, with ERα increasing secretion and ERβ decreasing it (Weiser and Handa, 2009; Weiser et al., 2010). ERβ is expressed in PVN neurons and shows colocalization with CRH (Lund et al., 2006; Oyola et al., 2017). Therefore, estradiol acting through ERβ could possibly be mediating the effects observed. Comparatively, ERα shows little to no expression in mouse PVN CRH neurons (Suzuki and Handa, 2005); however, it may regulate CRH neuron function indirectly via afferent inputs (Dayas et al., 2000). There are a number of different neural populations that express ERα and project to the PVN including neurons in the arcuate nucleus (Franceschini et al., 2006; Handa and Weiser, 2014), the bed nucleus of stria terminalis (Shughrue et al., 1997) and the peri-PVN region (Weiser and Handa, 2009). In addition, estradiol manipulations are known to regulate the signalling of other neurotransmitter systems in the PVN including serotonin (McAllister et al., 2012), oxytocin (Amico et al., 1981) and vasopressin (Lagunas et al., 2019; Vilhena-Franco et al., 2019). In summary, because estradiol modulates a number of neural circuits, neurotransmitter systems and hormone systems, it is likely that changes in CRH neuron function result from a combination of direct and indirect effects of estradiol. Although the relative importance of each of these pathways for mediating the changes in K+ channel function in CRH neurons is currently unclear, we can conclude that the initial trigger for these changes is the change in circulating estradiol. Despite changes in K+ channel activity, chronic estradiol manipulations did not influence specific parameters of CRH neuron intrinsic excitability. However, the correlation analysis showed that there is a significant relationship between the IA currents and CRH excitability when data from intact, cycling animals were included. Interestingly, although estradiol replacement increases IA current density in OvxLowE and OvxHighE animals compared with Ovx, IA current density does not reach the same level as that seen in intact diestrus or estrus mice. These data suggest that the magnitude of increase in K+ current density following estradiol replacement may not have been large enough to change CRH neuron intrinsic excitability. A second reason why estradiol may not have impacted CRH neuron intrinsic excitability is homeostatic plasticity. Past research has shown that chronic manipulations of K+ channel function can induce compensatory changes in excitability known as homeostatic plasticity (Burrone et al., 2002). This form of plasticity acts to return the activity of neural circuits to a homeostatic set point (Burrone et al., 2002; Hengen et al., 2013, 2016; Keck et al., 2013; Turrigiano et al., 1998). CRH neurons may similarly have a ‘homeostatic setpoint’ firing rate, which in the intact animal is subject to a dynamically changing hormonal environment, resulting in temporary changes in excitability across the estrous cycle (Power and Iremonger, 2021). However, in a static hormonal environment, such as that seen in Ovx+estradiol treated mice, CRH neuron spiking excitability may retune to the setpoint despite differences in K+ channel activity. In order for this to happen, the function of other ion channels would need to be regulated. This hypothesis would be interesting to address in future work. Together, data from the present study show that chronic estradiol elevations lead to enhanced K+ channel currents in CRH neurons. Although there were no obvious effects on spiking excitability, we predict that enhanced K+ channel function may affect how these neurons integrate and process stress-relevant synaptic inputs. ## References 1. Amico J. A., Seif S. M., Robinson A. G.. **Oxytocin in human plasma: correlation with neurophysin and stimulation with estrogen**. (1981) **52** 988-993. DOI: 10.1210/jcem-52-5-988 2. Aoki M., Shimozuru M., Kikusui T., Takeuchi Y., Mori Y.. **Sex differences in behavioral and corticosterone responses to mild stressors in ICR mice are altered by ovariectomy in peripubertal period**. (2010) **27** 783-789. DOI: 10.2108/zsj.27.783 3. Arroyo A., Kim B. S., Biehl A., Yeh J., Bett G. C.. **Expression of kv4. 3 voltage-gated potassium channels in rat gonadotrophin-releasing hormone (GnRH) neurons during the estrous cycle**. (2011) **18** 136-144. PMID: 20861393 4. Atkinson H. C., Waddell B. J.. **Circadian variation in basal plasma corticosterone and adrenocorticotropin in the rat: sexual dimorphism and changes across the estrous cycle**. (1997) **138** 3842-3848. DOI: 10.1210/endo.138.9.5395 5. Babb J. A., Masini C. V., Day H. E., Campeau S.. **Sex differences in activated corticotropin-releasing factor neurons within stress-related neurocircuitry and hypothalamic–pituitary–adrenocortical axis hormones following restraint in rats**. (2013) **234** 40-52. DOI: 10.1016/j.neuroscience.2012.12.051 6. Burrone J., O'byrne M., Murthy V. N.. **Multiple forms of synaptic plasticity triggered by selective suppression of activity in individual neurons**. (2002) **420** 414-418. DOI: 10.1038/nature01242 7. Chen Y., Molet J., Gunn B. G., Ressler K., Baram T. Z.. **Diversity of reporter expression patterns in transgenic mouse lines targeting corticotropin-releasing hormone-expressing neurons**. (2015) **156** 4769-4780. DOI: 10.1210/en.2015-1673 8. Clarke A. G., Kendall M. D.. **Histological changes in the thymus during mouse pregnancy**. (1989) **14** 65-78. PMID: 2623742 9. Daodee S., Monthakantirat O., Tantipongpiradet A., Maneenet J., Chotritthirong Y., Boonyarat C., Khamphukdee C., Kwankhao P., Pitiporn S., Awale S.. **Effect of Yakae-Prajamduen-Jamod traditional Thai remedy on cognitive impairment in an ovariectomized mouse model and its mechanism of action**. (2022) **27** 4310. DOI: 10.3390/molecules27134310 10. Daviu N., Füzesi T., Rosenegger D. G., Rasiah N. P., Sterley T.-L., Peringod G., Bains J. S.. **Paraventricular nucleus CRH neurons encode stress controllability and regulate defensive behavior selection**. (2020) **23** 398-410. DOI: 10.1038/s41593-020-0591-0 11. Dayas C., Xu Y., Buller K., Day T.. **Effects of chronic oestrogen replacement on stress–induced activation of hypothalamic–pituitary–adrenal axis control pathways**. (2000) **12** 784-794. DOI: 10.1046/j.1365-2826.2000.00527.x 12. Defazio R. A., Moenter S. M.. **Estradiol feedback alters potassium currents and firing properties of gonadotropin-releasing hormone neurons**. (2002) **16** 2255-2265. DOI: 10.1210/me.2002-0155 13. Desroziers E., Brock O., Bakker J.. **Potential contribution of progesterone receptors to the development of sexual behavior in male and female mice**. (2017) **90** 31-38. DOI: 10.1016/j.yhbeh.2016.05.008 14. Eid R. S., Lieblich S. E., Duarte-Guterman P., Chaiton J. A., Mah A. G., Wong S. J., Wen Y., Galea L. A.. **Selective activation of estrogen receptors α and β: implications for depressive-like phenotypes in female mice exposed to chronic unpredictable stress**. (2020) **119** 104651. DOI: 10.1016/j.yhbeh.2019.104651 15. Figueiredo H. F., Ulrich-Lai Y. M., Choi D. C., Herman J. P.. **Estrogen potentiates adrenocortical responses to stress in female rats**. (2007) **292** E1173-E1182. DOI: 10.1152/ajpendo.00102.2006 16. Franceschini I., Lomet D., Cateau M., Delsol G., Tillet Y., Caraty A.. **Kisspeptin immunoreactive cells of the ovine preoptic area and arcuate nucleus co-express estrogen receptor alpha**. (2006) **401** 225-230. DOI: 10.1016/j.neulet.2006.03.039 17. Füzesi T., Daviu N., Wamsteeker Cusulin J. I., Bonin R. P., Bains J. S.. **Hypothalamic CRH neurons orchestrate complex behaviours after stress**. (2016) **7** 11937. DOI: 10.1038/ncomms11937 18. Gerrits M., Grootkarijn A., Bekkering B. F., Bruinsma M., Den Boer J. A., Ter Horst G. J.. **Cyclic estradiol replacement attenuates stress-induced c-Fos expression in the PVN of ovariectomized rats**. (2005) **67** 147-155. DOI: 10.1016/j.brainresbull.2005.06.021 19. Ghobadi N., Sahraei H., Meftahi G. H., Bananej M., Salehi S.. **Effect of estradiol replacement in ovariectomized NMRI mice in response to acute and chronic stress**. (2016) **6** 176-184. DOI: 10.7324/JAPS.2016.601128 20. Gutman G. A., Chandy K. G., Grissmer S., Lazdunski M., Mckinnon D., Pardo L. A., Robertson G. A., Rudy B., Sanguinetti M. C., Stühmer W.. **International Union of Pharmacology. LIII. Nomenclature and molecular relationships of voltage-gated potassium channels**. (2005) **57** 473-508. PMID: 16382104 21. Handa R. J., Weiser M. J.. **Gonadal steroid hormones and the hypothalamo–pituitary–adrenal axis**. (2014) **35** 197-220. DOI: 10.1016/j.yfrne.2013.11.001 22. Hellier V., Brock O., Candlish M., Desroziers E., Aoki M., Mayer C., Piet R., Herbison A., Colledge W. H., Prévot V.. **Female sexual behavior in mice is controlled by kisspeptin neurons**. (2018) **9** 1-12. DOI: 10.1038/s41467-017-02797-2 23. Hengen K. B., Lambo M. E., Van Hooser S. D., Katz D. B., Turrigiano G. G.. **Firing rate homeostasis in visual cortex of freely behaving rodents**. (2013) **80** 335-342. DOI: 10.1016/j.neuron.2013.08.038 24. Hengen K. B., Pacheco A. T., Mcgregor J. N., Van Hooser S. D., Turrigiano G. G.. **Neuronal firing rate homeostasis is inhibited by sleep and promoted by wake**. (2016) **165** 180-191. DOI: 10.1016/j.cell.2016.01.046 25. Herman J. P., Cullinan W. E.. **Neurocircuitry of stress: central control of the hypothalamo–pituitary–adrenocortical axis**. (1997) **20** 78-84. DOI: 10.1016/S0166-2236(96)10069-2 26. Hu P., Liu J., Yasrebi A., Gotthardt J. D., Bello N. T., Pang Z. P., Roepke T. A.. **Gq protein-coupled membrane-initiated estrogen signaling rapidly excites corticotropin-releasing hormone neurons in the hypothalamic paraventricular nucleus in female mice**. (2016) **157** 3604-3620. DOI: 10.1210/en.2016-1191 27. Jamieson B., Nair B., Iremonger K.. **Regulation of hypothalamic corticotropin-releasing hormone neurone excitability by oxytocin**. (2017) **29** e12532. DOI: 10.1111/jne.12532 28. Karatsoreos I. N., Bhagat S. M., Bowles N. P., Weil Z. M., Pfaff D. W., Mcewen B. S.. **Endocrine and physiological changes in response to chronic corticosterone: a potential model of the metabolic syndrome in mouse**. (2010) **151** 2117-2127. DOI: 10.1210/en.2009-1436 29. Keck T., Keller G. B., Jacobsen R. I., Eysel U. T., Bonhoeffer T., Hübener M.. **Synaptic scaling and homeostatic plasticity in the mouse visual cortex in vivo**. (2013) **80** 327-334. DOI: 10.1016/j.neuron.2013.08.018 30. Kim J., Lee S., Fang Y.-Y., Shin A., Park S., Hashikawa K., Bhat S., Kim D., Sohn J.-W., Lin D.. **Rapid, biphasic CRF neuronal responses encode positive and negative valence**. (2019a) **22** 576-585. DOI: 10.1038/s41593-019-0342-2 31. Kim J. S., Han S. Y., Iremonger K. J.. **Stress experience and hormone feedback tune distinct components of hypothalamic CRH neuron activity**. (2019b) **10** 1-15. DOI: 10.1038/s41467-018-07882-8 32. Kitay J. I.. **Effects of estradiol on pituitary-adrenal function in male and female rats**. (1963) **72** 947-954. DOI: 10.1210/endo-72-6-947 33. Kitay J. I.. **Effect of oestradiol in adrenal corticoidogenesis: an additional step in steroid biosynthesis**. (1966) **209** 808-809. DOI: 10.1038/209808a0 34. Kreisman M. J., Mccosh R. B., Tian K., Song C. I., Breen K. M.. **Estradiol enables chronic corticosterone to inhibit pulsatile luteinizing hormone secretion and suppress Kiss1 neuronal activation in female mice**. (2020) **110** 501-516. DOI: 10.1159/000502978 35. Lagunas N., Marraudino M., De Amorim M., Pinos H., Collado P., Panzica G., Garcia-Segura L. M., Grassi D.. **Estrogen receptor beta and G protein-coupled estrogen receptor 1 are involved in the acute estrogenic regulation of arginine-vasopressin immunoreactive levels in the supraoptic and paraventricular hypothalamic nuclei of female rats**. (2019) **1712** 93-100. DOI: 10.1016/j.brainres.2019.02.002 36. Lee S. K., Ryu P. D., Lee S. Y.. **Estrogen replacement modulates voltage-gated potassium channels in rat presympathetic paraventricular nucleus neurons**. (2013) **14** 134. DOI: 10.1186/1471-2202-14-134 37. Lo M. J., Chang L. L., Wang P. S.. **Effects of estradiol on corticosterone secretion in ovariectomized rats**. (2000) **77** 560-568. DOI: 10.1002/(SICI)1097-4644(20000615)77:4<560::AID-JCB4>3.0.CO;2-D 38. Lund T. D., Hinds L. R., Handa R. J.. **The androgen 5α-dihydrotestosterone and its metabolite 5α-androstan-3β, 17β-diol inhibit the hypothalamo–pituitary–adrenal response to stress by acting through estrogen receptor β-expressing neurons in the hypothalamus**. (2006) **26** 1448-1456. DOI: 10.1523/JNEUROSCI.3777-05.2006 39. Mcallister C. E., Creech R., Kimball P., Muma N. A., Li Q.. **GPR30 is necessary for estradiol-induced desensitization of 5-HT1A receptor signaling in the paraventricular nucleus of the rat hypothalamus**. (2012) **37** 1248-1260. DOI: 10.1016/j.psyneuen.2011.12.018 40. Nilsson M. E., Vandenput L., Tivesten Å., Norlén A.-K., Lagerquist M. K., Windahl S. H., Börjesson A. E., Farman H. H., Poutanen M., Benrick A.. **Measurement of a comprehensive sex steroid profile in rodent serum by high-sensitive gas chromatography-tandem mass spectrometry**. (2015) **156** 2492-2502. DOI: 10.1210/en.2014-1890 41. Owens J. W., Ashby J.. **Critical review and evaluation of the uterotrophic bioassay for the identification of possible estrogen agonists and antagonists: in support of the validation of the OECD uterotrophic protocols for the laboratory rodent**. (2002) **32** 445-520. DOI: 10.1080/20024091064291 42. Oyola M. G., Thompson M. K., Handa A. Z., Handa R. J.. **Distribution and chemical composition of estrogen receptor β neurons in the paraventricular nucleus of the female and male mouse hypothalamus**. (2017) **525** 3666-3682. DOI: 10.1002/cne.24295 43. Patchev V. K., Hayashi S., Orikasa C., Almeida O. F.. **Implications of estrogen–dependent brain organization for gender differences in hypothalamo–pituitary–adrenal regulation**. (1995) **9** 419-423. DOI: 10.1096/fasebj.9.5.7896013 44. Porteous R., Haden P., Hackwell E. C., Singline A., Herde M. K., Desai R., Handelsman D. J., Grattan D. R., Herbison A. E.. **Reformulation of PULSAR for analysis of pulsatile LH secretion and a revised model of estrogen-negative feedback in mice**. (2021) **162** bqab165. DOI: 10.1210/endocr/bqab165 45. Power E. M., Iremonger K. J.. **Plasticity of intrinsic excitability across the estrous cycle in hypothalamic CRH neurons**. (2021) **11** 1-7. DOI: 10.1038/s41598-021-96341-4 46. Roepke T. A., Qiu J., Smith A. W., Rønnekleiv O. K., Kelly M. J.. **Fasting and 17β-estradiol differentially modulate the M-current in neuropeptide Y neurons**. (2011) **31** 11825-11835. DOI: 10.1523/JNEUROSCI.1395-11.2011 47. Roy B. N., Reid R. L., Van Vugt D. A.. **The effects of estrogen and progesterone on corticotropin-releasing hormone and arginine vasopressin messenger ribonucleic acid levels in the paraventricular nucleus and supraoptic nucleus of the rhesus monkey**. (1999) **140** 2191-2198. DOI: 10.1210/endo.140.5.6684 48. Seale J., Wood S., Atkinson H., Bate E., Lightman S., Ingram C., Jessop D., Harbuz M.. **Gonadectomy reverses the sexually diergic patterns of circadian and stress-induced hypothalamic-pituitary-adrenal axis activity in male and female rats**. (2004) **16** 516-524. PMID: 15189326 49. Serova L. I., Harris H. A., Maharjan S., Sabban E. L.. **Modulation of several responses to stress by estradiol benzoate and selective estrogen receptor agonists**. (2010) **205** 253. DOI: 10.1677/JOE-10-0029 50. Shughrue P. J., Lane M. V., Merchenthaler I.. **Comparative distribution of estrogen receptor–α and–β mRNA in the rat central nervous system**. (1997) **388** 507-525. DOI: 10.1002/(SICI)1096-9861(19971201)388:4<507::AID-CNE1>3.0.CO;2-6 51. Speert D. B., Mcclennen S. J., Seasholtz A. F.. **Sexually dimorphic expression of corticotropin-releasing hormone-binding protein in the mouse pituitary**. (2002) **143** 4730-4741. DOI: 10.1210/en.2002-220556 52. Sterley T.-L., Baimoukhametova D., Füzesi T., Zurek A. A., Daviu N., Rasiah N. P., Rosenegger D., Bains J. S.. **Social transmission and buffering of synaptic changes after stress**. (2018) **21** 393-403. DOI: 10.1038/s41593-017-0044-6 53. Suzuki S., Handa R. J.. **Estrogen receptor-β, but not estrogen receptor-α, is expressed in prolactin neurons of the female rat paraventricular and supraoptic nuclei: comparison with other neuropeptides**. (2005) **484** 28-42. DOI: 10.1002/cne.20457 54. Tang A. C., Nakazawa M., Romeo R. D., Reeb B. C., Sisti H., Mcewen B. S.. **Effects of long-term estrogen replacement on social investigation and social memory in ovariectomized C57BL/6 mice**. (2005) **47** 350-357. DOI: 10.1016/j.yhbeh.2004.10.010 55. Taniguchi H., He M., Wu P., Kim S., Paik R., Sugino K., Kvitsani D., Fu Y., Lu J., Lin Y.. **A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex**. (2011) **71** 995-1013. DOI: 10.1016/j.neuron.2011.07.026 56. Tantipongpiradet A., Monthakantirat O., Vipatpakpaiboon O., Khampukdee C., Umehara K., Noguchi H., Fujiwara H., Matsumoto K., Sekeroglu N., Kijjoa A.. **Effects of puerarin on the ovariectomy-induced depressive-like behavior in ICR mice and its possible mechanism of action**. (2019) **24** 4569. DOI: 10.3390/molecules24244569 57. Turrigiano G. G., Leslie K. R., Desai N. S., Rutherford L. C., Nelson S. B.. **Activity-dependent scaling of quantal amplitude in neocortical neurons**. (1998) **391** 892-896. DOI: 10.1038/36103 58. Ulrich-Lai Y. M., Herman J. P.. **Neural regulation of endocrine and autonomic stress responses**. (2009) **10** 397-409. DOI: 10.1038/nrn2647 59. Utsuyama M., Hirokawa K.. **Hypertrophy of the thymus and restoration of immune functions in mice and rats by gonadectomy**. (1989) **47** 175-185. DOI: 10.1016/0047-6374(89)90030-4 60. Vastagh C., Solymosi N., Farkas I., Liposits Z.. **Proestrus differentially regulates expression of ion channel and calcium homeostasis genes in GnRH neurons of mice**. (2019) **12** 137. DOI: 10.3389/fnmol.2019.00137 61. Viau V., Meaney M. J.. **Variations in the hypothalamic-pituitary-adrenal response to stress during the estrous cycle in the rat**. (1991) **129** 2503-2511. DOI: 10.1210/endo-129-5-2503 62. Vilhena–Franco T., Mecawi A. S., Almeida–Pereira G., Lucio-Oliveira F., Elias L. L. K., Antunes-Rodrigues J.. **Oestradiol acts through its beta receptor to increase vasopressin neuronal activation and secretion induced by dehydration**. (2019) **31** e12712. DOI: 10.1111/jne.12712 63. Wada T., Sameshima A., Yonezawa R., Morita M., Sawakawa K., Tsuneki H., Sasaoka T., Saito S.. **Impact of central and peripheral estrogen treatment on anxiety and depression phenotypes in a mouse model of postmenopausal obesity**. (2018) **13** e0209859. PMID: 30589890 64. Wamsteeker-Cusulin J. I., Füzesi T., Watts A. G., Bains J. S.. **Characterization of corticotropin-releasing hormone neurons in the paraventricular nucleus of the hypothalamus of Crh-IRES-Cre mutant mice**. (2013) **8** e64943. DOI: 10.1371/journal.pone.0064943 65. Weiser M., Handa R. J.. **Estrogen impairs glucocorticoid dependent negative feedback on the hypothalamic–pituitary–adrenal axis via estrogen receptor alpha within the hypothalamus**. (2009) **159** 883-895. DOI: 10.1016/j.neuroscience.2008.12.058 66. Weiser M. J., Foradori C. D., Handa R. J.. **Estrogen receptor beta activation prevents glucocorticoid receptor-dependent effects of the central nucleus of the amygdala on behavior and neuroendocrine function**. (2010) **1336** 78-88. DOI: 10.1016/j.brainres.2010.03.098 67. Young E. A., Altemus M., Parkison V., Shastry S.. **Effects of estrogen antagonists and agonists on the ACTH response to restraint stress in female rats**. (2001) **25** 881-891. DOI: 10.1016/S0893-133X(01)00301-3 68. Zoller A. L., Kersh G. J.. **Estrogen induces thymic atrophy by eliminating early thymic progenitors and inhibiting proliferation of β-selected thymocytes**. (2006) **176** 7371-7378. DOI: 10.4049/jimmunol.176.12.7371
--- title: Comparing Gastroesophageal Reflux Disease (GERD) and Non-GERD Patients Based on Knowledge Level of Acute Myocardial Infarction Symptoms, Risk Factors and Immediate Action Taken in Eastern Province, Saudi Arabia journal: Cureus year: 2023 pmcid: PMC10038176 doi: 10.7759/cureus.35309 license: CC BY 3.0 --- # Comparing Gastroesophageal Reflux Disease (GERD) and Non-GERD Patients Based on Knowledge Level of Acute Myocardial Infarction Symptoms, Risk Factors and Immediate Action Taken in Eastern Province, Saudi Arabia ## Abstract Introduction: A major cause of death globally is cardiovascular disease (CVD). Chest pain, nausea, vomiting, and heartburn are common symptoms of acute myocardial infarction (AMI). Chest pain is also the main symptom of gastroesophageal reflux disease (GERD). Therefore, the differential diagnosis of chest pain can become more challenging when GERD and AMI coincide. This study evaluated and compared the knowledge of the signs, symptoms, and immediate action that must be taken regarding AMI among GERD and non-GERD patients. Methodology: An observational cross-sectional study using an online questionnaire was created and published between October and November 2022 to collect data from Saudi males and females 18 or older willing to participate in the study. Participants who were not Saudi had declined to complete the survey or had not fully completed it was excluded. The questionnaire contained three sections; after collecting informed consent, it made inquiries regarding respondents’ GERD status, demographic information, and knowledge and attitudes regarding GERD. Results: This study included 691 responses from 300 non-GERD participants and 391 GERD participants. The study showed a high level of awareness ($75.5\%$) of GERD, with significant differences in the level of awareness according to marital status, education levels, and occupation status. There was no significant difference in the level of awareness according to gender and GERD diagnosis, where the p-value > 0.05. The most common source of information about AMI was the Internet, followed by health care professionals. The most commonly known symptoms of AMI were sudden pain or discomfort in the chest, followed by a sudden shortness of breath. Additionally, there was no significant association between the diagnosis of GERD and known risk factors. The association between GERD and other diseases (chi-square = 46.94, p-value 0.01). Obesity and smoking were the two main risk factors for heart attacks. Conclusion: This study demonstrated that there was no significant difference between GERD and non-GERD participants regarding the knowledge and awareness level of AMI. Moreover, it showed that there was a lack of general knowledge and awareness of AMI in Saudi Arabia. The authors recommend initiating more awareness programs in Saudi Arabia to inform people about AMI and cardiovascular disease. More research is required to determine whether other patients are aware of AMI. ## Introduction Cardiovascular diseases (CVD) are one of the leading causes of mortality worldwide, with approximately 17.9 million deaths reported annually [1]. Saudi Arabia reported a rise in CVD rates in recent years, with an overall prevalence of $5.5\%$ [2]. Additionally, acute coronary syndrome (ACS) death rates in Saudi Arabia were $4\%$, $5.8\%$, and $8.1\%$ in-hospital, at one month, and at one year, respectively [3]. Gastroesophageal reflux disease (GERD) results from the regurgitation of gastric contents into the esophagus [4]. More than $40\%$ of adults in the United States suffer from GERD each month [5]. In Saudi Arabia, the estimated range of GERD prevalence is between $23.47\%$ and $45.4\%$ [6]. The classic symptoms of acute myocardial infarction (AMI) include chest pain, nausea, vomiting, and heartburn. These symptoms are also associated with GERD [7]. Moreover, the coexistence of GERD and AMI at the same time can make the differential diagnosis of chest pain more complicated [8]. GERD shares similar risk factors with AMI that include obesity, diabetes, hypertension, and hyperlipidemia, along with behavioral risk factors, including smoking and alcohol use [9-12]. Recent studies in Saudi Arabia showed a suboptimal awareness level of coronary artery disease (CAD) risk factors [13,14]. Since GERD is the most common cause of non-cardiac chest pain, it is crucial to assess public knowledge about the differences between heart attack and heartburn. No such work appears to have been published regarding such awareness among GERD patients. This study assesses and compares the understanding of early signs, symptoms, and appropriate responses in cases of AMI among GERD and non-GERD patients in Saudi Arabia. ## Materials and methods Study design and selection criteria An observational cross-sectional study was carried out using an online questionnaire to obtain responses from male and female citizens of Saudi Arabia who are older than 18 years of age and willing to participate in the study. Non-Saudi citizen participants, people who declined participation, and people who did not complete the entire questionnaire were excluded. Questionnaire design The questionnaire was designed in Arabic, as it is the native language of Saudi Arabia, and distributed randomly via Google Forms using social media platforms such as WhatsApp, Twitter, and Telegram. The questionnaire consisted of three sections and was presumed to take approximately three minutes to be completed. The questionnaire was developed and published between October and November 2022 and generated 691 responses, with 391 suffering from GERD who were self-diagnosed based on their symptoms and 300 who have never experienced such symptoms. The first section began with gathering informed consent from participants, followed by a query regarding the respondent’s status with regard to GERD. The second section addressed demographic data, including age, gender, marital status, educational level, and occupational status of respondents. Finally, the third section contained questions regarding the knowledge and attitude of GERD respondents' knowledge of and attitude toward AMI among GERD and non-GERD patients in Saudi Arabia. Ethical consideration and statistical analysis This study was approved by the Ethics Committee of King Faisal University with an ethical approval code KFU-REC-2022-SEP-ETHICS172. Participants were given a statement guaranteeing that their confidential information and privacy would be protected before proceeding to the questionnaire. Completion and submission of the questionnaire were considered as consent for inclusion in the study. Data were extracted, reviewed, coded, and entered into IBM Statistical Package for the Social Sciences (SPSS) software, version 22 (SPSS, Inc., Chicago, IL). ## Results The results showed a significant difference in awareness level according to marital status; the highest awareness level was among single people and the lowest awareness was among widowed people ($F = 3.602$, p-value = 0.013). By educational level, the highest level of awareness was among bachelors and the lowest awareness level was among people with diplomas ($F = 8.328$, p-value < 0.01). By occupational status, the highest awareness level was among students and the lowest was among housewives ($F = 8.328$, p-value < 0.01). There were no significant differences in awareness according to gender or GERD diagnosis (p-value > 0.05) (Table 1). There were 691 participants between the ages of 15 and 86 (mean = 31.7, standard deviation = 13.66). The majority ($56.4\%$) were female, and $43.3\%$ were male. Regarding marital status: the majority ($55.9\%$) were single, $42.5\%$ were married, $1\%$ were divorced, and $0.6\%$ were widowed. On an educational level, the majority ($58.8\%$) had a bachelor’s degree, $21\%$ had secondary education, $8.4\%$ had a diploma, $5.2\%$ had postgraduate education, $4.8\%$ had primary education, and $1.9\%$ had intermediate education. The majority ($41.8\%$) of respondents were students, $29.2\%$ were employees, $11.3\%$ were retired, $9.7\%$ were housewives, and $8\%$ were looking for a job. The majority ($53.5\%$) earned monthly incomes of less than 3,000 SR, $23.4\%$ had incomes from 3,000 SR to 10,000 SR, $15.5\%$ had incomes from 10,000 SR to 20,000 SR, and $7.5\%$ had incomes of more than 20,000 SR (Table 2). Results show that $56.6\%$ have GERD diagnoses and $43.4\%$ do not. Among people with GERD diagnoses, $37.6\%$ had no other diseases, $5.2\%$ had diabetes, $4.1\%$ had dyslipidemia, $3.9\%$ had hypertension, $1.4\%$ had heart diseases, 0.45 had suffered a stroke, and $3.9\%$ had other diseases. $51.4\%$ of this cohort were aware of heart attacks and $5.2\%$ had not. $38.9\%$ of the people knew someone who had experienced a heart attack before and $17.7\%$ did not. $39.8\%$ had received information related to heart attacks and $16.8\%$ had not. $52\%$ knew that sudden heart attack requires prompt treatment and $4.6\%$ did not. $28.4\%$ would call an ambulance if they witnessed heart attack signs/symptoms, $24.6\%$ would take the patient to the hospital, $2.9\%$ will call a doctor and $0.7\%$ would contact the patient’s family. $47.95\%$ knew the phone number to contact an ambulance and $8.7\%$ did not. $44.4\%$ had heard about the risk factors of heart attack and $2.2\%$ had not. Among people without GERD/heartburn diagnoses, $37.9\%$ did not suffer from any diseases, $1.4\%$ had diabetes, $0.4\%$ had dyslipidemia, $0.7\%$ had hypertension, $0.9\%$ had heart disease, none had suffered a stroke, and $14\%$ had other diseases. $40.1\%$ knew about heart attacks and 3.3 % did not. $25.3\%$ knew someone who had experienced a heart attack and $18.1\%$ did not. $29.1\%$ had received information related to heart attacks and $14.3\%$ had not. $39.8\%$ knew sudden heart attacks require prompt treatment and $3.6\%$ did not. $21.1\%$ would call an ambulance if they witnessed heart attack signs/symptoms, $20.7\%$ would the patient to the hospital, $9\%$ would call a doctor, and $0.3\%$ would contact the patient’s family. 34.7 knew the phone number to contact an ambulance and $8.7\%$ did not. $32.4\%$ had heard about the risk factors of heart attack and $11\%$ had not (Table 3). Levels of awareness among respondents regarding heart attack information were high (mean: $75.5\%$). The most identified item was that sudden heart attacks require prompt treatment ($91.8\%$), then hearing about heart attacks ($91.5\%$), and the least-known item was what to do first when witnessing signs/symptoms of heart attack ($49.5\%$) (Table 4). There was a significant association between GERD diagnoses and other diseases (chi-square = 46.94, p-value < 0.01). There were also significant associations between this diagnosis and knowing someone who had experienced a heart attack before (chi-square = 8.094, p-value = 0.004). However, there were no significant associations between this diagnosis and other factors (hearing about heart attacks, receiving information related to heart attacks, knowing the need for prompt treatment for sudden heart attacks, knowing the proper actions to take when witnessing signs/symptoms of heart attack, knowing the ambulance phone number). The p-value for these factors was > 0.05 (Table 5). The most popular source of information about heart attack was the Internet ($28.1\%$), followed by health care professionals ($21\%$), books ($15.8\%$), TV ($13.9\%$), media ($13.6\%$), seminars ($4.8\%$), and promotional leaflets ($2.8\%$) (Table 6). The most identified risk factor for heart attack among respondents was obesity ($14.5\%$), followed by smoking ($13.9\%$), heart diseases ($11.6\%$), high cholesterol ($9.8\%$), stress ($9.1\%$), diabetes ($8.4\%$), unhealthy diet ($8\%$), lack of exercise ($7.2\%$), alcohol ($6.1\%$), genetics ($5.5\%$), atrial fibrillation ($4.6\%$), and exercise ($0.9\%$). $0.5\%$ were not aware of any risk factors (Table 7). The most frequently identified symptoms of heart attack were sudden pain or discomfort in the chest ($26.2\%$), sudden shortness of breath ($21.2\%$), sudden pain or discomfort in the arms or shoulders ($16.9\%$), sudden pain or discomfort in jaw, neck, or back ($15.2\%$), weakness or dizziness ($11.2\%$) and sudden disturbance of the vision in one or both eyes ($9.2\%$) (Table 8). ## Discussion The current study assessed the awareness of early signs and symptoms of AMI among GERD and non-GERD patients in Saudi Arabia and their knowledge of the best course of action to be taken from the onset of symptoms. The result showed that $43.6\%$ of participants were male and $56.4\%$ were female. Out of the respondents, $56.6\%$ had received a diagnosis of GERD, and $43.3\%$ had not received such a diagnosis. The result demonstrated a high level of awareness toward AMI in both groups of respondents (GERD and non-GERD), which was higher than expected. Moreover, there was a significant difference in the level of awareness according to an educational level only. Conversely, there were no significant differences between males and females and between GERD/non-GERD diagnoses. The knowledge of proper action in response to AMI was low. The result showed a high level of knowledge of AMI in GERD patients. This might be due to the disease itself; it could encourage patients to be more aware of complications and other diseases that share similar risk factors. The present findings seem to be consistent with other research, which found high levels of knowledge of AMI ($60.4\%$) among hypertensive patients [15]. Similarly, a study conducted in Nepal illustrated that more than half of the participants knew that AMI could be a complication of diabetes mellitus [16]. Additionally, there were no significant differences between participants who had GERD and those who did not have GERD regarding knowledge of AMI. This might be explained by the fact that most participants had a high level of education. The most common symptoms identified by the participants were sudden pain in the chest, followed by a sudden shortness of breath and sudden pain in the arm or shoulder. However, a study in the United States reported that the most common symptoms of AMI are chest pain and jaw, neck, or back pain. Furthermore, the knowledge of the most common symptoms of AMI was higher in that study compared to this one [17]. The most prevalent risk factors recognized among respondents were smoking ($13.90\%$) and obesity ($14.50\%$). This aligns with a study that was conducted in the western region of Saudi Arabia, where $66.7\%$ of respondents identified smoking as a risk factor [18]. Moreover, two other studies in Saudi Arabia and Kuwait reported very high awareness levels of smoking as a risk factor, reaching over $95\%$ of respondents [19,20]. The results showed that the level of knowledge regarding the appropriate action to be taken in the presence of AMI symptoms ($49.5\%$) was low compared with Korea ($67.0\%$) [21], Poland ($87.4\%$) [22], and the United States ($86.8\%$) [17]. Limitations This study’s strength was its high enrollment of participants who had been diagnosed with GERD or heartburn; this strengthened the result. However, there were some limitations in the study. First, the database did not contain data on the source of participants’ GERD diagnoses and whether they had been made by a doctor or by the participants themselves. Second, Saudi residents in Saudi Arabia were included in the study, but residents of other nationalities and Saudis residing abroad were not included. Third, the majority of the participants were of young age; the mean age was 31.6, with a standard deviation of 13.66. If there was a greater diversity in respondent age, the study might have shown additional information. Future studies should take these limitations into consideration. Recommendation Based on the findings of this study, annual awareness and education campaigns would be appropriate, with a focus on GERD patients and high-risk people, in order to raise knowledge of AMI symptoms and the appropriate course of action that should be taken if symptoms develop. ## Conclusions The study showed that the overall knowledge and awareness of AMI were suboptimal. The majority of participants had previously heard about AMI, but they were unsure regarding the proper first step to be taken when witnessing signs or symptoms. There was no significant difference in awareness levels among people who reported GERD diagnoses and those who did not. The authors suggest launching additional awareness campaigns about AMI and what to do next and educating people about cardiovascular disease in Saudi Arabia, especially among members of high-risk groups. More studies are needed to identify awareness of GERD complications and also to detect awareness regarding AMI among other patients. ## References 1. **Non-communicable diseases**. (2022) 2. Al-Nozha MM, Arafah MR, Al-Mazrou YY. **Coronary artery disease in**. *Saudi Med J* (2004) **25** 1165-1171. PMID: 15448760 3. Alhabib KF, Kinsara AJ, Alghamdi S. **The first survey of the Saudi Acute Myocardial Infarction Registry Program: main results and long-term outcomes (STARS-1 Program)**. *PLoS One* (2019) **14** 0 4. Clarrett DM, Hachem C. **Gastroesophageal reflux disease (GERD)**. *Mo Med* (2018) **115** 214-218. PMID: 30228725 5. Katz PO, Gerson LB, Vela MF. **Guidelines for the diagnosis and management of gastroesophageal reflux disease**. *Am J Gastroenterol* (2013) **108** 308-328. PMID: 23419381 6. Alkhathami AM, Alzahrani AA, Alzhrani MA, Alsuwat OB, Mahfouz ME. **Risk factors for gastroesophageal reflux disease in Saudi Arabia**. *Gastroenterology Res* (2017) **10** 294-300. PMID: 29118870 7. Khan S, Memon RA, Khan MF, Dar MI. **Misconception of heart attack pain with heart burn - how common is it?**. *Pak Heart J* (2017) **50** 89-93 8. Lei WY, Wang JH, Wen SH. **Risk of acute myocardial infarction in patients with gastroesophageal reflux disease: a nationwide population-based study**. *PLoS One* (2017) **12** 0 9. Hampel H, Abraham NS, El-Serag HB. **Meta-analysis: obesity and the risk for gastroesophageal reflux disease and its complications**. *Ann Intern Med* (2005) **143** 199-211. PMID: 16061918 10. Wu JC, Mui LM, Cheung CM, Chan Y, Sung JJ. **Obesity is associated with increased transient lower esophageal sphincter relaxation**. *Gastroenterology* (2007) **132** 883-889. PMID: 17324403 11. Moki F, Kusano M, Mizuide M. **Association between reflux oesophagitis and features of the metabolic syndrome in Japan**. *Aliment Pharmacol Ther* (2007) **26** 1069-1075. PMID: 17877514 12. Okwuosa TM, Klein O, Chan C, Jenny NS, Schreiner P, Green D, Liu K. **13-year long-term associations between changes in traditional cardiovascular risk factors and changes in fibrinogen levels: the Coronary Artery Risk Development in Young Adults (CARDIA) study**. *Atherosclerosis* (2013) **226** 214-219. PMID: 23177973 13. Almalki MA, AlJishi MN, Khayat MA, Bokhari HF, Subki AH, Alzahrani AM, Alhejily WA. **Population awareness of coronary artery disease risk factors in Jeddah, Saudi Arabia: a cross-sectional study**. *Int J Gen Med* (2019) **12** 63-70. PMID: 30666149 14. Basham K, Aldubaikhi A, Sulaiman I. **Public awareness of early symptoms of acute myocardial infarction among Saudi population**. *J Family Med Prim Care* (2021) **10** 3785-3790. PMID: 34934681 15. Adhikari A, Bhandari B. **Awareness and practice of prevention of myocardial infarction among hypertensive patients: cross-sectional study in a tertiary cardiac centre of Nepal**. *Nepalese Heart J* (2017) **14** 29-30 16. Karki RKC, Pandey A. **Awareness regarding myocardial infarction among diabetic patients attending in a tertiary level hospital**. *Acta Sci Med Sci* (2019) **3** 131-136 17. Fang J, Gillespie C, Keenan NL, Greenlund KJ. **Awareness of heart attack symptoms among US adults in 2007, and changes in awareness from 2001 to 2007**. *Future Cardiol* (2011) **7** 311-320. PMID: 21627473 18. Albugami S, Al-Husayni F, Bakhsh L. **The perception of coronary artery disease and cardiac catheterization in Saudi Arabia: “What the public know”**. *Cureus* (2020) **12** 0 19. Wartak SA, Friderici J, Lotfi A, Verma A, Kleppel R, Naglieri-Prescod D, Rothberg MB. **Patients' knowledge of risk and protective factors for cardiovascular disease**. *Am J Cardiol* (2011) **107** 1480-1488. PMID: 21414599 20. Awad A, Al-Nafisi H. **Public knowledge of cardiovascular disease and its risk factors in Kuwait: a cross-sectional survey**. *BMC Public Health* (2014) **14** 1131. PMID: 25367768 21. Kim HS, Lee H, Kim K. **The general public’s awareness of early symptoms of and emergency responses to acute myocardial infarction and related factors in South Korea: a national public telephone survey**. *J Epidemiol* (2016) **26** 233-241. PMID: 26853101 22. Kopec G, Sobien B, Podolec M. **Knowledge of a patient-dependant phase of acute myocardial infarction in Polish adults: the role of physician's advice**. *Eur J Public Health* (2011) **21** 603-608. PMID: 20709780
--- title: Prospective associations between accelerometry-derived physical activity and sedentary behaviors and mortality among cancer survivors authors: - Elizabeth A Salerno - Pedro F Saint-Maurice - Fei Wan - Lindsay L Peterson - Yikyung Park - Yin Cao - Ryan P Duncan - Richard P Troiano - Charles E Matthews journal: JNCI Cancer Spectrum year: 2023 pmcid: PMC10038185 doi: 10.1093/jncics/pkad007 license: CC BY 4.0 --- # Prospective associations between accelerometry-derived physical activity and sedentary behaviors and mortality among cancer survivors ## Body The number of individuals living beyond a cancer diagnosis, herein referred to as cancer survivors, in the United *States is* projected to reach 22.2 million by 2030 [1]. Consistent and compelling evidence demonstrates the benefits of recreational (ie, leisure-time) physical activity (PA) during long-term survivorship, including improved physical function, cardiorespiratory fitness, and psychosocial health [2-4]. Recreational PA is also statistically significantly associated with improved survival after cancer [5]; a recent roundtable report from the American College of Sports Medicine reported a consistent inverse association between higher levels of postdiagnosis PA and risk of all-cause mortality [4]. However, much of this work is confined to breast, prostate, and colorectal cancer and is based almost entirely on self-reported measures of PA [6]. Self-reported PA assessment is crucial for cost-effective and widespread surveillance and sufficiently estimates MVPA levels at the population level [7]. These assessments are prone to bias [8], however, which could contribute to overreporting MVPA and underreporting sedentary behaviors [9] and attenuate their associations with mortality [10]. PA questionnaires often focus on MVPA to mirror federal PA guidelines and therefore may not assess light-intensity PA behaviors, such as activities of everyday living (eg, household activities, shopping, caregiving). These lighter-intensity behaviors are often more difficult to recall and report than exercise-specific behaviors. They are also more common during survivorship given renewed cancer-specific recommendations to avoid inactivity if unable to meet PA guidelines [3,11] and high rates of cancer-related fatigue [12] that may limit participation in more strenuous activities. Understanding the potential survival benefits associated with multiple PA intensities (ie, light, MVPA, total) is an important and necessary step toward personalized PA recommendations after cancer. One solution to the limitations of self-reported PA assessments is to use accelerometers. Waist-worn monitors capture bodily acceleration that is summarized over specific epochs (eg, 1 minute), which can then characterize the intensity, duration, and total volume of daily activity. The National Health and Nutrition Examination Survey (NHANES) 2003-2004 and 2005-2006 cycles included accelerometry in representative samples of US adults and followed individuals for subsequent mortality. Thraen-Borowski and colleagues [13] recently reported that cancer survivors engaged in less accelerometry-derived light PA and MVPA and more sedentary behavior than matched adults, which is consistent with previous NHANES analyses in cancer survivors [14-18]. To date, there is no comprehensive analysis, to our knowledge, of the relationship between accelerometry-derived PA and sedentary behavior and mortality among cancer survivors in a representative sample of US adults. To address this gap, we investigated if PA of other intensities (ie, not just MVPA) and sedentary behavior were associated with mortality in NHANES cancer survivors. With over 10 years of follow-up, multiple indicators of health status, and a wide variety of cancer types, this analysis represents a prime opportunity to advance our understanding of the relationships between PA, sedentary behaviors, and all-cause mortality among long-term cancer survivors. ## Abstract ### Background Survival benefits of self-reported recreational physical activity (PA) during cancer survivorship are well-documented in common cancer types, yet there are limited data on the associations between accelerometer-derived PA of all domains, sedentary behavior, and mortality in large, diverse cohorts of cancer survivors. ### Methods Participants included adults who reported a cancer diagnosis in the National Health and Nutrition Examination Survey and wore an accelerometer for up to 7 days in 2003-2006. Participants were followed for subsequent mortality through 2015. We examined the association of light PA, moderate to vigorous PA, total PA, and sedentary behavior, with all-cause mortality. Cox proportional hazards models estimated hazard ratios (HRs) and $95\%$ confidence intervals (CIs), adjusting for demographics and health indicators. ### Results A total of 480 participants (mean age of 68.8 years [SD = 12.4] at the time of National Health and Nutrition Examination Survey assessment) reported a history of cancer. A total of 215 deaths occurred over the follow-up period. For every 1-h/d increase in light PA and moderate to vigorous PA (MVPA), cancer survivors had $49\%$ (HR = 0.51, $95\%$ CI = 0.34 to 0.76) and $37\%$ (HR = 0.63, $95\%$ CI = 0.40 to 0.99) lower hazards of all-cause mortality, respectively. Total PA demonstrated similar associations with statistically significantly lower hazards of death for each additional hour per day (HR = 0.68, $95\%$ CI = 0.54 to 0.85), as did every metabolic equivalents of task-hour per day increase in total PA estimations of energy expenditure (HR = 0.88, $95\%$ CI = 0.82 to 0.95). Conversely, more sedentary time (1 h/d) was not associated with statistically significantly higher hazards (HR = 1.08, $95\%$ CI = 0.94 to 1.23). ### Conclusions These findings reinforce the current recommendations for cancer survivors to be physically active and underscore the continued need for widespread PA promotion for long-term survival in older cancer survivors. ## Participants and study design NHANES collects extensive health data from a representative sample of US adults [19]. Details on data collection in NHANES have been reported elsewhere [20,21]. Briefly, NHANES 2003-2004 and 2005-2006 included a representative sample of noninstitutionalized US adults. Between 2003 and 2006, participants were assessed for PA levels and self-reported up to 4 cancer diagnoses and corresponding ages at diagnosis. The first chronological cancer reported was considered the primary cancer ($$n = 630$$). Nonmelanoma skin primary cancers were removed ($$n = 150$$), leaving a final analytic sample of 480 adult cancer survivors (Table 1). Participants were followed for subsequent mortality through December 2015. NHANES protocols were approved by the National Center for Health Statistics ethics review board, and all participants provided written informed consent. This analysis was not subject to institutional review board review based on National Institutes of Health policy because it consisted of deidentified data with no direct participant contact; thus, it is not human subjects research. **Table 1.** | Cancer type | Frequency No. (%) | | --- | --- | | Breast | 102 (21.3) | | Prostate | 93 (19.4) | | Othera | 63 (13.1) | | Colon | 40 (8.3) | | Melanoma | 40 (8.3) | | Cervical | 37 (7.7) | | Uterine | 25 (5.2) | | Lymphoma | 14 (2.9) | | Multiple primariesb | 14 (2.9) | | Lung | 13 (2.7) | | Bladder | 11 (2.3) | | Kidney | 11 (2.3) | | Thyroid | 11 (2.3) | | Unknown/missing | 6 (1.3) | ## Accelerometry From 2003 to 2006, participants were asked to wear an Actigraph model 7164 accelerometer on the nondominant hip during all waking hours for a 7-day period [19]. PA data retained for analysis met wear time validation criteria of at least 10 hours of wear time per day for at least 1 day, with nonwear time defined using an automated algorithm [19]. Sedentary time was defined as hours per day spent below 100 counts per minute (cpm), and total PA time was defined as hours per day spent at or above 100 cpm [22]. Light-intensity PA was defined as hours per day between 100 and 759 cpm, and MVPA was defined as hours per day spent at or greater than 760 cpm [23]. We further explored estimations of energy expenditure, calculated from metabolic equivalents of task (MET) hours using the Freedson equation [METS/min = 1.439008 + 0.000795 × count/min (vertical axis)] [24] and total PA as time recorded above 100 cpm. ## Mortality All-cause mortality was assessed through linkage to the National Death Index through December 31, 2015 [25]. International Classifications of Diseases (ICD)-9 and ICD-10 codes were used to classify deaths due to all causes. Person-years accrued from the interview date to the date of death or censoring (December 31, 2015), whichever came first. ## Covariates Covariates included age, sex, race, education, smoking and alcohol status, body mass index (BMI), diet quality, chronic conditions, mobility, health status, frailty, cancer type, and time since diagnosis. Demographic information (age, sex, education), health behaviors (smoking status), and diagnoses of chronic conditions (diabetes, heart disease, heart failure, stroke, chronic bronchitis, emphysema) were self-reported. Race and ethnicity were also self-reported using fixed categories (Mexican American, Non-Hispanic Black, Non-Hispanic White, Other Hispanic, and Other [including Alaska Native, Asian, other Hispanic, and other race and ethnicity including multiracial]) to characterize the population and oversample Mexican American and Non-Hispanic Black adults. Height and weight were measured, and BMI was calculated using the standard kg/m2 equation. Diet quality was measured using 24-hour recall measures of 12 dietary components from the Healthy Eating Index-2005 [26] (range 0-100; higher scores indicate healthier diet). Mobility limitations were assessed through reported difficulty walking 0.25 miles without special equipment or up 10 steps in adults aged at least 60 years. Participants younger than 60 years were assessed for mobility limitations if they reported limitations related to work, memory problems, or other physical or mental limitations. Self-reported health was measured with the question “Would you say your health in general is excellent, very good, good, fair, or poor?” A frailty index was created based on the concept of deficit accumulation [27] and has been described in detail elsewhere [28]. Briefly, this index was derived from 38 self-reported and clinically assessed health indicators. Items were summed and divided by the number of available items; individuals were then categorized as robust (≤0.10), vulnerable (0.11 to 0.21), frail (0.22 to 0.45), or most frail (>0.45) [29]. Time since diagnosis was calculated by subtracting current age from the age reported at primary cancer diagnosis. ## Statistical analysis Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and $95\%$ confidence intervals (CIs), adjusting for covariates based on previous research [28,30], and included age, sex, race and ethnicity, education, diet, smoking status, BMI, self-reported health, mobility limitations, and diagnoses of diabetes, stroke, heart disease, heart failure, chronic bronchitis, and emphysema. Missingness for any given covariate was minimal (≤$5\%$) and thus treated as missing in all models. The proportional hazards assumption for key exposures was graphically checked using Schoenfeld residual and Kaplan-Meier plots. PA variables were modeled both continuously via 1-hour intervals per day and categorically via quartiles. To examine possible effect modification, we conducted stratified analyses by sex (male, female), age at time of NHANES assessment (median split at 71 years), BMI (<25, 25 to <30, ≥30), frailty status (robust or vulnerable, frail or most frail), health status (very good or excellent, good, fair or poor), mobility limitations (yes, no), chronic conditions (0, ≥1), weight change over the past year (losers, gainers or maintainers), time from diagnosis to accelerometry monitoring (≤2 years, >2 years), and time from monitoring to death (<5 years, ≥5 years). Further, post hoc sensitivity analyses investigated the possibility of reverse causality, as recommended by Strain and colleagues [31], through exclusions of individuals who 1) reported 1 or more chronic condition; 2) reported 2 or more chronic conditions; 3) were most frail; 4) lost weight over the past year; 5) died within 1 year of monitoring, and; 6) died within 2 years of monitoring. Finally, we explored differential associations by cancer type (major: breast, prostate, colon; minor: all others) using both exclusions and stratifications [32]. Total PA was used for post hoc sensitivity analyses, measured continuously. All analyses were conducted in SAS 9.4 and SUDAAN, incorporating sample weights as recommended by the National Center for Health Statistics [33] to account for survey cycles, strata, and sampling units; statistical significance was set at.05. ## Results Table 2 details participant characteristics. On average, participants were overweight (BMI mean = 27.9 kg/m2, SD = 5.8) and frail ($63.9\%$ frail or most frail status) with a common cancer ($49.0\%$ breast, prostate, or colon) (Table 2). Participants were followed-up for an average of 12.0 (SD = 11.8) years after their cancer diagnosis, and average follow-up time between monitoring and death or censoring was 8.4 (SD = 3.7) years (Table 2). A total 215 deaths occurred between monitoring and death or censoring. Compared with those in the lowest quartile of MVPA, participants in the highest quartile were statistically significantly younger ($P \leq .001$), less frail ($P \leq .001$), and healthier (eg, fewer mobility limitations and chronic conditions). Participants in the highest quartile of sedentary behavior were more likely to be older (>71 years; $P \leq .001$), male ($$P \leq .001$$), frailer ($P \leq .001$), and less healthy (eg, more mobility limitations and chronic conditions) compared with those in the lowest quartile. **Table 2.** | Characteristic | Unnamed: 1 | Light PA | Light PA.1 | Light PA.2 | MVPA | MVPA.1 | MVPA.2 | Sedentary behavior | Sedentary behavior.1 | Sedentary behavior.2 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Characteristic | Full sample, % | Lowest quartile, %(n = 120) | Highest quartile, %(n = 120) | P | Lowest quartile, %(n = 121) | Highest quartile, %(n = 120) | P | Lowest quartile, %(n = 121) | Highest quartile, %(n = 120) | P | | Characteristic | (n = 480) | Lowest quartile, %(n = 120) | Highest quartile, %(n = 120) | P | Lowest quartile, %(n = 121) | Highest quartile, %(n = 120) | P | Lowest quartile, %(n = 121) | Highest quartile, %(n = 120) | P | | Mean age (SD), y | 68.8 (12.4) | 74.9 (9.0) | 65.9 (12.8) | <.001 | 75.9 (8.9) | 60.7 (11.7) | <.001 | 62.2 (12.8) | 73.3 (10.3) | <.001 | | Mean BMIa (SD), kg/m2 | 27.9 (5.8) | 28.1 (6.5) | 27.7 (5.8) | .58 | 27.5 (6.2) | 27.5 (5.1) | .46 | 27.9 (5.5) | 27.9 (6.6) | .36 | | Race and ethnicity | | | | .86 | | | .05 | | | .002 | | Mexican American | 5.4 | 4.2 | 7.5 | | 3.3 | 6.7 | | 10.7 | 1.7 | | | Non-Hispanic | | | | | | | | | | | | Black | 15.6 | 14.2 | 16.7 | | 18.2 | 14.2 | | 13.2 | 21.7 | | | White | 74.8 | 79.2 | 70.8 | | 72.7 | 75.8 | | 71.9 | 72.5 | | | Other Hispanic | 1.3 | 0.8 | 1.7 | | 0.8 | 1.7 | | 1.7 | 0.8 | | | Other groupsb | 2.9 | 1.7 | 3.3 | | 5.0 | 1.7 | | 2.5 | 3.3 | | | Sex | | | | .05 | | | .95 | | | .001 | | Female | 55.0 | 49.2 | 61.7 | | 56.2 | 55.8 | | 66.9 | 46.7 | | | Male | 45.0 | 50.8 | 38.3 | | 43.8 | 44.2 | | 33.1 | 53.3 | | | Self-reported health status | | | | .36 | | | .08 | | | .47 | | Excellent | 6.9 | 6.7 | 3.3 | | 6.6 | 9.2 | | 5.8 | 9.2 | | | Very good | 24.4 | 19.2 | 26.7 | | 15.7 | 28.3 | | 22.3 | 20.8 | | | Good | 37.5 | 31.7 | 44.2 | | 30.6 | 36.7 | | 37.2 | 35.8 | | | Fair/poor | 26.7 | 38.3 | 21.7 | | 41.3 | 20 | | 29.8 | 30.8 | | | Frailty index | | | | <.001 | | | <.001 | | | <.001 | | Robust | 4.8 | 1.7 | 3.3 | | 1.7 | 12.5 | | 9.9 | 1.7 | | | Vulnerable | 31.3 | 19.2 | 40.8 | | 11.6 | 50.0 | | 38.8 | 24.2 | | | Frail | 51.0 | 53.3 | 45.8 | | 57.9 | 34.2 | | 44.6 | 55.8 | | | Most frail | 12.9 | 25.8 | 10.0 | | 28.9 | 3.3 | | 6.6 | 18.3 | | | History of | | | | | | | | | | | | Congestive heart disease | 9.4 | 17.5 | 5.8 | <.01 | 18.2 | 5.8 | .01 | 7.4 | 14.2 | .09 | | Stroke | 8.5 | 14.2 | 8.3 | .15 | 12.4 | 6.7 | .13 | 4.1 | 8.3 | .18 | | Diabetes | 15.2 | 24.2 | 15.8 | .11 | 24.0 | 8.3 | .001 | 11.6 | 20.0 | .05 | | Heart failure | 8.3 | 13.3 | 6.7 | .59 | 15.7 | 4.2 | .59 | 5.8 | 12.5 | .90 | | Chronic bronchitis | 12.5 | 11.7 | 14.2 | .57 | 15.7 | 10.8 | .31 | 9.9 | 15.0 | .23 | | Emphysema | 7.5 | 12.5 | 4.2 | .04 | 14.0 | 5.0 | .27 | 4.1 | 10.0 | .08 | | Mobility limitations | 38.8 | 57.5 | 29.2 | <.001 | 65.3 | 16.7 | <.001 | 28.9 | 51.7 | .01 | | Chronic conditions | | | | <.001 | | | <.001 | | | .03 | | 0 | 66.9 | 54.2 | 67.5 | | 55.4 | 76.7 | | 72.7 | 59.2 | | | ≥1 | 33.1 | 45.8 | 32.5 | | 44.6 | 23.3 | | 27.3 | 40.8 | | | Weight change | | | | .19 | | | .09 | | | .09 | | Losers | 37.4 | 39.2 | 31.7 | | 43.0 | 32.5 | | 34.7 | 44.2 | | | Gainers/maintainers | 62.6 | 58.3 | 67.5 | | 57.0 | 67.5 | | 65.3 | 55.8 | | | Mean time between diagnosis and monitoring (SD), y | 12.0 (11.8) | 12.7 (12.7) | 12.8 (11.2) | .97 | 12.7 (12.0) | 12.1 (10.3) | .65 | 11.8 (11.0) | 12.9 (13.1) | .49 | | Mean time between monitoring and death (SD), mo | 8.4 (3.7) | 6.7 (3.9) | 9.3 (3.1) | <.001 | 6.2 (3.9) | 10.0 (2.6) | <.001 | 9.6 (3.0) | 7.7 (3.6) | <.001 | ## PA and sedentary behavior Table 3 details hazard ratios and $95\%$ confidence intervals for age-adjusted, multivariable, and MVPA-adjusted models. Figure 1 depicts the quartile associations with P trends. More light-intensity PA was associated with statistically significantly lower risk of death, such that mortality risk was lower with each increasing hour (HR = 0.51, $95\%$ CI = 0.34 to 0.76) and quartile (Ptrend =.002) of light PA. MVPA demonstrated similar associations, both continuously for every additional hour of MVPA (HR = 0.63, $95\%$ CI = 0.40 to 0.99) and categorically via quartiles (Ptrend =.004). When exploring these associations by total PA, each increasing hour per day of activity of any intensity was associated with statistically significantly lower hazards of death (continuous HR = 0.68, $95\%$ CI = 0.54 to 0.85, Pquartile trend =.001). These findings persisted for every additional MET-hour per day estimates of PA energy expenditure (continuous HR = 0.88, $95\%$ CI = 0.82 to 0.95, Pquartile trend =.002). Conversely, each 1-h/d increase in sedentary behavior was not associated with statistically significantly higher mortality hazards (HR = 1.08, $95\%$ CI = 0.94 to 1.23, Pquartile trend =.07). Further mutual adjustment for continuous daily hours of MVPA did not substantially attenuate mortality associations for sedentary behavior, light PA, or PA energy expenditure (Table 3). Finally, a model with sedentary behavior, light PA, and MVPA indicated a statistically significant independent effect of light PA on mortality (HR = 0.55, $95\%$ CI = 0.35 to 0.87). **Figure 1.:** *Mortality hazard ratios by quartiles of sedentary behavior and physical activity among US cancer survivors in National Health and Nutrition Examination Survey (NHANES) 2003-2006. All multivariable models adjusted for age, sex, race and ethnicity, education, diet, smoking status, body mass index, self-reported health, mobility limitations, frailty, time since diagnosis, primary cancer type, and diagnoses of diabetes, stroke, heart disease, heart failure, chronic bronchitis, and emphysema. MVPA = moderate to vigorous physical activity; PA = physical activity.* TABLE_PLACEHOLDER:Table 3. ## Sensitivity analyses Stratified analyses indicated statistically significantly different associations between total PA and mortality by several indicators. Specifically, cancer survivors who were frail or most frail, had 1 and more chronic conditions, lost weight in the past year, and were at least 5 years from monitoring demonstrated greater mortality protection with increasing levels of PA (Pinteractions <.01). No statistically significant interactions emerged for sex, age, BMI, health status, mobility limitations, or time between diagnosis and monitoring (Pinteractions >.10) (Figure 2). Further post hoc sensitivity analyses investigated the possibility of reverse causation through exclusionary analyses. Table 4 details hazard ratios and $95\%$ confidence intervals in the original full sample and after varying exclusions (eg, chronic conditions, died within 1 year). These analyses suggest that our findings of reduced mortality risk with increased levels of PA are robust, if slightly inflated (Table 4). Finally, sensitivity analyses in major cancer types only (eg, breast, prostate, colon; $$n = 235$$, deaths = 116) confirmed the association between total PA and all-cause mortality (HR = 0.74, $95\%$ CI = 0.58 to 0.94). Further stratification confirmed that this association did not differ by major vs minor cancer type ($$P \leq .35$$). **Figure 2.:** *Stratified associations between total physical activity and mortality among US cancer survivors in the National Health and Nutrition Examination Survey (NHANES) 2003-2006. All multivariable models adjusted for age, sex, race and ethnicity, education, diet, smoking status, body mass index, self-reported health, mobility limitations, frailty, time since diagnosis, primary cancer type, and diagnoses of diabetes, stroke, heart disease, heart failure, chronic bronchitis, and emphysema. BMI = body mass index; Dx = diagnosis.* TABLE_PLACEHOLDER:Table 4. ## Discussion To our knowledge, this is the first analysis exploring the associations between accelerometry-derived PA, sedentary behavior, and all-cause mortality in a national sample of cancer survivors. Light PA, MVPA, and total PA were statistically significantly associated with a lower risk of mortality. Higher levels of sedentary behavior were not statistically significantly associated with greater mortality risk. Overall, these findings confirm the benefits of varying PA intensities after a cancer diagnosis, particularly in older adult cancer survivors, and underscore the importance of continued PA promotion during long-term survivorship. Our results are consistent with the literature documenting lower mortality risk with higher levels of self-reported PA during cancer survivorship, both generally [4,5,32] and in NHANES specifically [6]. The current analysis extends this work to include accelerometry-derived measures and explores a wider range of PA intensities than are traditionally captured in self-report. Our findings highlight the importance of lighter-intensity activities, and total PA accumulation, for survival benefits, even after mutual adjustment for MVPA. More recent work has highlighted the positive relationship between light PA and survival in the general population [34-36], suggesting that a broader range of intensities may promote longevity. We further demonstrated that this association may be independent of sedentary behavior and MVPA in this sample of cancer survivors. Such findings are important in this older adult cancer population due to the unique confluence of barriers (eg, fatigue, neuropathy) to initiating and maintaining PA regimens [37]. Although separating out actionable domain- and intensity-specific PA will be important to guide survivors in meeting cancer-specific guidelines [3], it is encouraging to see survival associations with total accumulated PA, not just recreational. Future PA programs may focus on a wider range of intensities to tailor exercise prescriptions while maintaining survival benefits in this diverse population. Sedentary behavior has distinct facilitators and barriers from PA [38] and has emerged as a strong predictor of premature mortality [39-41]. In cancer survivors specifically, increased sedentary behavior has been associated with poor quality of life, pain, and fatigue [42,43]. The epidemiological association between sedentary behavior and cancer mortality is more limited, with only a few studies reporting a modest $12\%$-$13\%$ higher risk of cancer mortality in most vs least time spent sedentary [4]. However, these studies have included both adults with and without cancer and used self-reported measures of sedentary behavior. Our current findings did not demonstrate statistically significantly higher mortality risk with more sedentary time, which may be due to suboptimal power. Though not statistically significant, the elevated hazard ratio is consistent with self-reported sedentary behavior-mortality findings in NHANES cancer survivors [6]. Given its hip placement, it is possible that the accelerometer misclassified certain stationary but light-intensity activities (eg, standing) as sedentary. Other devices, such as the thigh-worn activPAL (PAL Technologies Ltd), may be better suited to measure sedentary behavior [44]. Regardless, the sedentary behavior hazard ratio is consistent with other studies [4,6] and warrants future confirmatory studies. These studies should also consider if and how sedentary behavior offsets PA during survivorship, as seen in the general population [45], and how the behavioral tradeoff of reducing sedentary behavior to increase light PA, and vice versa [46,47], may be associated with mortality. Given stronger associations for the protective role of PA in individuals with poorer health, overestimation of the strength of associations is a concern. Indeed, accelerometry-mortality analyses in the general population may be subject to reverse causation [28], particularly in samples with short follow-up (<6 years), older participants (≥65 years of age), and limited statistical adjustment for poor health at the time of accelerometry measurement. The current NHANES analysis includes over 10 years of follow-up and multiple indicators of health status, providing a unique opportunity to explore these relationships while minimizing potential confounding. We conducted several stratified and post hoc analyses to test for reverse causation and largely confirmed our findings with consistent strength of associations (Table 4). However, we have previously discussed [28] that the reverse causation phenomenon in the context of accelerometry-derived PA, also known as “confounding by health status,” is not directly quantifiable and can only be inferred through comparisons of hazards ratios across models where confounding may be more or less present. Given that our sample included a subgroup of cancer survivors in NHANES with further reduced sample sizes in stratified analyses, we cannot completely rule out the possibility of this bias. Replicating these findings in larger cancer cohorts of varying ages with additional confounding variables will be important to confirm the PA-mortality association. There are several strengths and limitations of this analysis. We leveraged a cohort of a representative sample of noninstitutionalized adults in the United States with prospective linkage to death indices, which allowed us to characterize the association between accelerometry-derived PA and sedentary behaviors with mortality in 480 adult cancer survivors. This analysis extends work that has previously been confined to self-reported PA measures [32,48] and common cancer sites [49,50]. Our sensitivity analyses suggest that the PA survival benefits are not limited to breast, prostate, and colorectal cancers [4]; however, these cancer types still represented most cases in the current sample. Due to small sample sizes after stratification, we were unable to robustly explore these associations within cancer types. This will be important to determine if these associations are consistent across disease sites, as suggested herein, or rather differential. Although 7 days of monitoring is reflective of usual PA levels of months and years [51,52], and the strength of the PA accelerometer-mortality association appears to hold over many years in the general NHANES population [28], it is possible that the lack of updated PA information during follow-up could reduce the strength of associations observed. Prediagnosis PA was also not available for analysis as a covariate. Cancer survivors in this sample were more than a decade postdiagnosis on average and thus had survived long enough to be part of such an analysis. There was a large presence of older age (eg, ≥65 years) and frailty in our sample, suggesting that these findings are most salient for older adult cancer survivors. Investigators should attempt to reproduce these findings in individuals closer to diagnosis and/or treatment to better understand these relationships in the acute stages of disease. We also had limited information on cancer stage, treatment regimen, or other cancer-specific clinical characteristics; it is possible that individuals who receive more intensive treatment have different behavioral and clinical profiles and thus different survival patterns. In conclusion, our results support the beneficial association between PA and survival in long-term, older adult cancer survivors. Importantly, a broad range of PA intensities was associated with reduced risk of all-cause mortality. Promoting PA and discouraging sedentary behaviors after a cancer diagnosis should be a priority for providers and researchers alike, with renewed focus on lighter-intensity activities for those individuals who may not be able or willing to engage in higher-intensity exercise. ## Data availability The data that support the findings of this study are available on the Centers for Disease Control and Prevention’s webpage at https://www.cdc.gov/nchs/nhanes/index.htm. These data were derived from the following resources available in the public domain: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2003; https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2005. ## Author contributions Elizabeth Ann Salerno, PhD, MPH (Conceptualization; Formal analysis; Methodology; Writing—original draft; Writing—review & editing); Pedro F. Saint-Maurice, PhD (Formal analysis; Methodology; Writing—review & editing); Fei Wan, PhD (Formal analysis; Writing—review & editing); Lindsay L. Peterson, MD, MSCR (Writing—review & editing); Yikyung Park, ScD (Writing—review & editing); Yin Cao, ScD, MPH (Writing—review & editing); Ryan P. Duncan, PT, DPT, MSCI (Writing—review & editing); Richard P. Troiano, PhD, MNS (Writing—review & editing); Charles E. Matthews, PhD (Conceptualization; Methodology; Supervision; Writing—review & editing). ## Funding This work was supported by the Intramural Research Program of the National Cancer Institute at the National Institutes of Health [no number]. RPD is supported by National Institutes of Health (NIH) K23 HD100569. ## Conflicts of interest The authors have no conflicts of interest to disclose. ## References 1. 1 American Cancer Society. Cancer treatment and survivorship facts and figures 2019-2021. 2019. https://www.cancer.org/research/cancer-facts-statistics/survivor-facts-figures.html. Accessed July 1, 2020.. (2019) 2. Mishra SI, Scherer RW, Snyder C, Geigle PM, Berlanstein DR, Topaloglu O.. **Exercise interventions on health-related quality of life for people with cancer during active treatment. Mishra SI, ed**. *Cochrane Database Syst Rev* (2012) **2012** CD008465. DOI: 10.1002/14651858.CD008465.pub2 3. Campbell KL, Winters-Stone KM, Wiskemann J. **Exercise guidelines for cancer survivors: consensus statement from international multidisciplinary roundtable**. *Med Sci Sports Exerc* (2019) **51** 2375-2390. DOI: 10.1249/MSS.0000000000002116 4. Patel AV, Friedenreich CM, Moore SC. **American College of Sports Medicine Roundtable report on physical activity, sedentary behavior, and cancer prevention and control**. *Med Sci Sports Exerc* (2019) **51** 2391-2402. DOI: 10.1249/MSS.0000000000002117 5. Holmes MD, Chen WY, Feskanich D, Kroenke CH, Colditz GA.. **Physical activity and survival after breast cancer diagnosis**. *JAMA* (2005) **293** 2479-2486. DOI: 10.1001/jama.293.20.2479 6. Cao C, Friedenreich CM, Yang L.. **Association of daily sitting time and leisure-time physical activity with survival among US cancer survivors**. *JAMA Oncol* (2022) **8** 395-403. PMID: 34989765 7. Amireault S, Godin G, Lacombe J, Sabiston CM.. **Validation of the Godin-Shephard Leisure-Time Physical Activity Questionnaire classification coding system using accelerometer assessment among breast cancer survivors**. *J Cancer Surviv* (2015) **9** 532-540. DOI: 10.1007/s11764-015-0430-6 8. Boyle T, Lynch BM, Courneya KS, Vallance JK.. **Agreement between accelerometer-assessed and self-reported physical activity and sedentary time in colon cancer survivors**. *Support Care Cancer* (2015) **23** 1121-1126. DOI: 10.1007/s00520-014-2453-3 9. Sallis JF, Saelens BE.. **Assessment of physical activity by self-report: status, limitations, and future directions**. *Res Q Exerc Sport* (2000) **71** 1-14. DOI: 10.1080/02701367.2000.11082780 10. Hutcheon JA, Chiolero A, Hanley JA.. **Random measurement error and regression dilution bias**. *BMJ* (2010) **340** 1402-1406. DOI: 10.1136/bmj.c2289 11. Piercy KL, Troiano RP, Ballard RM. **The physical activity guidelines for Americans**. *JAMA* (2018) **320** 2020-2028. DOI: 10.1001/jama.2018.14854 12. Ma Y, He B, Jiang M. **Prevalence and risk factors of cancer-related fatigue: a systematic review and meta-analysis**. *Int J Nurs Stud* (2020) **111** 103707. DOI: 10.1016/j.ijnurstu.2020.103707 13. Thraen-Borowski KM, Gennuso KP, Cadmus-Bertram L.. **Accelerometer-derived physical activity and sedentary time by cancer type in the United States**. *PLoS One* (2017) **12** e0182554. DOI: 10.1371/journal.pone.0182554 14. Loprinzi PD, Lee H, Cardinal BJ.. **Objectively measured physical activity among US cancer survivors: considerations by weight status**. *J Cancer Surviv* (2013) **7** 493-499. DOI: 10.1007/s11764-013-0293-7 15. Smith WA, Nolan VG, Robison LL, Hudson MM, Ness KK.. **Physical activity among cancer survivors and those with no history of cancer--- a report from the National Health and Nutrition Examination Survey 2003-2006**. *Am J Transl Res.* (2011) **3** 342-350. PMID: 21904654 16. Lynch BM, Dunstan DW, Healy GN, Winkler E, Eakin E, Owen N.. **Objectively measured physical activity and sedentary time of breast cancer survivors, and associations with adiposity: findings from NHANES (2003-2006)**. *Cancer Causes Control* (2010) **21** 283-288. DOI: 10.1007/s10552-009-9460-6 17. Lynch BM, Friedenreich CM, Winkler EAH. **Associations of objectively assessed physical activity and sedentary time with biomarkers of breast cancer risk in postmenopausal women: findings from NHANES (2003-2006)**. *Breast Cancer Res Treat* (2011) **130** 183-194. DOI: 10.1007/s10549-011-1559-2 18. Lynch BM, Dunstan DW, Winkler E, Healy GN, Eakin E, Owen N.. **Objectively assessed physical activity, sedentary time and waist circumference among prostate cancer survivors: findings from the National Health and Nutrition Examination Survey (2003-2006)**. *Eur J Cancer Care (Engl)* (2011) **20** 514-519. DOI: 10.1111/j.1365-2354.2010.01205.x 19. Troiano RP, Berrigan D, Dodd KW, Mâsse LC, Tilert T, McDowell M.. **Physical activity in the United States measured by accelerometer**. *Med Sci Sports Exerc* (2008) **40** 181-188. DOI: 10.1249/mss.0b013e31815a51b3 20. Skubisz C, Kern A.. *Encyclopedia of Human Services and Diversity* (2014). DOI: 10.4135/9781483346663 21. Saint-Maurice PF, Graubard BI, Troiano RP. **Estimated number of deaths prevented through increased physical activity among US adults**. *JAMA Intern Med* (2022) **182** 349-352. DOI: 10.1001/JAMAINTERNMED.2021.7755 22. Matthews CE, Chen KY, Freedson PS. **Amount of time spent in sedentary behaviors in the United States, 2003-2004**. *Am J Epidemiol* (2008) **167** 875-881. DOI: 10.1093/aje/kwm390 23. Matthew CE.. **Calibration of accelerometer output for adults**. *Med Sci Sports Exerc* (2005) **37** S512-S522. DOI: 10.1249/01.mss.0000185659.11982.3d 24. Freedson PS, Melanson E, Sirard J.. **Calibration of the Computer Science and Applications, Inc. accelerometer**. *Med Sci Sports Exerc* (1998) **30** 777-781. DOI: 10.1097/00005768-199805000-00021 25. 25 National Center for Health Statistics. Office of Analysis and Epidemiology. The linkage of National Center for Health Statistics survey data to the National Death Index – 2015 Linked Mortality File (LMF): methodology overview and analytic considerations. 2018. https://www.cdc.gov/nchs/data-linkage/mortality-methods.htm. Accessed September 25, 2020.. (2018) 26. Guenther PM, Reedy J, Krebs-Smith SM.. **Development of the Healthy Eating Index-2005**. *J Am Diet Assoc* (2008) **108** 1896-1901. DOI: 10.1016/J.JADA.2008.08.016 27. Rockwood K, Mitnitski A.. **Frailty in relation to the accumulation of deficits**. *J Gerontol A Biol Sci Med Sci* (2007) **62** 722-727. DOI: 10.1093/gerona/62.7.722 28. Matthews CE, Troiano RP, Salerno EA. **Exploration of confounding due to poor health in an accelerometer–mortality study**. *Med Sci Sports Exerc* (2020) **2** 2546-2553. DOI: 10.1249/mss.0000000000002405 29. Hoover M, Rotermann M, Sanmartin C, Bernier J.. **Validation of an index to estimate the prevalence of frailty among community-dwelling seniors**. *Heal Reports* (2013) **24** 10-17 30. Saint-Maurice PF, Troiano RP, Bassett DR. **Association of daily step count and step intensity with mortality among US adults**. *JAMA* (2020) **323** 1151-1160. DOI: 10.1001/jama.2020.1382 31. Strain T, Wijndaele K, Sharp SJ, Dempsey PC, Wareham N, Brage S.. **Impact of follow-up time and analytical approaches to account for reverse causality on the association between physical activity and health outcomes in UK Biobank**. *Int J Epidemiol* (2020) **49** 162-172. DOI: 10.1093/ije/dyz212 32. Cannioto RA, Dighe S, Mahoney MC. **Habitual recreational physical activity is associated with significantly improved survival in cancer patients: evidence from the Roswell Park Data Bank and BioRepository**. *Cancer Causes Control* (2019) **30** 1-12. DOI: 10.1007/S10552-018-1101-5/TABLES/5 33. Hamilton MT, Hamilton DG, Zderic TW.. **Role of low energy expenditure and sitting in obesity, metabolic syndrome, type 2 diabetes, and cardiovascular disease**. *Diabetes* (2007) **56** 2655-2667. DOI: 10.2337/db07-0882 34. Qiu S, Cai X, Wu T. **Objectively-measured light-intensity physical activity and risk of cancer mortality: a meta-analysis of prospective cohort studies**. *Cancer Epidemiol Biomarkers Prev* (2020) **29** 1067-1073. DOI: 10.1158/1055-9965.EPI-19-1446 35. Loprinzi PD.. **Light-intensity physical activity and all-cause mortality**. *Am J Health Promot* (2017) **31** 340-342. DOI: 10.4278/AJHP.150515-ARB-882 36. LaMonte MJ, Buchner DM, Rillamas-Sun E. **Accelerometer-measured physical activity and mortality in women aged 63 to 99**. *J Am Geriatr Soc* (2018) **66** 886-894. DOI: 10.1111/jgs.15201 37. Stone P, Richardson A, Ream E. **Cancer-related fatigue: Inevitable, unimportant and untreatable? Results of a multi-centre patient survey. Cancer Fatigue Forum**. *Ann Oncol Off J Eur Soc Med Oncol* (2000) **11** 971-975 38. Owen N, Healy GN, Matthews CE, Dunstan DW.. **Too much sitting: the population health science of sedentary behavior**. *Exerc Sport Sci Rev* (2010) **38** 105-113. DOI: 10.1097/JES.0b013e3181e373a2 39. Proper KI, Singh AS, Van Mechelen W, Chinapaw MJM.. **Sedentary behaviors and health outcomes among adults: a systematic review of prospective studies**. *Am J Prev Med* (2011) **40** 174-182. DOI: 10.1016/J.AMEPRE.2010.10.015 40. Young DR, Hivert M-F, Alhassan S. **Sedentary behavior and cardiovascular morbidity and mortality: a science advisory from the American Heart Association**. *Circulation* (2016) **134** e262-e279. DOI: 10.1161/CIR.0000000000000440 41. Koster A, Caserotti P, Patel KV. **Association of sedentary time with mortality independent of moderate to vigorous physical activity**. *PLoS One* (2012) **7** e37696. DOI: 10.1371/JOURNAL.PONE.0037696 42. Forsythe LP, Alfano CM, George SM. **Pain in long-term breast cancer survivors: the role of body mass index, physical activity, and sedentary behavior**. *Breast Cancer Res Treat* (2013) **137** 617-630. DOI: 10.1007/S10549-012-2335-7 43. Trinh L, Amireault S, Lacombe J, Sabiston CM.. **Physical and psychological health among breast cancer survivors: interactions with sedentary behavior and physical activity**. *Psychooncology* (2015) **24** 1279-1285. DOI: 10.1002/PON.3872 44. Giurgiu M, Bussmann JBJ, Hill H. **Validating accelerometers for the assessment of body position and sedentary behavior**. *J Meas Phys Behav* (2020) **3** 253-263. DOI: 10.1123/JMPB.2019-0068 45. Ekelund U, Brown WJ, Steene-Johannessen J. **Do the associations of sedentary behaviour with cardiovascular disease mortality and cancer mortality differ by physical activity level? A systematic review and harmonised meta-analysis of data from 850 060 participants**. *Br J Sports Med* (2019) **53** 886-894. DOI: 10.1136/bjsports-2017-098963 46. Keadle SK, Conroy DE, Buman MP, Dunstan DW, Matthews CE.. **Targeting reductions in sitting time to increase physical activity and improve health**. *Med Sci Sports Exerc* (2017) **49** 1572-1582. DOI: 10.1249/MSS.0000000000001257 47. Tudor-Locke C, Leonardi C, Johnson WD, Katzmarzyk PT.. **Time spent in physical activity and sedentary behaviors on the working day: the American Time Use Survey**. *J Occup Environ Med* (2011) **53** 1382-1387. DOI: 10.1097/JOM.0B013E31823C1402 48. Friedenreich CM, Stone CR, Cheung WY, Hayes SC.. **Physical activity and mortality in cancer survivors: a systematic review and meta-analysis**. *JNCI Cancer Spectr* (2020) **4** pkz080. DOI: 10.1093/JNCICS/PKZ080 49. Schmid D, Leitzmann MF.. **Association between physical activity and mortality among breast cancer and colorectal cancer survivors: a systematic review and meta-analysis**. *Ann Oncol.* (2014) **25** 1293-1311. DOI: 10.1093/annonc/mdu012 50. Irwin ML, Smith AW, McTiernan A. **Influence of pre- and postdiagnosis physical activity on mortality in breast cancer survivors: the health, eating, activity, and lifestyle study**. *J Clin Oncol* (2008) **26** 3958-3964. DOI: 10.1200/J.Clin.Oncol.2007.15.9822 51. Saint-Maurice PF, Sampson JN, Keadle SK, Willis EA, Troiano RP, Matthews CE.. **Reproducibility of accelerometer and posture-derived measures of physical activity**. *Med Sci Sports Exerc* (2020) **52** 876-883. DOI: 10.1249/MSS.0000000000002206 52. Keadle SK, Shiroma EJ, Kamada M, Matthews CE, Harris TB, Lee IM.. **Reproducibility of accelerometer-assessed physical activity and sedentary time**. *Am J Prev Med* (2017) **52** 541-548. DOI: 10.1016/J.AMEPRE.2016.11.010
--- title: 'YouTube as an information source in paediatric dentistry education: Reliability and quality analysis' authors: - İlhan Uzel - Behrang Ghabchi - Ayşe Akalın - Ece Eden journal: PLOS ONE year: 2023 pmcid: PMC10038246 doi: 10.1371/journal.pone.0283300 license: CC BY 4.0 --- # YouTube as an information source in paediatric dentistry education: Reliability and quality analysis ## Abstract ### Introduction In the era of Covid 19 pandemic, the audio-visual contents of YouTube™ could be an information source for dental students, practitioners, and patients. The aim of this study was to evaluate the quality, content, and demographics of YouTube™ videos about pediatric dentistry for the education of dentistry students. ### Materials and methods A search on YouTube™ was performed using the keywords "pediatric dentistry”, “pediatric dental treatments”, “primary teeth treatments" in Turkish. The first 50 videos selected for each keyword were evaluated. Parameters of the videos such as the number of views, the days since the upload, the duration of the video, and the number of likes and dislikes were recorded. Videos are categorized by upload source and content categories as an academic, dentist, physician, patient, reporter, and other, and average points are obtained for the Journal of American Medical Association (JAMA) benchmark. The normality of the data was evaluated with the Shapiro-Wilk test. The data were not distributed normally, compared with the Kruskal Wallis test between source and content groups. The Dunn’s Post Hoc was used to determine to find out which group caused the difference. The Spearman Correlation coefficient was calculated to assess a possible correlation between JAMA, GQS, and VPI scores. All significance levels were set at 0.05. ### Results The duplicates and non-related ones were removed from 150 videos and remaining 119 videos were evaluated. Most of the videos were uploaded by the dentists and other categories, and mainly the videos were uploaded for patient education. JAMA score was 1 out of 4 for 55 videos, 2 for 63 videos, and 3 for only 1 video. When the video source groups were compared, the difference was statistically significant ($$p \leq 0.01$$). The difference between academic and patient groups ($$p \leq 0.007$$); the dentist and patient groups were statistically significant ($$p \leq 0.02$$). ### Conclusion YouTube platform does not contain videos of appropriate quality to support the education of dentistry students in pediatric dentistry in Turkish. ## Introduction COVID-19 disease (or SARS-CoV-2 virus) is a public health problem that emerged towards the end of 2019 and affected the whole world. This disease, which was declared a “Pandemic” by the World Health Organization on March 11, 2020, has adversely affected many sectors and caused serious problems in both practicing medicine and receiving education in the dental field. With the decision of the government, face-to-face training and exams have been suspended for an indefinite period in our country, as in many countries, due to efforts to minimize social contact to prevent the spread of the new coronavirus. On the other hand, the students had to receive digitalized education and were led to educate themselves—even partially—by obtaining information from various platforms on their own means. Web sites and various social media have rushed to the assistance of students in this regard [1]. Social media is a Web 2.0 technology that was founded in 1979 when Tom Truscott and Jim Ellis of Duke University created a worldwide discussion system that enabled internet users to broadcast public messages [2]. Nowadays, with the development and spread of social media, people have been given the freedom to present their ideas and information in the format they want, and an environment where people are given the opportunity to produce content and access this information easily. In this wide range of topics, many contents related to health are also found. Traditionally, information about medicine and dentistry was available in direct consultation with experts trained in this field. But today, with the use of the internet in every field in developed countries, it has become popular for people to use online resources to access this information [3]. Although the tendency to search for medical information on the internet varies potentially according to age, habit, and location, it has been reported that up to $75\%$ of people use the internet for this purpose [4]. The content provided by social media on health is frequently used not only for informing patients but also for educating students. It has been determined that various educational approaches, especially in social media such as Wikipedia, YouTube ™, and Facebook, are frequently used by dentists and medical professionals to get information [5, 6]. Among the changes that have taken place in dental education in recent years, there is also the integration of electronic methods that support multimedia presentations and e-learning strategies in the education of faculty members. A related innovation has been the use of participatory Internet websites, called Web 2.0 content, as they allow academic institutions to provide students with appropriate information with $\frac{24}{7}$ accessibility [7]. Since the development and upload of health-related information on the internet are not limited to professionals and it can be done by everyone, it can be assumed that false information, as well as correct information, will also exist [7, 8]. YouTube ™, a social media platform where people from all over the world upload, share, and watch videos in a simple and integrated manner was founded in the United States in 2005 by 3 former PayPal employees, Chad Hurley, Steve Chen, and Jawed Karim [9–11]. While YouTube™ is the largest and most popular video hosting platform, it is a free video sharing service that is currently the second largest search engine after Google [12, 13]. YouTube ™ offers every user the opportunity to upload, watch and share videos, and the site generally includes content such as video clips, televised content, music videos, vlogs (video blogs), short original videos, and educational videos. In YouTube ™, both amateurs and professionals can produce content, have their own channel, and communicate with each other through comments [14]. There are many different methods for evaluating the information and content (text, picture, video) uploaded to the internet environment. These methods provide the opportunity to evaluate the content of the website according to different criteria. Among these, The Journal of American Medical Association (JAMA) comparison criteria, Global Quality Score (GQS), and Video Power Index (VPI) are the commonly used evaluation methods. The Journal of American Medical Association (JAMA) [15] is a non-specific and objective tool consisting of 4 separate criteria, whose benchmarks can be defined in online videos and resources. These criteria provide an assessment of who published the content when it was published, and whether the sources were available. 1 point is earned for each criterion in the videoThe Global Quality Score (GQS) [16] is a specialized, unverified but widely used scoring system in which scores are awarded to determine the quality of online videos [17, 18]. While this scaling system normally uses ordinary people as subjects, in the present study, the Modified Global Quality Score (mGQS), was developed for dentistry by Cesur Aydın et al. [ 12] was used. Video Power index (VPI) is a video popularity index based on the number of views and likes of the evaluated videos. Therefore, the aim of the present study was to evaluate the video contents for dental professionals or students uploaded on YouTube, in terms of the correct information and educational level. ## Materials and methods The study design has been done as a cross-sectional descriptive study. A search was made by a single researcher (A.A.) under the ’video’ filter using the keywords ’pediatric dentistry, pediatric dental treatments, primary teeth treatments’ on the YouTube ™ search engine (https://www.youtube.com/) on 23 November 2020 and it took about one month for analyze and evaluate the videos. The search was carried out by opening a new YouTube ™ account that has not been used before. The reason for this is that the YouTube algorithm offers content considering user interactions [19]. While the videos were included in the study, the conditions of being related to the subject were searched. The first 50 videos were evaluated for each keyword. The title, video source content type, video duration, number of views, days after uploading, and the number of likes and dislikes of each video evaluated were recorded. Using these data, view rate (number of views / days since upload), likes rate [likes / (likes + dislikes) * 100] and video power index (VPI) variables were calculated. The evaluated videos were examined under 6 groups in terms of uploading resources. These are [1] academic (loaders affiliated with research groups and universities); [2] the dentist; [3] physician; [4] the patient; [5] the messenger; [6] other. Videos were evaluated under 5 titles according to content as: [1] information on undergraduate / postgraduate dental education; [2] patient information; [3] inspection experience; [4] entertainment; [5] advertising. The Journal of American Medical Association (JAMA) comparison criteria [Table 1] were used to evaluate the accuracy and reliability of the videos. A total of 4 points indicates a high source of accuracy and reliability, while 0 points indicate poor source reliability [17]. JAMA comparison criteria are given in Table 1. **Table 1** | Criteria | Explanation | | --- | --- | | Authorship | The credentials and links of the author and contributors should be provided | | Attribution | It clearly lists all copyright information, citing references and sources for the content. | | Validity | The first date of the posted content and subsequent content updates should be specified. | | Explanation | Conflicts of interest, financing, sponsorship, advertising, support, and video ownership should be fully disclosed. | In addition, the quality assessment of the videos was made using the Modified Global Quality Score (mGQS) [Table 2] scaling. This scale scores the quality of the content from 1 to 5. The highest content quality score was determined as 5. More information about the mGQS system is shown in Table 2. **Table 2** | Rating | Definition of Quality | | --- | --- | | 1 | Poor quality and flow; most of the information is missing, not suitable for use by dentists | | 2 | Generally, low quality and flow; Limited use for dentists as only some information is available | | 3 | Medium quality and low standards of flow; contains some important information but does not provide enough information, useful to the basic level for dentists | | 4 | Good quality and flow; The vast majority of important information on the subject has been presented, useful for dentists | | 5 | Excellent quality and flow; very useful for dentists | Video Power *Index is* calculated with the formula [like rate * view rate / 100]. The rate of likes is calculated with the formulas [likes * 100 / (likes + dislikes)] and views (number of views/days) [20]. ## Statistical analysis The normality of the data was evaluated with the Shapiro-Wilk test. Since the data were not distributed normally, JAMA, GQS and VPI scores were compared with the Kruskal Wallis test between source groups and content groups. When the difference was statistically significant, the Dunn’s Post Hoc was used to determine to find out which group caused the difference. The Spearman Correlation coefficient was calculated to assess a possible correlation between JAMA, GQS, and VPI scores. The statistical significance level was set at $$p \leq 0.05.$$ ## Results When duplicates were removed from 150 videos that were reached with keywords, a total of 132 videos were evaluated. Among these videos, 13 videos that are not related to the subject were removed. According to the recorded data of the videos that were evaluated, a total of 44282 seconds of video was examined and the average of these videos was calculated as 372 seconds. It has been determined that the videos were liked 14708 times and 3896 disliked by the users in total. Other recorded and calculated properties of the videos are given in Table 3. **Table 3** | Unnamed: 0 | Average | Standard deviation | Minimum | Maximum | | --- | --- | --- | --- | --- | | Video duration (sec) | 372.0 | 674.0 | 8.0 | 5285.0 | | Viewed number | 46622.21 | 160953.42 | 10.0 | 1195982.0 | | Time from uploading (day) | 1222.4 | 783.45 | 40.0 | 4809.0 | | Viewing rate | 38.25 | 133.41 | 0.01 | 832.25 | | Like | 124.64 | 362.2 | 0.0 | 2068.0 | | Dislike | 33.01 | 146.83 | 0.0 | 1380.0 | | Like ratio (%) | 86.95 | 15.56 | 0.0 | 100.0 | According to the video source classification, it was determined that video uploaders had the most accounts in the categories of dentists ($33.61\%$ $$n = 40$$) and under the category of ‘other’ ($34.45\%$ $$n = 41$$). It was seen that the least uploaded video source was physicians ($1.68\%$ $$n = 2$$). The distribution of video sources is given in Fig 1. **Fig 1:** *Classification of YouTube videos about pediatric dentistry according to the uploaded source.* When the contents of the evaluated videos were examined, it was found that there was a high rate of videos for informing patients ($67.22\%$ $$n = 80$$) according to the classification made. It was observed that the least uploaded content area was the advertisement ($3.36\%$ $$n = 4$$) category. The graphic of content classification is given in Fig 2. **Fig 2:** *Classification of YouTube videos about pediatric dentistry according to their content.* ## Content When all sources and contents were evaluated, the mean JAMA score obtained was 1.54 (SD: 0.5), the mean mGQS score was 1.98 (SD: 0.8) and the mean VPI ratio was 34.2 (SD: 108, 2) as calculated. The means and standard deviations (Mean (SD)) of JAMA, mGQS, and VPI values calculated separately for all sources and contents are listed in Table 4. **Table 4** | Unnamed: 0 | JAMA | mGQS | VPI | | --- | --- | --- | --- | | GROUP VARIABLES | Mean (SD) | Mean (SD) | Mean (SD) | | Video Source | | | | | Academic | 1.87 (0.3) | 3(1.4) | 1.64(1.4) | | Dentist | 1.7 (0.5) | 2 (0.8) | 29.94 (115.7) | | Physician | 2 (0) | 2.5 (1.5) | 108.4 (0) | | Patient | 1.07 (0.2) | 1.07 (0.2) | 136.79 (205.05) | | Reporter | 1.53 (0.4) | 2.33 (0.6) | 15.75 (29.1) | | Other | 1.46 (0.4) | 1.90 (0.5) | 11.46 (22.9) | | Video content | | | | | Undergraduate/ graduate education | 1.87 (0,5) | 3.75 (0.8) | 22.54 (30.3) | | Patient disclosure | 1.63 (0.4) | 2.16 (0.6) | 7.55 (17.9) | | Examination experience | 1.19 (0.3) | 1.09 (0.2) | 105.02 (177.1) | | Entertainment | 1.33 (0.4) | 1 (0) | 144.06 (249.4) | | Advertisement | 1.25 (0.4) | 1 (0) | 5.28 (4.3) | When an analysis was made based on the JAMA criteria, it was seen that 55 videos got 1, 63 videos got 2 points and only 1 video got 3 points. None of the videos got a score of 4 indicating full reliability. Although the physicians received the highest score when the uploaded sources of the video were evaluated, these values were quite low to conclude that the videos were of high quality in the reliability criteria. The uploaded videos from academic and dentist sources did not exceed 2 points on average based on these criteria. When the video was examined in terms of content, even though the highest quality was in undergraduate/graduate education, again, no category exceeded 2 points on average. When the video source groups were compared, the difference was statistically significant ($$p \leq 0.01$$). The difference between academic and patient groups ($$p \leq 0.007$$); the dentist and patient groups were statistically significant ($$p \leq 0.02$$). The difference between examination experience and graduate/undergraduate education groups were statistically significant ($$p \leq 0.029$$) in terms of video content. When the mGQS score was evaluated, it was seen that academic sources uploaded the highest quality videos in terms of video sources. When the videos uploaded by dentists were considered in terms of educational quality, they have fallen behind the journalist and physician resources. Regarding the video content, the highest score in video quality was obtained in the undergraduate/graduate education category as expected. The lowest score was received by the content of entertainment and advertising videos uploaded by patients as a source. There was a statistically significant difference among video source groups in GQS scores ($p \leq 0.001$). The difference between patient and academic; dentist and reporter groups were statistically significant ($p \leq 0.001$; $$p \leq 0.002$$; $$p \leq 0$$<0.001, respectively). The difference between graduate/undergraduate content and patient disclosure ($$p \leq 0.007$$), examination experience ($p \leq 0.001$), entertainment ($p \leq 0.001$), advertisement ($p \leq 0.001$) were statistically significant. When the popularity of videos was examined over the VPI value, it was seen that the videos uploaded by academic education sources had very low popularity, and the most popular videos were uploaded by patients and for entertainment purposes. The difference between academic and reporter groups were statistically significant ($$p \leq 0.043$$) and, dentist and patient groups were significant ($$p \leq 0.039$$). The correlations between JAMA and GQS ($r = 0.426$); VPI (r = -0.242) scores were low. The correlation between GQS and VPI scores were -0.208. the difference between patient disclose and examinatişon experience was statistically significant ($$p \leq 0.002$$). ## Discussion Because of the extraordinary situations experienced due to COVID-19, education has shifted towards the home environment and internet-based information acquisition has become almost mandatory. Online videos are among the most preferred of these information resources. There are many reasons why students frequently consult online videos to support their education, such as ease of access, more catchy visual data than written ones, more fun, and no time limit [19]. A considerable part of these online videos is also available on YouTube ™, the well-known social media platform [21]. Data in C.H.Basch et al. paper show the potentially inaccurate and negative influence social media can have on population-wide education uptake that should be urgently addressed by agencies of each country’s Public Health Service as well as its global counterparts [22]. However, medical information, including YouTube ™, is broadly different from each other in terms of scientific quality, as it is broadcasted directly without any control over the content [23, 24]. As can be seen in this study, most online content uploaders on YouTube ™ are people who do not owe academic education and the information they provide is not peer-reviewed and/or validated. In this study, it has been shown that the sources providing academic education are shared less on the YouTube ™ platform than other sources and the average JAMA score of these posts is not sufficient to be considered safe. In terms of content, it has been revealed that although the academic group had, naturally, the highest mGQS score, they offered medium-quality content with an average of 3 points out of 5. When video popularity was examined, this group had a very low rate compared to the others. Although variable parameters such as the number of views and likes of the videos provide us some information about the benefits of the videos, it may be considered normal that the popularity of the academic group was lower than the other groups, since this platform was often a medium with a high demand for entertainment content and open to every user. It was observed that the videos evaluated within the scope of this study, which was conducted subjectively, were not of high quality and reliable to be used as a source of information in pediatric dentistry education, considering the determined criteria. Although there was not any similar study that evaluated this subject specifically in the literature, the findings of our study were in accordance with earlier studies that reported results of YouTube screening on other dentistry-related issues. Knösel et al. observed that the uploaded video content about dentistry on YouTube did not meet the reliability criteria to a large extent [7]. Similarly, in a study by Burns et al. [ 25]. $36\%$ of the students reported uncertainty and dissatisfaction in accessing evidence-based information on YouTube. It can be predicted that online resources will be used frequently during this period, and academic education resources should do their part in this regard. It is necessary to inform students about how to lead them to sources where they can obtain reliable and qualified information, to make more widespread online information sources for the purpose of providing education, and to develop content, with special attention to student education. ## Conclusion It has been observed that the YouTube ™ platform does not contain videos of enough quality and reliability to support the education of dentistry students in pediatric dentistry. The content offered by educators needs to become more widespread and popular. Students should be made aware of which sources to search for and that not all information they obtain from online video sources is completely correct. ## References 1. Basch CH, Hillyer GC, Meleo-Erwin ZC, Jaime C, Mohlman J, Basch CE. **Preventive Behaviors Conveyed on YouTube to Mitigate Transmission of COVID-19: Cross-Sectional Study**. *JMIR Public Health Surveill* (2020.0) **6** e18807. DOI: 10.2196/18807 2. Kaplan AM, Haenlein M. **Users of the world, unite! The challenges and opportunities of Social Media**. *Bus Horiz* (2010.0) **53** 59-68. DOI: 10.1016/j.bushor.2009.09.003 3. Biggs TC, Bird JH, Harries PG, Salib RJ. **YouTube as a source of information on rhinosinusitis: the good, the bad and the ugly**. *J Laryngol Otol* (2013.0) **127** 749-754. DOI: 10.1017/S0022215113001473 4. Nason K, Donnelly A, Duncan HF. **YouTube as a patient-information source for root canal treatment**. *Int Endod J* (2016.0) **49** 1194-1200. DOI: 10.1111/iej.12575 5. Green B, Hope A. **Promoting clinical competence using social media**. *Nurse Educ* (2010.0) **35** 127-129. DOI: 10.1097/NNE.0b013e3181d9502b 6. Vance K, Howe W, Dellavalle RP. **Social Internet Sites as a Source of Public Health Information**. *Dermatol Clin* (2009.0) **27** 133-136. DOI: 10.1016/j.det.2008.11.010 7. Knösel M, Jung K, Bleckmann A. **YouTube, Dentistry, and Dental Education**. *J Dent Educ* (2011.0) **75** 1558-1568. DOI: 10.1002/j.0022-0337.2011.75.12.tb05215.x 8. Leong AY, Sanghera R, Jhajj J, Desai N, Jammu BS, Makowsky MJ. **Is YouTube Useful as a Source of Health Information for Adults With Type 2 Diabetes? A South Asian Perspective**. *Can J Diabetes* (2018.0) **42** 395-403.e4. DOI: 10.1016/j.jcjd.2017.10.056 9. 9Ataman SN. Youtube Türkiye’ye Geliyor. https://medium.com/bidünyaiçerik/youtube-türkiyeye-geliyor-5778bc0c7691. Published 2020. 10. Duncan I, Yarwood-Ross L, Haigh C. **YouTube as a source of clinical skills education**. *Nurse Educ Today* (2013.0) **33** 1576-1580. DOI: 10.1016/j.nedt.2012.12.013 11. Burke SC, Snyder S, RC R. **An Assessment of Faculty Usage of YouTube as a Teaching Resource**. *Internet J Allied Health Sci Pract* (2009.0) **7** 1-8. DOI: 10.1353/cj.0.009 12. Cesur Aydin K, Gunec HG. **Quality of information on YouTube about artificial intelligence in dental radiology**. *J Dent Educ* (2020.0) **84** 1166-1172. DOI: 10.1002/jdd.12362 13. Basch CE, Basch CH, Hillyer GC, Jaime C. **The Role of YouTube and the Entertainment Industry in Saving Lives by Educating and Mobilizing the Public to Adopt Behaviors for Community Mitigation of COVID-19: Successive Sampling Design Study**. *JMIR Public Health Surveill* (2020.0) **6** e19145. DOI: 10.2196/19145 14. Mutlu B, Bazarcı S. **Marka İşbirlikleri İçin Yeni Bir Alan: Youtube İçerik Üreticileri ve Kanal Toplulukları Üzerine Netnografik Bir Araştırma**. *J Akdeniz Uni Fac Commun* (2017.0) 28-45. DOI: 10.31123/akil.437267 15. Silberg WM, Lundberg GD, Musacchio RA. **Assessing, controlling, and assuring the quality of medical information on the Internet: Caveat lector et viewer—Let the reader and viewer beware**. *JAMA* (1997.0) **277** 1244-1245. PMID: 9103351 16. Bernard A, Langille M, Hughes S, Rose C, Leddin D, Veldhuyzen Van Zanten S. **A systematic review of patient inflammatory bowel disease information resources on the world wide web**. *Am J Gastroenterol* (2007.0) **102** 2070-2077. DOI: 10.1111/j.1572-0241.2007.01325.x 17. Kunze KN, Cohn MR, Wakefield C. **YouTube as a Source of Information About the Posterior Cruciate Ligament: A Content-Quality and Reliability Analysis**. *Arthrosc Sport Med Rehabil* (2019.0) **1** e109-e114. DOI: 10.1016/j.asmr.2019.09.003 18. Li M, Yan S, Yang D, Li B, Cui W. **YouTube**. *BMC Public Health* (2019.0) **19** 952. DOI: 10.1186/s12889-019-7297-9 19. Dias da Silva MA, Pereira AC, Walmsley AD. **Who is providing dental education content via YouTube?**. *Br Dent J* (2019.0) **226** 437-440. DOI: 10.1038/s41415-019-0046-8 20. Erdem MN, Karaca S. **Evaluating the accuracy and quality of the information in kyphosis videos shared on youtube**. *Spine (Phila Pa 1976)* (2018.0) **43** E1334-E1339. DOI: 10.1097/BRS.0000000000002691 21. Atilla AO, Öztürk T. **Maksiller Ekspansiyon İçin Bilgi Kaynağı Olarak Youtube’un Video Analizi ile Değerlendirilmesi**. *Selcuk Dent J* (2020.0) **7** 494-499. DOI: 10.15311/selcukdentj.820932 22. Basch CE, Basch CH, Hillyer GC, Meleo-Erwin ZC, Zagnit EA. **YouTube Videos and Informed Decision-Making About COVID-19 Vaccination: Successive Sampling Study**. *JMIR Public Health Surveill* (2021.0) **7** e28352. DOI: 10.2196/28352 23. Ozdede M, Peker I. **Analysis of Dentistry YouTube V i d e o s R e l a t e d To C OV I D—1 9**. *Braz Dent J* (2020.0) **31** 392-398. DOI: 10.1590/0103-6440202003767 24. Mangan MS, Cakir A, Yurttaser Ocak S, Tekcan H, Balci S, Ozcelik Kose A. **Analysis of the quality, reliability, and popularity of information on strabismus on YouTube**. *Strabismus* (2020.0) **28** 175-180. DOI: 10.1080/09273972.2020.1836002 25. Burns LE, Abbassi E, Qian X, Mecham A, Simeteys P, Mays KA. **YouTube use among dental students for learning clinical procedures: A multi-institutional study**. *J Dent Educ* (2020.0) **84** 1151-1158. DOI: 10.1002/jdd.12240
--- title: Primary care provider uptake of intensive behavioral therapy for obesity in Medicare patients, 2013–2019 authors: - Mounira Ozoor - Mark Gritz - Rowena J. Dolor - Jodi Summers Holtrop - Zhehui Luo journal: PLOS ONE year: 2023 pmcid: PMC10038247 doi: 10.1371/journal.pone.0266217 license: CC BY 4.0 --- # Primary care provider uptake of intensive behavioral therapy for obesity in Medicare patients, 2013–2019 ## Abstract ### Background The delivery of Intensive Behavioral Therapy (IBT) for obesity by primary care providers (PCPs) has been covered by Medicare to support elderly individuals with obesity (BMI > 30 kg/m2) in managing their weight since 2011 for individual therapy and 2015 for group therapy. We conducted a cohort study of PCPs in an attempt to understand patterns of uptake of IBT for obesity services among PCPs serving the Medicare population across the U.S. ### Methods We used the Centers for Medicare and Medicaid Services Provider Utilization and Payment Data from 2013 to 2019 to identify all PCPs who served more than 10 Medicare beneficiaries in each of the seven-year period to form a longitudinal panel. The procedure codes G0447 and G0473 were used to identify PCPs who provided IBT; and the characteristics of these providers were compared by the IBT-uptake status. ### Results Of the 537,754 eligible PCPs who served Medicare patients in any of the seven years, only $1.2\%$ were found to be IBT service providers in at least one year from 2013 through 2019 (246 always users, 1,358 early adopters, and 4,563 late adopters). IBT providers shared a few common characteristics: they were more likely to be male, internal medicine providers, saw a higher number of Medicare beneficiaries, and practiced in the South and Northeast regions. Having higher proportion of patients with hyperlipidemia was associated with higher likelihood of a provider being an IBT-user. ### Conclusions Very few PCPs continuously billed IBT services for Medicare patients with obesity. Further investigation is needed to mitigate barriers to the uptake of IBT services among PCPs. ## Introduction The prevalence of obesity in the U.S. has been rising [1, 2] and is currently at an all-time high [3]. As of 2018, approximately $30\%$ of the adult population and $28.5\%$ of the senior population are living with obesity [4]. As they age, the baby-boomer generation has higher rates of obesity than previous generations [5]. Obesity is one of the major drivers of preventable health care costs, which is estimated between $147 billion and $210 billion [6]. In order to address this challenging public health issue, in 2011, the Centers for Medicare and Medicaid Services (CMS) established a Healthcare Common Procedure Coding System (HCPCS) code and authorized primary care providers (PCPs) to deliver and bill for Intensive Behavioral Therapy (IBT) for obesity in an attempt to help treat individuals with obesity (BMI ≥ 30 kg/m2) [7]. IBT is an evidence-based service with a “B” rating, meaning there is “high certainty that the net benefit is moderate or there is moderate certainty that the net benefit is moderate to substantial,” and is recommended by the United States Preventive Services Task Force to be provided by clinicians to eligible patients [8]. Some studies found that high intensity counseling, as well as behavioral interventions, will deliver continuous weight loss for individuals with obesity [9]. IBT consists of working closely with an approved provider to target behaviors that are contributing to a patient’s obesity condition and educate the patient and implement changes such as tracking dietary intake and creating an exercise plan. All Medicare beneficiaries with obesity are eligible for the service [10]. The benefit allows for 15-min visits once per week for 4 weeks, biweekly visits for months 2 to 6, and then once monthly for another 6 months if the patient lost ≥ 3 kg (6.6 lb) [11]. To encourage Medicare beneficiaries to receive IBT services, CMS waived the Medicare coinsurance and Part-B deductible for this service. However, CMS established several requirements that must be met for PCPs to receive reimbursement for delivering IBT services to beneficiaries with obesity, including: [1] the service must be provided in a primary care setting by a qualified PCP; [2] a total of 22 visits can be billed in a 12-month period; [3] patients are seen once per week in the first month and twice per month through month 6; and [4] patients must meet a weight loss goal of at least 3 kg (6.6 lbs.) during the first 6 months to continue treatment, which is expected to be once a month for the next 6 months. If the patient does not achieve the weight loss goal, they must wait at least 6 months and be assessed for eligibility based on BMI in order to begin a new treatment period once again [11]. IBT services are reimbursed through two HCPCS codes. The first code, G0447, was authorized for services beginning in November 2011 and is used for one-on-one face-to-face behavioral counseling for obesity for a 15-minute encounter. The second code, G0473, was authorized beginning in January 2015 for face-to-face group (2–10 patients) behavioral counseling for obesity for a 30-minute group session. The average reimbursement rates for each service is $24-$26 for G0447 and $12-$13 for G0473 in 2018 [12]. The uptake of IBT services among Medicare providers has been underwhelming ($0.1\%$ of eligible beneficiaries in 2012 [13] and $0.2\%$ in 2015 [14]) since the CMS authorized billing in 2011. The purpose of this paper is to examine the pattern of uptake of IBT for obesity services among PCPs serving the Medicare fee-for-service population across the U.S. to identify characteristics of PCPs who provided the services early, late, or never after the activation of the codes. Our analysis of provider characteristics complements and adds to our understanding of the utilization of IBT services by beneficiary characteristics [13, 14]. ## Methods We conducted a cohort analysis of PCPs who were authorized to provide IBT services and who billed Medicare for more than ten unique beneficiaries for any service between 2013 and 2019 to examine their IBT service uptake patterns. The study was approved by the Institutional Review Boards of University of Colorado and Michigan State University as non-human subject research. ## Data sources The publicly available Medicare Fee-for-service Provider Utilization and *Payment data* for Physician and Other Supplier Public Use File as well as the Provider Summary Tables from CMS covering calendar years 2013 through 2019 are used to identify the primary care providers eligible for delivering IBT [15]. The public use file includes providers who had a valid National Provider Identifier and submitted Medicare Part-B non-institutional claims in a calendar year. To protect the privacy of Medicare beneficiaries, any aggregated records that are derived from 10 or fewer beneficiaries are excluded from the public use data, which led to some providers being classified as a non-user even though they had used IBT for a handful of patients. The data elements include the provider’s identifier, provider type, gender, zip code, state, HCPCS codes billed for more than 10 unique beneficiaries by the provider, and number of claims, unique beneficiaries and average Medicare payment amount by HCPCS codes. The Provider Summary Tables include data pertaining to service utilization, payments, provider demographics, and beneficiary demographics. America’s Health Ranking’s obesity data [16] and Kaiser Family Foundation’s publications [17] are used to find Medicare beneficiary populations with obesity by state. ## Data analysis We restricted the PCP types to family practice, general practice, internal medicine, nurse practitioner and physician assistant as these were the PCPs who were authorized to bill under the two IBT HCPCS codes. If a provider who was included in the analysis did not have any of the two HCPCS codes in a given year, the provider was considered an IBT non-user for that year. We categorized the providers based on their IBT-uptake pattern over this period: always users include providers who had IBT claims for more than 10 unique Medicare beneficiaries every year during 2013 through 2019; early adopters include providers who had at least some utilization during 2013 or 2014 but not always in the other five years; late adopters include providers who did not have IBT claims reported during 2013 and 2014, but did during at least one year in 2015 through 2019, and lastly, never users include providers who had no reported payments for IBT claims from 2013 through 2019. Within each of these categories, we examined the following provider characteristics: gender, region, provider type, average count of unique Medicare beneficiaries, provider charges and beneficiaries with other relevant chronic conditions served by the provider. We examined the IBT service penetration in two ways: the number of Medicare beneficiaries per 1,000 elderly population with obesity in a state, and the number of PCPs who billed for IBT per 1,000 PCPs in a state. Choropleth maps were used to show utilization patterns between 2013 and 2019 within the measurement period across states. These maps can identify regions of strong and weak adoption of the services. Descriptive statistics (sample size, proportions for discrete variables, medians, means and standard deviations for continuous variables) were presented by provider user types. Multinomial logistic regressions were used to compare provider characteristics and patient-compositions related to utilization patterns. Two-sided tests at the 0.05 significance level were performed using Stata 17 (StataCorp LLC). ## Results Table 1 shows provider characteristics by their IBT-uptake patterns (never users, late adopters, early adopters and always users) for the 537,754 providers identified in the public use data. Of these providers, the majority of them were considered never users ($98.9\%$); of the never users, $61.1\%$ were females, mainly practicing in the South ($36.3\%$) and Midwest ($24.1\%$), and primarily made up of internal medicine providers ($24.8\%$), nurse practitioners ($34.0\%$), and family practice providers ($20.2\%$). The never users had a median count of 108 unique Medicare beneficiaries per year when considering all Medicare claims. **Table 1** | Unnamed: 0 | IBT Providers* | IBT Providers*.1 | IBT Providers*.2 | IBT Providers*.3 | IBT Providers*.4 | IBT Providers*.5 | IBT Providers*.6 | IBT Providers*.7 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | Never Users | Never Users | Late Adopters | Late Adopters | Early Adopters | Early Adopters | Always Users | Always Users | | | (n = 531,587) | (n = 531,587) | (n = 4,563) | (n = 4,563) | (n = 1,358) | (n = 1,358) | (n = 246) | (n = 246) | | | n | % | n | % | n | % | n | % | | Gender | | | | | | | | | | Males | 206059 | 38.9% | 1975 | 56.7% | 831 | 61.2% | 167 | 67.9% | | Females | 325059 | 61.1% | 2588 | 43.3% | 527 | 38.8% | 79 | 32.1% | | Region: | | | | | | | | | | Midwest | 128375 | 24.1% | 696 | 15.3% | 188 | 13.8% | 22 | 8.9% | | Northeast | 109266 | 20.6% | 1150 | 25.2% | 347 | 25.6% | 90 | 36.6% | | South | 193094 | 36.3% | 1951 | 42.5% | 606 | 44.6% | 104 | 42.3% | | West | 100717 | 18.9% | 776 | 17.0% | 217 | 16.0% | 30 | 12.2% | | Provider Type: | | | | | | | | | | Family practice | 107270 | 20.2% | 1692 | 37.1% | 485 | 35.7% | 78 | 31.7% | | General practice † | 7428 | 1.4% | 106 | 2.3% | 35 | 2.6% | 3 | 1.2% | | Internal medicine | 132094 | 24.8% | 1951 | 42.8% | 713 | 52.5% | 161 | 65.4% | | Nurse practitioner | 180882 | 34.0% | 611 | 13.4% | 93 | 6.8% | 2 | 0.8% | | Physician assistant | 103913 | 19.5% | 203 | 4.4% | 32 | 2.4% | 2 | 0.8% | | Median annual number of unique Medicare beneficiaries | 108 | 108 | 266 | 266 | 329 | 329 | 350 | 350 | | Median annual number of services/encounters per year | 264 | 264 | 1624 | 1624 | 2804 | 2804 | 4202 | 4202 | | Median annual number of all procedure codes | 16 | 16 | 42 | 42 | 55 | 55 | 65 | 65 | | Median annual submitted charge amount | $43,955 | $43,955 | $166,638 | $166,638 | $273,253 | $273,253 | $354,981 | $354,981 | | Percent of beneficiaries with obesity related chronic conditions ‡ | | | | | | | | | | Hypertension | 68.0% | 68.0% | 69.6% | 69.6% | 70.6% | 70.6% | 71.7% | 71.7% | | Diabetes | 38.2% | 38.2% | 38.8% | 38.8% | 39.0% | 39.0% | 42.9% | 42.9% | | Hyperlipidemia | 57.1% | 57.1% | 60.9% | 60.9% | 60.9% | 60.9% | 66.1% | 66.1% | Late adopters comprised just 4,563 ($0.9\%$) providers. Most of these providers were males ($56.7\%$), practicing mainly in the South ($42.5\%$) and Northeast ($25.2\%$). The late adopters were mainly internal medicine ($42.8\%$) and family practice ($37.1\%$) providers, with a median count of 266 unique Medicare beneficiaries per year. Early adopters consisted of only 1,358 ($0.3\%$) providers, with $61.2\%$ being males. The early adopters mainly practiced in the South ($44.6\%$) and the Northeast ($25.6\%$) and were primarily internal medicine ($52.5\%$) and family practice ($35.7\%$) providers. The early adopters had a median count of 329 unique Medicare beneficiaries per year. Lastly, the smallest group, the always users, consisted of only 246 providers ($0.05\%$), who shared a similar distribution of characteristics as the early and late adopters. These providers were $67.9\%$ male, practicing mainly in the South ($42.3\%$) and Northeast ($36.6\%$), and made up of mostly internal medicine ($65.4\%$) and family practice ($31.7\%$) providers. The always users had a median count of 350 unique Medicare beneficiaries per year. We also utilized the Provider Summary Tables to help further understand the distributions of several other variables among the different IBT-uptake patterns, including the median numbers of total annual services or encounters, annual unique billed HCPCS codes, and annual total submitted charges. Generally the services rendered were lowest among non-IBT providers and highest among the always users (Table 1). We also examined the distributions of obesity related chronic conditions, including each provider’s average proportions of Medicare patients with hypertension, diabetes, and hyperlipidemia by the IBT-uptake patterns. The proportions of patients with these chronic conditions were fairly similar but did slightly increase across the four IBT-uptake groups (lower panel Table 1), showing somewhat of a pattern between the prevalence of the relative chronic conditions in the populations that the providers served and whether or not they provided IBT obesity services to their patients. Table 2 showed the odds ratio (OR) of a multinomial logistic regression with the four uptake patterns as the outcome and factors in Table 1 as regressors (excluding the highly collinear number of annual billed HCPCS codes). The provider types other than internal medicine were grouped together to avoid the small cell problem. The $95\%$ confidence intervals (CIs) indicated that male, Northeast-, South- and West-region, and internal medicine providers had an increasing likelihood of being a late, early, or always adopters compared to never users. Compared with never users, for every 10 percentage points increase in patients with hypertension among later adopters, the OR for being a late adopter was 0.92 ($95\%$ CI 0.90, 0.96). Similarly, for every 10 percentage points increases in patients with diabetes the OR for being a late adopter was 0.93 ($95\%$ CI 0.90, 0.96) and the OR for being an early adopter was 0.87 ($95\%$ CI 0.82, 0.91). However, for every 10 percentage points increases in patients with hyperlipidemia, the OR for being a late adopter was 1.28 ($95\%$ CI 1.24, 1.32), the OR for being an early adopter was 1.50 ($95\%$ CI 1.41, 1.60), and the OR for being an always user was 1.92 ($95\%$ CI 1.63, 2.26). An alternative classification of providers showed similar results (see Supplemental materials). **Table 2** | Unnamed: 0 | Never Users | Late Adopters | Late Adopters.1 | Early Adopters | Early Adopters.1 | Always Users | Always Users.1 | | --- | --- | --- | --- | --- | --- | --- | --- | | | Never Users | OR (95% CI) | OR (95% CI) | OR (95% CI) | OR (95% CI) | OR (95% CI) | OR (95% CI) | | Male | 38.9% | Ref | Ref | Ref | Ref | Ref | Ref | | Female | 61.1% | 0.66 | (0.62, 0.71) | 0.64 | (0.57, 0.72) | 0.55 | (0.42, 0.74) | | Midwest | 24.1% | Ref | Ref | Ref | Ref | Ref | Ref | | Northeast | 20.6% | 1.76 | (1.60, 1.93) | 1.82 | (1.52, 2.18) | 3.54 | (2.22, 5.66) | | South | 36.3% | 1.67 | (1.53, 1.82) | 1.83 | (1.55, 2.16) | 2.49 | (1.57, 3.95) | | West | 18.9% | 1.46 | (1.32, 1.62) | 1.56 | (1.28, 1.90) | 1.89 | (1.09, 3.28) | | Other PCPs† | 75.2% | Ref | Ref | Ref | Ref | Ref | Ref | | Internal medicine | 24.8% | 1.49 | (1.40, 1.60) | 2.03 | (1.80, 2.28) | 2.84 | (2.12, 3.79) | | # Medicare beneficiaries /100 | 1.8 | 1.08 | (1.07, 1.09) | 1.08 | (1.07, 1.09) | 1.08 | (1.06, 1.09) | | # services encounters /100 | 9.4 | 1.00 | (1.00, 1.00) | 1.00 | (1.00, 1.00) | 1.00 | (1.00, 1.00) | | $ submitted charges /10,000 | 11.2 | 1.04 | (1.03, 1.04) | 1.04 | (1.04, 1.05) | 1.05 | (1.04, 1.06) | | % patients w/ hypertension /10 ‡ | 6.8 | 0.92 | (0.90, 0.96) | 0.95 | (0.86, 1.04) | 0.79 | (0.63, 1.17) | | % patients w/ diabetes /10 ‡ | 3.8 | 0.93 | (0.90, 0.96) | 0.87 | (0.82, 0.91) | 1.05 | (0.94, 1.17) | | % patients w/ hyperlipidemia patients /10 ‡ | 5.7 | 1.28 | (1.24, 1.32) | 1.50 | (1.41, 1.60) | 1.92 | (1.63, 2.26) | Fig 1 displayed the total number of claim counts for IBT services and the total number of unique Medicare beneficiaries reported in the public use data from 2013 to 2019 for providers who delivered IBT services to more than 10 fee-for-service Medicare beneficiaries in a year. Although over the years there had been a steady increase in utilization, both in the number of claims and number of beneficiaries, only about $1\%$ of the eligible Medicare population who met the criteria ever received IBT services from 2013 through 2019. **Fig 1:** *Intensive behavioral therapy (IBT) for obesity services utilization patterns (2013–2019).The number of claims for intensive behavioral therapy (IBT) for obesity services (solid line) and the number of beneficiaries who received the IBT services (dash line) documented in 2013 to 2019 Medicare Fee-for-Service Provider Utilization and Payment Data for Physicians and Other Suppliers.* Fig 2 presented choropleth maps of the number of Medicare beneficiaries who received IBT services per estimated 1,000 Medicare beneficiaries with obesity in each state for 2013 and 2019 (left panel), alongside the number of IBT providers delivering IBT services to more than 10 beneficiaries in a year per 1,000 eligible providers (right panel). Due to copyright issues, we had to use an open source tool at USGC.gov to recreate these maps which lost the formatting option of the states having filled colors that represented their values and only allowed us to manually select and plot different colored markers to represent the state’s value (For alternative format see S2 File). There was a faster increase in IBT beneficiaries occurring mainly in the South overall, however two states, New Jersey and New York, in the Northeast stood out and had the highest rate of IBT beneficiaries per 1,000 beneficiaries with obesity overall in 2019 (27.5 and 26.4, respectively). Although most providers were classified as never users, the choropleth maps confirmed the regional pattern that IBT uptake increased primarily in the South over the study time period, but once again New Jersey and New York had the overall highest rate of IBT Providers per 1,000 providers in 2019 (22.5 and 19.3, respectively). The state that saw the largest increase in the absolute number of IBT providers per 1,000 providers was Nevada (0 to 18.3 from 2013 to 2019); and New Hampshire and Vermont were the only two states that had a slight overall decrease in the number of IBT providers per 1,000 providers from 2013 to 2019 (1.0 to 0.0 and 2.4 to 2.0, respectively). **Fig 2:** *Choropleth maps for intensive behavioral therapy (IBT) utilization by Medicare beneficiaries and primary care providers.The number of Intensive behavioral therapy (IBT) beneficiaries per 1,000 Medicare population with Obesity (left panel) and the number of providers per 1,000 eligible primary care providers (right panel) who provided IBT services to more than 10 Medicare beneficiaries across the United States in 2013 and 2019.* ## Discussion This study was the first to classify providers by the IBT uptake timing to early, late, always or never adopter groups and examine differences between these groups. There was a pattern in the characteristics of providers delivering IBT services to more than 10 Medicare beneficiaries in a calendar year. Specifically, being female, non-internal medicine PCP, practicing in the Midwest region and small clinics were associated with lower probabilities of being an IBT user. Having higher proportions of patients with hyperlipidemia was associated with higher probabilities of being a late adopter, an early adopter, or an always user. This relationship between the patient composition and provider IBT-uptake was consistent with the need for obesity management in patients with high lipids. These patterns were consistent with practice-level characteristics associated with IBT uptake (e.g., larger practices are more likely to use IBT) [18], geographic distribution of the Medicare population with obesity (e.g., prevalence of beneficiaries with obesity is highest in the South) [19], and the complexity of obesity management (e.g., some PCPs such as nurse practitioners need more formal training for treating obesity) [20]. Although IBT utilization in the Medicare population across the U.S. has been steadily increasing over the years since the implementation of the CMS procedure codes, it is still a highly underutilized benefit among the Medicare population who qualify for the service. Approximately $28.5\%$ of Medicare beneficiaries are estimated to meet criteria for obesity, suggesting there are about 10 to 11 million fee-for-service beneficiaries that meet the criteria to receive IBT services per year [4, 21]. There are likely many factors affecting providers’ and Medicare beneficiaries’ decisions that result in the very low utilization of IBT for obesity treatment. Among the factors that may affect providers’ decisions to deliver IBT for obesity treatment are some of the structural features of the billing and payment requirements CMS established for the two HCPCS codes for individual and group sessions. For example, one potentially important factor is the low reimbursement rates for both individual and group sessions relative to primary care evaluation and management reimbursement levels. The rate of reimbursement for a 15-min consultation of IBT is $26, whereas payment for a typical evaluation and management code for an established patient for a level-2 visit (CPT code 99212) is $45 and a level-3 visit is $74 (CPT code 99213, Medicare payment for calendar year 2018). While CMS aimed to increase utilization of preventive service and screenings through procedure codes, the uptake rates of these services have been low. For example, $3\%$ of Medicare FFS beneficiaries received a visit specifically addressing depression screening in 2016 [22]. The G-codes for Annual Wellness Visits were introduced in 2011 with much higher allowable charge than obesity counseling services but the penetration was only 16~$17\%$ in 2014 [23, 24]. Another factor is likely the expectation of weekly visits for the first month and twice a month for the next five months, which for busy primary care practices could significantly curtail the available appointments for their entire patient panel. In addition, these frequent visits may be difficult for patients to achieve in certain populations with other comorbid conditions such as diabetes and patients with more resource constraints such as transportation barriers. Another important barrier to uptake is the restriction that the service needs to be provided in a primary care setting by a physician, nurse practitioner, physician assistant or a qualified provider under their direct supervision [25]. A non-physician auxiliary practitioner, such as a registered dietician (RD), may provide IBT but they must bill “incident to” the primary care physician, who must be physically present at the time services are provided [11]. Referral to RDs and other auxiliary providers who work outside of the primary care setting is not covered by CMS [11]. In the Medicare Fee-for-service data from 2013 to 2019, we found the following provider specialty types who submitted Medicare claims for IBT services: general practice, family practice, internal medicine, obstetrics/gynecology, pediatric medicine, geriatric medicine, nurse practitioner, certified nurse specialist, or physician assistant. Had other providers been certified to provide obesity service it might increase uptake of the benefit. In Table 3, we summarized a list of barriers as well as potential solutions for low IBT provider-uptake and patient-utilization rates based on our findings from provider interviews and surveys [18, 20]. **Table 3** | Barriers to IBT service provider uptake and patient utilization | Potential Solution | | --- | --- | | Reimbursement does not fully cover cost of clinician personnel, facility, and program (care coordination) expenses. | Add coverage for care coordination and facility expenses. | | Requirement for delivery under direct supervision of PCP. | Broaden the types of providers (registered dieticians, behavioral health workers, nurses, social workers) who can deliver IBT and bill independently. | | Requirement for weekly (first month) and biweekly (month 2 and beyond) in-person visits is burdensome to patients. | Allow telehealth delivery of IBT services. | | Requirement for specified weight loss of 3 kgs in the first 6 months to continue IBT billing. | Remove weight loss specification. Behavioral changes require time to take effect. Obesity is a chronic condition, not subacute or temporary one. | Two similar studies have examined the uptake rates of IBT services since its implementation in 2012 [13, 14]. The first study examined utilization in 2012 and 2013 [13] and the second study for 2012 through 2015 [14]. These two studies showed very similar results in terms of overall low IBT utilization. However, in contrast to our study, their analyses focused on Medicare beneficiary characteristics and not provider characteristics. Using patient-level claims data, both studies described the demographics of patients who received IBT services. Our study focused on provider characteristics by IBT-uptake status and geographic differences, which have not yet been evaluated in similar studies. This new information allows us to further understand where IBT uptake is extremely low and how it varies across different provider types. Also, these results examine the data over a longer period of time, indicating the trend towards low utilization has continued. At the time of the study, the 2020 data became available. However, we decided to exclude it from our analyses because the data may be misleading due to the effect on health care visits of the early part of the COVID-19 pandemic. The number of IBT claims dropped to 245,442 (a $15\%$ drop compared with 2019 data) and the number of unique beneficiaries dropped to 134,560 (a $11\%$ drop). Access to non-emergent health care services was limited during the last months of 2020. One limitation of this work is that we were not able to assess providers who did provide IBT for obesity but only for 10 or fewer beneficiaries because of data suppression in the CMS publicly available data, which may lead to a slight underestimation. Some late adopters may have been misclassified when they used some IBT services in the early years to a lesser extent. Compared with the previous report [14] our estimated number of beneficiaries in 2013 was slightly lower (46,821 from the previous report versus our finding of 44,572). Therefore, our results may underrepresent the true amount of IBT for obesity happening in the earlier years. However, our estimates in 2014 (46,171 from the previous report versus our finding of 57,488) and 2015 (57,576 from the previous report versus our finding of 73,234) were higher which might be due to the updates of final-action claim items over time. The differences in our estimates and previous reports were small and would not change the qualitative conclusions of the study. Another limitation of the public use data is that it only contains information on Medicare fee-for-service beneficiaries and as such it may not be representative of a physician’s entire practice pattern. A third limitation of the study is that when we classified providers into early, late, always or never adopters we did not exclude providers who were not in the National Plan and Provider Enumeration System until after 2014 or providers who were deactivated between 2013 and 2019. The number of providers in the cohort would be smaller if we had done so. We found 141,313 out of the 575,936 PCPs in our cohort had the first enumeration year after 2014, among whom 140,808 were classified as never users in our analysis and 505 as late adopters. Using the CMS Deactivated Providers list as of July 2022, we found 6,049 of the 575,936 PCPs in our cohort were deactivated in the study period, among whom 5,985 were never users, 35 were late adopters and 29 were early adopters. We do not expect these limitations to affect our results qualitatively. Our study provides the targeting regions and provider types to increase utilization in the future. Many structural changes may be needed to improve the uptake of IBT services for Medicare beneficiaries. Low reimbursement rates relative to other services provided by or under the direct supervision of PCPs are often cited as a reason PCPs do not offer IBT or similar types of services to Medicare beneficiaries [26]. Another structural change CMS could adopt to increase beneficiary access to IBT for obesity services is to remove the requirement for delivery by or under the direct supervision of a PCP. Registered dietitians are authorized Medicare Part B providers and CMS could change the coverage determination to permit them to independently deliver and bill for IBT for obesity services. As evidenced by the recent rapid transition to more telehealth services, IBT for obesity services are a good candidate for consideration of delivery via telehealth modes that would enhance the access to these services by Medicare beneficiaries who face transportation barriers in accessing in-person services. Finally, these results identify states with higher uptake that researchers can target for a further study regarding practice characteristics that are related to successful adoption of the service. ## Conclusion This study documents the extremely low use of IBT for obesity benefit in Medicare and regional variations in uptake rates with the South and Northeast U.S. regions having the highest number of IBT-billing providers. Adopters of the IBT services tend to differ systematically from the never users. Without CMS actively improving the benefit design this valuable service will continue to be underutilized. Further investigation is needed to identify barriers to the uptake of IBT services among PCPs. ## References 1. Doshi JA, Polsky D, Chang VW. **Prevalence And Trends In Obesity Among Aged And Disabled U.S. Medicare Beneficiaries, 1997–2002**. *Health Aff (Millwood)* (2007.0) **26** 1111-1117. DOI: 10.1377/hlthaff.26.4.1111 2. Chang VW, Alley DE, Dowd JB. **Trends in the Relationship Between Obesity and Disability, 1988–2012**. *Am J Epidemiol* (2017.0) **186** 688-695. DOI: 10.1093/aje/kwx092 3. 3United Health Group. Our Nation’s Health: Obesity Rate at All-Time High; Childhood Poverty Continues to Decline. [cited 25 Jul 2019]. https://www.unitedhealthgroup.com/newsroom/2018/2018-12-12-our-nations-health-annual-report.html 4. 4Explore Obesity in the United States | 2019 Senior Report. In: America’s Health Rankings [Internet]. [cited 25 Jul 2019]. https://www.americashealthrankings.org/explore/senior/measure/obesity_sr/state/ALL 5. 5Report to the Congress: Medicare and the Health Care Delivery System. [cited 26 Nov 2019]. http://www.medpac.gov/-documents-/reports 6. 6The Healthcare Costs of Obesity. In: The State of Obesity [Internet]. [cited 25 Jul 2019]. https://www.stateofobesity.org/healthcare-costs-obesity/ 7. Batsis JA, Huyck KL, Bartels SJ. **Challenges with the Medicare Obesity Benefit: Practical Concerns & Proposed Solutions**. *J Gen Intern Med* (2015.0) **30** 118-122. DOI: 10.1007/s11606-014-3031-6 8. Moyer VA. **Screening for and management of obesity in adults: U.S. Preventive Services Task Force recommendation statement**. *Ann Intern Med* (2012.0) **157** 373-378. DOI: 10.7326/0003-4819-157-5-201209040-00475 9. Batsis JA, Zagaria AB. **Addressing Obesity in Aging Patients**. *Med Clin North Am* (2018.0) **102** 65-85. DOI: 10.1016/j.mcna.2017.08.007 10. 10Intensive Behavioral Therapy for Obesity. 19 Nov 2019 [cited 28 Jul 2022]. https://www.hopkinsmedicine.org/health/treatment-tests-and-therapies/intensive-behavioral-therapy-for-obesity 11. Wadden TA, Tsai AG, Tronieri JS. **A Protocol to Deliver Intensive Behavioral Therapy (IBT) for Obesity in Primary Care Settings: The MODEL-IBT Program**. *Obesity* (2019.0) **27** 1562-1566. DOI: 10.1002/oby.22594 12. 12Ethicon. 2018 Obesity Counseling Reimbursement Fact Sheet. 2018. https://sa1s3.patientpop.com/assets/docs/51094.pdf 13. Batsis JA, Bynum JPW. **Uptake of the centers for Medicare and Medicaid obesity benefit: 2012–2013**. *Obesity* (2016.0) **24** 1983-1988. DOI: 10.1002/oby.21578 14. Dewar S, Bynum J, Batsis JA. **Uptake of Obesity Intensive Behavioral Treatment Codes in Medicare Beneficiaries, 2012–2015**. *J Gen Intern Med* (2019.0). DOI: 10.1007/s11606-019-05437-1 15. 15Medicare Provider Utilization and Payment Data: Physician and Other Supplier—Centers for Medicare & Medicaid Services. [cited 30 Jul 2019]. https://www.cms.gov/research-statistics-data-and-systems/statistics-trends-and-reports/medicare-provider-charge-data/physician-and-other-supplier.html 16. 16Explore Obesity in the United States | 2017 Annual Report. In: America’s Health Rankings [Internet]. [cited 31 Jul 2019]. https://www.americashealthrankings.org/explore/annual/measure/Obesity/state/ALL 17. 17Population Distribution by Age. In: The Henry J. Kaiser Family Foundation [Internet]. 29 Nov 2018 [cited 31 Jul 2019]. https://www.kff.org/other/state-indicator/distribution-by-age/ 18. Luo Z, Gritz M, Connelly L, Dolor RJ, Phimphasone-Brady P, Li H. **A Survey of Primary Care Practices on Their Use of the Intensive Behavioral Therapy for Obese Medicare Patients**. *J Gen Intern Med* (2021.0) **36** 2700-2708. DOI: 10.1007/s11606-021-06596-w 19. Gartner DR, Taber DR, Hirsch JA, Robinson WR. **The spatial distribution of gender differences in obesity prevalence differs from overall obesity prevalence among US adults**. *Ann Epidemiol* (2016.0) **26** 293-298. DOI: 10.1016/j.annepidem.2016.02.010 20. Nederveld A, Phimphasone-Brady P, Connelly L, Fitzpatrick L, Holtrop JS. **The Joys and Challenges of Delivering Obesity Care: a Qualitative Study of US Primary Care Practices**. *J Gen Intern Med* (2021.0) **36** 2709-2716. DOI: 10.1007/s11606-020-06548-w 21. 21Total Number of Medicare Beneficiaries. In: The Henry J. Kaiser Family Foundation [Internet]. 12 Mar 2019 [cited 27 Nov 2019]. https://www.kff.org/medicare/state-indicator/total-medicare-beneficiaries/ 22. Colligan EM, Cross-Barnet C, Lloyd JT, McNeely J. **Barriers and facilitators to depression screening in older adults: a qualitative study**. *Aging Ment Health* (2020.0) **24** 341-348. DOI: 10.1080/13607863.2018.1531376 23. Tao G.. **Utilization pattern of other preventive services during the US Medicare annual wellness visit**. *Prev Med Rep* (2018.0) **10** 210-211. DOI: 10.1016/j.pmedr.2017.12.014 24. Misra A, Lloyd JT, Strawbridge LM, Wensky SG. **Use of Welcome to Medicare Visits Among Older Adults Following the Affordable Care Act**. *Am J Prev Med* (2018.0) **54** 37-43. DOI: 10.1016/j.amepre.2017.08.030 25. 25Centers for Medicare & Medicaid Services. National Coverage Determination (NCD) for Intensive Behavioral Therapy for Obesity (210.12). 2012 [cited 14 Jul 2020]. https://www.cms.gov/medicare-coverage-database/details/ncd-details.aspx?NCDId=353 26. 26Academy of Nutrition and Dietetics. Meeting the Need for Obesity Treatment: A Toolkit for the RD/PCP Partnership. 2012 [cited 14 Jul 2020]. http://www.eatrightpro.org/~/media/eatrightpro%20files/practice/patient%20care/medical%20nutrition%20therapy/meeting%20the%20need%20for%20obesity%20treatment%20pdf.ashx
--- title: 'Epidemic efficacy of Covid-19 vaccination against Omicron: An innovative approach using enhanced residual recurrent neural network' authors: - Rakesh Kumar - Meenu Gupta - Aman Agarwal - Anustup Mukherjee - Sardar M. N. Islam journal: PLOS ONE year: 2023 pmcid: PMC10038250 doi: 10.1371/journal.pone.0280026 license: CC BY 4.0 --- # Epidemic efficacy of Covid-19 vaccination against Omicron: An innovative approach using enhanced residual recurrent neural network ## Abstract The outbreak of COVID-19 has engulfed the entire world since the end of 2019, causing tremendous loss of lives. It has also taken a toll on the healthcare sector due to the inability to accurately predict the spread of disease as the arrangements for the essential supply of medical items largely depend on prior predictions. The objective of the study is to train a reliable model for predicting the spread of Coronavirus. The prediction capabilities of various powerful models such as the Autoregression Model (AR), Global Autoregression (GAR), Stacked-LSTM (Long Short-Term Memory), ARIMA (Autoregressive Integrated Moving Average), Facebook Prophet (FBProphet), and Residual Recurrent Neural Network (Res-RNN) were taken into consideration for predicting COVID-19 using the historical data of daily confirmed cases along with Twitter data. The COVID-19 prediction results attained from these models were not up to the mark. To enhance the prediction results, a novel model is proposed that utilizes the power of Res-RNN with some modifications. Gated Recurrent Unit (GRU) and LSTM units are also introduced in the model to handle the long-term dependencies. Neural Networks being data-hungry, a merged layer was added before the linear layer to combine tweet volume as additional features to reach data augmentation. The residual links are used to handle the overfitting problem. The proposed model RNN Convolutional Residual Network (RNNCON-Res) showcases dominating capability in country-level prediction 20 days ahead with respect to existing State-Of-The-Art (SOTA) methods. Sufficient experimentation was performed to analyze the prediction capability of different models. It was found that the proposed model RNNCON-Res has achieved $91\%$ accuracy, which is better than all other existing models. ## 1. Introduction The first case of COVID-19 was encountered in Wuhan city, China, in Nov 2019. This deadly disease had a very bad impact on human life, and its fallout affected the socio-economic conditions around the world. Before being declared a pandemic by the World Health Organisation (WHO), the Public Health Emergency of International Concern (PHEIC) was announced on 11th March 2020 [1]. The pandemic was caused due to respiratory viruses such as SARS-CoV-2 [2]. In India, the first case of COVID-19 was reported in Thrissur, Kerala, on January 30th, 2020 [3]. The transmission was monitored using a mathematical model named “The Indian Supermodel” [4, 5]. The new variant, named the “Delta variant,” was discovered in India in late 2020 [6], which was the major factor leading to the second wave of infections in early April 2021 [7]. The number of active cases reached 2.5 million towards the end of April 2021, and the average number of cases reported every day touched the mark of 300,000 [8]. In early January 2022, the emergency use of vaccines like Covishield (i.e., developed by the University of Oxford) [9] and Covaxin (i.e., developed by Bharat Biotech in collaboration with the National Institute of Virology) [10] was approved by Drug Controller General of India. The Indian vaccination program began on 16 January 2021 and has crossed 1.8 billion cumulative dosages as of 14th April 2022 [11]. The double vaccine dosage system greatly decreased the number of COVID-19 cases. Still, after the advent of Omicron (a variant of SARS-CoV-2), it has been observed that many people have contracted corona even after proper vaccination. Due to the single-stranded RNA genome of SARS-CoV-2 has a high intrinsic mutation rate, which leads to immunological escape that can quickly spread in the population that has received vaccinations [12]. Omicron also displayed remarkable immunological escape power compared to the other Variants of Concern (VOC). The Omicron was first identified in November 2021 in Gauteng province, South Africa was promptly classified as a VOC on November 26, 2021 [13, 14]. This forced other nations in the world to take strict actions to stop the spread of this new deadly variant. After its identification in South Africa, this VOC rapidly disseminated to various parts of the world. According to Report 49 from the Imperial College London COVID-19 response team, the Omicron variation has a 5.4 times higher risk of reinfection than the Delta version. This suggests that the protection against reinfection provided by prior infection against Omicron reinfection may be as low as $19\%$ [15]. The number of mutations found in the omicron variant is much higher than in other variants. The omicron variant had 32 mutations, twice the number of mutations in the Delta variant. These mutations were connected to increased transmissibility or immune evasion and had been discovered in variations including Delta and Alpha [16]. The spike (S) protein (site for antibody binding) has passed through these alterations, which produced the Omicron strain in its high infectivity and transmissibility traits. There is still much to be clarified, despite numerous investigations having sought to comprehend this new obstacle in the COVID-19 strains [17]. A new argument has emerged in favor of natural vaccinations in response to the distinctive identification of the Omicron variation. Omicron is comparable to live attenuated vaccines in several ways, which leads a number of specialists to assume that it could function as a natural vaccination. The high rate of antibody production in those who had recovered from Omicron was also emphasized as supporting evidence for the hypothesis put forth by some researchers that Omicron functions as a natural vaccine. There have also been some controversies since, like the earlier variations, it has serious health consequences and a high infection rate. However, this idea was opposed by some experts [18]. According to recent research, *Omicron is* remarkably resistant to the neutralizing effects of vaccinations, convalescent serum, and the majority of antibody treatments. Additionally, while most antiviral medications under research are effective against Omicron, only a small number of neutralizing antibodies are potent against it [19]. It is observed that a lot of work has been done in analyzing Chest X-Ray images using CNNs to classify the Coronavirus disease. These analyses are majorly focused on classifying the COVID-19 outbreak. Hence, there is a need to develop a model which trains on historic COVID-19 data and accurately predicts the future of the disease compared to existing methods. The main aim of this proposed study is to provide a future prediction of COVID-19-confirmed cases using time-series data. This model will be able to accurately predict COVID-19 cases beforehand, which could help healthcare departments to take the required measures to arrange medical supplies. This could prevent unwanted loss of lives that are encountered due to a lack of infrastructure. The proposed model is based on RNN architecture using GRU and LSTM units. Furthermore, the power of residual learning is also exploited, which ultimately delivers better results when compared to other SOTA methods. This paper is further classified into different parts. Section 2 discusses the different researcher’s views and analyses of Omicron using different techniques. Section 3 discusses the data discerption and methodology used for analysis. The experiment results analysis over Omicron after vaccination is discussed in section 4. Finally, this paper is concluded in section 5 with its future aspects. ## 2. Literature review Medicine and epidemiology have extensively used Machine Learning (ML) and DL techniques in the past few years. The previous research work done by researchers is extensively discussed below. In [20], the authors gave a comparison of soft computing and ML models to anticipate the COVID-19 outbreak as an alternative to the SIR (Susceptible-Infected-Recovered) and SEIR (Susceptible-Exposed-Infectious-Removed) models. They showed Adaptive Network-Based Fuzzy Inference System (ANFIS), and Multi-Layered Perceptron (MLP) gave the best results. Further, they suggested ML is a potential tool to predict this outbreak as it is highly complex in nature. In [21], the authors used a DCNN model based on class decomposition known as "Decompose, Transfer, and Compose (DeTraC)" for the classification of COVID-19 in chest X-ray images. This model was applied to enhance the functionality of previously trained models. It added a class decomposition layer to the pre-trained models and considered each subset an independent class. As a result, the image dataset’s classes were broken down into multiple sub-classes, which were then reassembled to provide the final predictions. This allowed for the classification of medical images under the limited availability of annotated medical images, and hence the model achieved an accuracy of $93.1\%$. In [22], the author used a DCNN-based model (called Inception V3) with transfer learning to detect COVID-19 automatically from chest X-Ray images. The model achieved a validation accuracy is $93\%$. In [23], the authors compared recent DCNN algorithms such as VGG16, VGG19, DenseNet201, InceptionResNetV2, InceptionV3, Resnet50, and MobileNetV2 for the classification of X-Ray images to detect and classify coronavirus pneumonia. Fine-tuned versions of the DCNN mentioned above models were also designed in addition to weight decay and L2-regularizers, which were used to avoid over-fitting. These models were also tested on a CT dataset for multiclass classification. Finally, it was found that InceptionResNetV2 and DenseNet201 provided better results as compared to other models, with an accuracy of $92.18\%$ and $88.09\%$, respectively. In [24], the authors presented an alternative modeling framework to CNNs for identifying positive COVID-19 cases based on Capsule Networks (CapsNets), which can handle smaller datasets. CapsNets are alternative models that use ‘routing by agreement’ to capture spatial information. The proposed framework, termed ‘COVID-CAPS,’ achieved an accuracy of $95.7\%$, an Area Under the Curve (AUC) of 0.97, and sensitivity and specificity of $90\%$ and $95.8\%$, respectively. In [25], the authors discussed that AI could easily analyze symptoms and detect warning signs, thus helping in early detection, diagnosis of the infection, and monitoring. It can track the spread of the virus to a certain extent at different scales, such as epidemiological, medical, and molecular scales. AI helps in projecting mortality and active cases in any region. It can also be used in drug development and diagnostic test design, helping accelerate vaccine development and making clinical trials safer. In [26], the authors proposed a procedure for deriving features for the training of Res-RNN to predict confirmed cases of COVID-19 based on meteorological factors like humidity and temperature. The proposed procedure decomposes all features and the signal to be predicted into its stationary and non-stationary modes, which were used to train separate Res-RNN. The results were summed to derive the final forecast of COVID-19 cases. A MAPE value of $4.68\%$ was achieved, which confirms the applicability of the technique. In [27], the authors developed a nature-inspired Algorithm, a hybridized approach where an enhanced version of the Beetle Antennae Search (BAS) algorithm was used to improve the prediction model performance and to determine the parameters of the ANFIS. Antecedent and conclusion ANFIS parameters were taken into consideration. This method was then evaluated using WHO’s official data on the COVID-19 outbreak. ## 3. Proposed methodology and data set This section discusses the dataset description and proposed model formulation for analyzing the impact of Omicron after vaccination. ## 3.1. Dataset description The data is collected from the HDX Novel Coronavirus (COVID-19) Cases dataset, available online [28]. This data has been recorded since 22 January 2020 and is categorized country-wise, which is updated daily and maintained by Johns Hopkins University Centre for Systems Science and Engineering (JHU CSSE). The data is divided into the following categories: confirmed cases, vaccinated cases, recovered cases, unconfirmed cases, and daily tweet volume (which shows the change in the number of cases with several tweets). The data were divided into training and validation, $90\%$ and $10\%$, respectively. The details about the dataset are given in Table 1. **Table 1** | Dataset | Number of rows | Number of columns | Training | Testing | | --- | --- | --- | --- | --- | | HDX Covid-19 Dataset | 918 | 2 | 818 | 100 | ## 3.2. Proposed methodology This work proposes an RNN technique-based model that is prominent in handling sequential time series data. While training the RNN network, we encounter a major problem called vanishing gradients as the network goes deeper. This problem arises due to the continuous multiplication of weights while Backpropagation Through Time (BPTT) with increasing requirements of learning long-term dependencies. GRU and LSTM units were introduced into the network to solve this problem, as shown in Fig 1. These units focus on the important features that must be carried forward in the network. After using the GRU and LSTM units, the model faced the problem of learning from repeated data. It is very common for a time series dataset to have repeating data. As the GRU units tend to learn from repeated data, it becomes extremely harmful for the model as the model tends to get biased for a particular data point. Hence, a residual architecture is used in the network, combining the power of Residual architecture aided with GRU and LSTM units. These units are used as gates to regulate which input is being transferred to the residual network. **Fig 1:** *Framework architecture.* As mentioned, BPTT updates the gradients while learning long-term dependencies in RNNs. BPTT is similar to the backpropagation used in standard neural networks and CNNs. Because of continuous multiplication, the weights converge to near zero, preventing the model from learning or updating itself. The problem of vanishing gradients is also addressed using GRU units which handle it efficiently. GRU consists of two gates, i.e., update and resets gates, respectively. The 1st gate determines what amount of information is required to pass into the next unit, while the 2nd gate decides how much information is required to erase. In this way, GRU helps keep only a certain amount of important information. The different gates in GRU are described in Eq [1]: p=σ(Wpxt+Uiat−1) q=σ(Wqxt+Uqat−1) r=tanh(Wrxt+Ur(at−1⊙q)) at=(1−p)⊙r+p⊙at−1 [1] Here, p and q refer to reset and update gates, respectively. W is the weighted parameter, a is the hidden state, and ⊙ refers to element-wise multiplication. Next, Residual learning is used to reduce redundancy in learning. Residual Networks have recently achieved prominent results in developing CNN models and hence are introduced in RNNs too. Hence, residual links have been added to the GRU units. These residual links extract the unique information from the data and forward it to the network. The shortcut connections or the residual links do not add extra parameters and hence do not increase the complexity of the model. Finally, the output from the GRU and LSTM units is guided to a merged layer that assembles the output from the corresponding GRU units and the output from the tweet volume. The result is then passed through a linear layer where the assembled output provided by the merged layer is normalized. Residual Recurrent Network (RRN) is used in the model, as shown in Fig 1. The basic architecture of RRN is very similar to a simple RNN. Let each hidden state be represented by a = {a1,…,aT}, T being steps of hidden states and sequential input be represented by x = {x1,…,xT}. Identity connections are made from at−1 unit to at unit and recurrent transformation also takes place, which takes at and xt as input, then recurrent transformation takes place in the form given in Eq [2]. Here, F is the residual function, and W is the weighted parameters. g(at−1) is the identity function. h is the activation function which is traditionally set as tanh. As higher recurrent depth leads to better performance, the depth is taken as K. The state-wise output is shown in Eq [3]: y1t=σ(xtW1+at−1U1+b1) y2t=σ(xtW2+y1tU2+b2) … yKt=σ(xtWK+yK−1tUK+bk) F(at−1,xt)=yt [3] Also, xt is taken at every layer. LSTM is based on the basic structure of RNN, which avoids gradient vanishing and is a powerful tool for remembering long-range dependencies. At each time step t, inputs (xt, ct−1, at−1) are sent to three gates, namely, the input gate, output gate, and forget gate (ct−1 denotes previous memory state) which generates three signals, it, ot, ft respectively. The mathematical formulation is shown in Eq [4] ft=σ(xt*Uf+at−1*Wf) it=σ(xt*Ui+at−1*Wi) ot=σ(xt*U0+at−1*Wo) ct˜=tanh(xt*Uc+at−1*Wc) ct=ft*ct−1+it*ct˜ [4] ## 3.3. Metrics used The performance of the proposed model is analyzed based on different performance matrices such as True Positive (TP), False Positive (FP), True Negative (TN), and False Negative (FN). The accuracy, precision, recall, Negative Predictive Value, and sensitivity of the model are calculated using Eq [5] to Eq [9]. Mean Squared Error (MSE) gives a better idea when dealing with time-series or regression tasks. MSE is the mean of the square of the error between the predicted value and the ground truth as shown in Eq [10]. ## 4. Experimental results and analysis The results of the proposed RNNCON-Res and SOTA methods (i.e., RNN and Res-RNN) are presented and compared in this section. Res-RNN provided better results when compared to simple RNNs as the residual links in the network help in focusing important features in the input as a shortcut connection carries the information over the network. When the proposed model is compared with the Res-RNN model, it outperforms in terms of accuracy due to the involvement of the GRU unit in the network, as it solves the problem of vanishing gradients while remembering the long-term dependencies. The covid-19 dataset is combined with tweet volume as an additional feature used to reach data augmentation. The proposed model provides better results when compared with SOTA methods, as shown in Fig 2. **Fig 2:** *Comparison of the number of covid case predictions using RNNCON-Res and other models.* The Autoregressive Model (AR) model is very popular for forecasting time series data using a linear combination of past values of the variable [29]. An order-p autoregressive model can be mathematically described as shown in Eq [11]: xt=αpxt−p+εt+c [11] The prediction for a given time-stamp t is a weighted sum of past data points in a given period of size ‘w’ multiplied by a parameter ap. White noise is introduced, which is a small random noise represented by εt+h. The deviation between the true and linear values can be explained using this. c is intercept s. The GAR is a model which is trained with one set of αp and c for all the sources when the signal received from all the different sources shows the same patterns and data required to train the system is limited. Vector Autoregression (VAR) is a forecasting algorithm used in multivariate time series models, as shown in Eq [12]: x¯t+h=∑$$p \leq 0$$w−iApxt−p+εt+h+c [12] Where signal-wise αp in AR is replaced with a matrix Ap to capture the correlation information. Fig 3 shows the comparison of RNNCON-Res with ResRNN. In this, ResRNN can predict the rise and fall of the prediction line of the graph but fails to give accurate numbers of covid-19 cases. However, RNNCON-Res can predict accurately. Finally, it is concluded that the results of the proposed model are better than other existing models. **Fig 3:** *Comparison of the number of covid case predictions using RNNCON-Res and ResRNN with true value.* The proposed model is also compared with ARIMA and FBProphet time series models, as shown in Fig 4. It can be observed that ARIMA performs better in predicting the covid-19 cases compared to FBProphet. However, these models do not perform better than the proposed RNNCON-Res model. **Fig 4:** *Comparison of the number of covid-19 cases prediction using (a) ARIMA and (b) FBProphet with True Value.* The $91\%$ accuracy has been achieved using the RNNCON-Res model, as discussed in Fig 5. As RNNCON-Res gave promising results, Exploratory Data Analysis d(EDA) was performed to plot the graph of the number of probable cases based on ground truth. Fig 6(A) shows the total covid-19 cases with the ground truth values. Based on the present scenario, the model can forecast the total number of cases up to 20 days. **Fig 5:** *Graphs for Training and validation for RNNCON-Res.(a) loss (b) accuracy.* **Fig 6:** *Predictions based on RNNCON-Res (a) Shows total probable COVID-19 cases based on ground truth (b) Shows probable Omicron and Delta variant cases.* The probable cases of each Omicron and Delta variant were also estimated using the model shown in Fig 6(B) and the total number of recovered cases over the total number of covid-19 cases as shown in Fig 7. **Fig 7:** *Total number of confirmed, death, and recovered cases.* The exact details for the performance of the proposed model and the results generated for different classes are better described using a confusion matrix, as shown in Fig 8. It shows individual values of different metrics given in Eq [5] to Eq [9]. The value achieved for Precision, Negative Predictive Value, specificity, sensitivity, and accuracy are 0.904, 0.926, 0.914, 0.917, and $91.6\%$ respectively. **Fig 8:** *Confusion matrix for result analysis of RNNCON-Res.* ## 5. Discussion of performance The results received from RNNCON-Res are compared with other methods, such as Recurrent Neural Model [30] and Stacked-LSTM [31], as discussed in Table 2. Stacked-LSTM is advanced by using LSTM units instead of simple recurrent units used in the Recurrent Neural Model for better results. $90\%$ accuracy was achieved using Stacked-LSTM. In the proposed model, LSTM and GRU units are used along with a residual network, which provides better results than Stacked-LSTM and achieved the benchmark of $91\%$ accuracy. **Table 2** | S. No | Model | Dataset used | Accuracy (in %) | | --- | --- | --- | --- | | 1.0 | RNNCON-Res (Proposed) | HDX | 91 | | 2.0 | Recurrent Neural Model | UCI ML Repository | 88 | | 3.0 | Stacked-LSTM | HDX | 90 | The MSE for RNNCON-*Res is* calculated to be 4067567.11. The model is also compared with other models like ARIMA [32], BSTS [32], and NAR Neural Model [33]. The ARIMA and BSTS model delivered RMSE of 4391 and 3874 on the HDX dataset, and the NAR Neural Model delivered RMSE of 47366 on the entire data. RNNCON-Res performed the best among these models reaching RMSE of 2016. ## 6. Conclusion and future scope Predicting disease propagation is an essential part of disease management and control. Some successfully applied models include ANFIS, MLP, DeTraC, VGG16, VGG19, etc. But, in many cases, predictive neural networks may struggle due to small datasets. This paper predicted the outbreak of COVID-19 in the US using historic daily confirmed cases and Twitter data. The prediction capabilities of various powerful models such as the AR, GAR, Recurrent Neural Model, Stacked-LSTM, and Res-RNN were taken into consideration while predicting the outbreak of COVID-19 in the US using historic daily confirmed cases and Twitter data, which did not give very accurate results. The proposed model uses GRU and LSTM units with Residual links to tackle the over-fitting problem and to remember long-term dependencies. Data augmentation was facilized by introducing a merged layer before the linear layer to use tweet volume as an additional feature (as neural networks are known to require huge data). The RNNCON-Res model demonstrated dominating capability in country-level prediction 20 days ahead. Also, RNNCON-Res can be used to solve any time series forecasting problem. In the future, this model can be used in reinforcement learning with sequential data, making it a self-learning machine to better predict the spread of COVID. ## References 1. **Statement on the second meeting of the International Health Regulations (2005)**. *Emergency Committee regarding the outbreak of novel coronavirus (2019-nCoV)* (2020.0) 2. **Naming the coronavirus disease (COVID-19) and the virus that causes it**. (2020.0) 3. Andrews M. A., Areekal B., Rajesh K. R., Krishnan J., Suryakala R., Krishnan B.. **First confirmed case of COVID-19 infection in India: A case report**. (2020.0) **151** 490-492. DOI: 10.4103/ijmr.IJMR_2131_20 4. 4Department of Science and Technology. DST initiates COVID-19 India National Supermodel for monitoring infection transmission & aid decision-making by policymakers. https://dst.gov.in/dst-initiates-covid-19-india-national-supermodel-monitoring-infection-transmission-aid-decision. (Accessed: April 12, 2022). 5. Agrawal M., Kanitkar M., Vidyasagar M.. **Modelling the spread of SARS-CoV-2 pandemic-Impact of lockdowns & interventions**. (2021.0) **153** 175-181. DOI: 10.4103/ijmr.IJMR_4051_20 6. Peacock S.. **“Expert reaction to cases of variant B.1.617 (the ’Indian variant’) being investigated in the UK**. (2021.0) 7. **India’s ’double mutation’ covid virus variant is worrying the world.**. *Bloomberg.* (2021.0) 8. Gettleman J., Yasir S., Kumar H., Raj S., Loke A.. **As Covid-19 devastates India, deaths go undercounted**. (2021.0) **24** 9. Gaurav K.. **Govt’s expert panel approves AstraZeneca/Oxford Covid-19 vaccine for emergency use**. (2021.0) 10. **Expert panel recommends Bharat Biotech’s Covaxin for restricted emergency use"**. *News18* (2021.0) 11. **COVID-19 vaccination Ministry of Health and Family Welfare**. *Total Vaccination* 12. Mascola J. R.. **"SARS-CoV-2 viral variants-tackling a moving target [published on February 11, 2021]**. DOI: 10.1001/jama20212088 13. **Classification of Omicron (B.1.1.529)**. *SARS-CoV-2 variant of Concern* (2021.0) 14. Choudhary O. P., Dhawan M.. **Omicron variant (B. 1.1. 529) of SARS-CoV-2: threat assessment and plan of action**. (2022.0) **97** 106187. DOI: 10.1016/j.ijsu.2021.106187 15. Zhongming Z., Linong L., Xiaona Y., Wangqiang Z., Wei L.. **Omicron largely evades immunity from past infection or two vaccine doses**. (2021.0) 16. Song Y., Masaki F.. **Preparation for the challenge of heavily mutated Omicron variant**. (2021.0) **11**. DOI: 10.1002/ctm2.679 17. Arora S., Grover V., Saluja P., Algarni Y. A., Saquib S. A., Asif S. M.. **Literature review of Omicron: a grim reality amidst COVID-19**. (2022.0) **10** 451. DOI: 10.3390/microorganisms10020451 18. Das S., Samanta S., Banerjee J., Pal A., Giri B., Kar S. S.. **Is Omicron the end of pandemic or start of a new innings?**. (2022.0) 102332. DOI: 10.1016/j.tmaid.2022.102332 19. Fan Y., Li X., Zhang L., Wan S., Zhang L., Zhou F.. **SARS-CoV-2 Omicron variant: recent progress and future perspectives**. (2022.0) **7** 1-11. PMID: 34980881 20. Ardabili S. F., Mosavi A., Ghamisi P., Ferdinand F., Varkonyi-Koczy A. R., Reuter U.. **Covid-19 outbreak prediction with machine learning**. (2020.0) **13** 249 21. Abbas A., Abdelsamea M. M., Gaber M. M.. **Classification of COVID-19 in chest X-ray images using DeTraC deep convolutional neural network**. (2021.0) **51** 854-864. DOI: 10.1007/s10489-020-01829-7 22. Asif S., Wenhui Y., Jin H., Jinhai S.. **Classification of COVID-19 from chest X-ray images using deep convolutional neural network. In**. (2020.0) 426-433 23. El Asnaoui K., Chawki Y.. **Using X-ray images and deep learning for automated detection of coronavirus disease**. (2021.0) **39** 3615-3626. DOI: 10.1080/07391102.2020.1767212 24. Afshar P., Heidarian S., Naderkhani F., Oikonomou A., Plataniotis K. N., Mohammadi A.. **Covid-caps: A capsule network-based framework for identification of covid-19 cases from x-ray images**. (2020.0) **138** 638-643. DOI: 10.1016/j.patrec.2020.09.010 25. Vaishya R., Javaid M., Khan I. H., Haleem A.. **Artificial Intelligence (AI) applications for COVID-19 pandemic.**. (2020.0) **14** 337-339 26. Mousavi M., Salgotra R., Holloway D., Gandomi A. H.. **COVID-19 time series forecast using transmission rate and meteorological parameters as features**. (2020.0) **15** 34-50 27. Zivkovic M., Bacanin N., Venkatachalam K., Nayyar A., Djordjevic A., Strumberger I.. **COVID-19 cases prediction by using hybrid machine learning and beetle antennae search approach**. (2021.0) **66** 102669. DOI: 10.1016/j.scs.2020.102669 28. 28The Humanitarian Data Exchange (HDX). Novel Coronavirus (COVID-19) Cases Data. https://data.humdata.org/dataset/novel-coronavirus-2019-ncov-cases.. (2019.0) 29. Wu Y., Yang Y., Nishiura H., Saitoh M.. **Deep learning for epidemiological predictions**. *In* (2018.0) 1085-1088 30. Alamsyah A., Prasetiyo B., Al Hakim M. F., Pradana F. D.. **Prediction of COVID-19 using recurrent neural network model**. (2021.0) **8** 98-103 31. Shastri S., Singh K., Kumar S., Kour P., Mansotra V.. **Time series forecasting of Covid-19 using deep learning models: India-USA comparative case study**. (2020.0) **140** 110227. DOI: 10.1016/j.chaos.2020.110227 32. Feroze N.. **Forecasting the patterns of COVID-19 and causal impacts of lockdown in top five affected countries using Bayesian Structural Time Series Models**. (2020.0) **140** 110196. DOI: 10.1016/j.chaos.2020.110196 33. Melin P., Monica J. C., Sanchez D., Castillo O.. **A new prediction approach of the COVID-19 virus pandemic behavior with a hybrid ensemble modular nonlinear autoregressive neural network**. (2020.0) 1-10. DOI: 10.1007/s00500-020-05452-z
--- title: The impact of a standardized Enhanced Recovery After Surgery (ERAS) protocol in patients undergoing minimally invasive heart valve surgery authors: - Alexander Gebauer - Johanna Konertz - Johannes Petersen - Jens Brickwedel - Denise Köster - Leonie Schulte-Uentrop - Hermann Reichenspurner - Evaldas Girdauskas journal: PLOS ONE year: 2023 pmcid: PMC10038258 doi: 10.1371/journal.pone.0283652 license: CC BY 4.0 --- # The impact of a standardized Enhanced Recovery After Surgery (ERAS) protocol in patients undergoing minimally invasive heart valve surgery ## Abstract ### Background An enhanced recovery after surgery (ERAS) protocol is a multimodal and multi-professional strategy aiming to accelerate postoperative convalescence. Pre-, intra- and postoperative measures might furthermore reduce postoperative complications and hospital length of stay (LOS) in a cost-effective way. We hypothesized that our unique ERAS protocol leads to shorter stays on the intensive care unit (ICU) and a quicker discharge without compromising patient safety. ### Methods This retrospective single center cohort study compares data of $$n = 101$$ patients undergoing minimally invasive heart valve surgery receiving a comprehensive ERAS protocol and $$n = 111$$ patients receiving routine care. Hierarchically ordered primary endpoints are postoperative hospital length of stay (LOS), postoperative complications and ICU LOS. ### Results Patients risk profiles and disease characteristics were comparably similar. Age was relevantly different between the groups (56 [17] vs. 57.5 [13] years, $$p \leq 0.015$$) and therefore adjusted. Postoperative LOS was significantly lower in ERAS group (6 [2] days vs. 7 [1] days, $p \leq 0.01$). No significant differences, neither in intra- or postoperative complications, nor in the number of readmissions ($15.8\%$ vs. $9.9\%$, $$p \leq 0.196$$) were shown. In hospital LOS (7 [3] days vs. 8 [4] days, $p \leq 0.01$) and ICU LOS (18.5 [6] hours vs. 26.5 [29] hours, $p \leq 0.01$) a considerable difference was shown. ### Conclusion The ERAS protocol for minimally invasive heart valve surgery is safe and feasible in an elective setting and leads to a quicker hospital discharge without compromising patient safety. However, further investigation in a randomized setting is needed. ## Introduction Enhanced recovery after surgery (ERAS) protocols are multimodal and multi professional strategies in perioperative care, aiming to reduce hospital length of stay (LOS) and healthcare-associated complications by attenuating physiological and psychological stress responses [1–3]. ERAS was initially established in 1990 by danish surgeon Prof. Dr. H. Kehlet and was based on the assumption, that trauma-induced stress responses stimulate endocrine and metabolic changes, which play a crucial role in the appearance of postoperative morbidity and prolonged hospital stays [3]. Early key elements included the utilization of minimal invasive surgery to reduce trauma, early mobilization after surgery and a new nutritional intake regimen [3]. With growing insights into the development of multilayered stress responses, these protocols evolved to cover up the entire patient journey, starting from preoperative patient education up to the point of early ambulation and follow-up care [4]. This is why a successful implementation requires the interdisciplinary cooperation of surgeons, anesthetists, physiotherapists and nurses [5]. Thus, physiologic and psychologic stress responses are supposed to be reduced, resulting in an overall enhancement of recovery [6]. ERAS protocols proved to be efficient most notably in colorectal surgery [7, 8], demonstrating major advantages in healthcare-associated infections, hospital LOS, gastrointestinal morbidity [9, 10], and cost-effectiveness [11]. Subsequently, an employment in different surgical settings was suggested [12]. These implications led to a paradigm shift in perioperative care, as ERAS Society was established in 2010 and since then developed certified recommendations for numerous surgical disciplines. For the matter of cardiac surgery (CS), patients offer a wide variety of complex pathologies and cardiac disease typically coexists with multiple comorbidities. In addition, the use of cardiopulmonary bypass (CPB) can trigger systemic inflammatory response syndromes (SIRS) [13], which adds to the challenge of designing a protocol that addresses the effects on multiple organ systems and their impact on the postoperative recovery phase [14]. This is why ERAS Cardiac Society, a subgroup within ERAS Society, developed evidence based expert consensus recommendations in the spirit of basic ERAS principles through a systematic literature review process. This scalable guideline manuscript provides a summarization of key elements for deployment in CS, marked with class of evidence and level of recommendation [15]. Nevertheless, the majority of protocols in CS suggests feasibility of applying certain elements of ERAS [16, 17], but hardly any cover up the entire clinical process provided by ERAS and high quality data is missing. The program designed at the University Heart and Vascular Centre Hamburg (UHZ) seeks to serve the holistic approach of an integrative ERAS protocol and to contribute to the refinement of ERAS in CS. In Addition, unlike many other ERAS programs that are designed to safely accelerate recovery for elderly and vulnerable patients in particular, this protocol is focused on patients with low risk profiles and stable clinical conditions, in order to retain ICU capacities for complex surgeries and critically ill patients. ERAS at UHZ started in February 2018 in selected patients undergoing minimally invasive heart valve surgery. We hypothesized, that the protocol leads to a reduction of hospital LOS without compromising patient safety and without adversely affecting the clinical results. ## Study design and ethical approval In this retrospective single-center cohort analysis performed at the University Heart and Vascular Center Hamburg, Germany, data of 101 consecutive patients undergoing elective minimally invasive heart valve surgery led by ERAS protocol and 111 patients receiving routine care was analyzed. It is in full accordance with the Declaration of Helsinki, released 2008 and was approved by our local Ethics Committee, Ethikkommission Hamburg, PV7050. There was no patients or public involvement in the development of this study. All patients included in this study gave their verbal consent for anonymized retrospective data analysis during the admission interview. For this retrospective study, the Ethics Committee did not require written informed consent. However, every patient admitted to the UHZ, is systematically asked for their consent to use their anonymized health- and treatment-related data in ongoing research projects [18]. From February 2018 to September 2020, 101 eligible patients matching the inclusion criteria were screened from the pool of all scheduled minimally invasive heart valve surgeries. They were contacted by phone and asked to attend a preoperative consultation 2–3 weeks prior to their surgery. Inclusion criteria was minimally invasive aortic valve or mitral valve surgery as well as age <75 years, sufficient physical condition and willingness to participate in the ERAS program, including preoperative preconditioning. Mitral valve surgery included non-rib-spreading fully endoscopic 3D (Aesculap Einstein Vision, Tuttlingen, Germany) mitral valve repair (MVR) or replacement with or without concomitant tricuspid valve repair, left atrial ablation and left atrial appendage closure [19]. Procedures on the aortic valve included reconstructive techniques, such as bicuspid aortic valve repair and David- or Yacoub procedure [20], as well as aortic valve replacement including simultaneous supracoronary ascending aorta replacement or Bentall procedure [21]. Cardiac tumors, such as fibroelastomas or myxoma, that were accessible to minimally invasive surgery were also included. Exclusion criteria was age >75 years, the need for complete median sternotomy (i.e. due to prior chest radiation, concomitant coronary artery disease requiring bypass surgery or re-operations after prior heart surgery), unwillingness to participate in ERAS program and severe comorbidities or conditions that increase the risk for peri- or postoperative complications or made them unsuitable for ERAS protocol. Furthermore, preexisting risk scores and prediction models were utilized, that identified prior heart surgery, extracardiac arteriopathy, obesity, elevated serum creatinine >150 μmol/L, nonelective and complex surgery as independent risk factors for failure of enhanced recovery [22, 23]. The control group contained 111 patients with the same inclusion and exclusion criteria, who either had surgery in a previous time period before the implementation of the ERAS protocol or did not want to participate. The three hierarchically ordered primary endpoints are postoperative hospital length of stay (LOS), postoperative complications (beside nosocomial infections) and ICU LOS. The secondary endpoints are duration of the heart lung machine, aortic cross clamp time (XCT), number of transfusions, nosocomial infections, delirium appearance of postoperative atrial fibrillation, re-operation and readmissions. ## ERAS program This ERAS program was developed according to the Guidelines for Perioperative Care in Cardiac Surgery Enhanced Recovery After Surgery Society Recommendations as previously described in detail from our group [15, 18]. Core elements of contained a dedicated prehabilitation program with representatives of every involved profession one to three weeks before surgery. Present physical condition and frailty was assessed and patients were asked to perform daily exercising activities to sustain or improve physical capacity. A detailed patient education did not only focus on the forensic aspects of possible complications but served as psychologic preparation to improve mental readiness for surgery. An intensified physiotherapy regimen started three hours after surgery and contained at least two daily units. As limited rehab capacities might delay a timely discharge, rehab slots were organized during the first interview two weeks before surgery. Thus, a direct transfer to rehab was routinely performed in the majority of the patients unless wished otherwise. ## Patient characteristics A detailed overview of demographic and clinical baseline characteristics of all enrolled patients ($$n = 212$$, $73\%$ men, 54±11 years) is displayed in Table 1. Patients were predominantly men ($73.3\%$ in ERAS vs. $71.2\%$ in control-group, $$p \leq 0.734$$) without relevant differences in disease characteristics between the two groups. Perioperative risk was low in both groups, with a tendency to a lower risk score in ERAS group (EuroSCORE II 0.67 (0.28) vs. 0.73 (0.38) in control group, $$p \leq 0.06$$). The ERAS group tends to be younger compared to the control group (56 [17] vs. 57.5 [13] years, $$p \leq 0.015$$). The primary analysis is adjusted for age and EuroSCORE II due to differences between treatment groups. There was no missing data for variables of interest. **Table 1** | Baseline characteristic | ERAS group | Control group | p-value | | --- | --- | --- | --- | | Baseline characteristic | n = 101 | n = 111 | p-value | | Age (years) | 56 (17) | 57.5 (13) | 0.015 | | Gender | | | | | • male, n (%) | 74 (73.3%) | 79 (71.2%) | 0.734 | | • female, n (%) | 27 (26.7%) | 32 (28.8%) | | | LV-function, % | 58±8 | 59±8 | 0.501 | | BMI | 25.7±3.4 | 26.2±3.3 | 0.271 | | Creatinine (mg/dl) | 0.9 (0.2) | 0.94 (0.2) | 0.166 | | Atrial fibrillation, n (%) | 12 (11.9%) | 16 (14.4%) | 0.586 | | Diabetes mellitus Type II, n (%) | 2 (2.0%) | 7 (6.3%) | 0.119 | | Hyperlipidemia, n (%) | 20 (19.8%) | 24 (21.6%) | 0.744 | | Hypertension, n (%) | 46 (45.5%) | 60 (54.1%) | 0.216 | | Coronary artery disease, n (%) | 5 (5.0%) | 11 (9.9%) | 0.172 | | Chronic lung disease, n (%) | 9 (8.9%) | 13 (11.7%) | 0.504 | | Prior stroke, n (%) | 1 (1.0%) | 3 (2.7%) | 0.360 | | EuroSCORE II (%) | 0.67 (0.28) | 0.73 (0.38) | 0.06 | ## Statistical analysis Statistical analysis was performed in collaboration with the Institute of Medical Biometry and Epidemiology (IMBE) of the University Medical Center Hamburg Eppendorf using SPSS Version 27.0 (IBM Corp, New York, USA). Categorical variables are given as percentages and absolute or relative numbers. Continuous variables are presented as mean ± standard deviation or median (IQR). If histograms showed a curve of normal distribution for continuous variables, data was compared using the unpaired two-sided t-test. Data without normal distribution was analyzed using Mann-Whitney U-Test. Categorical variables were analyzed using the chi-square test or the Fisher’s exact test, as appropriate. All p-values regarding the baseline variables are considered descriptively. The three hierarchical primary endpoints were analyzed using a multivariable linear regression or a logistic regression with adjustment for the clinically relevant baseline variables (age, aortic cross-clamp time (XCT)). If the previous endpoint is significant between the treatment groups, the following endpoint is evaluated confirmatory. Given the hierarchy of the endpoints, an adjustment of the significance level is not necessary. A p-value of <0.05 was considered statistically significant. The secondary endpoints were analyzed descriptively as mentioned above. ## Procedural characteristics Receiving mitral valve or tricuspid valve surgery, atrial septal defect (ASD) closure, cryoablation or left atrial appendage (LAA) occlusion, $48.5\%$ of patients in ERAS-group underwent right lateral mini-thoracotomy vs. $55.9\%$ of patients in control group. $51.5\%$ of patients in ERAS group had partial upper sternotomy vs. $44.1\%$ in control group ($$p \leq 0.285$$), receiving aortic valve repair/replacement, aortic root or ascending aortic replacement. There is a relevant difference aortic cross clamp time between the treatment groups (77±28min vs. 107±41min, $p \leq 0.01$). Therefore, the primary analysis is adjusted accordingly. An overview of performed procedures and intraoperative characteristics is shown in Table 2. **Table 2** | Surgical variables | ERAS group | Control group | p-value | | --- | --- | --- | --- | | Surgical variables | n = 101 | n = 111 | p-value | | Mitral valve surgery | 49 (48.5%) | 62 (55.9%) | 0.285 | | • mitral valve replacement | 0 | 3 (2.7%) | | | • mitral valve repair | 48 (47.5%) | 59 (53.2%) | | | Aortic valve surgery | 51 (51.5%) | 49 (44.1%) | 0.285 | | • aortic valve replacement | 22 (21.8%) | 26 (23.4%) | | | • aortic valve repair | 26 (25.7%) | 20 (18%) | 0.173 | | • concomitant aortic root surgery | 4 (4%) | 7 (6.3%) | | | • ascending aorta replacement | 1 (1%) | 0 | | | • Bentall procedure | 0 | 2 (1.8%) | | | • fibroelastoma removal | 4 (4%) | 0 | | | CPB time (min) | 130.5 (61) | 147 (81) | 0.076 | | XCT (min) | 77±28 | 107±41 | <0.01 | ## Primary endpoints The first primary endpoint (postoperative LOS) was statistically significant ($p \leq 0.01$). Therefore, the second primary endpoint (postoperative complications) was confirmatory evaluated without showing a statistically significance. Hence, the last primary endpoint (ICU LOS) could only be evaluated descriptively. Here, a relevant difference between the treatment groups was shown. ## Clinical outcome No intra- or perioperative complications were associated with ERAS-protocol. There was no in-hospital or 30-day-mortality in both groups. Transfusion was necessary in $11.9\%$ of patients in ERAS group vs. $18.9\%$ in control group ($$p \leq 0.158$$). There was no relevant difference in the occurrence of nosocomial infections ($12.9\%$ vs. $15.3\%$, $$p \leq 0.61$$), or other postoperative complications, which are summarized in detail in Table 3 ($13.9\%$ vs. $18\%$, $$p \leq 0.41$$). Reintubation was necessary in 3 patients that underwent ERAS-protocol vs. 4 patients in control group ($$p \leq 0.797$$). $24.8\%$ of patients in ERAS group developed postoperative atrial fibrillation (poAF) compared to $15.3\%$ in control group ($$p \leq 0.09$$). There was no difference in the number of readmissions to ICU ($4\%$ vs. $2.7\%$, $$p \leq 0.609$$). **Table 3** | Postoperative data | ERAS group | Control group | p-value | | --- | --- | --- | --- | | Postoperative data | n = 101 | n = 111 | p-value | | Hospital LOS, days | 7 (3) | 8 (4) | <0.01 | | ICU LOS, hours | 18.5 (6) | 26.5 (29) | <0.01 | | Postoperative LOS, days | 6 (2) | 7 (1) | <0.01 | | ICU readmission, n (%) | 4 (4%) | 3 (2.7%) | 0.609 | | Reintubation necessary, n (%) | 3 (3%) | 4 (3.6%) | 0.797 | | Transfusion necessary, n (%) | 12 (11.9%) | 21 (18.9%) | 0.158 | | Redo surgery | 7 (6.9%) | 9 (8.1%) | 0.746 | | • Valve related | 4 (4%) | 1 (0.9%) | 0.143 | | • Bleeding | 3 (3%) | 8 (7.2%) | 0.165 | | • Pericardial tamponade | 0 | 2 (1.8%) | 0.175 | | Pacemaker, n (%) | 3 (3%) | 4 (3.6%) | 0.797 | | Nosocomial Infections, n (%) | 13 (12.9%) | 17 (15.3%) | 0.61 | | • Pneumonia | 3 (3%) | 9 (8.1%) | 0.106 | | Postoperative LV-function, % | 52±8 | 52±9 | 0.65 | | Postoperative complications, n (%) | 14 (13.9%) | 20 (18%) | 0.41 | | • Delirium, n (%) | 5 (5%) | 7 (6.3%) | 0.67 | | • AV-Block, n (%) | 4 (4%) | 6 (5.4%) | 0.62 | | • Acute coronary syndrome, n (%) | 2 (2.0%) | 1 (0.9%) | 0.506 | | • Stroke, n (%) | 2 (2.0%) | 2 (1.8%) | 0.924 | | • poAF, n (%) | 25 (24.8%) | 17 (15.3%) | 0.085 | | Readmission from rehab, n (%) | 16 (15.8%) | 11 (9.9%) | 0.196 | | • Pleura effusion | 4 (4%) | 3 (2.7%) | | | • Pericardial effusion | 5 (5%) | 2 (1.8%) | | | • Dyspnoe | 1 (1%) | 0 | | | • Valve dysfunction | 1 (1%) | 0 | | | • Organic psychosyndrome | 2 (2%) | 0 | | | • Atrial fibrillation | 2 (2%) | 1 (0.9%) | | | • Wound infection | 0 | 1 (0.9%) | | | • Other Reasons | 1 (1%) | 4 (3.6%) | | | • Intervention necessary | 8 (7.9%) | 4 (3.6%) | | Surgical revision for bleeding, valve malfunctioning or pericardial tamponade that was necessary immediately or in the course of the hospital stay is summarized under redo surgery and was necessary in $8.9\%$ in ERAS group vs. $11.7\%$ in control group ($$p \leq 0.504$$). $4\%$ in ERAS group were valve related redo surgeries vs. $0.9\%$ in control group ($$p \leq 0.143$$). Surgical revision because of bleeding occurred in $3\%$ in ERAS group vs. $7.2\%$ in control group ($$p \leq 0.165$$). Permanent pacemaker implantations were also comparably similar ($3\%$ vs. $3.6\%$, $$p \leq 0.797$$). There was no difference in the number of readmissions from rehab after 30 days ($15.9\%$ vs. $9.9\%$, $$p \leq 0.196$$). In ERAS group, a relevant decrease in hospital LOS (7 [3] days vs. 8 [4] days, $p \leq 0.01$) as well as a significant decrease in postoperative LOS (6 [2] days vs. 7 [1] days, $p \leq 0.01$) was shown. ICU LOS was significantly shorter (18.5 [6] hours vs. 26.5 [29] hours in control group, $p \leq 0.01$). ## Patient selection and education Since a small proportion of approximately 10–$15\%$ of patients develop 80–$90\%$ of complications, perioperative care is suggested to be further individualized [24]. In this program, a dedicated patient selection with exclusion of complex surgery and elderly patients >75 years, utilization of EuroSCOREII to predict in-hospital mortality, LOS and specific postoperative complications [25] reliably ensured hemodynamic stability after surgery and made way for uncomplicated immediate extubation in the OR. The physiotherapeutic assessment three weeks prior to surgery is meant to detect and exclude frail patients and ascertain suitability for our demanding protocol. At the same time, it gives motivational benchmarks that patients can try to get back to in the course of postoperative convalescence. Furthermore, daily exercising activities before surgery demonstrated a decline in sympathetic over-reactivity and an improved insulin sensitivity [26, 27]. Additionally, being familiar with the execution of some exercises prior to surgery is suggested to help overcome postoperative phlegm and reluctance. A number of randomized controlled trials was able to demonstrate that such prehabilitation programs result in an improved physical and mental readiness for surgery, a reduction in ICU and hospital LOS and improved transition from hospital to the community [28–30]. Nevertheless, open heart surgery remains to be a demanding turning point in many patients lives. Even though some patients might feel noticeable restrictions in their everyday life caused by the disease, most of them will live a more or less independent and self-regulated life up to the point of hospitalization. Tubes, catheters and restrictions might limit mobility, regulations and schedules need to be followed, which marks a severe interference in a patient’s autonomy. In this ERAS protocol education and empowerment of patients exceeds the traditional form of information and consent, which is usually focused on the forensic aspects of possible complications. It is rather designed to obtain the patient as an active and relevant actor in their own process of healing, to improve mental and physical readiness for surgery and to help them regain their autonomy as soon as possible. Working off the imbalance of knowledge and creating an environment of shared decision making allows for reasonable expectations on the patient side and has shown to reduce fear and postoperative analgetic use [6, 31]. In turn, postoperative convalescence might be improved by a reduction of fear and analgetic use [32, 33]. ## Immediate extubation and intensive care management Safety of an early extubation and its positive impact on ICU LOS was demonstrated by numerous studies [34–36]. A decisive factor in the reduction of shorter ventilation times goes back to the establishment of short acting narcotics that allow for quicker extubation [36, 37]. Even an immediate extubation after off-pump coronary artery bypass surgery proved to be safe and feasible [38], although Montes et al. demonstrated that an extubation in the OR might be safe but has no effect on ICU or hospital LOS [39]. However, positive end-expiratory pressure during mechanical ventilation impedes RV ejection [40]. Considering every patient after heart surgery at risk for cardiac complications, we still identified minimalizing mechanical ventilation to be a core element of our ERAS protocol. Out of three patients who had to be reintubated, one patient had a stroke related seizure and two patients suffered from respiratory depression due to extensive pleural effusion days after surgery. Compared to four patients undergoing reintubation in the control group for similar reasons (two stroke related seizures and two respiratory depressions), we found the advantages to be preponderant, as we found significantly shorter lengths of stay and a numeric reduction of possibly ventilator-associated pneumonia in ERAS group ($3\%$ vs. $8.1\%$, $$p \leq 0.106$$). Intubation for surgical revisions, e.g. before rethoracotomy, was not included in the cases mentioned above. Furthermore, transferring already extubated patients to the ICU might prevent the staff from keeping patients asleep who are respiratory stable enough to be extubated because of a potentially stressful work environment and reduces nursing tasks. Although immediate extubation was not utilized in the following case, Ender et al. demonstrated that for CS patients undergoing their unique fast-track concept, a direct transfer from the OR to a specifically opened postanesthetic care unit (PACU) without an intermediary stay on the ICU was feasible without compromising patient safety [35]. It is indeed worth mentioning, that $14\%$ of these patients had to be transferred to ICU eventually. On the one hand, a reason might be a greater variety of complex surgery with patients undergoing multiple valve and combined procedures, and on the other hand there were limited opening hours on the PACU in Leipzig, necessitating an admission to ICU if patients were not hemodynamically stabile enough to be transferred to an intermediate care ward on 6:30pm on the day of surgery. At UHZ, a careful patient selection and the new establishment of an overnight 24 hours PACU should allow for a safe patient transfer directly to the low care ward, entirely skipping ICU or intermediate care ward. It cannot be ignored, that exclusion of elderly or high risk patients is an unconventional approach for ERAS protocols, that are typically focused on vulnerable patients in particular. However, we want to shine light on the growing importance of limited ICU resources, which is why it is designed to let low risk patients skip ICU entirely and thus retain ICU capacities for the critically ill. Due to the relatively new establishment of the PACU24, resources for overnight care are still in development. Therefore, most patients included in this study were directly transferred to ICU. Conducted cases, however, provide an indication that skipping ICU entirely will potentially find its way into this ERAS protocol. ## Aortic cross clamp time Aortic cross clamp time (XCT) is an independent predictor of mortality, morbidity and prolonged hospital LOS in CS patients [41, 42]. In this study, a relevant difference in XCT between patients undergoing ERAS protocol and control group was shown and is in need of explanation. Patients with low risk profiles undergoing isolated valve surgery may often be operated by aspiring young surgeons. After implementation of our ERAS protocol, there was much importance attached to the procedures being performed by highly trained surgeons with great expertise to guarantee the best possible outcome. This is why XCT in control group probably better reflects real-world experience in a university teaching hospital with many different levels of skills. However, in this study univariate regression models showed a minimal effect on hospital LOS and no relevant effect on ICU LOS or any major postoperative complication to be significantly associated with XCT. ## Clinical implications The success story of ERAS in a number of surgical disciplines made comprehensive protocols in CS gradually emerge, that, along with a reduction of hospital LOS, also indicated potential medical benefits. Results from Williams et al., who contributed to ERAS in CS with a retrospective comparison of 443 patients undergoing ERAS protocol and 489 historic patients in routine care, showed a reduction of gastrointestinal complications and increased staff and patient satisfaction [43], while a randomized trial of Li et al. could demonstrate a reduction of major postoperative complications such as acute renal failure, stroke or heart block [44]. A potential economic benefit of up to 1900€ per patient was demonstrated during the pilot phase of this study at the UHZ [45]. Taking these findings into account, our confidence is strengthened that this unique protocol will hopefully lead to improved clinical outcomes and contribute to an extension of ERAS programs in CS. At the same time, decreasing lengths on CPB and emergence of less invasive techniques might furthermore increase the number of patients being eligible for enhanced recovery after CS. Sutureless aortic valve protheses offer an alternative for multiple valve or high risk surgery to reduce CPB time [46]. However, higher rates of paravalvular leaks and permanent pacemaker implantations question possible advantages over stented protheses or transcatheter valves and do not suggest a routine use in low risk patients at the moment [47]. On the other hand, encouraging early data from a novel beating heart mitral valve repair system indicates a proceeding development of “interventional” CS, that might be contributing to quicker convalescence and greater ICU capacities [48]. Even though Engelmann et al. elaborated all the different aspects of ERAS in CS and presented an inviting piece for educational and planning purposes, it soon becomes clear, that implementation of such a protocol is a complex and demanding task that requires permanent self-evaluation and the willingness to break with long-established practices [49]. To address this, a weekly feedback round with representatives of every involved profession and daily ward rounds for all included patients with subsequent case discussions was organized. As recommended by Salenger et al., who published a guideline for successful implementation of ERAS [49], it is intended to facilitate the way for necessary changes into clinical practice. In surgical disciplines in particular, a traditionally conservative culture of holding on to reliable habits is understandable. Nevertheless, the future of enhanced recovery and its feasibility is based on clinicians who can ignite enthusiasm over a permanent vision of how to provide better care. New measures of care are necessary, that do not only strive towards satisfying surgical results and an early discharge, but towards a patient-centered individualization of health care. ## Limitations Validity of data in a retrospective study design is limited by nature. Instead of randomization, patient selection was performed and differences in baseline variables occurred, which were adjusted using a multivariate regression model. However, results of the regression model demonstrated minimal influence of addressed variables for all of the primary and secondary endpoints. Indeed, the data collected in this study made way for the INCREASE-study (Interdisciplinary Perioperative Care in Minimally-invasive Heart Valve Surgery, NCT04977362), which is a randomized clinical trial that started in June 2021 and is expected to provide high quality data about the organization and execution of our ERAS protocol in the minimally invasive treatment of heart valve pathologies and their potential transfer into standard-of-care treatment. ## Conclusion The main finding of this study is that the ERAS protocol for minimally invasive heart valve surgery is safe and feasible in selected patients and an elective setting. Clinical outcomes demonstrated a non-inferiority compared to routine care while hospital LOS was significantly shortened and ICU LOS was relevantly reduced, indicating that this protocol might be transferred into standard of care treatment. If these effects persist in a randomized controlled trial, needs to be explored. ## References 1. Firoozfard B., Christensen T., Kristensen J. K., Mogensen S., Kehlet H.. **Fast-track open transperitoneal nephrectomy**. (2003) **37** 305-8. DOI: 10.1080/00365590310014742 2. Kehlet H.. **Organizing postoperative accelerated recovery programs**. (1996) **21** 149-51. PMID: 8956444 3. Kehlet H.. **Multimodal approach to control postoperative pathophysiology and rehabilitation**. (1997) **78** 606-17. DOI: 10.1093/bja/78.5.606 4. Ljungqvist O., Hubner M.. **Enhanced recovery after surgery-ERAS-principles, practice and feasibility in the elderly," (in eng)**. (2018) **30** 249-252. DOI: 10.1007/s40520-018-0905-1 5. Fearon K. C.. **Enhanced recovery after surgery: a consensus review of clinical care for patients undergoing colonic resection**. (2005) **24** 466-77. DOI: 10.1016/j.clnu.2005.02.002 6. Kehlet H., Wilmore D. W.. **Fast-track surgery**. (2005) **92** 3-4. DOI: 10.1002/bjs.4841 7. Eskicioglu C., Forbes S. S., Aarts M. A., Okrainec A., McLeod R. S.. **"Enhanced recovery after surgery (ERAS) programs for patients having colorectal surgery: a meta-analysis of randomized trials,"**. (2009) **13** 2321-9. DOI: 10.1007/s11605-009-0927-2 8. Feldheiser A.. **"Enhanced Recovery After Surgery (ERAS) for gastrointestinal surgery, part 2: consensus statement for anaesthesia practice,"**. (2016) **60** 289-334. DOI: 10.1111/aas.12651 9. Barbieux J.. **Does enhanced recovery reduce postoperative ileus after colorectal surgery?**. (2017) **154** 79-85. DOI: 10.1016/j.jviscsurg.2016.08.003 10. Grant M. C., Yang D., Wu C. L., Makary M. A., Wick E. C.. **Impact of Enhanced Recovery After Surgery and Fast Track Surgery Pathways on Healthcare-associated Infections: Results From a Systematic Review and Meta-analysis**. (2017) **265** 68-79. DOI: 10.1097/SLA.0000000000001703 11. Stone A. B.. **Implementation Costs of an Enhanced Recovery After Surgery Program in the United States: A Financial Model and Sensitivity Analysis Based on Experiences at a Quaternary Academic Medical Center**. (2016) **222** 219-25. DOI: 10.1016/j.jamcollsurg.2015.11.021 12. Langelotz C., Spies C., Muller J. M., Schwenk W.. **Fast-track"-rehabilitation in surgery, a multimodal concept**. (2005) **105** 555-9. DOI: 10.1080/00015458.2005.11679780 13. Hess P. J.. **Systemic inflammatory response to coronary artery bypass graft surgery**. (2005) **62** S6-9. DOI: 10.2146/ajhp050302 14. Ljungqvist O., Francis K. K., Urman R.. (2020) XXIII-654 15. Engelman D. T.. **Guidelines for Perioperative Care in Cardiac Surgery: Enhanced Recovery After Surgery Society Recommendations**. (2019) **154** 755-766. DOI: 10.1001/jamasurg.2019.1153 16. Fleming I. O.. **Aggregation of Marginal Gains in Cardiac Surgery: Feasibility of a Perioperative Care Bundle for Enhanced Recovery in Cardiac Surgical Patients**. (2016) **30** 665-70. DOI: 10.1053/j.jvca.2016.01.017 17. Toraman F., Evrenkaya S., Yuce M., Goksel O., Karabulut H., Alhan C.. **Fast-track recovery in noncoronary cardiac surgery patients**. (2005) **8** E61-4. DOI: 10.1532/HSF98.20041138 18. Kubitz J. C.. **Establishment of an enhanced recovery after surgery protocol in minimally invasive heart valve surgery**. (2020) **15** e0231378. DOI: 10.1371/journal.pone.0231378 19. Westhofen S.. **A matched pairs analysis of non-rib-spreading, fully endoscopic, mini-incision technique versus conventional mini-thoracotomy for mitral valve repair**. (2016) **50** 1181-1187. DOI: 10.1093/ejcts/ezw184 20. Bechtel J. F., Erasmi A. W., Misfeld M., Sievers H. H.. **[Reconstructive surgery of the aortic valve: the Ross, David, and Yacoub procedures]**. (2006) **31** 413-22. DOI: 10.1007/s00059-006-2836-4 21. Bentall H., De Bono A.. **A technique for complete replacement of the ascending aorta**. (1968) **23** 338-9. DOI: 10.1136/thx.23.4.338 22. Constantinides V. A.. **Fast-track failure after cardiac surgery: development of a prediction model**. (2006) **34** 2875-82. DOI: 10.1097/01.CCM.0000248724.02907.1B 23. Parlow J. L., Ahn R., Milne B.. **Obesity is a risk factor for failure of "fast track" extubation following coronary artery bypass surgery**. (2006) **53** 288-94. DOI: 10.1007/BF03022217 24. Boehm O., Baumgarten G., Hoeft A.. **Preoperative patient assessment: Identifying patients at high risk," (in eng)**. (2016) **30** 131-43. DOI: 10.1016/j.bpa.2016.04.005 25. Toumpoulis I. K., Anagnostopoulos C. E., Swistel D. G., DeRose J. J.. **Does EuroSCORE predict length of stay and specific postoperative complications after cardiac surgery?," (in eng)**. (2005) **27** 128-33. DOI: 10.1016/j.ejcts.2004.09.020 26. Snowden C. P.. **Cardiorespiratory fitness predicts mortality and hospital length of stay after major elective surgery in older people**. (2013) **257** 999-1004. DOI: 10.1097/SLA.0b013e31828dbac2 27. Valkenet K., van de Port I. G., Dronkers J. J., de Vries W. R., Lindeman E., Backx F. J.. **The effects of preoperative exercise therapy on postoperative outcome: a systematic review**. (2011) **25** 99-111. DOI: 10.1177/0269215510380830 28. Arthur H. M., Daniels C., McKelvie R., Hirsh J., Rush B.. **Effect of a preoperative intervention on preoperative and postoperative outcomes in low-risk patients awaiting elective coronary artery bypass graft surgery. A randomized, controlled trial**. (2000) **133** 253-62. DOI: 10.7326/0003-4819-133-4-200008150-00007 29. Sawatzky J. A.. **Prehabilitation program for elective coronary artery bypass graft surgery patients: a pilot randomized controlled study**. (2014) **28** 648-57. DOI: 10.1177/0269215513516475 30. Waite I., Deshpande R., Baghai M., Massey T., Wendler O., Greenwood S.. **Home-based preoperative rehabilitation (prehab) to improve physical function and reduce hospital length of stay for frail patients undergoing coronary artery bypass graft and valve surgery**. (2017) **12** 91. DOI: 10.1186/s13019-017-0655-8 31. Egbert L. D., Battit G. E., Welch C. E., Bartlett M. K.. **Reduction of Postoperative Pain by Encouragement and Instruction of Patients. A Study of Doctor-Patient Rapport**. (1964) **270** 825-7. DOI: 10.1056/NEJM196404162701606 32. Kehlet H., Holte K.. **Effect of postoperative analgesia on surgical outcome**. (2001) **87** 62-72. DOI: 10.1093/bja/87.1.62 33. Johnston M., Vögele C.. **Benefits of psychological preparation for surgery: a meta-analysis**. *Annals of Behavioral Medicine* (1993) 245-256 34. Engelman R. M.. **Mechanisms to reduce hospital stays," (in eng)**. (1996) **61** S26-9. DOI: 10.1016/0003-4975(95)01081-5 35. Ender J.. **Cardiac surgery fast-track treatment in a postanesthetic care unit: six-month results of the Leipzig fast-track concept**. (2008) **109** 61-6. DOI: 10.1097/ALN.0b013e31817881b3 36. Wong W. T., Lai V. K., Chee Y. E., Lee A.. **Fast-track cardiac care for adult cardiac surgical patients**. (2016) **9**. DOI: 10.1002/14651858.CD003587.pub3 37. Cheng D. C.. **Pro: early extubation after cardiac surgery decreases intensive care unit stay and cost," (in eng)**. (1995) **9** 460-4. DOI: 10.1016/s1053-0770(05)80105-3 38. Horswell J. L., Herbert M. A., Prince S. L., Mack M. J.. **Routine immediate extubation after off-pump coronary artery bypass surgery: 514 consecutive patients," (in eng)**. (2005) **19** 282-7. DOI: 10.1053/j.jvca.2005.03.002 39. Montes F. R.. **The lack of benefit of tracheal extubation in the operating room after coronary artery bypass surgery," (in eng)**. (2000) **91** 776-80. DOI: 10.1097/00000539-200010000-00003 40. Mahmood S. S., Pinsky M. R.. **Heart-lung interactions during mechanical ventilation: the basics," (in eng)**. (2018) **6** 349. DOI: 10.21037/atm.2018.04.29 41. Al-Sarraf N.. **Cross-clamp time is an independent predictor of mortality and morbidity in low- and high-risk cardiac patients," (in eng)**. (2011) **9** 104-9. DOI: 10.1016/j.ijsu.2010.10.007 42. Iino K.. **Prolonged Cross-Clamping During Aortic Valve Replacement Is an Independent Predictor of Postoperative Morbidity and Mortality: Analysis of the Japan Cardiovascular Surgery Database**. (2017) **103** 602-609. DOI: 10.1016/j.athoracsur.2016.06.060 43. Williams J. B.. **One-year results from the first US-based enhanced recovery after cardiac surgery (ERAS Cardiac) program**. (2019) **157** 1881-1888. DOI: 10.1016/j.jtcvs.2018.10.164 44. Li M.. **Enhanced recovery after surgery pathway for patients undergoing cardiac surgery: a randomized clinical trial**. (2018) **54** 491-497. DOI: 10.1093/ejcts/ezy100 45. Petersen J.. **Economic impact of enhanced recovery after surgery protocol in minimally invasive cardiac surgery," (in eng)**. (2021) **21** 254. DOI: 10.1186/s12913-021-06218-5 46. Sadowski J., Kapelak B., Pfitzner R., Bartuś K.. **Sutureless aortic valve bioprothesis ’3F/ATS Enable’—4.5 years of a single-centre experience," (in eng)**. (2009) **67** 956-63. PMID: 19784899 47. Lam K. Y.. **Self-deployment aortic valve prostheses: sutureless is never futureless," (in eng)**. (2022) **62**. DOI: 10.1093/ejcts/ezac383 48. Gammie J. S.. **Safety and performance of a novel transventricular beating heart mitral valve repair system: 1-year outcomes**. (2020) **59** 199-206. DOI: 10.1093/ejcts/ezaa256 49. Salenger R., Morton-Bailey V., Grant M., Gregory A., Williams J. B., Engelman D. T.. **Cardiac Enhanced Recovery After Surgery: A Guide to Team Building and Successful Implementation**. (2020) **32** 187-196. DOI: 10.1053/j.semtcvs.2020.02.029
--- title: Analysis of disease burden due to high body mass index in childhood asthma in China and the USA based on the Global Burden of Disease Study 2019 authors: - Chengyue Zhang - Qing Qu - Kaiyu Pan journal: PLOS ONE year: 2023 pmcid: PMC10038260 doi: 10.1371/journal.pone.0283624 license: CC BY 4.0 --- # Analysis of disease burden due to high body mass index in childhood asthma in China and the USA based on the Global Burden of Disease Study 2019 ## Abstract ### Background Currently, there is a growing concern about the disease burden of child asthma particularly due to high body mass index (BMI). The prevalence and disease burden of asthma differ between developing and developed countries, with implications on disease intervention. Therefore, we provide a comparative analysis of childhood asthma between China and the United States of America (USA). ### Methods Using the Global Burden of Disease (GBD) 2019 data, we estimated and compared the age-standardized prevalence, disability-adjusted life years (DALYs), years of life lost (YLLs), years of lost due to disability (YLDs), DALYs due to high BMI of asthma in children aged 1–14 years in China and the USA. Joinpoint regression analysis was applied to assess changes in temporal trends. ### Results DALYs due to high BMI and the ratio of DALYs to DALYs due to high BMI in children with asthma showed a significant upward trend in both countries and were higher in males than in females. Almost all epidemiological indicators of asthma showed a hump of curve from 2014 to 2019, and peaked in 2017. There was a decreasing trend of YLLs for asthma in children both countries, while China has a saliently greater decreasing trend. ### Conclusion The disease burden caused by high BMI of childhood asthma was on the rise in children with asthma in both China and the USA. High BMI needs to be taken more into account in the development of future policies for the prevention, control, and treatment of childhood asthma. However, the increasing trend of this disease burden in American children was significantly lower than that in Chinese children. We recommend learning from the American government to impose a high-calorie tax, increase physical exercise facilities, and provide better health care policies. ## Introduction Asthma is one of the most common chronic diseases in children, with wheezing, coughing, and airflow restriction as clinical manifestations, affecting children’s daily life [1]. The incidence, prevalence, and medical costs of this disease have been increasing in recent years [2,3]. A survey revealed that the prevalence of asthma in Chinese children increased from $0.91\%$ to $2.12\%$ between 1990 and 2010 [4]. Respiratory health during early life may have a lifelong impact on lung health and life expectancy; thus, prevention and control of childhood asthma is particularly crucial to promote individual health and reduce the societal burden of the disease [5]. However, the etiology of childhood asthma is yet to be elucidated. Therefore, identifying its risk factors and exploring possible mechanisms is necessary for early detection and intervention to prevent further adverse outcomes [6]. Currently, reported risk factors for asthma include genetic factors, tobacco exposure, dampness/humidity, animal contact, climate, and inhalation of small particles [6–8]. High body mass index (BMI), which is considered as the seventh-leading level 2 risk factor for attributable disability-adjusted life years (DALYs) of diseases in 2019, is also a risk factor for asthma [9]. It is thought to be associated with dietary habits, lifestyle, and food intake [10]. There are differences in the prevalence and disease burden of asthma between developing and developed countries [3]. The direct and indirect economic costs of childhood asthma are high, and there is a link between the disease burden of asthma and the economic level of the country [11]. It is well known that *China is* the largest developing country in the world and that the United States of America (USA) is the major developed country [12,13]. However, to the best of our knowledge, there has been no comparative analysis between China and the USA in these areas. Thus, this study aimed to investigate the prevalence of asthma, DALYs, and the effect of the risk factor high BMI on disease burden in children aged 1–14 years in China and the USA, to compare and analyze the differences between them, to provide information for resource allocation, and to learn from the prevention and control strategies of developed countries such as the USA, which can provide some prevention and control strategies to reduce the disease burden of childhood asthma in developing countries such as China. ## Data source The Global Burden of Disease (GBD) 2019 is a cross-border collaborative project covering 204 countries and regions. It collected data from disease surveillance sites, surveys of the National Health Service, and published literature data to estimate descriptive epidemiological information on the incidence, prevalence, disability-adjusted life years (DALYs), years of lost due to disability (YLDs), and years of life lost (YLLs) for 369 stratified diseases and injuries using the DisMod-MR 2.1 as a Bayesian meta-regression model [14,15]. The GBD estimation process uses 86,249 sources that are broad and representative, including censuses, household surveys, health service use, civil registration and vital statistics, air pollution testing, etc. The data is publicly available (http://ghdx.healthdata.org/gbd-results-tool). DALYs are a summary metric of YLDs calculated by multiplying the prevalence of individual sequelae by the disability weights, and YLLs that are the actual loss of life that occurs from death at each age modified by parameters such as the standard life expectancy at the corresponding age [16]. As a non-negligible risk factor for asthma, high BMI is defined as being overweight or obese by GBD 2019 for children aged 1–19 years according to International Obesity Task Force standards. In this study, we obtained data on the prevalence, DALYs, YLDs, YLLs, and DALYs due to high BMI of childhood asthma in children aged 1–14 years in China and the USA from GBD 2019. Furthermore, to evaluate the role played by high BMI in disease burden of childhood asthma, we calculated the ratio of DALYs to DALYs due to high BMI and plotted its temporal trend. To study the changes in each epidemiological indicator at different ages, we analyzed children aged 1–4 years, 5–9 years, and 10–14 years separately. All epidemiological data obtained were age-standardized to match the characteristics of the different national reference populations and finally expressed in terms of 100,000 population [14]. This study was conducted in compliance with the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) [17]. No ethical review board approval was required for this study. ## Statistical analysis Joinpoint Regression Program version 4.9.0.0 (National Cancer Institute, Rockville, MD, USA) was used to analyze the changes in trends in specific disease burden indicators from 1990 to 2019. The ordinary least squares method was used to fit the regression model under the assumption that the error random variables were homoskedastic. To increase the credibility of the results, we set the maximum number of joinpoints to 3 and determined the locations of the joinpoints and the corresponding p-values by Monte Carlo permutation test. Log-linear regression was used to calculate their annual percentage change (APC) and average annual percentage change (AAPC) [18]. T-distribution and normal distribution were used to assume sharp rocks for APC and AAPC, respectively, to evaluate whether the trend of curve changes and the overall trend of each segment are statistically significant, as well as to derive the $95\%$ confidence intervals (CI). The AAPC is calculated from the APC in children from different countries, age groups and sex weighted geometrically by the length of each period. Comparing AAPC with 0, the curve shows an increasing or decreasing trend with $95\%$ CI not including 0 when the AAPC value is positive or negative. When the $95\%$ CI of AAPC includes 0, the value is stable. The differences were considered statistically significant at $P \leq 0.05.$ ## Trends in prevalence of childhood asthma in China and the USA from 1990 to 2019 The temporal trends in the age-standardized prevalence of childhood asthma by sex and age strata in China and the USA are presented in Fig 1, which can be further confirmed by the results of the joinpoint regression analysis (Table 1). Overall, the prevalence of childhood asthma showed a modest downward trend in Chinese children except for the group of male children aged 1–4 years, whereas it showed a modest upward trend in American children with higher disease prevalence. Compared with the previous change of decreasing and then increasing curve, the prevalence of asthma in children of all three age groups in China increased dramatically from 2014 to 2017, followed by a sharp decreasing trend starting from 2017. The prevalence of asthma among male and female children aged 1–4 years in the USA showed a relatively downward trend from 2010–2019 (males: AAPC -0.2; $95\%$ CI -0.5, 0.1; females: AAPC -0.2; $95\%$ CI -0.3, 0.0), and a small upward trend among children in the other two age groups. However, the AAPC for prevalence of asthma in the USA during this time period was significantly lower than that from 2005 to 2009, which means that the upward trend was reduced. In a cross-sectional comparison of the three age groups, the prevalence of asthma was higher in males than in females in all age groups, and the values of the prevalence were apparently higher in children aged 5–9 years than in the remaining two age groups both in China and the USA. **Fig 1:** *Trends in the age-standardized prevalence per 100,000 population of children asthma by sex and age strata in China and the USA, 1990–2019.(a). 1–4 years old age group. (b). 5–9 years old age group. (c). 10–14 years old age group.* TABLE_PLACEHOLDER:Table 1 ## Trends in DALYs and DALYs due to high BMI of childhood asthma in China and the USA from 1990 to 2019 The temporal trends in the age-standardized DALYs and DALYs due to high BMI of childhood asthma by sex and age strata in China and the USA are shown in Fig 2, while their APC and APCC values are detailed in Table 1. Temporal trends in DALYs and morbidity in childhood asthma were broadly similar, except for a large downward trend in Chinese children of both sexes aged 1–4 years (males: AAPC -1.3; $95\%$ CI -2,8, 0.2; females: AAPC -2.2; $95\%$ CI -4.1, -0.1). DALYs due to high BMI in children with asthma in the USA showed a consistent and substantial upward trend, and gradually decreased in 2010. The same indicator in Chinese children of almost all age group showed a hump of curve from 2014 to 2019, and peaked in 2017, except for that of Chinese females aged 10–14 years who showed a considerable increase from 2012 to 2019 (AAPC 13.0; $95\%$ CI 10.5, 15.6). In terms of AAPC values, *China is* almost universally higher than the USA. Similarly, DALYs due to high BMI for asthma in children aged 5–9 years has the highest values among the three age groups and were higher in males than females in all age groups for both China and the USA. **Fig 2:** *Trends in the age-standardized DALYs rates and DALYs rates due to high BMI per 100,000 population of children asthma by sex and age strata in China and the USA, 1990–2019.(a). DALYs in 1–4 years old age group. (b). DALYs in 5–9 years old age group. (c). DALYs in 10–14 years old age group. (d). DALYs due to high BMI in 1–4 years old age group. (e). DALYs due to high BMI in 5–9 years old age group. (f). DALYs due to high BMI in 10–14 years old age group. Abbreviation: DALYs, disability-adjusted life years; BMI, body mass index.* ## Trends in YLDs and YLLs of childhood asthma in China and the USA from 1990 to 2019 The temporal trends in the age-standardized YLDs and YLLs of childhood asthma by sex and age strata in China and the USA are shown in Fig 3 and supported with Table 2. The trend of YLDs was relatively identical to the prevalence of childhood asthma. Overall, there was a decreasing trend of YLLs for asthma in children both in China and the USA. The decreasing trend was saliently greater in China than in the USA, with the Chinese children aged 1–4 years having the highest onset of YLLs in 1990 and the largest absolute values of AAPC (males: AAPC -10.4; $95\%$ CI -10.7, -10.0; females: AAPC -11.3; $95\%$ CI -11.8, -10.8). Moreover, the YLLs for children aged 10–14 years in the USA have the highest levels than the other groups. YLLs were also higher in males than that in females in all age groups of children. **Fig 3:** *Trends in the age-standardized YLDs rates and YLLs rates due to high BMI per 100,000 population of children asthma by sex and age strata in China and the USA, 1990–2019.(a). YLDs in 1–4 years old age group. (b). YLDs in 5–9 years old age group. (c). YLDs in 10–14 years old age group. (d). YLLs in 1–4 years old age group. (e). YLLs in 5–9 years old age group. (f). YLLs in 10–14 years old age group. Abbreviation: YLDs, years lived with disability; YLLs, years of life lost.* TABLE_PLACEHOLDER:Table 2 ## Trends in ratio of DALYs rates to DALYs rates due to high BMI of childhood asthma in China and the USA from 1990 to 2019 The ratio of DALYs to DALYs due to high BMI in children with asthma showed a consistent upward trend in both China and the USA (Table 3 and Fig 4). The ratio was higher in the USA than in China, and the AAPC in China was approximately 3–4 times higher than that in the USA, suggesting that DALYs due to high BMI of asthma had a higher increasing trend in Chinese children. In terms of age group, Chinese children aged 10–14 years and American children aged 5–9 years had the highest increasing trend of this ratio respectively. Notably, the increasing trend of the ratio is apparently higher in male children than in female children in China, while there was no obvious difference between the two sexes in the USA. **Fig 4:** *Trends in the in the ratio of DALYs due to high BMI/DALYs of children asthma by sex and age strata in China and the USA, 1990–2019.(a). 1–4 years old age group. (b). 5–9 years old age group. (c). 10–14 years old age group. Abbreviation: DALYs, disability-adjusted life years; BMI, body mass index.* TABLE_PLACEHOLDER:Table 3 ## Discussion Our findings revealed that the prevalence of asthma among children of all ages in China showed a slow increase from around 2006 to 2014, which can be attributed with the environmental pollution due to industrialization and urbanization in China, as well as the increase in children with high BMI [19,20]. The rapid increase of prevalence from 2016 to 2017 is related to the release of updated guidelines on the diagnosis and prevention of childhood asthma in China in 2016. In this guideline, clear quantitative indicators for the diagnosis of asthma in children are proposed, thus effectively improving the level of diagnosis [21]. The rapid decline from 2017 to 2019 may be the result of the first Chinese Children’s Asthma Action Plan (CCAAP) released in China in 2017, which combines doctors’ treatment decisions, education on basic asthma treatment and children’s compliance with medical advice, providing a standardized and personalized basis for child treatment and family management, thus enhancing asthma control [22]. DALYs of asthma in children of all ages in China were decreasing, which could be attributed to the release of asthma guidelines and the promotion of formal treatment protocols. These strategies can significantly reduce the disease burden by reducing disease severity and improving symptom control [21,23,24]. Overall, the prevalence and DALYs of childhood asthma in the USA showed an increasing trend from 1990 to 2019. Part of the reason for this is related to the significant increase in the consumption of sugary drinks such as nutritional/energy drinks, juice drinks, and sweet tea among American children, and sugary drinks are thought to be associated with childhood asthma [25]. However, the upward trend of both indicators from around 2010 to 2019 was significantly reduced compared to that in the period from 2005 to 2009. The prevalence and DALYs of childhood asthma decreased in children aged 1–4 years between 2010 and 2019. This may be related to the enactment of the Clean Air Act amendments in 2011 and the Clean Power Plan in 2015 in the USA. In particular, the Regional Greenhouse Gas Initiative (RGGI) has contributed significantly to the reduction of greenhouse gases in the power sector and toxic air pollutants associated with the onset of asthma, such as PM2.5, over the past decade [26]. Another possible reason is that most US states enacted a tax on sugar-sweetened beverages (SSB) by 2010, which plays a role in reducing SSB-induced high BMI [27]. The prevalence and DALYs of asthma were higher in children in the USA than those in China. The main reason may be related to the health hypothesis, indicating that less exposure to infection during childhood may result in a greater chance of developing asthma later in life [3]. YLLs in children in both countries account for a minimal proportion of DALYs, and are associated with a very low mortality rate in children with asthma [28]. High BMI is a risk factor of great concern for childhood asthma [29]. Our study found an increasing trend in DALYs due to high BMI and the ratio of DALYs rates to DALYs rates due to high BMI in Chinese and American children with asthma. One reason for this is that certain parents of children with asthma are concerned about exercise-induced bronchoconstriction (EIB), and thus impose restrictions on the physical activities of their children, consequently affecting weight management [30]. However, in reality, exercise can increase lung function, promote cardiopulmonary fitness, and control asthma [31]. Adequate warm-up before exercise is also recommended [32]. Another reason is that children with high BMI are more inclined to consume a high-fat diet, which can increase bronchial hyperresponsiveness and exacerbate the symptoms of asthma [33]. Positive effects of weight loss on asthma-related outcomes have been demonstrated [34]. The latest Global Initiative for Asthma (GINA 2021) lists obesity as a modifiable risk factor [29]. Since the global prevalence of high BMI is constantly rising, increasing exercise and reducing high-fat food intake to control high BMI are necessary to reduce the risk of asthma [35]. Although the DALYs rates due to high BMI and the ratio of DALYs rates to DALYs rates due to high BMI were both higher in American children than in Chinese children, the increasing trend of both indicators was significantly higher in Chinese children of almost all ages than in the American children. Chinese boys of all ages had a higher percentage of high BMI than girls, which is consistent with the findings of Guo et al. [ 36]. A study of disease burden, injury, and risk factors by state in the USA from 1990 to 2016 found that high BMI was the most important risk factor in the USA, and that exposure was steadily increasing [37]. They believe that renewed efforts to control weight at the community level are important, and that controlling high BMI needs to be a priority for all stakeholders such as physicians, nurses, policy makers, patients, and families. Liu et al. found that the American government partially eliminated the adverse effects of obesity on asthma by imposing a high-calorie tax, increasing the proportion of nutritious food advertisements, banning the sale of soft drinks, increasing opportunities and venues for physical activity, and implementing better health care policies [38]. Therefore, it is recommended that high BMI be taken more into account in the future development of policies for the prevention, control, and treatment of childhood asthma. Moreover, we recommend that children reduce their BMI by increasing physical activity and eating a healthy diet, which parents should encourage and safeguard [39]. The prevalence, DALYs and DALYs due to high BMI were higher in boys than in girls across all age groups in both countries. Boys also have a relatively narrow airway, and are more inclined to vigorous exercise with their greater range of motion, and thus, are more likely to get exposed to allergens. This corroborates the findings of Ellie et al. that boys are more likely to develop allergic diseases than girls based on blood-specific IgE assays and skin prick tests for common allergens [7]. In both China and the USA, the prevalence and DALYs were highest in the 5–9 years age group. This could be due to the fact that children in this age group are at high risk of upper respiratory tract infections because of their low self-management skills and immune levels [36]. In addition, upper respiratory tract infections are important triggers for asthma in children. Moreover, the YLLs for girls aged 10–14 years in the USA have the highest levels than that for girls in the other groups, which is considered to be associated with the increased levels of estrogen and progesterone during the luteal phase in girls of this age group, resulting in increased inflammation of the airway wall [40]. Our study has some limitations. The GBD 2019 lacks data on other risk factors such as high-fat diet and tobacco exposure for asthma in children aged 1–14 years, as well as interactions between risk factors in the estimates, these factors may have introduced bias in the study. In addition, the diagnosis of asthma in children aged 1–4 years is based primarily on clinical judgment and assessment of symptoms and physical findings, which may lead to a failure to reliably diagnose asthma in this age group [4,41]. ## Conclusion DALYs rates due to high BMI and ratio of DALYs rates to DALYs rates due to high BMI were on the rise in children with asthma in both China and USA. High BMI needs to be taken more into account in the development of future policies for the prevention, control, and treatment of childhood asthma. Although both indicators of asthma in children in the USA are higher than in those China, the increasing trend is significantly lower than that in Chinese children of almost all ages. Therefore, it is recommended to learn from the American government to impose a high-calorie tax, increase physical exercise facilities, and provide better health care policies. Besides, we appeal that children increase their physical activity and maintain a healthy diet and that parents encourage and safeguard it. Children aged 5–9 years had the highest prevalence and DALYs in both countries. Careful attention and targeted intervention should be considered in this population, who are particularly at high risk for asthma. ## References 1. von Mutius E, Smits HH. **Primary prevention of asthma: from risk and protective factors to targeted strategies for prevention**. *Lancet* (2020.0) **396** 854-66. DOI: 10.1016/S0140-6736(20)31861-4 2. Kim A, Lim G, Oh I, Kim Y, Lee T, Lee J. **Perinatal factors and the development of childhood asthma**. *Ann Allergy Asthma Immunol* (2018.0) **120** 292-9. DOI: 10.1016/j.anai.2017.12.009 3. Ferrante G, La Grutta S. **The Burden of Pediatric Asthma.**. *Front Pediatr.* (2018.0) **6** 186. DOI: 10.3389/fped.2018.00186 4. Li X, Song P, Zhu Y, Lei H, Chan KY, Campbell H. **The disease burden of childhood asthma in China: a systematic review and meta-analysis.**. *J Glob Health* (2020.0) **10** 010801. DOI: 10.7189/jogh.10.01081 5. Postma DS, Bush A, van den Berge M. **Risk factors and early origins of chronic obstructive pulmonary disease**. *Lancet* (2015.0) **385** 899-909. DOI: 10.1016/S0140-6736(14)60446-3 6. Yu H, Su F, Wang LB, Hemminki K, Dharmage SC, Bowatte G. **The Asthma Family Tree: Evaluating Associations Between Childhood, Parental, and Grandparental Asthma in Seven Chinese Cities.**. *Front Pediatr* (2021.0) **9** 720273. DOI: 10.3389/fped.2021.720273 7. Ellie AS, Sun Y, Hou J, Wang P, Zhang Q, Sundell J. **Prevalence of Childhood Asthma and Allergies and Their Associations with Perinatal Exposure to Home Environmental Factors: A Cross-Sectional Study in Tianjin, China.**. *Int J Environ Res Public Health* (2021.0) **18**. DOI: 10.3390/ijerph18084131 8. Yang X, Zhang Y, Zhan X, Xu X, Li S, Xu X. **Particulate matter exposure is highly correlated to pediatric asthma exacerbation.**. *Aging (Albany NY).* (2021.0) **13** 17818-29. DOI: 10.18632/aging.203281 9. Sinyor B, Concepcion Perez L. *Pathophysiology Of Asthma.* (2022.0) 10. Kunaratnam K, Halaki M, Wen LM, Baur LA, Flood VM. **Tracking Preschoolers’ Lifestyle Behaviors and Testing Maternal Sociodemographics and BMI in Predicting Child Obesity Risk**. *J Nutr* (2020.0) **150** 3068-74. DOI: 10.1093/jn/nxaa292 11. Soares LON, Theodoro EE, Angelelli MM, Lin LL, Carchedi GR, Silva CC. **Evaluating the effect of childhood and adolescence asthma on the household economy**. *J Pediatr (Rio J).* (2022.0) **98** 490-5. DOI: 10.1016/j.jped.2021.12.010 12. Qiu H, Cao S, Xu R. **Cancer incidence, mortality, and burden in China: a time-trend analysis and comparison with the United States and United Kingdom based on the global epidemiological data released in 2020**. *Cancer Commun (Lond).* (2021.0) **41** 1037-48. DOI: 10.1002/cac2.12197 13. Wen H, Xie C, Shi F, Liu Y, Liu X, Yu C. **Trends in Deaths Attributable to Smoking in China, Japan, United Kingdom, and United States From 1990 to 2019.**. *Int J Public Health.* (2022.0) **67** 1605147. DOI: 10.3389/ijph.2022.1605147 14. Diseases GBD, Injuries C. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet* (2020.0) **396** 1204-22. DOI: 10.1016/S0140-6736(20)30925-9 15. 15Metrics IfH, Evaluation. Protocol for the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD). IHME Seattle, WA; 2015.. *Protocol for the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD).* (2015.0) 16. Kobayashi Y, Peters GM, Ashbolt NJ, Shiels S, Khan SJ. **Assessing burden of disease as disability adjusted life years in life cycle assessment**. *Sci Total Environ* (2015.0) 530-531. DOI: 10.1016/j.scitotenv.2015.05.017 17. Stevens GA, Alkema L, Black RE, Boerma JT, Collins GS, Ezzati M. **[Guidelines for Accurate and Transparent Health Estimates Reporting: the GATHER Statement].**. *Epidemiol Serv Saude* (2017.0) **26** 215-22. DOI: 10.5123/S1679-49742017000100023 18. Kim HJ, Fay MP, Feuer EJ, Midthune DN. **Permutation tests for joinpoint regression with applications to cancer rates**. *Stat Med* (2000.0) **19** 335-51. DOI: 10.1002/(sici)1097-0258(20000215)19:3&lt;335::aid-sim336&gt;3.0.co;2-z 19. Mukherjee A, Agrawal M. **A Global Perspective of Fine Particulate Matter Pollution and Its Health Effects**. *Rev Environ Contam Toxicol* (2018.0) **244** 5-51. DOI: 10.1007/398_2017_3 20. Hjellvik V, Tverdal A, Furu K. **Body mass index as predictor for asthma: a cohort study of 118,723 males and females**. *Eur Respir J* (2010.0) **35** 1235-42. DOI: 10.1183/09031936.00192408 21. **[Guideline for the diagnosis and optimal management of asthma in children(2016)].**. *Zhonghua Er Ke Za Zhi.* (2016.0) **54** 167-81. DOI: 10.3760/cma.j.issn.0578-1310.2016.03.003 22. Zhang B, Jin R, Guan RZ, Lin RJ, Chang DY, Zhang LH. **[Evaluation of the efficacy of Chinese Children’s Asthma Action Plan on the long-term management of children with asthma at home].**. *Zhonghua Yi Xue Za Zhi* (2020.0) **100** 3702-5. DOI: 10.3760/cma.j.ch112137-20200408-01125 23. Zhou X, Hong J. **Pediatric Asthma Management in China: Current and Future Challenges.**. *Paediatr Drugs.* (2018.0) **20** 105-10. DOI: 10.1007/s40272-017-0276-7 24. Asher I, Haahtela T, Selroos O, Ellwood P, Ellwood E. **Global Asthma Network survey suggests more national asthma strategies could reduce burden of asthma.**. *Allergol Immunopathol (Madr).* (2017.0) **45** 105-14. DOI: 10.1016/j.aller.2016.10.013 25. Xie L, Atem F, Gelfand A, Delclos G, Messiah SE. **Association between asthma and sugar-sweetened beverage consumption in the United States pediatric population.**. *J Asthma* (2022.0) **59** 926-33. DOI: 10.1080/02770903.2021.1895210 26. Perera F, Cooley D, Berberian A, Mills D, Kinney P. **Co-Benefits to Children’s Health of the U.S. Regional Greenhouse Gas Initiative**. *Environ Health Perspect* (2020.0) **128** 77006. DOI: 10.1289/EHP6706 27. Chriqui JF, Eidson S, Chaloupka F. **State sales taxes on regular soda (as of January 2014)–BTG Fact Sheet.**. *Bridging the Gap Programme* (2014.0) 28. Ughasoro MD, Eze JN, Oguonu T, Onwujekwe EO. **Burden of childhood and adolescence asthma in Nigeria: Disability adjusted life years.**. *Paediatr Respir Rev.* (2021.0). DOI: 10.1016/j.prrv.2021.07.004 29. Reddel HK, Bacharier LB, Bateman ED, Brightling CE, Brusselle GG, Buhl R. **Global Initiative for Asthma Strategy 2021: executive summary and rationale for key changes**. *Eur Respir J* (2022.0) **59**. DOI: 10.1183/13993003.02730–2021 30. Kornblit A, Cain A, Bauman LJ, Brown NM, Reznik M. **Parental Perspectives of Barriers to Physical Activity in Urban Schoolchildren With Asthma.**. *Acad Pediatr.* (2018.0) **18** 310-6. DOI: 10.1016/j.acap.2017.12.011 31. 31Global Strategy for Asthma Management and Prevention. Available from: https://ginasthma.org/gina-reports/ (accessed on 1 May 2020). 32. Parsons JP, Hallstrand TS, Mastronarde JG, Kaminsky DA, Rundell KW, Hull JH. **An official American Thoracic Society clinical practice guideline: exercise-induced bronchoconstriction**. *Am J Respir Crit Care Med* (2013.0) **187** 1016-27. DOI: 10.1164/rccm.201303-0437ST 33. Wood LG. **Diet, Obesity, and Asthma.**. *Ann Am Thorac Soc* (2017.0) **14** S332-S8. DOI: 10.1513/AnnalsATS.201702-124AW 34. Juel CT, Ali Z, Nilas L, Ulrik CS. **Asthma and obesity: does weight loss improve asthma control? a systematic review**. *J Asthma Allergy* (2012.0) **5** 21-6. DOI: 10.2147/JAA.S32232 35. Li X, Cao X, Guo M, Xie M, Liu X. **Trends and risk factors of mortality and disability adjusted life years for chronic respiratory diseases from 1990 to 2017: systematic analysis for the Global Burden of Disease Study 2017**. *BMJ* (2020.0) **368** m234. DOI: 10.1136/bmj.m234 36. Guo X, Li Z, Ling W, Long J, Su C, Li J. **Epidemiology of childhood asthma in mainland China (1988–2014): A meta-analysis.**. *Allergy Asthma Proc* (2018.0) **39** 15-29. DOI: 10.2500/aap.2018.39.4131 37. Mokdad AH, Ballestros K, Echko M, Glenn S, Olsen HE. **The State of US Health, 1990–2016: Burden of Diseases, Injuries, and Risk Factors Among US States**. *JAMA* (2018.0) **319** 1444-72. DOI: 10.1001/jama.2018.0158 38. Liu J, Yuan M, Chen Y, Wang Y, Wang Q, Zhang Q. **Global burden of asthma associated with high body mass index from 1990 to 2019**. *Ann Allergy Asthma Immunol* (2022.0) **129** 720-30. DOI: 10.1016/j.anai.2022.08.013 39. Zhang D, Zheng J. **The Burden of Childhood Asthma by Age Group, 1990–2019: A Systematic Analysis of Global Burden of Disease 2019 Data.**. *Front Pediatr.* (2022.0) **10** 823399. DOI: 10.3389/fped.2022.823399 40. Melgert BN, Ray A, Hylkema MN, Timens W, Postma DS. **Are there reasons why adult asthma is more common in females?**. *Curr Allergy Asthma Rep* (2007.0) **7** 143-50. DOI: 10.1007/s11882-007-0012-4 41. Castro-Rodriguez JA, Holberg CJ, Wright AL, Martinez FD. **A clinical index to define risk of asthma in young children with recurrent wheezing**. *Am J Respir Crit Care Med* (2000.0) **162** 1403-6. DOI: 10.1164/ajrccm.162.4.9912111
--- title: Inferring feature importance with uncertainties with application to large genotype data authors: - Pål Vegard Johnsen - Inga Strümke - Mette Langaas - Andrew Thomas DeWan - Signe Riemer-Sørensen journal: PLOS Computational Biology year: 2023 pmcid: PMC10038287 doi: 10.1371/journal.pcbi.1010963 license: CC BY 4.0 --- # Inferring feature importance with uncertainties with application to large genotype data ## Abstract Estimating feature importance, which is the contribution of a prediction or several predictions due to a feature, is an essential aspect of explaining data-based models. Besides explaining the model itself, an equally relevant question is which features are important in the underlying data generating process. We present a Shapley-value-based framework for inferring the importance of individual features, including uncertainty in the estimator. We build upon the recently published model-agnostic feature importance score of SAGE (Shapley additive global importance) and introduce Sub-SAGE. For tree-based models, it has the advantage that it can be estimated without computationally expensive resampling. We argue that for all model types the uncertainties in our Sub-SAGE estimator can be estimated using bootstrapping and demonstrate the approach for tree ensemble methods. The framework is exemplified on synthetic data as well as large genotype data for predicting feature importance with respect to obesity. ## Author summary Artificial intelligence and machine learning have been increasingly popular tools for modelling complex relationships in medicine and genomics. For example a machine learning model for predicting the likelihood of a particular person developing some disease. The prediction model can for instance be based on genomics data, which consists of a large number of features for each single person. Such prediction models can be very complex and difficult to interpret, hence they are often denoted black-box models. However, to exploit the knowledge the prediction model has gained, we must be able to interpret it, and explain which features are important for the model, but also for the underlying data. We investigate a theoretical approach for extracting feature importance, even when the model input consists of many features. Lastly, we emphasize the need for estimating the uncertainty of the individual feature importance, and provide a bootstrap procedure for doing so. ## Introduction With the strong improvement of black-box machine learning models such as gradient boosting models and deep neural networks, the question of how to infer feature importance, including uncertainty estimates, in these types of models has become increasingly important. This is particularly important if the results from the model can be trustfully investigated further within medicine or genomics such as when applied to drug discovery [1]. The Shapley decomposition, a solution concept from cooperative game theory [2], has enjoyed a surge of interest in the literature on explainable artificial intelligence in recent years, (cf. [ 3–16]). A widely used Shapley-based framework for deriving feature importance in machine learning models post-training is Shapley additive explanations (SHAP) [4, 6], which explains individual predictions’ deviations from the average model prediction. As such, SHAP attributes feature importance as they are perceived by the model. The more recently introduced Shapley additive global importance (SAGE) is also based on the Shapley decomposition, but attributes feature importance by a global decomposition of the model loss across a whole data set [17]. The SAGE framework thus provides an explanation of the influence of the features taking into account not only the model, but also implicitly the data via the loss function, thus encapsulating that the model might not be—and most likely is not—a perfect description of the data [18]. The SAGE value needs to be estimated, and the SAGE estimator is itself a random variable as the corresponding SAGE estimate is based on data of finite size generated from some unknown probability distribution. As is the case for any feature importance score, we argue that the uncertainty in the estimate is equally important as the estimate itself for drawing conclusions. However, computation of the SAGE-estimate is infeasible even for moderate-sized data, and thus further approximations are needed [17]. To this end, we introduce Sub-SAGE, which is motivated by SAGE but can be estimated exactly for tree-ensemble models, by using a reduced subset of coalitions. Additionally, we describe how to estimate a confidence interval for the Sub-SAGE value. No calculation of such uncertainty exists in the SAGE package or the literature. We estimate the confidence interval using paired bootstrapping, and demonstrate its calculation for tree ensemble models on simulated as well as genotype data. We argue that this procedure provides a way to infer the true feature importance in the underlying data. The remainder of this paper is structured as follows. In Materials and methods we introduce the particular genotype data set to be used, as well as background concepts such as the Shapley value, SHAP and SAGE, before moving on to Sub-SAGE, and its uncertainty. The method is exemplified using synthetic data. In the Results section, the method is applied on the genotype data before we discuss the results in Discussion and conclusion. ## Data and use case In order to evaluate the Sub-SAGE feature importance score, we will apply it using collected genotype data from the UK Biobank [19, 20]. UK *Biobank is* a large prospective cohort study in the United Kingdom that began in 2006 consisting of about 500000 participants. As use case, we considered the aim of inferring the feature importance of single nucleotide polymorphisms (SNPs) with respect to a logistic regression model for predicting the susceptibility of obesity (BMI ≥ 30). The model includes a large number of SNPs as features, as well as accounting for non-linear effects. Obesity was selected since this particular trait has been extensively researched in previous genome-wide association studies (GWAS) providing a meaningful way to evaluate our method [21]. ## Ethics statement Ethical approval was obtained by the UK Biobank from the North West Multicentre Research Ethics Committee, the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland. All participants provided written informed consent. The research in this paper has been conducted using the UK Biobank Resource under Application Number 32285. The application for access to the UK Biobank Resource was approved on October 10, 2018. ## Shapley-based explanation methods We provide a brief introduction to the Shapley decomposition-based SHAP and SAGE frameworks to quantify feature importance in machine learning models. An advantage of such Shapley-based frameworks is that they in principle can be applied on any input-output model, parametric or non-parametric, including non-linear machine learning models such as neural networks or tree ensemble models. Consequently, non-linear effects can also be captured using these frameworks. The Shapley decomposition is a solution concept from cooperative game theory [2]. It provides a decomposition of any value function v(S) that characterises the game, and produces a single real number, or payoff, per set of players in the game (coalitions). The resulting decomposition satisfies the three properties of efficiency, monotonicity and symmetry, and is provably the only method to satisfy all three [22, 23, Thm. 2]. For details see S1 File. Consider a supervised learning task characterised by a set of M features xi and corresponding univariate responses yi, for $i = 1$, …, N, and a fitted model that is a mapping from feature values to response values, i.e. xi→y^(xi). As usual, uppercase letters denote random variables while lowercase letters denote observed data values. In this work, we assume independent features, implying E[Xj|Xk = xk] = E[Xj] ∀ j ≠ k. This assumption is further discussed in Discussion and conclusion section. ## The SHAP value Definition 1. Let S⊆M\{k}, with M={1,…,M}, denote a subset of all features not including feature k. S‾ denote the corresponding complement subset of excluded features (S∪S‾=M). The SHAP value, ϕkSHAP(x,y^), for a feature with index k with respect to feature values x and a corresponding fitted model y^, is defined as [6] ϕkSHAP(x,y^)=∑S⊆M\{k}|S|!(M-|S|-1)!M![vx,y^(S∪{k})-vx,y^(S)]. [ 1] Here, the value function vx,y^(S) is defined as the expected output of a prediction model conditioned that only a subset S of all features are included in the model, vx,y^(S)=EXS¯[y^(X|XS=xS)]. [ 2] For instance, if XS¯ is a continuous random vector and we assume all features to be mutually independent, we have EXS¯[y^(X|XS=xS)]=∫xS¯y^(XS=xS,XS¯=xS¯)p(XS¯=xS¯|XS=xS)dxS¯=∫xS¯y^(XS=xS,XS¯=xS¯)p(XS¯=xS¯)dxS¯. [3] The stochastic behaviour in y^(X|XS=xS) is due to the random vector XS‾ of unknown feature values. We can think of the difference vx,y^(S∪{k})−vx,y^(S) as the mean difference in a single model prediction when using feature k in the model compared to when the value of feature k is absent. The larger absolute SHAP value a feature k has in a single prediction, the more influence the feature is regarded to have on this particular prediction. ## The SAGE value Define a loss function ℓ(yi,y^(xi)) as a measure of how well the fitted model y^(xi) maps the features to a response, compared to the true response value yi. As defined in [17], we take the SAGE value function w(S) as the expected difference in the observed value of the loss function when the features in S are included in the model compared to excluding all features. Definition 2. Given a data generating process (X, Y), a function y^ to model the relationship between X and Y, and a loss function ℓ(yi,y^(xi)), we define wX,Y,y^(S) as: wX,Y,y^(S)=EX,Y[ℓ(Y,VX,y^(∅))]-EX,Y[ℓ(Y,VX,y^(S))]. [ 4] Here, ∅ denotes the empty set, while VX,y^(S) is the stochastic version of Eq [2]. Specifically, VX,y^(S) is a random variable since its observed value varies depending on the random vector XS. vx,y^(S) is a constant as we condition on the observed vector xS. For the case where x and y are continuous, the expected value of the loss function when only a subset S of feature values are known is given by EX,Y[ℓ(Y,VX,y^(S))]=∫y∫xSℓ(y(x),EXS¯[y^(X|XS=xS)])p(y|xS)p(xS)dxSdy. [ 5] Notice that the computation of vx,y^(S)=EXS‾[y^(X|XS=xS)] happens inside the loss function, which is usually non-linear. Also note that in Eq [5], we integrate over all possible values of XS. Definition 3. The SAGE value for a feature k is defined as [17] ϕkSAGE(X,Y,y^)=∑S⊆M\{k}|S|!(M-|S|-1)!M![wX,Y,y^(S∪{k})-wX,Y,y^(S)]. [ 6] We can think of the difference wX,Y,y^(S∪{k})−wX,Y,y^(S) as the expected difference in the loss function when including feature k in the model compared to excluding feature k with respect to the subset S of known feature values. SAGE is therefore a global feature importance score (as opposed to the local SHAP value) as it does not evaluate a single prediction, but rather the impact feature k has across all predictions. The use of the loss function in the SAGE definition also ensures that the feature importance is not only based on the model, as for the SHAP value, but also on the data itself. An interpretation of SAGE is that a positive SAGE value for a feature implies that including this feature in the model reduces the expected model loss compared to when not including the feature. The features and response can be continuous or discrete. In the discrete case, integrals must be replaced by sums in Eqs [3] and [5]. The expressions in Eqs. [ 2] and [4] are in general unknown and need to be estimated for each choice of model and loss function. Consequently, the SHAP and SAGE values become estimates as well. ## Tree ensemble models Consider a tree ensemble model consisting of several regression trees fτ(xi) with predicted response y^(xi), such that y^(xi)=∑τ=1Tfτ(xi) for T trees. By the linearity property of the expected value, we have vx,y^(S)=EXS¯[∑τ=1Tfτ(X|XS=xS)]=∑τ=1TEXS¯[fτ(X|XS=xS)]. [ 7] The computation of EXS‾[fτ(X|XS=xS)] can be understood through a simple example: The regression tree illustrated in Fig 1 has depth two and splits on the two features indexed 1 and 2, which are continuous and mutually independent. The regression tree has parameters such as splitting points, tj, for branch nodes, and leaf values vj, for leaf nodes. For an observed value of x2 = 3 we have EXS¯[fτ(X|XS=xS)]=EX1[fτ(X1|X2=3)]=P(X1≥20)v3+P(X1<20)v2. [ 8] **Fig 1:** *A regression tree including two features X1 and X2.* *In* general, we do not know the value of P(X1 ≤ 20), and need to estimate it. Consider N data instances with recorded feature values from feature k. An unbiased estimate of P(Xk ≤ t) is then P^(Xk≤t)=1N∑$i = 1$NI(xi,k≤t), [9] where xi,k is the observed value of feature k for data instance i, and I(⋅) is the indicator function. Using this estimate, we can also get an unbiased estimate for Eq [8]. An unbiased estimate of EXS‾[fτ(X|XS=xS)] for any regression tree can be achieved by a recursive algorithm [4] with running time O(L2M), where L is the number of leaves, see Algorithm 1. This requires the estimated probabilities of ending at a particular node j given previous information from the ancestor nodes. If the feature used for splitting at a particular node j is not used in any of the ancestor nodes of node j, an estimate such as in Eq [9] can be used. If the feature is used for splitting in any of the ancestor nodes, this must be accounted for by restricting to the interval of possible values the particular feature can take at node j. Algorithm 1. Recursive algorithm for computation of EXS‾[fτ(X|XS=xS)]. 1: Input: Tree fτ with depth d, leaf values v=(v1,…,v2d), feature used for splitting f=(f1,…,f2d−1) and corresponding splitting points t=(t1,…,t2d−1). Estimated probabilities of ending at a node j given previous information, for all nodes in the tree, p=(p1,…,p2d−1), by using some data (x1, y1), …, (xN, yN) of size N. The subset of features S with corresponding known values xS. The left and right descendant node for each internal node l=(l1,…,l2d−1) and r=(r1,…,r2d−1). The index of a node j in the tree fτ. 2: Function CondExpTree(j, fτ, v, t, f, l, r, p) 3: if IsLeaf(j) then 4: return vj 5: else 6: if fj∈S then 7: if xj ≤ tj then 8: return CondExpTree(lj, fτ, v, t, f, l, r, p) 9: else 10: return CondExpTree(rj, fτ, v, t, f, l, r, p) 11: end if 12: else 13: return CondExpTree(lj, fτ, v, t, f, l, r, p) plj + 14: CondExpTree(rj, fτ, v, t, f, l, r, p) prj 15: end if 16: end if 17: End Function 18: CondExpTree(1, fτ, v, t, f, l, r, p) ▹ Start at root node. ## SAGE in practice As the expressions in Eqs [2] and [4] must be estimated, in practice we get a SAGE estimator rather than a SAGE value. However, since the SAGE estimator requires summing over all 2M−1 subsets S⊆M\{k}, for each feature, computing the SAGE estimator for observed data with many features becomes infeasible. In [17], the SAGE estimate is approximated through a Monte Carlo simulation process. Specifically, instead of iterating over all 2M−1 subsets, a subset S is randomly sampled with replacement in each iteration out of I iterations in total. The differences wX,Y,y^(S∪{k})−wX,Y,y^(S) for each S are estimated by sampling data instances with replacement and computing sample means (see [17], S1 File for details). For an arbitrarily large data set, the authors show convergence to the true SAGE estimate as I → ∞. Among other things, both the accuracy and convergence speed of the algorithm naturally depend on the number of features in the prediction model. Notice that [17] provides the degree of convergence of the approximation of the estimate, not the uncertainty in the estimate. Keeping in mind that the SAGE estimator is a random variable, we argue that its uncertainty is equally important as the estimate itself. No calculation of this inherent uncertainty exists in the SAGE package nor the literature. To this end, we introduce Sub-SAGE, which is inspired by the SAGE framework, but consists of a reduced number of subsets S∈Q. While applicable to any number of features, it is best suited for interpreting a small number of features, or a small subset of features in a large feature set. ## Sub-SAGE Given hundreds or thousands of features in a model, the computation time required to get a satisfactory accurate estimate of SAGE [17] for each feature, quickly becomes unacceptable. A hybrid approach is to select a reduced subset of features of particular interest to investigate. For instance in a GWAS, such a filtering process can be achieved via a generalized linear mixed model [24], and rank importance based on the computed p-values. An alternative filtering procedure that also accounts for non-linear effects is described in [25]. The association between the reduced subset of promising features, and the response can then be more thoroughly investigated via a non-linear machine learning model together with SHAP values. See [25], Figure 10 for an example. To infer whether the model-based importance of the features investigated via SHAP values is also reflected in the underlying data generating process, one can compute SAGE values. However, we typically want the reduced subset of features to be sufficiently large in order to reduce the chance of missing out on important features. Even in this case, computation of SAGE values may be impractical or even infeasible. For this purpose, we introduce Sub-SAGE, where only a selection of the in total 2M−1 subsets are involved in the computation for each feature. If we want to measure the importance of a feature k based on its marginal effect, as well as potential pairwise interactions it may be involved in, computing S={∅} and S={m} for $m = 1$, …, k − 1, k + 1, …, M is sufficient. In addition, by including S={1,…,k−1,k+1,…,M}, the set of all features except feature k, can be used to measure the importance of feature k in the presence of all features at the same time. Definition 4. Let Qk denote the set of subsets including S={∅}, S={m} for $m = 1$, …, k − 1, k + 1, …, M, and S={1,…,k−1,k+1,…,M}. We define the Sub-SAGE value, ψk, for feature k as ψk(X,Y,y^)=∑S∈Qk|S|!(M-|S|-1)!3(M-1)![wX,Y,y^(S∪{k})-wX,Y,y^(S)], [10] Each subset is weighted such that the sum of the weights of all subsets with equal size is the same for each subset size. In addition, the sum of all weights is equal to one. Hence, the construction is similar to the weights defined for Shapley values. See S1 File for details. In this particular case, there are three possible subset sizes, and so the sum of the weights for each subset size is 13. Shapley properties such as symmetry, dummy property and monotonicity still hold for Sub-SAGE. However, as the sum is not over all possible subsets, the Sub-SAGE values do no longer satisfy the efficiency axiom of the Shapley decomposition, which SHAP and SAGE do (see S1 File). However, we still consider the Sub-SAGE to be informative with respect to feature importance via the computed differences wX,Y,y^(S∪{k})−wX,Y,y^(S). In addition, the purpose is only to evaluate a small fraction of all features, not all of them. By only considering a reduced number of subsets S, compared to SAGE, and only considering a reduced number of features to evaluate, both computing the Sub-SAGE estimate as well as the uncertainty in the corresponding Sub-SAGE estimator become feasible for black-box models, such as tree ensemble models and neural networks. ## Using Sub-SAGE to infer true relationships in the data As the goal is to infer feature importance from a black-box model using Sub-SAGE values, similar to calculating p-values without taking into account the effect of model selection, we must be extra careful. Any model selection procedure using training data is likely to overfit, resulting in a model consisting of possibly false relationships that are not a general property of the population from which the data was sampled. It is therefore essential that the Sub-SAGE value is calculated using independent data the model was not fitted on. We denote such independent data as test data, (X10,Y10),…,(XNI0,YNI0), with NI samples in total. The following observation is proven S1 File. ## Observation: Sub-SAGE value in multiple linear regression Consider a fitted linear regression model y^i=β^Txi. By using test data independent of the data used for constructing the linear regression model, and using the squared error loss, one can show that for a feature k, and any S∈Qk: wX,Y,y^(S∪{k})-wX,Y,y^(S)=2β^kCov(Y,Xk)-β^k2Var(Xk). [ 11] As the expression is independent of the subset S, this is also equal to the Sub-SAGE value of feature k. The first term in Eq [11] can be interpreted as the extent to which the influence of feature k based on the model constructed using training data, is reflected in the independent test data. If the signs of β^k and Cov(Y, Xk) are identical, the first term is positive. If they differ, the Sub-SAGE value will be negative since the second term in Eq [11] is always negative. The second term βk^2Var(Xk) is equal to the increased variance in the model by including feature k. If the model regards the feature as important (resulting in non-zero β^k), and even if the covariance between Xk and Y has the same sign as β^k, the benefit of including feature k in the model will depend on the increased variance of the model. This is by construction in line with the bias-variance trade-off [26]. ## Sub-SAGE applied on tree ensemble models SHAP values can be estimated efficiently for tree ensemble models, even with hundreds of thousands of features [25], by improving Algorithm 1 to get a significantly reduced running time of O(TLD2), for T trees each of tree depth D [4]. Unfortunately, there is no similar way to reduce the running time for estimation of SAGE values, as well as Sub-SAGE values, for tree ensemble models with non-linear choices of loss functions [4]. We consider a tree ensemble model consisting of T trees. Consider a particular feature k to compute the Sub-SAGE value as well as a subset S∈Qk. We separate the trees in the model into two groups τk and the complement group (τ¯k) where τk is the set of trees including feature k used at least once for splitting. ## Regression with squared error loss For regression a common loss function is the squared error between the response and prediction per sample, i.e. ℓ=(y(x)−y^(x))2. Then one can show that (see S1 File for the derivation), wX,Y,y^(S∪{k})-wX,Y,y^(S)=EX,Y[(Y(X)-VX,y^(S))2]-EX,Y[(Y(X)-VX,y^(S∪{k}))2]=EX,Y[2Y(X)(∑j∈τkVX,fj(S∪{k})-VX,fj(S))+(∑j∈τkVX,fj(S))2-(∑j∈τkVX,fj(S∪{k}))2+2(∑j∉τkVX,fj(S))(∑j∈τkVX,fj(S∪{k})-VX,fj(S))]. [ 12] ## Classification with binary cross-entropy loss A commonly used loss function for binary classification problems is binary cross-entropy, ℓ=−y(x)logy^(x)−(1−y(x))log(1−y^(x))=(1−y(x))∑$j = 1$Tfj(x) +log(1+e−∑$j = 1$Tfj(x)). For this loss function, one can show that (see S1 File) wX,Y,y^(S∪{k})-wX,Y,y^(S) [13] =EX,Y[(1-Y(X))∑$j = 1$TVX,fj(S)+log(1+exp(-∑$j = 1$TVX,fj(S)))] [14] -EX,Y[(1-Y(X))∑$j = 1$TVX,fj(S∪{k})+log(1+exp(-∑$j = 1$TVX,fj(S∪{k})))] [15] =EX,Y[(1-Y(X))(∑j∈τkVX,fj(S)-VX,fj(S∪{k})) [16] +log(1+exp(-∑j∈τkVX,fj(S)-∑j∉τkVX,fj(S))1+exp(-∑j∈τkVX,fj(S∪{k})-∑j∉τkVX,fj(S∪{k})))]. [ 17] ## Plug-in estimates As discussed earlier, the expression wX,Y,y^(S∪{k})−wX,Y,y^(S) needs to be estimated for each S∈Qk, and based on data, (x10,y10),…,(xNI0,yNI0), never used during training of the model. Let v^x0,y0,fτ(S) for a particular observation (x0, y0) and regression tree fτ denote the estimate of vx0,fτ(S)=EXS‾[fτ(X0|XS0=xS0)] as described in Algorithm 1. A plug-in estimate of ψk, denoted ψ^k, for a regression problem with continuous response for a tree ensemble model using the squared error loss is given by ψ^k=∑S∈Q|S|!(M-|S|-1)!3(M-1)![2NI∑$i = 1$NIyi0(∑j∈τkv^xi0,fj(S∪{k})-v^xi0,fj(S))+1NI∑$i = 1$NI(∑j∈τkv^xi0,fj(S))2-1NI∑$i = 1$NI(∑j∈τkv^xi0,fj(S∪{k}))2+2NI∑$i = 1$NI(∑j∉τkv^xi0,fj(S))(∑j∈τkv^xi0,fj(S∪{k})-v^xi0,fj(S))]. [ 18] The corresponding plug-in estimate for the binary cross-entropy loss given in Eq [17] can be found in a similar fashion, basically by estimating expected values as their corresponding sample means. For tree ensemble models with tree stumps (maximum depth of one for each tree), the estimate in [18] is further reduced and can be expressed as sample variance and covariance terms, see S1 File. ## Inference of Sub-SAGE via bootstrapping The importance of any feature may be evaluated by estimating Sub-SAGE values. Similar to SAGE, a positive Sub-SAGE value for a feature k indicates that including the feature in the model is expected, based on the subsets S∈Qk, to reduce the loss function. However, the corresponding Sub-SAGE plug-in estimator given the data generating process (X10,Y10),…,(XNI0,YNI0) from some unknown probability distribution includes uncertainty, and this should be evaluated before making any assumptions about feature importance. The complexity of the Sub-SAGE plug-in estimators makes nonparametric bootstrapping a tempting approach. One possible procedure is to iteratively, given independent data points (x10,y10),…,(xNI0,yNI0), resample the data points with replacement to get a bootstrapped sample (x1*,y1*),…,(xNI*,yNI*), and train a model for each bootstrapped sample (with potential hyperparameters fixed). For each such generated model, a corresponding plug-in estimate, ψ^b*, can be computed, and after B iterations, the sample (ψ^1*,…,ψ^B*) can approximate B realizations arising from the true distribution of the plug-in estimator. However, in a high-dimensional setting, generating even one model may be time-consuming, and there may be circumstances where only one particular model is available for the user together with a test data sample of insufficient size to train additional models. Another option is to leave the model fixed, and only bootstrap the data repeatedly to get the sample (ψ^1*,…,ψ^B*). In this paper, we focus on the latter procedure. A (1 − 2α)$100\%$ confidence interval can be approximated by the percentile interval given by [ψ^*(α),ψ^*(1−α)], where ψ^*(α) is the 100α empirical percentile, meaning the B ⋅ αth least value in the ordered list of the samples (ψ^1*,…,ψ^B*). The accuracy in the percentile interval increases for larger number of bootstrap samples. The algorithm of the paired bootstrap applied specifically to tree ensemble models is given in Algorithm 2. Notice that for each bootstrap sample, the probability estimates in the trees need to be updated according to Eq [9]. In situations where the plug-in estimator is biased, or there is skewness in the corresponding distribution, the bias-corrected and accelerated bootstrap [27], may give even more accurate confidence intervals at the cost of considerable increase in computational efforts. Algorithm 2 Paired bootstrap of Sub-SAGE value with percentile interval 1: Given independent test data (x10,y10),…,(xNI0,yNI0), model y^(x)=∑τ=1Tfτ(x), feature k, a loss function and α to estimate (1 − 2α)$100\%$ confidence interval: 2: Pre-allocate vector BootVec of length B, the total number of bootstrap samples. 3: for $b = 1$, 2, …, B do 4: *Resample data* NI times with replacement to get 5: (x1*,y1*),…,(xNI*,yNI*) 6: Update probabilities estimates in all the trees in y^(x) to get p* 7: BootVec[b] = ψ^k* 8: end for 9: Percentile interval given by [ψ^*(α),ψ^*(1−α)] ## Proof of concept—With known underlying data generating process In this section, we exemplify the Sub-SAGE method on synthetic data with a known relationship defined as f(Xi)=a0+a1Xi,1+a2Xi,2+a21Xi,1eXi,2+a3Xi,32+a4sin(Xi,4)+a5log(1+Xi,5)-Xi,5I(Xi,6>7)+ϵi, [19] with a0 = −0.5, a1 = 0.03, a2 = −0.05, a21 = 0.3, a3 = 0.02, a4 = 0.35, a5 = −0.2, and where the features are sampled from the following distributions X1∼Binom(size=2,$$p \leq 0.4$$)X2∼Binom(size=2,$$p \leq 0.04$$)X3∼Γ(shape=10,rate=2)X4∼Unif(0,π)X5∼Poisson(λ=15)X6∼N(μ=0,σ=10)ϵ∼N(μ=0,σ=2). [ 20] In addition, we generate 94 noise variables: $j = 7$, …, 47 with a normal distribution Xj ∼ N(μj, σj) and $j = 48$, …, 100 with a binomial distribution Xj ∼ Binom(2, pj) where μj, σj and pj are sampled from a uniform distribution. In realistic applications, the data distributions and relations are unknown and the purpose of model fitting is to estimate the relations between variables. Data is generated to give a total of 16000 samples, and then separated randomly in three disjoint subsets: Data for training ($50\%$), data for evaluation during training ($30\%$) and independent test data ($20\%$) used for estimating Sub-SAGE values. We fit an ensemble tree model using XGBoost [28] to the true influential features 1, …, 6 together with the noise variables 7, …, 100. The hyperparameters are fixed to max_depth = 2, learning rate η = 0.05, subsample = 0.7, regularization parameters λ = 1, γ = 0 and colsample_bytree = 0.8 with early_stopping_rounds = 20 using training data ($$n = 8000$$) and validation data ($$n = 4800$$), and a squared error loss. See [28] for details about the hyperparameters. This results in a final model including a total of 230 trees and 62 unique features out of the 100 input-features. From the trained model, each feature is given a score to evaluate its feature importance based on the model. We apply the expected relative feature contribution (ERFC) [25], given data of size N, which is basically a summary score from the corresponding SHAP values for each feature and individual data point, κk=∑$i = 1$N|ϕi,kSHAP(xi,y^)||ϕ0SHAP|+∑$j = 1$K|ϕi,jSHAP(xi,y^)|, [21] with ϕ0SHAP=vx,y^(∅). The ERFCs scores can be computed based on the data used to construct the model, as we only need to measure what the model considers important. The features with the largest ERFC-values are then considered the most promising ones based on the model. Depending on your hypothesis of interest, one can evaluate the uncertainty in the feature importance by computing Sub-SAGE estimates with corresponding bootstrap-derived percentile intervals. However, it is important that the Sub-SAGE estimates are calculated based on independent test data never used during training. From the trained model, we compute the ERFC based on the training data and validation data together ($$n = 12800$$), and Table 1 shows the top 10 features with the largest ERFC-values. **Table 1** | Feature | ERFC | | --- | --- | | x 6 | 0.48 | | x 5 | 0.06 | | x 3 | 0.026 | | x 1 | 0.022 | | x 4 | 0.0036 | | x 2 | 0.003 | | x 12 | 0.0028 | | x 30 | 0.0022 | | x 40 | 0.0019 | We see that the XGBoost model has accurately ranked the most influential features among the top 10 list, for this rather simple relationship. These scores, based on SHAP values, are only with respect to what the model considers important. The Sub-SAGE can now be applied to infer whether feature importance from the model is also reflected in the data. As an example, let us consider features 6, 1, 2 and 12 where feature 6 has a strong influence, feature 1 has a weaker influence, and feature 2 has the weakest influence, while feature 12 has no influence with respect to f(xi) in Eq [19]. Their Sub-SAGE estimate along with histograms to estimate the corresponding distribution of the Sub-SAGE estimators, using $B = 1000$ bootstrap samples, are shown in Fig 2 for training plus validation data as well as for independent test data. **Fig 2:** *The estimate of the Sub-SAGE, and the corresponding bootstrap distribution for the synthetic data for features x6, x2, x1 and x12, when applying data used during training (orange), and independent test data (blue).* We see that Sub-SAGE values inferred using training data overestimate the false influence of feature 12, while using the test data correctly indicates that feature 12 has a weak or no influence. We also see from the other histograms that using the training data underestimates the uncertainty in the Sub-SAGE estimate. By using the test data for computation of the Sub-SAGE estimates, the estimated $95\%$ percentile intervals of the Sub-SAGE values for each feature are 6: (39.45, 44.15), 1: (−0.038, 0.14), 2: (−0.043, 0.040) and 12: (−0.030, 0.0050). These ranges allow us to conclude that feature 6, correctly, is highly influential, while feature 12 is highly unlikely to have any influence. Moreover, feature 1 is likely to be influential, while the influence of feature 2 is very uncertain. To correct for a potential bias in the plug-in estimator of the Sub-SAGE as well as potential changes in the standard deviation of the estimator at different levels, the bias-corrected and accelerated bootstrap confidence interval may give more accurate bootstrap confidence intervals [27]. This results in the intervals 6: (39.45, 44.13), 1: (−0.034, 0.14), 2: (−0.047, 0.037) and 12: (−0.031, 0.0040), with only negligible changes from the percentile confidence intervals. As the data generating process is known, we can compare the true SHAP value at each point with the corresponding SHAP value from the fitted model. Fig 3 shows that the influence of feature 6 is quite accurately modelled, while the effect of feature 1 and particularly feature 2 is highly underestimated when x1 = 1 and x2 = 2. Since there is an interaction effect involving features 1 and 2, the SHAP value of feature 1 depends on the value of feature 2. It also becomes clear that feature 12, according to the model, has a negative trend in the SHAP value, but the true SHAP value is equal to zero (no importance), regardless of the value of feature 12. See S1 File for derivations. This shows an example where feature 12 is erroneously attributed high predictive importance by SHAP, while the corresponding Sub-SAGE value correctly indicates it has no importance. **Fig 3:** *Comparison of true SHAP value for each data point with the estimated SHAP value from the model fitted on the synthetic data, Eq (19). The deviations explain the reasons behind under- and overestimation of feature importance using SHAP values.* We can explore the results even further by comparing the estimated Sub-SAGE values from the tree ensemble model with the exact Sub-SAGE value from the true model in [19], as well as by comparing with the exact SAGE value. The results are given in Table 2 and the details of the computations in S1 File. **Table 2** | Feature | Sub-SAGE (TM) | Sub-SAGE, (XGB) | SAGE (TM) | | --- | --- | --- | --- | | 6.0 | 42.37 | 41.83 | 42.64 | | 1.0 | 0.071 | 0.057 | 0.072 | | 2.0 | 0.017 | 0.00073 | 0.018 | | 12.0 | 0.0 | -0.0133 | 0.0 | The comparison shows that the XGBoost model has captured almost all of the relationship between the response and feature 6, while the influence of feature 1, and particularly feature 2 has been underestimated. It also shows the small difference between the true Sub-SAGE and SAGE value with respect to the model in [19], and therefore the small gain of computing the SAGE value in this particular case. However, as the true relationship in this case is restricted to pairwise interactions, the insignificant difference between SAGE and Sub-SAGE cannot in general be anticipated for instance when the true relationship includes higher-order interactions. ## Application on genetic data using the UK Biobank resource To demonstrate the ability of Sub-SAGE on observed data, we consider a realistic machine learning problem using both genetic and non-genetic data of moderate size from UK Biobank. The aim is to infer the influence of specific features with respect to obesity (BMI ≥ 30), by training an XGBoost model and computing Sub-SAGE values. We treat this as a classification problem between the categories obese and non-obese (see [25] for details). Of particular interest is whether any genetic markers are important. The most used method in this setting is a so-called genome-wide association study (GWAS), where each genetic variant is tested individually in a general linear (mixed-effects) regression model [24, 29]. A corresponding p-value of less than 5 × 10−8 is often considered statistically significant. This tiny significance level is chosen due to the multiple comparison problem [30]. When the same association is replicated in an independent data set, the association is considered to be robust. We study a particular XGBoost model constructed in [25]. The model was trained to predict the probability of an individual being obese, p(Yi = 1|xi), given genetic and non-genetic data, xi, such that logit(p(Yi=1|xi))=∑τ=1Tfτ(xi), consisting of T regression trees. *The* genetic data consists of so-called minor allele counts or genotype values from single nucleotide polymorphisms (SNPs) filtered to limit dependence between the SNPs without significant loss of information [25]. In detail, the particular XGBoost model mentioned above was constructed after the so-called ranking process explained in [25], which ranks features by importance and filter for correlation. Using a sample of 207 015 individuals, a total of 529 024 SNPs were split into 50 randomly selected subsets, each consisting of SNPs with mutually small correlation (Pearson’s correlation r2 < 0.2) and unrelated individuals, with the purpose of limiting the correlation between features due to linkage disequilibrium [31], as well as reducing the effect of population stratification and cryptic relatedness [32]. For each such subset, XGBoost models in combination with cross-validation were fitted, and the importance of each SNP, taking into account all the generated models, was measured according to the model-agnostic ERFC score introduced in [25] based on SHAP values. The ranked list of features by importance was again filtered for correlation (Pearson’s correlation r2 < 0.2), and used during the so-called model fitting process based on data not used during the ranking process (see Figure 3 in [25]). The aim in the model fitting process is to find how large the portion of top-ranked features in the training data must be to get the strongest prediction model, using the PR-AUC metric [33], restricted to an ensemble model consisting of XGBoost models constructed via cross-validation (see Figure 9 in [25]). From the best-performing ensemble model, and for simplicity restricted to regression trees of maximum depth equal to two, we picked one of the XGBoost models from this ensemble model as the particular model to investigate further in this paper. Non-genetic features included during the process are sex, age, physical activity frequency, intake of saturated fate, sleep duration, stress and alcohol consumption. Not surprisingly, these non-genetic features are considered most important, and therefore also included in our particular XGBoost model to investigate. During the model fitting process, the hyperparameters were optimized within a restricted region of the hyperparameter space (given in Table 4 in [25]). When restricting to regression trees of maximum depth equal to two, the best performing ensemble model resulted in the following hyperparameter values: Learning rate η = 0.05, colsample = subsample = colsample_by_tree = 0.8, max_depth = 2, λ = 1, γ = 1, early_stopping_rounds = 20, with binary cross-entropy loss. The particular XGBoost model to be investigated is constructed based on 64 000 individuals from UK Biobank, and includes 532 features spread along a total of 607 trees. Computing SAGE values for all features would be very time-consuming, if not infeasible. Other feature importance scores that are faster to compute, yet less trustworthy, such as SHAP or ERFC must typically be used instead when pre-evaluating the importance of each feature. The features with the largest ERFC-scores based on the training data are given in Table 3. We consider those to be the most relevant for further investigation. **Table 3** | Feature | ERFC | | --- | --- | | Alcohol intake frequency | 0.088 | | Genetic sex | 0.086 | | Physical activity frequency | 0.073 | | Intake of saturated fat | 0.044 | | Sleep duration | 0.036 | | Stress | 0.034 | | Age at recruitment | 0.033 | | rs17817449 | 0.017 | | rs489693 | 0.012 | | rs1488830 | 0.011 | | rs13393304 | 0.01 | | rs10913469 | 0.01 | | rs2820312 | 0.0086 | Before we compute the Sub-SAGE values, we check each feature in the context of domain knowledge. While the non-genetic features are considered the most important, the most important SNP according to the model is rs17817449. This SNP is connected to the FTO gene at chromosome 16, and has previously been (statistically significant) associated with obesity in a large number of genome-wide association studies including different independent data sets [21]. The SNP rs13393304 at chromosome 2 has previously been associated with obesity using UK *Biobank data* [34]. From the PheWeb platform [35], a generalized linear mixed model [24], based on TopMed imputation on each individual [36], was constructed separately on each trait out of a total of 1419 traits in UK Biobank. In this case, the SNP rs489693 is second most associated with obesity, yet not statistically significant with p-value = 2.3 × 10−7. Likewise, for the SNPs rs1488830, rs10913469 and rs2820312 the computed p-values are 2.2 × 10−3, 7.1 × 10−5 and 1.1 × 10−2 respectively, and therefore not declared statistically significant with respect to obesity. However, the association between the SNP rs2820312 and hypertension is in fact statistically significant (2.5 × 10−9) in the PheWeb platform, and obesity is known to be a risk factor for hypertension [37]. The uncertainty of the feature importance of the SNPs rs17817449, rs13393304 and rs2820312 in Table 3 are explored more thoroughly by computing Sub-SAGE estimates including paired bootstrap-derived percentile intervals, with $B = 1000$ bootstrap samples, by using 20000 (unrelated White-British) participants from UK Biobank not used while training the model. We also compute Sub-SAGE for the randomly selected SNP rs7318381, which has never been associated with obesity, and with a small ERFC in the XGBoost model (0.0016). The results are given in Fig 4. **Fig 4:** *The estimates and corresponding uncertainties in the Sub-SAGE values for the four SNPs agree with previous studies (GWAS) regarding SNP-association with obesity.* The Sub-SAGE values do indicate that both rs17817449 and rs13393304 are highly likely to be associated with obesity. The $95\%$ percentile interval of the Sub-SAGE value for rs17817449 is (0.0006, 0.0016), and (0.00014, 0.00073) for rs13393304. The SNPs rs2820312 and rs7318381 are less likely to be associated with obesity, and if they are true associations, the uncertainties in the estimates indicate that the effects are microscopic. The $95\%$ percentile intervals for rs2820312 is (−7.08 × 10−5, 2.95 × 10−4), and (−1.13 × 10−5, 6.32 × 10−5) for rs7318381. When dealing with relatively large data sizes such as for the genetic example above, the bias-corrected and accelerated bootstrap interval can become infeasible due to the estimation of the acceleration parameter. However, as the acceleration parameter is proportional to the skewness of the bootstrap distribution, and if the bootstrap distribution indeed has a small skewness, as is the case here, it is often sufficient to set the acceleration parameter equal to zero. This gives no change in the percentile intervals of rs17817449 and rs13393304, but the bias-corrected $95\%$ bootstrap intervals of rs2820312 and rs7318381 become (−6.10 × 10−5, 3.0 × 10−4) and (−1.18 × 10−5, 6.19 × 10−5) respectively. These are negligible changes, indicating that the plug-in estimates are low-biased. With the number of individuals and size of the model explained above the provided R and Rcpp code performs a single Sub-SAGE estimate in around 15–20 minutes using CPUs available at the Farnam Cluster from Yale Center for Research Computing. The bootstrap samples were accomplished using job arrays in a high-performance computing environment, and were completed within around 1.5 hours for each feature. ## Discussion and conclusion We present a Shapley-value-based framework for inferring the importance of individual features, including uncertainty in the estimator. We argue that SAGE values, or Sub-SAGE values, are more appropriate for quantifying global feature importance than SHAP values, as SHAP values only depend on the fitted model itself, good or bad, while SAGE and Sub-SAGE values additionally account for the performance with respect to the true data generating process via the loss function. Effectively, using SAGE and Sub-SAGE for inferring feature importance reduces the false positive rate compared to when using SHAP. As the computation of SAGE values quickly becomes challenging for increasing number of features, we introduce the Sub-SAGE value as an appropriate alternative. We demonstrate how to infer feature influence for a tree ensemble model with high-dimensional data using Sub-SAGE and paired bootstrapping. As an example, we use XGBoost, a gradient tree-boosting model, applied to both a known data generating process, as well as realistic high-dimensional data. We emphasize the importance of using test data, independent of data used to construct the model, to compute Sub-SAGE estimates. The particular choice of Qk in the definition of Sub-SAGE was based on the fact that marginal effects and pairwise interaction effects are accounted for. An alternative is to include all subsets with cardinality restricted to some value. Yet another approach is to sample a restricted number of subsets S following the same probability distribution as for the Kernel SHAP method, see [3] for details. The main idea is that the Shapley consistency property is not a necessity if the question is whether a feature k is regarded as important with respect to a particular prediction model. It is important to notice that the percentile intervals, constructed to evaluate the uncertainty in the Sub-SAGE estimate, themselves include uncertainty. The uncertainty of the percentile intervals depends on the number of bootstraps, B, as well as the size n of data samples. However, in addition, the uncertainty also depends on the ratio p/n, where p is the total number of features used in the model (not necessarily the number of input features for constructing the model). This fact is particularly important in high-dimensional problems, and it has been discussed for instance in [38]. When applied to linear models, one observation from a simulation is for instance that the paired bootstrap becomes more conservative (loss of power) the larger the ratio p/n is. Observe that for the simulation example above, p/$$n = 62$$/3200 = 0.019, while for the genetic data, the ratio is p/$$n = 533$$/20000 = 0.027, deliberately chosen to be small in order to account for the problems arising when p/n becomes too large. For the genetic data, a filtering process is first needed as the data from UK Biobank originally includes around 530000 SNPs and 207000 individuals (p/$$n = 2$.56$). The applied filtering method and potential pitfalls are described in [25]. It seems reasonable to apply the same loss function in the Sub-SAGE estimate as the loss function that was used to construct the model. However, there may be situations where it is meaningful to compute the Sub-SAGE values for a different loss function than the loss function used during training in order to make more objective interpretations. This may for instance be the case when the model is provided ‘as is’, and you do not know the training loss function, or when using adapted loss functions, e.g. weighted binary cross-entropy, but the interpretation is relevant for a standard cross-entropy. If we use Sub-SAGE to infer importance of a feature, let the null hypothesis be that the corresponding Sub-SAGE value is less than or equal to zero. A simple procedure to investigate this is to construct a one-sided (1 − α)$100\%$ percentile interval, and reject the null hypothesis if the corresponding lower bound is greater than zero. However, the use of a bootstrap confidence interval to construct a hypothesis test often has a low statistical power [39]. If in addition several features are tested simultaneously, a multiple testing procedure would be necessary in order to control the false positive rate [30]. We leave it to future research as to how to construct a more powerful hypothesis testing procedure, and how to control the false positive rate in a multiple testing procedure. The statistical power when inferring feature importance based on Sub-SAGE will rely on the model uncertainty, the degree to which the prediction model has captured the true relationship between a particular feature and the response. In this work we have assumed all features to be mutually statistically independent, an unrealistic scenario in most cases. If many features are statistically dependent, one is required to estimate conditional expected values (see [3] for details). Even for medium-size data sets this often becomes very tedious and even infeasible in most cases. One possibility is to use principal component analysis for dimensionality reduction, but this is not straightforward if we need the features of the model to be meaningful, and thereby explainable. In addition, principal component analysis is based on variance in the features and not explanatory power. An important line of future research to allow for evaluation of feature importance in a high-dimensional setting is dimensionality reduction that preserves interpretability. The estimates provided by Sub-SAGE, as for SHAP values, will be more reliable the better the overall predictions from the model. Recent research has shown the strong benefit of including individual polygenic risk scores [40] as a covariate in the XGBoost model for greater performance in predictions of susceptibility for several phenotypes [41]. Computation of Sub-SAGE values in this case would depend on the correlation between PRS and individuals SNPs used in the XGBoost model, and if so the need for estimating conditional expected values. In this paper, we have focused on the marginal effect of each feature, which is the total effect of the feature including the isolated effect of the feature as well as possible interaction effects the feature may be involved in. In principle, the construction of SHAP interaction values to quantify pairwise interaction effects, as introduced in [4] and applied in [25], could be extended to also include SAGE or Sub-SAGE interaction values. In this setting, the running time is even more computationally demanding. In that respect, the idea of reducing the number of subsets to include in the Shapley-based computations can be one approach to reduce the computation time. Interesting future research would be how to apply the idea of Sub-SAGE to search for interaction effects. ## References 1. Jiménez-Luna J, Grisoni F, Schneider G. **Drug discovery with explainable artificial intelligence**. *Nature Machine Intelligence* (2020.0) **2** 573-584. DOI: 10.1038/s42256-020-00236-4 2. Shapley LS. **A Value for n-Person Games**. *Contributions to the Theory of Games (AM-28)* (1953.0) **Volume II** 3. Aas K, Jullum M, Løland A. **Explaining individual predictions when features are dependent: More accurate approximations to Shapley values**. *Artificial Intelligence* (2021.0) **298**. DOI: 10.1016/j.artint.2021.103502 4. Lundberg SM, Erion G, Chen H, DeGrave A, Prutkin JM, Nair B. **From local explanations to global understanding with explainable AI for trees**. *Nature Machine Intelligence* (2020.0) **2**. DOI: 10.1038/s42256-019-0138-9 5. Sellereite N, Jullum M. **shapr: An R-package for explaining machine learning models with dependence-aware Shapley values**. *Journal of Open Source Software* (2019.0) **5** 2027. DOI: 10.21105/joss.02027 6. Lundberg SM, Lee SI, Guyon I, Luxburg UV, Bengio S, Wallach H, Fergus R, Vishwanathan S. *Advances in Neural Information Processing Systems 30* (2017.0) 4765-4774 7. Strumbelj E, Kononenko I. **Explaining prediction models and individual predictions with feature contributions**. *Knowledge and Information Systems* (2013.0) **41** 647-665. DOI: 10.1007/s10115-013-0679-x 8. Strumbelj E, Kononenko I. **An Efficient Explanation of Individual Classifications using Game Theory**. *Journal of Machine Learning Research* (2010.0) **11** 1-18 9. Lundberg SM, Erion GG, Lee SI. *Consistent Individualized Feature Attribution for Tree Ensembles* (2019.0) 10. Redelmeier A, Jullum M, Aas K. *Explaining predictive models with mixed features using Shapley values and conditional inference trees* (2020.0) 11. Kwon Y, Rivas MA, Zou J. *Efficient computation and analysis of distributional Shapley values* (2021.0) 12. Song E, Nelson B, Staum J. **Shapley Effects for Global Sensitivity Analysis: Theory and Computation**. *SIAM/ASA Journal on Uncertainty Quantification* (2016.0) **4** 1060-1083. DOI: 10.1137/15M1048070 13. Moehle N, Boyd S, Ang A. *Portfolio Performance Attribution via Shapley Value* (2021.0) 14. Covert I, Lundberg S, Lee SI. *Explaining by Removing: A Unified Framework for Model Explanation* (2020.0) 15. Keinan A, Hilgetag CC, Meilijson I, Ruppin E. **Fair Attribution of Functional Contribution in Artificial and Biological Networks**. *Neural Computation* (2003.0) **16** 1887-1915. DOI: 10.1162/0899766041336387 16. Fryer DV, Strumke I, Nguyen H. **Model independent feature attributions: Shapley values that uncover non-linear dependencies**. *PeerJ Computer Science* (2021.0) **7** e582. DOI: 10.7717/peerj-cs.582 17. Covert I, Lundberg S, Lee SI. *Understanding Global Feature Contributions With Additive Importance Measures* (2020.0) 18. Fryer D, Strümke I, Nguyen H. *Shapley values for feature selection: The good, the bad, and the axioms* (2021.0) 19. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J. **UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age**. *PLoS medicine* (2015.0) **12** e1001779. DOI: 10.1371/journal.pmed.1001779 20. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K. **The UK Biobank resource with deep phenotyping and genomic data**. *Nature* (2018.0) **562** 203-209. DOI: 10.1038/s41586-018-0579-z 21. Locke AE, Kahali B, Berndt SI. **Genetic studies of body mass index yield new insights for obesity biology**. *Nature* (2015.0) **518** 197-206. DOI: 10.1038/nature14177 22. Young HP. **Monotonic solutions of cooperative games**. *International Journal of Game Theory* (1985.0) **14** 65-72. DOI: 10.1007/BF01769885 23. Huettner F, Sunder M. **Axiomatic arguments for decomposiing goodness of fit according to Shapley and Owen values**. *Electronic Journal of Statistics* (2012.0) **6** 1239-1250. DOI: 10.1214/12-EJS710 24. Zhou W, Nielsen JB, Fritsche LG, Dey R, Gabrielsen ME, Wolford BN. **Efficiently controlling for case-control imbalance and sample relatedness in large-scale genetic association studies**. *Nature Genetics* (2018.0) **50**. DOI: 10.1038/s41588-018-0184-y 25. Johnsen PV, Riemer-Sørensen S, DeWan AT, Cahill ME, Langaas M. **A new method for exploring gene–gene and gene–environment interactions in GWAS with tree ensemble methods and SHAP values**. *BMC Bioinformatics* (2021.0) **22**. DOI: 10.1186/s12859-021-04041-7 26. Hastie T, Tibshirani R, Friedman J. *The Elements of Statistical Learning* (2009.0) 223-228 27. Efron B. **Better Bootstrap Confidence Intervals**. *Journal of the American Statistical Association* (1987.0) **82** 171-185. DOI: 10.2307/2289153 28. 28Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining—KDD’16. 2016; p. 785–794. 29. Visscher PM, Wray NR, Zhang Q, Sklar P, McCarthy MI, Brown MA. **10 Years of GWAS Discovery: Biology, Function, and Translation**. *American Journal of Human Genetics* (2017.0) **101** 5-22. DOI: 10.1016/j.ajhg.2017.06.005 30. Goeman JJ, Solari A. **Multiple hypothesis testing in genomics**. *Statistics in Medicine* (2014.0) **33** 1946-1978. DOI: 10.1002/sim.6082 31. Reich DE, Cargill M, Bolk S, Ireland J, Sabeti PC, Richter DJ. **Linkage disequilibrium in the human genome**. *Nature* (2001.0) **411** 199-204. DOI: 10.1038/35075590 32. Sillanpää MJ. **Overview of techniques to account for confounding due to population stratification and cryptic relatedness in genomic data association analyses**. *Heredity* (2011.0) **106** 511-519. DOI: 10.1038/hdy.2010.91 33. Grau J, Grosse I, Keilwagen J. **PRROC: computing and visualizing precision-recall and receiver operating characteristic curves in R**. *Bioinformatics* (2015.0) **31** 2595-2597. DOI: 10.1093/bioinformatics/btv153 34. Karlsson T, Rask-Andersen M, Pan G, Höglund J, Wadelius C, Ek WE. **Contribution of genetics to visceral adiposity and its relation to cardiovascular and metabolic disease**. *Nature medicine* (2019.0) **25** 1390-1395. DOI: 10.1038/s41591-019-0563-7 35. Gagliano Taliun SA, VandeHaar P, Boughton AP, Welch RP, Taliun D, Schmidt EM. **Exploring and visualizing large-scale genetic associations by using PheWeb**. *Nature Genetics* (2020.0) **52**. DOI: 10.1038/s41588-020-0622-5 36. Taliun D, Harris DN, Kessler MD. **Sequencing of 53,831 diverse genomes from the NHLBI TOPMed Program**. *Nature* (2021.0) **590** 290-299. DOI: 10.1038/s41586-021-03205-y 37. Hall JE, do Carmo JM, da Silva AA, Wang Z, Hall ME. **Obesity-Induced Hypertension**. *Circulation Research* (2015.0) **116** 991-1006. PMID: 25767285 38. Karoui NE, Purdom E. **Can We Trust the Bootstrap in High-dimensions? The Case of Linear Models**. *Journal of Machine Learning Research* (2018.0) **19** 66 39. Givens GH, Hoeting JA. *Computational Statistics* (2012.0) 287-321 40. Torkamani A, Wineinger NE, Topol EJ. **The personal and clinical utility of polygenic risk scores**. *Nature Reviews Genetics* (2018.0) **19** 581-590. DOI: 10.1038/s41576-018-0018-x 41. Elgart M, Lyons G, Romero-Brufau S, Kurniansyah N, Brody JA, Guo X. **Non-linear machine learning models incorporating SNPs and PRS improve polygenic prediction in diverse human populations**. *Communications Biology* (2022.0) **5** 1-12. DOI: 10.1038/s42003-022-03812-z
--- title: 'Anemia prevalence and associated factors among school-children of Kersa Woreda in eastern Ethiopia: A cross-sectional study' authors: - Kabtamu Gemechu - Haftu Asmerom - Lealem Gedefaw - Mesay Arkew - Tilahun Bete - Wondimagegn Adissu journal: PLOS ONE year: 2023 pmcid: PMC10038290 doi: 10.1371/journal.pone.0283421 license: CC BY 4.0 --- # Anemia prevalence and associated factors among school-children of Kersa Woreda in eastern Ethiopia: A cross-sectional study ## Abstract ### Background Anemia in school children is a worldwide public health problem, affecting about a quarter of this population. It also remains a significant problem in developing countries, with multifactorial causes. Anemia in school children has adverse effects on the development of the physical, cognitive, immunity of affected children, and subsequently their educational achievement which may lead to loss of productivity at a later age in life. Regular surveillance that could provide evidence-based local data is required to intervene in the problems. Therefore, this study aimed to determine the prevalence and associated factors of anemia among school children in primary schools of eastern Ethiopia. ### Methods School-based cross-sectional study was conducted by recruiting 482 school- children. Data on socio-demographic and dietary habits were collected from parents/legal guardians. Capillary blood for blood film preparation and hemoglobin measurement and stool sample for the diagnosis of intestinal parasites infection was collected. Hemoglobin concentration was measured using a hemoglobinometer HemoCue® 301+, and stool examination by direct wet mount and concentration technique. Data were entered into epi-data and exported into SPSS for analysis. Bivariate and multivariate logistic regression was run to identify associated factors. Association was described using adjusted OR (AOR) along with $95\%$ CI and variables with a p-value<0.05 were considered statistically significant. ### Results The overall prevalence of anemia was $24.5\%$. Being female (AOR = 2.88, $95\%$ CI: 1.69, 4.92), family size of more than 5 (AOR = 2.78, $95\%$ CI: 1.60, 4.81), not consuming green leafy vegetables (AOR = 4.09, $95\%$ CI: 2.42, 6.94), consumption of milk (AOR = 2.22, $95\%$ CI: 1.27, 3.88), being stunting (AOR = 3.17, $95\%$ CI: 1.70, 5.91) and parasite infections (AOR = 5.23, $95\%$ CI: 2.77, 9.85) were significantly associated with anemia. ### Conclusion In this study nearly one-fourth of children were anemic. Anemia was a moderate public health problem among schoolchildren in the study area. Thus, school-based interventions targeting nutritional factors and intestinal parasite infection need to be implemented. ## Introduction Anemia is a significant worldwide public health problem affecting both developed and developing nations [1]. Globally, it affects about 1.74 billion ($22.8\%$) of the world’s population [2] of which 305 million ($25.4\%$) are school-age children. In Ethiopia, anemia remains a mild to severe public health problem with prevalence ranging from $7.6\%$ [3] to $43.7\%$ [4]. The finding of a Nationally conducted systematic review and meta-analysis indicated $23\%$ of schoolchildren were anemic [5]. Anemia occurs at all ages, however; reproductive-age women, preschool, and school-age children are affected more [6]. School children are one of the most vulnerable groups because they are at the age where physical growth and physiological change are fast enough which leads to the high demand for nutritional requirements [7]. Moreover, they are more vulnerable to intestinal parasite infections which are a major contributor to anemia [8,9]. Anemia has significant adverse health consequences and impacts on social and economic development. Globally, it causes 68.36 and 58.6 million years lived with a disability in 2010 [10] and 2019 [2] respectively. Anemia harms the physical, and cognitive development of affected children and subsequently their educational achievement which may lead to loss of productivity at a later age in life [11]. It also affects the function of immune systems and consequently increases the risk of infections by impacting both humoral and cellular immunity [12,13]. There are multiple and complex contributing factors to anemia including nutritional deficiency (iron, folate, and vitamin B12), genetic hemoglobin disorder, intestinal parasitic infections, and acute and chronic infections [14]. Pathophysiology of anemia is diverse based on its underlying etiologies which may attribute to decreased production from micronutrient deficiency because of poor dietary diversity [15] and increased destruction and/or blood loss due to parasite infections [16]. Generally, anemia is not a disease, but instead the manifestation of an underlying pathological process that occurs due to multiple contributing factors. Given the multifactorial nature, control, and prevention of anemia require an integrated approach based on the identification and addressing of specific risk factors. This requires regular surveillance that could provide evidence-based local data to policymakers to design plans and policies to intervene in the problems. Although the previous study in the local area assessed the magnitude of anemia with association to other factors, intestinal parasites, and nutritional status were not considered and the morphological type of anemia was not assessed. Therefore, this study aimed to assess the current magnitude, morphological type, and associated factors of anemia among school- children in Kersa Woreda primary school, eastern Ethiopia. ## Study setting and population A school-based cross-sectional study was conducted among primary school children in Kersa woreda, eastern Ethiopia, from December 7, 2020, to January 7, 2021. Kersa *Woreda is* located 478 km to the east of Addis Ababa and 42 km to the west of Harar city. The *Woreda is* located at an altitude ranging from 1400 to 3200 m above sea level. There are 38 kebeles, of which 35 are rural and 3 are town kebeles. According to the 2007 population and housing census of Ethiopia, the projected total population of the Kersa district was 172,626, of whom 87,029 were men and 85,597 were women [17]. According to the Office of Education of Kersa Woreda, there are a total of 101 primary schools, of which 94 and 7 are public and private, respectively. Out of 101, 27 schools have all grade levels from 1–8. There are approximately 45597 primary school students. Primary school children aged 6–17 who were willing and whose parents gave consent were included in the study. On the other hand, children who had been receiving hematinic factors such as iron, vitamin B12, and foliate for three months and anti-helminthic medication two weeks before the study were excluded. ## Sample size determination and sampling technique A total of 500 sample sizes was calculated using single population proportion statistical formula n = (Zα /2)2 p (1-p)/d2 by considering assumptions: 0.27 P (proportion) [18], $95\%$ level of confidence, $5\%$ margin of error, 1.5 design effect, and $10\%$ non-response. The study participants were chosen using a multi-stage sampling method. First, 27 primary schools with grade levels 1 through 8 were found, and $30\%$ of those were included. Then, eight primary schools were chosen using simple random sampling. The K-value was computed by dividing the total number of the study population by the entire sample size ($K = 9651$/500 = 19). The total sample size was proportionally allocated among the selected schools based on the number of students. The proportionally allocated sample size was further distributed to grade level considering the number of students (Fig 1). Finally, a systematic random sampling technique was used to select study participants, using each school student’s registration detail for an academic year as a sampling frame. **Fig 1:** *Schematic diagram showing sampling technique of school-children attending primary school in Kersa woreda, eastern Ethiopia, 2020/2021.* ## Socio-demographic and related data Schools were sampled and children were selected at school. Then, children’s parents/ legal guardians were contacted by Health Extension workers, and data on socio-demographic, and dietary habits were collected through face-to-face interviews using a semi-structured questionnaire that is extracted from different literature. Dietary factors were collected and assessed by using a modified version of the Helen Keller International Food Frequency Questionnaire (FFQ) that was used previously in Ethiopia [19]. Consumption of commonly consumed food items in the area was grouped into five as vegetables, fruit, meat, dairy product, and coffee/ tea was assessed. ## Anthropometric measurement Anthropometric measurements were taken according to WHO guidelines for anthropometric measurements [20]. Weight was measured to the nearest 0.1 kg using a portable digital weight measuring scale. Each child was weighed with light clothing and barefoot. The weighing scale was calibrated using the standard calibration weight of 2kg iron bars. Height was measured in the Frankfurt position using a locally manufactured stadiometer with a sliding head bar. All measurements were performed in duplicate and the average value was used for anthropometry data analysis. Then anthropometric measurements were converted into Height-for-Age Z scores (HAZ) and Body Mass Index for-Age Z scores (BAZ) using WHO Anthro Plus version 1.0.4. ## Sample collection, processing, and analysis Capillary blood samples for hemoglobin measurement and peripheral blood film preparation were collected from each study participant by finger prick using a sterile single-use disposable lancet. Hemoglobin concentration was measured using a portable digital hemoglobinometer (HemoCue®301+, Angel Holm, Sweden). After correcting for altitude, adjusted Hb concentration was used to define anemia. Blood films were prepared, stained with wright stain, and examined microscopically for evaluation of red cell morphology. Approximately 2 grams of a stool sample for parasite examination were collected following the standard procedures for stool sample collection. Collected samples were preserved using $10\%$ formalin and intestinal parasites were examined by both direct wet mount and formol-ether concentration techniques. ## Data quality assurance and quality control To assure the quality of data, all phases of quality assurance were maintained. The questionnaire prepared in English was translated into the local language, Afaan Oromo. Pre-tested was conducted on 25 ($5\%$) of the total sample size. The training was given to data collectors and supervisors for two consecutive days on the objective of the study, confidentiality of information, and the data collection process. Standard operating procedures for hemoglobin measurement, blood film preparation, staining [23], and formol-ether concentration technique [24] were strictly followed. All laboratory reagents were checked for their expiry date before use. The accuracy of HemoCue and micro cuvettes were checked by comparing Hb values with those measured on an automated hematology analyzer (UniCel DxH800 Beckman Coulter, USA) on 20 samples of patients at Hiwot Fana Specialized University Hospital. The quality of the Wright stain was checked on thin blood film stained at different staining times by comparing the staining characteristics of RBC and WBC. Laboratory results were properly recorded on report formats using participants’ identification numbers. ## Data processing, analysis, and interpretation All data were checked for completeness and coded. Data entry was done using Epi Data version 3.1 and exported into SPSS version 20 for analysis. Descriptive statistics were used to describe study variables. Both bivariable and multivariable binary logistic regression were computed to identify associated factors. Variables with a p-value <0.25 in bivariate analysis were considered a candidate for multivariable analysis. Multi-co-linearity was checked using the variance inflation factor (VIF) test and all candidate variables were included in the final model of multivariable analysis. The model goodness of fit was tested using the Hosmer-Lemeshow goodness of fit test ($$p \leq 0.46$$) Adjusted odds ratio (AOR) with the corresponding $95\%$ confidence interval (CI) was used to indicate the strength of the association and a variable with a P-value <0.05 was considered statistically significant. ## Ethical consideration Ethical clearance was obtained from Jimma University Institutional Review Board (Ref. No: IRB$\frac{00937}{2020}$). A letter of support was submitted to Kersa Woreda Health Bureau and Education office to obtain official permission and obtained official permission was submitted to each school director. Informed written consent from the parents/legal guardians of children and oral assent from children below the age of 18 years were obtained after describing the benefits and risks of the study. The following measures were taken to ensure confidentiality: Participant names and other identifiers were not used; only participant codes were used during data collection, entry, and analysis on the computer. The data was only accessible to authorized people, including the data collector and investigator. To protect their privacy, the children’s parents were also interviewed separately. All collected data were exclusively utilized for the study. Children who tested anemic and positive for intestinal parasites were connected to nearby health institutions for proper treatment. ## Socio-demographic characteristics Out of the total sample size [500] school children recruited for the study, 482 were enrolled in this study with a response rate of $96.4\%$, and 7 children took anti-parasite medication in the last two weeks, 3 children with hematin factors treatment in the last three months before the study and 8 children who were not volunteer to give stool and blood sample were excluded. The age of the study participant ranges from 6 to 17 years, with a median age of 10 years. More than half, $52.7\%$ ($$n = 254$$) of study participants were male. About $57.7\%$ ($$n = 278$$) of study participants were in the grade level of 1–4 and around $61\%$ ($$n = 294$$) were rural residents. Regarding the occupational status of parents, the majority, $63.1\%$ ($$n = 304$$) of children’s fathers were farmers, and most, $73\%$ ($$n = 352$$) of the mother were housewives, whereas about $44.8\%$ ($$n = 216$$) of child fathers and majority, $62.7\%$ ($$n = 302$$) mothers are unable to read and write. About $46.7\%$ ($$n = 225$$) of children’s parents have monthly household incomes <500 Ethiopian Birr and nearly more than half have a family size of more than five (Table 1). **Table 1** | Variable | Category | Frequency | Percentage | | --- | --- | --- | --- | | Age (in years) | 6–9 | 207 | 42.9 | | Age (in years) | 10–14 | 219 | 45.4 | | Age (in years) | 15–17 | 56 | 11.6 | | Sex | Male | 254 | 52.7 | | Sex | Female | 228 | 47.3 | | Students Grade Level | 1–4 | 278 | 57.7 | | Students Grade Level | 5–8 | 204 | 42.3 | | Residence | Rural | 294 | 61.0 | | Residence | Urban | 188 | 39.0 | | Father’s occupation | Farmer | 304 | 63.1 | | Father’s occupation | Merchant | 35 | 7.3 | | Father’s occupation | Private employee | 33 | 6.8 | | Father’s occupation | Government employee | 110 | 22.8 | | Mothers Occupation | Housewife | 352 | 73.0 | | Mothers Occupation | Merchant | 51 | 10.6 | | Mothers Occupation | Private employee | 17 | 3.5 | | Mothers Occupation | Government employee | 62 | 12.9 | | Fathers educational status | unable to read and write | 216 | 44.8 | | Fathers educational status | Primary education | 104 | 21.6 | | Fathers educational status | Secondary and above | 162 | 33.6 | | Mothers educational status | unable to read and write | 302 | 62.7 | | Mothers educational status | Primary education | 76 | 15.8 | | Mothers educational status | Secondary and above | 104 | 21.6 | | Average monthly income in Ethiopia birr | <500 | 225 | 46.7 | | Average monthly income in Ethiopia birr | 500–1999 | 66 | 17.8 | | Average monthly income in Ethiopia birr | ≥2000 | 171 | 35.5 | | Family Size | ≤5 | 233 | 48.3 | | Family Size | >5 | 249 | 51.7 | ## Dietary factors Food consumption is classified based on reviewed previous literature [25–27]. About one-third, $33.2\%$ ($$n = 160$$) of study participants responded not to have access to green leafy vegetables, while about $47.8\%$ ($$n = 154$$) of consumers had access more than once a week. Of the total study participants, nearly more than half, $55.2\%$ ($$n = 266$$) of children consumed milk, and the majority, $64.5\%$ ($$n = 311$$) of them had a habit to take coffee/tea after the meal (Table 2). **Table 2** | Variable | Category | Frequency | Percentage | | --- | --- | --- | --- | | Consumption of green leafy vegetables | Yes | 322 | 66.8 | | Consumption of green leafy vegetables | No | 160 | 33.2 | | Frequency of green leafy vegetable consumption | Daily | 93 | 28.9 | | Frequency of green leafy vegetable consumption | Once a week | 75 | 23.3 | | Frequency of green leafy vegetable consumption | More than once a week | 154 | 47.8 | | Consumption of citrus fruit | Yes | 195 | 40.5 | | Consumption of citrus fruit | No | 287 | 59.5 | | Frequency of citrus fruit consumption | Daily | 8 | 4.1 | | Frequency of citrus fruit consumption | Once a week | 168 | 86.2 | | Frequency of citrus fruit consumption | More than once a week | 19 | 9.7 | | Consumption of red meat | Yes | 197 | 40.9 | | Consumption of red meat | No | 285 | 59.1 | | Frequency of red meat consumption | Once a week | 161 | 81.7 | | Frequency of red meat consumption | More than once a week | 36 | 18.3 | | Consumption of milk | Yes | 266 | 55.2 | | Consumption of milk | No | 216 | 44.8 | | Frequency of Milk consumption | Daily | 58 | 21.8 | | Frequency of Milk consumption | More than once a week | 136 | 51.1 | | Frequency of Milk consumption | Once a week | 72 | 27.1 | | Consumption of coffee/tea after a meal | Yes | 311 | 64.5 | | Consumption of coffee/tea after a meal | No | 171 | 35.5 | ## Nutritional status and clinical factors Of a total of study participants, about $16\%$ ($$n = 77$$) were stunted for their age (HAZ <- 2 SD) and $11.8\%$ ($$n = 57$$) were thin for their age (BAZ<-2 SD). Among study participants examined for intestinal parasite infection, $16.2\%$ ($$n = 78$$) were infected with at least one intestinal parasite. A total of six species of intestinal parasites were identified. Schistosoma mansoni 28($16.2\%$), *Giardia lamblia* 22 ($4.6\%$), and *Haymenolepsis nana* 13($2.7\%$) were predominant parasites. In addition to this, two cases with double infection (*Giardia lamblia* and Haymenolepsis nana) and one with triple infection (Ascaris lumbricoid, Schistosoma mansoni, and Giardia lamblia) were also identified (Table 3). **Table 3** | Variable | Frequency | Percentage | | --- | --- | --- | | Stunting (HAZ <-2SD) | | | | Yes | 77.0 | 16 | | No | 405.0 | 84 | | Thinness (BAZ<-2 SD) | | | | Yes | 57.0 | 11.8 | | No | 425.0 | 88.2 | | Intestinal parasite infection | | | | Yes | 78.0 | 16.2 | | No | 404.0 | 83.8 | | Types of Intestinal parasite | | | | S. mansoni | 28.0 | 16.2% | | G.lamblia | 22.0 | 4.6% | | H. nana | 13.0 | 2.7% | | A.lumbaricoids | 4.0 | | | Hookworm | 6.0 | | | E.histolytica/dispar | 5.0 | | ## Prevalence, severity, and types of anemia The mean hemoglobin value of school- children was 13.4±1.83g/dl, ranging from 6.3–17.8 g/dl. The overall prevalence of anemia was $24.5\%$ ($\frac{118}{482}$ $95\%$ CI: 20.6–28.6) and high prevalence was detected in the females and age group of 10–14 years, $\frac{58}{118}$ ($49.2\%$) followed by age 5–9, $\frac{44}{118}$ ($37.3\%$) and 15–17 years, $\frac{16}{118}$ ($13.6\%$). Among those who were anemic, 65 ($55.1\%$) ($95\%$ CI: 45.8–63.6) had mild, 49 ($41.5\%$) ($95\%$ CI: 33.1–50.8) had moderate anemia, and four cases of severe anemia were also identified. Regarding types of anemia, the examined blood film showed 68 ($57.6\%$) microcytic hypochromic, 28 ($23.7\%$) normocytic normochromic, and 22 ($18.6\%$) macrocytic normochromic cells. ## Factors associated with anemia Independent variables including Sex (being female) (AOR = 2.88 $95\%$ CI: 1.69, 4.92), family size of more than five (AOR = 2.78, $95\%$ CI: 1.60, 4.81), non-consumption of green leafy vegetables (AOR = 4.09, $95\%$ CI: 2.42, 6.94), consumption of milk (AOR = 2.22, $95\%$ CI: 1.27, 3.88), nutritional status (being stunted) (AOR = 3.17, $95\%$ CI: 1.70, 5.91) and being positive for intestinal parasite infection (AOR = 5.23, $95\%$ CI: 2.77, 9.85) were remain as independent predictors of anemia in the final model of multivariate analysis ($p \leq 0.05$) (Table 4). **Table 4** | Variable | Category | Anemia | Anemia.1 | COR (95% CI) | AOR (95% CI) | | --- | --- | --- | --- | --- | --- | | Variable | Category | Anemic | Non-Anemic | COR (95% CI) | AOR (95% CI) | | Sex: | Female | 76 (33.3%) | 152 (66.7%) | 2.52 (1.64,3.88) | 2.88 (1.69, 4.92) ** | | Sex: | Male | 42 (16.5%) | 212 (83.5%) | 1 | 1 | | Mother- | Unable to read and | 90 (29.8%) | 212 (70.2%) | 1.839(0.95, 3.51) | 1.14 (0.58, 2.26) | | Mother- | write Primary | 11(14.5%) | 65 (85.5%) | 0.28 (0.30, 1.76) | 0.34 (0.125, 0.92) | | Mother- | Secondary and above | 17 (16.3%) | 87 (83.7%) | 1 | 1 | | Family size | >5 | 88 (35.3%) | 161 (64.7%) | 3.7 (2.33, 5.88) | 2.78 (1.603, 4.81) ** | | Family size | ≤5 | 30 (12.9%) | 203 (87.1%) | 1 | 1 | | Consumption of green leafy veg. | No | 73 (45.6%) | 87 (54.4%) | 5.17 (3.32,8.04) | 4.09 (2.42, 6.94) ** | | Consumption of green leafy veg. | Yes | 45 (14%) | 277 (86%) | 1 | 1 | | Consumption of citrus fruit | No | 83 (28.9%) | 204 (71.1%) | 1.86 (1.19,2.91) | 1.52 (0.86, 2.70) | | Consumption of citrus fruit | Yes | 35 (18%) | 160 (82%) | 1 | 1 | | Consumption of red meat | No | 87 (30.5%) | 198 (69.5%) | 2.35 (1.49,3.72) | 1.45 (0.74, 2.82) | | Consumption of red meat | Yes | 31 (15.7%) | 166 (84.3%) | 1 | 1 | | Consumption of milk | Yes | 89 (33.5%) | 177 (66.5%) | 3.24 (2.03, 5.17) | 2.22 (1.27, 3.88) * | | Consumption of milk | No | 29 (13.4%) | 187 (86.6%) | 1 | 1 | | Consumption of coffee/ tea | Yes | 87 (28%) | 224 (72%) | 1.75 (1.11,2.78) | 1.40 (0.80, 2.46) | | Consumption of coffee/ tea | No | 31 (18.1%) | 140 (81.9%) | 1 | 1 | | Stunting | Yes | 41 (53.25%) | 36 (46.75%) | 4.85 (2.91,8.09) | 3.17 (1.70, 5.91) ** | | Stunting | No | 77 (19%) | 328 (81%) | 1 | 1 | | Thinness | Yes | 18 (31.6%) | 39 (68.4%) | 1.5 (0.82, 2.74) | 1.762 (0.84, 3.71) | | Thinness | No | 100 (23.5%) | 325 (76.5%) | 1 | 1 | | IPI | Yes | 48 (61.5%) | 30 (38.5%) | 7.67 (4.49,13.08) | 5.23 (2.77, 9.85) ** | | IPI | No | 70 (17.3%) | 334 (82.7%) | 1 | 1 | ## Discussion The overall prevalence of anemia among schoolchildren was $24.5\%$, indicating a moderate public health problem according to the WHO classification of anemia as a problem of public health significance [28]. The prevalence of anemia found in this study is comparable to that seen in Angola, $21.6\%$ [29], Cape Verde, $23.8\%$ [30], and two studies reported in Ethiopia [31,32]. Nonetheless, higher than the results from research conducted in Northeastern Brazil, $9.3\%$ [33], Vietnam, $12.9\%$ [34], Cameron, $5\%$ [35], and other parts of Ethiopia [27,36,37]. This could be due to variations in the geographical area, study setting, socioeconomic status, and age group included in the study. This study primarily concerns children who are attending schools in predominantly rural areas, while studies in Durbete, Bonga, Jimma, and Gondar *Town focus* on those individuals who are found in urban areas. In this study, we considered a broad age range of 6–17 years. This may be accounted for by the high prevalence of anemia in our findings as compared to the lower frequency observed in other studies with comparable designs. In contrast, it is lower than the findings reported from Northwestern Nigeria, $40.3\%$ [38], Egypt, $59.3\%$ [39], and five studies conducted in various sections of Ethiopia [25,26,40–42]. The discrepancy might be due to differences in the study setting, sample size, and epidemiological distribution of parasite infection. This study is schools based whereas a study that was carried out in the Arba Minch area, Pawe, and Jimma town was community-based. Another explanation for this variance in magnitude could be the epidemiological spread of intestinal parasite infection. According to a study from northwest Nigeria, hookworm was the most common parasite and $53\%$ of infected children had anemia [38]. Studies in other parts of Ethiopia also indicated that anemia is substantially associated with intestinal parasites, where the incidence of these parasites ranges from 30 to $46.5\%$, with hookworm being the most common and contributing significantly to anemia [26,40,42]. The prevalence of anemia in this study was significantly associated with being female, family size of more than five, non-consumption of green leafy vegetables, consumption of milk, stunting, and intestinal parasite infection. The likelihood of anemia in females was 2.88 (AOR = 2.88) times higher as compared to males. This finding is in line with the study reported from Egypt [39] and southwest Ethiopia [27]. This is might be related to the combined effect of both rapid physical growth and the occurrence of menarche in adolescent girls, which might make females at the most risk of anemia. In addition to this, a high prevalence of anemia is detected in females and the age group 10–14 years in our finding which might be an early age of menarche and the production of estrogen hormone in females. Studies indicated that the occurrence of menarche significantly increase female’s iron requirement due blood loss [43,44], whereas the estrogen hormone has an antagonist effect on erythropoiesis. This condition with other factors may put females at more risk of anemia. On the other hand, male at the age of puberty starts to produce testosterone hormone having an agonist effect on erythropoiesis and this condition may put them at less risk of anemia. Studies suggested that testosterone increases erythropoiesis via increased erythropoietin [45,46]. The chance of being anemic among children from a family size of more than five was 2.78 (AOR = 2.78) times compared to those with a family size of less than five. A similar finding was reported from a study done in northwest Ethiopia [47], and southwest Ethiopia [27] which showed that children from large family sizes were more likely to be anemic than those from low family sizes. The risk of having anemia was 4.09 (AOR = 4.09) times higher among children who did not consume green leafy vegetables as compared to the consumer. This is consistent with the study conducted in Jimma town [25,42] which reported that children who consume food from plant sources less are more likely to be anemic. This suggested that limited access to green leafy vegetables which are rich in micronutrients may primarily lead to anemia. The green leafy vegetable is a rich source of micronutrients such as iron and vitamins [48]. Unfortunately, the presence of antinutritional factors such as phytate and polyphenol acts to reduce the bioavailability of iron [49]. However, heat cooking has been shown to reduce the level of antinutritional factors in vegetables and increase the bioavailability of iron [50]. The odds of anemia in children who consumed milk were 2.22 (AOR = 2.22) times higher as compared to non-consumer. The possible justification for this might be the inhibitory effect of mineral calcium and protein found in milk on iron absorption in the absence of iron absorption enhancers [51]. There is scientific evidence that suggested mineral calcium and protein found in dairy products affect the absorption of iron in the diet. Calcium inhibits iron absorption by blocking the divalent metal transporter-I (DMT-I) uptake of iron [52,53]. Similarly, Casein protein binds iron with high affinity and forms a complex, and makes iron non-absorbable [54]. Consumption of food rich in vitamin c such as fruit and green leafy vegetable is advantageous to enhancing iron absorption. The other variable significantly associated with anemia was the nutritional status of children. Stunted children were 3.17 (AOR = 3.17) times more likely to be anemic. This finding is consistent with the finding of previous studies which showed that stunted children are more at risk of anemia than non-stunted children [32,37,40]. This might be due to inadequate intake of a diversified diet and/or increased nutrient loss due to intestinal parasite infection. Addressing the nutritional issue of school children through implementing a program like a school-based feeding practice might be very important to improve the nutritional status and prevent the continuation of stunting. Community education to create awareness on adequate dietary intake and prevention of intestinal parasites is also important. School children who were positive for intestinal parasite infection were, 5.23 (AOR = 5.23, $95\%$ CI: 2.77, 9.85) times more likely to be anemic than non-infected children. Comparable with the studies conducted in Northwestern Nigeria [38], Egypt [39], and Ethiopia [25,26,32,37,40] which indicated intestinal parasite infection as one of the factors that increase the risk of anemia. This could be explained by the contribution of the identified intestinal parasites to anemia through different mechanisms which could be blood loss, impaired nutrient absorption, reduction in nutrient intake due to induced loss of appetite, autoimmune hemolysis, and inflammation [55–57]. Intervention approaches including periodical school-based deworming, health education, enhancing sanitation services, and encouraging methods to maintain personal hygiene will help to control the extent of intestinal parasite infection [58,59]. This study is interpreted with some strengths and limitations. It provides information on the current prevalence and identifies associated factors of anemia among school children in the study area. In addition to the measurement of hemoglobin concentration, we try to assess the morphological type of anemia through a detailed morphological examination of a red blood cell. Additionally, an adjustment of hemoglobin for variation of altitude was made to avoid underestimating the prevalence of anemia. Despite this, the study has some limitations. One limitation is that we didn’t measure micronutrients like serum ferritin, folate, and vitamin B12 to identify the specific cause of anemia. The other limitation is the cross-sectional nature of the study design which makes the inference impossible to determine a cause-effect relationship. ## Conclusion and recommendation Nearly one-fourth of the schoolchildren in the study area were anemic. Females and children aged 10 to 14 years were more likely to be anemic. More than half of anemic children had mild anemia. Overall, anemia was a moderate public health problem among schoolchildren in the study area. Being female, having a family size greater than five, a non-consumption of green leafy vegetables, consuming milk, stunting, and intestinal parasite infection was significantly associated with anemia. Thus, school-based interventions targeting intestinal parasites and nutritional factors need to be implemented. To minimize the risk of intestinal parasite infection, health education to promote personal hygiene and environmental sanitation should be given. In addition, school-based regular deworming should be provided to students. Setting policy and implementing a program like school-based feeding practices might be very important to improving the nutritional status of schoolchildren. ## References 1. 1WHO. The global prevalence of anaemia in 2011. Geneva: World Health Organization; 2015.. *The global prevalence of anaemia in 2011* (2015.0) 2. Gardner W, Kassebaum N. **Global, Regional, and National Prevalence of Anemia and Its Causes in 204 Countries and Territories, 1990–2019**. *Curr Dev Nutr* (2020.0) **4** 830. DOI: 10.1093/cdn/nzaa053_035 3. Feleke BE, Derbie A, Zenebe Y, Mekonnen D, Hailu T, Adem Y. **Burden And Determinant Factors Of Anemia Among Elementary School Children In Northwest Ethiopia: A Comparative Cross Sectional Study**. *Afr J Infect Dis* (2017.0) **12** 1-6. DOI: 10.21010/ajid.v12i1.1 4. Desalegn A, Mossie A, Gedefaw L. **Nutritional Iron Deficiency Anemia: Magnitude and Its Predictors among School Age Children, Southwest Ethiopia: A Community Based Cross-Sectional Study**. *PLoS One* (2014.0) **9** 1-13. DOI: 10.1371/journal.pone.0114059 5. Tezera R, Sahile Z, Yilma D, Misganaw E, Mulu E. **Prevalence of anemia among school-age children in Ethiopia: a systematic review and meta-analysis**. *Syst Rev* (2018.0) **7** 1-7. PMID: 29304876 6. McLean E, Cogswell M, Egli I, Wojdyla D, De Benoist B. **Worldwide prevalence of anaemia, WHO vitamin and mineral nutrition information system, 1993–2005**. *Public Health Nutr* (2009.0) **12** 444-454. DOI: 10.1017/S1368980008002401 7. Lestari S, Fujiati II, Eyanoer PC, Sari DK, Daulay M, Nasution NH. **The Prevalence and Risk Factors of Anemia among Primary School Children in North Sumatera, Indonesia**. *Age* (2020.0) **8** 119 8. Getnet A, Worku S. **The association between major helminth infections (soil-transmitted helminthes and schistosomiasis) and Anemia among school children in Shimbit elementary school, Bahir Dar, Northwest Ethiopia**. *Am J Heal Res* (2015.0) **3** 97-104 9. Allali S, Brousse V, Sacri A-S, Chalumeau M, de Montalembert M. **Anemia in children: prevalence, causes, diagnostic work-up, and long-term consequences**. *Expert Rev Hematol* (2017.0) **10** 1023-1028. DOI: 10.1080/17474086.2017.1354696 10. Kassebaum NJ, Jasrasaria R, Naghavi M, Wulf SK, Johns N, Lozano R. **A systematic analysis of global anemia burden from 1990 to 2010**. *Blood, J Am Soc Hematol* (2014.0) **123** 615-624. DOI: 10.1182/blood-2013-06-508325 11. AbuKishk N, Turki Y, Saleh S, Albaik S, Hababeh M, El-Khatib Z. **Anaemia prevalence in children newly registered at UNRWA schools: a cross-sectional study**. *BMJ Open* (2020.0) **10** e034705. DOI: 10.1136/bmjopen-2019-034705 12. Hassan TH, Badr MA, Karam NA, Zkaria M, El Saadany HF, Rahman DMA. **Impact of iron deficiency anemia on the function of the immune system in children**. *Medicine (Baltimore)* (2016.0) **95**. DOI: 10.1097/MD.0000000000005395 13. Hill DL, Carr EJ, Rutishauser T, Moncunill G, Campo JJ, Innocentin S. **Immune system development varies according to age, location, and anemia in African children**. *Sci Transl Med* (2020.0) **12**. DOI: 10.1126/scitranslmed.aaw9522 14. **Iron deficiency anemia. assessment, prevention, and control**. *A Guid Program Manag* (2001.0) 15. Agustina R, Nadiya K, Andini EA, Setianingsih AA, Sadariskar AA, Prafiantini E. **Associations of meal patterning, dietary quality and diversity with anemia and overweight-obesity among Indonesian school-going adolescent girls in West Java**. *PLoS One* (2020.0) **15** e0231519. DOI: 10.1371/journal.pone.0231519 16. Chaparro CM, Suchdev PS. **Anemia epidemiology, pathophysiology, and etiology in low-and middle-income countries**. *Ann N Y Acad Sci* (2019.0) **1450** 15. DOI: 10.1111/nyas.14092 17. 17Central Statistical Agency, Summary and statistical report of the 2007 population and housing census: population size by age and sex, Addis Ababa, Ethiopia, 2008.. *Summary and statistical report of the 2007 population and housing census: population size by age and sex, Addis Ababa, Ethiopia* (2008.0) 18. Mesfin F, Berhane Y, Worku A. **Anemia among Primary School Children in Eastern Ethiopia**. *PLoS One* (2015.0) **10** 1-10. DOI: 10.1371/journal.pone.0123615 19. Amare B, Moges B, Moges F, Fantahun B, Admassu M, Mulu A. **Nutritional status and dietary intake of urban residents in Gondar, Northwest Ethiopia**. *BMC Public Health* (2012.0) **12** 1-10. PMID: 22214479 20. 20National Health And Nutrition Examination Survey (NHANES). Anthropometry Procedures Manual. 2007. 21. 21World Health O. Haemoglobin Concentrations for the Diagnosis of Anaemia and Assessment of Severity. Vitamin and Mineral Nutrition Information System. Document Reference WHO. 2011.. *Vitamin and Mineral Nutrition Information System* (2011.0) 22. Onis M De Onyango AW, Borghi E, Siyam A, Siekmann J. **Development of a WHO growth reference for school-aged children and adolescents**. *Bull World Health Organ* (2007.0) **85** 660-667. DOI: 10.2471/blt.07.043497 23. 23Bain BJ, Bates I, Laffan MA, Lewis SM. Basic haematological techniques. 11th ed. In: Houston M, editor. Dacie and Lewis Practical Haematology. 11th ed. 2012. pp. 11–15. 24. Cheesbrough M.. *Direct examination of faeces and concentration techniques* (2009.0) 191-2000 25. Assefa S, Mossie A, Hamza L. **Prevalence and severity of anemia among school children in Jimma Town, Southwest Ethiopia**. *BMC Hematol* (2014.0) **14** 1-9. DOI: 10.1186/2052-1839-14-3 26. Tariku EZ, Abebe GA, Melketsedik ZA, Gutema BT, Megersa ND, Sorrie MB. **Anemia and its associated factors among school-age children living in different climatic zones of Arba Minch Zuria District, Southern Ethiopia**. *BMC Hematol* (2019.0) **19** 1-9. PMID: 30637107 27. Tesfaye M, Yemane T, Adisu W, Asres Y, Gedefaw L. **Anemia and iron deficiency among school adolescents: burden, severity, and determinant factors in southwest Ethiopia**. *Adolesc Health Med Ther* (2015.0) **6** 189. DOI: 10.2147/AHMT.S94865 28. 28World Health Organization. Worldwide prevalence of anaemia 1993–2005: WHO global database on anaemia. World Health Organization; 2008.. *Worldwide prevalence of anaemia 1993–2005: WHO global database on anaemia* (2008.0) 29. Oliveira D, Ferreira FS, Atouguia J, Fortes F, Guerra A, Centeno-Lima S. **Infection by intestinal parasites, stunting and anemia in school-aged children from southern Angola**. *PLoS One* (2015.0) **10** e0137327. DOI: 10.1371/journal.pone.0137327 30. Semedo MRL, Santos M, Baião MR, Luiz RR, da Veiga G V. **Anemia and associated factors among school-age children in Cape Verde, West Africa**. *African J Food, Agric Nutr Dev* (2014.0) **14** 9511-9528 31. Mesfin F, Berhane Y, Worku A. **Anemia among primary school children in Eastern Ethiopia**. *PLoS One* (2015.0) **10** e0123615. DOI: 10.1371/journal.pone.0123615 32. Gutema B, Adissu W, Asress Y, Gedefaw L. **Anemia and associated factors among school-age children in Filtu Town, Somali region, Southeast Ethiopia**. *BMC Hematol* (2014.0) **14** 1-6. DOI: 10.1186/2052-1839-14-13 33. da Silva Ferreira H, de Assunção Bezerra MK, de Assunção ML, de Menezes RCE. **Prevalence of and factors associated with anemia in school children from Maceió, northeastern Brazil**. *BMC Public Health* (2016.0) **16** 1-12. PMID: 26728978 34. Hoang NTD, Orellana L, Le TD, Gibson RS, Worsley A, Sinclair AJ. **Anaemia and its relation to demographic, socio-economic and anthropometric factors in rural primary school children in Hai Phong City, Vietnam**. *Nutrients* (2019.0) **11** 1478. DOI: 10.3390/nu11071478 35. Tabi ESB, Cumber SN, Juma KO, Ngoh EA, Akum EA, Eyong EM. **A cross-sectional survey on the prevalence of anaemia and malnutrition in primary school children in the Tiko Health District, Cameroon**. *Pan Afr Med J* (2019.0) **32** 1-16. DOI: 10.11604/pamj.2019.32.111.15728 36. Mahmud MA, Spigt M, Mulugeta Bezabih A, Lopez Pavon I, Dinant G-J, Blanco Velasco R. **Risk factors for intestinal parasitosis, anaemia, and malnutrition among school children in Ethiopia**. *Pathog Glob Health* (2013.0) **107** 58-65. DOI: 10.1179/2047773213Y.0000000074 37. Getaneh Z, Enawgaw B, Engidaye G, Seyoum M, Berhane M, Abebe Z. **Prevalence of anemia and associated factors among school children in Gondar town public primary schools, northwest Ethiopia: A school-based cross-sectional study**. *PLoS One* (2017.0) **12** e0190151. DOI: 10.1371/journal.pone.0190151 38. Bello-Manga H, Mamman A, Suleiman H, Olayinka A, Musa A, Musa K. **Prevalence and pattern of parasitic infestations among nomadic Fulani children in a grazing reserve in Northwestern Nigeria**. *Ann Trop Med Public Heal* (2017.0) **10** 39. Salama RA, Labib MR. **The Prevalence of anemia among informal primary school children: a community based study in Rural Upper Egypt**. *Epidemiol Biostat Public Heal* (2016.0) **13** 40. Birhanu M, Gedefaw L, Asres Y. **Anemia among school-age children: magnitude, severity and associated factors in Pawe town, Benishangul-Gumuz region, Northwest Ethiopia**. *Ethiop J Health Sci* (2018.0) **28** 259-266. DOI: 10.4314/ejhs.v28i3.3 41. Birmeka M, Urga K, Petros B. **Assessment of Anemia and Iron Status of Primary School children aged 7–19 in Enemore-Ener and Abeshege Districts, Gurage Zone, Ethiopia**. *EC Nutr* (2017.0) **10** 18-25 42. Desalegn A, Mossie A, Gedefaw L. **Nutritional iron deficiency anemia: magnitude and its predictors among school age children, southwest Ethiopia: a community based cross-sectional study**. *PLoS One* (2014.0) **9** e114059. DOI: 10.1371/journal.pone.0114059 43. Hafeez MA, Rafique HM, Shaikh AW, Laghari ZA. **Prevalence of anaemia and its association with diet among the adolescent students of university of Sindh Jamshoro**. *Pakistan J Med Heal Sci* (2016.0) **10** 383-386 44. 44(WHO) WHO. Prevention of iron deficiency anemia in adolescents. Role Wkly Iron Folic Acid Suppl SEA-CAH-02; World Heal Organ Reg Off South-East Asia Geneva, Switz. 2011. 45. Bachman E, Travison TG, Basaria S, Davda MN, Guo W, Li M. **Testosterone induces erythrocytosis via increased erythropoietin and suppressed hepcidin: evidence for a new erythropoietin/hemoglobin set point**. *Journals Gerontol Ser A Biomed Sci Med Sci* (2014.0) **69** 725-735. DOI: 10.1093/gerona/glt154 46. Maggio M, Snyder PJ, Ceda GP, Milaneschi Y, Luci M, Cattabiani C. **Is the haematopoietic effect of testosterone mediated by erythropoietin? The results of a clinical trial in older men**. *Andrology* (2013.0) **1** 24-28. DOI: 10.1111/j.2047-2927.2012.00009.x 47. Feleke BE, Derbie A, Zenebe Y, Mekonnen D, Hailu T, Tulu B. **Burden and determinant factors of anemia among elementary school children in northwest Ethiopia: a comparative cross sectional study**. *African J Infect Dis* (2018.0) **12** 1-6 48. Amagloh FK, Atuna RA, McBride R, Carey EE, Christides T. **Nutrient and total polyphenol contents of dark green leafy vegetables, and estimation of their iron bioaccessibility using the in vitro digestion/Caco-2 cell model**. *Foods* (2017.0) **6** 54. DOI: 10.3390/foods6070054 49. Hurrell R, Egli I. **Iron bioavailability and dietary reference values**. *Am J Clin Nutr* (2010.0) **91** 1461S-1467S. DOI: 10.3945/ajcn.2010.28674F 50. Nomkong RF, Ejoh RA, Dibanda RF, Gabriel MN. **Bioavailability of Iron and Related Components in Cooked Green Leafy Vegetables Consumed in Cameroon**. *Food Nutr Sci* (2019.0) **10** 1096-1111 51. Thompson BA V, Sharp PA, Elliott R, Fairweather-Tait SJ. **Inhibitory effect of calcium on non-heme iron absorption may be related to translocation of DMT-1 at the apical membrane of enterocytes**. *J Agric Food Chem* (2010.0) **58** 8414-8417. DOI: 10.1021/jf101388z 52. Beck KL, Coad J. **Dairy product (calcium) consumption and iron nutrition**. *Nutrients in Dairy and their Implications on Health and Disease* (2017.0) 149-160 53. Lönnerdal B. **Calcium and iron absorption—mechanisms and public health relevance**. *Int J Vitam Nutr Res* (2010.0) **80** 293. PMID: 21462112 54. Bondi SA, Lieuw K. **Excessive Cow’s Milk Consumption and Iron Deficiency in Toddlers: Two Unusual Presentations and Review**. *ICAN Infant, Child, Adolesc Nutr* (2009.0) **1** 133-139 55. Shaw JG, Friedman JF, Gorakshakar AC. **Iron Deficiency Anemia: Focus on Infectious Diseases in Lesser Developed Countries**. *Anemia* (2011.0) **2011** 260380. DOI: 10.1155/2011/260380 56. Darlan DM, Ananda FR, Sari MI, Arrasyid NK, Sari DI. **Correlation between iron deficiency anemia and intestinal parasitic infection in school-age children in Medan**. *IOP Conf Ser Earth Environ Sci* (2018.0) **125** 12059 57. Friedman JF, Kanzaria HK, McGarvey ST. **Human schistosomiasis and anemia: the relationship and potential mechanisms**. *Trends Parasitol* (2005.0) **21** 386-392. DOI: 10.1016/j.pt.2005.06.006 58. Banu Hasina, K H, H MA. **Relationships Between Anaemia And Parasitic Infections In Adolescent Girls Of Bangladesh**. *Bangladesh J Zool* (2014.0) **42** 91-103 59. Dessie A, Gebrehiwot TG, Kiros B, Wami SD, Chercos DH. **Intestinal parasitic infections and determinant factors among school-age children in Ethiopia: a cross-sectional study**. *BMC Res Notes* (2019.0) **12** 1-6. PMID: 30602384
--- title: 'Knowledge, attitudes, and practices (KAP) of nutrition among school teachers in Bangladesh: A cross-sectional study' authors: - Mohammad Asadul Habib - Mohammad Rahanur Alam - Tanjina Rahman - Akibul Islam Chowdhury - Lincon Chandra Shill journal: PLOS ONE year: 2023 pmcid: PMC10038292 doi: 10.1371/journal.pone.0283530 license: CC BY 4.0 --- # Knowledge, attitudes, and practices (KAP) of nutrition among school teachers in Bangladesh: A cross-sectional study ## Abstract ### Background Teachers play a pivotal role in imparting nutritional knowledge to their students. This research aimed to evaluate the nutrition-related knowledge, attitudes, and practices (KAP) of selected Bangladeshi school teachers across the country. ### Methods A cross-sectional study was performed using a multistage sampling method. A pretested and structured questionnaire was used to collect data. Statistical analyses, including descriptive statistics, multiple regression analysis, and ANOVA tests, were performed to carry out the study. ### Results Among the 280 participants, only $9.9\%$ were happy with their understanding of childhood nutrition requirements, around $54.2\%$ were familiar with basic nutrition-related knowledge, and overall, $97.7\%$ of participants had a positive attitude towards learning about nutrition-related knowledge focusing on the well-being of children. Moreover, only $38.7\%$ had training in pediatric nutrition. Age, type of school, type of residence, professional training of school teachers, and the intent of having ever taken part or paying attention to nutrition-related knowledge significantly impacted the respondents’ nutrition knowledge score ($p \leq 0.05$). ### Conclusion Adequate planning and intervention measures should be developed to improve teachers’ understanding, behavior, and practice that encourage the growth of optimal nutrition-related behavior among school-aged children to establish a healthy nation. ## Introduction "Children are like buds in a garden and should be carefully and lovingly nurtured, as they are the nation’s future and the citizens of tomorrow", is a famous quote by Jawaharlal Nehru, the first prime minister of India. Recent decades have seen a dramatic shift in social infrastructure and the rise of industrialized society, which have contributed to an alarming increase in childhood obesity rates across the developing world [1, 2]. Obesity is perhaps the most prevalent form of malnutrition in affluent countries, affecting adults and children. Children with poor eating habits, physical inactivity, and a poor social and economic environment are more likely to be overweight, malnourished, and develop chronic disease, cognitive disability, and non-communicable diseases (NCDs) in adulthood, such as diabetes, obesity, hypertension, metabolic syndrome, coronary heart disease, and so on [3, 4]. According to the recent data from the World Health Organization (WHO), childhood obesity has become one of this century’s critical public health concerns [5]. It is reported that 20 to 40 percent of children and adolescents in wealthy countries are affected [6]. As of 2010, over 42 million children (<5 years) were facing overweight and nearly 35 million of those were from developing countries [7]. More than 22 million children under five are considered obese worldwide, with one out of every ten children being overweight. The prevalence of obesity in African and Asian countries is less ($10\%$) than the European & American countries ($20\%$) [8]. Both education and health are inextricably linked to one another [9]. The developing mind is more receptive to new information and experiences during childhood, making this a prime time for education that can increase the likelihood that knowledge and abilities will be retained and used later in life [10]. It is widely assumed that a child’s capacity to reach their full potential is strongly tied to the beneficial effects of excellent health, proper nutrition, physical activity, and quality education [9]. Schools play a critical role in the community regarding health education and awareness. School does have the most important impact on the lives of children [11, 12]. Studies have also shown that school teachers can implement a successful student health-related education program [9]. Student behavior regarding diet can be changed by the teachers’ ability to deliver nutrition education properly [13]. Trained teachers can quickly develop healthy eating habits among students, such as consuming fruits and vegetables, avoiding sugar-related products and unhealthy street foods, etc. [ 13, 14]. Some schools run education and health-related programs simultaneously to produce a "health-promoting" atmosphere that encourages learners to study [15]. According to several studies, healthy eating information in secondary school curricula is sparse, and even if it exists, it plays a minor role in the curriculum [5]. A nutrition-focused health education program at the school level is essential to address nutrition and health issues as the number of school-aged children, and health-related complications have risen in developing countries [16]. Various factors influence the development of excellent nutrition habits, and the knowledge–attitude–practice (KAP) paradigm provides a framework for bringing a constructive change in nutrition practices [17]. Improving the attitudes, knowledge, and practices regarding nutrition in children and adolescent is critical because it will result in a more food-conscious and healthier society. Teachers play an essential role in students’ healthy dietary habits at an early life that may track down in adulthood via educating about more nutritious food choices, regular physical activity, and a sedentary lifestyle. They may also incorporate basic food and nutrition courses in the syllabus [13, 18]. In Bangladesh, there were no data about the teacher’s knowledge regarding nutrition, however, the knowledge, attitude and practice of child, adolescent and mothers are evaluated [19–21]. So, there is a need to evaluate the knowledge, attitude and practice towards building an intense nutrition related perceptions among school teachers. The current study aimed to evaluate the knowledge, attitudes, and practices (KAP) regarding nutrition among school teachers in Bangladesh. Determining their level of KAP for developing well-balanced nutritious eating habits and regular physical activity towards students will provide concrete evidence from which it could be possible to guide an intervention plan to improve the current condition. ## Study design A cross-sectional study was conducted from February 2022 to June 2022 in different regions of Bangladesh. Multistage stratified random sampling was applied in collecting data from school teachers of nine [9] districts in Bangladesh (Fig 1). Initially, all respondents were given a self-administered standardized questionnaire [22]. Informed consent was taken from each participant and responses were only collected from school teachers who filled the questionnaire completely. A total of 280 school teachers were finally involved in this study. School teachers who did not provide their consent and did not fill the questionnaire were excluded. Additionally, no monetary reward or tangible prize was offered for their participation into the study. **Fig 1:** *Sample areas of the study (Prepared by using QGIS software version 3.10.2).* ## Data collection tools This study used a pre-designed questionnaire based on the KAP model [22]. An android-based "KoBo Collect" software was used to collect the data. There were four sections of the questionnaire: (a) demographic characteristics, (b) nutrition-related knowledge, (c) nutrition-related attitude, and (d) nutrition-related practices or behaviors. A total of 66 questions were asked, of which 8 questions were demographic-related, 49 were nutrition knowledge related, 5 were attitude related, and 4 were practice related. The knowledge-related questions were divided into seven sections containing questions related to (i) protein, (ii) fat, (iii) vitamin, (iv) calcium, (v) dietary fiber, (vi) nutrient elements, and (vii) the child’s nutrition. ## Scoring The assignment technique was used to determine the nutrition-related knowledge score by explicitly judging each question as true or false. Answers were graded with a 1 for the correctness and a 0 for incorrectness. This section had a maximum score of 49 points. Each attitude question was categorized into three parts; (i) satisfied/confident/ willing, (ii) neutral, and (iii) unsatisfied/no confidence/ unwilling. ## Internal validity The internal validity of the final questionnaire was measured by internal reliability, and Cronbach’s alpha (α) was 0.742. The validity of three segments that evaluated the nutrition-related knowledge, attitude, and practice among school teachers was also measured by internal reliability, and Cronbach’s alpha (α) was 0.773. ## Data analysis The statistical analyses were performed using IBM SPSS 23.0 software. The demographic features of survey respondents were used to compute descriptive statistics: frequency & percentage. To evaluate the mean differences among the continuous variables, ANOVA tests were used. A multiple linear regression analysis was performed to determine the influences on nutrition knowledge. Multiple linear regression analysis includes variables with statistical significance in single component analysis. A two-sided test was used to verify all statistics, and the statistical significance level was set at 0.05. ## Ethical approval The study protocol was approved by the ethical board of Noakhali Science and Technology University, and respective institutes contributed to this study (NSTU/SCI/EC/$\frac{2022}{113}$). Informed written consent was also obtained from all participants before the study. ## Characteristics of school-teachers Table 1 demonstrated the demographic characteristics of study participants, where we found that school teachers aged 17 to 57 years were mainly distributed at the age of >40 years ($41.1\%$). Most of the teachers were male ($72.9\%$); $81.8\%$ of participants were married; $52.4\%$ were from government schools, $9\%$ were from urban areas, $35.4\%$ have served as teachers for more than 10 years, and $45.4\%$ have served as a school teacher for 3–10 years. About $57.1\%$ had higher secondary education, and $42.9\%$ had received above higher secondary education. Moreover, $78.9\%$ of the respondents were professionally trained. **Table 1** | Variables | N (%) | Variables.1 | N (%).1 | | --- | --- | --- | --- | | Age | | Education Level | | | 17–30 years | 54 (19.3) | Primary | 0 | | 31–40 years | 111 (39.6) | Secondary | 0 | | >40 years | 115 (41.1) | Higher | 160 (57.1) | | >40 years | 115 (41.1) | Above | 120 (42.9) | | Gender | | Type of School | | | Male | 204 (72.9) | Government | 148 (52.9) | | Female | 76 (27.1) | Private | 132 (47.1) | | Marital Status | | Working Time | | | Unmarried | 46 (16.4) | <3 years | 54 (19.2) | | Married | 229 (81.8) | 3–10 years | 127 (45.4) | | Divorced | 5 (1.8) | >10 years | 99 (35.4) | | Residence | | Professional Training | | | Urban | 162 (57.9) | Yes | 221 (78.9) | | Rural | 118 (42.1) | No | 59 (21.1) | ## Nutrition-related knowledge Table 2 demonstrated the average scores of the school teachers on nutrition-related knowledge based on the demographic profile. The average score of nutrition-related knowledge was 34.33 (SD: 18.16), which was $70.07\%$ of the total score of 49.00. The average scores (%) of the seven sections (Dimension 1 to Dimension 7) in Table 2 were as follows: (i) 63.23±15.15, (ii) 74.18± 17.22, (iii) 72.50 ± 13.22, (iv) 85.72±18.16, (v) 73.44± 15.55, (vi) 65.40±22.09, and (vii) 56.03±25.77. Interestingly, the lowest average score was found in "Child nutrition-related knowledge," and the highest average score was in "Calcium-related knowledge." *In this* analysis, gender (dimension 6), marital status (dimension 2), types of residence (dimension 2), an education level (dimension 1,2,3,4,5,6), type of school (dimension 5), working time (dimension 5,7), professional training (dimension 2,6) of school teachers, and total nutrition-related knowledge scores showed statistically significant differences ($p \leq 0.05$). **Table 2** | Variables | Dimensions | Dimensions.1 | Dimensions.2 | Dimensions.3 | Dimensions.4 | Dimensions.5 | Dimensions.6 | Dimensions.7 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Variables | 1 | 2 | 2 | 3 | 4 | 5 | 6 | 7 | | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | Age (Years) | | 17–30 31–40 >40 | 62.04±14.6364.56±11.2664.93±14.20 | 73.01±15.8676.45±16.4276.02±15.89 | 73.01±15.8676.45±16.4276.02±15.89 | 74.84±11.9573.80±12.3573.77±13.30 | 86.11±15.1090.09±17.2887.61±19.13 | 76.34±12.2272.87±14.6576.52±15.70 | 72.22±24.0170.27±20.7768.98±21.04 | 56.01±30.1161.71±28.4155.87±30.0 | | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender | | Male Female | 64.22±13.2164.25±13.25 | 75.35±16.8676.32±13.91 | 75.35±16.8676.32±13.91 | 74.51±13.0272.59±11.53 | 87.99±18.8189.14±14.34 | 75.05±14.9175.00±14.30 | 71.73±20.43*65.79±23.71* | 58.95±29.7756.25±28.61 | | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | | Unmarried Married Divorced | 62.32±51.0964.62±12.8963.33±7.45 | 72.05±18.5576.29±15.5977.14±12.77 | 72.05±18.5576.29±15.5977.14±12.77 | 74.09±12.6974.05±12.7470.00±7.45 | 89.13±14.5888.43±18.3675.00±00.00 | 76.09±12.6175.16±14.8560.00±21.66 | 71.74±24.3170.16±20.8853.33±18.25 | 52.17±30.9959.60±29.1550.00±23.39 | | Residence | Residence | Residence | Residence | Residence | Residence | Residence | Residence | Residence | | Urban Rural | 64.71±12.7863.56±13.77 | 73.28±15.28*78.81±16.68* | 73.28±15.28*78.81±16.68* | 73.25±12.4175.00±12.94 | 88.27±16.0588.35±19.79 | 73.79±15.3976.74±13.63 | 69.75±21.6070.62±21.43 | 56.40±30.1160.69±28.41 | | Education level | Education level | Education level | Education level | Education level | Education level | Education level | Education level | Education level | | Primary Secondary Higher Above | 44.44±19.2465.35±14.4165.10±12.6565.10±12.81 | 61.90±41.2361.90±23.1874.20±16.0681.40±9.75 | 61.90±41.2361.90±23.1874.20±16.0681.40±9.75 | 58.33±28.8761.11±16.3174.01±11.1177.26±11.56 | 58.33±52.0478.57±26.5687.66±16.8292.45±13.13 | 66.67±29.4059.26±19.9873.96±13.0780.56±12.59 | 55.56±38.4953.97±26.8270.00±20.8874.30±19.03 | 33.33±19.0954.17±23.8361.09±27.1155.08±33.76 | | Type of School | Type of School | Type of School | Type of School | Type of School | Type of School | Type of School | Type of School | Type of School | | Government Private | 64.18±12.3564.27±14.14 | 76.83±14.0974.24±18.03 | 76.83±14.0974.24±18.03 | 75.06±12.6272.79±12.60 | 88.51±16.8588.07±18.66 | 76.66±14.47*73.23±14.85* | 17.72±18.2069.44±24.73 | 56.25±31.4060.42±27.02 | | Working time | Working time | Working time | Working time | Working time | Working time | Working time | Working time | Working time | | <3 years 3–10 years >10 years | 61.42±14.4164.57±12.2465.32±13.61 | 73.01±15.8675.36±17.7077.34±13.84 | 73.01±15.8675.36±17.7077.34±13.84 | 74.38±10.8474.28±13.8173.40±12.0 | 87.50±14.3788.00±19.3889.14±17.19 | 76.13±13.02*71.92±15.32*78.45±14.09* | 72.84±25.9668.77±20.4670.37±20.14 | 59.72±30.68*62.20±27.70*52.27±30.21* | | Professional training of teacher | Professional training of teacher | Professional training of teacher | Professional training of teacher | Professional training of teacher | Professional training of teacher | Professional training of teacher | Professional training of teacher | Professional training of teacher | | Yes No | 64.47±12.6763.27±15.09 | 64.47±12.6763.27±15.09 | 77.18±14.70*69.73±19.50* | 74.67±11.5871.46±15.87 | 89.37±17.0284.32±19.64 | 75.87±14.4471.94±15.49 | 71.64±19.34*64.40±27.58* | 58.89±29.4755.72±29.39 | Table 2 also showed that the highest score was among participants aged 31–40, whereas the lowest score was found among teachers aged 17–30, except for "Protein-related knowledge." Participants who were married, residing in urban areas, or were more educated had a higher average score than other participants. Compared to teachers at private schools, government school teachers had higher average test scores. Participants in government schools received lower average scores than the total mean scores. Except for "Vitamin-related information," the average scores of the respondents who worked for more than ten years were the highest, while participants who worked for less than three years scored the lowest. School teachers having professional training perform better than those who don’t have. However, untrained teachers’ average scores were better than the scores for all students. " Protein-related knowledge" were statistically significant with only the education level of the demographic profile ($p \leq 0.05$). " Fat-related knowledge" were statistically significant with demographic profile (marital status, residence, education level, professional training of teacher) ($p \leq 0.05$). Vitamin-related knowledge and Calcium-related knowledge were statistically significant with only the education level of the demographic profile ($p \leq 0.05$). Dietary fiber-related information was statistically significant with the education level, school type, and working time ($p \leq 0.05$). Nutrient element-related knowledge was statistically substantial with demographic profile (gender, education level, professional training of teacher) ($p \leq 0.05$) (Table 2). Child nutrition-related knowledge was statistically significant with only working time of the demographic profile. ## Nutrition knowledge learning-related attitudes Approximately $61.8\%$ of the school teachers reported having "the confidence to conduct the childhood nutrition task properly." Teachers who felt confident in themselves scored better than those who did not. Only $24.3\%$ of interviewees reported being "satisfied" with their prior understanding of child nutrition. Teachers who were content with their present nutrition knowledge received lower marks than those who weren’t. Regarding their current dietary expertise, the majority need to be more convinced. Approximately $81.8\%$ of school teachers were "ready to get information linked to child nutrition knowledge using new media (Google, online journal, Facebook, what’s app, Messenger and others)" in addition to $78.9\%$ who were "willing to attend more childhood nutrition education and training." There were statistically significant variations with p-value <0.05 between attitude and behavior and nutrition-related knowledge (Table 3). **Table 3** | Variables | N (%) | A score of attitudes (mean± SD) | P value | | --- | --- | --- | --- | | A.1. Do you have the confidence to do the childhood nutrition work well? | A.1. Do you have the confidence to do the childhood nutrition work well? | A.1. Do you have the confidence to do the childhood nutrition work well? | A.1. Do you have the confidence to do the childhood nutrition work well? | | No confidence | 26 (9.3%) | 49.62±18.86 | .000* | | Neutral | 81 (28.9%) | 67.78±14.23 | .000* | | Confidence | 173 (61.8%) | 89.08±12.99 | .000* | | A.2. Are you willing to learn more knowledge about children-related nutrition? | A.2. Are you willing to learn more knowledge about children-related nutrition? | A.2. Are you willing to learn more knowledge about children-related nutrition? | A.2. Are you willing to learn more knowledge about children-related nutrition? | | Unwilling | 10 (3.6%) | 39.00±20.79 | .000* | | Neutral | 45 (16.1%) | 58.89±17.99 | .000* | | Willing | 225 (80.4%) | 85.11±13.99 | .000* | | A.3. Are you satisfied with the childhood nutrition knowledge you already have? | A.3. Are you satisfied with the childhood nutrition knowledge you already have? | A.3. Are you satisfied with the childhood nutrition knowledge you already have? | A.3. Are you satisfied with the childhood nutrition knowledge you already have? | | Satisfied | 68 (24.3%) | 62.21±18.84 | .000* | | Neutral | 116 (41.3%) | 76.03±14.79 | .000* | | Unsatisfied | 96 (34.3%) | 95.21±10.25 | .000* | | A.4. Are you willing to attend more young childhood nutrition education and training? | A.4. Are you willing to attend more young childhood nutrition education and training? | A.4. Are you willing to attend more young childhood nutrition education and training? | A.4. Are you willing to attend more young childhood nutrition education and training? | | Unwilling | 10 (3.6%) | 37.00±21.10 | .000* | | Neutral | 49 (17.5%) | 58.37±13.59 | .000* | | Willing | 221 (78.9%) | 85.79±13.95 | .000* | | A.5. Are you willing to get information related to children’s nutrition knowledge through new media (Google, online journal, Facebook, what’s app, Messenger and others)? | A.5. Are you willing to get information related to children’s nutrition knowledge through new media (Google, online journal, Facebook, what’s app, Messenger and others)? | A.5. Are you willing to get information related to children’s nutrition knowledge through new media (Google, online journal, Facebook, what’s app, Messenger and others)? | A.5. Are you willing to get information related to children’s nutrition knowledge through new media (Google, online journal, Facebook, what’s app, Messenger and others)? | | Unwilling | 11 (3.9%) | 40.91±24.68 | .000* | | Neutral | 40 (14.3%) | 58.25±17.95 | .000* | | Willing | 229 (81.8%) | 84.76±14.04 | .000* | ## Nutrition knowledge learning-related practices or behaviors Only $42.5\%$ of the participants took courses or received training in children’s nutrition. Teachers who took classes or received training in childhood nutrition performed, on average better than those who did not. Concerning $67.9\%$ of participants, nutrition awareness among youngsters was a focus. Teachers who were aware of the nutrition-related information of the students had better average results than those who weren’t. Less than half ($47.9\%$) of the participants have occasionally attempted to get their family members or acquaintances to learn more about the nutritional aspect of the children. The average scores of teachers were greater than those of those who frequently ($12.1\%$), occasionally ($47.9\%$), or never ($32.9\%$) took the initiative to convey students’ nutritional information to their families or friends. The vast majority of participants ($53.4\%$) occasionally took the initiative to receive more knowledge on child nutrition. The likelihood that teachers will attempt to know about children’s nutrition is higher than that of other adults: always ($6.8\%$), frequently ($15.4\%$), occasionally ($52.6\%$), and never ($24.3\%$). The following questions for the total nutrition-related knowledge scores revealed statistically significant differences with p value <0.05: "Have you participated in childhood nutrition knowledge courses or training?" " Have you paid attention to children’s nutrition knowledge?" " Have you ever taken the initiative to promote children’s nutrition knowledge to your relatives or friends?" " Have you ever taken the initiative to learn about child nutrition in your spare time?" ( Table 4). **Table 4** | Variables | N (%) | A score of Practice (mean ± SD) | P Value | | --- | --- | --- | --- | | P1. Have you ever participated in Childhood nutrition education knowledge courses or training? | P1. Have you ever participated in Childhood nutrition education knowledge courses or training? | P1. Have you ever participated in Childhood nutrition education knowledge courses or training? | P1. Have you ever participated in Childhood nutrition education knowledge courses or training? | | No | 161 (57.5%) | 25.62±18.48 | .000 | | Yes | 119 (42.5%) | 56.09±22.48 | .000 | | P2. Have you ever paid attention to children’s nutrition knowledge? | P2. Have you ever paid attention to children’s nutrition knowledge? | P2. Have you ever paid attention to children’s nutrition knowledge? | P2. Have you ever paid attention to children’s nutrition knowledge? | | No | 90 (32.1%) | 16.53±16.10 | .000 | | Yes | 190 (67.9%) | 49.01±21.85 | .000 | | P3. Have you ever taken the initiative to promote children’s nutrition knowledge to your relatives or friends? | P3. Have you ever taken the initiative to promote children’s nutrition knowledge to your relatives or friends? | P3. Have you ever taken the initiative to promote children’s nutrition knowledge to your relatives or friends? | P3. Have you ever taken the initiative to promote children’s nutrition knowledge to your relatives or friends? | | Never | 92 (32.9%) | 14.13±13.89 | .000 | | Sometimes | 134 (47.9%) | 40.86±11.38 | .000 | | Often | 34 (12.1%) | 63.97±10.55 | .000 | | Always | 20 (7.1%) | 92.50±13.69 | .000 | | P4. Have you ever taken the initiative to learn about child nutrition in your spare time? | P4. Have you ever taken the initiative to learn about child nutrition in your spare time? | P4. Have you ever taken the initiative to learn about child nutrition in your spare time? | P4. Have you ever taken the initiative to learn about child nutrition in your spare time? | | Never | 68 (24.3%) | 9.93±12.51 | .000 | | Sometimes | 150 (53.6%) | 38.67±12.32 | .000 | | Often | 43 (15.4%) | 60.17±16.42 | .000 | | Always | 19 (6.8%) | 91.44±15.62 | .000 | ## Factors affecting nutrition-related knowledge Age, school type, practice score, and nutrition-related knowledge among school instructors were all found significantly different in multiple linear regression models ($p \leq 0.05$). With age, teachers became more knowledgeable regarding nutrition. Teachers from public schools had a greater understanding of nutrition than private schools. The teachers’ working hours, professional backgrounds, and overall knowledge scores were statistically significant compared to their attitudes toward nutrition. Compared to teachers with less work time, those with more work time showed a more nutrition-related mindset. Those who had a more nutrition-related attitude scored higher in all areas of nutrition knowledge. School type, knowledge score overall, and nutrition-related practice among instructors were shown to differ significantly ($p \leq 0.05$). Teachers who had attended public schools had more experience with nutrition. Those who had experience with nutrition-related activities scored higher overall in nutrition-related knowledge (Table 5). **Table 5** | Demographic Variables | Knowledge | Knowledge.1 | Knowledge.2 | Knowledge.3 | Attitude | Attitude.1 | Attitude.2 | Attitude.3 | Practice | Practice.1 | Practice.2 | Practice.3 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Demographic Variables | Β | SE | t | P | Β | SE | t | P | β | SE | t | p | | Age (Years) | -.723 | 1.140 | -.634 | .000* | 3.851 | 2.420 | 1.591 | .113 | -5.645 | 3.438 | .115 | .908 | | Gender | .268 | 1.134 | .236 | .527 | -1.963 | 2.413 | -.814 | .417 | -2.617 | 3.430 | -1.642 | .102 | | Marital Status | -.178 | 1.434 | -.124 | .813 | 2.151 | 3.053 | .705 | .482 | 5.915 | 4.328 | -.763 | .446 | | Residence | .442 | 1.005 | .440 | .901 | 2.491 | 2.135 | 1.166 | .244 | -1.225 | 3.042 | 1.367 | .173 | | Education Level | 5.165 | .820 | 6.296 | .660 | -.964 | 1.871 | -.515 | .607 | -5.619 | 2.639 | -.403 | .687 | | Type of School | 1.591 | 1.065 | 1.494 | .000* | -.710 | 2.278 | -.312 | .755 | -2.928 | 3.233 | -2.129 | .034* | | Working Time | .255 | 1.149 | .222 | .136 | -5.216 | 2.427 | -2.150 | .032* | .008 | 3.479 | -.905 | .366 | | Professional training of teacher | -.519 | 1.378 | -.377 | .824 | -10.797 | 2.861 | -3.774 | .000* | .257 | 4.173 | .002 | .998 | | Total Knowledge Score | Ref. | Ref. | Ref. | Ref. | .742 | .122 | 6.090 | .000* | .789 | .178 | 4.428 | .000* | | Attitude Score | .163 | .027 | 6.090 | .707 | Ref. | Ref. | Ref. | Ref. | .105 | .086 | 1.213 | .226 | | Practice Score | .086 | .019 | 4.428 | .000* | .052 | .043 | 1.213 | .226 | Ref. | Ref. | Ref. | Ref. | ## Discussion The KAP model has been widely employed in public health concerning behavior-changing communication (BCC) strategies to encourage healthy behavior based on a society’s local context. The guiding premise of this model was to increase knowledge since doing so will change attitudes and behaviors and lessen the socioeconomic toll that diseases have on society. According to this paradigm, an ideal diet could help with disease prevention, health promotion, and risk reduction [23]. The ability to adjust one’s eating habits may be significantly influenced by knowledge [24]. This study revealed that the participating number of teachers with adequate understanding of nutrition were not satisfactory and overall, their correct score for all nutrition-related information was $70.07\%$. Without enough nutrition expertise, teachers won’t be able to advocate health-related programs that may reduce childhood overweight and obesity related issues [25, 26]. Unlike educators in other nations, Bangladeshi teachers are found to bear better concerned about nutrition. Additionally, this study found that all domains had poor average scores, with "Child’s nutrition-related knowledge" receiving the lowest score. The findings indicated that most of the participants with poor scores in this domain were deficient in nutrition related knowledge. Our study findings also revealed the existence of false beliefs and taboos on various nutrition related topics for children among school teachers, for instance, the state of child’s health. The lack of understanding among teachers about child’s nutrition seem to be a problem that may hinder their ability to raise awareness among their pupils. This may lead to poor health, which could hamper cognitive development and damage the learning environment, as those who have nutrition education and wellness opportunities can better teach their students about nutrition and aid in improving their health [27]. This survey is the first cross-sectional study to assess the association between nutrition-related knowledge and other aspects among school instructors, including attitudes and actions connected to nutrition knowledge learning and demographic variables. Here, we showed that factors including age, residence status, type of school, training of school teachers, the behavior of having ever taken part in training, and intent of ever paying attention to children’s nutrition affect the overall nutrition-related knowledge. The study further showed an increased awareness of school teachers with their age, which may be related to their extensive teaching experience and the ups and downs in various students’ performances that they have seen. The study also found that private school teachers had the lowest scores in all domains of nutrition related knowledge. Previous studies on food-related views among rural American elementary and middle school teachers revealed that many lacked nutrition-related understanding regarding important foods [28]. Our investigations found that teachers with professional training (excluding nutrition courses) had less nutrition-related knowledge than the instructors who did not have a sort of training, probably because of their age and year of experience in service. According to our study, practices related to nutrition knowledge are associated with the total scores of nutrition related knowledge. Lack of adequate facilities and equipment for healthy food consumption in many schools may reduce the practice of healthy eating. Developing food and nutrition-based curricula in schools may improve the nutrition related KAP for students and teachers [29]. According to the results of the current study, $21.01\%$ of the participants had never taken any training or course related to child’s nutrition. Based on a recent survey, the percentage of child caregivers who attended workshops or played a role in service nutrition-related education was only one-third [30]. The current study also demonstrated that the teachers who took part in nutrition awareness training or courses had higher marks than those who did not. Some studies showed that teachers’ training on nutrition and its related methods increases their nutrition-related knowledge [31, 32]. However, better nutrition-related understanding cannot possibly be predicted by teachers’ individualized attitudes and beliefs [33]. The current study also showed that all school instructors, regardless of their age with full coverage, must have their early training in strengthening nutrition-related knowledge. Most school teachers in our study, especially those with more nutrition-related expertise, exhibited good behaviors. Teachers’ knowledge and food-related beliefs and behaviors can influence the habits of their student to adopt the healthy eating practices as teachers themselves could be role model for his students [34, 35]. School-based nutrition programs are required to encourage teachers to learn about nutrition actively. Our study also found that most respondents ($81.8\%$) were willing to use social media (such as Facebook, WhatsApp, Google, online journals, and Messenger) to learn about child’s nutrition knowledge. Additionally, people who were open to using these media had a higher likelihood of knowing something more related to nutrition than those who were neutral or opposed to it [36, 37]. However, further research is required on how preschool teachers could be informed through social media on issues related to public health, more particularly child health and nutrition. This way, it is possible to prevent childhood obesity and malnutrition, the two most contemporary public health concerns, with the help of teachers. This study had several limitations. First, we could not determine the causality due to the data from a cross-sectional survey. Second, the sample size of the study was small as the subjects were randomly selected based on $100\%$ response rate that which may impact the precision and dependability of some of the findings. Furthermore, data on the knowledge, attitudes and nutrition practices among school instructors were self-reported. Finally, because there were only four and three items in the "nutrient element-related knowledge" and "calcium-related knowledge" sections of the questionnaire, perhaps, it is possible that the participants were nor familiar with the nutrition related topics that could bring information bias. ## Conclusion A fair amount of nutrition expertise among school teachers was found only in a few districts (3 Out of 9 in this study) of Bangladesh. Most teachers were eager to learn about nutrition-related topics yet needed more formal training. In addition, nutrition-related knowledge among school teachers was correlated with age, sex, marital status, residence type, school type, educational qualification, working hours, professional training, and attention to child health and nutrition information. Significant association was also found between the knowledge and behavior of the teachers about nutrition. The KAP model was used in this study to observe the existing condition and find out the gap in this field, which will help in planning effective interventions to improve Bangladeshi school-teachers current understanding of child health and nutrition in future. School teachers can help in improving the nutritional status of children by adopting convenient, simple, cost-effective educational tools. They can also monitor both the physical (change in body weight for consuming a balanced diet) and psychological (change in attention span while studying) outcomes of their students. Thus, teachers could play a pivotal role in the overall well-being of our society. ## References 1. Arunachalam S., Kandasami M.. **Mother’s attitude on childhood obesity and its prevention**. *Current Pediatric Research* (2019) **23** 117-121 2. Aranceta J.. **Prevention of overweight and obesity: a Spanish approach.**. *Public health nutrition,* (2007) **10** 1187-1193. DOI: 10.1017/S1368980007000699 3. Khor G.L.. **Update on the prevalence of malnutrition among children in Asia.**. *Nepal Med Coll J* (2003) **5** 113-22. PMID: 15024783 4. Kaufman-Shriqui V.. **Effect of a school-based intervention on nutritional knowledge and habits of low-socioeconomic school children in Israel: a cluster-randomized controlled trial.**. *Nutrients* (2016) **8** 234. DOI: 10.3390/nu8040234 5. Garrido-Fernández A.. **Attitude towards the Promotion of Healthy Eating among Secondary School Teachers&mdash;Construction and Validation of a Questionnaire.**. *Nutrients* (2022) **14** 2271. PMID: 35684071 6. Wang Y., Lobstein T.. **Worldwide trends in childhood overweight and obesity**. *International journal of pediatric obesity* (2006) **1** 11-25. PMID: 17902211 7. Sahoo K.. **Childhood obesity: causes and consequences**. *Journal of family medicine and primary care* (2015) **4** 187 8. Kosti R.I., Panagiotakos D.B.. **The epidemic of obesity in children and adolescents in the world**. *Central European journal of public health* (2006) **14** 151. PMID: 17243492 9. Htun Y.M.. **Knowledge, attitude and reported practice of primary school teachers on specified school health activities in Danuphyu Township, Ayeyarwaddy Region, Myanmar.**. *South East Asia Journal of Public Health* (2013) **3** 24-29 10. David M.D., Kimiywe J., Waudo J.. **Nutritional knowledge, attitude and practices and nutritional status of school-going children in Machakos District, Kenya.**. *in International Conference on Nutrition and Food Sciences, IPCBEE.* (2012) 11. Laforêt-Fliesser Y.. **School Health Guidance Document (2010), Standards, Programs and Community Development Branch.**. *Ontario: Ministry of health promotion* (2010) 12. Saraswathy G.. (2007) 13. Hall E., Chai W., Albrecht J.A.. **A qualitative phenomenological exploration of teachers’ experience with nutrition education**. *American Journal of Health Education* (2016) **47** 136-148. PMID: 27226814 14. Laguna M.C.. **Teachers as healthy beverage role models: relationship of student and teacher beverage choices in elementary schools**. *Journal of community health* (2020) **45** 121-127. PMID: 31407126 15. Organization W.H.. **A guide for establishing health promoting schools**. *WHO Regional Office for South-East Asia* (2003) 16. Sanyaolu A.. **Childhood and adolescent obesity in the United States: a public health concern.**. *Global pediatric health* (2019) **6** 2333794X19891305. DOI: 10.1177/2333794X19891305 17. Mamba N.P., Napoles L., Mwaka N.M.. **Nutrition knowledge, attitudes and practices of primary school children in Tshwane Metropole, South Africa.**. *African Journal of Primary Health Care and Family Medicine* (2019) **11** 1-7. DOI: 10.4102/phcfm.v11i1.1846 18. Nanayakkara J., Margerison C., Worsley A.. **Ways to improve secondary school teachers’ confidence in teaching food and nutrition subjects**. *Education Inquiry,* (2022) 1-16 19. Ghosh S.. **Balanced diet related knowledge, attitude and practices (KAP) among adolescent school girls in Noakhali district, Bangladesh: A cross sectional study.**. *International Journal of Adolescent Medicine and Health* (2020) **34** 319-325. DOI: 10.1515/ijamh-2020-0106 20. Kundu S.. **Sources of nutrition information and nutritional knowledge among school-going adolescents in Bangladesh.**. *Public Health in Practice,* (2020) **1** 100030. DOI: 10.1016/j.puhip.2020.100030 21. Hoddinott J.. **Behavior change communication activities improve infant and young child nutrition knowledge and practice of neighboring non-participants in a cluster-randomized trial in rural Bangladesh.**. *PloS one* (2017) **12** e0179866. DOI: 10.1371/journal.pone.0179866 22. Liu H.. **Nutrition-related knowledge, attitudes, and practices (KAP) among kindergarten teachers in Chongqing, China: A cross-sectional survey.**. *International journal of environmental research and public health* (2018) **15** 615. DOI: 10.3390/ijerph15040615 23. Kamp B.. **Position of the American Dietetic Association, American Society for Nutrition, and Society for Nutrition Education: food and nutrition programs for community-residing older adults**. *Journal of the American Dietetic Association* (2010) **110** 463-472. PMID: 20213956 24. Bookari K., Yeatman H., Williamson M.. **Exploring Australian women’s level of nutrition knowledge during pregnancy: a cross-sectional study**. *International journal of women’s health* (2016) **8** 405. DOI: 10.2147/IJWH.S110072 25. Sharma S.. **Nutrition-related knowledge, attitudes, and dietary behaviors among head start teachers in Texas: a cross-sectional study**. *Journal of the Academy of Nutrition and Dietetics* (2013) **113** 558-562. PMID: 23415503 26. Anderson L.M.. **The effectiveness of worksite nutrition and physical activity interventions for controlling employee overweight and obesity: a systematic review.**. *American journal of preventive medicine* (2009) **37** 340-357. DOI: 10.1016/j.amepre.2009.07.003 27. Saha S., Zahid M.K., Rasheed S.. **The study of the level of knowledge, attitude, practices (KAP) as well as the effects of school environment on the nutritional status of children (7–12) coming from affluent families in the Dhaka City in Bangladesh**. *Bangladesh Journal of Nutrition* (2011) 31-48 28. Findholt N.E.. **Food-related practices and beliefs of rural US elementary and middle school teachers.**. *Rural and remote health,* (2016) **16** 1-9. PMID: 27287617 29. Hyska J.. **Assessing nutritional status and nutrition-related knowledge, attitudes and practices of Albanian schoolchildren to support school food and nutrition policies and programmes.**. *Food Policy* (2020) **96** 101888 30. Nahikian-Nelms M.. **Influential factors of caregiver behavior at mealtime: a study of 24 child-care programs**. *Journal of the American Dietetic Association* (1997) **97** 505-509. PMID: 9145088 31. Motamedrezaei O.. **The effect of nutrition and food hygieneeducation on the knowledge of female elementary school teachers in city of ferdows**. *Journal of education and health promotion* (2013) **2** 10. PMID: 24083260 32. Chen Y.-H.. **Significant effects of implementation of health-promoting schools on schoolteachers’ nutrition knowledge and dietary intake in Taiwan.**. *Public Health Nutrition* (2010) **13** 579-588. DOI: 10.1017/S1368980009991017 33. Katsagoni C.N.. **Schoolteachers’ nutrition knowledge, beliefs, and attitudes before and after an e-learning program**. *Journal of nutrition education and behavior* (2019) **51** 1088-1098. PMID: 31402288 34. He L.. **Association of children’s eating behaviors with parental education, and teachers’ health awareness, attitudes and behaviors: a national school-based survey in China.**. *The European Journal of Public Health* (2014) **24** 880-887. DOI: 10.1093/eurpub/ckt177 35. Horne P.J.. **Increasing pre-school children’s consumption of fruit and vegetables. A modelling and rewards intervention**. *Appetite* (2011) **56** 375-385. PMID: 21112361 36. Wu T.. **Evaluation of chronic disease prevention and control public service advertisement on the awareness and attitude change among urban population in Chongqing, China: A cross-sectional study.**. *International journal of environmental research and public health* (2017) **14** 1515. DOI: 10.3390/ijerph14121515 37. Wang F.. **Evaluation of a WeChat-based dementia-specific training program for nurses in primary care settings: A randomized controlled trial.**. *Applied Nursing Research,* (2017) **38** 51-59. DOI: 10.1016/j.apnr.2017.09.008
--- title: 'Presence and severity of migraine is associated with development of primary open angle glaucoma: A population-based longitudinal cohort study' authors: - Kyoung Ohn - Kyungdo Han - Jung Il Moon - Younhea Jung journal: PLOS ONE year: 2023 pmcid: PMC10038297 doi: 10.1371/journal.pone.0283495 license: CC BY 4.0 --- # Presence and severity of migraine is associated with development of primary open angle glaucoma: A population-based longitudinal cohort study ## Abstract ### Purpose To examine the association between the presence and severity of migraine and development of primary open-angle glaucoma (POAG) using a nationwide population-based longitudinal cohort data. ### Methods Data were retrieved from the Korean National Health Insurance Service for 2,716,562 individuals aged ≥ 40 years and assessed for the development of POAG from 2009 through 2018. Subjects were classified into the following 3 groups: healthy control subjects, subjects with mild migraine, and those with severe migraine. Hazard ratios (HR) of glaucoma development were calculated for each group. Subgroup analyses of subjects stratified by age, sex, lifestyle factors (smoking, drinking, and body mass index (BMI)), and comorbidities (diabetes, hypertension, and dyslipidemia). ### Results During the 9-year follow-up period, the incidence rate of POAG per 1000 person-years was 2.41 and 3.25 in subjects without and with migraine, respectively. Among the migraine group, the incidence rate was 3.14 and 3.89 in mild and severe subgroups, respectively. The HR was 1.355 ($95\%$ CI, 1.300–1.412) and 1.188 ($95\%$ CI, 1.140–1.239) before and after adjusting for potential confounding factors in the migraine group per se. Regarding the severity of migraine, the adjusted HRs were 1.169 ($95\%$ CI, 1.117–1.224) in the mild migraine group, and 1.285 ($95\%$ CI, 1.166–1.415) in the severe migraine group compared to the control group. The results were consistent in subgroup analyses after stratifying by age, sex, lifestyle factors, and comorbidities. ### Conclusions Migraine is associated with increased risk of POAG development. Furthermore, chronic and severe migraine is associated with greater risk of POAG development. ## Introduction Migraine affects more than $10\%$ of the adult population worldwide and it ranks among the world’s most disabling medical illnesses [1]. In addition, the economic and societal effect of migraine is substantial: it affects patients’ quality of life and impairs work, social activities, and family life [2–4]. Although the nature and mechanism of migraine are complex and remain incompletely understood, potential mechanisms include vasospasm, vascular endothelium-related hypercoagulability, and vascular changes related with cortical spreading depression [5, 6]. Migraine is considered a systemic vasculopathy and is associated with ischemic heart disease, stroke, and other cardiovascular diseases [6]. Glaucoma is also a multifactorial disease characterized by progressive optic neuropathy and distinctive visual field loss [7, 8]. While intraocular pressure has been identified as the most important risk factor for its development, other risk factors have been identified [9, 10]. It has been known that female [11], older age [8, 12, 13], smoking [14, 15], drinking [14, 16], underexercising [14, 17], lower BMI [14, 18–20], and CKD [21] increase the risk for glaucoma. Both systemic vascular factors, such as hypertension and diabetes, and ocular vascular factors, such as ocular blood flow and ocular perfusion pressure, have been identified as risk factors, emphasizing the role of vascular mechanisms in its pathophysiology [9, 10, 22]. This association seems even stronger in those with normal tension glaucoma (NTG) [23], which is the most prevalent type of glaucoma in Korea. The proportion of NTG among POAG patients in Korea was $77\%$ in the Namil epidemiologic study [24]. Therefore, we included the patients with NTG into the subject population of the present study to better represent the epidemiologic situation in Korea. Considering this common etiology, a potential association between migraine and primary open angle glaucoma (POAG) has been previously studied, but the results are inconclusive. The Blue Mountains Eye Study demonstrated a positive association between migraine and POAG among those aged 70–79 [25]. In contrast, the Beaver Dam Eye Study found that there was no evidence of a relationship between open-angle glaucoma and migraine [26]. In a meta-analysis investigating the association between migraine and glaucoma, a significant association was found in case-control design studies, but not in cohort design studies, and the authors concluded the association is still controversial [27]. In addition, no study has explored the risk of glaucoma according to the severity of migraine. Therefore, this study investigated the association between migraine and increased risk of developing POAG over a 9-year follow-up period using a nationwide longitudinal cohort data. We also examined whether this potential association was proportional with the chronicity and the severity of migraine. ## Database Data were retrieved from the Korean National Health Insurance Service (KNHIS) database, namely, the National Health Information Database (NHID) and the Korean Health Examination Database (KHED). All Koreans residing in the Republic of Korea are obliged to join the KNHIS since 1989. Collectively, the NHID contains medical data (e.g., personal information, diagnosis, medical treatment), history of comorbid diseases (e.g., hypertension, dyslipidemia, stroke, diabetes mellitus), and demographics (e.g., age, sex, household income) of patients. In addition, KNHIS also provides a standardized health screening program to all Koreans aged 40 or more. The KHED includes anthropometric data and laboratory tests. The KNHIS provides de-identified data to medical researchers who meet the access criteria. This study was approved by the Institutional Review Board of Yeouido St. Mary’s Hospital, Seoul, Korea (SC21ZISE0175). Informed consent was waived by the IRB. This study was a retrospective cohort study, and all data were fully anonymized by the KNHIS before assessed by researchers. This study adhered to the Declaration of Helsinki. ## Exposure determination and outcome measurement The exposure of interest was migraine diagnosis. Diagnosis for migraine was defined as at least one KNHIS claim under International Classification of Diseases (ICD)-10 code for migraine (G43) in 2009 (index year). Regarding the severity of migraine, subjects without chronic and severe migraine were further stratified into the “mild” group and those with chronic and severe migraine were categorized into the “severe” group. Chronic migraine was defined as diagnosis of migraine more than twice at least 3 months apart. Severe migraine was defined as follows: i) at least one hospitalization for migraine as the main diagnosis, ii) at least one visit to the emergency room for migraine as the main diagnosis, iii) Triptans (N02CC) and ergots (N02CA) are administered at least once within 1 year from the time of diagnosis, iv) perform the following procedures at least once within 1 year from the time of diagnosis: MM070, LA210, LA222, LA223, LA224, LA225, LA226, LA227, LA228, S0471, S0472, S0474, S0475, S0476, S0477, S0478, S4730, SZ636, SZ637, SZ638, or SZ639. The outcome of interest was POAG. Development of POAG was defined as KNHIS claims with ICD-10 code for POAG (H401). The ICD-10 code H401 includes POAG and NTG. To strengthen the diagnostic validity, only those with at least three outpatient visits for glaucoma a year were included [28]. To establish a temporal and causal relationship between migraine and POAG, those with prior diagnosis of POAG before the index year and those who were diagnosed within one year after migraine diagnosis were excluded. ## Statistical analysis The Student’s T-test was used to compare continuous variables between groups, and the Chi-square test was used to compare categorical variables. Cox’s proportional hazard regression was performed to estimate the hazard ratios for various risk factors. Hazard ratio for exposure of interest, history of migraine, was calculated using three models: model-1, un-adjusted; model-2, adjusted for age, sex, smoking habits, drinking habits, frequency of exercise, household income; model-3, adjusted for comorbid disease status (diabetes, hypertension, dyslipidemia, body mass index, and glomerular filtration rate), along with all risk factors addressed in model-2. Incidence probabilities of POAG to migraine was calculated using Kaplan-Meier survival analysis. All statistical analyses were performed using SAS (Version 9.4; SAS Institute, Cary, NC, USA). p-values ≤ 0.05 were considered statistically significant. ## Study population Fig 1 shows our study population. We first identified 2,896,383 individuals aged 40 years or more who had undergone national health screening program in 2009. After exclusion of 59,327 individuals who had a previous history of POAG, additional 107,591 individuals were excluded due to missing data. In addition, we excluded patients who developed POAG within one year from the index date. Finally, 2,716,562 individuals were enrolled and assessed for the development of POAG until 2018. **Fig 1:** *Selection of study population.* Table 1 presents the demographic characteristics, medical conditions, and laboratory test results of subjects in study population. Compared to subjects without history of migraine, subjects diagnosed with migraine were more likely to be female, non-smoker, non-drinker, had higher prevalence of hypertension, dyslipidemia, myocardial infarction, chronic heart failure, and stroke. Apart from that, we found small but statistically significant differences in age, sex, exercise, income, glomerular filtration rate (GFR), body mass index (BMI), waist circumference, systolic/diastolic blood pressure, and blood level of glucose, total cholesterol, high/low density lipoprotein, triglyceride, γ-glutamyl transpeptidase, aspartate transaminase, and alanine transferase between those with and without migraine. **Table 1** | Unnamed: 0 | Unnamed: 1 | Migraine, n(%) | Migraine, n(%).1 | p-value | | --- | --- | --- | --- | --- | | | | No | Yes | p-value | | Parameters | Parameters | 2628753 | 87809 | | | Sex, male | Sex, male | 1,343,046 (51.09) | 24,289 (27.66) | < .0001 | | Smoking status | Smoking status | | | < .0001 | | | Non | 1,651,662 (62.83) | 69,972 (79.69) | | | | Ex-Smoker | 414,682 (15.77) | 8,376 (9.54) | | | | Current | 562,409 (21.39) | 9,461 (10.77) | | | Drinking status | Drinking status | | | < .0001 | | | Non | 1,521,777 (57.89) | 64,916 (73.93) | | | | Mild | 910,785 (34.65) | 19,682 (22.41) | | | | Heavy | 196,191 (7.46) | 3,211 (3.66) | | | Regular exercise | Regular exercise | 529,358 (20.14) | 15,980 (18.2) | < .0001 | | Monthly income, lower 25% | Monthly income, lower 25% | 469,348 (17.85) | 16,572 (18.87) | < .0001 | | Diabetes | Diabetes | 303,486 (11.54) | 9,836 (11.2) | 0.0017 | | Hypertension | Hypertension | 919,150 (34.97) | 37,515 (42.72) | < .0001 | | Dyslipidemia | Dyslipidemia | 595,416 (22.65) | 25,500 (29.04) | < .0001 | | Myocardial infarction | Myocardial infarction | 10,901 (0.41) | 506 (0.58) | < .0001 | | Chronic heart failure | Chronic heart failure | 21,294 (0.81) | 1,410 (1.61) | < .0001 | | Stroke | Stroke | 54,308 (2.07) | 4,978 (5.67) | < .0001 | | Migraine | Migraine | | | | | | Chronic | 0 (0) | 19,987 (22.76) | < .0001 | | | Severe | 0 (0) | 49,387 (56.24) | < .0001 | | Migraine Group | Migraine Group | | | < .0001 | | | Non | 2,628,753 (100) | 0 (0) | | | | Mild | 0 (0) | 74,795 (85.18) | | | | Severe | 0 (0) | 13,014 (14.82) | | | Age, year | Age, year | 54.14±10.4 | 56.46±10.91 | < .0001 | | GFR, mL/min/1.73m2 | GFR, mL/min/1.73m2 | 84.91±37.51 | 83.77±30.84 | < .0001 | | Body mass index, Kg/m2 | Body mass index, Kg/m2 | 23.97±3.03 | 23.99±3.1 | 0.0769 | | Waist Circumference ratio, % | Waist Circumference ratio, % | 81.24±8.89 | 80.49±9.4 | < .0001 | | Total cholesterol, mg/dl | Total cholesterol, mg/dl | 199.34±42.04 | 201.56±45.76 | < .0001 | | Blood glucose | Blood glucose | 100.01±25.91 | 98.58±23.39 | < .0001 | | Systolic BP | Systolic BP | 124.22±15.52 | 123.97±15.6 | < .0001 | | Diastolic BP | Diastolic BP | 77.21±10.23 | 76.89±10.08 | < .0001 | | HDL, mg/dl | HDL, mg/dl | 56.16±34.12 | 56.82±35.21 | < .0001 | | LDL, mg/dl | LDL, mg/dl | 119.11±83.6 | 121.8±82.94 | < .0001 | ## The risk of POAG among the study subjects during the 9-year follow up Table 2 shows the incidence and hazard ratio (HR) of POAG according to migraine status. The incidence rate per 1000 person-years was 2.408 and 3.249 in subjects without and with migraine, respectively. Among the migraine group, the incidence rate was 3.137 and 3.893 in mild and severe groups, respectively. The HRs calculated using model-1, model-2, and model-3 were 1.355 ($95\%$ confidence interval [CI]: 1.300–1.412), 1.202 ($95\%$ CI: 1.153–1.253), and 1.188 ($95\%$ CI: 1.140–1.239), respectively. Cumulative incidence of POAG according to history of migraine is illustrated in Fig 2. The log-rank test showed that subjects with migraine had significantly higher POAG development rates compared with those without ($P \leq 0.001$). In addition, a subgroup analysis was performed to further elucidate the impact of migraine severity on development of POAG. HR was greater when migraine was chronic and severe. HRs for subjects with mild migraine compared to those without migraine were 1.306 ($95\%$ CI: 1.248–1.367), 1.181 ($95\%$ CI: 1.128–1.236), and 1.169 ($95\%$ CI: 1.117–1.224) and those with severe migraine were 1.628 ($95\%$ CI: 1.478–1.792), 1.312 ($95\%$ CI: 1.191–1.445), and 1.285 ($95\%$ CI: 1.166–1.415) when using model-1, model-2, and model-3, respectively. **Fig 2:** *Cumulative incidence of POAG according to migraine diagnosis.* TABLE_PLACEHOLDER:Table 2 We further conducted a risk-stratified analysis to further quantify the impact of migraine in specific patient subgroups. Tables 3 and 4 presents HRs of POAG in subgroups stratified according to subjects’ age, sex, BMI, smoking/drinking habits or presence of diabetes/hypertension/dyslipidemia. The association between migraine and subsequent POAG was consistent in all subgroup analyses. The association of migraine with development of POAG differed according to age (P for interaction = 0.004) and HTN (P for interaction = 0.001) (Table 3). The impact of migraine on POAG development was greater in the younger group (aHR: 1.235, $95\%$ CI: 1.168–1.305) compared to those aged 65 or older (aHR: 1.139, $95\%$ CI: 1.069–1.214). In addition, the aHR of POAG diagnosis during the 9-year follow-up period was greater in those without hypertension (aHR: 1.261, $95\%$ CI:1.186–1.341) compared to those with hypertension (aHR: 1.128, $95\%$ CI:1.066–1.194) Similar pattern was noted when subjects with chronic and severe migraine were classified into a subgroup (Table 4). ## Discussion We found that subjects with migraine were independently associated with a 1.19 times increased risk of POAG diagnosis after adjusting for age, sex, smoking, drinking, exercise, income, diabetes, hypertension, dyslipidemia, body mass index, and glomerular filtration rate within 9 years after their diagnosis. While those with mild migraine showed 1.17 times greater risk, those with severe and chronic migraine showed 1.29 times greater risk of POAG after adjusting for confounding factors. Migraine is an episodic neurologic disorder characterized by recurrent attacks of throbbing headache and typical premonitory features such as nausea or sensory hypersensitivity. The development of symptoms are related with the activation of trigeminovascular system and cortical spreading depression, although underlying mechanisms are still up for discussion [5]. It is noteworthy that glaucoma shares certain characteristics with migraine in that both are possibly associated with vascular dysregulation in terms of pathophysiology of the diseases [29, 30]. Moreover, vasospasm intensity, calculated based on the presence or absence of migraine, cold extremities and vasospastic response to temperature change, is strongly associated with the development of the disease in particular subgroups of glaucoma [31]. Broadway and Drance [31] discovered that glaucoma patients with focal ischemia of the optic nerve head and subsequent focal loss of the neuroretinal rim had a higher prevalence of vasospasm and migraine than glaucoma patients without these forms of optic tissue destruction. Whether migraines can significantly increase the risk of developing POAG is still controversial. For example, Lin et al. [ 32]. reported that subjects with migraine were more likely to have POAG compared with those without migraine, even after adjusting for gender, age, monthly income, and level of urbanization of the community. The Collaborative Normal-Tension Glaucoma Study Group analyzed the risk factors for deterioration of visual field defects in NTG and found migraine to be an independent risk factor for more rapid progression [11]. Huang et al. [ 33] found that migraine is associated with a higher risk of open angle glaucoma in patients with no comorbidity who are aged under 50 years. Motsko et al. reported migraines were found to be independent risk factors for the development of OAG [34]. Also, other previous studies found significant association between migraine and POAG [35, 36]. On the other hand, some studies reported that migraines did not increase the risk of POAG [37–39]. Landers et al. found there was no significant association between migraine and POAG after adjusting for confounders [39]. We believe that diverse study designs and ethnic populations may be contributing to the disparities among studies. Prevalence of migraine and POAG both increase along with the subjects’ age. It was consistently reported that migraine remained to be a significant risk factor for the development of POAG when the relative risk was adjusted for the subjects’ age [27, 33]. To corroborate this observation, we performed a risk-stratified analysis in our study (Tables 3 and 4). Notably, when the subjects were divided into two age groups, the impact of migraine on the development of POAG was greater in those who were younger than 65 compared to those aged 65 or older. In addition, we also found that migraineurs without comorbid hypertension showed higher risk of POAG than those with hypertension. Further research is warranted to clarify our findings. Previously, Huang et al. [ 33] also reported that the cumulative incidence of POAG was significantly higher among the migraineurs compared to normal control subjects. However, because it was not possible to perform a subclassification of migraine group in their study, the relationship between glaucoma progression and migraine severity was not addressed. To shed light on this, we subclassified subjects with migraine based on chronicity and severity of migraine based on the frequency and type of insurance claims. It is worth to note that we saw stronger evidence of association between migraine and POAG, when the course of migraine was worse (Table 2). Our study has the following strengths. First, the sample size is large enough to jointly evaluate the impact of migraine on the development of POAG, with the effect of other variables including age, sex, smoking or drinking habits, socioeconomic status, and comorbid diseases. In addition, we aimed to establish a temporal and causal relationship between migraine and POAG onset using this longitudinal database by excluding those previously diagnosed with POAG before or within one year after the index year. Secondly, we only included those with medically diagnosed migraine and POAG rather than self-questionnaires, thereby increasing the validity of the study subjects. Thirdly, although confined to Koreans, the study population of the present study well represents the actual population composition of the entire country, given that it is an obligation to join the KNHIS for all individuals in South Korea. Our study also has some limitations. All data were collected from KHED and NHID, which originally were not established with the sole intention of being used for research purpose which might have led to some limitations. First, people with undiagnosed glaucoma were not included in our estimates of glaucoma prevalence, which may have resulted in an underestimated prevalence. Second, national claims data do not always match hospital chart records. It has been shown that national claims data and hospital records only have a 70–$80\%$ concordance rate [40]. We attempted to mitigate this by including only those with at least three visits under the diagnosis of glaucoma, but this may have affected our results. Third, we haven’t excluded or controlled for different classes of antihypertensive medication which can be associated with glaucoma development or progression. Some studies found that antihypertensive drugs seem to reduce the risk of developing glaucoma [41]. On the other hand, others found that antihypertensive medication compromises optic nerve head blood flow which may lead to a higher risk of glaucomatous progression [42, 43]. Finally, the degree of migraine can vary depending on the medication, however, we have not included the effect of migraine medication on chronicity and severity of migraine in the subanalysis. Also, the clinical information about glaucoma medication, glaucomatous optic neuropathy and visual field defects were unavailable and measurement and adjustment for factors a priori known to be associated with the POAG, such as intraocular pressure, might strengthen the importance of our findings. ## Conclusion In conclusion, our study suggests that presence and severity of migraine are both associated with increased risk of subsequent development of POAG. We found that the hazard of POAG development was 1.19 times greater in subjects with migraine compared to those without after adjusting for age, sex, lifestyle factors, and comorbidities. Furthermore, we also found that subjects with chronic and severe migraine showed 1.29 times greater hazard of POAG development. It is recommended that neurologists refer migraineurs for glaucoma assessment, especially when migraine is chronic and severe. ## References 1. Stovner L, Hagen K, Jensen R, Katsarava Z, Lipton R, Scher A. **The global burden of headache: a documentation of headache prevalence and disability worldwide.**. *Cephalalgia* (2007) **27** 193-210. DOI: 10.1111/j.1468-2982.2007.01288.x 2. Breslau N, Davis GC, Schultz LR, Peterson EL. **Joint 1994 Wolff Award Presentation. Migraine and major depression: a longitudinal study.**. *Headache* (1994) **34** 387-93. DOI: 10.1111/j.1526-4610.1994.hed3407387.x 3. Lipton RB, Hamelsky SW, Kolodner KB, Steiner TJ, Stewart WF. **Migraine, quality of life, and depression: a population-based case-control study**. *Neurology* (2000) **55** 629-35. DOI: 10.1212/wnl.55.5.629 4. Silberstein SD. **Migraine.**. *Lancet* (2004) **363** 381-91. DOI: 10.1016/S0140-6736(04)15440-8 5. Pietrobon D, Moskowitz MA. **Pathophysiology of migraine**. *Annu Rev Physiol* (2013) **75** 365-91. DOI: 10.1146/annurev-physiol-030212-183717 6. Tietjen EG. **Migraine and ischaemic heart disease and stroke: potential mechanisms and treatment implications**. *Cephalalgia* (2007) **27** 981-7. DOI: 10.1111/j.1468-2982.2007.01407.x 7. Resnikoff S, Pascolini D, Etya’ale D, Kocur I, Pararajasegaram R, Pokharel GP. **Global data on visual impairment in the year 2002**. *Bull World Health Organ* (2004) **82** 844-51. PMID: 15640920 8. Tham YC, Li X, Wong TY, Quigley HA, Aung T, Cheng CY. **Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis**. *Ophthalmology* (2014) **121** 2081-90. DOI: 10.1016/j.ophtha.2014.05.013 9. Quigley HA. **Glaucoma.**. *Lancet* (2011) **377** 1367-77. DOI: 10.1016/S0140-6736(10)61423-7 10. Weinreb RN, Khaw PT. **Primary open-angle glaucoma**. *Lancet* (2004) **363** 1711-20. DOI: 10.1016/S0140-6736(04)16257-0 11. Drance S, Anderson DR, Schulzer M. **Risk factors for progression of visual field abnormalities in normal-tension glaucoma**. *Am J Ophthalmol* (2001) **131** 699-708. DOI: 10.1016/s0002-9394(01)00964-3 12. Heijl A, Bengtsson B, Hyman L, Leske MC. **Natural history of open-angle glaucoma**. *Ophthalmology* (2009) **116** 2271-6. DOI: 10.1016/j.ophtha.2009.06.042 13. Rudnicka AR, Mt-Isa S, Owen CG, Cook DG, Ashby D. **Variations in primary open-angle glaucoma prevalence by age, gender, and race: a Bayesian meta-analysis**. *Invest Ophthalmol Vis Sci* (2006) **47** 4254-61. DOI: 10.1167/iovs.06-0299 14. Perez CI, Singh K, Lin S. **Relationship of lifestyle, exercise, and nutrition with glaucoma.**. *Curr Opin Ophthalmol* (2019) **30** 82-8. DOI: 10.1097/ICU.0000000000000553 15. Law SM, Lu X, Yu F, Tseng V, Law SK, Coleman AL. **Cigarette smoking and glaucoma in the United States population.**. *Eye (Lond).* (2018) **32** 716-25. DOI: 10.1038/eye.2017.292 16. Wise LA, Rosenberg L, Radin RG, Mattox C, Yang EB, Palmer JR. **A prospective study of diabetes, lifestyle factors, and glaucoma among African-American women.**. *Ann Epidemiol.* (2011) **21** 430-9. DOI: 10.1016/j.annepidem.2011.03.006 17. Meier NF, Lee DC, Sui X, Blair SN. **Physical Activity, Cardiorespiratory Fitness, and Incident Glaucoma.**. *Med Sci Sports Exerc* (2018) **50** 2253-8. DOI: 10.1249/MSS.0000000000001692 18. Pasquale LR, Willett WC, Rosner BA, Kang JH. **Anthropometric measures and their relation to incident primary open-angle glaucoma**. *Ophthalmology* (2010) **117** 1521-9. DOI: 10.1016/j.ophtha.2009.12.017 19. Lin SC, Pasquale LR, Singh K, Lin SC. **The Association Between Body Mass Index and Open-angle Glaucoma in a South Korean Population-based Sample**. *J Glaucoma* (2018) **27** 239-45. DOI: 10.1097/IJG.0000000000000867 20. Marshall H, Berry EC, Torres SD, Mullany S, Schmidt J, Thomson D. **Association Between Body Mass Index and Primary Open Angle Glaucoma in Three Cohorts**. *Am J Ophthalmol* (2022) **245** 126-33. DOI: 10.1016/j.ajo.2022.08.006 21. Liu W, Guo R, Huang D, Ji J, Gansevoort RT, Snieder H. **Co-occurrence of chronic kidney disease and glaucoma: Epidemiology and etiological mechanisms.**. *Surv Ophthalmol.* (2022). DOI: 10.1016/j.survophthal.2022.09.001 22. Yanagi M, Kawasaki R, Wang JJ, Wong TY, Crowston J, Kiuchi Y. **Vascular risk factors in glaucoma: a review.**. *Clin Exp Ophthalmol* (2011) **39** 252-8. DOI: 10.1111/j.1442-9071.2010.02455.x 23. Cursiefen C, Wisse M, Cursiefen S, Jünemann A, Martus P, Korth M. **Migraine and tension headache in high-pressure and normal-pressure glaucoma**. *Am J Ophthalmol* (2000) **129** 102-4. DOI: 10.1016/s0002-9394(99)00289-5 24. Kim CS, Seong GJ, Lee NH, Song KC, Namil Study Group KGS. **Prevalence of primary open-angle glaucoma in central South Korea the Namil study**. *Ophthalmology* (2011) **118** 1024-30. DOI: 10.1016/j.ophtha.2010.10.016 25. Wang JJ, Mitchell P, Smith W. **Is there an association between migraine headache and open-angle glaucoma? Findings from the Blue Mountains Eye Study**. *Ophthalmology* (1997) **104** 1714-9. DOI: 10.1016/s0161-6420(97)30075-x 26. Klein BE, Klein R, Meuer SM, Goetz LA. **Migraine headache and its association with open-angle glaucoma: the Beaver Dam Eye Study**. *Invest Ophthalmol Vis Sci* (1993) **34** 3024-7. PMID: 8360033 27. Xu C, Li J, Li Z, Mao X. **Migraine as a risk factor for primary open angle glaucoma: A systematic review and meta-analysis.**. *Medicine (Baltimore).* (2018) **97** e11377. DOI: 10.1097/MD.0000000000011377 28. Jung Y, Han K, Ohn K, Kim DR, Moon JI. **Association between diabetes status and subsequent onset of glaucoma in postmenopausal women**. *Sci Rep* (2021) **11** 18272. DOI: 10.1038/s41598-021-97740-3 29. Flammer J, Konieczka K, Flammer AJ. **The primary vascular dysregulation syndrome: implications for eye diseases.**. *Epma* (2013) **4** 14. DOI: 10.1186/1878-5085-4-14 30. Killer HE, Forrer A, Flammer J. **Retinal vasospasm during an attack of migraine.**. *Retina* (2003) **23** 253-4. DOI: 10.1097/00006982-200304000-00023 31. Broadway DC, Drance SM. **Glaucoma and vasospasm**. *Br J Ophthalmol* (1998) **82** 862-70. DOI: 10.1136/bjo.82.8.862 32. Lin HC, Chien CW, Hu CC, Ho JD. **Comparison of Comorbid Conditions between Open-Angle Glaucoma Patients and a Control Cohort A Case-Control Study**. *Ophthalmology* (2010) **117** 2088-95. DOI: 10.1016/j.ophtha.2010.03.003 33. Huang JY, Su CC, Wang TH, Tsai IJ. **Migraine and increased risk of developing open angle glaucoma: a population-based cohort study.**. *BMC Ophthalmol.* (2019) **19** 50. DOI: 10.1186/s12886-019-1062-9 34. Motsko SP, Jones JK. **Is there an association between hypothyroidism and open-angle glaucoma in an elderly population? An epidemiologic study**. *Ophthalmology* (2008) **115** 1581-4. DOI: 10.1016/j.ophtha.2008.01.016 35. Girkin CA, McGwin G, McNeal SF, Lee PP, Owsley C. **Hypothyroidism and the development of open-angle glaucoma in a male population**. *Ophthalmology* (2004) **111** 1649-52. DOI: 10.1016/j.ophtha.2004.05.026 36. Welinder LG, Riis AH, Knudsen LL, Thomsen RW. **Diabetes, glycemic control and risk of medical glaucoma treatment: A population-based case-control study.**. *Clin Epidemiol.* (2009) **1** 125-31. DOI: 10.2147/clep.s6831 37. Lin HC, Kang JH, Jiang YD, Ho JD. **Hypothyroidism and the risk of developing open-angle glaucoma: a five-year population-based follow-up study**. *Ophthalmology* (2010) **117** 1960-6. DOI: 10.1016/j.ophtha.2010.02.005 38. Chen HY, Lin CL, Kao CH. **Does Migraine Increase the Risk of Glaucoma?: A Population-Based Cohort Study.**. *Medicine (Baltimore).* (2016) **95** e3670. DOI: 10.1097/MD.0000000000003670 39. Landers J, Goldberg I, Graham SL. **Analysis of risk factors that may be associated with progression from ocular hypertension to primary open angle glaucoma**. *Clin Exp Ophthalmol* (2002) **30** 242-7. DOI: 10.1046/j.1442-9071.2002.00528.x 40. Park I, Kim DJ, Kim J, Kim H, Kim H, Min K. **Current Status of Aspirin User in Korean Diabetic Patients Using Korean Health Insurance Database.**. *The Journal of Korean Diabetes Association* (2006) 30. DOI: 10.4093/jkda.2006.30.5.363 41. Horwitz A, Klemp M, Jeppesen J, Tsai JC, Torp-Pedersen C, Kolko M. **Antihypertensive Medication Postpones the Onset of Glaucoma: Evidence From a Nationwide Study**. *Hypertension* (2017) **69** 202-10. DOI: 10.1161/HYPERTENSIONAHA.116.08068 42. Topouzis F, Wilson MR, Harris A, Founti P, Yu F, Anastasopoulos E. **Association of open-angle glaucoma with perfusion pressure status in the Thessaloniki Eye Study**. *Am J Ophthalmol* (2013) **155** 843-51. DOI: 10.1016/j.ajo.2012.12.007 43. Harris A, Topouzis F, Wilson MR, Founti P, Kheradiya NS, Anastasopoulos E. **Association of the optic disc structure with the use of antihypertensive medications: the thessaloniki eye study**. *J Glaucoma* (2013) **22** 526-31. DOI: 10.1097/IJG.0b013e31824d1e12
--- title: 'Factors correlated with pain after total knee arthroplasty: A systematic review and meta-analysis' authors: - Unni Olsen - Maren Falch Lindberg - Christopher Rose - Eva Denison - Caryl Gay - Arild Aamodt - Jens Ivar Brox - Øystein Skare - Ove Furnes - Kathryn A. Lee - Anners Lerdal journal: PLOS ONE year: 2023 pmcid: PMC10038299 doi: 10.1371/journal.pone.0283446 license: CC BY 4.0 --- # Factors correlated with pain after total knee arthroplasty: A systematic review and meta-analysis ## Abstract ### Main objective Systematically review and synthesize preoperative and intraoperative factors associated with pain after total knee arthroplasty (TKA) in patients with osteoarthritis. ### Methods Based on a peer-reviewed protocol, we searched Medline, Embase, CINAHL, Cochrane Library, and PEDro for prospective observational studies (January 2000 to February 2023) investigating factors associated with pain after TKA. The primary outcome was pain twelve months after TKA. Pain at three and six months were secondary outcomes. Multivariate random-effects meta-analyses were used to estimate mean correlation ($95\%$ CIs) between factors and pain. Sensitivity analysis was performed for each risk of bias domain and certainty of evidence was assessed. ### Results Of 13,640 studies, 29 reports of 10,360 patients and 61 factors were analysed. The mean correlation between preoperative factors and more severe pain at twelve months was estimated to be 0.36 ($95\%$ CI, 0.24, 0.47; $P \leq .000$; moderate-certainty evidence) for more catastrophizing, 0.15 ($95\%$ CI; 0.08, 0.23; $P \leq .001$; moderate-certainty evidence) for more symptomatic joints, 0.13 ($95\%$ CI, 0.06, 0.19; $P \leq .001$; very low-certainty evidence) for more preoperative pain. Mean correlation between more severe radiographic osteoarthritis and less pain was -0.15 ($95\%$ CI; -0.23, -0.08; $P \leq .001$; low-certainty evidence). In sensitivity analysis, the estimated correlation coefficient for pain catastrophizing factor increased to 0.38 ($95\%$ CI 0.04, 0.64). At six and three months, more severe preoperative pain was associated with more pain. Better preoperative mental health was associated with less pain at six months. ### Conclusion and relevance More pain catastrophizing, more symptomatic joints and more pain preoperatively were correlated with more pain, while more severe osteoarthritis was correlated with less pain one year after TKA. More preoperative pain was correlated with more pain, and better mental health with less pain at six and three months. These findings should be further tested in predictive models to gain knowledge which may improve TKA outcomes. ## Introduction Total knee arthroplasty (TKA) is one of the most common surgical procedures [1, 2], and is considered as an effective procedure in relieving pain and restore physical function in patients with end-stage osteoarthritis (OA). Although TKA surgery is effective for most, one in five patients may experience chronic postsurgical pain [3, 4]. Chronic postsurgical pain is typically defined as pain that develops after a surgical procedure and persists at least three months [5, 6]. Chronic postsurgical pain is associated with lower patient satisfaction and higher societal and health care expenses due to resource-intensive revision surgery and long-term recovery [4, 7–10]. A comprehensive understanding of factors associated with poor pain outcomes is imperative for the development of a prediction model needed to identify patients at higher risk for chronic postsurgical pain [11, 12]. Although numerous preoperative and intra-operative factors have been studied, synthesizing the available evidence has yielded contradictory findings, perhaps related to certainty of evidence, merging data from short- and long-term outcomes, or pooling estimates from prospective and retrospective study designs [13–21]. Some authors did not perform meta-analysis due to heterogeneity in design and methods [14, 22–24]. Thus, we aimed to build from previous reviews and synthesize current evidence between preoperative and intraoperative factors associated with pain twelve months (primary outcome) and three and six months (secondary outcomes) after TKA. ## Methods We performed our systematic review and meta-analysis according to an a priori peer-reviewed protocol and a preprint [25, 26]. The study was registered in International Prospective Register of Systematic Reviews (PROSPERO; CRD42018079069) [26]. We followed Cochrane Handbook guidelines [27], and reported the study using the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) reporting guideline (S1 Checklist). ## Search strategy and data sources Two researchers (UO, MFL) and research librarians developed the search strategy with input from the research team [25]. The research librarian performed a systematic search for publications in MEDLINE (Ovid), Embase (Ovid), Cumulative Index to Nursing and Allied Health Literature (CINAHL; EBSCO), Cochrane Library and Physiotherapy Evidence Database between January 1, 2000, and February 6, 2023. No language restrictions were set. References were imported to Endnote X8 Software version 20.2.1 (Clarivate Analytics). ## Eligibility criteria We included peer-reviewed published studies that reported estimates of association between preoperative or intraoperative factors and pain at three, six and twelve months after TKA. Studies were eligible if participants were 18 years or older, diagnosed with osteoarthritis, and scheduled for primary TKA. Eligible study designs were prospective longitudinal observational studies and randomized clinical trials that provided estimates of association. Conference abstracts, retrospective studies, case-control studies, studies of uni-compartmental surgery and studies that lacked clear pain outcome measures were not eligible. Studies that merged data from mixed patient populations or did not report separate data for the osteoarthritis or TKA population were excluded ## Outcomes The primary outcome was pain at twelve months following TKA. Secondary outcomes were pain at three and six months. ## Study selection and data extraction We used a standardized data extraction form customized to the research question as explained in the published protocol [25] which included study design, country, participant characteristics, sample size, measures and outcomes, statistical analyses, and estimates of association. Two reviewers (UO, MFL) independently screened titles and abstracts for relevance, assessed full-text publications against eligibility criteria and assessed risk of bias. Disagreements were resolved by consensus or by consulting a third author (ED). ## Methodological quality The Quality in Prognosis Studies (QUIPS) tool [28] was used to systematically evaluate risk of bias in the retrieved studies according to the protocol [25]. The six QUIPS domains include study participation, study attrition, prognostic factor measurement, outcome measurement, confounding, and statistical analysis and reporting [27]. ## Certainty of evidence We assessed certainty of evidence using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) framework [29]. Two researchers (UO and MFL) judged certainty of evidence, with a third researcher involved in discussing cases of disagreement (ED). GRADEpro GDT (McMaster University) was used to manage and summarize the evidence. ## Statistical analysis We synthesised results from all included studies at three, six, and twelve months post-surgery according to our pre-specified protocol [25], with the exception that we used a multivariate random-effects meta-analysis that accounts for the sparse data (many factors relative to the number of studies), as in our recent review of factors for post-surgical function [30]. Further protocol deviations are noted below, in the discussion and in the Methods in the Supplement. The included studies reported associations as odds ratios (ORs), risk ratios (RRs), linear model coefficients (including differences), or correlations using discrete or continuous scales to measure factors and outcomes. Correlation coefficients were meta-analyzed on the arctangent scale [31], and estimates were back-transformed to the correlation scale for reporting. We expected within-study correlation and between-study heterogeneity and therefore used a multivariate random-effects model to estimate mean rather than common correlations between factors and pain. Heterogeneity was quantified by using I2 statistics. P scores were calculated to evaluate the certainty that the mean correlation for each factor is larger in magnitude than the mean correlations for all other factors [32]. We also explored how estimates may depend on the choice of model: we removed factors supported by few studies (to decrease the impact of sparsity) and compared estimates from the two multivariate models and univariate meta-analyses for consistency. We then performed sensitivity analyses on pain at twelve months, and excluded studies judged to have high risk of bias for each QUIPS domain and re-ran the multivariate meta-analysis (S4 Appendix). Statistical analyses were performed using Stata 16 (StataCorp LLC, College Station, Texas, USA). Mean correlations with $95\%$ confidence intervals (CIs) are reported. Hypothesis testing was not predefined, but 2-sided P values are reported for completeness. ## Results The search yielded 13,640 studies. After title and abstract screening, 406 studies were assessed in full text and 374 were ineligible, leaving 29 studies [33–61] with a total sample of 10,360 patients (Fig 1). Sample sizes ranged from 26 [43] to 5309 [50]. We excluded eight studies from analysis because attempts to obtain missing data from authors were unsuccessful or insufficient [62–69]. The search strategy, subject headings and keywords customized for all databases is presented in S8 Appendix and reasons for study exclusion are in S9 Appendix. **Fig 1:** *Flow chart of included studies.* In all, 61 preoperative and intraoperative factors were identified in the 29 studies [33–61]. All studies used prospective longitudinal observational designs, and most were single-center studies [33, 36–45, 48–51, 54, 55, 58–61] and conducted in European countries [33, 37, 39–48, 52, 57, 61]. No randomized trial met inclusion criteria. Mean age ranged from 63 [40] to 73 years [48], and the percentages of women in the samples varied from $49\%$ [58] to $95\%$ [40]. As shown in the Table 1, most studies used the Western Ontario and McMaster Universities Arthritis Index (WOMAC) to measure pain [34, 35, 37, 38, 47, 49, 51, 52, 58, 61]. **Table 1** | Study, country | Country | Design | Patients analyzed, No. | Data collection | Follow-up, mo | Baseline Age, y | Patients, No./Total No (%) Female Male | Patients, No./Total No (%) Female Male.1 | Analysis | Factors measured | Outcome measure | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Cremeans-Smith et al, 2016a [49] | United States | PC | 101 | | 3 | Mean, 69 | 75/110 (68) | 35/110 (32) | Hierarchical linear regression | Education (level), pain (WOMAC), Cortisol (level), anaesthesia type (general vs spinal) | WOMAC | | Lindner et al, 2018 [61] | Germany | PC | 61 | | 3 | Mean, 67 | 37/61 (61) | 24/61 (39) | Stepwise multiple linear regression | Pain (WOMAC) | WOMAC | | Lingard et al, 2007 [35] | UK, US, Canada, Australia | PC | 676 | 1997–1999 | 3 | Distress: median, 70 Non-distress: median, 71 | 574/676 (85) | 102/676 (15) | Repeated measures | Psychological distress (SF-36) | WOMAC | | Luo et al,2019 [59] | PC | PC | 471 | 2017–2018 | 3 | Mean, 64 | 357/471 (76) | 114/471 (24) | Pearson correlation | Sleep dysfunction (PSQI), daytime sleepiness (ESS), sleep quality (self-developed scale) | KSS | | Perruccio et al,2019 [60] | Canada | PC | 477 | 2014–2016 | 3 | Mean, 65 | 279/477 (58) | 198/477 (42) | Linear regression | Age (y), sex (men/women), BMI, comorbidity (AAOs comorbidity Scale), symptomatic joint count, pain (KOOS), low back pain (yes/no), depression (HADS) | KOOS | | Attal et al, 2012a [33] | France | PC | 81 | 2008–2011 | 6 | Mean, 69 | 58/89 (65) | 31/89 (35) | Stepwise logistic regression | Trail Making Time (TMT-B time) | Brief Pain Inventory (BPI) | | Bossman et al, 2017 [52] | Germany | PC | 47 | | 6 | Mean, 69 | 37/56 (66) | 19/56 (34) | Analysis of variance (bootstrap) | Age (y), sex (men/women), BMI, pain (WOMAC), conditioned pain modulation (pressure pain algometry), heart rate variability (SDNN), temporal summation (pin-prick stimulator), pain catastrophizing (PCS), Sympathetic/ parasympathetic activity (LogLF) | WOMAC | | Bruehl et al, 2023 [54] | US | PC | 91 | NR | 6 | Mean, 67 | 57 (63) | 34 (37) | Generalized linear density ratio model | Ischemia duration (blood sample), oxidative stress (blood sample) | MPQ-2 | | Bugada et al, 2017 [57] | Italy | PC | 563 | 2012–2015 | 6 | Median, 72 | 421/606 (69) | 185/606 (31) | Logistic regression | Age (y), | NRS | | Chen et al, 2021 [55] | China | PC | 220 | 2019–2020 | 6 | Pain ≥4:median, 70Pain <4: median,71 | 102/220 (46) | 118/220 (54) | Logistic regression | Age (y), serum angiotensin II Type 2 receptor (AT2R), temporal summation (PD-Q), Anxiety and depression (HADS), disability (WOMAC). pain expectation (NRS), pain sites (count) | VAS | | Edwards et al, 2022 [56] | US | PC | 248 | | 6 | Mean, 65 | 147 (59.5) | 101 (40.5) | Backwards selection regression | Pain (BPI), State catastrophizing (PCS), catastrophizing (PCS), opioid use, sleep efficiency (PSQI), other chronic pain sites (count), painful areas (count), anxiety (PROMIS), agreeableness (NEO Inventory) | BPI | | Engel et al, 2014 [58] | US | Case-control | 54 | | 6 | Mean, 68 | 36/74 (49) | 38/74 (51) | Multiple hierarchical regression | Arthritis helplessness (AHI), coping efficacy (scale) | WOMAC | | Escobar et al, 2007 [47] | Spain | PC | 640 | 1999–2000 | 6 | Mean, 72 | 473/640 (74) | 167/640 (26) | General linear model | Age (y), sex (men/women), social support (yes/no), comorbidity (CCI), pain (WOMAC), low back pain (yes/no), mental health (SF-36) | WOMAC | | Fitz-simmons et al, 2018 [53] | Canada | PC | 74 | 2014 | 6 | Mean, 65 | 67/99 (68) | 32/99 (32) | Multiple linear regression | Suspected neuropathic pain (SNEP), Preoperative pain (ICOAP), Pain catastrophizing (PCS), depression (PHQ, comorbidity (count) | ICOAP | | Pua et al, 2019 [50] | Singapore | PC | 4026 | 2013–2017 | 6 | Mean, 68 | 3003/4026 (75) | 1023/4026 (25) | Proportion-al odds regression | Age (y), Sex (Men/women), BMI, education (primary, secondary, tertiary), ethnicity (Chinese, Indian, Malay, other), social support (yes/no), comorbidities (yes/no), contralateral knee pain (KSS), pain (OKQ), Knee extension and flexion (goniometer), physical function (categories), depression (SF-36) | OKQ | | Yang et al, 2019 [51] | US | PC | 107 | 2010–2011 | 6 | Mean, 65 | 55/107 (51) | 52/107 (49) | Multiple logistic regression | Mental health (SF-36), Pain catastrophizing (PCS), use device (yes/no) | WOMAC | | Attal et al, 2012a [33] | France | PC | 69 | 2008–2011 | 12 | Mean, 69 | 58/89 (65) | 31/89 (35) | Stepwise logistic regression | Recall (ROCF) | BPI | | Dave et al, 2017 [34] | United States | PC | 241 | 2012–2014 | 12 | Mean, 67 | 146/241 (61) | 95/241 (39) | Poisson regression | Painful body regions (count), pain (WOMAC), pain catastrophizing (PCS) | WOMAC | | Dowsey et al, 2012 [36] | Australia | PC | 473 | 2006–2007 | 12 | Mean, 71 | 331/473 (70) | 142/473 (30) | Multivariate linear regression | Age (y), sex (men/women), BMI, comorbidity (CCI), pain (IKSS), physical function (IKSS), mental health (SF-12), Osteoarthritis severity (K-L grade), cruciate retaining, patella resurface | IKSS | | Getachew et al, 2020 [39] | Norway | PC | 185 | 2012–2014 | 12 | Mean, 68 | 137/202 (68) | 65/202 (32) | Multiple logistic regression | Age (y), Sex (men/women), Pain (NRS), fatigue (LFS)Sleep quality (PSQI), depression (HAD) | BPI | | Giordiano et al, 2020 [41] | Denmark | PC | 136 | NR | 12 | High pain relief: mean, 69Low pain relief: mean, 68 | 82/136 (60) | 54/136 (40) | Linear regression | Pain (VAS), circulating micromiRna-146a-5p (venous blood) | VAS | | Hardy et al, 2022 [48] | France | PC | 103 | 2014–2015 | 12 | Mean, 73 | 67/36 | 65/35 | Logistic regression | Catastrophizing (PCS) | VAS | | Kornilov et al, 2018 [40] | Russia | PC | 79 | 2014 | 12 | Mean, 63 | 75/79 (95) | 4/79 (5) | Logistic regression | Pain (BPI),physical activity (HUNT 2 physical activity score) | BPI | | Lingard et al,2007a [35] | UK, US, Canada, Australia | PC | 676 | 1997–1999 | 12 | Distress: median, 70Non-distress: median, 71 | 574/676 (85) | 102/676 (15) | Repeated measures | Psychological distress (SF-36) | WOMAC | | Petersen et al, 2015 [42] | Denmark | PC | 78 | | 12 | Low pain: mean, 68High pain group: mean, 72 | 50/78 (59) | 28/78 (41) | Multi-variate logistic regression | Pain (VAS),temporal summation (von Frey stimulator) | VAS | | Petersen et al, 2017 [44] | Denmark | PC | 130 | | 12 | Chronic pain: mean, 69 Normal recovery: mean, 68 | Chronic pain: 14/19 (74)Normal recovery: 59/105 (56) | Chronic pain:5/19 (26)Normal recovery: 46/105 (44) | Linear regression | Temporal summation (von Frey stimulator), K-L grade, warm detection-/heat pain threshold | VAS | | Petersen et al, 2020 [43] | Denmark | PC | 26 | 2011–2012 | 12 | High pain: Mean, 64Low pain: mean, 70 | 14/26 (54) | 12/26 (46) | Pearson correlation | Synovial membrane thickness (CE-MRI), degree perfusion (voxels*ME), volume perfusion (IRE), synovitis severity | VAS | | Tilbury et al, 2018 [45] | Netherlands | PC | 146 | 2011–2012 | 12 | Mean, 67 | 101/146 (69) | 87/146 (31) | Multi-variate linear regression | BMI, mental health (SF-36), outcome expectancies (HSS) | KOOS | | Sullivan et al, 2011 [38] | Canada | PC | 120 | | 12 | 67 (mean) | 73/120 (61) | 47/120 (39) | Multiple regression | Age (y), sex (men/women), BMI, comorbidity (CCI), pain (WOMAC), pain catastrophizing (PCS), depression (PHQ-9), kinesophobia (TSK), surgery duration (minutes) | WOMAC | | Van de Water et al, 2019 [46] | Netherlands | PC | 559 | 2012–2015 | 12 | Mean, 67 | 378/559 (68) | 181/559 (32) | Multi-variate linear regression | Pain (KOOS),K-L grade | KOOS | | Wylde et al, 2012 [37] | United Kingdom | PC | 220 | | 12 | Median, 70 | 136/220 (62) | 84/220 (38) | Ordinary least squares regression | Age (y), sex (men/women), comorbidity (SCQ), pain (WOMAC), depression (HADS), anxiety (HADS), pain-self efficacy (PSEQ) | WOMAC | We present separate estimates of mean correlations between preoperative and intraoperative factors and the three-, six- and twelve- month pain outcomes in multivariate meta-analysis (Figs 2–4). Multivariate meta-analytical estimates of correlation at each postoperative follow-up time are shown in Fig 2 and S1 Appendix. Descriptions of potential inconsistencies at three, six and twelve months are in S2 Appendix, and univariate meta-analyses for associations between individual factors and the outcomes are in S4 Appendix. Results from sensitivity analysis are presented in S1 Appendix. We provide a full glossary of labels for included factors in the Table in S5 Appendix. We report all estimates between preoperative and intraoperative factors and pain during the year (three, six and twelve months) after TKA as mean correlations, with positive correlations indicating more postoperative pain. **Fig 2:** *Forest plot of factors associated with pain at twelve months.* **Fig 3:** *Forest plot of factors associated with pain at six months.* **Fig 4:** *Forest plot of factors associated with pain at three months.* A total of 15 studies with 3,241 participants [33–46, 48] reported estimates for 34 factors correlated with pain twelve months after TKA (Fig 2). The two most common factors were preoperative pain [34, 36–42, 46] reported in nine studies and mental health (including anxiety, depression, psychological distress) reported in six studies [35–39, 45]. Most of these studies were judged as having high risk of bias on one or more domain (S6 Appendix). Mean correlation between preoperative pain catastrophizing and pain twelve months after TKA was estimated to be 0.36 ($95\%$ CI, 0.24 to 0.47; $P \leq .001$; P score = $80.2\%$; three studies [34, 38, 48]; moderate-certainty evidence and substantial heterogeneity among reported estimates of association [I2 = $72.4\%$], while mean correlation for more temporal summation was estimated as 0.21 ($95\%$ CI, 0.05 to 0.36; $P \leq .000$; P score = $61.1\%$; two studies [42, 44]; very low-certainty evidence and heterogeneity among reported estimates of association might not be important [I2 = $0\%$]), more symptomatic joints was estimated to be 0.15 ($95\%$ CI, 0.08 to 0.23; $P \leq .001$; P score = $51.3\%$; two studies [34, 37]; moderate-certainty evidence and heterogeneity among reported estimates of association might not be important [I2 = $0\%$]), and more preoperative pain was estimated to be 0.13 ($95\%$ CI, 0.06 to 0.19; $P \leq .001$; P score = $44.6\%$; nine studies [34, 36–42, 46]; very low-certainty evidence and considerable heterogeneity among reported estimates of association [I2 = $97.0\%$]). In contrast, mean correlation for more severe osteoarthritis and pain at twelve months was negative. The estimated correlation was -0.15 ($95\%$ CI, -0.23 to -0.08; $P \leq .001$; P score = $51.6\%$; three studies [36, 44, 46]; low-certainty evidence and heterogeneity among reported estimates of association might not be important [I2 = $0\%$]), Results from the prespecified sensitivity analysis (S4 Appendix), estimated a mean correlation of 0.38 ($95\%$ CI, 0.04 to 0.64) between pain catastrophizing and more pain, compared to 0.28 ($95\%$ CI, 0.11 to 0.43) when including all studies. The mean correlation estimate was 0.15 ($95\%$ CI 0.06 to 0.24) for symptomatic joints compared to 0.15 ($95\%$ CI 0.07 to 0.23) when including all studies. The mean correlation estimate was 0.16 ($95\%$ CI -0.00 to 0.25) for level of pain compared to 0.13 ($95\%$ CI 0.06 to 0.19) when including all studies. Mean correlation estimate was -0.15 ($95\%$ CI -0.24 to -0.06) for more severe osteoarthritis compared to -0.15 ($95\%$ CI -0.23 to -0.08) when including all studies. The association for temporal summation identified in the multivariate meta-analysis was obscured in the sensitivity analysis as the statistical analysis domain was judged high risk of bias. There was 11 studies with 6,078 participants that included estimates for 34 potential factors associated with pain six months after TKA (Fig 3) [33, 47, 50–58]. Mean correlation with preoperative pain was 0.20 ($95\%$ CI 0.12 to 0.28; $P \leq .000$; P score = $66.1\%$; five studies [47, 50, 52, 53, 56]; low-certainty evidence and heterogeneity among reported estimates of association may not be important [I2 = $37.6\%$]). Mean correlation with better mental health was -0.13 ($95\%$ CI -0.24 to -0.02; $$P \leq 0.01$$; P score = $49.1\%$; six studies [47, 50, 52, 53, 56]; moderate-certainty evidence and heterogeneity among reported estimates of association may not be important [I2 = $29.1\%$]). For the other secondary outcome, pain three months after TKA, five studies with 1786 patients provided pain outcome data at three months after TKA for 14 potential factors (Fig 4) [35, 49, 59–61]; Mean correlation with preoperative pain was 0.27 ($95\%$ CI 0.13 to 0.39; $p \leq .001$; P score = $81.0\%$; three studies [49, 60, 61]; low-certainty evidence and heterogeneity among reported estimates of association may not be important [I2 = $0\%$]). Meta-analytical estimates for the other factors do not exclude the possibility of no correlation with pain at three, six, and twelve months. It is plausible that these factors are uncorrelated with pain, but also possible that important correlations exist but cannot be estimated with much precision. We compared meta-analytic estimates from three models and there was reasonable consistency between the univariate and multivariate meta-analysis for all factors with respect to direction of association (S2 Appendix). Decisions regarding risk of bias for each QUIPS domain are shown in S4 Fig in S1 Appendix. We judged the included studies to be generally low risk of bias for prognostic factor measurement ($$n = 16$$) and outcome measurement ($$n = 21$$). In contrast, some studies were judged high risk of bias for study participation ($$n = 12$$), study attrition ($$n = 16$$), and statistical analysis ($$n = 13$$). Full details of our certainty of evidence (GRADE) judgements are provided in S7 Appendix. Risk of bias and imprecision were the most common reasons for downgrading the certainty of evidence. ## Discussion To our knowledge this is the first systematic review and meta-analysis examining factors correlated with pain at three, six and twelve months after TKA that also evaluated certainty of evidence. For the primary outcome at twelve months and based on at total sample of 3,241 patients, we estimated that pain catastrophizing, more symptomatic joints, and higher level of preoperative pain were correlated with worse pain outcomes, while more severe radiographic osteoarthritis were correlated with better pain outcome twelve months later. Our findings suggest that more severe preoperative pain is correlated with worse pain outcomes and that better mental health is associated with better pain outcomes at three and six months. It is worth noting that our findings do not indicate that the individual patient with a poor risk profile will experience chronic postsurgical pain if they undergo TKA surgery. Findings simply suggest that the identified factors were correlated with less or worse pain in an absolute sense. Thus, our results should be interpreted accordingly. We estimated moderate-certainty evidence that pain catastrophizing is correlated with worse pain outcomes at twelve months. The correlation was larger in sensitivity analysis where we removed a study with high risk of bias. Our findings are similar to results from prior systematic reviews or meta-analyses [18, 22, 70]. However, our study differs in two critical ways: our results are entirely based on prospective studies, and we did not pool results from studies with short-term and longer-term follow-up. Efficacy for cognitive behavioral therapy to enhance skills for coping with pain remains unknown [71, 72], and still TKA surgery may be the most effective intervention, giving more pain relief, than non-operative treatment. We found moderate-certainty evidence that a higher number of symptomatic joints was associated with more pain twelve months after TKA, with equal correlation in the sensitivity analysis. This result is supported by findings from a previous univariate meta-analysis that identified multiple painful sites as a factor influencing the pain outcome [18] but the association was not significant in the multivariate meta-analysis. Degenerated cartilage and subchondral bone are removed during surgery; however, pain may also be generated from other structures or tissue surrounding the knee, which might influence pain outcome. We found positive correlations between more preoperative pain and pain severity at twelve months (very-low certainty evidence). Positive correlations were also identified for the secondary outcomes at three and six months (low-certainty evidence). Our findings are in consistency with other reviews and meta-analysis [13, 18]. There is emerging evidence that improvement in pain for most patients usually follows a steep trajectory in the first three to six postoperative months, before pain levels seems to plateau at twelve months [73–75]. Accordingly, we have added new evidence on preoperative factors correlated with adverse pain outcomes at three, six and twelve months after TKA. There were no intraoperative factors that correlated with pain outcomes at three, six or twelve months. We found a negative correlation between severity of osteoarthritis and pain at twelve months, i.e., the more severe the osteoarthritis before surgery, the lower the pain severity twelve months later. Although the evidence was rated as low-certainty, the correlation persisted in the sensitivity analysis. Another meta-analysis has shown that patients with mild radiographic osteoarthritis reported more pain after TKA [16]. In contrast to our study, evidence was not graded and retrospective study designs with follow-up from one to six years were included. Results from our and their meta-analyses indicate that patients with severe osteoarthritis might gain more from TKA surgery than patients with less severe osteoarthritis. Non-operative treatment options should be considered to all patients with low-grade radiographic OA findings before surgery [76]. This study had many strengths, including up-to-date robust methods that followed Cochrane Handbook guidelines with descriptions in a pre-specified peer-reviewed protocol [25], a preprint [26], assessing risk of bias using QUIPS, and judging certainty of evidence using GRADE. We included only longitudinal prospective studies with associations reported at pre-defined time points in the first postoperative year and applied multivariate meta-analysis when the number of variables was large relative to number of studies [26]. There are some limitations that need to be addressed. First, we included studies that were largely heterogeneous for measurement of factors. Less heterogeneity existed in postoperative pain measures, with WOMAC being the most common. We used a number of exploratory statistics to estimate associations. Researchers either opt for narrow eligibility criteria and risk excluding potentially useful evidence, or wider eligibility criteria that require appropriate methods to address the heterogeneity [27]. We chose the latter, but results should be interpreted carefully due to underlying heterogeneity. Some included studies had large sample sizes that resulted in narrow CIs, and I2 for the pooled results tend to be very high and might be misleading [29]. Our estimates may also be biased by including several studies judged high risk of bias. To address this issue, we performed pre-specified sensitivity analyses excluding studies with high risk of bias for each QUIPS domain. We were unable to perform planned analyses of non-reporting bias and small study effects, or planned subgroup analyses, because the number of included studies did not meet our pre-specified criterion. We had also planned leave-one-study-out sensitivity analysis to explore the influence of each study on meta-analysis results, but this was not feasible. Many of the studies in our review had limitations that resulted in downgrading our certainty of the evidence. This does not necessarily indicate that those studies were of poor quality, but that important areas requiring documentation according to methodological standards were not reported. The importance of consistent reporting following these standards should be stressed so that evidence can be evaluated with high certainty. We suggest that researchers design studies using tools such as QUIPS to minimize risk of bias. We did not address the magnitude of change in pain score, which probably would be the most interesting for the patients, but only the degree of pain at twelve months. ## Conclusions Our findings suggest that the preoperative factors of pain catastrophizing, symptomatic joints, pain, and radiographic osteoarthritis are correlated with pain one year after TKA. Pain are correlated with the six- and three- months pain outcomes, while mental health is correlated with pain at six months. However, our result highlights the need for further investigation on several factors that have been evaluated only once or in studies with small sample sizes. These factors should be considered when developing predictive models to identify patients most likely to experience chronic post-surgical pain. Accurately identifying factors associated with the pain outcome will be crucial for the development of effective predictive models. ## References 1. Price AJ, Alvand A, Troelsen A, Katz JN, Hooper G, Gray A. **Knee replacement**. *Lancet* (2018.0) **392** 1672-82. DOI: 10.1016/S0140-6736(18)32344-4 2. **National Inpatient Sample (NIS), 2019 [Internet].** 3. Beswick AD, Wylde V, Gooberman-Hill R, Blom A, Dieppe P. **What proportion of patients report long-term pain after total hip or knee replacement for osteoarthritis? A systematic review of prospective studies in unselected patients**. *BMJ Open* (2012.0) **2** e000435. DOI: 10.1136/bmjopen-2011-000435 4. Petersen KK, Simonsen O, Laursen MB, Nielsen TA, Rasmussen S, Arendt-Nielsen L. **Chronic postoperative pain after primary and revision total knee arthroplasty**. *Clin J Pain* (2015.0) **31** 1-6. DOI: 10.1097/AJP.0000000000000146 5. Schug SA, Lavand’homme P, Barke A, Korwisi B, Rief W, Treede RD. **The IASP classification of chronic pain for ICD-11: chronic postsurgical or posttraumatic pain**. *Pain* (2019.0) **160** 45-52. DOI: 10.1097/j.pain.0000000000001413 6. Treede RD, Rief W, Barke A, Aziz Q, Bennett MI, Benoliel R. **Chronic pain as a symptom or a disease: the IASP Classification of Chronic Pain for the International Classification of Diseases (ICD-11).**. *Pain* (2019.0) **160** 19-27. DOI: 10.1097/j.pain.0000000000001384 7. Kahlenberg CA, Nwachukwu BU, McLawhorn AS, Cross MB, Cornell CN, Padgett DE. **Patient Satisfaction After Total Knee Replacement: A Systematic Review**. *HSS J* (2018.0) **14** 192-201. DOI: 10.1007/s11420-018-9614-8 8. Pinedo-Villanueva R, Kolovos S, Maronga C, Delmestri A, Howells N, Judge A. **Primary care consultations and pain medicine prescriptions: a comparison between patients with and without chronic pain after total knee replacement**. *BMC Musculoskelet Disord* (2022.0) **23** 548. DOI: 10.1186/s12891-022-05492-6 9. Nakano N, Shoman H, Olavarria F, Matsumoto T, Kuroda R, Khanduja V. **Why are patients dissatisfied following a total knee replacement? A systematic review.**. *Int Orthop* (2020.0) **44** 1971. DOI: 10.1007/s00264-020-04607-9 10. Zaffagnini S, Di Paolo S, Meena A, Alesi D, Zinno R, Barone G. **Causes of stiffness after total knee arthroplasty: a systematic review**. *Int Orthop* (2021.0) **45** 1983-99. DOI: 10.1007/s00264-021-05023-3 11. Kent P, Cancelliere C, Boyle E, Cassidy JD, Kongsted A. **A conceptual framework for prognostic research**. *BMC Med Res Methodol* (2020.0) **20** 172. DOI: 10.1186/s12874-020-01050-7 12. Riley RD, Moons KGM, Snell KIE, Ensor J, Hooft L, Altman DG. **A guide to systematic review and meta-analysis of prognostic factor studies**. *BMJ* (2019.0) **364** k4597. DOI: 10.1136/bmj.k4597 13. Harmelink KEM, Zeegers A, Hullegie W, Hoogeboom TJ, Nijhuis-van der Sanden MWG, Staal JB. **Are There Prognostic Factors for One-Year Outcome After Total Knee Arthroplasty? A Systematic Review**. *J Arthroplasty* (2017.0) **32** 3840-53 e1. DOI: 10.1016/j.arth.2017.07.011 14. Vissers MM, Bussmann JB, Verhaar JA, Busschbach JJ, Bierma-Zeinstra SM, Reijman M. **Psychological factors affecting the outcome of total hip and knee arthroplasty: a systematic review**. *Semin Arthritis Rheum* (2012.0) **41** 576-88. DOI: 10.1016/j.semarthrit.2011.07.003 15. Santaguida PL, Hawker GA, Hudak PL, Glazier R, Mahomed NN, Kreder HJ. **Patient characteristics affecting the prognosis of total hip and knee joint arthroplasty: a systematic review**. *Can J Surg* (2008.0) **51** 428-36. PMID: 19057730 16. Shohat N, Heller S, Sudya D, Small I, Khawalde K, Khatib M. **Mild radiographic osteoarthritis is associated with increased pain and dissatisfaction following total knee arthroplasty when compared with severe osteoarthritis: a systematic review and meta-analysis**. *Knee Surg Sports Traumatol Arthrosc* (2022.0) **30** 965-81. DOI: 10.1007/s00167-021-06487-x 17. Pozzobon D, Ferreira PH, Blyth FM, Machado GC, Ferreira ML. **Can obesity and physical activity predict outcomes of elective knee or hip surgery due to osteoarthritis? A meta-analysis of cohort studies**. *BMJ Open.* (2018.0) **8** e017689. DOI: 10.1136/bmjopen-2017-017689 18. Lewis GN, Rice DA, McNair PJ, Kluger M. **Predictors of persistent pain after total knee arthroplasty: a systematic review and meta-analysis**. *Br J Anaesth* (2015.0) **114** 551-61. DOI: 10.1093/bja/aeu441 19. Khatib Y, Madan A, Naylor JM, Harris IA. **Do Psychological Factors Predict Poor Outcome in Patients Undergoing TKA? A Systematic Review**. *Clin Orthop Relat Res* (2015.0) **473** 2630. DOI: 10.1007/s11999-015-4234-9 20. Duan G, Liu C, Lin W, Shao J, Fu K, Niu Y. **Different Factors Conduct Anterior Knee Pain Following Primary Total Knee Arthroplasty: A Systematic Review and Meta-Analysis.**. *J Arthroplasty* (2018.0) **33** 1962-71 e3. DOI: 10.1016/j.arth.2017.12.024 21. Youlden DJ, Dannaway J, Enke O. **Radiographic severity of knee osteoarthritis and its relationship to outcome post total knee arthroplasty: a systematic review**. *ANZ J Surg* (2020.0) **90** 237-42. DOI: 10.1111/ans.15343 22. Burns LC, Ritvo SE, Ferguson MK, Clarke H, Seltzer Z, Katz J. **Pain catastrophizing as a risk factor for chronic pain after total knee arthroplasty: a systematic review**. *J Pain Res* (2015.0) **8** 21-32. DOI: 10.2147/JPR.S64730 23. Lungu E, Vendittoli PA, Desmeules F. **Preoperative Determinants of Patient-reported Pain and Physical Function Levels Following Total Knee Arthroplasty: A Systematic Review.**. *Open Orthop J* (2016.0) **10** 213-31. DOI: 10.2174/1874325001610010213 24. van Jonbergen HP, Reuver JM, Mutsaerts EL, Poolman RW. **Determinants of anterior knee pain following total knee replacement: a systematic review**. *Knee Surg Sports Traumatol Arthrosc* (2014.0) **22** 478-99. DOI: 10.1007/s00167-012-2294-x 25. Olsen U, Lindberg MF, Denison EM, Rose CJ, Gay CL, Aamodt A. **Predictors of chronic pain and level of physical function in total knee arthroplasty: a protocol for a systematic review and meta-analysis**. *BMJ Open* (2020.0) **10** e037674. DOI: 10.1136/bmjopen-2020-037674 26. Rose CJ, Olsen U, Lindberg M, Denison E-L, Aamodt A, Lerdal A. **A new multivariate meta-analysis model for many variates and few studies.**. *arXiv. GIT revision: 72ca65b. arXiv*. DOI: 10.48550/arXiv.2009.11808 27. Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page M. **version 6.2**. *Cochrane Handbook for Systematic Reviews of Interventions* (2020.0) 28. Hayden JA, van der Windt DA, Cartwright JL, Cote P, Bombardier C. **Assessing bias in studies of prognostic factors**. *Ann Intern Med* (2013.0) **158** 280-6. DOI: 10.7326/0003-4819-158-4-201302190-00009 29. Iorio A, Spencer FA, Falavigna M, Alba C, Lang E, Burnand B. **Use of GRADE for assessment of evidence about prognosis: rating confidence in estimates of event rates in broad categories of patients**. *BMJ* (2015.0) **350** h870. DOI: 10.1136/bmj.h870 30. Olsen U, Lindberg MF, Rose C, Denison E, Gay C, Aamodt A. **Factors Correlated With Physical Function 1 Year After Total Knee Arthroplasty in Patients With Knee Osteoarthritis: A Systematic Review and Meta-analysis**. *JAMA Netw Open* (2022.0) **5** e2219636. DOI: 10.1001/jamanetworkopen.2022.19636 31. Borenstein M, Hedges LV, Higgins JP, Rothstein HR. (2009.0) 32. Rucker G, Schwarzer G. **Ranking treatments in frequentist network meta-analysis works without resampling methods**. *BMC Med Res Methodol* (2015.0) **15** 58. DOI: 10.1186/s12874-015-0060-8 33. Attal N, Masselin-Dubois A, Martinez V, Jayr C, Albi A, Fermanian J. **Does cognitive functioning predict chronic pain? Results from a prospective surgical cohort**. *Brain* (2014.0) **137** 904-17. DOI: 10.1093/brain/awt354 34. Dave AJ, Selzer F, Losina E, Usiskin I, Collins JE, Lee YC. **The association of pre-operative body pain diagram scores with pain outcomes following total knee arthroplasty**. *Osteoarthritis Cartilage* (2017.0) **25** 667-75. DOI: 10.1016/j.joca.2016.12.013 35. Lingard EA, Riddle DL. **Impact of psychological distress on pain and function following knee arthroplasty**. *J Bone Joint Surg Am* (2007.0) **89** 1161-9. DOI: 10.2106/JBJS.F.00914 36. Dowsey MM, Nikpour M, Dieppe P, Choong PF. **Associations between pre-operative radiographic changes and outcomes after total knee joint replacement for osteoarthritis**. *Osteoarthritis Cartilage* (2012.0) **20** 1095-102. DOI: 10.1016/j.joca.2012.05.015 37. Wylde V, Dixon S, Blom AW. **The role of preoperative self-efficacy in predicting outcome after total knee replacement**. *Musculoskeletal Care* (2012.0) **10** 110-8. DOI: 10.1002/msc.1008 38. Sullivan M, Tanzer M, Reardon G, Amirault D, Dunbar M, Stanish W. **The role of presurgical expectancies in predicting pain and function one year following total knee arthroplasty**. *Pain* (2011.0) **152** 2287-93. DOI: 10.1016/j.pain.2011.06.014 39. Getachew M, Lerdal A, Smastuen MC, Gay CL, Aamodt A, Tesfaye M. **High levels of preoperative pain and fatigue are red flags for moderate-severe pain 12 months after total knee arthroplasty-A longitudinal cohort study**. *Musculoskeletal Care* (2021.0) **19** 186-92. DOI: 10.1002/msc.1522 40. Kornilov N, Lindberg MF, Gay C, Saraev A, Kuliaba T, Rosseland LA. **Higher physical activity and lower pain levels before surgery predict non-improvement of knee pain 1 year after TKA**. *Knee Surg Sports Traumatol Arthrosc* (2018.0) **26** 1698-708. DOI: 10.1007/s00167-017-4713-5 41. Giordano R, Petersen KK, Andersen HH, Lichota J, Valeriani M, Simonsen O. **Preoperative serum circulating microRNAs as potential biomarkers for chronic postoperative pain after total knee replacement**. *Mol Pain* (2020.0) **16** 1744806920962925. DOI: 10.1177/1744806920962925 42. Petersen KK, Arendt-Nielsen L, Simonsen O, Wilder-Smith O, Laursen MB. **Presurgical assessment of temporal summation of pain predicts the development of chronic postoperative pain 12 months after total knee replacement**. *Pain* (2015.0) **156** 55-61. DOI: 10.1016/j.pain.0000000000000022 43. Petersen KK, Arendt-Nielsen L, Vela J, Skou ST, Eld M, Al-Mashkur NM. **Less Severe Preoperative Synovitis is Associated With Higher Self-reported Pain Intensity 12 Months After Total Knee Arthroplasty-An Exploratory Prospective Observational Study**. *Clin J Pain.* (2020.0) **36** 34-40. DOI: 10.1097/AJP.0000000000000768 44. Petersen KK, Simonsen O, Laursen MB, Arendt-Nielsen L. **The Role of Preoperative Radiological Severity, Sensory Testing, and Temporal Summation on Chronic Postoperative Pain following Total Knee Arthroplasty**. *Clin J Pain* (2017.0) **34** 193-7 45. Tilbury C, Haanstra TM, Verdegaal SHM, Nelissen R, de Vet HCW, Vliet Vlieland TPM. **Patients’ pre-operative general and specific outcome expectations predict postoperative pain and function after total knee and total hip arthroplasties**. *Scand J Pain* (2018.0) **18** 457-66. DOI: 10.1515/sjpain-2018-0022 46. van de Water R, Leichtenberg C, Nelissen R, Kroon H, Kaptijn H, Onstenk R. **Preoperative Radiographic Osteoarthritis Severity Modifies the Effect of Preoperative Pain on Pain/Function After Total Knee Arthroplasty: Results at 1 and 2 Years Postoperatively.**. *J Bone Joint Surgery Am* (2019.0) **101** 879 47. Escobar A, Quintana JM, Bilbao A, Azkarate J, Guenaga JI, Arenaza JC. **Effect of patient characteristics on reported outcomes after total knee replacement.**. *Rheumatology (Oxford).* (2007.0) **46** 112-9. DOI: 10.1093/rheumatology/kel184 48. Hardy A, Sandiford MH, Menigaux C, Bauer T, Klouche S, Hardy P. **Pain catastrophizing and pre-operative psychological state are predictive of chronic pain after joint arthroplasty of the hip, knee or shoulder: results of a prospective, comparative study at one year follow-up**. *Int Orthop* (2022.0) **46** 2461-9. DOI: 10.1007/s00264-022-05542-7 49. Cremeans-Smith JK, Greene K, Delahanty DL. **Physiological Indices of Stress Prior to and Following Total Knee Arthroplasty Predict the Occurrence of Severe Post-Operative Pain.**. *Pain Med* (2016.0) **17** 970-9. DOI: 10.1093/pm/pnv043 50. Pua YH, Poon CL, Seah FJ, Thumboo J, Clark RA, Tan MH. **Predicting individual knee range of motion, knee pain, and walking limitation outcomes following total knee arthroplasty**. *Acta Orthop* (2019.0) **90** 179-86. DOI: 10.1080/17453674.2018.1560647 51. Yang HY, Losina E, Lange JK, Katz JN, Collins JE. **Longitudinal Trajectories of Pain and Function Improvement Following Total Knee Replacement.**. *ACR Open Rheumatol* (2019.0) **1** 308-17. DOI: 10.1002/acr2.1041 52. Bossmann T, Brauner T, Wearing S, Horstmann T. **Predictors of chronic pain following total knee replacement in females and males: an exploratory study**. *Pain Manag* (2017.0) **7** 391-403. DOI: 10.2217/pmt-2017-0023 53. Fitzsimmons M, Carr E, Woodhouse L, Bostick GP. **Development and Persistence of Suspected Neuropathic Pain After Total Knee Arthroplasty in Individuals With Osteoarthritis.**. *PM R* (2018.0) **10** 903-9. DOI: 10.1016/j.pmrj.2018.01.010 54. Bruehl S, Milne G, Schildcrout J, Shi Y, Anderson S, Shinar A. **Perioperative oxidative stress predicts subsequent pain-related outcomes in the 6 months after total knee arthroplasty**. *Pain* (2023.0) **164** 111-8. DOI: 10.1097/j.pain.0000000000002670 55. Chen F, Gao W, Hu J, Yang X, Chai X, Wang D. **Preoperative angiotensin II type 2 receptor is a predictor for developing chronic post-surgical pain after total knee arthroplasty surgery**. *Life Sci* (2021.0) **278** 119654. DOI: 10.1016/j.lfs.2021.119654 56. Edwards RR, Campbell C, Schreiber KL, Meints S, Lazaridou A, Martel MO. **Multimodal prediction of pain and functional outcomes 6 months following total knee replacement: a prospective cohort study**. *BMC Musculoskelet Disord* (2022.0) **23** 302. DOI: 10.1186/s12891-022-05239-3 57. Bugada D, Allegri M, Gemma M, Ambrosoli AL, Gazzerro G, Chiumiento F. **Effects of anaesthesia and analgesia on long-term outcome after total knee replacement: A prospective, observational, multicentre study**. *Eur J Anaesthesiol* (2017.0) **34** 665-72. DOI: 10.1097/EJA.0000000000000656 58. Engel C, Hamilton NA, Potter PT, Zautra AJ. **Impact of two types of expectancy on recovery from total knee replacement surgery (TKR) in adults with osteoarthritis**. *Behav Med* (2004.0) **30** 113-23. DOI: 10.3200/BMED.30.3.113-123 59. Luo ZY, Li LL, Wang D, Wang HY, Pei FX, Zhou ZK. **Preoperative sleep quality affects postoperative pain and function after total joint arthroplasty: a prospective cohort study**. *J Orthop Surg Res* (2019.0) **14** 378. DOI: 10.1186/s13018-019-1446-9 60. Perruccio A, Fitzpatrick J, Power J, Gandhi R, Rampersaud Y, Mahomed N. **The effects of depression, low back pain and comorbidities on pain after total knee arthroplasty for osteoarthritis are modified by sex**. *Arthritis Care Res* (2019.0) 61. Lindner M, Nosseir O, Keller-Pliessnig A, Teigelack P, Teufel M, Tagay S. **Psychosocial predictors for outcome after total joint arthroplasty: a prospective comparison of hip and knee arthroplasty**. *BMC Musculoskelet Disord* (2018.0) **19** 159. DOI: 10.1186/s12891-018-2058-y 62. Amusat N, Beaupre L, Jhangri GS, Pohar SL, Simpson S, Warren S. **Diabetes that impacts on routine activities predicts slower recovery after total knee arthroplasty: an observational study**. *J Physiother* (2014.0) **60** 217-23. DOI: 10.1016/j.jphys.2014.09.006 63. Lingard EA, Katz JN, Wright EA, Sledge CB, Kinemax Outcomes G. **Predicting the outcome of total knee arthroplasty**. *J Bone Joint Surg Am* (2004.0) **86** 2179-86. DOI: 10.2106/00004623-200410000-00008 64. Papakostidou I, Dailiana ZH, Papapolychroniou T, Liaropoulos L, Zintzaras E, Karachalios TS. **Factors affecting the quality of life after total knee arthroplasties: a prospective study.**. *BMC Musculoskelet Disord* (2012.0) **13** 116. DOI: 10.1186/1471-2474-13-116 65. Belford K, Gallagher N, Dempster M, Wolfenden M, Hill J, Blaney J. **Psychosocial predictors of outcomes up to one year following total knee arthroplasty**. *Knee.* (2020.0) **27** 1028-34. DOI: 10.1016/j.knee.2020.03.006 66. Sharma S, Kumar V, Sood M, Malhotra R. **Effect of Preoperative Modifiable Psychological and Behavioural Factors on Early Outcome Following Total Knee Arthroplasty in an Indian Population**. *Indian J Orthop* (2021.0) **55** 939-47. DOI: 10.1007/s43465-020-00325-x 67. Barroso J, Wakaizumi K, Reckziegel D, Pinto-Ramos J, Schnitzer T, Galhardo V. **Prognostics for pain in osteoarthritis: Do clinical measures predict pain after total joint replacement?**. *PLoS One* (2020.0) **15** e0222370. DOI: 10.1371/journal.pone.0222370 68. Carriere JS, Martel MO, Loggia ML, Campbell CM, Smith MT, Haythornthwaite JA. **The Influence of Expectancies on Pain and Function Over Time After Total Knee Arthroplasty.**. *Pain Med* (2022.0) **23** 1767-76. DOI: 10.1093/pm/pnac067 69. Chodor P, Kruczynski J. **Preoperative Risk Factors of Persistent Pain following Total Knee Arthroplasty**. *BioMed Res Int* (2022.0) **2022** 4958089. DOI: 10.1155/2022/4958089 70. Sorel JC, Veltman ES, Honig A, Poolman RW. **The influence of preoperative psychological distress on pain and function after total knee arthroplasty: a systematic review and meta-analysis.**. *Bone Joint J* (2019.0) 7-14. DOI: 10.1302/0301-620X.101B1.BJJ-2018-0672.R1 71. Birch S, Stilling M, Mechlenburg I, Hansen TB. **No effect of cognitive behavioral patient education for patients with pain catastrophizing before total knee arthroplasty: a randomized controlled trial**. *Acta Orthop* (2020.0) **91** 98-103. DOI: 10.1080/17453674.2019.1694312 72. Bay S, Kuster L, McLean N, Byrnes M, Kuster MS. **A systematic review of psychological interventions in total hip and knee arthroplasty**. *BMC Musculoskelet Disord* (2018.0) **19** 201. DOI: 10.1186/s12891-018-2121-8 73. Dowsey MM, Smith AJ, Choong PFM. **Latent Class Growth Analysis predicts long term pain and function trajectories in total knee arthroplasty: a study of 689 patients**. *Osteoarthritis Cartilage* (2015.0) **23** 2141-9. DOI: 10.1016/j.joca.2015.07.005 74. Lindberg MF, Rustoen T, Miaskowski C, Rosseland LA, Lerdal A. **The relationship between pain with walking and self-rated health 12 months following total knee arthroplasty: a longitudinal study**. *BMC Musculoskelet Disord* (2017.0) **18** 75. DOI: 10.1186/s12891-017-1430-7 75. Sayah SM, Karunaratne S, Beckenkamp PR, Horsley M, Hancock MJ, Hunter DJ. **Clinical Course of Pain and Function Following Total Knee Arthroplasty: A Systematic Review and Meta-Regression.**. *J Arthroplasty* (2021.0) **36** 3993-4002 e37. DOI: 10.1016/j.arth.2021.06.019 76. Bannuru RR, Osani MC, Vaysbrot EE, Arden NK, Bennell K, Bierma-Zeinstra SMA. **OARSI guidelines for the non-surgical management of knee, hip, and polyarticular osteoarthritis**. *Osteoarthritis Cartilage* (2019.0) **27** 1578-89. DOI: 10.1016/j.joca.2019.06.011
--- title: 'Prevalence of undernutrition and associated factors among adults taking antiretroviral therapy in sub-Saharan Africa: A systematic review and meta-analysis' authors: - Awole Seid - Omer Seid - Yinager Workineh - Getenet Dessie - Zebenay Workneh Bitew journal: PLOS ONE year: 2023 pmcid: PMC10038308 doi: 10.1371/journal.pone.0283502 license: CC BY 4.0 --- # Prevalence of undernutrition and associated factors among adults taking antiretroviral therapy in sub-Saharan Africa: A systematic review and meta-analysis ## Abstract ### Background Undernutrition (Body Mass Index < 18.5 kg/m2) is a common problem and a major cause of hospital admission for patients living with HIV. Though sub-Saharan *Africa is* the most commonly affected region with HIV and malnutrition, a meta-analysis study that estimates the prevalence and correlates of undernutrition among adults living with HIV has not yet been conducted. The objective of this study was to determine the pooled prevalence of undernutrition and associated factors among adults living with HIV/AIDS in sub-Saharan Africa. ### Methods Studies published in English were searched systematically from databases such as PubMed, Google Scholar, and gray literature, as well as manually from references in published articles. Observational studies published from 2009 to November 2021 were included. The data extraction checklist was prepared using Microsoft Excel and includes author names, study area, publication year, sample size, prevalence/odds ratio, and confidence intervals. The results were presented and summarized in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) standard. Heterogeneity was investigated using the Q test, I2, τ2, τ and predictive interval. STATA version 17 was used to analyze the data. A meta-analysis using a random-effects model was used to determine the overall prevalence and adjusted odds ratio. The study has been registered in PROSPERO with a protocol number of CRD42021268603. ### Results In this study, a total of 44 studies and 22,316 participants were included. The pooled prevalence of undernutrition among adult people living with HIV (PLWHIV) was $23.72\%$ ($95\%$ CI: 20.69–26.85). The factors associated with undernutrition were participants’ age (AOR = 0.5, $95\%$ CI: 0.29–0.88), gender (AOR = 2.08, $95\%$ CI: 0.22–20.00), World Health Organization (WHO) clinical stage (AOR = 3.25, $95\%$ CI: 2.57–3.93), Cluster of Differentiation 4 (CD4 count) (AOR = 1.94, $95\%$ CI: 1.53–2.28), and duration of ART (AOR = 2.32, $95\%$ CI: 1.6–3.02). ### Conclusion The pooled prevalence of undernutrition among adult PLWHIV in sub-Saharan Africa remained high. WHO clinical stage, CD4 count, duration of ART treatment, age, and sex were found to be the factors associated with undernutrition. Reinforcing nutrition counseling, care, and support for adults living with HIV is recommended. Priority nutritional screening and interventions should be provided for patients with advanced WHO clinical stages, low CD4 counts, the male gender, younger age groups, and ART beginners. ## Introduction According to the United Nations Program on HIV/AIDS (UNAIDS) 2021 fact sheet, there were 38.4 million people living with HIV, of whom 36.7 million were adults (15 years of age or older). Seventy-six percent of adults living with HIV had access to ART treatment [1]. Sub-Saharan Africa accounted for $57\%$ of all new HIV infections in 2019 [2]. Similarly, Eastern and Southern Africa constituted the largest number of AIDS-related deaths [280,000] globally in 2021 [1]. Despite progress in HIV care, the overall life expectancy among adults living with HIV is 5 to 10 years less than that of uninfected adults [3]. Malnutrition takes the lion’s share in increasing the risk of mortality and the occurrence of other opportunistic infections among adults living with HIV [4, 5]. Poor nutrition and HIV have bidirectional relationships and exacerbate one another [6]. In resource-limited countries like sub-Saharan Africa, many people living with HIV (PLWHIV) on long-term ART follow-up lack adequate nutrition [7, 8], and, undernutrition is an indicator of a poor prognosis of HIV care [9]. Similarly, malnutrition is among the major causes of hospital admission in PLWHIV [10]. Meanwhile, HIV affects nutritional status in three distinct ways. It decreases food consumption (through poor appetite and inability to eat and swallow), raises energy needs (up to $20\%$ more energy), and hinders the body’s ability to absorb nutrients. All these factors predispose patients to undernutrition and finally to wasting syndrome [11]. Thus, identifying and treating malnutrition in people living with HIV can fasten recovery from infection, enhance immunity, and possibly slow the progression to AIDS [12]. A meta-analysis study in Ethiopia showed the pooled prevalence of undernutrition among adults receiving ART was $26\%$. Undernutrition among people living with HIV is associated with socio-demographic and clinical factors such as age, WHO clinical stage, CD4 count, duration of ART treatment, and food security [13, 14]. There is a need to estimate the overall prevalence of undernutrition in Sub-Saharan Africa, as the region remains the world’s epicenter of HIV transmission. Similarly, several studies indicated that the region is affected by food insecurity, which is directly linked to undernutrition for individuals, households, and communities affected by HIV [15–18]. Beside this, several nutritional programs to address undernutrition in adults living with HIV in sub-Saharan Africa have been instituted without consolidated evidence on the overall estimates of the prevalence and correlates of undernutrition [19]. Moreover, the available primary studies in sub-Saharan Africa lack consistency and are not conclusive. Therefore, the purpose of this meta-analysis study was to determine the pooled (overall) prevalence of undernutrition and its associated factors among adults living with HIV in sub-Saharan Africa. This study will help policymakers devise evidence-based nutrition intervention programs for patients living with HIV in sub-Saharan Africa. ## Inclusion and exclusion criteria Both published and unpublished observational studies (i.e., cross-sectional, case-control, and cohort) conducted among HIV-positive adults in SSA countries were included. Articles published only in the English language were included. On the other hand, studies with no free full texts, and studied only pregnant women were excluded. For clarity, the topic is described using a PICO format as follows: ## Information source and search strategy We used the databases mainly PubMed, Google Scholar, and Gray (unpublished) literature, university repositories, and manual searches of references from a list of included articles. Articles published from 2009 to November 10, 2021, were included. We used 2009 as there was a previous study among women living with HIV and published in 2008. However, our study also included male adults. Articles identified through the electronic searches were exported and managed using EndNote Version 8 reference manager. Articles from PubMed were accessed using the following keywords (Table 1). **Table 1** | Key variables | Searching words in PubMed | | --- | --- | | Prevalence of undernutrition | “Proportion” or “Prevalence” or “Magnitude” or “Burden” AND “Malnutrition” OR “Undernutrition” OR “Under-weight” OR “Wasting” OR “Malnourished” AND “Adult” AND “Living with HIV” OR “HIV-positive” OR “HIV-infected” OR “Anti-Retroviral Therapy” AND (Each country) sub-Saharan Africa”. | | Associated factors | “Associated factors” OR “Determinants” OR “Predictors” OR “Correlates” AND “Malnutrition” OR “Undernutrition” OR “Under-weight” OR “Wasting” OR “Malnourished” AND “Adult” AND “Living with HIV” OR “HIV-positive” OR “HIV-infected” OR “Anti-Retroviral Therapy AND (Each country) in sub-Saharan Africa”. | ## Data collection process and data items The authors prepared data extraction using Microsoft Excel. All relevant data for this review were extracted by two reviewers (AS and OS). The disparities between reviewers at the time of data abstraction were resolved through discussion with the third author (ZWB). The data extraction sheet included primary authors, publication year, country, study design, sample size, response rate, study setting, study population, proportion, $95\%$ confidence interval, and the logarithm of proportion S1 Table. ## Effect measures We include studies that measure under-weight (undernutrition) using BMI < 18.5 Kg/m2. Underweight was utilized as an indicator of advanced malnutrition, despite the fact that it does not reliably indicate the nutritional status of adults [21]. The proportion of undernutrition was calculated by dividing the number of individuals under-nourished by the total sample of study subjects included in the final analysis. We used the adjusted odds ratio (AOR) as an effect measure to find associated factors of undernutrition. ## Risk of bias assessment Newcastle Ottawa Scale (NOS) adapted for cross-sectional studies was used to assess the quality of the studies. NOS has three categories and has a maximum score of 10 for cross-sectional studies. The categories are selection (maximum of 5 stars), comparability (maximum of 2 stars), and study outcome (maximum of 3 stars). Each study was independently appraised by two authors. Disagreements between authors were resolved through discussion with a third author. Finally, the quality score of each study was calculated as the sum of scores, thus ranging from zero to ten for cross-sectional studies, and zero to nine for cohort and case-control studies. A score of greater or equal to 6 points was considered “good” and included in the study [22]. Additionally, publication bias was assessed using Egger’s regression test, funnel plot, and sensitivity analysis. ## Synthesis methods Data analysis was performed using STATA (version 17) software. We employed a random effect model to find the pooled prevalence and associated factor estimates of under-nutrition. Heterogeneity of effect sizes was assessed using I2, τ2,τ and prediction interval. Subgroup and trim and fill were performed to deal with the potential source of heterogeneity. Sensitivity analysis was also performed. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist is used for data presentation S2 Table [23]. ## Study selection and characteristics A total of 3,477 articles were identified from PubMed, Google scholar, and gray literature in the initial search. After the removal of 1,220 duplicates, 2,257 articles were screened for title and abstract. In the next step, 2,194 articles were excluded based on titles and abstracts. The full texts of 62 articles were downloaded and assessed against inclusion criteria. Thus, 18 articles were excluded for the following reasons: 7 studies did not report data on the outcome variable [5, 24–29], six studies were review papers [4, 14, 30–33], three studies focused on children [30, 34, 35], one study was conducted out of SSA [36], one study focused on pregnant women [37], and one study lacks full text [38]. Finally, 44 studies were included in final systematic review and meta-analysis (Fig 1) [23]. No studies were excluded after appraising the quality using NOS. **Fig 1:** *Flow chart showing the sequence of study selection on undernutrition among adult PLWHIV in sub-Saharan Africa, 2009–2021.* In this study, a total of 22,316 adults living with HIV were included. The sample size of the included studies ranged from 145 in Botswana to 3,993 in Tanzania. Of the 44 included studies, only one was a retrospective cohort and the rest were carried out using cross-sectional study designs. The largest number of articles (33 studies) were reported from Ethiopia [8, 39–70] and only one article is reported from Kenya [71], Botswana [72], Ghana [73], Democratic Republic of Congo (DRC) [74], Senegal [75], Uganda [76], and Zimbabwe [77]. Four articles were from each of South Africa [78, 79] and Tanzania [13, 80]. Regarding the publication year, studies were published from 2009 to 2021 and the majority of the studies [11], were reported in 2020, followed by nine in 2017, and five in each of 2018 and 2015 (Tables 2 and 3). ## Quality appraisal results The heterogeneity test of the study revealed I2 = $96.8\%$, τ2 = 98.83, τ = 9.94, prediction interval (13.8–$33.68\%$), and $95\%$ confidence interval of the average estimate (20.69–$26.79\%$). The source of high I2 is not identified, though it is expected to rise in a meta-analysis of proportions in different countries, and the result should be interpreted conservatively [81]. Furthermore, the study also demonstrated a wide prediction interval, a direct and easily interpretable indicator as compared to the confidence interval, implying evidence of high heterogeneity. Regarding the publication bias, Egger’s regression test (B1 = 5.72, $$p \leq 0.002$$) showed there was publication bias but studies looked relatively symmetrical in the funnel plot (Fig 2). However, these indicators of publication bias were developed in the context of comparative data and may not be reliable indicators of publication bias in a meta-analysis of proportions [82]. **Fig 2:** *Funnel plot of studies on undernutrition among adult PLWHIV in SSA, 2009–2021.* Sub-group analysis and trim and fill analysis were also performed to deal with the publication bias and heterogeneity S3 Table. A sensitivity analysis was performed and all estimates were within the confidence interval limit, and no individual study contributed to the publication bias. Consequently, it is unnecessary to exclude studies from the final meta-analysis. ## Prevalence of undernutrition among adult PLWHIV in sub-Saharan Africa Of the 44 studies that reported a proportion of undernutrition, the highest prevalence ($60\%$) was reported in Ethiopia [58], whereas the lowest ($8.3\%$) was reported in Kenya [71]. Majority of the studies are reported from Ethiopia (33 studies), cross-sectional design (43 studies), carried out during 2016–2021 (33 studies), and on patients taking ART (38 studies). In this study, the pooled prevalence of undernutrition using a random-effect model meta-analysis was found to be $23.74\%$ ($95\%$ CI: 20.77–26.73) (Fig 3). Sub-group analysis by country showed $25.8\%$ ($95\%$ CI: 22.4–29.3) in Ethiopia, $14.5\%$ ($95\%$ CI: 14.9–17.2) in South Africa, and $24.3\%$ ($95\%$ CI: 16.3–32.3) in Tanzania. Additional sub-group analysis by ART status, by study design, and by publication year also carried out (Table 3). **Fig 3:** *Forest plot of pooled prevalence of undernutrition among adult PLWHIV in sub-Saharan Africa 2021.* ## Factors associated with undernutrition among adults of PLWHIV in sub-Saharan Africa From the searched published articles 16 reported “WHO clinical stage” [13, 39, 41–43, 45, 47, 50, 52, 57, 59, 60, 64–66, 68, 83], four studies reported “CD4 count” [40, 43, 47, 55, 57, 59, 60, 80, 83], four studies reported patient’s “age” [13, 39, 55, 56, 60, 77], four studies reported “sex” [40, 56, 64, 80, 83], and four articles reported “duration of ART treatment” [8, 59, 64, 67, 68] as factors associated with undernutrition among adult PLWHIV in sub-Saharan Africa (Table 4). Patients living with HIV and WHO clinical stage III/IV were 3.25 (AOR, $95\%$ CI: 2.57–3.93) times higher odds of developing undernutrition as compared to WHO clinical stage I/II (Fig 4). Similarly, patients whose CD4 count was less than 200cells/mm3 were 1.94 times (AOR = 1.94, $95\%$ CI: 1.53–2.28) with higher odds of developing undernutrition as compared to their counterparts (CD4 ˃500cells/mm3) (Fig 5). **Fig 4:** *The pooled effect of “WHO clinical stage” on undernutrition among adult PLWHIV in sub-Saharan Africa, 2021.* **Fig 5:** *Pooled effect of “CD4 count” on undernutrition among adult PLWHIV in sub-Saharan Africa, 2021.* TABLE_PLACEHOLDER:Table 4 Regarding the age of study participants, patients aged 40 years and above had $49\%$ lower odds of developing undernutrition as compared to those aged 19 to 30 years (AOR = 0.51, $95\%$ CI: 0.26–0.76) (Fig 6). Furthermore, the odds of developing undernutrition among males living with HIV were 2 times (AOR = 2.11, $95\%$ CI: 1.52–2.7) higher as compared to female patients (Fig 7). Similar to this, patients receiving ART for less than 12 months had 2.68 times the risk of developing undernutrition compared to individuals taking it for more than 12 months (Fig 8). This implies that the longer the duration of patients’ taking ART, the lower the risk of developing under-nutrition. **Fig 6:** *Pooled effect of “age” of study participants on undernutrition among adult PLWHIV in sub-Saharan Africa, 2021.* **Fig 7:** *Pooled effect of “sex” on undernutrition among adult PLWHIV in sub-Saharan Africa, 202.* **Fig 8:** *Pooled effect of “ART duration” on undernutrition among adult PLWHIV in sub-Saharan Africa, 2021.* ## Discussion Despite the improvement of comprehensive HIV care, sub-Saharan Africa continues to be an epicenter of HIV transmission and has a high prevalence of malnutrition among adults living with HIV. This study aimed to investigate the pooled prevalence and correlates of undernutrition among adults living with HIV/AIDS. The result shows a significant number of adults living with HIV are malnourished, and several socio-demographic and clinical factors have been associated with undernutrition. In this meta-analysis, the pooled prevalence of undernutrition (adult BMI <18.5kg/m2) among adults living with HIV was high ($23.74\%$) and interpreted as a serious situation according to WHO nutrition landscape information system cut-off values (20–$39\%$) [84]. It is serious because studies have shown that undernutrition increases the risk of opportunistic infections (OIs) and mortality [4, 85]. The pooled prevalence slightly decreased from $25.8\%$ in 2009–2015 to $23\%$ in 2015–2021. This might be attributed to the expansion of nutrition intervention programs and the improvement of comprehensive HIV care in Africa. This finding is in line with a previous meta-analysis study in Ethiopia, in which the pooled prevalence was reported as $26\%$ ($95\%$ CI: 22–$30\%$) [14]. In contrast, our finding is higher as compared to a previous meta-analysis study conducted on women living with HIV in sub-Saharan Africa: $10.3\%$ ($95\%$ CI: $7.4\%$–$14.1\%$). The study used secondary data from the DHS and analyzed the reported estimate of only 11 sub-Saharan African countries. The disparity might be explained by the fact that over the last 13 years, there have been changes in socio-demography, the trend of HIV incidence, food insecurity, and other population factors that might have been associated with undernutrition. Moreover, the study was carried out only among women, which may have affected the result. Nevertheless, our study also asserts that men living with HIV are at a higher risk of developing malnutrition as compared to women [37], even though the biological mechanism is not clear. We also found that patients with an advanced WHO clinical stage, a lower CD4 count, being of male sex, a younger age, and a shorter duration of ART treatment had a higher likelihood of developing undernutrition in adults living with HIV. The factor regarding the WHO clinical stage was also reported in a meta-analysis report in Ethiopia [86]. This may be due to the advanced WHO clinical stage and low CD4 count, which are indicators of severe immune deficiency are directly linked to undernutrition, especially protein and energy malnutrition. Thus, it is important to give due emphasis to nutrition counseling and supplementation with high-energy and protein foods during the follow-up visit. Furthermore, young adults and ART beginners were identified as being at risk for malnutrition. The reason for the younger ages might be due to the poor emotional readiness to accept the disease condition and the failure to receive comprehensive HIV care at an early age. On the other hand, increasing age may improve acceptance and the perceived benefits of adherence to recommendations by health care providers. However, the result for younger ages requires further exploration. In spite of this, although it is acknowledged that receiving ART improves nutritional status, the impact of HIV on nutrition begins even before diagnosis and needs a longer course of therapy in order to be reversed, and noticed by anthropometric measurements. ## Limitation of the study The possible limitation for this review was the inability of accessing some databases like EMBASE, CINHAL, and Scopus. This was compensated by searching for published articles in broad databases like Google scholar. The other limitation is the absence of similar meta-analysis studies for comparison of our result. There is an uneven distribution of included studies among countries, a large number of which were reported from Ethiopia. Additionally, there is high heterogeneity, and studies published only in English were included. ## Conclusion The pooled prevalence of undernutrition among adult PLWHIV in sub-Saharan Africa remained high. WHO clinical stage, CD4 count, duration of ART treatment, age, and sex were found to be the factors associated with undernutrition. Reinforcing nutrition counseling, care, and support for adults living with HIV is recommended. Priority nutritional screening and interventions should be provided for patients with advanced WHO clinical stages, low CD4 counts, the male gender, younger age groups, and ART beginners. ## References 1. 1UNAIDS. Fact Sheet. 2022.. *Fact Sheet* (2022.0) 2. 2UNAIDS. 2020 Global AIDS Update: Seizing the moment. Geneva: Joint United Nations Programme on HIV/AIDS (UNAIDS. 2020.. (2020.0) 3. Cawley C, Wringe A, Todd J, Gourlay A, Clark B, Masesa C. **Risk factors for service use and trends in coverage of different HIV testing and counselling models in northwest T anzania between 2003 and 2010.**. *Tropical Medicine & International Health.* (2015.0) **20** 1473-87. PMID: 26223900 4. Alebel A, Demant D, Petrucka P, Sibbritt D. **Effects of undernutrition on mortality and morbidity among adults living with HIV in sub-Saharan Africa: a systematic review and meta-analysis.**. *BMC infectious diseases.* (2021.0) **21** 1-20. DOI: 10.1186/s12879-020-05706-z 5. Masiira B, Baisley K, Mayanja BN, Kazooba P, Maher D, Kaleebu P. **Mortality and its predictors among antiretroviral therapy naïve HIV-infected individuals with CD4 cell count≥ 350 cells/mm3 compared to the general population: data from a population-based prospective HIV cohort in Uganda.**. *Global health action.* (2014.0) **7** 21843. PMID: 24433941 6. Piwoz E, Preble E. **Policy Brief: HIV, Food Security and Nutrition.**. *Retrieved* (2000.0) **21** 2009 7. Osei Akumiah P, Akuamoah Sarfo L, Agyekum O, Asare K, Azumah Nayembil D, Ofosu Agyapong G. *Nutritional Knowledge, Attitude and Intake among HIV Patients: A Study at Agogo Presbyterian Hospital, Ghana.* (2015.0) 8. Dedha M, Damena M, Egata G, Negesa L. **Undernutrition and associated factors among adults human immunodeficiency virus positive on antiretroviral therapy in hospitals, East Hararge Zone, Oromia, Ethiopia: A cross-sectional study.**. *International journal of health sciences* (2017.0) **11** 35. PMID: 29114193 9. Fadeiye E, Adegbenro C, Olumakaiye M. **Assessment of nutritional knowledge of HIV positive adult patients of obafemi awolowo university teaching hospitals complex ile-ife, nigeria**. *J Nutr Health Sci* (2016.0) **3** 102 10. Ford N, Shubber Z, Meintjes G, Grinsztejn B, Eholie S, Mills EJ. **Causes of hospital admission among people living with HIV worldwide: a systematic review and meta-analysis**. *The lancet HIV* (2015.0) **2** e438-e44. DOI: 10.1016/S2352-3018(15)00137-X 11. **Nutrition and HIV/AIDS: A training manual for nurses and midwives.**. *ECSA-HC (East, Central, and Southern African Health Community) Arusha;* (2008.0) 12. 12Health FMo. Ethiopian Guide to Clinical Nutrition Care for Children and Adults with HIV. 2008.. *Ethiopian Guide to Clinical Nutrition Care for Children and Adults with HIV.* (2008.0) 13. Sunguya BF, Ulenga NK, Siril H, Puryear S, Aris E, Mtisi E. **High magnitude of under nutrition among HIV infected adults who have not started ART in Tanzania—a call to include nutrition care and treatment in the test and treat model.**. *BMC nutrition.* (2017.0) **3** 1-9 14. Alebel A, Kibret GD, Petrucka P, Tesema C, Moges NA, Wagnew F. **Undernutrition among Ethiopian adults living with HIV: a meta-analysis.**. *BMC nutrition.* (2020.0) **6** 1-10. PMID: 32153975 15. Aibibula W, Cox J, Hamelin A-M, Mamiya H, Klein MB, Brassard P. **Food insecurity and low CD4 count among HIV-infected people: a systematic review and meta-analysis.**. *AIDS care* (2016.0) **28** 1577-85. DOI: 10.1080/09540121.2016.1191613 16. Drammeh W, Hamid NA, Rohana A. **Determinants of household food insecurity and its association with child malnutrition in Sub-Saharan Africa: A review of the literature**. *Current Research in Nutrition and Food Science Journal* (2019.0) **7** 610-23 17. Giller KE. **The food security conundrum of sub-Saharan Africa.**. *Global Food Security* (2020.0) **26** 100431 18. Saha S, Behnke A, Oldewage-Theron W, Mubtasim N, Miller M. **Prevalence and Factors Associated with Food Insecurity among Older Adults in Sub-Saharan Africa: A Systematic Review**. *Journal of Nutrition in Gerontology and Geriatrics* (2021.0) **40** 171-96. DOI: 10.1080/21551197.2021.1988027 19. Tesfay FH, Javanparast S, Gesesew H, Mwanri L, Ziersch A. **Characteristics and impacts of nutritional programmes to address undernutrition of adults living with HIV in sub-Saharan Africa: a systematic review of evidence**. *BMJ Open* (2022.0) **12** 20. Dwyer-Lindgren L, Cork MA, Sligar A, Steuben KM, Wilson KF, Provost NR. **Mapping HIV prevalence in sub-Saharan Africa between 2000 and 2017**. *Nature* (2019.0) **570** 189-93. DOI: 10.1038/s41586-019-1200-9 21. Koethe JR, Heimburger DC. **Nutritional aspects of HIV-associated wasting in sub-Saharan Africa**. *The American journal of clinical nutrition* (2010.0) **91** 1138S-42S. DOI: 10.3945/ajcn.2010.28608D 22. Peterson J, Welch V, Losos M, Tugwell P. **The Newcastle-Ottawa scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses.**. *Ottawa: Ottawa Hospital Research Institute* (2011.0) 1-12 23. Page MJ MJ, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. *BMJ* (2021.0) **372**. DOI: 10.1136/bmj.n71 24. Tadele M, Tesfa M, Tsegaye G, Temesgen H, Alamirew NM. **Determinants of chronic energy deficiency among adults living with HIV in Shebel Berenta District, East Gojjam, Amhara region, North West Ethiopia, 2017: case control study.**. *BMC research notes.* (2019.0) **12** 1-6. PMID: 30602384 25. Oluma A, Abadiga M, Mosisa G, Etafa W, Fekadu G. **Food Insecurity among people living with HIV/AIDS on ART follower at public hospitals of Western Ethiopia**. *International Journal of Food Science* (2020.0) **2020**. DOI: 10.1155/2020/8825453 26. Kassie AM, Abate BB, Kassaw MW. **Prevalence of overweight/obesity among the adult population in Ethiopia: a systematic review and meta-analysis.**. *BMJ open* (2020.0) **10** e039200. DOI: 10.1136/bmjopen-2020-039200 27. Kalil FS, Kabeta T, Jarso H, Hasen M, Ahmed J, Kabeta S. **Determinants of Undernutrition Among Adult People on Antiretroviral Therapy in Goba Hospital, Southeast Ethiopia: A Case–Control Study.**. *Nutrition and Dietary Supplements.* (2020.0) **12** 223-36 28. Getahun MB, Teshome GS, Fenta FA, Bizuneh AD, Mulu GB, Kebede MA. **Determinants of Severe Acute Malnutrition Among HIV-positive Children Receiving HAART in Public Health Institutions of North Wollo Zone, Northeastern Ethiopia: Unmatched Case–Control Study.**. *Pediatric Health, Medicine and Therapeutics.* (2020.0) **11** 313. DOI: 10.2147/PHMT.S267892 29. Gedle D, Mekuria G, Kumera G, Eshete T, Feyera F, Ewunetu T. **Food insecurity and its associated factors among people living with HIV/AIDS receiving anti-retroviral therapy at Butajira Hospital, Southern Ethiopia.**. *Journal of Nutrition & Food Sciences.* (2015.0) **5** 2-6 30. Nigussie J, Girma B, Molla A, Mareg M. **Under-nutrition and associated factors among human immunodeficiency virus-infected children in sub-Saharan Africa: A systematic review and meta-analysis.**. (2020.0) 31. Mengie T, Dejen D, Muche T, Getacher L, Kindie B, Tamiru C. *Under Nutrition and Its Determinants Among Adults Receiving Antiretroviral Therapy in Ethiopia: A Systematic Review and Meta-analysis.* 32. Hidru HD, Gebremedhine H, Gebretsadik A, Teame H, Negash H, Mengesha MB. **Prevalence of Food Insecurity and Its Associated Factors among Adult People with Human Immunodeficiency Virus in Ethiopia: A Systematic Review and Meta-Analysis.**. *International Journal of Food Science* (2021.0) **2021**. DOI: 10.1155/2021/7816872 33. Abate BB, Aragie TG, Tesfaw G. **Magnitude of underweight, wasting and stunting among HIV positive children in East Africa: A systematic review and meta-analysis.**. *PloS one.* (2020.0) **15** e0238403. DOI: 10.1371/journal.pone.0238403 34. Tiruneh CM, Walle BG, Emiru TD, Tibebu NS, Abate MW, Nigat AB. **Under-nutrition and associated factors among children on ART in Southern Ethiopia: a facility-based cross-sectional study**. *Italian Journal of Pediatrics* (2021.0) **47** 1-10. PMID: 33407754 35. Gondar E.. **PREVALENCE OF MALNUTRITION AND ASSOCIATED FACTORS AMONG HIVINFECTED CHILDREN AGED 6–59 MONTHS AT GONDAR UNIVERSITY HOSPITAL.**. *NORTHWEST ETHIOPIA: DEPARTMENT OF PEDIATRICS AND CHILDHEALTH, COLLEGE OF MEDICINE AND HEATLH.* (2014.0) 36. Thapa R, Amatya A, Pahari DP, Bam K, Newman MS. **Nutritional status and its association with quality of life among people living with HIV attending public anti-retroviral therapy sites of Kathmandu Valley, Nepal.**. *AIDS research and therapy* (2015.0) **12** 1-10. PMID: 25598834 37. Uthman OA. **Prevalence and pattern of HIV-related malnutrition among women in sub-Saharan Africa: a meta-analysis of demographic health surveys.**. *BMC public health.* (2008.0) **8** 1-8. DOI: 10.1186/1471-2458-8-226 38. Debebe L.. *Under nutrition and Associated Factors among Adults Living with Hiv/Aids Attending in Selected Public Hospitals in Addis Ababa, Ethiopia: A Cross-Sectional Study* (2019.0) 39. Adal M, Howe R, Kassa D, Aseffa A, Petros B. **Malnutrition and lipid abnormalities in antiretroviral naive HIV-infected adults in Addis Ababa: a cross-sectional study.**. *PloS one.* (2018.0) **13** e0195942. PMID: 29672576 40. Amza L, Demissie T, Halala Y. **Under nutrition and associated factors among adult on highly active antiretroviral therapy in Wolaita Sodo teaching and referral hospital, southern nations nationalities peoples region, Ethiopia.**. *International Journal of Nutrition and Metabolism* (2017.0) **9** 10-9 41. Asnakew M. **Malnutrition and associated factors among adult individuals receiving highly active antiretroviral therapy in health facilities of Hosanna Town, Southern Ethiopia.**. *Open Access Library Journal* (2015.0) **2** 1 42. Birhane M, Loha E, Alemayehu FR. **Nutritional status and associated factors among adult HIV/AIDS patients receiving ART in Dilla University Referral Hospital, Dilla, Southern Ethiopia.**. *Research Gate.* (2017.0) 43. Daka DW, Ergiba MS. **Prevalence of malnutrition and associated factors among adult patients on antiretroviral therapy follow-up care in Jimma Medical Center, Southwest Ethiopia.**. *PloS one* (2020.0) **15** e0229883. DOI: 10.1371/journal.pone.0229883 44. Daniel M, Mazengia F, Birhanu D. **Nutritional status and associated factors among adult HIV/AIDS clients in Felege Hiwot Referral Hospital, Bahir Dar, Ethiopia.**. *Science Journal of Public Health* (2013.0) **1** 24-31 45. Fentie M, Wassie MM, Tesfahun A, Alemu K, Mequanent M, Ayele TA. **Chronic energy deficiency and associated factors among adults living with HIV in Gondar University Referral Hospital northwest Ethiopia.**. *BMC Nutrition.* (2017.0) **3** 1-9 46. Fufa H, Umeta M, Taffesse S, Mokhtar N, Aguenaou H. **Nutritional and immunological status and their associations among HIV-infected adults in Addis Ababa, Ethiopia.**. *Food and nutrition bulletin.* (2009.0) **30** 227-32. DOI: 10.1177/156482650903000303 47. Gebru TH, Mekonen HH, Kiros KG. **Undernutrition and associated factors among adult HIV/AIDS patients receiving antiretroviral therapy in eastern zone of Tigray, Northern Ethiopia: a cross-sectional study.**. *Archives of Public Health.* (2020.0) **78** 1-8. DOI: 10.1186/s13690-020-00486-z 48. Gedle D, Gelaw B, Muluye D, Mesele M. **Prevalence of malnutrition and its associated factors among adult people living with HIV/AIDS receiving anti-retroviral therapy at Butajira Hospital, southern Ethiopia.**. *BMC nutrition.* (2015.0) **1** 1-11 49. Girma M, Motuma A, Negasa L. **Assessment of Nutritional Status and Associated Factors among Adult People Living With HIV/AIDS in Hiowt Fana Specialized University Hospital, Eastern Ethiopia.**. *Edited and Compiled by.* (2017.0) **199** 50. Gebremichael DY, Hadush KT, Kebede EM, Zegeye RT. **Food insecurity, nutritional status, and factors associated with malnutrition among people living with HIV/AIDS attending antiretroviral therapy at public health facilities in West Shewa Zone, Central Ethiopia.**. *BioMed research international.* (2018.0) **2018**. DOI: 10.1155/2018/1913534 51. Hadgu TH, Worku W, Tetemke D, Berhe H. **Undernutrition among HIV positive women in Humera hospital, Tigray, Ethiopia, 2013: antiretroviral therapy alone is not enough, cross sectional study.**. *BMC public health.* (2013.0) **13** 1-10. PMID: 23280303 52. Hailemariam S, Bune GT, Ayele HT. **Malnutrition: Prevalence and its associated factors in People living with HIV/AIDS, in Dilla University Referral Hospital.**. *Archives of Public Health.* (2013.0) **71** 13. DOI: 10.1186/0778-7367-71-13 53. Kenea MA, Garoma S, Gemede HF. **Assessment of adult nutritional status and associated factors among ART users in Nekemte referral hospital and health center, east Wollega zone, Ethiopia.**. *Journal of Food and Nutrition Sciences* (2015.0) **3** 56 54. Kumie G.. *PREVALENCE OF MALNUTRITION AND ASSOCIATED FACTR AMONG ADULT PEOPLE LIVING WITH HIV/AIDS RECEIVING ANTI-RETROVIRAL THERAPY AT ARADA SUB CITY, ADDIS ABABA, ETHIOPIA* (2020.0) 55. Mitiku A, Ayele TA, Assefa M, Tariku A. **Undernutrition and associated factors among adults living with Human Immune Deficiency Virus in Dembia District, northwest Ethiopia: an institution based cross-sectional study.**. *Archives of Public Health* (2016.0) **74** 1-8. DOI: 10.1186/s13690-016-0143-y 56. Motuma A, Abdeta T. **Undernutrition and Associated Factors Among Seropositive Adults in ART Clinic Treatment Centre, Hiwot Fana Specialized University Hospital, Eastern Ethiopia.**. *Pathology and Laboratory Medicine.* (2021.0) **5** 10 57. Mulu H, Hamza L, Alemseged F. **Prevalence of Malnutrition and Associated Factors among Hospitalized Patients with Acquired Immunodeficiency Syndrome in Jimma University Specialized Hospital, Ethiopia.**. *Ethiop J Health Sci* (2016.0) **26** 217-26. DOI: 10.4314/ejhs.v26i3.4 58. Nigusso FT, Mavhandu-Mudzusi AH. **High magnitude of food insecurity and malnutrition among people living with HIV/AIDS in Ethiopia: A call for integration of food and nutrition security with HIV treatment and care Programme**. *Nutrition and Health* (2021.0) **27** 141-50. DOI: 10.1177/0260106020971855 59. Oumer B, Boti N, Hussen S, Gultie T. **Prevalence Of Under nutrition And Associated Factors Among Adults Receiving First-Line Antiretroviral Treatment In Public Health Facilities Of Arba Minch Town, Southern Ethiopia.**. *HIV/AIDS (Auckland, NZ).* (2019.0) **11** 313 60. Regassa TM, Gudeta TA. *Under nutrition and Associated Factors Among Adult People Attending Highly Active Anti-Retroviral Therapy in Health Facilities of Bench Maji Zone, South West Ethiopia, 2018.* (2020.0) 61. Sahile AT, Ayehu SM, Fanta SF. **Underweight and Its Predictors Among Patients on Anti Retroviral Therapy at Selected Health Facilities of Addis Ababa, Ethiopia, 2020.**. *HIV/AIDS (Auckland, NZ).* (2021.0) **13** 99 62. Saliya MS, Azale T, Alamirew A, Tesfaye DJ. **Assessment of nutritional status and its associated factors among people affected by human immune deficiency virus on antiretroviral therapy: A cross sectional study in Siltie zone, South Ethiopia.**. *Healthcare in Low-resource Settings.* (2021.0) **9** 63. Shifera N, Molla A, Mesafint G, Gichew S, Matiyas R, Yosef T. **Malnutrition status, and its Associated Factors among Adult HIV Positive Attending ART at Public Hospitals of Bench Sheko Zone, South West, Ethiopia, 2020.**. (2021.0) 64. Takele A, Engida ARA. **Under nutrition and Associated Factors among Adult Individuals Receiving Highly Active Antiretroviral Therapy in Health Facilities of Nekemte Town.**. *Eastern Ethiopia.* 65. Teklu T, Chauhan NM, Lemessa F, Teshome G. **Assessment of Prevalence of Malnutrition and Its Associated Factors among AIDS Patients from Asella, Oromia, Ethiopia.**. *BioMed Research International.* (2020.0) **2020** 7360190. DOI: 10.1155/2020/7360190 66. Teshome MS, Gissa SB, Tefera BZ, Lema TB. **Undernutrition and its predictors among people living with HIV/AIDS attending antiretroviral therapy clinic in Jimma University Specialized Hospital**. *International Journal of Nutrition and Metabolism* (2017.0) **9** 67-74 67. Wasie B, Kebede Y, Yibre A. **Nutritional status of adults living with HIV/AIDS at the University of Gondar Referral Hospital, Northwest Ethiopia.**. *Ethiopian Journal of Health and Biomedical Sciences.* (2010.0) **3** 3-14 68. Wasihun Y, Yayehrad M, Dagne S, Menber Y, Awoke T, Nega G. *Prevalence of Undernutrition and Associated Factors among Adult People Living with HIV/AIDS and on ART in Achefer District, North West Ethiopia: Cross-Sectional study.* (2020.0) 69. Yitbarek GY, Engidaw MT, Ayele BA, Tiruneh SA, Alamir MT. **Magnitude of obesity/overweight and its associated factors among HIV/AIDS patients on antiretroviral therapy in Jimma Zone Hospitals, South West Ethiopia: hospital-based cross-sectional study**. *Diabetes, metabolic syndrome and obesity: targets and therapy.* (2020.0) **13** 1251. DOI: 10.2147/DMSO.S247221 70. Zemede Z, Tariku B, Kote M, Estifanos W. **Undernutrition and associated factors among HIV-positive adult patients enrolled in antiretroviral therapy (ART) clinics in the Arba Minch area, southern Ethiopia.**. *HIV/AIDS (Auckland, NZ).* (2019.0) **11** 147. DOI: 10.2147/HIV.S200120 71. Saito A, Karama M, Kamiya Y. **HIV infection, and overweight and hypertension: a cross-sectional study of HIV-infected adults in Western Kenya**. *Tropical medicine and health* (2020.0) **48** 1-12. DOI: 10.1186/s41182-020-00215-w 72. Nnyepi M.. **The risk of developing malnutrition in people living with HIV/AIDS: Observations from six support groups in Botswana.**. *South African Journal of Clinical Nutrition* (2009.0) **22** 89-93. DOI: 10.1080/16070658.2009.11734224 73. Nanewortor BM, Saah FI, Appiah PK, Amu H, Kissah-Korsah K. **Nutritional status and associated factors among people living with HIV/AIDS in Ghana: cross-sectional study of highly active antiretroviral therapy clients.**. *BMC nutrition* (2021.0) **7** 1-8. PMID: 33413686 74. Akilimali P, Musumari P, Kashala-Abotnes E, Tugirimana P, Mutombo P. **Food insecurity and undernutrition in treated HIV patients a (post-) conflict setting: A cross sectional study from Goma, Eastern Democratic Republic of Congo.**. *J Nutrition Health Food Sci.* (2016.0) **4** 1-9 75. Benzekri NA, Sambou J, Diaw B, Sall EHI, Sall F, Niang A. **High prevalence of severe food insecurity and malnutrition among HIV-infected adults in Senegal, West Africa.**. *PloS one* (2015.0) **10** e0141819. DOI: 10.1371/journal.pone.0141819 76. Odwee A, Kasozi KI, Acup CA, Kyamanywa P, Ssebuufu R, Obura R. **Malnutrition amongst HIV adult patients in selected hospitals of Bushenyi district in southwestern Uganda.**. *African Health Sciences* (2020.0) **20** 122-31. DOI: 10.4314/ahs.v20i1.17 77. Takarinda KC, Mutasa-Apollo T, Madzima B, Nkomo B, Chigumira A, Banda M. **Malnutrition status and associated factors among HIV-positive patients enrolled in ART clinics in Zimbabwe.**. *BMC Nutrition* (2017.0) **3** 15. DOI: 10.1186/s40795-017-0132-8 78. Mahlangu K, M P, Madiba S. **The nutritional status of adult antiretroviral therapy recipients with a recent HIV diagnosis; a cross-sectional study in primary health facilities in Gauteng, South Africa.**. *InHealthcare.* (2020.0) **8** 290. DOI: 10.3390/healthcare8030290 79. Naidoo K, Yende-Zuma N, Augustine S. **A retrospective cohort study of body mass index and survival in HIV infected patients with and without TB co-infection.**. *Infectious diseases of poverty.* (2018.0) **7** 1-6. PMID: 29335021 80. Kabalimu TK, Sungwa E, Lwabukuna WC. **Malnutrition and associated factors among adults starting on antiretroviral therapy at PASADA Hospital in Temeke District, Tanzania.**. *Tanzania Journal of Health Research* (2018.0) **20** 81. Barker TH, Migliavaca CB, Stein C, Colpani V, Falavigna M, Aromataris E. **Conducting proportional meta-analysis in different types of systematic reviews: a guide for synthesisers of evidence.**. *BMC Medical Research Methodology* (2021.0) **21** 1-9. PMID: 33397292 82. Cumpston M, Li T, Page MJ, Chandler J, Welch VA, Higgins JP. **Updated guidance for trusted systematic reviews: a new edition of the Cochrane Handbook for Systematic Reviews of Interventions.**. *Cochrane Database Syst Rev.* (2019.0) **10** 14651858. DOI: 10.1002/14651858.ED000142 83. Tekelehaimanot AN, Lemma TB, Gudina EK, Getnet M, Amdisa D, Dadi LS. **Predictors of Under Nutrition and Its Implication toward HIV Continuum Care among Adult People Living with HIV in Jimma Zone Public Hospitals, Southwest Ethiopia: A Mixed Method Study**. *Journal of the International Association of Providers of AIDS Care (JIAPAC).* (2020.0) **19** 2325958220976254. DOI: 10.1177/2325958220976254 84. 84Organization WH. Nutrition Landscape Information System (NLIS) country profile indicators: interpretation guide. 2019.. *Nutrition Landscape Information System (NLIS) country profile indicators: interpretation guide.* (2019.0) 85. Alebel A, Demant D, Petrucka P, Sibbritt D. **Effects of undernutrition on opportunistic infections among adults living with HIV on ART in Northwest Ethiopia: Using inverse-probability weighting.**. *PLOS ONE.* (2022.0) **17** e0264843. DOI: 10.1371/journal.pone.0264843 86. Tesfa Mengie, D D, Temesgen Muche, Lemma Getacher. **Under Nutrition and Its Determinants Among Adults Receiving Antiretroviral Therapy in Ethiopia: A Systematic Review and Meta-analysis.**. *International Journal of Homeopathy & Natural Medicines.* (2021.0) **7** 1-6. DOI: 10.11648/j.ijhnm.20210701.11
--- title: A stretchable wireless wearable bioelectronic system for multiplexed monitoring and combination treatment of infected chronic wounds authors: - Ehsan Shirzaei Sani - Changhao Xu - Canran Wang - Yu Song - Jihong Min - Jiaobing Tu - Samuel A. Solomon - Jiahong Li - Jaminelli L. Banks - David G. Armstrong - Wei Gao journal: Science Advances year: 2023 pmcid: PMC10038347 doi: 10.1126/sciadv.adf7388 license: CC BY 4.0 --- # A stretchable wireless wearable bioelectronic system for multiplexed monitoring and combination treatment of infected chronic wounds ## Abstract Chronic nonhealing wounds are one of the major and rapidly growing clinical complications all over the world. Current therapies frequently require emergent surgical interventions, while abuse and misapplication of therapeutic drugs often lead to an increased morbidity and mortality rate. Here, we introduce a wearable bioelectronic system that wirelessly and continuously monitors the physiological conditions of the wound bed via a custom-developed multiplexed multimodal electrochemical biosensor array and performs noninvasive combination therapy through controlled anti-inflammatory antimicrobial treatment and electrically stimulated tissue regeneration. The wearable patch is fully biocompatible, mechanically flexible, stretchable, and can conformally adhere to the skin wound throughout the entire healing process. Real-time metabolic and inflammatory monitoring in a series of preclinical in vivo experiments showed high accuracy and electrochemical stability of the wearable patch for multiplexed spatial and temporal wound biomarker analysis. The combination therapy enabled substantially accelerated cutaneous chronic wound healing in a rodent model. A soft smart bandage is capable of multiplexed metabolic monitoring, antimicrobial treatment, and electrical stimulation. ## INTRODUCTION Chronic wounds are characterized by impaired or stagnant healing, prolonged and uncontrolled inflammation, as well as compromised extracellular matrix (ECM) function (1–3). Over 6.7 million people in the United States alone suffer from chronic nonhealing wounds including diabetic ulcers, nonhealing surgical wounds, burns, and venous-related ulcerations [4, 5], causing a staggering financial burden of over $25 billion per year on the health care system [6]. Chronic wound healing is a highly complex biological process consisting of four integrated and overlapping phases: hemostasis, inflammation, proliferation, and remodeling (1–3). Current therapies including skin grafts, skin substitutes, negative pressure wound therapy, and others can be beneficial but frequently require procedures or surgical intervention [7]. Microbial infection at the wound site can severely prolong the healing process and lead to necrosis, sepsis, and even death [3]. Both topical and systemic antibiotics are increasingly prescribed to patients suffering from chronic nonhealing wounds, but the overuse, abuse, and misapplication of antibiotics often lead to an escalating drug resistance in bacteria, causing a drastic increase in morbidity and mortality rates [8]. As an alternative therapeutic approach, electrical stimulation has shown to have a substantial effect on the wound healing process, including stimulating fibroblast proliferation and differentiation into myofibroblasts and collagen formation, keratinocyte migration, angiogenesis, and attracting macrophages [9, 10]. However, currently reported electrical stimulation devices usually require bulky equipment and wire connections, making them highly challenging for practical clinical use. More effective, fully controllable, and easy-to-implement therapies are critically needed for personalized treatment of chronic wounds with minimal side effects. At each stage of healing process, the chemical composition of the wound exudate changes substantially, indicating the stage of healing and even the presence of an infection (11–13). For example, increased temperature is associated with bacterial infection, and changes in temperature can provide information on various factors relevant to healing, inflammation, and oxygenation in the wound bed; acidity (pH) indicates a healing state with balanced protease activities and effective ECM remodeling, moreover, elevated pH in wound environment can be a sign of infection; elevated uric acid (UA) indicates wound severity with excessive reactive oxygen species and inflammation and shows immune system responding to inflammatory cytokines [14]; lactate and ammonium are crucial markers for soft-tissue infection diagnosis and angiogenesis in diabetic foot ulcers [15]; wound exudate glucose has a strong correlation with blood glucose and bacterial activities [16], providing crucial therapeutical guidance for clinical diabetic wound treatment. Recent advances in digital health and flexible electronics have transformed conventional medicine into remote at-home health care (17–23). Wearable biosensors could allow real-time and continuous monitoring of physical vital signs and physiological biomarkers in various biofluids such as sweat, saliva, and interstitial fluids (18–21, 24–30). *In* general, an ideal wound dressing should provide a moist wound environment, offer protection from secondary infections, remove wound exudate, and promote tissue regeneration. Despite the promising prospects opened by the wearable technologies (31–37), major challenges exist to realize their full potential toward practical chronic wound management applications: the chronic nonhealing wounds pose high requirement on the flexibility, breathability, and biocompatibility of the wearable devices to protect the wound bed from bacterial infiltrations and infection and modulate wound exudate level; the complex wound exudate matrix could substantially affect the biosensor performance, and thus, there are few reports on prolonged evaluation of biosensors in vivo [13, 31]; personalized wound management demands both effective wound therapy and close monitoring of crucial wound healing biomarkers in the wound exudate; the absence of miniaturized user-interactive fully integrated closed-loop wearable systems and the evaluation of such systems in vivo impede their practical use. To address these challenges, here, we introduce a fully integrated wireless wearable bioelectronic system that effectively monitors the physiological conditions of the wound bed via multiplexed and multimodal wound biomarker analysis and performs combination therapy through electro-responsive controlled drug delivery for anti-inflammatory antimicrobial treatment and exogenous electrical stimulation for tissue regeneration (Fig. 1, A and B). The wearable patch is mechanically flexible, stretchable, and can conformally adhere to the skin wound throughout the entire wound healing process, preventing any undesired discomfort or skin irritation. Because of the wound’s complex pathophysiological environment, compared to previously reported single-analyte sensing, multiplexing analysis of wound exudate biomarkers can provide more comprehensive and personalized information for effective chronic wound management. In this regard, a panel of wound biomarkers including temperature, pH, ammonium, glucose, lactate, and UA were chosen on the basis of their importance in reflecting the infection, metabolic, and inflammatory status of the chronic wounds. Real-time selective monitoring of these biomarkers in complex wound exudate could be realized in situ using custom-engineered electrochemical biosensor arrays (Fig. 1C). The wearable system’s capabilities of multiplexed monitoring, biomarker mapping, and combination therapy were evaluated in vivo over prolonged periods of time in rodent models with infected diabetic wounds. The multiplexed biomarker information collected by the wearable patch revealed both spatial and temporal changes in the microenvironment as well as inflammatory status of the infected wound during different healing stages. In addition, the combination of electrically modulated antibiotic delivery with electrical stimulation on the wearable technology enabled substantially accelerated chronic wound closure. **Fig. 1.:** *A wireless stretchable wearable bioelectronic system for multiplexed monitoring and treatment of chronic wounds.(A) Schematic of a soft wearable patch on an infected chronic nonhealing wound on a diabetic foot. (B) Schematic of layer assembly of the wearable patch, showing the soft and stretchable poly[styrene-b-(ethylene-co-butylene)-b-styrene] (SEBS) substrate, the custom-engineered electrochemical biosensor array, a pair of voltage-modulated electrodes for controlled drug release and electrical stimulation, and an anti-inflammatory and antimicrobial drug-loaded electroactive hydrogel layer. (C) Schematic layout of the smart patch consisting of a temperature (T) sensor, pH, ammonium (NH4+), glucose (Glu), lactate (Lac), and UA sensing electrodes, reference (Ref) and counter electrodes, and a pair of voltage-modulated electrodes for controlled drug release and electrical stimulation. (D and E) Photographs of the fingertip-sized stretchable and flexible wearable patch. Scale bars, 1 cm. (F and G) Schematic diagram (F) and photograph (G) of the fully integrated miniaturized wireless wearable patch. Scale bar, 1 cm. ADC, analog to digital converter; AFE, analog front end; PSoC, programmable system on chip; MUX, multiplexer; BLE, Bluetooth Low Energy. (H) Photograph of a fully integrated wearable patch on a diabetic rat with an open wound. Scale bar, 2 cm.* ## Design of the fully integrated stretchable wearable bioelectronic system The disposable wearable patch consists of a multimodal biosensor array for simultaneous and multiplexed electrochemical sensing of wound exudate biomarkers, a stimulus-responsive electroactive hydrogel loaded with a dual-function anti-inflammatory and antimicrobial peptide (AMP), as well as a pair of voltage-modulated electrodes for controlled drug release and electrical stimulation (Fig. 1, B and C). The multiplexed sensor array patch is fabricated via standard microfabrication protocols on a sacrificial layer of copper followed by transfer printing onto a poly[styrene-b-(ethylene-co-butylene)-b-styrene] (SEBS) thermoplastic elastomer substrate (figs. S1 and S2). The serpentine-like design of electronic interconnects, and the highly elastic nature of SEBS enables high stretchability and resilience of the sensor patch against undesirable physical deformations (Fig. 1, D and E). The flexible bandage seamlessly interfaces with a flexible printed circuit board (FPCB) for electrochemical sensor data acquisition, wireless communication, and programmed voltage modulation for controlled drug delivery and electrical stimulation (Fig. 1, F to H, and figs. S3 to S5). The wireless wearable device can be attached to the wound area with firm adhesion, allowing the animals to move freely over a prolonged period (movie S1 and figs. S6 and S7). ## Design and characterization of the soft sensor array for multiplexed biomarker analysis The array of flexible biosensors was custom developed to allow real-time multiplexed monitoring of the biomarkers in complex wound exudate. The continuous and selective measurement of glucose, lactate, and UA is based on amperometric enzymatic electrodes with glucose oxidase, lactate oxidase, and uricase immobilized in a highly permeable, adhesive, and biocompatible chitosan film, respectively (Fig. 2A). Electrodeposited Prussian blue (PB) serves as the electron-transfer redox mediator for the enzymatic reaction, which allows the biosensors to operate at a low potential (~0.0 V) to minimize the interferences of oxygen and other electroactive molecules. Because of the complex and heterogeneous composition of wound fluid (e.g., high protein levels, local and migrated cells, and exogenous factors such as bacteria) [13], previously reported enzymatic sensors suffer from severe matrix effects and fail to accurately measure the target metabolite levels in untreated wound fluid (figs. S8 and S9 and note S1). Moreover, high levels of metabolites in diabetic wound fluid, especially glucose (up to 50 mM), pose another major challenge to obtain linear sensor response in the physiological concentration ranges. To address these issues and achieve accurate wound fluid metabolic monitoring, increase sensor range, and minimize biofouling effects, we explored the use of an outer porous membrane that serves as a diffusion limiting layer to protect the enzyme, tune response, increase operational stability, as well as enhance the linearity and sensitivity magnitude of the sensor. We fabricated our enzymatic glucose oxidase/chitosan/single-walled carbon nanotubes (GOx/CS/MWCNT) glucose sensor with additional porous membrane coatings including CS, poly(ethylene glycol) diglycidyl ether (PEGDGE), Nafion, and polyurethane (PU) (fig. S9). As expected, the addition of diffusion layers indeed improves the sensor’s linear range in simulated wound fluid (SWF). However, CS-, PEGDGE-, and Nafion-coated sensors did not show reliable responses in wound fluid upon the addition of glucose. The PU-based enzymatic sensors showed the highest linearity over the wide physiological concentration range as well as high reproducibility in complex wound fluid matrix (fig. S10). The amperometric current signals generated from the PU-coated enzymatic glucose, lactate, and UA sensors are proportional to the physiologically relevant concentrations of the corresponding metabolites in SWF with sensitivities of 16.34, 41.44, and 189.60 nA mM−1, respectively (Fig. 2, B to D). Continuous monitoring of ammonium is based on a potentiometric ion-selective electrode where the binding of ammonium with its ionophore results in an electrode potential log-linearly corresponding to the target ion concentration with a sensitivity of 59.7 mV decade−1 (Fig. 2, E and F). Similarly, the pH sensor uses an electrodeposited polyaniline film as the pH-sensitive membrane and shows a sensitivity of 59.7 mV per pH (Fig. 2G). For all chemical sensors, a polyvinyl butyral (PVB)–coated Ag/AgCl electrode was used as the reference electrode that provides a stable voltage independent of the variations of wound fluid compositions [24]. A gold microwire-based resistive temperature sensor is integrated as part of the sensor array and shows a sensitivity of approximately $0.21\%$ °C−1 in the physiological temperature range of 25° to 45°C (Fig. 2H). **Fig. 2.:** *Design and characterization of the sensor array for multiplexed wound analysis.(A to D) Schematic (A) and chronoamperometric responses of the enzymatic glucose (B), lactate (C), and UA (D) sensors in SWF. Insets in (B) to (D), the calibration plots with a linear fit. PB, Prussian blue; Sub, substrate; Prod, product; CE, counter electrode; WE, working electrode; RE, reference electrode; I, current. (E and F) Schematic (E) and potentiometric response (F) of an NH4+ sensor in SWF. Insets in (F), the calibration plot with a linear fit. ISE, ion-selective electrode; PEDOT, poly(3,4-ethylenedioxythiophene); U, potential. (G) Potentiometric response of a polyaniline-based pH sensor in McIlvaine buffer. Insets, the calibration plot with a linear fit. (H) Resistive response of an Au microwire–based temperature sensor under temperature changes in physiologically relevant range in SWF. Insets, schematic of a temperature sensor and the calibration plot with a linear fit. All error bars in (A) to (H) represent the SD from three sensors. (I) Selectivity study of the multiplexed sensor array in SWF. Ten millimolar glucose, 50 μM UA, 1 mM lactate, and 1 mM NH4+ were added sequentially to the SWF. (J) Responses of the multiplexed sensor array before and during mechanical stretching (15%) in SWF (pH 8) containing 10 mM glucose, 50 μM UA, 1 mM lactate, and 0.25 mM NH4+. (K and L) Representative live (green)/dead (red) images of human dermal fibroblasts (HDFs) (K) and normal human epidermal keratinocytes (NHEKs) (L) cells seeded on the multiplexed sensor array and in PBS (control) after 1-day and 7-day culture. Scale bars, 200 μm. (M and N) Quantitative analysis of cell viability images (M) and cell metabolic activity (N) over a 7-day period after culture. RFUs, relative fluorescence units. Error bars represent the SD (n = 4).* Considering that other electrolytes and metabolites present in wound fluid may negatively affect the sensor outputs, we examined the selectivity of the sensor array consisting of all six sensors. As illustrated in Fig. 2I, the addition of nontarget electrolytes and metabolites did not trigger any substantial interference to the sensor response. Moreover, all biosensors showed high selectivity over nonspecific compounds when evaluated in SWF (fig. S11). It should be noted that while temperature has negligible effects on the potentiometric sensors, it substantiallyinfluences the performance of the enzymatic sensors due to the temperature-dependent enzyme activities (fig. S12). Moreover, our data show that the medium pH could also affect the performance of enzymatic sensors (fig. S13). With pH and temperature sensors integrated into the wearable patch, we are able to perform real-time adjustments and calibration of the enzymatic biosensors based on temperature and pH variations to realize accurate wound metabolite analysis. Owing to the soft SEBS substrate and the serpentine-like design of electronic interconnects, the wound patch showed excellent mechanical flexibility and stretchability, which are essential to maintaining good contact with the skin in vivo during the chronic wound healing process. Negligible alterations in the sensor responses before and under unidirectional tensile stretching (Fig. 2I) and after repetitive mechanical bending (fig. S14) were observed, indicating highly consistent sensor performance under various physical deformations. As the sensor patch is designed for long-term in vivo use, its cytocompatibility and biocompatibility are of great importance. Cell viability and metabolic activity of the cells seeded on a multiplexed sensor array were analyzed using a commercial live/dead kit and PrestoBlue assay, respectively (Fig. 2, K to N, and fig. S15). The high cell viabilities shown in the representative live/dead staining images of human dermal fibroblasts (HDFs) and normal human epidermal keratinocytes (NHEK) cells (Fig. 2, K to M, and fig. S15), along with the consistently increased cell metabolic activities (Fig. 2N) over multiday culture periods, indicate the high cytocompatibility of the soft sensor patch. ## Characterization of the therapeutic capabilities of the wearable patch in vitro In addition to the multiplexed and multimodal biosensing, the wearable patch is able to perform combination treatment of chronic wounds through drug release from an electroactive hydrogel layer and electrical stimulation under an exogenic electric field, both controlled by a pair of voltage-modulated electrodes (Fig. 3, A to C). The electroactive hydrogel consists of chondroitin 4-sulfate (CS), a sulfated glycosaminoglycan composed of units of glucosamine, cross-linked with 1,4-butanediol diglycidyl ether (fig. S16). Because of the shear-thinning behavior of the prepolymer solution, the hydrogel can be precisely fabricated via three-dimensional (3D) printing (fig. S17). The negatively charged CS hydrogel is an ideal choice for loading and controlled release of positively charged large biological drug molecules based on an electrically modulated “on/off” drug release mechanism (Fig. 3B). Here, an AMP, thrombin-derived c-terminal peptide-25 (TCP-25) [38], was loaded within the CS hydrogel network through the electrostatic interactions with the polymer backbone, with up to $15\%$ loading efficiency (Fig. 3D). The highly porous hydrogel network under equilibrium swelling could further enhance the drug loading efficiency (fig. S18). Under an applied positive voltage, the electroactive hydrogels will be rapidly protonated, resulting in anisotropic and microscopic contraction followed by syneresis/expelling of water from the gel [39] and consequently allowing a controlled release of the TCP-25 AMP (Fig. 3, E and F, and figs. S19 and S20). In addition, the electrical field will also facilitate the diffusion of positively charged AMP out of the stimuli-sensitive CS hydrogel toward the cathode due to electrophoretic flow [40]. **Fig. 3.:** *Characterization of the therapeutic capabilities of the wearable patch in vitro.(A to C) Schematic illustration of the therapeutic modules of the wearable patch (A) and the working mechanisms of the controlled drug delivery for antimicrobial treatment (B) and electrical stimulation for tissue regeneration (C). (D) Loading efficiency of dual-functional TCP-25 anti-inflammatory and AMP into CS electroactive hydrogel after 0.5- to 24-hour incubation. (E) Release amount of AMP from the hydrogel under programmed on-off electrical voltage (1 V, 10 min each step). (F) Long-term cumulative release of the AMP under programmed electrical modulation. (G and H) In vitro antimicrobial tests including zone of inhibition (G) and colony forming units (H) assays for electroactive hydrogels with and without TCP-25 AMP against multidrug-resistant Escherichia coli (MDR E. coli), P. aeruginosa, and methicillin-resistant Staphylococcus aureus (MRSA). (I to K) In vitro cytocompatibility assessment of TCP-25–loaded electroactive hydrogels using live/dead staining (I) and quantification of cell viability (J) and metabolic activity (K) for HDF and NHEK cells cultured in the presence of hydrogels. Scale bar, 100 μm. (L and M) Fluorescence images (L) and quantitative wound closure analysis (M) to evaluate the wearable patch’s therapeutic capability via electrical stimulation using an in vitro circular wound healing assay created by HDF cells. ES, electrical stimulation. A pulsed voltage was applied for electrical stimulation (1 V at 50 Hz, 0.01 s voltage on for each cycle). Scale bar, 500 μm. (N) Numerical simulation of the electrical field generated by the custom-designed electrical stimulation electrodes during operation. E, electrical field. Scale bar, 500 μm. Error bars represent the SD (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; n ≥ 3). ns, not significant.* The antimicrobial activity of the TCP-25 AMP–loaded hydrogel was evaluated against Gram-positive methicillin-resistant *Staphylococcus aureus* (MRSA) and Pseudomonas aeruginosa, and Gram-negative multidrug-resistant *Escherichia coli* (MDR E. coli) and Staphylococcus epidermidis, the most common pathogenic bacteria associated with microbial colonization of chronic nonhealing wounds (Fig. 3, G and H, and fig. S21). The zone of inhibition assay indicates the susceptibility of the MDR E. coli, P. aeruginosa, and MRSA toward TCP-25 AMP (Fig. 3G), while the standard colony-forming units (CFU) showed that the drug-loaded hydrogel was effectively protected from all pathogenic colonization (Fig. 3H). For cells cultured on antimicrobial peptide (AMP)-loaded hydrogels, the viability of HDF and NHEK cells remained >$90\%$, and their metabolic consistently increased during the 7-day culture (Fig. 3, I to K, and fig. S22), indicating that the TCP-25–loaded gels are highly cytocompatible and support cell proliferation. The wearable patch’s therapeutic capability toward enhanced tissue regeneration via electrical stimulation was assessed using an in vitro wound healing assay (Fig. 3, L and M, and fig. S23). The model wound treated with electrical stimulation showed substantially faster and more consistent migration of HDF cells toward the wound area for four consequent days after wounding as compared to the control group without electrical stimulation (Fig. 3L). Quantitative analysis of the model wound closure indicates higher wound closure rates in the wounds treated with electrical stimulation (Fig. 3M). The enhanced tissue regeneration is attributed to the directional electrical field generated from our custom-designed electrical stimulation electrodes (Fig. 3N), which plays a crucial role in cell behavior modulation including cell-cell junctions, cell division orientation, and cell migration trajectories (galvanotaxis or electrotaxis) (41–43). The electrical potential was applied directly to a pair of insulated electrodes to generate electrical field for electrical stimulation. It should also be noted that continuous electrical stimulation did not cause substantial temperature increase (fig. S24). ## Evaluation of the wearable patch in vivo for multiplexed wound biomarker monitoring To validate the capability and efficacy of our wearable patch, in vivo preclinical evaluations are essential. In this regard, the in vivo biocompatibility of the wearable patch was assessed. The immunohistofluorescent staining of subcutaneously implanted hydrogel and electrodes in rats showed negligible signs of leukocyte (CD3) and macrophage (CD68) antigens after 56 days, indicating the high biocompatibility of the wearable patch (fig. S25). All custom-developed biosensors on the wearable patch displayed consistent sensitivity during a 6-hour continuous measurement in SWF, indicating the high electrochemical stability of the sensors for wound analysis (fig. S26). In vivo multiplexed sensing study was then performed using an infected excisional wound model in diabetic mice. The wound fluid composition was assessed by the wearable patch before infection (day 1), after infection (day 4), and after treatment (day 7) (Fig. 4A). Substantially elevated UA, temperature, pH, lactate, and ammonium levels were observed as compared to those before infection. The increase in temperature can be potentially linked to inflammation [44]. The elevated levels of UA after infection can be due to up-regulation of xanthine oxidase, a component of the innate immune system responding to inflammatory cytokines in chronic ulcers that plays a key role in purine metabolism to produce UA [45]. pH, lactate, and ammonium are all acidity related, and their elevation during the bacteria infection has also been widely reported [46]. In contrast, the glucose level in infected wound fluid showed >$35\%$ decrease after infection, attributing to the increased glucose consumption of bacteria activities [16]. Upon wound treatment, the temperature, pH, lactate, UA, and ammonium decreased toward the levels before the infection, while the glucose level increased significantly after treatment, indicating the successful bacterial elimination (Fig. 4A). **Fig. 4.:** *In vivo evaluation of the wearable patch for multiplexed wound biomarker monitoring in a wound model in diabetic mice.(A) In vivo multiplexed analysis of the chemical composition of wound fluid using a wearable patch in an infected excisional wound model in a diabetic mouse. Infection and treatment were performed after the sensor recording on days 1 and 4, respectively. (B) In vivo continuous and multiplexed evaluation of wound parameters in a 24-hour fasted mouse before and after glucose administration via tail vein. (C) In vivo assessment of metabolic changes in wound microenvironment in response to fasting and food feeding in a diabetic mouse.* Considering that dietary intake may have major impact on the composition of diabetic wound fluid, we evaluated the metabolic changes in wound fluid in response to tail vein glucose administration (Fig. 4B) and food feeding (Fig. 4C). Glucose administration via tail vein into the 24-hour fasted mice sparked ~10 mM increase in the blood glucose level. The in vivo sensor readings from the wearable patch were recorded from 30 min before injection and continued until 270 min after injection (Fig. 4B). The glucose level in wound fluid showed a gradual increase throughout the 4 hours after injection, indicating a protracted delay with respect to blood glucose. A similar trend was observed for temperature values, attributing to an increased metabolic rate to facilitate digestion. No apparent change in UA level after injection was detected because of the absence of purine intake in the glucose administration. For the food feeding study, the wearable patch was tested before fasting, after 24-hour fasting, and 6 hours after feeding (Fig. 4C). The lactate and ammonium levels increased substantially after fasting, while glucose and UA levels decreased after fasting, consistent with the trend of observed blood level changes [47]. In the meantime, temperature decreased due to the fasting-induced hypothermia [48]. As expected, 6 hours after feeding, the glucose and UA levels increased from 11.9 to 20.3 mM and from 45.9 to 60.3 μM, respectively. These results indicate that wearable patch-enabled wound fluid analysis could be a promising approach to realize continuous and personalized metabolic monitoring. ## Spatial and temporal monitoring of critical-sized wounds using the wearable patch The wearable patch is mass producible and readily reconfigurable for various wound care applications. In the case of large chronic ulcers, the wound parameters and microenvironment may vary from site to site, making localized monitoring crucial for optimized assessment and treatment of chronic wound infection. As a proof of concept, we demonstrate customized wearable patches for spatial mapping of physiological conditions of critical-sized wounds during the healing process. As illustrated in Fig. 5 (A and B), we could incorporate a sensor array containing seven pH sensors and nine temperature sensors onto our wearable platform for monitoring and mapping critical-sized full-thickness infected chronic wounds in diabetic rats. The pH and temperature sensor arrays showed high reproducibility and stability in SWF solutions before and after in vivo application (Fig. 5, C and D, and fig. S27). **Fig. 5.:** *Spatial and temporal monitoring of critical-sized full-thickness infected wound defects in diabetic rats using the wearable patch.(A and B) Schematic (A) and photograph (B) of a soft sensor patch with pH and temperature sensor arrays designed for spatial and temporal monitoring of large and irregular wounds. Scale bar, 1 cm. (C and D) The characterization of pH (C) and temperature (D) sensor arrays on a wearable patch in SWF solutions. (E and F) Dynamic changes in pH (E) and temperature (F) values of each biosensor on a wearable patch for critical-sized noninfected and infected wounds. (G and H) The mapping of daily local pH (G) and temperature (H) sensor readings in the wound area for infected and noninfected wounds on each day over the 7-day study period.* On-body validation of the sensor array for spatial and temporal wound monitoring was conducted on critical-sized full-thickness wounds (35 mm in diameter) in Zucker diabetic fatty (ZDF) rats before infection, after infection, and after treatment. The dynamic changes in pH and temperature values for each biosensor on the wearable patch in noninfected and infected critical-sized wounds are illustrated in Fig. 5 (E and F). For noninfected wound studies, the pH and temperature values did not notably change over the 7-day period. However, for infected wound studies, the pH and temperature values increased daily upon applying a mixed infection (MRSA and P. aeruginosa) on day 1 and reached the peak value on days 3 and 4. Upon treatment on day 4, the pH and temperature values for each sensor decreased substantially and recovered toward the levels before the infection on day 7. The spatial mapping plots of pH (Fig. 5G) and temperature (Fig. 5H) in the chronic wound area on each day over the 7-day period were successfully generated on the basis of localized sensor readings. These results are in agreement with previous literature on the changes in the pH and temperature values during the healing progress [46]. A wide variation was observed in both pH and temperature in different regions of the wound upon bacterial infiltration on day 2, showing a higher bacteria growth in the wound edges. The infected wound showed a more uniform pH and temperature at different regions 2 and 3 days after infection due to the formation of uniform biofilm. Upon treatment, the variations increased in the treated wounds on days 5 and 6, indicating the disruption and eventually elimination of the biofilm after treatment (Fig. 5, G and H). ## Evaluation of the therapeutic efficacy of the wearable patch in chronic wound healing in vivo The wearable patch-facilitated combination therapy and wound healing were evaluated in a splinted excisional wound model in ZDF diabetic rats (Fig. 6A). Four different groups were tested: negative control, drug release, electrical stimulation, and combination therapy. The drug treatment was primarily used to eliminate bacterial infections and regulate immune response in early stages of healing. The electrical stimulation was used to facilitate ion channel up-regulation and redistribution, resulting in accelerated cell migration and wound healing. The wearable patch’s high flexibility and stretchability provided intact and comfortable contact with the animal’s back curvature. Over a 14-day period, the animals were routinely weighed where infected rats showed a nonsubstantially lower body weight compared to noninfected animals (fig. S28), indicating that the study procedures did not have any substantial influence on the animals’ health. Moreover, the standard CFU on the mixed infection isolated from the wound beds 3 days after drug and combination therapy groups showed a significant reduction in bacterial growth as compared to control and electrical stimulation groups, suggesting effectiveness of the wearable patch in the elimination of pathogenic species from the wound (fig. S29). Substantially higher rates of wound closure were observed in the treated wounds as compared to the control untreated group, where the group that received combination therapy showed the highest wound closure rate, collagen deposition, and granulation tissue formation, suggesting the recovery of the wound toward the unwounded state (Fig. 6, B and C, and fig. S30). We also evaluated the use of the wearable system for multiplexed biosensing and the combination therapy on the same diabetic rats (fig. 31): Compared to the individual evaluation as shown in Figs. 5 and 6 (B and C), similar sensing results and therapeutic effects to the individual evaluation were observed: *The continuous* sensing data were obtained up to 8 days until the wound dried after therapy while the wound fully closed 14 days after surgery. **Fig. 6.:** *In vivo evaluation of wearable patch-facilitated chronic wound healing in full-thickness infected wounds in ZDF diabetic rats.(A) Schematic of the wearable patch on a diabetic wound and the working diagram of combination therapy. (B and C) Representative images (B) and quantitative analysis of wound closure (C) for the control wound and wounds treated with drug, ES, and combination therapy on days 3 and 14 after application. Scale bar, 500 μm. (D) Representative images of Masson’s trichrome (MTC)–stained sections of the full-thickness skin wounds after 14 days of combination treatment. Scale bars, 500 μm. (E) Representative immunofluorescent stained images for nuclear factor κB (NF-κB) (purple), keratin 14 (Krt14) (green), and phosphatase and tensin homolog (Pten) (red) 14 days after the treatment. Scale bars, 500 μm. (F and G) Quantitative analysis of scar elevation index (SEI) based on MTC images (F) and Krt14 marker based on immunofluorescent images (G). (H) Quantitative real-time polymerase chain reaction (qRT-PCR) analysis of a library of wound biomarkers for wound biopsies after 3 and 14 days of treatment. (I to M) Relative expression of Pdgfa (I), Fgf (J), Serpine1 (K), IL-6 (L), and Stat3 (M) genes after 3 and 14 days of treatment. Error bars represent the SD (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; n = 3).* In addition, the histopathological analysis of sections of the full-thickness skin wounds via Masson’s trichrome (MTC) staining (Fig. 6D) and immunofluorescent staining (Fig. 6E) was performed. The MTC images showed a significantly higher collagen deposition and granulation tissue formation for treated groups compared to the control group on day 14 (Fig. 6D). Moreover, the control group on day 14 showed a significantly higher scar elevation index (SEI) of 175 ± $59\%$, indicating the formation of hypertrophic scars; in contrast, the SEI for combination treatment group was 100 ± $4\%$, showing uniform dermis repair after treatment (Fig. 6F). The combination therapy was able to accelerate the wound-induced hair follicle neogenesis with adjoining sebaceous glands within the wound bed (Fig. 6, D and E, insets) resembling structurally similar glands to those of the uninjured skin [49]. The immunohistochemical analysis of keratin 14 (Krt14), a marker of undifferentiated keratinocytes, revealed a delayed re-epithelization in the control group ($16\%$) as compared to the substantially accelerated re-epithelization in combination treatment group ($99\%$) after 14 days (Fig. 6, E and G, and fig. S30). We further observed a significant growth in the expression of tumor suppressor phosphatase and tensin homolog (Pten), an indicator of higher electrotactic responses [42], among electrical stimulation and combination treatment groups as a direct result of electrical stimulation (Fig. 6E and fig. S30). A higher expression of nuclear factor κB (NF-κB) enhancer binding protein, a key signaling factor that promotes remodeling of cellular junctions, cell proliferation, and adhesion [50], was also observed in the combination therapy group on day 14, indicating a higher cutaneous wound healing (Fig. 6E and fig. S30). We further studied the molecular mechanism behind the beneficial effects of our wearable patch’s combination treatment on wound healing using quantitative real-time polymerase chain reaction (qRT-PCR) analysis. A library of the most relevant genes associated with wound healing was screened. A substantially elevated expression of growth factors was confirmed in the combination-treated wounds (Fig. 6H, fig. S32, and note S2), while the control group on day 14 showed a reduced level of these growth factors due to compromised cutaneous wound healing that resulted in impaired re-epithelization and the formation of granulation tissue and ECM. Considering that platelet-derived growth factor subunit A (Pdgfa) plays crucial roles in stimulation of fibroblast proliferation (early function) and induces the myofibroblast phenotype (later function) [51], its elevation supports the higher rate of dermis and granulation tissue formation in the combination treatment group on day 14, while lower or delayed *Pdgfa* gene expression resulted in impaired wound healing in drug and electrical stimulation groups in the same period (Fig. 6I). The overexpression of *Pdgfa* gene in control group after 14 days might be due to the pathogenesis of hypertrophic scars and increased responsiveness of keloid fibroblasts to Pdgf [52]. In addition, the higher expression of fibroblast growth factor (Fgf) genes can be due to higher rate of epidermis regeneration, renewed capillaries, and in cells infiltrating in the granulation tissue (Fig. 6J) [53]. There was also a significant up-regulation of serine protease inhibitor clade E member 1 (Serpine1) genes in electrical stimulation and combination treatment groups as compared to the control and drug groups primarily due to the applied electrical field (Fig. 6K) [54]. Serpine1 regulates the extent and location of matrix restructuring and collagen remodeling while facilitating cell motility and proliferation in the process of wound regeneration. Moreover, significant up-regulations of proinflammatory cytokine interleukins-6 (IL-6) (Fig. 6L) and signal transducers and activators of transcription 3 (Stat3) (Fig. 6M) were observed in electrical stimulation and combination therapy groups on day 3. The IL-6 can positively influence different processes at the wound site, including stimulation of keratinocyte and fibroblast proliferation, synthesis and breakdown of ECM proteins, fibroblast chemotaxis, and regulation of the immune response [38], while Stats are cytoplasmic proteins that can transduce signals from a variety of growth factors and regulate target gene expression. Stat3 can be activated upon binding of IL-6 to its receptor and thus plays a key role in wound healing [55]. These results further confirmed the powerful combinatorial therapeutic capabilities of the wearable patch to accelerate chronic wound healing. ## DISCUSSION We present the development of a wireless wearable bioelectronic system consisting of a multimodal biosensor array for multiplexed monitoring of wound exudate biomarkers, a stimulus-responsive drug-loaded electroactive hydrogel, and a pair of voltage-modulated electrodes for controlled drug release and electrical stimulation. The wearable patch is fully biocompatible, mechanically flexible, stretchable, skin-conformal, and is capable of real-time selective monitoring of a panel of crucial wound biomarkers including temperature, pH, ammonium, glucose, lactate, and UA in multiple rodent models. The wearable patch demonstrated here represents a versatile platform for evaluating wound conditions and intelligent therapy and can be easily reconfigured to monitor several other metabolic and inflammatory biomarkers for various chronic wound care applications. Despite remarkable progress in developing wearable electrochemical biosensors for continuous monitoring of circulating metabolites in interstitial fluid and human sweat (24, 26, 56–58), in situ wound fluid analysis remains a major clinical challenge. This is mainly due to the complex and heterogeneous composition of wound fluid (e.g., high protein levels, local and migrated cells, and exogenous factors such as bacteria) that leads to severe and unique matrix effects for most previously reported biosensors and failure in accurate measurement of the target metabolite levels in wound fluid [31]. To mitigate such issue, here, we introduced the use of an outer porous PU-based membrane that serves as an analyte diffusion limiting layer to protect the electrode, tune response, increase long-term operational stability, linearity, and sensitivity magnitude as well as biocompatibility and mechanical stability of the sensor [59]. Our results indicate that the wearable patch-enabled wound fluid analysis could be a promising approach to realize continuous and personalized wound metabolic monitoring in both a temporal and spatial fashion. In addition to the multiplexed biosensors, the wearable patch is equipped with an on-demand electro-responsive drug release system, loaded with an antimicrobial and anti-inflammatory peptide. Under an applied positive voltage, the electroactive hydrogels will rapidly release the dual-function peptide that could effectively eliminate bacteria and modulate inflammatory responses in the wound site during the initial stages of healing, in a splinted excisional wound model in diabetic rats. The on-demand drug delivery can be readily modified with different electroactive hydrogels to deliver several other positively or negatively charged drugs and biomolecules (e.g., proteins, peptides, and growth factors). The integration of an electrical stimulation therapeutic module could facilitate cell motility and proliferation and ECM deposition and remodeling in the process of wound regeneration resulting in rapid and effective cutaneous wound healing. We demonstrated promising preliminary data for multiplexed in situ metabolic monitoring. However, one limitation of the current study is the lack of a continuous wound fluid sampling and circulation system. The mixing of newly secreted analytes with the old ones delayed the sensor response, leading to a compromised temporal sensing resolution. In addition, the long-term continuous operation stability of the biosensors in wound fluid in vivo may need further improvement. Additional antifouling protective membranes could be explored to minimize the influence of complex would fluid on sensor performance during in situ use. Compared to large critical-sized wounds in large animals (e.g., pigs), the spatial biomarker mapping of wounds in rodent models did not reveal substantial spatial variations. Future investigations can focus on using a microfluidic wound fluid sampling system for efficient capture and continuous delivery of wound fluid to the sensor chamber to improve the temporal resolution of in situ biomarker detection [27, 56]. Moreover, to improve wearable patch’s durability, low-power electronics or energy-harvesting modules could also be implemented into the wearable platform (60–63). The clinical technology transfer of this product will require multiple further in-depth studies including preclinical biocompatibility evaluation, long-term multiplexed sensor analysis, and efficacy assessment of the closed-loop therapeutic and regenerative modules in pig models due to anatomical, physiological, and functional similarities of pig and human skin wound healing. Further in-depth studies of the cellular and molecular mechanisms behind wound regeneration upon applying the wearable patch’s combination therapy via single-cell RNA sequencing would be beneficial. Another further device development direction to benefit future evaluation is scale-up manufacturing, packaging, and reliability assessment toward first-in-human studies. We envision that the custom-engineered fully integrated wearable patch could serve as a more effective, fully controllable, and easy-to-implement platform for personalized monitoring and treatment of chronic wounds with minimal side effects. ## Materials and reagents Tetrahydrofuran (THF), PVB resin BUTVAR B-98, sodium chloride, ammonium chloride, gelatin (from bovine skin), sodium thiosulfate pentahydrate, sodium bisulfite, ammonium ionophore I, aniline, CS, 1,4-butanediol diglycidyl ether (BDDE), 3,4-ethylenedioxythiophene (EDOT), poly(sodium 4-styrenesulfonate) (NaPSS), PU, glucose oxidase, uricase, chitosan, iron (III) chloride, potassium ferricyanide (III), paraformaldehyde, UA, and multiwalled carbon nanotubes (CNTs) were obtained from Sigma-Aldrich. l-lactate oxidase was purchased from Toyobo Corp. Hydrochloric acid, acetic acid, methanol, ethanol, acetone, urea, dextrose (d-glucose), and Dulbecco’s phosphate-buffered saline (DPBS) were acquired from Thermo Fisher Scientific. The SEBS polymer was obtained from Asahi Kasei Corporation. TCP-25 (GKYGFYTHVFRLKKWIQKVIDQFGE) ($98\%$ purity, acetate salt) and tetramethylrhodamine-labeled TCP-25 were purchased from CPC Scientific. ## Fabrication of the soft wearable patch Briefly, a 300-nm-thick sacrificial layer of copper was first deposited on the silicon wafer using e-beam evaporation (CHA Industries Mark 40) at a speed of 2.5 Å s−1, followed by standard photolithography (Microchemicals GmbH, AZ 5214) to define the connection wires. Cr/Au/Cr ($\frac{1}{100}$/20 nm) was deposited on the sacrificial copper through e-beam evaporation of at a speed of 0.2, 0.5, and 0.2 Å/s, respectively, followed by lift-off in acetone. SEBS (200 mg ml−1 in toluene) was then spin-coated with a speed of 300 revolutions per minute (rpm) for 30 s. The SEBS film was cured at 70°C for 1 hour to remove toluene, and the resulting SEBS film had a thickness of ~300 μm. The copper sacrificial layer was then chemically removed by immersing the silicon wafer in copper etchant (APS-100) for 12 hours. The patch was then picked up by a polydimethylsiloxane (PDMS) stamp and rinsed with deionized (DI) water thoroughly. A thin layer of parylene (ParaTech LabTop 3000 Parylene coater) was deposited (200 nm), followed by photolithography and reactive-ion etching (RIE) (Oxford Plasmalab, 100 ICP/RIE, 30 SCCM of O2, 100 W, 50 mtorr, 90 s) to expose openings for sensor modifications and pin connections. Laser patterning via a 50-W CO2 laser cutter (Universal Laser Systems; power, $20\%$; speed, $50\%$; points per inch, 1000; and vector mode) was used to define the patch shape and outline. After sensor modification, a water-soluble tape (AQUASOL) was used to pick up the wound patch from PDMS backings for further use. ## Enzymatic sensors To increase the electrode surface area for enzymatic sensors, a nanostructured Au film was electrodeposited on Au electrodes in a solution containing 50 mM chloroauric acid and 0.1 M HCl using multipotential deposition for 1500 cycles (for each cycle, −0.9 V for 0.02 s and 0.9 V for 0.02 s). For glucose and lactate sensors, a PB layer was deposited onto the Au electrodes by 10 cycles of cyclic voltammograms (CVs) (−0.2–0.6 V versus Ag/AgCl) with a scan rate of 50 mV s−1 in a freshly made solution containing 2.5 mM FeCl3, 2.5 mM K3[Fe (CN)6], 100 mM KCl, and 100 mM HCl. For the UA sensor, a PB layer was deposited using the same approach except only one CV cycle of electrodeposition. Next, a chitosan solution was prepared by dissolving $1\%$ chitosan in a $2\%$ acetic acid solution followed by vigorous magnetic stirring for 1 hour. The resulting solution was then mixed with CNTs (2 mg ml−1) by ultrasonic agitation over 30 min to prepare a chitosan/CNT solution. To prepare all enzymatic sensors, the chitosan/CNT solution was mixed thoroughly with an enzyme solution [10 mg ml−1 in PBS (pH 7.2)] with a volume ratio of 2:1. Next, 1 μl of the enzyme/chitosan/CNT cocktail was drop-casted onto the PB/Au electrode and dried under 4°C. Last, the PU layer was prepared by drop-casting 4.5 μl of 15 mg ml−1 of PU solution in a solvent mixture containing THF and N,N′-dimethylformamide (volume ratio, 98:2) on the enzyme layer and air-dried overnight under 4°C. ## pH sensor The pH sensor was based on pH-sensitive polyaniline film deposited on a Au electrode. First, the working electrode was electrochemically cleaned via 10 cycles of CVs with a scan rate of 0.1 V s−1 in 0.5 M HCl (−0.1–0.9 V). Next, the polyaniline electro-polymerization was performed in a 50-μl solution containing 0.1 M aniline and 1 M HCl via 12 CV cycles (−0.2–1.0 V) with a scan rate of 0.1 V s−1. The fresh solution was then used for another 12 CV cycles. Last, pH electrodes were air-dried overnight. ## Ammonium sensor A PEDOT:PSS film was electrodeposited using a constant current of 0.2 mA cm−2 for 10 min in a solution prepared by dissolving ferrocyanide (30 mg), NaPSS (206.1 mg), and EDOT (10.7 μl) in 10 ml of DI water. Next, an NH4+ selective membrane cocktail solution was prepared by dissolving 1 mg of ammonium ionophore I, 33 mg of polyvinyl chloride, and 66 mg of bis(2-ethylhexyl)sebacate (DOS) in 660 μl of THF. A 1.5 μl of the cocktail solution was then drop-casted on the PEDOT layer to create an ammonium-selective membrane and air-dried overnight. ## Reference electrode To prepare the Ag/AgCl reference electrode, silver was electrodeposited at −0.2 mA for 100 s using a plating solution containing 250 mM silver nitrate, 750 mM sodium thiosulfate, and 500 mM sodium bisulfite. Ten-microliter solution of 0.1 M FeCl3 was dropped on the Ag electrode for 90 s. Next, a solid-state reference membrane cocktail was prepared by dissolving 78.1 mg of PVB and 50 mg of NaCl in 1 ml of methanol followed by vigorous agitation in an ultrasonic bath for 30 min. Next, a 2.5 μl of the reference cocktail membrane was drop-casted on the Ag/AgCl electrode surface and air-dried overnight. ## The characterization of multiplex biosensors The multiplex sensor patches were characterized to evaluate their sensitivity, stability, and reproducibility in solutions of target analytes in SWF using a 1000C Multi-Potentiostat (8-channel) (CH Instruments Inc., Austin, TX, USA). The SWF solution was prepared by dissolving 584.4 mg of NaCl, 336.0 mg of NaHCO₃, 29.8 mg of KCl, 27.8 mg of CaCl2, and 3.30 g of bovine serum albumin in 100 ml of DI water. The enzymatic sensors were characterized chronoamperometrically in 0 to 40 mM glucose, 0 to 4 mM lactate, and 0 to 150 μM UA, at a potential of 0 V. The pH sensor calibration was performed in McIlvaine buffer solutions. Both pH and ammonium sensors were characterized electrochemically using open circuit potential. ## Electroactive hydrogel synthesis and 3D printing Three hundred milligrams of CS was dissolved in 1.14 ml of 1 M NaOH under vigorous stirring. Next, 279 μl of BDDE cross-linker was added and mixed thoroughly for another 30 min. An Anton Paar MCR302 rheometer equipped with a parallel plate to perform rheological characterization. Dynamic viscosity of the samples was measured as a function of shear rate. 3D hydrogel printing was performed on the basis of a custom-designed 3D printer based on a gantry system (A3200, Aerotech) and a benchtop dispenser (Ultimus V, Nordson EFD). One hundred fifty–micrometer nozzles were used for the printing. The pump pressure was set to be 14 kPa, and the nozzle moving speed was set at 5 mm s−1. The printed hydrogel was placed under 60°C for 60 min to form the cross-linked network of the electroactive hydrogels. The cross-linked hydrogels were then left in DI water at 4°C for 48 hours (with water replacement every 12 hours) to obtain equilibrium swelling. ## Drug loading and release studies The AMP was loaded into the hydrogel by incubating swollen electroactive gel in 1.5 ml of AMP solution (2 mg ml−1 in DPBS) in a sealed 12-well plate under 4°C for 24 hours. Passive as well as electro-stimulated release was examined at room temperature in a DPBS solution using the wearable patch. The AMP release was quantified by measuring fluorescence signals using a Synergy HTX Multi-Mode Reader (BioTek Instruments) spectrophotometer at 570 and 583 nm. ## Swelling studies The initial wet weight of each prepared hydrogel was documented. The samples were then immersed in DI water, and the hydrated samples were temporarily taken out of the water and weighed at 1, 4, 8, 24, and 48 hours. The swelling ratio was calculated as the weight gain divided by the original weight before hydration. ## Cell lines Normal Adult HDF cells (Lonza) and NHEKs (Lonza) were cultured under 37°C and $5\%$ CO2. Cells were passaged at $70\%$ confluency, and a passage number of 3 to 5 was used for all studies. ## In vitro cytocompatibility studies The electroactive hydrogels were washed and transferred to 24-well cell culture inserts (cell culture on permeable supports). The wells were seeded with HDFs and NHEKs (1 × 105 cells per well). The inserts were then placed in cell seeded 24-well plates, and cells were treated with appropriate media and incubated under 37°C and $5\%$ CO2 for the course of study. A similar study was performed for wearable patches with the cells directly seeded on the patches. ## Evaluation of cell proliferation and viability A commercial calcein AM/ethidium homodimer-1 live/dead kit (Invitrogen) and commercial PrestoBlue assays (Thermo Fisher Scientific) were used to evaluate cell viability and cell metabolic activity, respectively. In the live/dead assay, the samples were imaged with an Axio Observer inverted microscope (ZEISS); live cells were stained green with calcein-AM, whereas dead cells were stained red with ethidium homodimer-1. Using ImageJ software, cell viability was calculated as the percentage ratio of number of live cells to the number of total cells (live + dead). ## In vitro wound healing assay For in vivo wound healing assay (circular wound), first, a gelatin solution was prepared by dissolving gelatin in DI water (300 mg ml−1) and filtered with a sterile polyethersulfone syringe filter (0.22 μm in pore size). Then, 50 μl of the solution was dropped in the center of each well in 12-well plates. Before cell seeding, the plates were kept at room temperature under sterile conditions to keep gelatin in solid condition. Next, HDF cells with a density of 1 × 105 cells per well were seeded in each well and incubated at 37°C and $5\%$ CO2. The inherent thermoresponsive properties of gelatin allowed slow dissolving of the gel into the media, creating a uniform-sized wound in the center of cells adhered to the plate. The medium was then replaced by fresh media after 4 hours, and the wound closure was assessed daily for up to 4 days. ## Bacterial cells Methicillin-resistant S. aureus [American Type Culture Collection (ATCC) BAA-2313], P. aeruginosa (ATCC 15442), MDR E. coli (ATCC BAA-2452), and S. epidermidis (ATCC 12228) were used for antimicrobial tests. ## Minimum inhibitory concentration The minimum inhibitory concentration (MIC) of TCP-25 AMP against different pathogens was evaluated by measuring bacterial optical density. First, bacteria colonies were grown on agar plates containing 15 g l−1 agar and 30 g l−1 Bacto BD tryptic soy broth (TSB) under $5\%$ CO2 at 37°C for 24 hours. Next, the colonies were transferred and dispersed gently to TSB media, grown overnight in a shaker incubator at 37°C. A bacteria solution of 106 CFU ml−1 was prepared for all antimicrobial tests. For MIC test, 200 μl of bacteria solution in TSB was cultured in 96-well plates in the presence of different AMP concentrations (0, 5, 25, 50, 100, 250, 500, and 750 μg ml−1) and incubated at 37°C for 24 hours. Next, the optical density of the solutions was measured, and the relative optical density (as compared to the optical density of the control sample incubated in the absence of AMP) was reported to calculate MIC. ## CFU test Electroactive hydrogels with and without TCP-25 AMP were placed in 24-well plates and incubated with 1 ml of bacteria solution (106 CFU ml−1) in TSB media under 37°C and $5\%$ CO2 for 18 hours. Next, each sample was removed from bacteria solution, washed gently with DPBS (3×), and then placed in microcentrifuge tubes containing 1 ml of DPBS. The tubes were vortexed vigorously at 3000 rpm for 15 min to release bacteria trapped inside the hydrogels. A series of logarithmic dilutions (10, 102, 103, and 104) was then prepared from each solution. Twenty-microliter diluted solutions were then seeded on agar plates, followed by incubation under 37°C and $5\%$ CO2 for 18 hours. The number of colonies was then recorded and reported as CFU. ## Zone of inhibition A 100-μl bacteria solution (106 CFU ml−1) was dispersed uniformly each agar plate. Next, sterilized electroactive hydrogel disks (6 mm in diameter) loaded with AMP or without AMP were placed into 9-mm holes created in agar plates. The zone of inhibition was measured after 18 hours. ## Numerical electrical field simulation Simulation of the electric field generated during electrical stimulation was conducted by using the commercial software COMSOL Multiphysics through finite element method. Tetrahedral elements allowed modeling of the electric field in 3D space with testified accuracy. The electric field is described by∇⋅D=ρE=−∇Vwhere D, ρ, E, and V denote the electric displacement field, charge density, electric field, and electric potential. The device was fixed at the middle of a cubic computational domain. The side length of the computational domain was 100 mm. The relative permittivity above and below the device was set to be 1 and 76.8, respectively. The boundary condition for the computational domain was set byn⋅$D = 0$where n indicates the normal to surface of the boundary. The potential of the anode was set to be 1 V, and the potential of the ground electrode was set to be 0 V. ## Wireless system integration of the wearable patch A four-layer FPCB with a rounded rectangle (36.5 mm by 25.5 mm) geometry was designed using EAGLE CAD. The sensor patch was interfaced directly underneath the FPCB through a rectangular cutout (12 mm by 3.8 mm). The power management circuitry consists of a magnetic reed switch (MK24-B-3, Standex-Meder Electronics) and a voltage regulator (ADP162, Analog Devices). The electrical stimulation and drug delivery circuitry consist of a series voltage reference (ISL60002, Renesas Electronics), an operational amplifier square wave generator circuit (MAX40108, Maxim Integrated), and a switch array (TMUX1112, Texas Instruments). The potentiometric, amperometric, and temperature sensor interface circuitry consists of a voltage buffer array (MAX40018, Maxim Integrated), a switch array (TMUX1112, Texas Instruments), a voltage divider, and an electrochemical analog front-end (AD5941, Analog Devices). A programmable system on chip Bluetooth Low Energy (BLE) module (CYBLE-222014, Infineon Technologies) was used for data processing and wireless communication. The fully integrated wearable device was attached to the mice or rats using a 3M double sided tape and fixed with Mastisol liquid adhesive to enable strong adhesion, allowing the animals to move freely over a prolonged period. ## Characterization of adhesion of wearable patch The wearable patch was attached to chicken skin (2 cm by 2 cm) using Mastisol liquid adhesive and 3M double sided tape as described previously. A standard T-peel test was then performed according to American Society for Testing and Materials D1876 using a mechanical tester to evaluate patch adhesion to skin. Tegaderm adhesive (3M) was used as control. ## In vivo biodegradation and biocompatibility To assess biodegradation and biocompatibility of the wearable patch, a rat subcutaneous implantation model was used. After anesthesia and analgesia using $2.5\%$ (v/v) isoflurane, buprenorphine (1 mg kg−1), ketoprofen (5 mg kg−1), and bupivacaine (1 mg kg−1), 10-mm incisions in dorsal skin were created to form subcutaneous pockets on the back of Wistar rats (200 to 250 g; Charles River Laboratories, Wilmington, MA, USA). Next, samples were implanted into each pocket according to the protocol approved by the Institutional Animal Care and Use Committee (protocol no. IA20-1800) at California Institute of Technology. Animals were then euthanized after 14 and 56 days, and the samples were explanted with their surrounding tissues for further analysis. ## Multiplexed wound biomarker monitoring in vivo The on-body multiplex wound biomarker monitoring was performed using a diabetic wound model in db/db mice (BKS.Cg-Dock7m +/+ Leprdb/J mice, The Jackson Laboratory, Bar Harbor, ME, USA). After anesthesia and analgesia, a 10-mm full-thickness wounds (through to the level of the panniculus carnosus muscle) was created on the dorsum of mice using a surgical blade. A silicon 12-mm-diameter splint (Grace Bio-Labs, Bend, OR, USA) was placed on the wound area, secured with cyanoacrylate glue, and then fixed using Ethilon 5-0 sterile sutures (Nylon). The wearable patch was then placed on the wound and secured on the wound area using Tegaderm transparent film dressing (3M). The data from the wearable patch were wirelessly recorded. In the case of the infected wound, a mixture of bacteria solution (50-μl solution, 106 CFU ml−1 MRSA, and 106 CFU ml−1 P. aeruginosa) was applied into the wound area on day 4 after surgery. For the fasting experiments, the animals were fasted for 24 hours (only water was provided). One group of fasting animals were used for injection study. In this case, a 400 mM glucose solution in DBPS (based on body weight) was administered into the mouse tail vein to spark ~10 mM increase in blood glucose level. The in vivo sensor readings from the wearable patch were obtained from 30 min before injection and continued until 270 min after injection. For the fasting/feeding experiment, the animals were fasted for 24 hours, followed by feeding with protein rich laboratory rodent diet 5001 (LabDiet). For the food feeding study, the wearable patch was tested before fasting, after 24-hour fasting, and 6 hours of fasting/feeding. ## Spatial and temporal monitoring of critical-sized wounds Similar to multiplexed wound biomarker monitoring, critical-sized wounds (35 mm in diameter) were created in ZDF obese fa/fa diabetic rats (The Jackson Laboratory, Bar Harbor, ME, USA). Next, the sensor array patch was applied on the wound and secured by using 3M Tegaderm dressing. Simultaneous sensor readings were recorded daily for both infected and noninfected wounds before and after treatment. For the infection, a similar mixture of bacteria solution (100 μl solution, 106 CFU ml−1 MRSA, and 106 CFU ml−1 P. aeruginosa) was applied into the wound area on day 2 after surgery. During the in vivo trial, the data from the wearable patch were wirelessly recorded. ## Evaluation of wearable patch-facilitated chronic wound healing in vivo A 10-mm full-thickness wound was created in the ZDF obese fa/fa rat’s dorsal skin, and the wearable patch was placed on the wound. Four different rat groups were tested with different treatments offered by the wearable patch: negative control, drug release, electrical stimulation, and combination therapy. The animals were euthanized, and the tissue samples were explanted on days 4 and 14 after surgery and processed for further analysis. The adhesion of patches on animals during the course of study was monitored. ## In vivo antimicrobial, histological, and immunohistofluorescent evaluations For in vivo biocompatibility assessment, upon explantation, samples were fixed in $4\%$ paraformaldehyde under 4°C overnight, washed thoroughly with DPBS (5×), and then incubated in $30\%$ sucrose overnight (4°C). The samples were then mounted in optimal cutting temperature compound (Thermo Fisher Scientific) followed by flash freezing in liquid nitrogen (N2) and cryosectioning (10-μm sections). Hematoxylin and eosin and immunohistochemistry (IHC) staining were performed on cryosections. For IHC staining, two primary antibodies [anti-CD3 [SP7] (ab16669) and anti-CD68 (ab31630), Abcam] and two secondary antibodies [donkey anti-mouse, Alexa Fluor 568– and goat anti-rabbit, Alexa Fluor 488–conjugated antibodies; Invitrogen] were used. Upon antibody staining, the samples were counterstained against 4′,6-diamidino-2-phenylindole for cell nuclei visualization. The stained slides were then mounted with ProLong Diamond Antifade Mountant (Invitrogen) and imaged using an LSM 800 confocal laser scanning microscope (ZEISS). For regeneration studies, the bacteria samples were first isolated from the wound bed and assessed via CFU assay as described earlier. The wound samples were then explanted with the adjacent tissue, processed, sectioned, and stained via MTC staining and IHC. For IHC staining, different primary antibodies including recombinant anti-cytokeratin 5 antibody [SP27] (ab64081, Abcam), anti–NF-κB p65 (phospho S276) antibody (ab194726, Abcam), human/mouse/rat PTEN Alexa Fluor 647–conjugated antibody (IC847R, R&D Systems), and cytokeratin 14 monoclonal antibody (LL002, Thermo Fisher Scientific) and similar secondary antibodies were used. Upon staining, the samples were mounted with antifade mountant and visualized with a confocal microscope. ## qRT-PCR analysis RNA was isolated from wound tissue samples using the RNeasy Plus Micro Kit (QIAGEN). The RNA quantity and quality were assessed using a NanoDrop $\frac{2000}{2000}$c spectrophotometer at $\frac{260}{280}$ nm wavelengths. Next, the complementary DNA (cDNA) was synthesized using the QuantiTect Reverse Transcription Kit (QIAGEN). Gene expression was performed using a TaqMan Universal PCR Master Mix (Thermo Fisher Scientific). TaqMan Array Plates for rat wound healing gene expression were used where a library of genes was screened. The cDNAs synthesized in the previous step were then added to each plate and followed by quantitative analysis using a QuantStudio 3 Real-Time PCR system (Applied Biosystems). ## Unknown Funding: This work was supported by National Institutes of Health grant R01HL155815 (W.G.), National Institutes of Health grant R21DK13266 (W.G.), National Science Foundation grant 2145802 (W.G.), Office of Naval Research grant N00014-21-1-2483 (W.G.), Office of Naval Research grant N00014-21-1-2845 (W.G.), Heritage Medical Research Institute at the California Institute of Technology (Caltech) (W.G.), Donna and Benjamin M. Rosen Bioengineering Center at Caltech (W.G.), Rothenberg Innovation Initiative at Caltech (W.G.), Sloan Research Fellowship. Author contributions: Conceptualization: W.G. and E.S.S. Supervision: W.G. Methodology: W.G., E.S.S., and C.X. Investigation: E.S.S., C.X., C.W., Y.S., J.M., J.T., S.A.S., J.L., J.L.B., and D.G.A. Funding acquisition: W.G. Writing—original draft: W.G., E.S.S., and C.X. Writing—review and editing: C.W., Y.S., J.M., J.T., S.A.S., J.L., J.L.B., and D.G.A. Competing interests: E.S.S. and W.G. are inventors of a patent application through California Institute of Technology describing the system and technology under the U.S. provisional patent application no. $\frac{17}{993}$,743 (filed on 23 November 2022). The authors declare that they have no other competing interests. Data and materials availability: *All data* needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. ## References 1. Gurtner G. C., Werner S., Barrandon Y., Longaker M. T.. **Wound repair and regeneration**. *Nature* (2008.0) **453** 314-321. PMID: 18480812 2. Eming S. A., Martin P., Tomic-Canic M.. **Wound repair and regeneration: Mechanisms, signaling, and translation**. *Sci. Transl. Med.* (2014.0) **6** 265sr6. PMID: 25473038 3. Sun B. K., Siprashvili Z., Khavari P. A.. **Advances in skin grafting and treatment of cutaneous wounds**. *Science* (2014.0) **346** 941-945. PMID: 25414301 4. Fife C. E., Carter M. J.. **Wound care outcomes and associated cost among patients treated in US outpatient wound centers: Data from the US wound registry**. *Wounds* (2012.0) **24** 10-17. PMID: 25875947 5. Armstrong D. G., Boulton A. J. M., Bus S. A.. **Diabetic foot ulcers and their recurrence**. *N. Engl. J. Med.* (2017.0) **376** 2367-2375. PMID: 28614678 6. Frykberg R. G., Banks J.. **Challenges in the treatment of chronic wounds**. *Adv. Wound Care* (2015.0) **4** 560-582 7. Eriksson E., Liu P. Y., Schultz G. S., Martins-Green M. M., Tanaka R., Weir D., Gould L. J., Armstrong D. G., Gibbons G. W., Wolcott R., Olutoye O. O., Kirsner R. S., Gurtner G. C.. **Chronic wounds: Treatment consensus**. *Wound Repair Regen.* (2022.0) **30** 156-171. PMID: 35130362 8. Blair J. M. A., Webber M. A., Baylay A. J., Ogbolu D. O., Piddock L. J. V.. **Molecular mechanisms of antibiotic resistance**. *Nat. Rev. Microbiol.* (2015.0) **13** 42-51. PMID: 25435309 9. Thakral G., LaFontaine J., Najafi B., Talal T. K., Kim P., Lavery L. A.. **Electrical stimulation to accelerate wound healing**. *Diabet. Foot Ankle* (2013.0) **4** 22081 10. Kloth L. C.. **Electrical stimulation technologies for wound healing**. *Adv. Wound Care* (2014.0) **3** 81-90 11. Patel S., Maheshwari A., Chandra A.. **Biomarkers for wound healing and their evaluation**. *J. Wound Care* (2016.0) **25** 46-55. PMID: 26762498 12. Lindley L. E., Stojadinovic O., Pastar I., Tomic-Canic M.. **Biology and biomarkers for wound healing**. *Plast. Reconstr. Surg.* (2016.0) **138** 18S-28S. PMID: 27556760 13. Cutting K. F.. **Wound exudate: Composition and functions**. *Br. J. Community Nurs.* (2003.0) **8** S4-S9. PMID: 12819589 14. Nery R. A., Kahlow B. S., Skare T. L., Tabushi F. I., Castro A. d. A. e.. **Uric acid and tissue repair**. *Arq. Bras. Cir. Dig.* (2015.0) **28** 290-292. PMID: 26734804 15. Britland S., Ross-Smith O., Jamil H., Smith A. G., Vowden K., Vowden P.. **The lactate conundrum in wound healing: Clinical and experimental findings indicate the requirement for a rapid point-of-care diagnostic**. *Biotechnol. Prog.* (2012.0) **28** 917-924. PMID: 22581665 16. Hirsch T., Spielmann M., Zuhaili B., Koehler T., Fossum M., Steinau H.-U., Yao F., Steinstraesser L., Onderdonk A. B., Eriksson E.. **Enhanced susceptibility to infections in a diabetic wound healing model**. *BMC Surg.* (2008.0) **8** 5. PMID: 18312623 17. Xu C., Yang Y., Gao W.. **Skin-interfaced sensors in digital medicine: From materials to applications**. *Matter* (2020.0) **2** 1414-1445. PMID: 32510052 18. Ray T. R., Choi J., Bandodkar A. J., Krishnan S., Gutruf P., Tian L., Ghaffari R., Rogers J. A.. **Bio-integrated wearable systems: A comprehensive review**. *Chem. Rev.* (2019.0) **119** 5461-5533. PMID: 30689360 19. Kim J., Campbell A. S., de Ávila B. E.-F., Wang J.. **Wearable biosensors for healthcare monitoring**. *Nat. Biotechnol.* (2019.0) **37** 389-406. PMID: 30804534 20. Yang Y., Gao W.. **Wearable and flexible electronics for continuous molecular monitoring**. *Chem. Soc. Rev.* (2019.0) **48** 1465-1491. PMID: 29611861 21. Someya T., Bao Z., Malliaras G. G.. **The rise of plastic bioelectronics**. *Nature* (2016.0) **540** 379-385. PMID: 27974769 22. Zhou Z., Chen K., Li X., Zhang S., Wu Y., Zhou Y., Meng K., Sun C., He Q., Fan W., Fan E., Lin Z., Tan X., Deng W., Yang J., Chen J.. **Sign-to-speech translation using machine-learning-assisted stretchable sensor arrays**. *Nat. Electron.* (2020.0) **3** 571-578 23. Zhou Y., Zhao X., Xu J., Fang Y., Chen G., Song Y., Li S., Chen J.. **Giant magnetoelastic effect in soft systems for bioelectronics**. *Nat. Mater.* (2021.0) **20** 1670-1676. PMID: 34594013 24. Gao W., Emaminejad S., Nyein H. Y. Y., Challa S., Chen K., Peck A., Fahad H. M., Ota H., Shiraki H., Kiriya D., Lien D.-H., Brooks G. A., Davis R. W., Javey A.. **Fully integrated wearable sensor arrays for multiplexed in situ perspiration analysis**. *Nature* (2016.0) **529** 509-514. PMID: 26819044 25. Ray T. R., Ivanovic M., Curtis P. M., Franklin D., Guventurk K., Jeang W. J., Chafetz J., Gaertner H., Young G., Rebollo S., Model J. B., Lee S. P., Ciraldo J., Reeder J. T., Hourlier-Fargette A., Bandodkar A. J., Choi J., Aranyosi A. J., Ghaffari R., McColley S. A., Haymond S., Rogers J. A.. **Soft, skin-interfaced sweat stickers for cystic fibrosis diagnosis and management**. *Sci. Transl. Med.* (2021.0) **13** eabd8109. PMID: 33790027 26. Lee H., Choi T. K., Lee Y. B., Cho H. R., Ghaffari R., Wang L., Choi H. J., Chung T. D., Lu N., Hyeon T., Choi S. H., Kim D.-H.. **A graphene-based electrochemical device with thermoresponsive microneedles for diabetes monitoring and therapy**. *Nat. Nanotechnol.* (2016.0) **11** 566-572. PMID: 26999482 27. Koh A., Kang D., Xue Y., Lee S., Pielak R. M., Kim J., Hwang T., Min S., Banks A., Bastien P., Manco M. C., Wang L., Ammann K. R., Jang K.-I., Won P., Han S., Ghaffari R., Paik U., Slepian M. J., Balooch G., Huang Y., Rogers J. A.. **A soft, wearable microfluidic device for the capture, storage, and colorimetric sensing of sweat**. *Sci. Transl. Med.* (2016.0) **8** 366ra165 28. Tehrani F., Teymourian H., Wuerstle B., Kavner J., Patel R., Furmidge A., Aghavali R., Hosseini-Toudeshki H., Brown C., Zhang F., Mahato K., Li Z., Barfidokht A., Yin L., Warren P., Huang N., Patel Z., Mercier P. P., Wang J.. **An integrated wearable microneedle array for the continuous monitoring of multiple biomarkers in interstitial fluid**. *Nat. Biomed. Eng.* (2022.0) **6** 1214-1224. PMID: 35534575 29. Sempionatto J. R., Lin M., Yin L., De la paz E., Pei K., Sonsa-ard T., de Loyola Silva A. N., Khorshed A. A., Zhang F., Tostado N., Xu S., Wang J.. **An epidermal patch for the simultaneous monitoring of haemodynamic and metabolic biomarkers**. *Nat. Biomed. Eng.* (2021.0) **5** 737-748. PMID: 33589782 30. Wang C., Li X., Hu H., Zhang L., Huang Z., Lin M., Zhang Z., Yin Z., Huang B., Gong H., Bhaskaran S., Gu Y., Makihata M., Guo Y., Lei Y., Chen Y., Wang C., Li Y., Zhang T., Chen Z., Pisano A. P., Zhang L., Zhou Q., Xu S.. **Monitoring of the central blood pressure waveform via a conformal ultrasonic device**. *Nat. Biomed. Eng.* (2018.0) **2** 687-695. PMID: 30906648 31. Wang C., Shirzaei Sani E., Gao W.. **Wearable bioelectronics for chronic wound management**. *Adv. Funct. Mater.* (2022.0) **32** 2111022. PMID: 36186921 32. 32A. A. Trotsyuk, S. Niu, D. Henn, K. Chen, C.-C. Shih, M. R. Larson, A. M. Mermin-Bunnell, S. Mittal, J.-C. Lai, A. Saberi, E. Beard, S. Jing, D. Zhong, S. R. Steele, K. Sun, T. Jain, E. Zhao, C. R. Neimeth, W. G. Viana, J. Tang, D.Sivaraj, J. Padmanabhan, M. Rodrigues, D. P. Perrault, A. Chattopadhyay, Z. N. Maan, M. C. Leeolou, C. A. Bonham, S. H. Kwon, H. C. Kussie, K. S. Fischer, G. Gurusankar, K. Liang, K. Zhang, R. Nag, M. P. Snyder, M. Januszyk, G. C. Gurtner, Z. Bao, Wireless closed-loop smart bandage for chronic wound management and accelerated tissue regeneration. bioRxiv 2022.01.16.476432 (2022). 33. Dong R., Guo B.. **Smart wound dressings for wound healing**. *Nano Today* (2021.0) **41** 101290 34. Xiong Z., Achavananthadith S., Lian S., Madden L. E., Ong Z. X., Chua W., Kalidasan V., Li Z., Liu Z., Singh P., Yang H., Heussler S. P., Kalaiselvi S. M. P., Breese M. B. H., Yao H., Gao Y., Sanmugam K., Tee B. C. K., Chen P.-Y., Loke W., Lim C. T., Chiang G. S. H., Tan B. Y., Li H., Becker D. L., Ho J. S.. **A wireless and battery-free wound infection sensor based on DNA hydrogel**. *Sci. Adv.* (2021.0) **7** eabj1617. PMID: 34797719 35. Kalidasan V., Yang X., Xiong Z., Li R. R., Yao H., Godaba H., Obuobi S., Singh P., Guan X., Tian X., Kurt S. A., Li Z., Mukherjee D., Rajarethinam R., Chong C. S., Wang J.-W., Ee P. L. R., Loke W., Tee B. C. K., Ouyang J., Charles C. J., Ho J. S.. **Wirelessly operated bioelectronic sutures for the monitoring of deep surgical wounds**. *Nat. Biomed. Eng.* (2021.0) **5** 1217-1227. PMID: 34654900 36. Gao Y., Nguyen D. T., Yeo T., Lim S. B., Tan W. X., Madden L. E., Jin L., Long J. Y. K., Aloweni F. A. B., Liew Y. J. A., Tan M. L. L., Ang S. Y., Maniya S. D., Abdelwahab I., Loh K. P., Chen C.-H., Becker D. L., Leavesley D., Ho J. S., Lim C. T.. **A flexible multiplexed immunosensor for point-of-care in situ wound monitoring**. *Sci. Adv.* (2021.0) **7** eabg9614. PMID: 34020961 37. Kassal P., Kim J., Kumar R., de Araujo W. R., Steinberg I. M., Steinberg M. D., Wang J.. **Smart bandage with wireless connectivity for uric acid biosensing as an indicator of wound status**. *Electrochem. Commun.* (2015.0) **56** 6-10 38. Puthia M., Butrym M., Petrlova J., Strömdahl A.-C., Andersson M. Å., Kjellström S., Schmidtchen A.. **A dual-action peptide-containing hydrogel targets wound infection and inflammation**. *Sci. Transl. Med.* (2020.0) **12** eaax6601. PMID: 31894104 39. Tanaka T., Nishio I., Sun S.-T., Ueno-Nishio S.. **Collapse of gels in an electric field**. *Science* (1982.0) **218** 467-469. PMID: 17808541 40. Proctor C. M., Slézia A., Kaszas A., Ghestem A., del Agua I., Pappa A.-M., Bernard C., Williamson A., Malliaras G. G.. **Electrophoretic drug delivery for seizure control**. *Sci. Adv.* (2018.0) **4** eaau1291. PMID: 30167463 41. Song B., Gu Y., Pu J., Reid B., Zhao Z., Zhao M.. **Application of direct current electric fields to cells and tissues in vitro and modulation of wound electric field in vivo**. *Nat. Protoc.* (2007.0) **2** 1479-1489. PMID: 17545984 42. Zhao M., Song B., Pu J., Wada T., Reid B., Tai G., Wang F., Guo A., Walczysko P., Gu Y., Sasaki T., Suzuki A., Forrester J. V., Bourne H. R., Devreotes P. N., McCaig C. D., Penninger J. M.. **Electrical signals control wound healing through phosphatidylinositol-3-OH kinase-γ and PTEN**. *Nature* (2006.0) **442** 457-460. PMID: 16871217 43. Cohen D. J., James Nelson W., Maharbiz M. M.. **Galvanotactic control of collective cell migration in epithelial monolayers**. *Nat. Mater.* (2014.0) **13** 409-417. PMID: 24608142 44. Chanmugam A., Langemo D., Thomason K., Haan J., Altenburger E. A., Tippett A., Henderson L., Zortman T. A.. **Relative temperature maximum in wound infection and inflammation as compared with a control subject using long-wave infrared thermography**. *Adv. Skin Wound Care* (2017.0) **30** 406-414. PMID: 28817451 45. Fernandez M. L., Upton Z., Edwards H., Finlayson K., Shooter G. K.. **Elevated uric acid correlates with wound severity**. *Int. Wound J.* (2012.0) **9** 139-149. PMID: 21973196 46. Tegl G., Schiffer D., Sigl E., Heinzle A., Guebitz G. M.. **Biomarkers for infection: Enzymes, microbes, and metabolites**. *Appl. Microbiol. Biotechnol.* (2015.0) **99** 4595-4614. PMID: 25952112 47. Digirolamo M., Newby F. D., Lovejoy J.. **Lactate production in adipose tissue; a regulated function with extra-adipose implications**. *FASEB J.* (1992.0) **6** 2405-2412. PMID: 1563593 48. Sakakibara I., Fujino T., Ishii M., Tanaka T., Shimosawa T., Miura S., Zhang W., Tokutake Y., Yamamoto J., Awano M., Iwasaki S., Motoike T., Okamura M., Inagaki T., Kita K., Ezaki O., Naito M., Kuwaki T., Chohnan S., Yamamoto T. T., Hammer R. E., Kodama T., Yanagisawa M., Sakai J.. **Fasting-induced hypothermia and reduced energy production in mice lacking acetyl-CoA synthetase 2**. *Cell Metab.* (2009.0) **9** 191-202. PMID: 19187775 49. Griffin D. R., Archang M. M., Kuan C.-H., Weaver W. M., Weinstein J. S., Feng A. C., Ruccia A., Sideris E., Ragkousis V., Koh J., Plikus M. V., Di Carlo D., Segura T., Scumpia P. O.. **Activating an adaptive immune response from a hydrogel scaffold imparts regenerative wound healing**. *Nat. Mater.* (2021.0) **20** 560-569. PMID: 33168979 50. Park Y. R., Sultan M. T., Park H. J., Lee J. M., Ju H. W., Lee O. J., Lee D. J., Kaplan D. L., Park C. H.. **NF-κB signaling is key in the wound healing processes of silk fibroin**. *Acta Biomater.* (2018.0) **67** 183-195. PMID: 29242162 51. Leask A.. **Potential therapeutic targets for cardiac Fibrosis**. *Circ. Res.* (2010.0) **106** 1675-1680. PMID: 20538689 52. Profyris C., Tziotzios C., Do Vale I.. **Cutaneous scarring: Pathophysiology, molecular mechanisms, and scar reduction therapeutics Part I. The molecular basis of scar formation**. *J. Am. Acad. Dermatol.* (2012.0) **66** 1-10. PMID: 22177631 53. Gerling I., Nejman C., Chatterjee N. K.. **Effect of coxsackievirus B4 infection in mice on expression of 64,000-Mr autoantigen and glucose sensitivity of islets before development of hyperglycemia**. *Diabetes* (1988.0) **37** 1419-1425. PMID: 2843410 54. Simone T. M., Higgins C. E., Czekay R.-P., Law B. K., Higgins S. P., Archambeault J., Kutz S. M., Higgins P. J.. **SERPINE1: A molecular switch in the proliferation-migration dichotomy in wound-"activated" keratinocytes**. *Adv. Wound Care* (2014.0) **3** 281-290 55. Yu H., Jove R.. **The STATs of cancer--new molecular targets come of age**. *Nat. Rev. Cancer* (2004.0) **4** 97-105. PMID: 14964307 56. Wang M., Yang Y., Min J., Song Y., Tu J., Mukasa D., Ye C., Xu C., Heflin N., McCune J. S., Hsiai T. K., Li Z., Gao W.. **A wearable electrochemical biosensor for the monitoring of metabolites and nutrients**. *Nat. Biomed. Eng.* (2022.0) **6** 1225-1235. PMID: 35970928 57. Yang Y., Song Y., Bo X., Min J., Pak O. S., Zhu L., Wang M., Tu J., Kogan A., Zhang H., Hsiai T. K., Li Z., Gao W.. **A laser-engraved wearable sensor for sensitive detection of uric acid and tyrosine in sweat**. *Nat. Biotechnol.* (2020.0) **38** 217-224. PMID: 31768044 58. Bandodkar A. J., Gutruf P., Choi J., Lee K., Sekine Y., Reeder J. T., Jeang W. J., Aranyosi A. J., Lee S. P., Model J. B., Ghaffari R., Su C. J., Leshock J. P., Ray T., Verrillo A., Thomas K., Krishnamurthi V., Han S., Kim J., Krishnan S., Hang T., Rogers J. A.. **Battery-free, skin-interfaced microfluidic/electronic systems for simultaneous electrochemical, colorimetric, and volumetric analysis of sweat**. *Sci. Adv.* (2019.0) **5** eaav3294. PMID: 30746477 59. Quinn C. A. P., Connor R. E., Heller A.. **Biocompatible, glucose-permeable hydrogel for in situ coating of implantable biosensors**. *Biomaterials* (1997.0) **18** 1665-1670. PMID: 9613815 60. Yu Y., Nassar J., Xu C., Min J., Yang Y., Dai A., Doshi R., Huang A., Song Y., Gehlhar R., Ames A. D., Gao W.. **Biofuel-powered soft electronic skin with multiplexed and wireless sensing for human-machine interfaces**. *Sci. Robot.* (2020.0) **5** eaaz7946. PMID: 32607455 61. Song Y., Min J., Yu Y., Wang H., Yang Y., Zhang H., Gao W.. **Wireless battery-free wearable sweat sensor powered by human motion**. *Sci. Adv.* (2020.0) **6** eaay9842. PMID: 32998888 62. Park S., Heo S. W., Lee W., Inoue D., Jiang Z., Yu K., Jinno H., Hashizume D., Sekino M., Yokota T., Fukuda K., Tajima K., Someya T.. **Self-powered ultra-flexible electronics via nano-grating-patterned organic photovoltaics**. *Nature* (2018.0) **561** 516-521. PMID: 30258137 63. Chen J., Huang Y., Zhang N., Zou H., Liu R., Tao C., Fan X., Wang Z. L.. **Micro-cable structured textile for simultaneously harvesting solar and mechanical energy**. *Nat. Energy* (2016.0) **1** 16138 64. Veidal S. S., Vassiliadis E., Barascuk N., Zhang C., Segovia-Silvestre T., Klickstein L., Larsen M. R., Qvist P., Christiansen C., Vainer B., Karsdal M. A.. **Matrix metalloproteinase-9-mediated type III collagen degradation as a novel serological biochemical marker for liver fibrogenesis**. *Liver Int.* (2010.0) **30** 1293-1304. PMID: 20666994 65. Kessenbrock K., Plaks V., Werb Z.. **Matrix metalloproteinases: Regulators of the tumor microenvironment**. *Cell* (2010.0) **141** 52-67. PMID: 20371345 66. Elbialy Z. I., Assar D. H., Abdelnaby A., Asa S. A., Abdelhiee E. Y., Ibrahim S. S., Abdel-Daim M. M., Almeer R., Atiba A.. **Healing potential of Spirulina platensis for skin wounds by modulating bFGF, VEGF, TGF-ß1 and α-SMA genes expression targeting angiogenesis and scar tissue formation in the rat model**. *Biomed. Pharmacother.* (2021.0) **137** 111349. PMID: 33567349 67. Presta M., Andrés G., Leali D., Dell’Era P., Ronca R.. **Inflammatory cells and chemokines sustain FGF2-induced angiogenesis**. *Eur. Cytokine Netw.* (2009.0) **20** 39-50. PMID: 19541589 68. Wang D., Sai J., Richmond A.. **Cell surface heparan sulfate participates in CXCL1-induced signaling**. *Biochemistry* (2003.0) **42** 1071-1077. PMID: 12549928 69. Bao P., Kodra A., Tomic-Canic M., Golinko M. S., Ehrlich H. P., Brem H.. **The role of vascular endothelial growth factor in wound healing**. *J. Surg. Res.* (2009.0) **153** 347-358. PMID: 19027922
--- title: Polyisoprenylated cysteinyl amide inhibitors deplete singly polyisoprenylated monomeric G-proteins in lung and breast cancer cell lines authors: - Nada Tawfeeq - Jassy Mary S. Lazarte - Yonghao Jin - Matthew D. Gregory - Nazarius S. Lamango journal: Oncotarget year: 2023 pmcid: PMC10038354 doi: 10.18632/oncotarget.28390 license: CC BY 3.0 --- # Polyisoprenylated cysteinyl amide inhibitors deplete singly polyisoprenylated monomeric G-proteins in lung and breast cancer cell lines ## Abstract Finding effective therapies against cancers driven by mutant and/or overexpressed hyperactive G-proteins remains an area of active research. Polyisoprenylated cysteinyl amide inhibitors (PCAIs) are agents that mimic the essential posttranslational modifications of G-proteins. It is hypothesized that PCAIs work as anticancer agents by disrupting polyisoprenylation-dependent functional interactions of the G-Proteins. This study tested this hypothesis by determining the effect of the PCAIs on the levels of RAS and related monomeric G-proteins. Following 48 h exposure, we found significant decreases in the levels of KRAS, RHOA, RAC1, and CDC42 ranging within 20–$66\%$ after NSL-YHJ-2-27 (5 μM) treatment in all four cell lines tested, A549, NCI-H1299, MDA-MB-231, and MDA-MB-468. However, no significant difference was observed on the G-protein, RAB5A. Interestingly, 38 and $44\%$ decreases in the levels of the farnesylated and acylated NRAS were observed in the two breast cancer cell lines, MDA-MB-231, and MDA-MB-468, respectively, while HRAS levels showed a $36\%$ decrease only in MDA-MB-468 cells. Moreover, after PCAIs treatment, migration, and invasion of A549 cells were inhibited by 72 and $70\%$, respectively while the levels of vinculin and fascin dropped by 33 and $43\%$, respectively. These findings implicate the potential role of PCAIs as anticancer agents through their direct interaction with monomeric G-proteins. ## INTRODUCTION Small G-proteins, monomeric GTPases, or the RAS (Rat sarcoma) superfamily are a large family of small guanine nucleotide-binding proteins with molecular weights ranging from 20 to 30 kDa [1, 2]. These proteins share a core structure, the conserved G-box (GDP/GTP) binding domain, of approximately 170 residues [3]. Small GTPases function as binary molecular switches, transmitting extracellular signals to an intracellular environment [4]. The superfamily is classified into five subfamilies based on the cellular processes that they regulate [2]. These include the founding member of the superfamily RAS [5], RHO (Ras homology) [6], ARF/SAR (adenosine diphosphate (ADP) ribosylation factor) [7], the largest subfamily RAB (RAS-like in brain or RAS-related in brain) [8] and RAN (RAS-like nuclear or RAS-related nuclear protein) [9]. The small GTPase families regulate a wide range of processes in the cell; however, each family performs different functions within the cell due to differences in their structures, post-translational modifications (PTMs), and subcellular localization (Table 1) [10]. **Table 1** | Subfamily | Cellular functions | G-protein | Genetic alteration in cancer | Major cancer type | Incidence rates (%) | References | | --- | --- | --- | --- | --- | --- | --- | | RAS | Mediate and activate some major pathways that control cell proliferation, survival, and cell cycle progression [4]. | KRAS4b | Mutation | Pancreatic cancer | 90 | [28–31] | | RAS | Mediate and activate some major pathways that control cell proliferation, survival, and cell cycle progression [4]. | KRAS4b | Mutation | Non-small cell lung cancer | 30–35 | [28–31] | | RAS | Mediate and activate some major pathways that control cell proliferation, survival, and cell cycle progression [4]. | KRAS4a | Mutation | Colorectal Colon cancer | 30–45 | [28–31] | | RAS | Mediate and activate some major pathways that control cell proliferation, survival, and cell cycle progression [4]. | HRAS | Mutation | Bladder urothelial | 57 | [32] | | RAS | Mediate and activate some major pathways that control cell proliferation, survival, and cell cycle progression [4]. | NRAS | Mutation | Melanoma | 94 | [33, 34] | | RAS | Mediate and activate some major pathways that control cell proliferation, survival, and cell cycle progression [4]. | | | Leukemia | 59 | [33, 34] | | RHO | Regulate vesicle transport and assembly and disassembly of actin cytoskeleton required for cell migration and invasion [35]. | RHOA | Overexpression | Colon | 95 | [35] | | RHO | Regulate vesicle transport and assembly and disassembly of actin cytoskeleton required for cell migration and invasion [35]. | RHOA | Overexpression | Lung | 95 | [35] | | RHO | Regulate vesicle transport and assembly and disassembly of actin cytoskeleton required for cell migration and invasion [35]. | RAC1 | Mutation | Breast | 50 | [14, 33, 36–39] | | RHO | Regulate vesicle transport and assembly and disassembly of actin cytoskeleton required for cell migration and invasion [35]. | RAC1 | Overexpression | Breast | 70 | [14, 33, 36–39] | | RHO | Regulate vesicle transport and assembly and disassembly of actin cytoskeleton required for cell migration and invasion [35]. | RAC1 | Overexpression | Lung | 50 | [14, 33, 36–39] | | RHO | Regulate vesicle transport and assembly and disassembly of actin cytoskeleton required for cell migration and invasion [35]. | CDC42 | Overexpression | Breast | 95 | [14, 33, 40] | | RHO | Regulate vesicle transport and assembly and disassembly of actin cytoskeleton required for cell migration and invasion [35]. | CDC42 | Overexpression | Colorectal | 60 | [14, 33, 40] | | ARF | Regulate different steps in intracellular membrane transport [7]. | ARF1-ARF6 | Unknown | – | | [7] | | RAB | Regulate vesicular membrane trafficking events such as early endosomal membrane tracking, fusion, and sorting [41]. | RAB5 | Unknown | – | | [8, 41, 42] | | RAN | Facilitates transport into and out of the nucleus [9]. | RAN | Unknown | – | | [9] | Over the last decades, several small GTPases were found to be involved in the development of human carcinomas; hence they have become an interesting subject in cancer research [11, 12]. RAS superfamily of G-proteins genes is the most frequently mutated in cancers accounting for up to $30\%$ of human tumors [13], with KRAS being the most rampantly mutated, accounting for up to $86\%$ of RAS mutations in cancer [13]. Other monomeric G-proteins such as RHOA, CDC42, and RAC1 contribute to cancer progression and are found to be mutated or overexpressed in some tumors [11, 14–17]. Summarized in Table 1 are the incidence rates of some of the aberrant G-proteins in various cancers. Moreover, these proteins are involved in the organization and assembly of the F-actin cytoskeleton and thus regulate the cellular processes of migration and invasion that are commonly dysregulated during cancer development and progression to drive metastasis [18–21]. For most G-proteins, their proper functioning is strongly dependent on post-translational modifications (PTMs) [2, 22]. They differ in PTMs, which are characterized by their C-terminal sequence motifs known as hypervariable region (HVR) [23]. The RAS, RHO, and RAB family members are generally C-terminally modified by polyisoprenylation, palmitoylation, or myristoylation [24–26]. However, the ARF/SAR family is mostly modified at their N-terminus by myristoylation, while the Ran family is not modified at all (Table 2) (Figure 1) [2, 27]. The PTMs play numerous roles, such as allowing proper folding, membrane binding and localization, protein-protein interactions, and signaling [46, 47]. There are various protein-protein interactions that involve the polyisoprenyl moiety. For example, KRAS trafficking between various subcellular compartments is facilitated by interactions with several chaperone proteins such as phosphodiesterase δ (PDE-δ), galectins, calmodulin, tubulin, and prenylated RAB-acceptor protein 1 (PRA1) [48]. It has been reported that RAS function can be inhibited using S-farnesyl derivatives of rigid carboxylic acids such as S-trans,trans farnesyl thiosalicylic acid (FTS) [49, 50]. The mechanism of action of FTS involves the displacement of RAS from the plasma membrane, thereby accelerating its degradation [51]. It was suspected that the polyisoprenylated cysteinyl amide inhibitors (PCAIs) may function by displacing polyisoprenylated G-proteins from their associations with other macromolecules and cellular structures as membranes. To begin understanding the roles that the PCAIs may play in G-protein function, we determined the effect of PCAIs on two types of monomeric G-proteins, those that are modified with a single polyisoprenyl group (KRAS, RHOA, CDC42 and RAC1) and those that are either doubly polyisoprenylated (RAB5A) or polyisoprenylated and acylated (HRAS and NRAS). Moreover, we also report on the effects of PCAIs on cell migration and invasion, which is controlled by some of these monomeric G-proteins, as well as the PCAIs effect on F-actin cross-linking proteins, vinculin and fascin. ## PCAIs suppress the viability of the cancer cell lines in a concentration- and polyisoprenylation-dependent manner A significant feature of NSL-YHJ-2-27 is the presence of farnesyl tail which is also found in most G-proteins. Exposure of MDA-MB-468 cells to NSL-YHJ-2-27 resulted in physical changes in the cells as the concentration of the PCAIs increased. Cell viability results show EC50 values of 4.4 μM for NSL-YHJ-2-27 and >50 μM for NSL-YHJ-2-62 (Figure 2). The non-farnesylated compound, NSL-YHJ-2-62, did not elicit any changes to the cells. As shown in Table 3, the results obtained from MDA-MB-468 cells corroborate the results presented in a previous paper [52]. Prominent cell rounding started to become visible in cells that were treated with 5 μM NSL-YHJ-2-27, while higher concentrations completely killed the cells. Nonetheless, there were no changes observed on cells treated with NSL-YHJ-2-62 at all the concentrations used (Supplementary Figure 1). **Figure 2:** *Concentration-response curves of PCAIs against MDA-MB-468 cells.Cells were treated with varying concentrations of NSL-YHJ-2-27 (potent compound) or NSL-YHJ-2-62 (compound lacking the polyisoprenyl moiety used as control) at the onset and after 24 h. After 48 h, resazurin reduction assay was performed to determine the residual cell viability. The EC50 values were determined by plotting the relative fluorescence intensities (expressed percentages of the control values) against concentration in a non-linear regression curve fit using GraphPad Prism version 8.0 for Windows (San Diego, CA, USA).* TABLE_PLACEHOLDER:Table 3 ## PCAIs deplete singly polyisoprenylated but not doubly polyisoprenylated or polyisoprenylated and acylated G-protein levels To investigate the hypothesized anticancer mechanisms of the PCAIs through disruption of G-protein function, we checked the effects of the PCAIs on the G-protein levels in lung cancer (A549 and NCI-H1299) and breast cancer (MDA-MB-231 and MDA-MB-468) cell lines. When A549 cells were treated with 5 μM of NSL-YHJ-2-27 for 48 h, the KRAS, RHOA, RAC1, and CDC42 protein levels dropped by 46, 45, 57, 66, and $57\%$, respectively, compared to the control. However, no significant difference was observed in the levels of RAB5A, HRAS, and NRAS (Figure 3A). The same was observed in the NCI-H1299 cells, after treatment with 5 μM of NSL-YHJ-2-27 for 48 h revealed decreased levels of KRAS, RHOA, RAC1, and CDC42 of 40, 27, 20, and $21\%$, respectively, but no significant changes in RAB5A, HRAS, and NRAS levels (Figure 3B). Furthermore, both breast cancer cell lines exhibited significant reductions in the levels of KRAS, RHOA, RAC1, and CDC42 proteins of 49, 40, 50, and $48\%$, respectively in MDA-MB-468 cells (Figure 3C). In MDA-MB-231 cells, the levels of the same proteins were reduced by 38, 26, 37, and $36\%$, respectively (Figure 3D). Contrary to what we initially expected, significant decreases in the levels of HRAS and NRAS by 36 and $44\%$, respectively were observed in MDA-MB-468 cells (Figure 3C). On the other hand, MDA-MB-231 cells showed a $38\%$ reduction in NRAS protein levels at 5 μM treatment with NSL-YHJ-2-27 (Figure 3D). Of all the proteins tested, the non-farnesylated analog, NSL-YHJ-2-62, did not elicit any significant effects on the cell lines, confirming that the farnesyl moiety is required for the effects of the PCAIs (Figure 3A–3D). **Figure 3:** *The effect of PCAIs on G-protein levels.Cells were treated for 48 h with 0−5 μM of NSL-YHJ-2-27 (or NSL-YHJ-x-xx where x-xx is 2-27) or 10 μM of its non-farnesylated analog, NSL-YHJ-2-62 (or NSL-YHJ-x-xx where x-xx is 2-62). These were then lysed and subjected to western blot analysis for G-protein levels as described in the methods. (A–D) Western blot images and densitometry plots of bands following quantification using Image Lab Software were normalized against GAPDH or α-Actinin. The samples were analyzed for G-protein levels of expression in (A) A549 (B) NCI-H1299 (C) MDA-MB-468 and (D) MDA-MB-231 cell lines. Data are representative of three independent experiments. Statistical significance (* p < 0.05, ** p < 0.01, and *** p < 0.001) was determined by 1-way ANOVA with post hoc Dunnett’s test.* ## RAC1 and RHOA are secreted out of the cell after PCAIs treatment To understand the mechanism of G-protein depletion following treatment of cells with PCAIs, we set out to determine whether these proteins are degraded and/or secreted from the cell. The results show that RAC1 and RHOA were both secreted into the culture medium, while CDC42 and KRAS were not (Figure 4). **Figure 4:** *PCAIs induce RAC1 and RHOA secretion from cells.MDA-MB-468 cells were treated with either 0 (control) or 5 μM PCAIs. After 48 h, the media were collected and concentrated using vacuum concentrator (Labconco, USA). The concentrated media were subjected to western blot analysis using Jess Simple western assay, the target proteins (~25 kDA) were probed using the respective antibodies targeting RAC1, CDC42, RHOA, or KRAS.* ## PCAIs inhibit cancer cell migration and invasion For cells to metastasize, they need to migrate from a primary tumor through the extracellular matrix and invade distal tissues. To better understand the potential of the PCAIs at inhibiting metastasis, we tested their effect on cancer cells migration and invasion. Treatment with 5 μM of compound NSL-YHJ-2-27 inhibited the number of migrated cells in A549, NCI-H1299, MDA-MB-468, and MDA-MB-231 cell lines by 72, 41, 46, and $68\%$ respectively, after 24 h as compared to controls (Figure 5A). Furthermore, the effect of PCAIs on cellular invasion was determined using the trans-well invasion assays and the number of cells that were able to invade the extracellular matrix (ECM) and make their way to the other side of the membrane were quantified. Treatment of A549 cells with compound NSL-YHJ-2-27 yielded a significant reduction in the number of cells that invaded the ECM as compared to controls. We observed a concentration-dependent decrease in the number of cells that invaded through Matrigel following exposure to NSL-YHJ-2-27. Exposure to 1, 2, and 5 μM of NSL-YHJ-2-27 suppressed the invasion of A549 cells by 40, 44, and $70\%$ respectively, after 24 h compared to controls (Figure 5B). **Figure 5:** *NSL-YHJ-2-27 suppresses cancer cell migration and invasion.(A) Confluent monolayers of cancer cells separated by a “wound” generated using cell culture inserts (ibidi) were treated with the indicated concentrations of NSL-YHJ-2-27 and closure of the wounds was monitored, and images captured at 0 and 24 h after treatment using a Nikon Ti Eclipse microscope at 4X magnification. The number of cells that migrated into the wounds were counted. (B) A549 cells were plated onto the inserts of 24-well Matrigel invasion chambers after treatment and incubated for 24 h as indicated in the Methods. Cells that invaded from the top chamber of inserts through Matrigel were trapped on the membrane in the lower chamber of the inserts. These invading cells were fixed and then stained with 1% crystal violet. Bright field images were obtained using Nikon Eclipse microscope at 4X magnification. The results are the means of three independents experiments. Statistical significance (** p < 0.01, and *** p < 0.001) was determined using 1-way ANOVA with post hoc Dunnett’s tests.* ## PCAIs decrease the levels of vinculin and fascin in A549 cells To understand the effect of PCAIs on cell migration and invasion more, we investigated their effect on the F-actin cross-linking proteins vinculin and fascin that bridge integrins to the actin cytoskeleton [53]. The levels of vinculin protein decreased by $33\%$ and the levels of fascin protein dropped by $43\%$ after exposure of A549 cells to 5 μM of NSL-YHJ-2-27 (Figure 6). **Figure 6:** *PCAIs decrease the levels of vinculin and fascin proteins.Cells were treated for 48 h with 0−5 μM of NSL-YHJ-2-27 or 10 μM of NSL-YHJ-2-62. These were then lysed and subjected to western blot analysis for vinculin and fascin protein levels as described in the Materials and Methods. Densitometry of bands and quantification were performed using Image Lab Software and normalized to GAPDH. Data are representative of three independent experiments. Statistical significance (* p < 0.05 and ** p < 0.01) was determined by 1-way ANOVA with post hoc Dunnett’s test.* ## DISCUSSION The notion that the PCAIs may directly impact G-protein functions was predicted by their structural similarities to the PTMs on those modified with a single polyisoprenyl moiety and the numerous reports indicating that the farnesylation or geranylgeranylation directly contribute to protein-protein interactions [54–57]. One of the examples of prenylation dependent protein–protein interactions is the interaction with chaperone proteins such as galectin 3, 8, 14-3-3, PRA1, and calmodulin (CALM) in subcellular trafficking [54–57]. For example, galectin 8 isoforms have been shown to bind to farnesylated but not to unfarnesylated KRAS [58]. Moreover, it was found that inhibiting PDEδ with small molecules that bind to the farnesyl-binding pocket of PDEδ can impair KRAS localization to the plasma membrane [59, 60]. While exploring the mechanism of action of FTS, that was reported to inhibit RAS function, it was determined that it releases RAS from the membrane, displacing it into the cytosol, indicating that the major site of action of the FTS is at the anchoring point in the membrane [51]. The PTMs pathway is vital for the functions of most small GTPases where it is essential for their membrane binding and localization which is an essential step in their activation [61]. These modifications facilitate the proteins’ association with the inner surfaces of the plasma membranes, where they interact with upstream activators and downstream effectors in various signaling pathways [57]. Small G-proteins differ in PTMs, which are defined by the varying signal sequences in the HVR [23]. Of the RAS isoforms, KRAS is the only one in which a single C-terminal cysteine is modified by polyisoprenylation [24], while NRAS and HRAS are polyisoprenylated and additionally modified by one or two palmitoyl groups [24]. RHOA, RAC1, and CDC42 also have a single C-terminal polyisoprenylated cysteine, while RAB5 undergoes a double GG modification [44, 45] (Table 2) (Figure 1). These PTM differences may help explain the differences between the effects of the PCAIs on these small G-proteins. The significant suppression of the PCAIs on the levels of KRAS, RHOA, RAC1, and CDC42 that are modified only at one site through polyisoprenylation suggests that the PCAIs are more capable of dislodging them from polyisoprenylation-dependent interactions than they would RAB5A that has an additional modified cysteine. Geranylgeranylation results in higher affinities than farnesylation [62, 63]. This implies that PCAIs competitive displacement of a doubly geranylgeranylated RAB5A would be improbable. The effect of PCAIs on NRAS and HRAS varied depending on cell lines. In A549 and NCI-H1299, the levels of NRAS and HRAS were not affected, while in MDA-MB-468 cells, the said G-proteins decreased, and in MDA-MB-231 cells only NRAS showed a reduction in its protein levels. It isn’t clear how the PCAIs would be able to suppress the NRAS and HRAS levels since the additional acyl modifications that contribute to anchoring the proteins to the membranes would make dislodgement more difficult. That differences in the levels of these G-proteins were only observed in some cell lines is an indication that other unique cellular factors that do not involve direct competitive effects at the protein-protein interaction level may be in play. In fact, palmitoylation is readily reversible under physiologic conditions [64]. The binding and dissociation of RAS proteins modified only through farnesylation from membranes have been reported to occur rapidly than those attached through farnesylation and palmitoylation [25, 26]. RAS isomers that are both farnesylated and palmitoylated have more than 100-fold higher affinity for membranes than only farnesylated RAS [26, 65]. In addition to the PTMs, an accumulation of positively charged amino acids in the polybasic region are also essential for membrane attachments and protein-protein interactions [24, 66] (Table 2) (Figure 1). KRAS, RHOA, RAC1, and CDC42 which all showed decreased levels upon treatment with the PCAIs contain adjacent clusters of basic amino acid residues to bind negatively charged phospholipid headgroups in membranes [24]. The tethered positive charges of the ionized piperizinyl moiety in the PCAIs may somewhat mimic the positive charges of the polybasic regions of G-proteins and may play a similar role when the PCAIs uncouple G-proteins from their polyisoprenyl-dependent interactions. Overexpression and/or hyper-activation of some members of the RHO family of small GTPases enhance F-actin remodeling, which is central to cell migration and invasion processes involved in metastasis [67]. Therefore, the observed decreases in RhoA, vinculin and fascin levels upon PCAIs treatment explain the significant inhibition of A549 cells migration and invasion given the F-actin cross-linking roles of vinculin and fascin that bridge integrins to the actin cytoskeleton [52]. It has been reported that depletion of vinculin disrupts cell adhesion and promotes apoptosis [52]. Moreover, fascin has been reported to be overexpressed in various cancer types [68]. PCAIs-mediated depletion of vinculin and fascin may be through weakening of integrin-F-actin linkages and enhancing F-actin loss, thereby inhibiting cell migration and invasion. Other PCAIs were shown to disrupt vinculin punctates in NCI-H1299 cells [69, 70]. The proposed mechanism of action on how the PCAIs interact with the respective G-proteins were mainly based on their structure and how they can displace the target proteins. The effects of the PCAIs on the G-proteins may be due to direct physical competitive displacement that may result in more rapid degradation than when they are in complex with other proteins. It is still uncertain what happens to the proteins after they got displaced but based on our results, some of them appear to be degraded and secreted out. We hypothesize that the secretory pathway may be involved in the depletion of RAC1 and RHOA since the PCAIs treatment resulted in significant amounts of both proteins in the experimental media. Other processes such as feedback control on G-proteins may more accurately explain the changes in NRAS and HRAS only in some cells. Furthermore, our previous results show that PCAIs induce the phosphorylation activation of the MAP kinase pathway enzymes resulting in the phosphorylation of p90RSK [56]. Phosphorylated p90RSK is known to inhibit Son-of-Sevenless (SoS) [71]. p90RSK regulates cAMP-response element binding protein (CREB) [72, 73], thereby affecting gene transcription. It may also alter translation by phosphorylating ribosomal proteins [71]. The latter two effects would alter intracellular protein levels that may include changes in the NRAS and HRAS proteins. In conclusion, mutations in G-proteins have been associated in the progress of several cancers, thus, a new approach on developing new anticancer therapies by targeting these proteins will be tantamount to finding the cure. Our results show that PCAIs deplete the protein levels of some significant G-proteins which are known to be involved in the migration and invasion of cells (i.e., metastasis) such as RAC1, RHOA, and CDC42. Furthermore, the PCAIs also affect the expression of vinculin and fascin which are both important for cell motility by forming F-actin linkages. The initial findings presented here indicate how PCAIs can be used as potent agents in developing new anticancer therapeutics, therefore, more extensive studies need to be done to elucidate on its potency. Although we cannot conclusively explain the exact mechanism of action of PCAIs on how they affect the levels of some G-proteins yet, but we can say that these PCAIs have the ability to affect the progression of cancer. ## Materials Cell lines were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA). Antibodies specific to KRAS (Cat. # 53270), RHOA (Cat. # 2117), RAC$\frac{1}{2}$/3 (Cat. # 2465), RAB5A (Cat. # 46449), CDC42 (Cat. # 2462), Vinculin (Cat. # 18799), Fascin (Cat. # 54545) GAPDH (HRP Conjugate) (Cat. # 8884), α-Actinin (HRP Conjugate) (Cat. # 12413), anti-mouse IgG, HRP-linked Antibody (Cat. # 7076), and anti-rabbit IgG, HRP-linked Antibody (Cat. # 7074) were purchased from Cell Signaling Technology (Danvers, MA, USA). Antibodies specific to HRAS (MAB3617) and NRAS (MAB10009) were purchased from Thermo Scientific (Waltham, MA, USA). The PCAIs used in this study (Table 4) were synthesized in our lab as previously described [74, 75]. **Table 4** | Compound | | --- | | NSL-YHJ-2-27 | | NSL-YHJ-2-62 | ## Cell culture A549 (CCL-185)- collected from a 58-year old Caucasian male, it is a hypotriploid human cell line with the modal chromosome number 66 which can be found in $24\%$ of cells, NCI-H1299 (CRL-5803)- established from a lymph node of 43-year old White male patient with lung cancer who received prior radiation therapy, MDA-MB-231(HTB-26)- obtained from a 51-year old White female, it is an aneuploid female with a modal chromosome number 64, and MDA-MB-468 (HTB-132)- isolated from a pleural effusion of a 51-year old Black woman with a metastatic breast adenocarcinoma, an aneuploid female with most chromosome counts in the hypertriploid range with a modal chromosome number 64. The A549, MDA-MB-231, and MDA-MB-468 cells were cultured in high glucose Modified Eagle Medium (DMEM) (Genesee Scientific, San Diego, CA, USA) while NCI-H1299 cells were cultured in RPMI 1640 (Genesee Scientific, San Diego, CA, USA). All media were supplemented with $10\%$ heat-inactivated fetal bovine serum (Genesee Scientific, San Diego, CA, USA), 100 U/mL penicillin and 100 μg/mL streptomycin (Genesee Scientific, San Diego, CA, USA). The cultures were incubated at 37ºC in $5\%$ CO2/$95\%$ humidified air. In all cases, treatment with experimental compounds was done in basal medium supplemented with $5\%$ heat-inactivated fetal bovine serum. ## Effects of PCAIs on MDA-MB-468 cell line To determine the potency of the PCAIs, NSL-YHJ-2-27 and the control analog, NSL-YHJ-2-62, cell viability assay was conducted. Briefly, 1 × 104 cells/well of MDA-MB-468 cells were plated into 96-well culture plates (Genesee Scientific, San Diego, CA, USA) in experimental medium. When the cells adhered to the wells, the respective analogs were added to final concentrations of 0.5, 1, 2, 5, 10, 20, 50 μM at the beginning and after 24 h of incubation. Acetone ($1\%$ final concentration) used as the vehicle solvent was used for the control treatment. The cells were exposed to the compounds for 48 h after which bright-field microscope images (10× magnification) were captured using the Nikon Eclipse microscope to evaluate physical changes on the cells. Then resazurin reduction assay was conducted by adding $0.02\%$ of resazurin reagent dissolved in PBS into the cells. They were then incubated at 37ºC in $5\%$ CO2/$95\%$ humidified air for 2 h. Using SoftMax Pro Reader version 5.4 for Windows (Molecular Devices, CA, USA), the fluorescence intensities were determined by setting the excitation frequency at 544 nm and emission at 590 nm. The cell viability was expressed as the percentage of the fluorescence in the cells treated with the compounds relative to the control (0 μM). These were then plotted in a non-linear regression curve fit using GraphPad Prism version 8.0 for Windows (San Diego, CA, USA) to determine the EC50 value for each compound. ## Effect of PCAIs on the G-proteins Cells in complete medium were plated into 60.8 cm2 tissue culture dishes (Genesee Scientific, San Diego, CA, USA) at a cell density of 7 × 105 (or 1 × 106) cells/dish and then incubated for 24 h to allow the cells to adhere. Adherent cells were treated with varying concentrations of PCAIs (0–5 μM) in experimental medium (supplemented with $5\%$ heat-inactivated FBS). After 24 h, equivalent amounts of PCAIs were used to treat the cells for the 48 h exposure. Cells were washed with PBS and then lysed with RIPA buffer (Genesee Scientific, San Diego, CA, USA) supplemented with $0.1\%$ v/v protease/phosphatase inhibitor cocktail (Cell Signaling Technology, Danvers, MA, USA). The amount of protein in lysates was determined using the Quick Start™ Bradford protein assay (Bio-Rad, Hercules, CA, USA). Cell lysates containing equal protein amounts (20–30 μg) were boiled in XT sample buffer with XT reducing agent (Bio-Rad, Hercules, CA, USA). The samples were separated by SDS-PAGE on $12\%$ Criterion™ XT Bis-Tris protein gels and transferred onto Trans-Blot turbo midi 0.2 μm nitrocellulose membranes (Bio-Rad, Hercules, CA, USA). Membranes were blocked for 1 h at room temperature with OneBlock™ western-CL blocking buffer (Genesee Scientific, San Diego, CA, USA) and incubated overnight in blocking buffer containing respective monoclonal antibodies against the target proteins at 4°C. Membranes were then washed with 1X TBST and incubated with anti-rabbit or anti-mouse IgG, HRP-linked antibodies at room temperature for 2 h. Immunoreactive bands were then visualized using ProSignal® Pico (Genesee Scientific, San Diego, CA, USA) or Radiance Plus (Azure Biosystems, Dublin, CA, USA) ECL reagents per manufacturer’s recommendations using the ChemiDoc XRS+ System (Bio-Rad, Hercules CA, USA). Protein levels as judged by the chemiluminescent intensities were quantified using Image Lab 6.0 (Bio-Rad, Hercules CA, USA), normalized against the corresponding band intensities of either GAPDH or α-Actinin. The results from three independent trials were then plotted using GraphPad Prism version 8.0 for Windows (San Diego, CA, USA). ## Effect of PCAIs on the degradation of G-proteins MDA-MB-468 cells were plated into 60.8 cm2 tissue culture dishes (Genesee Scientific, San Diego, CA, USA) at a cell density of 1.5 × 106 cells/dish in complete medium and then incubated for 24 h to allow the cells to adhere. Adherent cells were treated with 0 (control) or 5 μM concentrations of PCAIs in experimental medium. After 24 h, PCAIs treatments were repeated followed by a further 24 h incubation. The presence of G-proteins in the incubation media was then determined as follows. Using a vacuum concentrator (Labconco, USA), 1.5 mL of collected media mixed with 4× loading buffer and 20× reducing agent were vacuum concentrated to 200 μL over 1 h at 30°C. The concentrated media were subjected to Western blotting using respective antibodies to detect the target proteins, RAC1, CDC42, RHOA, and KRAS. Simple Western Blotting using Jess assay (Protein Simple, Bio-Techne, MN, USA) was used to analyze the results. Experiments were done according to the manufacturer’s protocol. Briefly, the concentrated media were mixed with 0.1× sample buffer and 5× master mix in 600 μL tube. The samples were denatured for 5 minutes at 95°C and were kept on ice during loading. Running conditions were set in the machine – 30 min blocking, 1 h primary antibody and 1 h secondary antibody. ## Effect of PCAIs on cell migration The effects of PCAIs on cellular migration were determined using the wound healing assay. Cell culture inserts purchased from ibidi (Martinsried, GE) were used according to the manufacturer’s protocol to generate two confluent monolayers of cells separated by a “wound” for this assay as previously described [70]. Briefly, cells (2–4 × 105 cells/mL) were seeded into each side of the ibidi-cell culture inserts attached onto the wells of a 12-well plate. The plate was then incubated (37°C/$5\%$ CO2) overnight to allow the cells to attach onto the plate and form two adherent confluent monolayers of cells on either side of the tissue culture insert. The next day, the insert was gently removed to generate a gap or “wound” between the two confluent layers of cells. The monolayers were washed with experimental media once and then fresh experimental media containing varying concentrations of NSL-YHJ-2-27 (0–5 μM) were added. Bright-field microscope images around the “wound” were captured at 0 and 24 h using the Nikon Eclipse microscope. The number of cells that migrated into the “wounds” were counted for control and treated cells. Data were analyzed using GraphPad Prism version 8.0 for Windows (San Diego, CA, USA). ## Effect of PCAIs on cell invasion The effects of PCAIs on cellular invasion were determined using the transwell invasion assay. The 24-well BD Biocoat Matrigel invasion chambers and inserts (catalog #354480) (Corning, Bedford, MA, USA) were used according to the manufacturer’s protocol. Briefly, A549 cells were harvested and plated in T-25 flasks at 2.0 × 105 cells per flask and allowed to attach overnight. The following day media was replaced with media containing NSL-YHJ-2-27 (0–5 μM) and incubated for 24 h. The 24-well BD Biocoat Matrigel invasion chambers and inserts were rehydrated per manufacturer’s protocol with serum free media and incubated at 37°C for 2 h. The treated cells were harvested to create a cell suspension of 5.0 × 104 cells/mL. Rehydrating media was removed from the wells and inserts. Media containing $10\%$ FBS was added to the wells while 500 μL of the cell suspension was added into the inserts. Invasion inserts containing cell suspensions were then carefully transferred into the wells containing media with $10\%$ FBS. The cells were incubated for 24 h (37 C/$5\%$ CO2) to allow the cells to invade from the upper chamber through the Matrigel into the lower chambers of the inserts. After incubation, non-invasive cells were quickly removed with a cotton swab. Invasive cells remaining in the inserts were fixed with $4\%$ formaldehyde in PBS for 20 min and then stained with $1\%$ crystal violet. Invading cells were imaged using Nikon Eclipse microscope, quantified using the Nikon NIS-Elements software, analyzed using GraphPad Prism, and plotted as the means numbers of migrated cells against NSL-YHJ-2-27 concentration. ## Effect of PCAIs on vinculin and fascin Cells in complete medium were plated into 60.8 cm2 tissue culture dishes (Genesee Scientific, San Diego, CA, USA) at a cell density of 7 × 105 cells/dish and then incubated for 24 h to allow the cells to adhere. Adherent cells were treated with 0–5 μM of PCAIs in experimental medium. After 24 h, equivalent amounts of PCAIs were used to treat the cells for the 48 h exposure. Cells were washed with PBS and then lysed with RIPA buffer (Genesee Scientific, San Diego, CA, USA) supplemented with $0.1\%$ v/v protease/phosphatase inhibitor cocktail (Cell Signaling Technology, Danvers, MA, USA). The amount of protein in lysates was determined using the Quick Start™ Bradford protein assay (Bio-Rad, Hercules, CA, USA). Cell lysates containing equal protein amounts (20–30 μg) were boiled in XT sample buffer with XT reducing agent (Bio-Rad, Hercules, CA, USA). The samples were separated by SDS-PAGE on a $4\%$–$12\%$ Criterion™ XT Bis-Tris protein gels and transferred onto Trans-Blot turbo midi 0.2 μm nitrocellulose membranes (Bio-Rad, Hercules, CA, USA). Membranes were blocked for 1 h at room temperature with OneBlock™ western-CL blocking buffer (Genesee Scientific, San Diego, CA, USA) and incubated overnight in blocking buffer containing monoclonal antibodies against the target proteins vinculin and fascin at 4°C. Membranes were then washed with 1X TBST and incubated with anti-rabbit or anti-mouse IgG, HRP-linked antibodies at room temperature for 2 h. Immunoreactive bands were then visualized using ProSignal® Pico (Genesee Scientific, San Diego, CA, USA) or Radiance Plus (Azure Biosystems, Dublin, CA, USA) ECL reagents per manufacturer’s recommendations using the ChemiDoc XRS+ System (Bio-Rad, Hercules CA, USA). Protein levels as judged by the chemiluminescent intensities were quantified using Image Lab 6.0 (Bio-Rad, Hercules CA, USA), normalized against the corresponding band intensities of either GAPDH or α-Actinin. The results were then plotted using GraphPad Prism version 8.0 for Windows (San Diego, CA, USA). ## Statistical analysis All results are the means ± SEM. To determine statistical significance, the values of each treatment group were compared to the respective controls by One-Way ANOVA with Dunnett’s post-hoc test using GraphPad Prism version 8.0 for Windows (San Diego, CA, USA) and p values less than 0.05 were considered significant. ## References 1. Reiner DJ, Lundquist EA. **Small GTPases**. *WormBook* (2018) **2018** 1-65. DOI: 10.1895/wormbook.1.67.2 2. Wennerberg K, Rossman KL, Der CJ. **The Ras superfamily at a glance**. *J Cell Sci* (2005) **118** 843-46. DOI: 10.1242/jcs.01660 3. Vetter IR, Wittinghofer A. **The guanine nucleotide-binding switch in three dimensions**. *Science* (2001) **294** 1299-304. DOI: 10.1126/science.1062023 4. Cox AD, Der CJ. **Ras history: The saga continues**. *Small GTPases* (2010) **1** 2-27. DOI: 10.4161/sgtp.1.1.12178 5. Malumbres M, Pellicer A. **RAS pathways to cell cycle control and cell transformation**. *Front Biosci* (1998) **3** d887-912. DOI: 10.2741/a331 6. Reuther GW, Der CJ. **The Ras branch of small GTPases: Ras family members don’t fall far from the tree**. *Curr Opin Cell Biol* (2000) **12** 157-65. DOI: 10.1016/s0955-0674(99)00071-x 7. Donaldson JG, Jackson CL. **Regulators and effectors of the ARF GTPases**. *Curr Opin Cell Biol* (2000) **12** 475-82. DOI: 10.1016/s0955-0674(00)00119-8 8. Li G, Stahl PD. **Structure-function relationship of the small GTPase rab5**. *J Biol Chem* (1993) **268** 24475-80. DOI: 10.1016/S0021-9258(20)80550-X 9. Kuersten S, Ohno M, Mattaj IW. **Nucleocytoplasmic transport: Ran, beta and beyond**. *Trends Cell Biol* (2001) **11** 497-503. DOI: 10.1016/s0962-8924(01)02144-4 10. Heider D, Hauke S, Pyka M, Kessler D. **Insights into the classification of small GTPases**. *Adv Appl Bioinform Chem* (2010) **3** 15-24. DOI: 10.2147/aabc.s8891 11. Orgaz JL, Herraiz C, Sanz-Moreno V. **Rho GTPases modulate malignant transformation of tumor cells**. *Small GTPases* (2014) **5** e29019. DOI: 10.4161/sgtp.29019 12. Karnoub AE, Weinberg RA. **Ras oncogenes: split personalities**. *Nat Rev Mol Cell Biol* (2008) **9** 517-31. DOI: 10.1038/nrm2438 13. Prior IA, Hood FE, Hartley JL. **The Frequency of Ras Mutations in Cancer**. *Cancer Res* (2020) **80** 2969-74. DOI: 10.1158/0008-5472.CAN-19-3682 14. Fritz G, Just I, Kaina B. **Rho GTPases are over-expressed in human tumors**. *Int J Cancer* (1999) **81** 682-87. DOI: 10.1002/(sici)1097-0215(19990531)81:5<682::aid-ijc2>3.0.co;2-b 15. Liu Y, Wang Y, Zhang Y, Miao Y, Zhao Y, Zhang PX, Jiang GY, Zhang JY, Han Y, Lin XY, Yang LH, Li QC, Zhao C, Wang EH. **Abnormal expression of p120-catenin, E-cadherin, and small GTPases is significantly associated with malignant phenotype of human lung cancer**. *Lung Cancer* (2009) **63** 375-82. DOI: 10.1016/j.lungcan.2008.12.012 16. Regala RP, Weems C, Jamieson L, Copland JA, Thompson EA, Fields AP. **Atypical protein kinase Ciota plays a critical role in human lung cancer cell growth and tumorigenicity**. *J Biol Chem* (2005) **280** 31109-15. DOI: 10.1074/jbc.M505402200 17. Lang W, Wang H, Ding L, Xiao L. **Cooperation between PKC-alpha and PKC-epsilon in the regulation of JNK activation in human lung cancer cells**. *Cell Signal* (2004) **16** 457-67. DOI: 10.1016/j.cellsig.2003.09.002 18. Hall A. **Rho GTPases and the actin cytoskeleton**. *Science* (1998) **279** 509-14. DOI: 10.1126/science.279.5350.509 19. Nobes CD, Hall A. **Rho, rac and cdc42 GTPases: regulators of actin structures, cell adhesion and motility**. *Biochem Soc Trans* (1995) **23** 456-59. DOI: 10.1042/bst0230456 20. Nobes CD, Hall A. **Rho, rac, and cdc42 GTPases regulate the assembly of multimolecular focal complexes associated with actin stress fibers, lamellipodia, and filopodia**. *Cell* (1995) **81** 53-62. DOI: 10.1016/0092-8674(95)90370-4 21. Hall A. **The cytoskeleton and cancer**. *Cancer Metastasis Rev* (2009) **28** 5-14. DOI: 10.1007/s10555-008-9166-3 22. Colicelli J. **Human RAS superfamily proteins and related GTPases**. *Sci STKE* (2004) **2004** RE13. DOI: 10.1126/stke.2502004re13 23. Rajalingam K, Schreck R, Rapp UR, Albert S. **Ras oncogenes and their downstream targets**. *Biochim Biophys Acta* (2007) **1773** 1177-95. DOI: 10.1016/j.bbamcr.2007.01.012 24. Ahearn IM, Haigis K, Bar-Sagi D, Philips MR. **Regulating the regulator: post-translational modification of RAS**. *Nat Rev Mol Cell Biol* (2011) **13** 39-51. DOI: 10.1038/nrm3255 25. Silvius JR, l’Heureux F. **Fluorimetric evaluation of the affinities of isoprenylated peptides for lipid bilayers**. *Biochemistry* (1994) **33** 3014-22. DOI: 10.1021/bi00176a034 26. Schroeder H, Leventis R, Rex S, Schelhaas M, Nägele E, Waldmann H, Silvius JR. **S-Acylation and plasma membrane targeting of the farnesylated carboxyl-terminal peptide of N-ras in mammalian fibroblasts**. *Biochemistry* (1997) **36** 13102-9. DOI: 10.1021/bi9709497 27. Donaldson JG. **Arfs and membrane lipids: sensing, generating and responding to membrane curvature**. *Biochem J* (2008) **414** e1-2. DOI: 10.1042/BJ20081438 28. Friday BB, Adjei AA. **K-ras as a target for cancer therapy**. *Biochim Biophys Acta* (2005) **1756** 127-44. DOI: 10.1016/j.bbcan.2005.08.001 29. Cox AD, Fesik SW, Kimmelman AC, Luo J, Der CJ. **Drugging the undruggable RAS: Mission possible?**. *Nat Rev Drug Discov* (2014) **13** 828-51. DOI: 10.1038/nrd4389 30. Sánchez-Muñoz A, Gallego E, de Luque V, Pérez-Rivas LG, Vicioso L, Ribelles N, Lozano J, Alba E. **Lack of evidence for KRAS oncogenic mutations in triple-negative breast cancer**. *BMC Cancer* (2010) **10** 136. DOI: 10.1186/1471-2407-10-136 31. Neumann J, Zeindl-Eberhart E, Kirchner T, Jung A. **Frequency and type of KRAS mutations in routine diagnostic analysis of metastatic colorectal cancer**. *Pathol Res Pract* (2009) **205** 858-62. DOI: 10.1016/j.prp.2009.07.010 32. **Comprehensive molecular characterization of gastric adenocarcinoma**. *Nature* (2014) **513** 202-9. DOI: 10.1038/nature13480 33. Hodis E, Watson IR, Kryukov GV, Arold ST, Imielinski M, Theurillat JP, Nickerson E, Auclair D, Li L, Place C, Dicara D, Ramos AH, Lawrence MS. **A landscape of driver mutations in melanoma**. *Cell* (2012) **150** 251-63. DOI: 10.1016/j.cell.2012.06.024 34. Adams KF, Schatzkin A, Harris TB, Kipnis V, Mouw T, Ballard-Barbash R, Hollenbeck A, Leitzmann MF. **Overweight, obesity, and mortality in a large prospective cohort of persons 50 to 71 years old**. *N Engl J Med* (2006) **355** 763-78. DOI: 10.1056/NEJMoa055643 35. Parri M, Chiarugi P. **Rac and Rho GTPases in cancer cell motility control**. *Cell Commun Signal* (2010) **8** 23. DOI: 10.1186/1478-811X-8-23 36. Fritz G, Brachetti C, Bahlmann F, Schmidt M, Kaina B. **Rho GTPases in human breast tumours: expression and mutation analyses and correlation with clinical parameters**. *Br J Cancer* (2002) **87** 635-44. DOI: 10.1038/sj.bjc.6600510 37. Forbes SA, Bindal N, Bamford S, Cole C, Kok CY, Beare D, Jia M, Shepherd R, Leung K, Menzies A, Teague JW, Campbell PJ, Stratton MR, Futreal PA. **COSMIC: mining complete cancer genomes in the Catalogue of Somatic Mutations in Cancer**. *Nucleic Acids Res* (2011) **39** D945-50. DOI: 10.1093/nar/gkq929 38. Schnelzer A, Prechtel D, Knaus U, Dehne K, Gerhard M, Graeff H, Harbeck N, Schmitt M, Lengyel E. **Rac1 in human breast cancer: overexpression, mutation analysis, and characterization of a new isoform, Rac1b**. *Oncogene* (2000) **19** 3013-20. DOI: 10.1038/sj.onc.1203621 39. Shieh DB, Godleski J, Herndon JE, Azuma T, Mercer H, Sugarbaker DJ, Kwiatkowski DJ. **Cell motility as a prognostic factor in Stage I nonsmall cell lung carcinoma: the role of gelsolin expression**. *Cancer* (1999) **85** 47-57. DOI: 10.1002/(sici)1097-0142(19990101)85:1<47::aid-cncr7>3.0.co;2-l 40. Gómez Del Pulgar T, Valdés-Mora F, Bandrés E, Pérez-Palacios R, Espina C, Cejas P, García-Cabezas MA, Nistal M, Casado E, González-Barón M, García-Foncillas J, Lacal JC. **Cdc42 is highly expressed in colorectal adenocarcinoma and downregulates ID4 through an epigenetic mechanism**. *Int J Oncol* (2008) **33** 185-93. DOI: 10.3892/IJO.33.1.185 41. Zerial M, McBride H. **Rab proteins as membrane organizers**. *Nat Rev Mol Cell Biol* (2001) **2** 107-17. DOI: 10.1038/35052055 42. Stenmark H, Valencia A, Martinez O, Ullrich O, Goud B, Zerial M. **Distinct structural elements of rab5 define its functional specificity**. *EMBO J* (1994) **13** 575-83. DOI: 10.1002/j.1460-2075.1994.tb06295.x 43. Shimizu H, Toma-Fukai S, Kontani K, Katada T, Shimizu T. **GEF mechanism revealed by the structure of SmgGDS-558 and farnesylated RhoA complex and its implication for a chaperone mechanism**. *Proc Natl Acad Sci U S A* (2018) **115** 9563-68. DOI: 10.1073/pnas.1804740115 44. Wennerberg K, Der CJ. **Rho-family GTPases: it’s not only Rac and Rho (and I like it)**. *J Cell Sci* (2004) **117** 1301-12. DOI: 10.1242/jcs.01118 45. Münzberg E, Stein M. **Structure and Dynamics of Mono- vs. Doubly Lipidated Rab5 in Membranes**. *Int J Mol Sci* (2019) **20** 4773. DOI: 10.3390/ijms20194773 46. Cox AD, Der CJ. **Ras family signaling: therapeutic targeting**. *Cancer Biol Ther* (2002) **1** 599-606. DOI: 10.4161/cbt.306 47. Wright LP, Philips MR. **Thematic review series: lipid posttranslational modifications. CAAX modification and membrane targeting of Ras**. *J Lipid Res* (2006) **47** 883-91. DOI: 10.1194/jlr.R600004-JLR200 48. Abdelkarim H, Banerjee A, Grudzien P, Leschinsky N, Abushaer M, Gaponenko V. **The Hypervariable Region of K-Ras4B Governs Molecular Recognition and Function**. *Int J Mol Sci* (2019) **20** 5718. DOI: 10.3390/ijms20225718 49. Marciano D, Ben-Baruch G, Marom M, Egozi Y, Haklai R, Kloog Y. **Farnesyl derivatives of rigid carboxylic acids-inhibitors of ras-dependent cell growth**. *J Med Chem* (1995) **38** 1267-72. DOI: 10.1021/jm00008a004 50. Marom M, Haklai R, Ben-Baruch G, Marciano D, Egozi Y, Kloog Y. **Selective inhibition of Ras-dependent cell growth by farnesylthiosalisylic acid**. *J Biol Chem* (1995) **270** 22263-70. DOI: 10.1074/jbc.270.38.22263 51. Haklai R, Weisz MG, Elad G, Paz A, Marciano D, Egozi Y, Ben-Baruch G, Kloog Y. **Dislodgment and accelerated degradation of Ras**. *Biochemistry* (1998) **37** 1306-14. DOI: 10.1021/bi972032d 52. Tawfeeq N, Jin Y, Lamango NS. **Synthetic Optimization and MAPK Pathway Activation Anticancer Mechanism of Polyisoprenylated Cysteinyl Amide Inhibitors**. *Cancers (Basel)* (2021) **13** 5757. DOI: 10.3390/cancers13225757 53. Goldmann WH, Auernheimer V, Thievessen I, Fabry B. **Vinculin, cell mechanics and tumour cell invasion**. *Cell Biol Int* (2013) **37** 397-405. DOI: 10.1002/cbin.10064 54. Riou P, Kjær S, Garg R, Purkiss A, George R, Cain RJ, Bineva G, Reymond N, McColl B, Thompson AJ, O’Reilly N, McDonald NQ, Parker PJ, Ridley AJ. **14-3-3 proteins interact with a hybrid prenyl-phosphorylation motif to inhibit G proteins**. *Cell* (2013) **153** 640-53. DOI: 10.1016/j.cell.2013.03.044 55. Hoffman GR, Nassar N, Cerione RA. **Structure of the Rho family GTP-binding protein Cdc42 in complex with the multifunctional regulator RhoGDI**. *Cell* (2000) **100** 345-56. DOI: 10.1016/s0092-8674(00)80670-4 56. Bhagatji P, Leventis R, Rich R, Lin CJ, Silvius JR. **Multiple cellular proteins modulate the dynamics of K-ras association with the plasma membrane**. *Biophys J* (2010) **99** 3327-35. DOI: 10.1016/j.bpj.2010.10.001 57. Wang M, Casey PJ. **Protein prenylation: unique fats make their mark on biology**. *Nat Rev Mol Cell Biol* (2016) **17** 110-22. DOI: 10.1038/nrm.2015.11 58. Meinohl C, Barnard SJ, Fritz-Wolf K, Unger M, Porr A, Heipel M, Wirth S, Madlung J, Nordheim A, Menke A, Becker K, Giehl K. **Galectin-8 binds to the Farnesylated C-terminus of K-Ras4B and Modifies Ras/ERK Signaling and Migration in Pancreatic and Lung Carcinoma Cells**. *Cancers (Basel)* (2019) **12** 30. DOI: 10.3390/cancers12010030 59. Zimmermann G, Papke B, Ismail S, Vartak N, Chandra A, Hoffmann M, Hahn SA, Triola G, Wittinghofer A, Bastiaens PI, Waldmann H. **Small molecule inhibition of the KRAS-PDEδ interaction impairs oncogenic KRAS signalling**. *Nature* (2013) **497** 638-42. DOI: 10.1038/nature12205 60. Leung EL, Luo LX, Li Y, Liu ZQ, Li LL, Shi DF, Xie Y, Huang M, Lu LL, Duan FG, Huang JM, Fan XX, Yuan ZW. **Identification of a new inhibitor of KRAS-PDEδ interaction targeting KRAS mutant nonsmall cell lung cancer**. *Int J Cancer* (2019) **145** 1334-45. DOI: 10.1002/ijc.32222 61. Gelb MH, Brunsveld L, Hrycyna CA, Michaelis S, Tamanoi F, Van Voorhis WC, Waldmann H. **Therapeutic intervention based on protein prenylation and associated modifications**. *Nat Chem Biol* (2006) **2** 518-28. DOI: 10.1038/nchembio818 62. Duverna R, Ablordeppey SY, Lamango NS. **Biochemical and docking analysis of substrate interactions with polyisoprenylated methylated protein methyl esterase**. *Curr Cancer Drug Targets* (2010) **10** 634-48. DOI: 10.2174/156800910791859443 63. Lamango NS, Duverna R, Zhang W, Ablordeppey SY. **Porcine Liver Carboxylesterase Requires Polyisoprenylation for High Affinity Binding to Cysteinyl Substrates**. *Open Enzym Inhib J* (2009) **2** 12-27. DOI: 10.2174/1874940200902010012 64. Magee AI, Gutierrez L, McKay IA, Marshall CJ, Hall A. **Dynamic fatty acylation of p21N-ras**. *EMBO J* (1987) **6** 3353-57. DOI: 10.1002/j.1460-2075.1987.tb02656.x 65. Shahinian S, Silvius JR. **Doubly-lipid-modified protein sequence motifs exhibit long-lived anchorage to lipid bilayer membranes**. *Biochemistry* (1995) **34** 3813-22. DOI: 10.1021/bi00011a039 66. Hancock JF, Paterson H, Marshall CJ. **A polybasic domain or palmitoylation is required in addition to the CAAX motif to localize p21ras to the plasma membrane**. *Cell* (1990) **63** 133-39. DOI: 10.1016/0092-8674(90)90294-o 67. Stricker J, Falzone T, Gardel ML. **Mechanics of the F-actin cytoskeleton**. *J Biomech* (2010) **43** 9-14. DOI: 10.1016/j.jbiomech.2009.09.003 68. Machesky LM, Li A. **Fascin: Invasive filopodia promoting metastasis**. *Commun Integr Biol* (2010) **3** 263-70. DOI: 10.4161/cib.3.3.11556 69. Ntantie E, Fletcher J, Amissah F, Salako OO, Nkembo AT, Poku RA, Ikpatt FO, Lamango NS. **Polyisoprenylated cysteinyl amide inhibitors disrupt actin cytoskeleton organization, induce cell rounding and block migration of non-small cell lung cancer**. *Oncotarget* (2017) **8** 31726-44. DOI: 10.18632/oncotarget.15956 70. Ntantie E, Allen MJ, Fletcher J, Nkembo AT, Lamango NS, Ikpatt OF. **Suppression of focal adhesion formation may account for the suppression of cell migration, invasion and growth of non-small cell lung cancer cells following treatment with polyisoprenylated cysteinyl amide inhibitors**. *Oncotarget* (2018) **9** 25781-95. DOI: 10.18632/oncotarget.25372 71. Houles T, Roux PP. **Defining the role of the RSK isoforms in cancer**. *Semin Cancer Biol* (2018) **48** 53-61. DOI: 10.1016/j.semcancer.2017.04.016 72. Bruning JC, Gillette JA, Zhao Y, Bjorbaeck C, Kotzka J, Knebel B, Avci H, Hanstein B, Lingohr P, Moller DE, Krone W, Kahn CR, Muller-Wieland D. **Ribosomal subunit kinase-2 is required for growth factor-stimulated transcription of the c-Fos gene**. *Proc Natl Acad Sci U S A* (2000) **97** 2462-67. DOI: 10.1073/pnas.97.6.2462 73. De Cesare D, Jacquot S, Hanauer A, Sassone-Corsi P. **Rsk-2 activity is necessary for epidermal growth factor-induced phosphorylation of CREB protein and transcription of c-fos gene**. *Proc Natl Acad Sci U S A* (1998) **95** 12202-7. DOI: 10.1073/pnas.95.21.12202 74. Aguilar BJ, Nkembo AT, Duverna R, Poku RA, Amissah F, Ablordeppey SY, Lamango NS. **Polyisoprenylated methylated protein methyl esterase: a putative biomarker and therapeutic target for pancreatic cancer**. *Eur J Med Chem* (2014) **81** 323-33. DOI: 10.1016/j.ejmech.2014.05.018 75. Tawfeeq N, Jin Y, Lamango NS. *Synthetic optimization of polyisoprenylated cysteinyl amide inhibitors to target cancers with hyperactive G-proteins 111th Annual Meeting of the American Association for Cancer Research* (2020)
--- title: The Prevalence and Characteristics of Body Dysmorphic Disorder Among Adults in Makkah City, Saudi Arabia. A Cross-Sectional Study journal: Cureus year: 2023 pmcid: PMC10038359 doi: 10.7759/cureus.35316 license: CC BY 3.0 --- # The Prevalence and Characteristics of Body Dysmorphic Disorder Among Adults in Makkah City, Saudi Arabia. A Cross-Sectional Study ## Abstract Background: Body dysmorphic disorder (BDD) is a mental health condition where a person spends much time worrying about flaws in their appearance. The international prevalence of BDD had been reported, and it was about 1.9-$2.2\%$. Objectives: The current study aims to explore the prevalence of BDD among the general population in Makkah, Saudi Arabia. Methods: *This is* a descriptive cross-sectional study that used an electronic questionnaire. It was distributed to the general population using the convenience sample technique between September 2021 to November 2021. BDD was assessed among the participants using an Arabic-validated tool. The sample size was calculated to be 385 participants. Results: The study included a total of 392 participants. Most of them were female ($59.7\%$), 18-27 years old ($62.8\%$), and had bachelor or post-graduate degrees ($67.6\%$). Among all the included participants, only 28 met the criteria of BDD ($7.1\%$). The BDD population had an equal gender ratio, and most included respondents between 18 and 27 years old ($78.6\%$), college students ($60.7\%$), those with the lowest income level (< 5,000 SR) ($78.6\%$), and who had a normal body mass index ($46.4\%$). Conclusion: The prevalence of BDD in Makkah, Saudi Arabia, was $7.1\%$. No significant differences were noticed between BDD and non-BDD groups in age, obesity, and gender. ## Introduction Body dysmorphic disorder (BDD) is a psychological condition in which the patients have an excessive pathological concern about their appearance in addition to excessive fear of ugliness regarding certain aspects that are considered "not right" or even "not noticeable" by others; it is important to know that in the presence of observable defects on appearance, BDD should not be diagnosed [1]. BDD patients have constant worries, anxiety, and sadness about their looks in a way that causes a significant impairment of function and interferes with their lives [1]. They strongly believe they have a defect in their appearance, which makes them ugly or deformed. As a result, they become stressed [2-4]. Reviewing the literature, the risk factors associated with BDD are currently unclear. Nevertheless, multiple risk factors are hypothesized for the development of BDD, including genetic predisposition and childhood adversity, such as bullying or teasing [2,4]. Moreover, lack of family support or sexual abuse may be non-specific factors, in addition to being more aesthetically sensitive than average [2]. A systematic review conducted in 2016 showed that the estimated prevalence of BDD in the general population is around 1.9-$2.2\%$ [5]. Locally in Saudi Arabia, the reported prevalence of BDD among the general population is around 4.2-$8.8\%$, and among female students is 4.4-$12.3\%$ [6-9]. Since there is no available up-to-date data concerning the BDD prevalence in Makkah City, Saudi Arabia, to address this, the current study explores the prevalence of body dysmorphic disorder among the general population in Makkah City, Saudi Arabia. ## Materials and methods Study design, population, and sampling This descriptive cross-sectional study was conducted in public places (malls, cafes, and gardens) in Makkah City, Saudi Arabia, between September 2021 and November 2021. A convenience sample technique was used to select the participants. We included in our study all adult individuals above 18 years old. The exclusion criteria included non-Arabic speakers and participants with congenital anomalies or dermatological diseases. The minimum sample size required for this study was calculated by OpenEpi version 3.0 [10] in consideration of the following: the population size is about 8,325,304 individuals (according to the General Authority for Statistics), keeping the confidence interval (CI) level at $95\%$, and considering $50\%$ anticipated frequency. The sample size was calculated to be 385 participants. In case of possible data loss, the total sample size required is 400 participants. However, the final collected data was 392 participants. Data collection and instrument survey We used a valid, reliable, simple, understandable, Arabic-based self-report questionnaire that reserves participants' privacy. The questionnaire was distributed to all participants throughout the study using a standard way. Data was collected by giving the participants an electronic iPad containing the questionnaire, designed using Google forms. Informed consent was obtained before filling out the study questionnaire. In addition, the respondents received an iPad device and a request to participate in the study. We grouped the questionnaire's items into four main sections: Section 1 contains the consent form. Section 2 contains questions of the exclusion criteria, such as (do you have a congenital anomaly or dermatological disease); the participants who answered yes to these questions were referred to a submit page and excluded from the study. Section 3 addressed socio-demographic characteristics (e.g., age, gender, and level of education). Section 4 assessed BDD using the Arabic version of the Body Dysmorphic Disorder Questionnaire (BDDQ). Permission to use the BDDQ was with the corresponding author [6], which was a brief, self-reported measure derived from the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) diagnostic criteria for BDD [11]. The BDDQ comprises five closed-ended questions asking respondents if their appearance concerns are a source of worry. If yes, how much anguish are they in, or how much this interferes with their social or vocational functioning? *The diagnosis* is established when a person answers yes to the first two questions, any of the four items of the third question, b or c items of the fourth question, and no to the last question). Section 5 collects information about the participants' concerns, harassment, and psychiatric problems. Statical analysis Data was transferred into the Statistical Package for the Social Sciences (SPSS) software, version 26, for statistical analyses. Frequency, percentage, and mean ± SD were used to describe the participants' demographic data, the prevalence of BDD, and participants' concerns, harassment, and psychiatric problems. Crosstabulation was performed between the socio-demographic data among BDD and non-BDD participants based on Pearson's chi-square test. All statistical analyses were performed using two-tailed tests with an alpha error of 0.05. Ethical considerations Ethical approval was obtained from the Biomedical Ethics Committee of the Faculty of Medicine at Umm Al-Qura University (UQU), Makkah Al-Mukarramah, Saudi Arabia (approval number AHQP070921). ## Results Among 392 participants included in the study, only 28 met the criteria of BDD ($7.1\%$). Most BDD participants were between 18 and 27 years old ($78.6\%$), and the gender was distributed equally between males and females. Most participants were Saudis ($96.4\%$), and the most common educational level was a Bachelor's or Post-graduate degree, followed by a high school certificate or below ($50\%$ and $46.4\%$, respectively). Regarding occupation, $60.7\%$ were college students. Most participants earn less than 5000 Saudi Riyals monthly ($78.6\%$). Nearly half of the BDD participants had a normal range of Body Mass Index (BMI) ($46.4\%$). Most BDD individuals spend an average of more than 4 h on their mobile phones daily ($75\%$). There is no significant association between the BDD group or the non-BDD group with socio-demographic variables, as demonstrated in (Table 1). **Table 1** | Unnamed: 0 | BDD, N=28 (7.1%) | Non-BDD, N=364 (92.9%) | Total, N= 392 (100%) | P value** | | --- | --- | --- | --- | --- | | Age in years | | | | 0.19 | | 18-27 | 22 (78.6%) | 224 (61.5%) | 246 (62.8%) | 0.19 | | 28-37 | 4 (14.3%) | 102 (28.0%) | 106 (27.0%) | 0.19 | | ≥ 38 | 2 (7.1%) | 38 (10.4%) | 40 (10.2%) | 0.19 | | Gender | | | | 0.28 | | Male | 14 (50.0%) | 144 (39.6%) | 158 (40.3%) | 0.28 | | Female | 14 (50.0%) | 220 (60.4%) | 234 (59.7%) | 0.28 | | Nationality | | | | 0.38 | | Saudi | 27 (96.4%) | 334 (91.8%) | 361 (92.1%) | 0.38 | | Non-Saudi | 1 (3.6%) | 30 (8.2%) | 31 (7.9%) | 0.38 | | Educational level | | | | 0.08 | | High school certificate or below | 13 (46.4%) | 97 (26.6%) | 110 (28.1%) | 0.08 | | Diploma degree | 1 (3.6%) | 16 (4.4%) | 17 (4.3%) | 0.08 | | Bachelor's or Post-graduate degree | 14 (50%) | 251 (69%) | 265 (67.6%) | 0.08 | | Occupation | | | | 0.14 | | Employee* | 9 (32.1%) | 163 (44.8%) | 172 (43.9%) | 0.14 | | Unemployed | 2 (7.1%) | 49 (13.5%) | 51 (13%) | 0.14 | | College student | 17 (60.7%) | 152 (41.8%) | 169 (43.1%) | 0.14 | | Income | | | | 0.13 | | < 5,000 SR | 22 (78.6%) | 235 (64.6%) | 257 (65.6%) | 0.13 | | 5,000 – 9,999 SR | 5 (17.9%) | 62 (17.0%) | 67 (17.1%) | 0.13 | | ≥ 10,000 | 1 (3.6%) | 67 (18.4%) | 68 (17.4%) | 0.13 | | BMI*** | | | | 0.07 | | Underweight (<18.5) | 7 (25%) | 38 (10.4%) | 45 (11.5%) | 0.07 | | Normal range (18.5-24.9) | 13 (46.4%) | 152 (41.8%) | 165 (42.1%) | 0.07 | | Overweight (25.0-29.9) | 6 (21.4%) | 124 (34.1%) | 130 (33.2%) | 0.07 | | Obese (≥ 30) | 2 (7.1%) | 50 (13.7%) | 52 (13.3%) | 0.07 | | Average hours of phone use in one week | | | | 0.63 | | 2 hours or less | 1 (3.6%) | 5 (1.4%) | 6 (1.5%) | 0.63 | | 3 - 4 hours | 6 (21.4%) | 72 (19.8%) | 78 (19.9%) | 0.63 | | More than 4 hours | 21 (75%) | 287 (78.8%) | 308 (78.6%) | 0.63 | The detailed criteria of BDD according to DSM-IV among the included participants ($$n = 392$$) (Table 2). Exactly 246 of 392 individuals ($62.76\%$) were preoccupied and concerned with their appearance, and the most frequent body part concern was skin ($27.1\%$), and the minor concern was weight ($0.2\%$) (Table 3). Of those preoccupied, 139 wish to think less about their concerns. Out of those who answered yes to the previous two questions, 42 individuals answered (no) to the third question (the question that meant to exclude eating disorders), and of those, a final 28 participants answered (yes) to one of the four questions (Table 2). The most reported concern among the respondents was what they think about themself, not how others think about them ($46.4\%$). Additionally, the majority believe that their appearance looks normal to other people ($71.4\%$). When the participants were asked about the standards by which they judged others, most judged others by factors other than their appearance ($96.4\%$). Regarding the psychiatric history of the included participants, only one reported a history of depression, and another reported a history of anxiety disorders after visiting a psychiatrist (Table 4). **Table 4** | Item | Frequency | Percent (%) | | --- | --- | --- | | To what extent are you primarily concerned about? | | | | Primary concern is what others think | 8.0 | 28.6% | | Primary concern is what I think | 13.0 | 46.4% | | I do not have a concern | 7.0 | 25.0% | | How much do you believe others think your appearance is abnormal? | | | | Others think my appearance normal | 20.0 | 71.4% | | Others think my appearance is severely abnormal | 8.0 | 28.6% | | Standards by which you judge others? | | | | Mainly by appearance | 1.0 | 3.6% | | By other factors (for example: morals, way of thinking, the way by which a person treats) | 27.0 | 96.4% | | Have you ever tried to consult a psychiatrist? | | | | Yes | 2.0 | 7.1% | | No | 26.0 | 92.9% | | If answered with “yes” to the previous question, what was your diagnosis? | | | | Depression | 1.0 | 50.0% | | Anxiety | 1.0 | 50.0% | ## Discussion This is the first study to detect the prevalence of body dysmorphic disorder in the general population of Makkah, Saudi Arabia. The number of participants who met the criteria of BDD was 28 ($7.1\%$). Our prevalence rate is much higher than in Germany, the United States, and Turkey ($1.7\%$, $2.4\%$, and $6.3\%$, respectively) [12-14]. The international prevalence of BDD has been reported as 1.9-$2.2\%$ [5]. This discrepancy in the BDD prevalence rate could be attributed to the effect of sociocultural differences on body image, for example, between Arab and western populations. Different BDD screening instruments may produce different results [15]. In Saudi Arabia, multiple studies were conducted to assess the prevalence of BDD; the reported prevalence of BDD among the general population is between $4.2\%$ and $8.8\%$ and 4.4-$12.3\%$ among female students [6-9]. A study done in the AL-Qassim region of Saudi Arabia on dermatology patients shows that the prevalence of BDD is $18.6\%$ [16]. Most of the participants were young, from 18-27 years old ($89.8\%$), which is consistent with what was published in 2020 by The General Authority of Statistics in Saudi Arabia, in which the majority of the Saudi society is between 15 and 34 years old [17]. The current study revealed an equal distribution between males and females in the BDD population. Compared to previous studies included in a systematic review, there was female predominance with a ratio of 1.27 for the community of adults. In students, there was female predominance, with a ratio of 1.64; however, there was male predominance in cosmetic surgery (with a ratio of 0.71) and rhinoplasty (with a ratio of 0.91) [5]. Several reports have shown that BDD is associated with obsessive-compulsive disorder (OCD), anorexia nervosa, major depressive disorder (MDD), generalized anxiety disorder (GAD), and social anxiety disorder [18,19]. One study showed a prevalence rate of $12.1\%$ in OCD patients [18]. Our study found one participant with a history of depression and another with a history of anxiety disorders after visiting a psychiatrist. In comparison, Wei Li et al. found that two BDD patients suffered from comorbid MDD, three suffered from dysthymic disorder, one had panic disorder, one suffered from MDD and GAD, and one suffered from MDD, GAD, and social anxiety disorder [19]. According to the findings, there is no strong association between BDD and other mental disorders. A strong relationship between body dysmorphic disorder and obesity has been reported in the literature; one exciting finding is that most patients who visited clinics to do abdominoplasty present with mild to moderate BDD, with significant concerns about their weight and shape. This could be explained by the fact that approximately $70\%$ of the population in the western world is obese [20]. In addition, another study found a significant relationship between obesity/overweight and BDD [9]. However, our study shows an insignificant difference in obesity between BDD and non-BDD groups ($p \leq 0.05$). This finding is supported by previous studies, which have suggested that people with no BDD are more concerned about the shape or size of the belly than those with BDD [21]. There are still many unanswered questions about the relationship between obesity and BDD, and further research should be undertaken to investigate the association between BDD and obesity. This study is the first study conducted among the general population in Makkah. Additionally, BDDQs were collected in an in-person self-reported fashion using an electronic device to minimize the risk of bias. Limitations Our study has certain limitations. First, self-report bias is probable since the participants were the only reporters for all study variables, which participants' opinions might influence. Second, the present study used a cross-sectional design which precludes the ability to make causal conclusions. Despite these limitations, our study's findings emphasize the presence of remarkable BDD among the population of Makkah, Saudi Arabia, and clarify the aspects that may need further investigation. ## Conclusions In this survey, we found the prevalence of BDD patients among the Makkah population to be $7.1\%$. Most BDD patients were 18-27 years old with no gender predominance. No significant differences were noticed between BDD and non-BDD groups in age, obesity, and gender. Further investigations are needed to demonstrate the characteristics and determinants of BDD. ## References 1. Ahluwalia R, Bhatia NK, Kumar PS, Kaur P. **Body dysmorphic disorder: diagnosis, clinical aspects and treatment strategies**. *Indian J Dent Res* (2017) **28** 193-197. PMID: 28611331 2. Veale D. **Body dysmorphic disorder**. *Postgrad Med J* (2004) **80** 67-71. PMID: 14970291 3. França K, Roccia MG, Castillo D. **Body dysmorphic disorder: history and curiosities**. *Wien Med Wochenschr* (2017) **167** 5-7. PMID: 28220373 4. Eskander N, Limbana T, Khan F. **Psychiatric comorbidities and the risk of suicide in obsessive-compulsive and body dysmorphic disorder**. *Cureus* (2020) **12** 0 5. Veale D, Gledhill LJ, Christodoulou P, Hodsoll J. **Body dysmorphic disorder in different settings: a systematic review and estimated weighted prevalence**. *Body Image* (2016) **18** 168-186. PMID: 27498379 6. Alsaidan MS, Altayar NS, Alshmmari SH, Alshammari MM, Alqahtani FT, Mohajer KA. **The prevalence and determinants of body dysmorphic disorder among young social media users: a cross-sectional study**. *Dermatol Reports* (2020) **12** 8774. PMID: 33408841 7. Alghamdi WA, Subki AH, Khatib HA. **Body dysmorphic disorder symptoms: prevalence and risk factors in an Arab Middle Eastern population**. *Int J Gen Med* (2022) **15** 2905-2912. PMID: 35300125 8. Shaffi Ahamed S, Enani J, Alfaraidi L, Sannari L, Algain R, Alsawah Z, Al Hazmi A. **Prevalence of body dysmorphic disorder and its association with body features in female medical students**. *Iran J Psychiatry Behav Sci* (2016) **10** 0 9. Alomari AA, Makhdoom YM. **Magnitude and determinants of body dysmorphic disorder among female students in Saudi public secondary schools**. *J Taibah Univ Med Sci* (2019) **14** 439-447. PMID: 31728142 10. **OpenEpi: Open Source Epidemiologic Statistics for Public Health**. (2022) 11. Phillips KA. **The broken mirror: Understanding and treating body dysmorphic disorder**. (2005) 12. Koran LM, Abujaoude E, Large MD, Serpe RT. **The prevalence of body dysmorphic disorder in the United States adult population**. *CNS Spectr* (2008) **13** 316-322. PMID: 18408651 13. Rief W, Buhlmann U, Wilhelm S, Borkenhagen A, Brähler E. **The prevalence of body dysmorphic disorder: a population-based survey**. *Psychol Med* (2006) **36** 877-885. PMID: 16515733 14. Dogruk Kacar S, Ozuguz P, Bagcioglu E, Coskun KS, Uzel Tas H, Polat S, Karaca S. **The frequency of body dysmorphic disorder in dermatology and cosmetic dermatology clinics: a study from Turkey**. *Clin Exp Dermatol* (2014) **39** 433-438. PMID: 24758305 15. AlShahwan MA. **Prevalence and characteristics of body dysmorphic disorder in Arab dermatology patients**. *Saudi Med J* (2020) **41** 73-78. PMID: 31915798 16. Alonazi HG, Alharbi M, Alyousif LAM, Alialaswad W, Alharbi JM, Almalki MA, Alrashedee BF. **Prevalence of body dysmorphic disorder in patients attending dermatology clinic in Saudi Arabia/Qassim region**. *J Med Sci Clin Res* (2017) **5** 17. **Population Estimates in the Midyear of 2021**. (2021) 18. Conceição Costa DL, Chagas Assunção M, Arzeno Ferrão Y. **Body dysmorphic disorder in patients with obsessive-compulsive disorder: prevalence and clinical correlates**. *Depress Anxiety* (2012) **29** 966-975. PMID: 22815241 19. Li W, Lai TM, Bohon C. **Anorexia nervosa and body dysmorphic disorder are associated with abnormalities in processing visual information**. *Psychol Med* (2015) **45** 2111-2122. PMID: 25652023 20. Sarwer DB. **Commentary on: Prevalence of body dysmorphic disorder symptoms and body weight concerns in patients seeking abdominoplasty**. *Aesthet Surg J* (2016) **36** 333-334. PMID: 26879298 21. Johnson S, Williamson P, Wade TD. **A systematic review and meta-analysis of cognitive processing deficits associated with body dysmorphic disorder**. *Behav Res Ther* (2018) **107** 83-94. PMID: 29935380
--- title: Balloon dilatation is superior to CO2 laser excision in the treatment of subglottic stenosis authors: - Eleftherios Ntouniadakis - Josefin Sundh - Anders Magnuson - Mathias von Beckerath journal: European Archives of Oto-Rhino-Laryngology year: 2023 pmcid: PMC10038384 doi: 10.1007/s00405-023-07926-w license: CC BY 4.0 --- # Balloon dilatation is superior to CO2 laser excision in the treatment of subglottic stenosis ## Abstract ### Introduction Endoscopic treatment of subglottic stenosis (SGS) is regarded as a safe procedure with rare complications and less morbidity than open surgery yet related with a high risk of recurrence. The abundance of techniques and adjuvant therapies complicates a comparison of the different surgical approaches. The primary aim of this study was to investigate disease recurrence after CO2 laser excisions and balloon dilatation in patients with SGS and to identify potential confounding factors. ### Materials and methods In a tertiary referral center, two cohorts of previously undiagnosed patients treated for SGS were retrospectively reviewed and followed for 3 years. The CO2 laser cohort (CLC) was recruited between 2006 and 2011, and the balloon dilatation cohort (BDC) between 2014 and 2019. Kaplan‒Meier and multivariable Cox regression analyzed time to repeated surgery and estimated hazard ratios (HRs) for different variables. ### Results Nineteen patients were included in the CLC, and 31 in the BDC. The 1-year cumulative recurrence risk was $63.2\%$ for the CLC compared with $12.9\%$ for the BDC (HR 33.0, $95\%$ CI 6.57–166, $p \leq 0.001$), and the 3-year recurrence risk was $73.7\%$ for the CLC compared with $51.6\%$ for the BDC (HR 8.02, $95\%$ CI 2.39–26.9, $p \leq 0.001$). Recurrence was independently associated with overweight (HR 3.45, $95\%$ CI 1.16–10.19, $$p \leq 0.025$$), obesity (HR 7.11, $95\%$ CI 2.19–23.04, $$p \leq 0.001$$), and younger age at diagnosis (HR 8.18, $95\%$ CI 1.43–46.82, $$p \leq 0.018$$). ### Conclusion CO2 laser treatment is associated with an elevated risk for recurrence of SGS compared with balloon dilatation. Other risk factors include overweight, obesity, and a younger age at diagnosis. ## Introduction Subglottic stenosis (SGS) is a rare condition of mucosal scarring, compromising the extrathoracic part of the tracheal airway below the vocal folds. An inflammatory response leading to fibrosis can be triggered by prolonged intubation or tracheostomy, gastroesophageal reflux disease (GERD), or autoimmune conditions, such as vasculitis, sarcoidosis, and relapsing polychondritis [1]. The idiopathic type of SGS is considered to be very rare with an incidence of up to 1:200,000, affecting otherwise healthy perimenopausal females of Caucasian origin [1, 2]. Since SGS presents with common and relatively unspecific symptoms, such as exertional dyspnea, wheezing, chronic cough, or dysphonia, it is frequently misinterpreted as a difficult-to-treat lower airway obstruction resulting in a diagnostic delay of up to 2 years; thus, occasionally manifesting with stridor at rest [3]. Given that recurrence of SGS is regarded as the natural course of the condition, the main treatment goal is to restore durable airway patency without the need for tracheostomy. Open surgical procedures are considered to have the lowest incidence of recurrence; thus, a chance for a permanent treatment. However, these procedures are quite demanding with respect to institutional resources and are associated with increased perioperative and postoperative morbidity in terms of voice and swallowing deterioration [4–6]. Endoscopic techniques are low-risk, voice-sparing procedures that are safe to perform in an outpatient surgery setting; thus, have high patient acceptance [7–9]. However, they are considered to have a significantly higher recurrence rate than open surgery, reported to be approximately $30\%$ within 1 year postoperatively, $50\%$ within 2 years, and $80\%$ within 3 years [10, 11]. Resection of quadrants of the fibrotic tissue with carbon dioxide (CO2) laser and balloon dilatation alone or following cold knife incisions in the stenotic part of the airway have frequently been used, among others, as the endoscopic treatment of SGS [12]. The rarity of SGS combined with the different types and concepts of endoscopic procedures, the divergence of volumes and resources in different institutions, and other unmeasured confounding factors leading to a selection bias make the comparison of these two techniques complicated [11, 13]. The aim of this study was to describe the disease characteristics of the patient cohort treated for SGS in our institution, a tertiary referral center in Sweden, to retrospectively assess whether balloon dilatation is a superior treatment compared to CO2 laser excision of the scar tissue, and to identify potential confounding factors in terms of time to disease recurrence. ## Study subjects Previously undiagnosed adult patients treated primarily for isolated SGS at the Örebro University Hospital, a tertiary academic referral center in Sweden, between 1 January 2006 and 31 December 2019 were identified based on a retrospective chart review of relevant ICD-10 codes, in particular J38.6, J95.5, and J95.8. Patients with SGS caused by malignant tumors, external compression of the airway, or a damaged laryngotracheal cartilaginous framework, and those previously treated for stenosis in the laryngotracheal part of the airway, or with multilevel and distal tracheal strictures, were excluded from the study. ## Surgical techniques From the early 1990s until 2011, patients with SGS had traditionally been treated with endoscopic CO2 laser excision of the scar tissue by every laryngologist in our institution. The procedure was performed under general anesthesia with high-frequency positive pressure ventilation (HFPPV, Monsoon™ ventilation, Acutronic Medical Systems AG, Fabrik im Schiffli, CH-8816, Hirzel, Switzerland) through a steel, laser-resistant catheter. Stenosis was then either vaporized or divided with radial incisions through suspension microlaryngoscopy, depending on the nature of the cicatrix and its craniocaudal length. During 2012, Superimposed High-Frequency Jet Ventilation (SHFJV®, Twinstream™, Mariannengasse 17, 1090 Wien, Austria) was introduced at our institution as a promising method for airway surgery. Concurrently, the absence of a ventilation catheter in the trachea favored the switch of our surgical approach from CO2 laser excisions to balloon dilatation of the stenotic part of the airway, which became the surgical method of choice by the end of that year and has exclusively been used since. Through suspension laryngoscopy under general anesthesia with SHJV®, a balloon catheter is advanced in the airway and gently dilates the stenotic part of the airway, following radial incisions with cold steel if appropriate. An INSPIRA AIR® Balloon Dilatation System (Acclarent, Inc., 33 Technology Drive Irvine, CA 92618, USA) sized 14 mm at 10 atm pressure was used until 2017. It was then substituted by Continuous Radial Expansion™ balloons (Boston Scientific Corporation, 300 Boston Scientific Way, Marlborough, MA 01752, USA) for dilations of up to 15 mm at 8 atm pressure in females and 18 mm at 7 atm pressure in males. The pressure was applied during a short period of apnea aiming for a total of three-to-four dilatation attempts, with a duration between 1 and 2 min or until the patient started desaturating, and up to the maximum possible balloon expansion. ## Data collection This sharp switch in the surgical approach of treating SGS in our department generated the two patient groups we utilized in this study: the cohort of patients treated with CO2 laser excisions (CLC) between 1 January 2006 and 31 December 2011, and the cohort of patients treated with balloon dilatation (BDC) between 1 January 2014 and 31 December 2019. The period from 1 January 2012 to 31 December 2013 was considered an adaptation period for both the surgeons and the anesthesiology staff to acquaint themselves with the novel techniques. The follow-up time for both cohorts was set to 3 years postoperatively. The natural history of the disease after an endoscopic procedure is commonly implicated with a recurrence. In our study, this was defined as significant dyspnea requiring a new surgical treatment when assessed clinically with laryngotracheoscopy by an airway surgeon. Thus, the primary outcome of the study was determined as the time interval from the first surgery until the repeat surgery at recurrence (if it occurred), and the endpoints were a recurrence-free status at 3 years postoperatively or a surgical procedure for recurrence within the follow-up period. Demographic data extracted from the patients’ records included sex, age, time to SGS diagnosis, body mass index (BMI), SGS etiology, smoking history, the presence of diagnosed or self-reported GERD, and tracheal trauma from previous history of tracheostomy at any age or intubation within 2 years prior to the date of diagnosis. Other conditions registered from the patients’ records were diabetes, conditions of the lower airway or the lungs, and cardiovascular comorbidities, including ischemic heart disease, heart failure, arrhythmia, or cerebrovascular condition. ## Statistical analysis A power calculation was made prior to performing the statistical analysis. A total of 72 patients were required to have an $80\%$ chance of detecting a reduction in the recurrence rate from $80\%$ in the CLC group to $50\%$ in the BDC at 3 years postoperatively, which was significant at the $5\%$ level [10, 11]. Continuous variables were analyzed by the Mann‒Whitney U test and are presented as medians and the 25th-to-75th percentiles, whereas categorical variables were analyzed by the Chi-square test or Fisher’s exact test when appropriate and are presented as numbers and percentages. We visualized time to recurrence with the Kaplan‒Meier (KM) method and presented it as cumulative recurrence risk (1-KM). All patients were followed up after the initial operation to the first reoperation or censored at 3 years. Cox proportional hazard models were applied, estimating hazard ratios (HRs) with $95\%$ confidence intervals (CIs) to compare disease recurrence for the two treatment groups. Models were both crude and adjusted for sex, age (categorized as 18–39, 40–49, 50–59, and ≥ 60 years), cause of SGS, smoking, positive intubation history within 2 years prior to the initial SGS diagnosis, BMI according to the World Health Organization (WHO) classification (< 25 kg/m2 defined as normal weight, 25–29.9 kg/m2 defined as overweight, and ≥ 30 kg/m2 defined as obese), presence of self-reported or diagnosed GERD, and diabetes. Confounders were chosen prior to data analysis and in accordance with the previous studies [5, 11, 12]. The proportional hazard assumption was tested by the phtest command in STATA. A p value less than 0.05 was considered statistically significant. IBM® SPSS® Statistics, version 27 (IBM Corp. Armonk, NY, USA) and STATA release 17 (StataCorp. 2021. College Station, TX: StataCorp LLC.) were used for the statistical analysis. ## Ethics This human study was performed in accordance with the Declaration of Helsinki Guidelines and was approved by the Ethics Review Board in Uppsala (diary number $\frac{2016}{193}$) and the Swedish Ethical Review Authority (diary numbers 2020-05509 and 2022-02708-02). All adult participants provided written informed consent to participate. ## Results The study population consisted of 19 patients in the CLC and 31 patients in the BDC. We excluded 16 patients in total: Eight of them were previously treated for SGS outside our inclusion period, 3 subjects were found to have multilevel stenosis engaging other parts of the airway (2 with glottic, 1 with bronchial stenosis), 3 cases had a damaged cricotracheal cartilaginous framework and were not appropriate for endoscopic treatment, and in 2 cases treated with CO2 laser, we could not establish contact and receive an informed consent. Both groups had a similar mean time to diagnosis, yet the mean age at diagnosis was significantly lower in the BLC. The most predominant SGS cause was the idiopathic type, followed by trauma, in both cohorts, and none of the patients were tracheostomized at any age. Table 1 lists the demographic data and comorbidities of the study population at baseline. Only one patient presented with ischemic heart disease. None of them were diagnosed with conditions of the lower airway or the lungs, yet 7 patients had been prescribed steroid inhalers by general practitioners suspecting asthma prior to the diagnosis of SGS. No readmissions or other complications were observed postoperatively for either surgical technique. Because of the relatively small sample size, the SGS cause variable was converted into a binary variable to consolidate the regression analysis. Table 1Demographic data of the study populationCO2 laserBalloon dilatationp ValueN1931Age at diagnosis—median (25th–70th percentile)58.0 (50.8–63.5)47.0 (38.0–63.0)0.033 18–30 years old—n (%)0 [0]9 [29]0.014 40–49 years old—n (%)2 (10.5)7 (22.6) 50–59 years old—n (%)7 (36.8)5 (16.1) ≥ 60 years old—n (%)10 (52.6)10 (32.3) Years undiagnosed—median (25th–70th percentile)1.8 (1.0–3.0)2.0 (1.0–5.0)0.59Sex—n (%) Male04 (12.9)0.28 Female19 [100]27 (87.1)Cause—n (%) Idiopathic12 (63.2)25 (80.6)0.49 Traumatic3 (15.6)3 (9.7) GPA1 (5.3)2 (6.5) IgG4-mediated disease1 (5.3)1 (3.2) Rheumatoid arthritis1 (5.3)0 Other vasculitis1 (5.3)0BMI—n (%) Underweight/normal weight (< 25 kg/m2)6 (31.6)13 (41.9)0.36 Overweight (25–29.9 kg/m2)8 (42.1)7 (22.6) Obese (≥ 30 kg/m2)5 (26.3)11 (25.5)Smoking history—n (%) Never smoker16 (84.2)27 (87.1)0.99 Ever smoker3 (12.9)4 (12.9)Intubation history within 2 years prior to initial diagnosis—n (%) Positive4 (21.1)5 (16.1)0.72 Negative15 (78.9)26 (83.9)GERD—n (%) Positive6 (31.6)7 (22.6)0.52 Negative13 (68.4)24 (77.4)Diabetes—n (%) Positive2 (10.5)4 (12.9)0.99 Negative17 (89.5)27 (81.1)Hypertension—n (%) Positive6 (31.6)5 (16.1)0.29 Negative13 (68.4)26 (83.9)Bold values are statistically significant ($p \leq 0.05$)SD standard deviation, IQR interquartile range, BMI body mass index, GPA granulomatosis with polyangiitis, GERD gastroesophageal reflux disease A total of 30 events were observed in the study population, 14 in the CLC and 16 in the BDC. The 3-year recurrence risk was $73.7\%$ for the CLC (14 of 19 study subjects at risk) compared with $51.6\%$ for the BDC (16 of 31 patients at risk, Fig. 1). As seen in the KM plot, the association between study groups was different during the 3-year follow-up, with a tendency for disease recurrence within the first year in the CLC and after the first year in the BDC. Since the proportional hazard assumption was violated, we modeled the group variable interact with the follow-up time (0–1 vs. 1–3 years) as an indicator variable to estimate time-dependent analysis [14, 15]. In the first year, the cumulative risk of recurrence was $63.2\%$ for the CLC compared to $12.9\%$ for the BDC, with a crude HR of 7.55 ($95\%$ CI 2.42–25.6, $p \leq 0.001$) and an adjusted HR of 33.0 ($95\%$ CI 6.57–166, $p \leq 0.001$). Among patients without a recurrence during the first year, the follow-up period from 1 to 3 years showed a crude HR of 0.55 ($95\%$ CI 0.12–2.46) and an adjusted HR of 1.85 ($95\%$ CI 0.32–10.8).Fig. 1Kaplan‒Meier curves showing the cumulative risk for disease recurrence in patients undergoing CO2 laser (blue line) compared to balloon dilatation (red line) treatment for SGS The adjusted model showed an elevated risk of recurrence in overweight (adjusted HR 3.44, $95\%$ CI 1.16–10.2, $$p \leq 0.025$$) and obese patients (adjusted HR 7.11, $95\%$ CI 2.19–23.0, $$p \leq 0.001$$) compared to normal-weight patients. The group of patients aged 40 years and below was also found to have a higher risk of recurrence (adjusted HR 8.18, $95\%$ CI 1.43–46.8, $$p \leq 0.018$$) than the group of patients aged 50–59 years (Table 2).Table 2Time to recurrence evaluated with Cox regression to compare the CO2 laser treatment and balloon dilatation groupsNOutcomesCrude ($$n = 50$$)Adjusted ($$n = 50$$)n (%)HR ($95\%$ CI)pHR ($95\%$ CI)pTreatmenta Balloon dilatation3116 (51.6)RefRef CO2 laser1914 (73.7)2.48 (1.20–5.10)0.0148.02 (2.39–26.9) < 0.0010–1 Year follow-upa Balloon dilatationRefRef CO2 laser7.55 (2.42–25.6) < 0.00133.0 (6.57–166) < 0.0011–3 year follow-upa Balloon dilatationRefRef CO2 laser0.55 (0.12–2.46)0.441.85 (0.32–10.8)0.49Gender Male41 (25.0)RefRef Female4629 (63.0)3.14 (0.43–23.1)0.263.28 (0.36–29.7)0.29BMI Normal/underweight197 (36.8)RefRef Overweight1511 (73.3)2.73 (1.06–7.06)0.0383.44 (1.16–10.2)0.025 Obese1612 (75.0)3.52 (1.37–9.04)0.00897.11 (2.19–23.0)0.0011Smoking history Never smoker4325 (58.1)RefRef Ever smoker75 (71.4)1.46 (0.56–3.82)0.441.75 (0.44–6.94)0.43Intubation history Negative4124 (58.5)RefRef Positive96 (66.7)1.34 (0.55–3.29)0.521.83 (0.44–7.60)0.40GERD Negative3721 (56.8)RefRef Positive139 (69.2)1.48 (0.67–3.23)0.331.33 (0.33–5.36)0.68Cause Nonidiopathic139 (69.2)RefRef Idiopathic3721 (56.8)0.60 (0.27–1.32)0.201.17 (0.22–6.27)0.85Age < 4097 (77.8)1.30 (0.45–3.72)0.638.18 (1.43–46.8)0.018 40–4997 (77.8)1.46 (0.51–4.20)0.482.28 (0.54–9.65)0.26 50–59127 (58.3)RefRef ≥ 60209 (45.0)0.71 (0.26–1.91)0.500.73 (0.24–2.20)0.58Diabetes Negative4425 (56.8)RefRef Positive65 (83.3)2.09 (0.79–5.52)0.142.26 (0.55–9.23)0.26Bold values are statistically significant ($p \leq 0.05$)Patients are followed from the initial surgery up to 3 yearsHR hazard ratio, CI confidence interval, Ref reference, BMI body mass index, GERD gastroesophageal reflux diseaseaTreatment showed a nonproportional hazard and was evaluated as a time-dependent association by interaction with follow-up time (0–1 vs. 1–3 years ## Discussion The primary findings of our study indicate a superiority of balloon dilatation compared to CO2 laser excisions in short-term disease recurrence, particularly within the first year postoperatively. Furthermore, patients who were overweight or obese or had a disease presentation at a younger age were independently found to have a statistically significant increased risk of SGS recurrence. The diversity of surgical approaches in the endoscopic treatment of SGS, such as different dilation instruments (e.g., rigid endoscopes or inflatable balloons), scar excision instruments (e.g., cold steel or CO2 laser), and adjuvant therapies (e.g., mitomycin C or steroids), complicates the comparison of these procedures. The homogeneity of the two surgical techniques used in our study population facilitates, in essence, the comparison of the thermal effect of a CO2 laser excision with the cold tissue expansion of balloon dilatation, minimizing the confounding impact of different endoscopic treatments. This is reflected by the $51.6\%$ risk of recurrence at 3 years for our BDC group, which is consistent with other studies investigating the outcomes of balloon dilatation without CO2 laser-assisted excisions [5, 6, 13]. There is indisputable evidence that open surgical techniques prevail regarding the durability of maintaining a patent airway without the need for tracheostomy or repeated surgery, eliminating dyspnea. However, they are associated with substantial perioperative risks (e.g., anastomotic complications or temporary tracheostomy). Postoperative morbidity, including poor voice outcomes or even an eventual delayed disease recurrence of up to $30\%$ between 5 and 10 years postoperatively, cannot be overlooked [4, 6, 11, 16–18]. Thus, endoscopic treatment still has an important role in the treatment of SGS with its excellent convalescence and despite the higher recurrence rate when compared to open surgical procedures [5, 19]. Our results encourage the use of balloon dilatation instead of CO2 laser excisions considering the longer time to recurrence, since this is ultimately considered the natural course of the condition. We showed that there is a particular propensity for recurrence in the CLC during the first year postoperatively, whereas stenoses treated with balloon dilatation tend to recur during the second year of follow-up. Interestingly, there seems to be a trend of stability in the relapsing manner of the condition in both groups within the third year ($73.7\%$ for both the 2-year and 3-year recurrence risk for the CLC compared to $42.9\%$ and $51.6\%$, respectively, for the BDC; Fig. 1). These findings could be considered in the context of preoperative patient counseling and the individual selection of an endoscopic treatment. The vigilant perspective of an exceptional increase in the incidence of laryngotracheal stenosis during the COVID-19 outbreak [20] led to a prioritized handling of patients with airway problems. Therefore, the treatment of patients with airway obstruction, in particular SGS recurrence, was never delayed. Although stenoses related to iatrogenic trauma are regarded to be more prevalent [1, 21, 22], the profile of our study population matches the idiopathic type of the condition. Previously published studies have discussed potential environmental or hereditary factors related to the high prevalence of idiopathic SGS [12, 23–26]. However, this finding might also reflect the anticipative policy in our institution of striving for either tracheostomy in patients with expected prolonged intubation or prompt decannulation combined with noninvasive ventilation to minimize mucosal trauma and scarring predisposing for traumatic SGS. Furthermore, the idiopathic type consists predominantly of otherwise healthy, middle-aged, nonsmoking females experiencing symptoms of dyspnea for approximately 2 years before given the correct diagnosis of SGS [11, 13, 27]. An elevated BMI is also identified as a factor associated with disease recurrence [17, 28, 29]. This view is supported by our findings with a relatively low incidence of comorbidities, and HRs of 3.5 and 7.1 for overweight and obese patients, respectively, compared to normal or underweight patients. The large CIs observed apparently depend on our study’s small sample size. The theory of a hormonal imbalance in perimenopausal females has been previously studied to explain the onset of idiopathic SGS in that age group. Estrogen receptors are thought to be expressed either unproportionally compared to progesterone receptors and more extensively in females with idiopathic SGS compared to patients with a nonidiopathic type of SGS [30, 31]. Moreover, there is evidence of an age-related elevation in peripheral estrogen formation occurring in adipose tissue [32]. Thus, being overweight or obese could potentially affect and complicate the hormonal equilibrium in premenopausal females contributing to the development of idiopathic SGS before menopause. Pregnancy-associated idiopathic SGS, although a rare entity, further supports the hypothesis of a hormonal origin or blossoming of symptoms in an established and occult stenosis due to the physiological vascular and respiratory changes of pregnancy [33]. These are concepts requiring further studies that could potentially explain the idiopathic prevalence in our cohort and the higher risk of recurrence in the fertile age group (18–39 years old) than in the peri- or postmenopausal age groups. The major strength of our study is the segmentation of the inclusion period into nonoverlapping timeframes where the physicians in our department performed only one of two interventions, including a distinct learning period in between. In this manner, we sought to minimize performance bias, since nonrandom intervention assignment is a well-described disadvantage in all retrospective studies. Furthermore, only previously untreated patients with isolated stenosis of the subglottic region were included to eliminate the potential confounding effect of scar transformation by previous surgery and potential selection bias. Due to the relapsing nature of SGS and in conformity with results from previous reports [6, 11, 13], the follow-up time was set to 3 years for both cohorts, ensuring an equal and homogenous assessment of the survival analysis. The absence of an objective and subjective severity grading of stenosis both before the initial intervention and at the clinical assessment upon recurrence is the main limitation of our study. An anatomical classification made by the surgeon was absent from the entire CLC, as neither the Cotton–Myer nor McCaffrey system had been used by the physicians in our department at that time. Although these scales have been widely proposed to assess SGS disease severity and prognosis, the former does not address the length and complexity of the lesion, and the latter does not justify the cross-sectional degree of stenosis [1]. Song et al. [ 34] showed the poor interrater reliability of a visual estimation in Cotton–Myer grading among physicians and further discussed the difficulty in identifying cricoid cartilage when assessing stenosis length endoscopically. Moreover, neither of the two systems correlates with functional airway assessment with spirometry, as shown by several studies [35–37]. Since there is evidence that several measurements of pulmonary function could be used in the diagnosis and postoperative monitoring of patients with SGS [34, 35], the lack of a preoperative functional evaluation with spirometry in our study population is considered another shortcoming of our study. Moreover, it would be interesting to quantify patient-experienced dyspnea using questionnaires specifically developed for upper airway obstruction [38, 39]. However, these data were missing from the entire CLC, since a routine assessment with spirometry and the validated Swedish version of the Dyspnea Index was not introduced as part of the preoperative workup in our department until 2016. Our study indicates that balloon dilatation is superior to CO2 laser treatment in SGS patients, which is in conformity with several other retrospective studies [5, 6, 13]. Future prospective multicenter randomized control trials are recommended to achieve a sufficient sample size to further evaluate this evidence and examine the effect of adjuvant therapies and the associations of different patient-specific confounders predisposing patients to SGS recurrence. ## Conclusion Endoscopic treatment for SGS with balloon dilatation is more favorable regarding short-term SGS recurrence compared to CO2 laser treatment, and patients with a younger age of SGS onset, overweight, or obesity showed a higher risk for SGS recurrence. ## References 1. Aravena C, Almeida FA, Mukhopadhyay S, Ghosh S, Lorenz RR, Murthy SC. **Idiopathic subglottic stenosis: a review**. *J Thorac Dis* (2020) **12** 1100-1111. DOI: 10.21037/jtd.2019.11.43 2. Aarnæs MT, Sandvik L, Brøndbo K. **Idiopathic subglottic stenosis: an epidemiological single-center study**. *Eur Arch Otorhinolaryngol* (2017) **274** 2225-2228. DOI: 10.1007/s00405-017-4512-0 3. Gnagi SH, Howard BE, Anderson C, Lott DG. **Idiopathic Subglottic and tracheal stenosis: a survey of the patient experience**. *Ann Otol Rhinol Laryngol* (2015) **124** 734-739. DOI: 10.1177/0003489415582255 4. Clunie GM, Roe JWG, Alexander C, Sandhu G, McGregor A. **Voice and swallowing outcomes following airway reconstruction in adults: a systematic review**. *Laryngoscope* (2021) **131** 146-157. DOI: 10.1002/lary.28494 5. Dwyer CD, Qiabi M, Fortin D, Inculet RI, Nichols AC, MacNeil SD. **Idiopathic subglottic stenosis: an institutional review of outcomes with a multimodality surgical approach**. *Otolaryngol Head Neck Surg Off J Am Acad Otolaryngol Head Neck Surg* (2021) **164** 1068-1076. DOI: 10.1177/0194599820966978 6. Gelbard A, Anderson C, Berry LD, Amin MR, Benninger MS, Blumin JH. **Comparative treatment outcomes for patients with idiopathic subglottic stenosis**. *JAMA Otolaryngol Head Neck Surg* (2020) **146** 20-29. DOI: 10.1001/jamaoto.2019.3022 7. Naunheim MR, Naunheim ML, Rathi VK, Franco RA, Shrime MG, Song PC. **Patient preferences in subglottic stenosis treatment: a discrete choice experiment**. *Otolaryngol Head Neck Surg Off J Am Acad Otolaryngol Head Neck Surg.* (2018) **158** 520-526. DOI: 10.1177/0194599817742851 8. Allen CT, Lee CJ, Meyer TK, Hillel AD, Merati AL. **Risk stratification in endoscopic airway surgery: is inpatient observation necessary?**. *Am J Otolaryngol* (2014) **35** 747-752. DOI: 10.1016/j.amjoto.2014.06.014 9. Hoffman MR, Brand WT, Dailey SH. **Effects of balloon dilation for idiopathic laryngotracheal stenosis on voice production**. *Ann Otol Rhinol Laryngol* (2016) **125** 12-19. DOI: 10.1177/0003489415595425 10. Hseu AF, Benninger MS, Haffey TM, Lorenz R. **Subglottic stenosis: a ten-year review of treatment outcomes**. *Laryngoscope* (2014) **124** 736-741. DOI: 10.1002/lary.24410 11. Gelbard A, Donovan DT, Ongkasuwan J, Nouraei SA, Sandhu G, Benninger MS. **Disease homogeneity and treatment heterogeneity in idiopathic subglottic stenosis**. *Laryngoscope* (2016) **126** 1390-1396. DOI: 10.1002/lary.25708 12. Feinstein AJ, Goel A, Raghavan G, Long J, Chhetri DK, Berke GS. **Endoscopic management of subglottic stenosis**. *JAMA Otolaryngol Head Neck Surg* (2017) **143** 500-505. DOI: 10.1001/jamaoto.2016.4131 13. Lavrysen E, Hens G, Delaere P, Meulemans J. **Endoscopic treatment of idiopathic subglottic stenosis: a systematic review**. *Front Surg* (2019) **6** 75. DOI: 10.3389/fsurg.2019.00075 14. Dekker FW, de Mutsert R, van Dijk PC, Zoccali C, Jager KJ. **Survival analysis: time-dependent effects and time-varying risk factors**. *Kidney Int* (2008) **74** 994-997. DOI: 10.1038/ki.2008.328 15. Jachno KM, Heritier S, Woods RL, Mahady S, Chan A, Tonkin A. **Examining evidence of time-dependent treatment effects: an illustration using regression methods**. *Trials* (2022) **23** 857. DOI: 10.1186/s13063-022-06803-x 16. Yamamoto K, Kojima F, Tomiyama KI, Nakamura T, Hayashino Y. **Meta-analysis of therapeutic procedures for acquired subglottic stenosis in adults**. *Ann Thorac Surg* (2011) **91** 1747-1753. DOI: 10.1016/j.athoracsur.2011.02.071 17. Carpenter PS, Pierce JL, Smith ME. **Outcomes after cricotracheal resection for idiopathic subglottic stenosis**. *Laryngoscope* (2018) **128** 2268-2272. DOI: 10.1002/lary.27263 18. Menapace DC, Modest MC, Ekbom DC, Moore EJ, Edell ES, Kasperbauer JL. **Idiopathic subglottic stenosis: long-term outcomes of open surgical techniques**. *Otolaryngol Head Neck Surg Off J Am Acad Otolaryngol Head Neck Surg.* (2017) **156** 906-911. DOI: 10.1177/0194599817691955 19. Richards BA, Xie KZ, Bowen AJ, Aden A, Wiedermann J, Rutt AL. **Complications following laser wedge excision for idiopathic subglottic stenosis**. *Am J Otolaryngol* (2022) **43** 103629. DOI: 10.1016/j.amjoto.2022.103629 20. Piazza C, Filauro M, Dikkers FG, Nouraei SAR, Sandu K, Sittel C. **Long-term intubation and high rate of tracheostomy in COVID-19 patients might determine an unprecedented increase of airway stenoses: a call to action from the European Laryngological Society**. *Eur Arch Otorhinolaryngol* (2021) **278** 1-7. DOI: 10.1007/s00405-020-06112-6 21. Gelbard A, Francis DO, Sandulache VC, Simmons JC, Donovan DT, Ongkasuwan J. **Causes and consequences of adult laryngotracheal stenosis**. *Laryngoscope* (2015) **125** 1137-1143. DOI: 10.1002/lary.24956 22. Fiz I, Monnier P, Koelmel JC, Di Dio D, Fiz F, Missale F. **Multicentric study applying the european laryngological society classification of benign laryngotracheal stenosis in adults treated by tracheal or cricotracheal resection and anastomosis**. *Laryngoscope* (2020) **130** 1640-1645. DOI: 10.1002/lary.28274 23. Woliansky J, Paddle P, Phyland D. **Laryngotracheal stenosis management: a 16-year experience**. *Ear Nose Throat J* (2021) **100** 360-367. DOI: 10.1177/0145561319873593 24. Chan RK, Ahrens B, MacEachern P, Bosch JD, Randall DR. **Prevalence and incidence of idiopathic subglottic stenosis in southern and central Alberta: a retrospective cohort study**. *J Otolaryngol Head Neck Surg* (2021) **50** 64. DOI: 10.1186/s40463-021-00544-8 25. Anis MM, Zhao Z, Khurana J, Krynetskiy E, Soliman AM. **Translational genomics of acquired laryngotracheal stenosis**. *Laryngoscope* (2014) **124** E175-E179. DOI: 10.1002/lary.24382 26. Drake VE, Gelbard A, Sobriera N, Wohler E, Berry LL, Hussain LL. **Familial aggregation in idiopathic subglottic stenosis**. *Otolaryngol Head Neck Surg Off J Am Acad Otolaryngol Head Neck Surg.* (2020) **163** 1011-1017. DOI: 10.1177/0194599820935402 27. Berges AJ, Lina IA, Chen L, Ospino R, Davis R, Hillel AT. **Delayed diagnosis of idiopathic subglottic stenosis**. *Laryngoscope* (2022) **132** 413-418. DOI: 10.1002/lary.29783 28. Menapace DC, Ekbom DC, Larson DP, Lalich IJ, Edell ES, Kasperbauer JL. **Evaluating the association of clinical factors with symptomatic recurrence of idiopathic subglottic stenosis**. *JAMA Otolaryngol Head Neck Surg* (2019) **145** 524-529. DOI: 10.1001/jamaoto.2019.0707 29. Dang S, Shinn JR, Campbell BR, Garrett G, Wootten C, Gelbard A. **The impact of social determinants of health on laryngotracheal stenosis development and outcomes**. *Laryngoscope* (2020) **130** 1000-1006. DOI: 10.1002/lary.28208 30. Fiz I, Bittar Z, Piazza C, Koelmel JC, Gatto F, Ferone D. **Hormone receptors analysis in idiopathic progressive subglottic stenosis**. *Laryngoscope* (2018) **128** E72-E77. DOI: 10.1002/lary.26931 31. Fiz I, Antonopoulos W, Kölmel JC, Rüller K, Fiz F, Piazza C. **Hormone pathway comparison in non-idiopathic and idiopathic progressive subglottic stenosis**. *Eur Arch Otorhinolaryngol* (2022) **280** 775-780. DOI: 10.1007/s00405-022-07615-0 32. Nelson LR, Bulun SE. **Estrogen production and action**. *J Am Acad Dermatol* (2001) **45** S116-S124. DOI: 10.1067/mjd.2001.117432 33. Tapias LF, Rogan TJ, Wright CD, Mathisen DJ. **Pregnancy-associated idiopathic laryngotracheal stenosis: presentation, management and results of surgical treatment**. *Eur J Cardiothorac Surg* (2020) **59** 122-129. DOI: 10.1093/ejcts/ezaa296 34. Song SA, Santeerapharp A, Choksawad K, Franco RA. **Reliability of peak expiratory flow percentage compared to endoscopic grading in subglottic stenosis**. *Laryngoscope Investig Otolaryngol* (2020) **5** 1133-1139. DOI: 10.1002/lio2.492 35. Ntouniadakis E, Sundh J, von Beckerath M. **Monitoring adult subglottic stenosis with spirometry and dyspnea index: a novel approach**. *Otolaryngol Head Neck Surg Off J Am Acad Otolaryngol Head Neck Surg.* (2022) **167** 517-523. DOI: 10.1177/01945998211060817 36. Crosby T, McWhorter A, McDaniel L, Kunduk M, Adkins L. **Predicting need for surgery in recurrent laryngotracheal stenosis using changes in spirometry**. *Laryngoscope* (2020) **131** 2199-2203. DOI: 10.1002/lary.29239 37. Tie K, Buckmire RA, Shah RN. **The role of spirometry and dyspnea index in the management of subglottic stenosis**. *Laryngoscope* (2020) **130** 2760-2766. DOI: 10.1002/lary.28337 38. Noud M, Hovis K, Gelbard A, Sathe NA, Penson DF, Feurer ID. **Patient-reported outcome measures in upper airway-related dyspnea: a systematic review**. *JAMA Otolaryngol Head Neck Surg* (2017) **143** 824-831. DOI: 10.1001/jamaoto.2017.0348 39. Ntouniadakis E, Brus O, von Beckerath M. **Dyspnea Index: An upper airway obstruction instrument; translation and validation in Swedish**. *Clin Otolaryngol Off J ENT UK Off J Netherlands Soc Otorhinolaryngol Cervicofacial Surg* (2021) **46** 380-387
--- title: 'Indicators of Cardiometabolic Function in Pregnancy and Long-Term Risk of COVID-19: Population-Based Cohort Study' journal: Cureus year: 2023 pmcid: PMC10038685 doi: 10.7759/cureus.35325 license: CC BY 3.0 --- # Indicators of Cardiometabolic Function in Pregnancy and Long-Term Risk of COVID-19: Population-Based Cohort Study ## Abstract Background: Pregnancy increases a woman’s susceptibility to severe COVID-19, especially those with metabolic dysfunction. It is unknown if markers of metabolic dysfunction commonly assessed around pregnancy are associated with COVID-19 illness after pregnancy. Aim: The aim of this study is to evaluate the indicators of metabolic dysfunction collected in pregnancy and the future risk of severe COVID-19 after pregnancy. Methods: This population-based cohort study was completed in all of Ontario, comprising 417,713 women aged 15-49 years with a hospital birth between April 2007 and March 2018. The main exposure was each 1-kg/m2 higher body mass index (BMI), 1-mmol/L higher glucose concentration at the 50-g glucose challenge test, and one-week earlier gestational week at delivery. The main outcome was severe COVID-19 illness or death, from the start of the pandemic period on March 1, 2020, till December 31, 2021. Results: The adjusted hazard ratio (aHR) of COVID-19 illness increased per 1-kg/m2 higher BMI (1.05, $95\%$ CI 1.04-1.06), per 1-mmol/L higher serum glucose concentration (1.16, $95\%$ CI 1.10-1.22), and for each one-week earlier gestational week at delivery (1.12, $95\%$ CI 1.03-1.23). Relative to women with no dichotomized risk factors, the aHR for severe COVID-19 was 1.60 ($95\%$ CI 1.28-2.01) with one factor, 3.34 ($95\%$ CI 2.51-4.44) with two factors, and 4.52 ($95\%$ CI 2.11-9.67) with three factors. Conclusions: The number, and degree, of standard metabolic indicators measured around pregnancy predict the future risk of severe COVID-19 remotely after that pregnancy. ## Introduction Gestational diabetes mellitus (DM), higher body mass index (BMI) in pregnancy, and preterm delivery partly predict the onset of metabolic syndrome (MetSyn) in women years after pregnancy [1,2]. After the emergence of the SARS-COV-2 pandemic around March 2020, it became evident that pregnant women were prone to severe COVID-19 illness and adverse perinatal outcomes, such as preterm labor and preterm birth [3]. The risk of adversity was most pronounced in women with a high BMI or gestational DM within the index pregnancy [4]. What is not known, however, is whether the aforementioned factors, when measured in pregnancy, are associated with the onset of COVID-19 illness well after that pregnancy has ended. The availability of three standardized continuous measures, namely, pre-pregnancy BMI, glucose concentration at the time of gestational DM screening, and gestational week at delivery, enabled us to address this question. This study evaluated the future risk of severe COVID-19 in relation to prior pregnancy BMI, serum glucose concentration, and gestational age at birth - both as continuous and dichotomized risk factors. ## Materials and methods This study considered all women aged 15-49 years with a hospital live birth or stillbirth in Ontario, Canada, from April 1, 2007, to March 31, 2018. Ontario has universal healthcare under the Ontario Health Insurance Plan (OHIP), and most women undergo gestational DM screening with a 50-g glucose challenge test (GCT) at about 27 weeks of gestation (Appendix table). All women with known pre-pregnancy BMI and a GCT result in the index pregnancy were included. Those with pre-pregnancy DM and women not alive or eligible for OHIP on March 1, 2020 (the start of the SARS-CoV-2 pandemic) were excluded. If a woman had more than one eligible delivery during the study period, then her latest birth was considered. The last birth was considered up to March 31, 2018, to minimize the chance that a woman was pregnant, or had recently given birth, at the onset of the COVID-19 pandemic and because BMI data were only available up to that date. Births and outpatient and inpatient encounters were captured in province-wide administrative datasets that were linked using unique encoded identifiers and analyzed at ICES (https://datadictionary.ices.on.ca/Applications/DataDictionary/Default.aspx), as described by Catov et al. and Li et at. [ 1,5] and in Appendix table. BMI was identified in the Better Outcomes Registry and Network (BORN) Information System and Niday Perinatal Databases with a 50-g GCT in the Ontario Laboratory Information System [6]. All SARS-CoV-2 vaccinations in Ontario are also captured at ICES (https://data.ontario.ca/dataset/covid-19-vaccine-data-in-ontario). More than $95\%$ of pregnancies have an ultrasound enabling accurate pregnancy dating [7]. The main study outcome was severe COVID-19 illness arising from the start of the pandemic period of March 1, 2020 (i.e., time zero), to December 31, 2021 (the latest complete data). Severe COVID-19 illness was based on a positive SARS-CoV-2 PCR test within seven days preceding, or up to three days after, hospitalization or death [6]. Analyses Time-to-event analyses started at time zero. Separate Cox proportional hazard models generated hazard ratios (HRs) and $95\%$ CI for the respective relationships between the study outcomes and each 1-kg/m2 incremental higher BMI; 1-mmol/L higher GCT glucose concentration, and one-week earlier gestational week at delivery, from ≥37 weeks (term birth) declining weekly down to ≤24 weeks. Next, each study outcome was assessed about having 0 (referent), 1, 2, or 3 dichotomized risk factors (i.e., BMI ≥ 30 kg/m2, positive 50-g GCT ≥ 7.8 mmol/L, and/or preterm birth < 37 weeks). HRs were adjusted for age, rural residence, and area income quintile at the index birth; a woman’s age at time zero; chronic hypertension in the index pregnancy or up to two years before that pregnancy; and the time-varying SARS-CoV-2 first vaccination date. Censoring was based on death occurring prior to either outcome, loss of OHIP eligibility, or arrival at the end of the study period of December 31, 2021. A further analysis was censored at the start of any subsequent pregnancy during the SARS-CoV-2 pandemic. Sample size estimations were not performed as the current study used a fixed population-based data sample. Ethics approval Datasets were linked using unique encoded identifiers and analyzed at ICES. The use of data in this project was authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require the approval of a research ethics board. ## Results A total of 417,713 women were included, followed by a median (IQR) of 1.8 (1.8, 1.8) years after time zero, totaling 756,471 person-years of follow-up. Their mean (SD) age at the index delivery was 31.1 (5.2) years, and $83.4\%$ received at least one SARS-CoV-2 vaccination during the follow-up period (Table 1). **Table 1** | Characteristics | Measures | | --- | --- | | In the index delivery | In the index delivery | | Mean (SD) age, y | 31.1 (5.2) | | Income quintile (Q) | | | Q1 (lowest) | 86,965 (20.8) | | Q2 | 82,607 (19.8) | | Q3 | 87,443 (20.9) | | Q4 | 90,081 (21.6) | | Q5 (highest) | 69,744 (16.7) | | Missing | 873 (0.2) | | Rural residence | 36,000 (8.6) | | Unknown residence | 441 (0.1) | | Median (IQR) parity | 1 (0-1) | | Multifetal pregnancy | 7202 (1.7) | | Chronic hypertension in the index pregnancy, or up to 2 years before | 19,591 (4.7) | | Stillbirth | 831 (0.2) | | Measures in the index pregnancy | Measures in the index pregnancy | | Mean (SD) pre-pregnancy body mass index, kg/m2 | 25.5 (6.2) | | Mean (SD) glucose concentration at the 50-g glucose challenge test, mmol/La | 6.6 (1.8) | | Mean (SD) gestational age at delivery, weeks | 38.9 (1.6) | | Preterm birth < 37 weeks of gestation | 26,188 (6.3) | | From the start of the COVID-19 pandemic on March 1, 2020 (time zero), to the end of the study follow-up on December 31, 2021 | From the start of the COVID-19 pandemic on March 1, 2020 (time zero), to the end of the study follow-up on December 31, 2021 | | Mean (SD) age, years | 35.8 (5.6) | | Any subsequent pregnancy | 64,141 (15.4) | | Median (IQR) no. of years of follow-up, from time zero | 1.8 (1.8-1.8) | | Received at least one SARS-CoV-2 vaccination | 348,261 (83.4) | The adjusted HR of severe COVID-19 increased per 1-kg/m2 higher BMI (1.05, $95\%$ CI 1.04-1.06), per 1-mmol/L higher serum glucose concentration (1.16, $95\%$ CI 1.10-1.22), and for each one-week earlier gestational week at delivery (1.12, $95\%$ CI 1.03-1.23) (Table 2). **Table 2** | Risk factor assessed | Total no. of person-years of follow-up | No. of outcomesa (rate per 1000 person-years) | Unadjusted hazard ratio (95% CI) | Adjusted hazard ratio (95% CI)b | Adjusted hazard ratio (95% CI)b and further censoring on any future pregnancyb,c | | --- | --- | --- | --- | --- | --- | | Per 1-unit change in each risk factor | Per 1-unit change in each risk factor | Per 1-unit change in each risk factor | Per 1-unit change in each risk factor | Per 1-unit change in each risk factor | Per 1-unit change in each risk factor | | Each 1-kg/m2 higher body mass index | 756471 | - | 1.06 (1.04 to 1.07) | 1.05 (1.04 to 1.06) | 1.06 (1.04 to 1.07) | | Each 1-mmol/L higher serum glucose concentration | 756471 | - | 1.17 (1.12 to 1.23) | 1.16 (1.10 to 1.22) | 1.18 (1.12 to 1.25) | | Each 1-week earlier gestational age at delivery | 756471 | - | 1.14 (1.05 to 1.24) | 1.12 (1.03 to 1.23) | 1.07 (0.95 to 1.20) | | Number of dichotomized risk factors (number of women exposed) | Number of dichotomized risk factors (number of women exposed) | Number of dichotomized risk factors (number of women exposed) | Number of dichotomized risk factors (number of women exposed) | Number of dichotomized risk factors (number of women exposed) | Number of dichotomized risk factors (number of women exposed) | | 0 (N = 255,416)d | 462611 | 164 (0.4) | 1.00 (referent) | 1.00 (referent) | 1.00 (referent) | | 1 (N = 130,157)d | 235691 | 141 (0.6) | 1.69 (1.35 to 2.11) | 1.60 (1.28 to 2.01) | 1.53 (1.17 to 1.99) | | 2 (N = 30,019)d | 54334 | 70 (1.3) | 3.64 (2.75 to 4.81) | 3.34 (2.51 to 4.44) | 3.57 (2.60 to 4.90) | | 3 (N = 2121)d | 3835 | 7 (1.8) | 5.16 (2.43 to 10.99) | 4.52 (2.11 to 9.67) | 4.70 (2.06 to 10.75) | Relative to women with no risk factors at their conventional cut points, the aHR for severe COVID-19 was 1.60 ($95\%$ CI 1.28-2.01) with one factor, 3.34 ($95\%$ CI 2.51-4.44) with two factors, and 4.52 ($95\%$ CI 2.11-9.67) in women with three factors (Table 2). There were 64,141 women ($15.4\%$) who were pregnant during the SARS-CoV-2 pandemic, of whom 93 (1.5 per 1000) had severe COVID-19. Further censoring on pregnancy generated roughly the same aHRs (Table 2). ## Discussion Elevated BMI and gestational DM are known risk factors for acquiring COVID-19 in pregnancy [4]. However, unlike prior studies, these two risk factors were handled here as continuous measures, and they were also assessed in relation to acquiring COVID-19 well beyond pregnancy. In a novel manner, gestational age at delivery was also assessed while avoiding any potential reverse causation introduced when COVID-19 and timing of birth were assessed in the same pregnancy because a severe infection may either precipitate or necessitate preterm birth [3]. The risk of severe COVID-19 was higher when each measure was analyzed continuously and even more so when combined together at conventional cut points. Recent data among middle-aged non-pregnant adults demonstrated a higher risk of COVID-19-related ICU admission, invasive mechanical ventilation, and mortality in the presence of the MetSyn [8]. While the current study could not formally assess all MetSyn criteria, maternal lipid profile, glucose intolerance in pregnancy, and pre-pregnancy BMI have been shown to predict the onset of MetSyn three months postpartum [9], as do a history of preterm delivery, gestational DM, and BMI years after pregnancy [1,2]. These study findings align with a body of work that considers in-pregnancy measures like BMI, glucose handling, and timing of delivery as a means to predict, and potentially modify, a woman’s future health. To date, studies have largely focused on future cardiometabolic health [9-11], but not new-onset infection. The current study further suggests that metabolic measures may offer a future perspective on a woman’s vulnerability to COVID-19. Even so, it remains to be determined whether these and other metabolic factors (e.g., blood pressure) influence her susceptibility to other types of viral and bacterial infections, or whether metabolic modification after birth can mitigate the onset of severe infectious illness. Strengths and limitations *As a* limitation, the current study adjusted for some confounders, such as rural residence and income level, but not race or ethnicity. Chronic hypertension was also accounted for, but blood pressure measures were not available. It is unlikely that caregiver burden can explain the relationship between preterm delivery and COVID-19 risk after pregnancy as parents of a preterm-born infant exhibit only slightly higher levels of stress than parents of term-born children [12]. Rather, women with the MetSyn are more likely to experience preterm delivery, especially provider-initiated preterm birth [13]. The study was conducted within a jurisdiction with relatively high vaccine uptake; so, the risk of severe COVID-19 may be even more pronounced in settings with lower vaccine use. While the first SARS-COV-2 vaccination was handled as a time-varying covariate, emerging SARS-CoV-2 variants were not differentiated. ## Conclusions The risk of severe COVID-19 remotely after pregnancy is higher in the presence of metabolic indicators standardly measured around the time of pregnancy. ## References 1. Catov JM, Althouse AD, Lewis CE, Harville EW, Gunderson EP. **Preterm delivery and metabolic syndrome in women followed from prepregnancy through 25 years later**. *Obstet Gynecol* (2016) **127** 1127-1134 2. Tranidou A, Dagklis T, Tsakiridis I. **Risk of developing metabolic syndrome after gestational diabetes mellitus - a systematic review and meta-analysis**. *J Endocrinol Invest* (2021) **44** 1139-1149. PMID: 33226626 3. Villar J, Ariff S, Gunier RB. **Maternal and neonatal morbidity and mortality among pregnant women with and without COVID-19 Infection: The INTERCOVID Multinational Cohort Study**. *JAMA Pediatr* (2021) **175** 817-826. PMID: 33885740 4. Allotey J, Stallings E, Bonet M. **Clinical manifestations, risk factors, and maternal and perinatal outcomes of coronavirus disease 2019 in pregnancy: living systematic review and meta-analysis**. *BMJ* (2020) **370** 3320 5. Li T, Wang Y, Wu L. **The association between ABO blood group and preeclampsia: a systematic review and meta-analysis**. *Front Cardiovasc Med* (2021) **8** 665069. PMID: 34235185 6. Ray JG, Schull MJ, Vermeulen MJ, Park AL. **Association between ABO and Rh blood groups and SARS-CoV-2 infection or severe COVID-19 illness: a population-based cohort study**. *Ann Intern Med* (2021) **174** 308-315. PMID: 33226859 7. You JJ, Alter DA, Stukel TA, McDonald SD, Laupacis A, Liu Y, Ray JG. **Proliferation of prenatal ultrasonography**. *CMAJ* (2010) **182** 143-151. PMID: 20048009 8. Denson JL, Gillet AS, Zu Y. **Metabolic syndrome and acute respiratory distress syndrome in hospitalized patients with COVID-19**. *JAMA Netw Open* (2021) **4** 0 9. Retnakaran R, Qi Y, Connelly PW, Sermer M, Zinman B, Hanley AJ. **Glucose intolerance in pregnancy and postpartum risk of metabolic syndrome in young women**. *J Clin Endocrinol Metab* (2010) **95** 670-677. PMID: 19926711 10. Retnakaran R, Shah BR. **Abnormal screening glucose challenge test in pregnancy and future risk of diabetes in young women**. *Diabet Med* (2009) **26** 474-477. PMID: 19646185 11. Jowell AR, Sarma AA, Gulati M. **Interventions to mitigate risk of cardiovascular disease after adverse pregnancy outcomes: a review**. *JAMA Cardiol* (2022) **7** 346-355. PMID: 34705020 12. Schappin R, Wijnroks L, Venema MMU, Jongmans MJ. **Rethinking stress in parents of preterm infants: a meta-analysis**. *PLoS One* (2013) **8** 0 13. Chatzi L, Plana E, Daraki V. **Metabolic syndrome in early pregnancy and risk of preterm birth**. *Am J Epidemiol* (2009) **170** 829-836. PMID: 19713286
--- title: Phytochemical Screening, Antioxidant Activity, and Acute Toxicity Evaluation of Senna italica Extract Used in Traditional Medicine authors: - Rodrigue Towanou - Basile Konmy - Mahudro Yovo - Christian C. Dansou - Victorien Dougnon - Frédéric S. Loko - Casimir D. Akpovi - Lamine Baba-Moussa journal: Journal of Toxicology year: 2023 pmcid: PMC10038741 doi: 10.1155/2023/6405415 license: CC BY 4.0 --- # Phytochemical Screening, Antioxidant Activity, and Acute Toxicity Evaluation of Senna italica Extract Used in Traditional Medicine ## Abstract Medicinal plants such as *Senna italica* are increasingly used for their purgative virtues to treat stomach aches, fever, and jaundice. This study aims to screen the phytochemical compounds and to assess the antioxidant activity in vitro and the acute oral toxicity in vivo of *Senna italica* leaves. The plant was harvested, dried, pulverized, and preserved. Phytochemical screening was performed using different laboratory protocols. Ethanolic and aqueous extracts were, respectively, obtained by maceration and decoction technics. The assay for free radical scavenging was used to examine the antioxidant activity using DPPH. Acute oral toxicity was performed with aqueous and ethanolic extracts at 5000 mg/kg of body weight on female albinos Wistar rats, weighing 152.44 ± 3.68 g. Subjects were checked for any signs of mortality and macroscopy toxicity during the 14 days of the study. Biochemical and hematological parameters were measured to assess liver and kidney functions, and histological analysis of these organs was conducted. Phytochemical analysis highlighted the presence of total phenols, flavones, tannins, alkaloids, and quinone derivatives. Semiethanolic (78 μg/mL), ethanolic (9.7 μg/mL), and aqueous extract (9.2 μg/mL) showed an interesting antioxidant activity. Biochemical and hematological parameters were normal and not significantly different ($p \leq 0.05$). The plant extracts did not produce any toxic effect or mortality at the provided dose. Senna italica extracts induced an increase in the volume of liver and kidney tissues but no necrosis. Thus, lethal dose 50 of *Senna italica* leaf extract is probably higher than 5000 mg/kg. ## 1. Introduction According to the World Health Organization, various plant fractions and their dynamic components are used as traditional medicines by $80\%$ of the global population [1]. One of these plants is Senna italica, colloquially known as agouema, an herbaceous plant or small deciduous shrub reaching 60 cm high generally with prostrate stems. Its leaves are organized in a spiral as encountered in Leguminosae-Caesalpinioideae family. They are glaucous, glabrous, alternate, and paripinnate with 5 to 6 pairs of obliquely oblong, obovate leaflets, and adnate at both ends with pointed apices [2]; the leaves, the ripe seeds, and the pods of *Senna italica* (S.italica) have always been consumed for their purgative virtues. Taken as a maceration or decoction, they may be useful to treat jaundice, stomach aches, venereal diseases, fever, and bilious crises as well as intestinal affections. The leaves are dried, pulverized, and used as a remedy for skin troubles such as ulcers and burns. The flowers are employed to purge and induce childbirth. The macerated roots are taken to treat colics and flu. The macerated roots are used to treat colic and flu. Boiled roots are used to dress wounds [3]. The infusion is smeared on eyes to relieve sores. S. italica roots are also a constituent of various medicines taken against indigestion, liver problems, spleen disorders, dysmenorrhea, vomiting, and nausea. In Malawi, the root infusion is used to treat infantile diarrhea. The value of Senna italica. as a grazing plant is not unanimous [4]. In East Africa, most domestic animals eat it, while in West Africa, they seem to avoid it. On top of the purgative effects of the mature seeds, the young seeds are consumed as an appetizer or a vegetable in the region of Sahel. In Mauritania, the seeds of *Senna italica* Mill. are smoked. Sold as “neutral henna” or “blond henna,” the leaves are also exploited as conditioner to make hair shiny. It can give hair a yellowish tint [5]. Several anthraquinones such as aloe emodin, chrysophanol, rhein, sennosides, and sennidines have been isolated from the leaves and pods of Senna italica. These different compounds are responsible for the purgative effect of this plant. Chrysophanol is another active component of “neutral henna.” The proportion in the leaves varies from 1.1 to $3.8\%$ of dry weight. The pods show a much lower content than those of the leaves. In addition, the leaves contain steroids (α-amyrin, β-sitosterol, and stigmasterol) and flavonoids (kaempferol, quercetin, and apigenin) [3]. The ethanolic extract of S. italica plant possesses some interesting antipyretic and anti-inflammatory properties. 1,5-dihydroxy-3-methoxy-7-methylanthraquinone obtained from *Senna italica* Mill. can be used against several Gram-negative and Gram-positive bacteria. It also shows an anticarcinogenic activity in vitro [5]. Toxicity assays performed on rabbits and goats fed with the plant foliage were negative. Chicks and rats nourished with $10\%$ of seed diet presented symptoms of toxicity, but no mortality sign during the 6-week trial. A $2\%$ seed diet stimulated chick growth. Seeds produce a gum soluble in water and mainly composed of D-galactose and D-mannose [6, 7]. The ethanolic and aqueous extracts of S. italica leaves have free radical scavenging, antioxidant, and secondary metabolite effects and are nontoxic for rats. To explore the phytochemical constituents of Senna italica, we performed phytochemical screening and determined the phenolic composition. To assess the extracts capacity for free radical scavenging, we measured their antiradical activity. To check if the extracts are toxic or not for rats, we executed the acute toxicity test on Wistar rats following the OECD protocol. ## 2.1. Medicinal Plant Extracts An extraction of S. italica leaves was conducted using traditional medicine techniques, utilizing aqueous, ethanolic, and semiethanolic methods. The leaves were freshly collected from the Grand Popo (Agoué) commune in the south of Benin and identified in the Benin National Herbarium under the identification number N° YH 722/HNB. Following collection, the leaves were dried at a controlled temperature of 20 to 25°C, ground into powder, and stored in a hermetically sealed container using a proper protocol [8]. For the aqueous extract, 50 grams of S. italica leaf powder was boiled in 500 mL of water at 100°C. The ethanolic extracts were prepared by mixing 50 grams of leaf powder with ethanol or $50\%$ ethanol and continuously agitating the mixture for 72 hours. The resulting macerate was filtered using the Whitman paper N°1, and the filtrates were concentrated in a rotavapor before being dried in a proofer at 50°C. The extraction yield was calculated using the following formula:[1]% Yield=Dry weight of extractDry weight of extract x 100. ## 2.2. Phytochemical Identification Phytochemicals present in the leaf powder were identified using a chemical method. Various tests including alkaloids, tannins, phenols, flavonoids, quinone derivatives, saponins, cyanogenic compounds, and coumarins were conducted using previously described methods [9]. ## 2.3. Determination of Total Phenolic, Tannins, and Flavonoid Contents The total phenolic compounds in the extracts obtained from $\frac{1}{15}$ sample (w: v) of fresh nuts were measured using Folin–Ciocalteu's method, as described by singleton, six solvents (water, methanol, ethanol, water-HCl $1\%$, ethanol-HCl $1\%$, and methanol-HCl $1\%$) were used to determine the extraction capacity of C. nitida. The nuts were ground with a grinder, adapted to 96 well-plates, and 25 μl of Folin–Ciocalteu's reagent ($50\%$ v/v) were combined with 10 μl of 1 mg/ml (w/v) of the nuts extract. After incubation at room temperature for 5 min, 25 μl of $20\%$ (w/v) sodium carbonate (Na2CO3) and water were added to reach a volume of 200 μl per well. Blanks were prepared using water instead of the reagent to minimize the impact of interfering compounds. The absorbance was measured at 760 nm after incubation (30 min) using a multiwell plate reader. The assays were performed in triplicate at least, and the results were expressed as microgram gallic acid equivalent per 100 grams of extract using gallic acid (0–500 μg/ml) as a standard. The total flavonoids in each sample were quantified using the aluminum trichloride method adapted to 96 well-plates [10]. Hundred microliters of methanolic AlCl3 ($2\%$) were mixed with 100 μl of appropriate dilution of the extract solution. After incubation (15 min), the absorbance was measured at 415 nm using a multiplate Epoch spectrophotometer Biotech connected to a computer with the help of “Gen5” software against a blank (mixture of 100 μl methanolic extract solution and 100 μl methanol) and compared to a quercetin (0–50 μg/ml) calibration curve (R2 = 0.99). The flavonoid content was expressed as mg of quercetin substitutes per 100 g of extract. The condensed tannin content was determined using the vanillin assay recommended by Belyagoubi et al. [ 11]. 1500 μL of vanillin/methanol solution ($4\%$, w/v) was added to 50 μL of extract (S1 or S2) and followed by the addition of 750 μL of $37\%$ HCl. The sample was incubated at room temperature for 20 min, and the absorbance was measured at 550 nm against a blank. Catechin was used as a standard for calibration curve, and the total proportion of condensed tannins was calculated as mg of catechin equivalents per g of dry matter (mg CE/g DM). ## 2.4. Antioxidant Assay The antioxidant assay of different extracts was evaluated by the 2,2-diphenyl-1-picrylhydrazyl radical (DPPH.) method, which was often used for its simplicity. It is a technic based on the reduction of an alcoholic solution of DPPH. Provided an antioxidant that gives hydrogen or a proton, the nonradical form DPPH-H was formed [12, 13]. The decoloration of DPPH radical depends on the concentration of the different extracts used. The extract-free radical scavenging activity, expressed as IC50, defines the effective concentration of the substrate that causes loss of $50\%$ of DPPH radical activity [14, 15]. The optic densities were measured at 517 nm and used to calculate the percentage of scavenging of the DPPH radical, which was proportional to the antiradical power of the sample [16]. These IC50 are determined from curves that estimate the antioxidant activity of extract as a percentage of crude extract concentration [12]. A volume of 100 μL of each extract at different concentrations was added to 1900 μL of the ethanolic solution of DPPH (40 μg/mL). The negative control was prepared in parallel by mixing 100 μL of extraction solvent with 1900 μL of DPPH solution. After incubation at room temperature and in the darkness for one hour, absorbances are taken at 517 nm using a HACH LANGE DR 3900 spectrophotometer. The percentage of trapping was calculated by the formula:[2]Scavenging DPPH radi cal=ODControl−ODTest sampleODControl×100. ## 2.5. Acute Toxicity Study Nine male Wistar rats, aged 12 weeks and weighing an average of 152.44 ± 3.68 g, were housed in wire mesh cages measuring 50 × 30 × 20 cm3 upon receipt. They were kept in accordance with the experimental conditions outlined by the OECD (Organization for Economic Cooperation and Development) Guideline-423, adopted on December 23, 2001, as per previous studies [17]. The temperature of the room was maintained at 25 degrees Celsius, and artificial lighting was provided in alternating cycles of 12 hours of light and 12 hours of darkness. The rats were provided with standard granulated food from a commercial food supplier in Benin and were given access to drilling water ad libitum. ## 2.5.1. Experimental Design The acute toxicity study was conducted in compliance with the OECD Guidelines (Organization for Economic Cooperation and Development, Guideline-423, adopted on 17 December 2001). Senna italica extract was administered once to the rats as a single dose of 5000 mg per kg of body weight in accordance with the acute oral toxicity protocol of the OECD Guidelines for Chemicals Testing 423, adopted December 17, 2001, and in accordance with good laboratory practice. The rats were divided into three groups of three rats each: the first group of control rats, the second group of normal rats which received the aqueous extract of *Senna italica* via tube-feeding at a dose of 5000 mg/kg body weight, and finally, the third group of normal rats which received *Senna italica* ethanolic extract via tube-feeding at a dose of 5000 mg/kg body weight. The animals were fasted for 12 hours prior to the administration of the extracts, after which they were weighed and then the extracts were given based on their fasting body weight. The rats were observed for 14 days during the experiment to monitor their behavior. The extracts were considered toxic if $50\%$ of the experimental rats had died. ## 2.5.2. Hematological Analysis Blood samples were collected in EDTA-coated tubes, and hematological parameters were determined using a Mindray hematology analyzer. The following hematological parameters were analyzed: total and differential white blood cell count (WBC), red blood cells number (RBC), red cell distribution width (RCDW), hematocrit (HCT), hemoglobin (Hb), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), platelet count (PLT), and mean platelet volume (MPV). ## 2.5.3. Biochemical Analysis After allowing blood samples in non-EDTA coated tubes to clot for 5 minutes, they were immediately centrifuged at 3000 rpm for 10 minutes to separate the serum for analysis. Aspartate aminotransferase (AST), alanine aminotransferase (ALT), urea, creatinine, alkaline phosphatase, gamma GT, and total cholesterol were analyzed using a chemistry analyzer (Hu-mastar 200, Germany). Electrolytes were analyzed using an electrolyte analyzer (*Humate plus* 5, Germany). ## 2.5.4. Histopathological Analysis On the 14th day of the experiment, the rats were sacrificed with thiopental (30 mg/kg) after sampling. The kidney and liver were removed, fixed in a $10\%$ formalin solution, and paraffin-embedded. Microtome sections of approximately 3 to 5 μm were made and mounted on glass slides. The sections were dewaxed in toluene and hydrated in decreasing alcohol baths. For histological analysis, the sections were stained using hematoxylin and eosin (H and E) following the standard protocol [18]. ## 2.6. Ethical Considerations The animal research guideline adopted by the ethics committee of the Research Unit in Applied Microbiology and Pharmacology of Natural Substances-University of Abomey-Calavi was followed to ensure adherence to experimental guidelines and animal welfare. ## 2.7. Statistical Analysis The variance was analyzed using the procedure of generalized linear model with the R software version 4.2.0. The significance ($p \leq 0.05$) of the group factor was determined using the F test, and the averages were compared two-by-two using the student test. ## 3.1. Phytochemical Identification The results of the phytochemical analysis of S. italica leaves powder highlighted several secondary metabolites such as gallic tannins, catechic tannins, flavonols, leuco-anthocyanins, saponosides, triterpenoids, quinone derivatives, steroids, coumarins, mucilages, reducing compounds, o-heterosides, and c-heterosides. On the other hand, secondary metabolites such as anthracenes, anthocyanins, and cyanogenic compounds are absent from *Senna italica* leaves (Table 1). ## 3.2. Analysis of Flavonoid and Total Phenolic Contents The ethanolic extract contains more total phenols (3179.91 ± 223.11) mg GAE/100 g and tannins (2.74 ± 0.15) mg·EC/g than aqueous extract polyphenol = 1497.24 ± 21.55 mg·GAE/100 g and tannins = 1.10 ± 0.04 mg·EC/g. The flavonoid contents in aqueous extract were higher than flavonoid of ethanolic extract (Table 2). ## 3.3. Antioxidant Activity Free radical scavenging depends on the concentrations of gallic acid, butylated hydroxytoluene, and quercetin. These curves allowed the determination of the concentrations of each synthetic compound which allows the scavenging of $50\%$ (IC50) of DPPH free radicals. It turns out that the concentrations allowed to trap $50\%$ of the DPPH radicals by the reference compounds are, respectively, 10.45 ± 1.59 μg/mL; 29.98 ± 1.91 μg/mL; and 71.67 ± 2.52 μg/mL for quercetin, gallic acid, and butylhydroxytoluene. Then, it makes sense that quercetin had higher activity than gallic acid and BHT (Table 3). The antioxidant activities of S. italica leaves were different according to extracts. The IC50 values of different extracts revealed medium scavenging activity from 9.33 ± 1.31 μg/mL (AE) to 9.65 ± 0.28 μg/mL (EE). The high values ($p \leq 0.05$) of IC50 78.9 ± 2.29 μg/mL were obtained with $50\%$ AEE (Table 3). ## 3.4. Acute Toxicity The mean values of several hematological parameters, namely, white blood cell (WBC) number, red blood cell (RBC) count, hemoglobin (Hb), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), packed cell volume (PCV) did not ($p \leq 0.05$) diverge significantly in all three groups. MCV of group 2 varied significantly at day 14. Similarly, the thrombocytes were significantly increased on day 14 in all three groups, especially in the two experimental groups (Table 4). We found a significant difference (p ≤ 0.05) in ALT levels and a very significant difference (p ≤ 0.001) in alkaline phosphatase (ALP) levels. However, AST and ALT levels at the 14th day for all groups were slightly lower than the ones at the beginning. At the same time, creatinine and urea levels showed no significant ($p \leq 0.05$) change from day 0 to day 14. However, we noted a slight increase in the creatinine concentration at day 14 in group 1 and a slight decrease in groups 2 and 3; whereas, urea levels showed a decrease at day 14 in groups 1 and 3 and an increase in group 2 (Table 5). ## 3.4.1. Weight of Rats Figure 1 shows the weight variation between the experimental rats and the controls. At the beginning of the experience, the average weight in the three groups did not vary significantly. However, we found a significant increase in the average weights during the test. The rats that received aqueous extract showed the highest weight gain (198 g). The control rats showed the lowest weight gain (175 g). ## 3.5. Histological Sections Figures 2 and 3 show the results of histological sections of the kidney and of the liver, respectively. The diameter of the centrilobular vein was enlarged in rats having received the S. italica aqueous extract, whereas in rats treated with the ethanolic extract, this vein was narrowed. In all experimental rats, the volume of hepatocytes increased (Figure 2). Analysis of the Figure 3 shows that the renal cortex was normal in all rats. The glomeruli are normal (compact) in the controls but are altered (red arrow) with the retrieval of some podocytes and increase of Bowman's space in the batches that received the extracts. The renal tubules (black arrows) in the controls had a normal morphology. Their volume had increased in the experimental rats. The tubular epithelial cells (yellow arrow) were detached from the basal lamina and several of them were necrotic (Figure 3). ## 4. Discussion The main organ managing vital functions like digestion and detoxification of molecular compounds in the body is the liver [1]. In developing countries, the use of pharmacopeia therapies is very common because of the flora wealth. However, the chemical content and safety of certain plants used remain unknown. This work has looked for the chemical composition, the antioxidant power, and the acute oral toxicity of S. italica leaves extracts. Phytochemical analysis of S. italica leaves powder showed steroids, alkaloids, tannins, flavonoids, saponosides, mucilages, quinone derivatives, and cardiotonic derivatives. On the other side, we did not find any traces of free anthracenes and cyanogenic derivatives. These results support the ones of [19] who also detected flavonoids, alkaloids, and steroids presence in the aqueous and methanolic extracts of S. italica. Tannins were not found in Dabai et al. [ 19], as opposed to our study. This could be due to the variations observed in the nature of the plant material, phenological state, and geographical location [19]. The present work is not fully in line with the work of Alqethami and Aldhebiani [20] who did not detect flavonoids presence in S. italica fruit, but did find tannins and saponins as in our study [20]. The nonappearance of certain secondary metabolites in our results could be justified by the seasonal variations. They can affect chemical composition but also biological activity of plants [21]. The results of this work agree with those of [22] who did not detect the presence of cyanogenic derivatives in the leaves of S. italica. Phenolic substances are recognized for their numerous health benefits. Various pharmacological activities (antioxidant, anticancer, and antimicrobial activities) are attributed to them [23]. The study showed that the aqueous extract ($p \leq 0.001$) has the highest proportion of total phenol ($1497.24\%$ ± 21.55) and of flavonoids ($1278.11\%$ ± 21.23) ($p \leq 0.0001$). Besides, the highest percentage of condensed tannin (2.74 ± 0.15) was attributed to the ethanolic extract (Table 2). Our results are similar to those obtained by Dah-Nouvlessounon et al. [ 10]. They obtained the highest content of total flavonoids (561.69 ± 22.10 μgQE/100 g) in G. kola. These findings prove that total flavonoids represent an important part of the total phenolic composition of Senna italica. Previous studies have reported many herbs as an excellent source of phenolic molecules recognized for their antioxidant effects [24, 25]. The present work shows a progressive rise in the trapped DPPH percentage. It goes up to $80\%$. The aqueous extract of this plant was the most active with an IC50 equal to 10 μg/mL and ensued by the semiethanolic extract (66 μg/mL) and by the ethanolic extract with an IC50 value equal to 78 μg/mL. The concentrations in semiethanolic and ethanolic extracts of S. italica leaves diverged from 9.2 μg/mL to 78 μg/mL. This is not in line with the results of [26] which pointed the ethanolic extract as the best source of antioxidants. The highest IC50 value was given by the semiethanolic extract. This proves the low antiradical activity of semiethanolic extract compared to others. The aqueous extract showed the highest activity with an IC50 value = 9.2 μg/mL tailed by the ethanolic extract. The ethanolic and aqueous extracts of S. italica leaves showed higher antiradical activities than the controls used in this work. Flavonoids, which pertain to phenolic compounds, are considered as potential antioxidant sources. They have the ability to reduce free radical species and reactive forms of oxygen. The reducing power of free radicals is explained by the shield effect of flavonoids. The low redox potential that makes flavonoids thermodynamic has also been attributed to their shielding effect. The transfer of the hydrogen atom generated by the antioxidant reaction gives rise to a peroxyl radical. According to the work of Mokgotho et al. [ 27], the antioxidant power of S. italica is linked to its content in resveratrol. For other authors, the use of S. italica in traditional care was due to its antibacterial, antioxidant, antidiabetic, and hypertensive properties [28–31]. For the acute oral toxicity of S. italica leaves extracts, after gavage of the rats with the ethanolic and aqueous extracts of Senna italica, we found neither mortality nor morbidity signs. Not a single moribund animal was obtained throughout the 14 days of experiment. This supports the results of [6]. We did not observe any change in the functioning of the skin, eye, hair, and respiratory system. Behavioral and physical signs of toxicity such as sleep disturbance, seizure, breathing, restlessness, or hyperactivity were also absent. These results give evidence for the nontoxicity of the ethanolic and aqueous extracts of S. italica after administration at 5000 mg/kg of body weight. The oral ingestion of ethanolic and aqueous extracts of S. italica at 5000 mg/kg did not affect the normal growth of the experimental rats as shown by the evolution of the weight gain (Figure 1). However, a gradual change in their weight was underlined. According to mean weight values analysis, the weights of the rats of the three groups at day 0 and day 14 did not differ significantly. Considering the extract impact on weight gain, our results are similar to the ones of Frimpong and Nlooto [29]. To assess the extracts effects on the function of the rats's vital organs, certain hematological parameters were measured. The white blood cells (WBCs) number, hemoglobin (Hb), red blood cells (RBCs), hematocrit (Hte), the mean corpuscular volume (MCV), mean corpuscular hemoglobin concentration (MCHC), mean corpuscular hemoglobin content (HCM), platelets (PLTs) were checked for the rats. Table 3 sums up the results obtained about the hematological parameters. We found that the count of white blood cells (WBCs), hemoglobin (Hb), red blood cells (RBCs), hematocrit (Hte), and mean corpuscular hemoglobin concentration (MCHC) did not increase significantly ($p \leq 0.05$) in all three groups as opposed to mean corpuscular volume (MCV) which increased on day 14 compared to day 0. There is a decline in the count of red blood cells (RBCs), an anemia and an increase in case of exaggerated production or loss of liquid [32]. Platelet count (PLT) increased significantly on day 14 compared to day 1 in group 1 (p ≤ 0.05). It also increased very significantly (p ≤ 0.0001) in the two other groups on day 14. The platelet count allows the detection of a bleeding, infectious, or inflammatory risk after huge bleeding [32]. The thrombocytes count variations observed in our case would be linked to physiological growth of the rats. Statistical analysis showed no significant difference ($p \leq 0.05$) between the hematological parameters of experimental rats having received the aqueous extract or the ethanolic extract and control rats for the other parameters. There was also no significant variation in these parameters in each group between the first day and the fourteenth, despite the first observations. The high values of MCV, CHM, and MCHC indicate the presence of macrocytic normochromic red blood cells, while a decrease points to the presence of hypochromic microcytic RBC [30]. Apart from the significant variation in MCVs observed in batch 2 on day 14, in our case, there was no other significant behavior. This shows that the RBC of the rats were normocytic-normochromic and, therefore, that the ethanolic extract had no noxious effect on the RBC of the rats at 5000 mg/kg of body weight. In conclusion, the ethanolic and aqueous extracts of S. italica had no toxic effect on blood platelets and the aqueous extract of S. italica had no toxic effect on MCV. However, various factors related to the subject and its environment could be responsible for the nonsignificant variability recorded in this study and the observations related to the cell variability, in particular lifespan [33]. Several biochemical markers, namely, glucose, transaminases (AST and ALT), urea, creatinine, alkaline phosphatase (ALP), and gamma GT were also measured to assess the impact of ethanolic and aqueous extracts of S. italica on rats' vital organs. Table 5 shows the average concentrations of these parameters in the experimental group with the time. We obtained no significant difference ($p \leq 0.05$) between the levels of these markers in the groups on day 0 and 14, except on day 14 for group 2 (Table 5). There was a substantial variation (p ≤ 0.05) in ALT level and a highly significant difference (p ≤ 0.001) in alkaline phosphatase (ALP). However, AST and ALT levels on day 14 for all groups of rats were slightly lower than on day 0, as were creatinine and urea levels. The findings revealed no significant change ($p \leq 0.05$) between day 0 and 14. However, we noted a slight increase by day 14 in the creatinine level in group 1 and a slight decrease in groups 2 and 3, while urea levels decreased in groups 1 and 3 and increased in group 2. Creatinine and urea levels are good indicators of kidney function [34]. The data collected about biochemical parameters revealed no significant variation between the levels of AST, blood glucose, urea, gamma GT, and creatinine for the experimental rats exposed to the extracts compared to controls. Among the biochemical markers covered, transaminases (AST and ALT) are normally found in many cells' cytoplasm and mitochondria, mainly in the liver, heart muscle, and skeletal muscle. However, their concentrations are lower in the pancreas, kidney, and erythrocytes [35]. Therefore, increased serum AST and ALT levels indicate liver toxicity. This occur generally in the blood when the permeability of liver cells is impaired or when necrosis take place. AST and ALT are hepatic enzymes insuring the chemical transfer of an amine group to other molecules in the liver [34]. The activity of these enzymes is relative to the degree of damage [35]. Therefore, they are two relevant indicators of hepatotoxicity [36, 37]. An increase in two-or three-folds range of ALT levels implies hepatic cytolysis [32]. An increase in serum AST activity indicates a traumatic, an inflammatory or degeneracy due to plasma membrane damage and cellular necrosis [37, 38]. The significant change in ALT levels seen in group 2 highlights a hepatic cytolysis. Inflammation and tissue degeneration noticed in rats treated with S. italica aqueous extract at 5000 mg/kg of body weight did not affect hepatocytes. Therefore, we can deduce that the S. italica ethanolic extract did not cause any damage on the liver of rats at 5000 mg per kg of body weight. Renal function was assessed through serum ureal and creatinine concentrations. Creatinine and urea are important markers of the kidney function [39, 40]. These metabolism products have a constant level under normal conditions [41]. Renal impairment is reflected by their decrease or increase [42]. Pritchard and his colleagues had shown that a reduction in serum creatinine could indicate cachexia. Looking at the serum urea concentration, its rise can sign a nephropathy, dehydration, electrolyte imbalance, hypo-albuminemia, and tissue catabolism [42]. Because these parameters did not vary significantly in experimental rats as opposed to controls, we came to the conclusion of a normal kidney function. In sum, since no significant variation in AST, gamma GT, blood glucose, urea, and creatinine levels was noted, we can deduce that the ethanolic extract of S. italica is not toxic to the liver and kidneys in rats at 5000 mg/kg of body weight. S. italica ethanolic extract administered to rats in a proportion of 5000 mg/kg of body weight have not caused concomitant changes in the number of white blood cells and red blood cells. The first are essential to fight infection and develop resistance to infection after a prior exposure or vaccination. They consist of monocytes, lymphocytes, and granulocytes [43]. In the case of infection, inflammation, cancer or leukemia, the leucocytes count is increased and can be reduced by the bone marrow failure, liver disease, splenomegaly, autoimmune diseases, and certain drugs. Their levels are not very different in treated rats in comparison to the control group. This demonstrates that the S. italica ethanolic extract presents no toxicity for the white blood cells of rats at 5000 mg/kg of body weight. S. italica aqueous extract administered at 5000 mg/kg body weight produced significant change in ALT levels indicating liver cytolysis, inflammation, tissue degeneration, and thrombocytosis in administered rats. This explains the switch of the hemostasis system. Moreover, there was a significant variation in the alkaline phosphatase level of the rats having received 5000 mg/kg of body weight of S. italica aqueous extract. We can then summarize that the ethanolic extract was not toxic at a limit dose of 5000 mg/kg of body weight, while the aqueous extract of S. italica would be toxic only at a dose lower than 5000 mg/kg of body weight. ## 5. Conclusions The *Senna italica* leaves contain various pharmacologically active compounds including phenolics, derivated quinone, and flavonoids, which are good antioxidant, anti-inflammatory, and protective. S. italica leaves are nontoxic at the dose of 5000 mg/kg of body weight. Thus, the leaves of S. italica have a powerful antioxidant activity. They can be used orally as traditional medicine. S. italica leaves constituted then a useful phytobiotic resource which can promote human health. Dose studies are, however, necessary depending on the type of pathology. ## Data Availability No additional information is available for this paper. ## Conflicts of Interest The authors declare that they have no conflicts of interest. ## Authors' Contributions All authors conceptualized the study; RT. BK. MY. CCD. performed methodology; BK performed software CDA, VD, LB-M, and FSL validated the study; BK. made formal analysis; RT investigated the study; RT made resources; RT. curated the data; RT. written and prparedthe original draft; BK. RT. MY. written, reviewed, and edited the manuscript; CDA.VD. LB-M visualized the study; CDA, LB-M, and FSL. supervised the study All authors have read and agreed to the published version of the manuscript. ## References 1. Shahat A. A., Ullah R., Alqahtani A. S., Alsaid M. S., Husseiny H. A., Al Meanazel O. T. R.. **Hepatoprotective effect of eriobotrya japonica leaf extract and its various fractions against carbon tetra chloride induced hepatotoxicity in rats**. (2018) **2018** 2. Arbonnier M.. (2009) 3. El Agamy E. S. I., Ruppanner R., Ismail A., Champagne C. P., Assaf R.. **Antibacterial and antiviral activity of camel milk protective proteins**. (1992) **59** 169-175. DOI: 10.1017/s0022029900030417 4. Alqahtani A. S., Ullah R., Shahat A. A.. **Bioactive constituents and toxicological evaluation of selected antidiabetic medicinal plants of Saudi arabia**. (2022) **2022** 1-23. DOI: 10.1155/2022/7123521 5. Sermakkani M., Thangapandian V.. **GC-MS analysis of**. (2012) **5** 90-94 6. Anadón A., Martínez M. A., Ares I.. **Acute oral safety study of sodium caseinate glycosylated via maillard reaction with galactose in rats**. (2014) **77** 472-479. DOI: 10.4315/0362-028x.jfp-13-237 7. Ullah R., Alqahtani A. S., Noman O. M. A., Alqahtani A. M., Ibenmoussa S., Bourhia M.. **A review on ethno-medicinal plants used in traditional medicine in the Kingdom of Saudi Arabia**. (2020) **27** 2706-2718. DOI: 10.1016/j.sjbs.2020.06.020 8. Dougnon T. V., Hounsa E., Agbodjento E.. **Toxicological characterization of ten medicinal plants of the Beninese flora used in the traditional treatment of diarrheal diseases**. (2021) **2021** 1-13. DOI: 10.1155/2021/6676904 9. Evans W. C.. (2009) 10. Dah-Nouvlessounon D., Adjanohoun A., Sina H.. **Nutritional and Anti-Nutrient Composition of Three Kola Nuts (<i&amp;gt; Cola nitida, Cola acuminata </i&amp;gt; and <i&amp;gt; Garcinia kola </i&amp;gt;) Produced in Benin**. (2015) **06** 1395-1407. DOI: 10.4236/fns.2015.615145 11. Belyagoubi L., Loukidi B., Belyagoubi-Benhammou N.. **Valorization of Algerian saffron: stigmas and flowers as source of bioactive compounds**. (2021) **12** 6671-6683. DOI: 10.1007/s12649-021-01454-6 12. Bortolomeazzi R., Sebastianutto N., Toniolo R., Pizzariello A.. **Comparative evaluation of the antioxidant capacity of smoke flavouring phenols by crocin bleaching inhibition, DPPH radical scavenging and oxidation potential**. (2007) **100** 1481-1489. DOI: 10.1016/j.foodchem.2005.11.039 13. Fanou B. A., Klotoé J. R., Dougnon V., Soha A., Yovo M., Loko F.. **Etude comparative de la composition chimique, de l’activité antiradicalaire et de la toxicité de quatre plantes utilisées dans le traitement traditionnel des candidoses au Benin**. (2022) **20** 54-64 14. Ahmad N., Fazal H., Abbasi B. H., Anwar S., Basir A.. **DPPH free radical scavenging activity and phenotypic difference in hepatoprotective plant (Silybum marianum L.)**. (2013) **29** 460-467. DOI: 10.1177/0748233712436637 15. Alain K. Y., Yovo M., Boniface Y.. **Chemical study, antiradical and antibacterial potential of the extracts of Ximenia americana and Cussonia arborea of Benin**. (2014) **2** 1626-1635 16. Agbangnan D., Noudogbessi J., Chrostowska A., Tachon C., Fouquet E., Sohounhloue D.. **Phenolic compound of Benin’s red sorghum and their antioxidant properties**. (2013) **6** 277-280 17. Hounsa E., Dougnon T. V., Agbankpe A. J.. **Fetotoxicity and subacute toxicity of some plants involved in the treatment of infectious diarrhea in Benin**. (2022) **3**. DOI: 10.3389/fitd.2022.868645 18. Sènou M., Tchogou A., Agbangnan D.. **Efficacy of ethyl acetate fraction of Cocos nucifera aqueous extract on the treatment of anemia**. (2017) **6** 44-51 19. Dabai Y., Kawo A., Aliyu R.. **Phytochemical screening and antibacterial activity of the leaf and root extracts of**. (2012) **6** 914-918 20. Alqethami A., Aldhebiani A. Y.. **Medicinal plants used in jeddah, Saudi arabia: phytochemical screening**. (2021) **28** 805-812. DOI: 10.1016/j.sjbs.2020.11.013 21. Walt G.. **WHO’s World health report 2003**. (2004) **328** p. 6. DOI: 10.1136/bmj.328.7430.6 22. Adjou E. S. n., Koudoro A. Y., Nonviho G., Ahoussi E. D., Sohounhloue D. C. K.. **Phytochemical profile and potential pharmacological properties of leaves extract of**. (2021) **9** 36-39 23. Omer H. A. A., Caprioli G., Abouelenein D.. **Phenolic profile, antioxidant and enzyme inhibitory activities of leaves from two Cassia and two Senna species**. (2022) **27** p. 5590. DOI: 10.3390/molecules27175590 24. Nkosi C. Z., Opoku A. R., Terblanche S. E.. **Antioxidative effects of pumpkin seed (Cucurbita pepo) protein isolate in CCl4-induced liver injury in low-protein fed rats**. (2006) **20** 935-940. DOI: 10.1002/ptr.1977 25. Teissedre P. L., Waterhouse A. L.. **Inhibition of oxidation of human low-density lipoproteins by phenolic substances in different essential oils varieties**. (2000) **48** 3801-3805. DOI: 10.1021/jf990921x 26. Kolar F. R., Gogi C. L., Khudavand M. M., Choudhari M. S., Patil S. B.. **Phytochemical and antioxidant properties of some Cassia species**. (2018) **32** 1324-1328. DOI: 10.1080/14786419.2017.1342085 27. Mokgotho M. P., Gololo S. S., Masoko P.. **Isolation and chemical structural characterisation of a compound with antioxidant activity from the roots of**. (2013) **2013** 1-6. DOI: 10.1155/2013/519174 28. Masoko P., Gololo S. S., Mokgotho M. P., Eloff J. N., Howard R. I., Mampuru L. J.. **Evaluation of the antioxidant, antibacterial, and antiproliferative activities of the acetone extract of the roots of**. (2009) **7** 138-148. DOI: 10.4314/ajtcam.v7i2.50873 29. Frimpong E., Nlooto M.. **Tswana traditional health practitioners’ perspectives on the management of diabetes and hypertension: a qualitative study using focus group discussions**. (2019) **34** p. 93. DOI: 10.11604/pamj.2019.34.93.19112 30. Yovo M., Dedome S.-L. O., Sessou P.. **Etude phytochimique et activites biologiques des extraits de deux plantes medicinales utilisees pour traiter les infections cutanees et les septicemies au Benin**. (2020) **28** 507-514 31. Semerdjieva I., Zheljazkov V. D., Radoukova T.. **Biological activity of essential oils of four juniper species and their potential as biopesticides**. (2021) **26** p. 6358. DOI: 10.3390/molecules26216358 32. Michelson A. D.. **The clinical approach to disorders of platelet number and function**. (2019) 701-705 33. Reis A. B., Martins-Filho O. A., Teixeira-Carvalho A.. **Parasite density and impaired biochemical/hematological status are associated with severe clinical aspects of canine visceral leishmaniasis**. (2006) **81** 68-75. DOI: 10.1016/j.rvsc.2005.09.011 34. Yovo M., Towanou R., Ayihou Y.. **Acute toxicity of Aeollanthus pubescens essential oil with high antimicrobial potential against multidrug resistant bacteria isolated in poultry farms in Benin**. (2019) **6** 17-26 35. Costa-Silva J., Lima C., Silva E.. **Acute and subacute toxicity of the Carapa guianensis Aublet (Meliaceae) seed oil**. (2008) **116** 495-500. DOI: 10.1016/j.jep.2007.12.016 36. Pratt D. S., Kaplan M. M.. **Evaluation of abnormal liver-enzyme results in asymptomatic patients**. (2000) **342** 1266-1271. DOI: 10.1056/nejm200004273421707 37. Da Silva A. S., Wolkmer P., Costa M. M.. **Biochemical changes in cats infected with Trypanosoma evansi**. (2010) **171** 48-52. DOI: 10.1016/j.vetpar.2010.02.047 38. Sow A., Sidibé I., Bengaly Z.. **Field detection of resistance to isometamidium chloride and diminazene aceturate in Trypanosoma vivax from the region of the Boucle du Mouhoun in Burkina Faso**. (2012) **187** 105-111. DOI: 10.1016/j.vetpar.2011.12.019 39. El Khasmi M., Farh M.. **Impact des plantes médicinales sur le rein**. (2022) **2** 32-40 40. Gondjout O. L. V.. (2022) 41. Li X., Hui S., Mirek E. T.. **Circulating metabolite homeostasis achieved through mass action**. (2022) **4** 141-152. DOI: 10.1038/s42255-021-00517-1 42. Pritchard J. C., Burn C. C., Barr A. R., Whay H. R.. **Haematological and serum biochemical reference values for apparently healthy working horses in Pakistan**. (2009) **87** 389-395. DOI: 10.1016/j.rvsc.2009.05.003 43. Ibrahim A.. **Responses of white blood cells to killed**. (2021) **75**
--- title: 'Coverage of a Population-Based Non-Communicable Disease Screening Program Using Lot Quality Assurance Sampling in Rural North India: A Mixed Methods Study' journal: Cureus year: 2023 pmcid: PMC10038759 doi: 10.7759/cureus.35330 license: CC BY 3.0 --- # Coverage of a Population-Based Non-Communicable Disease Screening Program Using Lot Quality Assurance Sampling in Rural North India: A Mixed Methods Study ## Abstract Aim: We aimed to estimate the coverage of a population-based Non-communicable Disease (NCD) screening program using lot quality assurance sampling (LQAS) and identify factors affecting its implementation in district Nuh of Haryana, India. Method: A mixed-methods study was conducted with an initial LQAS coverage survey, followed by in-depth interviews. Thirty lots (villages or towns) were sampled in the district, and 20 people aged ≥ 30 years were randomly sampled from each lot. Participants were asked about receiving services under the program. Weighted coverage estimates, which is the proportion of people who had received screening services, were estimated. Using a decision value of more than nine negative responses out of 20 persons, all 30 lots were classified as good or poor performing. In-depth interviews of healthcare providers of good performing lots and district-level health officials were conducted, and factors affecting program implementation were identified. Findings: Six hundred participants were interviewed (mean age of 44.8 years, $57.2\%$ women). The proportion of people who reported having undergone screening for diabetes or hypertension was $2.1\%$, and all lots performed poorly based on decision value. Key factors affecting the program were leadership, prioritization of NCD activities, ensuring human resource and material requirements, regular incentives, qualities of workers, and community engagement. Conclusion: The screening coverage under the population-based NCD screening program was low in district Nuh, Haryana. This needs to be improved by addressing the identified health system and community-related factors. ## Introduction The Population-Based Screening (PBS) program was rolled out through the National Program for the Prevention and Control of Cancer, Diabetes, Cardiovascular Diseases, and Stroke (NPCDCS) in 2017 to integrate health promotion, screening, and management of five common non-communicable diseases (NCDs): diabetes, hypertension, oral cancer, breast cancer, and cervical cancer, for people who are ≥ 30 years of age [1]. This program is integrated with the latest Health & Wellness Centres (HWCs) initiative as part of the 12 expanded range of services to provide Comprehensive Primary Health Care (CPHC). As per the health ministry's annual report [2019-20], the PBS program has been implemented in 219 of the 742 districts in India [2]. As of 2020, more than $80\%$ of Medical Officers (MO), Auxiliary Nurse Midwife (ANM), Staff nurses (SN), and Accredited Social Health Activists (ASHA) in Haryana have undergone training for implementing PBS [3]. Nuh (previously Mewat) is a high-priority district (commonly known as an aspirational district) in Haryana State with a population of 10,89,406, which is predominantly rural ($90\%$) [4,5]. Nuh is prioritized for the operationalization of HWCs and has the highest number of functional HWCs in Haryana [3]. In the district Nuh, the NPCDCS-PBS program was implemented in 2018. The targets for achieving coverage for screening under the program are $50\%$ in year one, $65\%$ in year two, and $80\%$ in year three post-implementation [1]. It is desirable to design coverage surveys that utilize existing resources and can give rapid results to aid local health officials in making implementation decisions. Lot Quality Assurance Sampling (LQAS) is one such type of Rapid Epidemiological Assessment (REA) method, where the population under survey is divided into non-overlapping subpopulations (lots) for sampling and surveyed for coverage of a program [6]. It has been used to estimate immunization coverage, evaluate communicable disease control status, assess the performance of community health workers, etc. [ 7,8]. This method can provide coverage estimates for the whole population under survey and for each such sub-population or lot to redirect attention to where it is needed [9,10]. Being a rapid evaluation tool and amenable to use by health staff, it can be adapted for periodic coverage monitoring of the NCD screening program [11]. To estimate the coverage of population enumeration and the Community Based Assessment Checklist (CBAC) by ASHA under the NPCDCS PBS program in Nuh (Mewat) district, Haryana, using the LQAS technique, and to determine the factors facilitating and limiting implementation and coverage of the program. ## Materials and methods Study characteristics A mixed methods study consisting of two phases was designed to meet the objectives: an initial quantitative phase to estimate coverage and a qualitative phase where in-depth interviews were conducted to explain and supplement the quantitative survey findings. Lot quality assurance sampling (LQAS) was used to perform the coverage estimation of population enumeration, administration of CBAC, and screening activities by ASHA and ANM in the villages and towns of district Nuh, Haryana, India. The study was performed during the period December 2019 and January 2020. The study population was people who were ≥ 30 years of age and residents of the village/town for > six months. Sample size and sampling For the desired level of confidence and accuracy for the coverage estimate of $4\%$ and $95\%$, respectively, World Health Organization (WHO) manual for using LQAS was used to estimate a sample size of 600 [10]. A lot was defined as a village or town in the district. Out of the 443 villages and four towns in Nuh as per Census 2011, 27 villages and three towns were selected by probability proportional to size, giving a total of 30 lots (20 people to be sampled in each lot) [5]. The map-grid method was used in each lot to select 20 sampling areas as sampling frames or household lists were unavailable [10]. Each sampling area was approached to its approximate center using Geographical Positioning System (GPS); roads were counted north → east → south → west, and one was randomly selected. If the sampling area didn’t have a house, another sampling area was selected. In the selected road, a total number of houses were counted, and one house was randomly selected. In the selected house, a total number of eligible participants was counted, and one was randomly selected and approached to participate in the study. In case the selected house was locked or did not have any eligible person, or when the selected person was not available after two visits or didn’t consent to the study, an immediately adjacent house in the forward direction was approached. Separate random tables were used for sampling at different levels. These selected participants were asked whether they had received services under the program, using cues and specifiers to elicit the status of enumeration & CBAC administration by ASHA and the status of screening by ANM. A pretested semi-structured interview schedule in Hindi and Epicollect5 application were used for data collection. Quantitative analysis The proportion for categorical variables and mean with standard deviation for continuous variables were calculated for descriptive analyses. Sample weight was calculated as the inverse of the probability of selection (the product of the probability of selecting a lot and the probability of choosing an individual in a lot) and applied to all the individuals in the dataset before analysis. Coverage estimates with a $95\%$ confidence interval for various indicators were calculated using the command for estimating proportions for survey data in StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP. In addition to evaluating the overall coverage, lots were classified as good and poor performing for various indicators using a decision value of nine, obtained from Lemeshow & Taber LQAS tables [10]. A decision value of nine meant that when more than nine participants in a lot gave a negative response for an indicator, the lot was classified as poor performing. Qualitative methods The qualitative part was designed to supplement and explain the survey's quantitative findings and determine the factors facilitating and limiting the program. In-depth interviews were conducted over the telephone or in person using interview guides from July to September 2020. One of the good-performing lots was selected for elucidating the facilitating factors. As the investigator couldn’t contact the health workers in poor-performing lots due to COVID-19 lockdown restrictions, interviews with district health officials were used to identify barriers to program implementation. A total of three interviews were conducted, one each for an ASHA, a CHO, and a MO from one of the good-performing lots. And two interviews were conducted with one district NCD Officer and the district's Chief Medical Officer (CMO). The persons in charge during the first year of program implementation, were contacted and interviews were arranged. Qualitative analysis Interviews were conducted in Hindi and English. Interview audio records were transcribed into Hindi and translated into English using Microsoft Word (Redmond, USA). Scripts were coded manually by the deductive coding method. Facilitators and barriers to program implementation and coverage were then identified. Themes were formed by grouping the codes to identify critical areas of focus for the overall improvement of program implementation based on the framework of Health System Building Blocks and the role of the community. Ethics approval Ethical approval for conducting the study was obtained from the Institute Ethics Committee, AIIMS New Delhi (Number: IECPG-$\frac{522}{14.11.2018}$, OT-$\frac{1}{29.08.2019}$). Informed consent was obtained from the village or town representatives and study participants. ## Results LQAS survey results A total of 622 houses were approached for the study. Four houses were locked, and five had no eligible persons for the study. Of the 613 houses with at least one eligible person, four houses did not have the selected participants available for the interview even after two visits. Nine people did not consent to participate in the study. Finally, a total of 600 participants were interviewed for the survey. The mean ± standard deviation of the age of the participants was 44.8 ± 12.2 years. Most of the participants were women ($57.2\%$), housewives ($55.2\%$), followed by laborers ($15.8\%$), had no schooling ($68.5\%$), and belonged to the Muslim religion ($80.2\%$). According to the Udai Pareek scale [12], most families were classified as lower class ($55.7\%$), followed by the lower middle class ($44\%$). Of the 600 participants, 12 ($2.0\%$) reported having diabetes mellitus, 22 ($3.7\%$) reported having hypertension, and three ($0.5\%$) reported having both diabetes and hypertension. Of the 12 diabetes and 22 hypertensives, eight and 12 said they were currently taking medications for these conditions. Coverage of population-based screening program activities The coverage of the key activities under the population-based screening program is shown in Table 1. The proportion of people for whom ASHA had asked details related to population enumeration, i.e., line listing of family members and their identity details, fuel, and water usage, was $66.5\%$ ($95\%$ CI: 60.9-72.0), with 25 of the 30 lots performing acceptably on this indicator. The estimated proportion of people who self-reported to have undergone screening for diabetes and hypertension was only $2.1\%$ ($95\%$ CI: 0.01-4.3), and all 30 lots performed poorly for the same. **Table 1** | No. | Indicator | Estimated overall coverage among eligible persons in Nuh district, % (95% CI) | Good performing lots (> 10/20 positive response/lot) | Poor performing lots (< 11/20 positive response/lot) | | --- | --- | --- | --- | --- | | 1 | Recognizing ASHA(s) in their village | 89.2 (85.6–92.7) | 30/30 | | | 2 | ASHA asked population enumeration related details | 66.5 (60.9–72.0) | 25/30 | 5/30 | | 3 | ASHA asked disease status details | 8.9 (5.8–12.0) | | 30/30 | | 4 | ASHA asked details related to CBA checklist | 2.5 (1.0–4.0) | | 30/30 | | 5 | ASHA measured waist circumference | 0.9 (0–1.8) | | 30/30 | | 6 | ASHA informed the need for screening for Diabetes & Hypertension | 4.5 (0.1–8.9) | | 30/30 | | 7 | ASHA provided information about a screening camp | 5.3 (0–11.3) | 1/30 | 29/30 | | 8 | Screened for diabetes and/or hypertension in the camp conducted through ASHA and/or ANM | 2.1 (0.01–4.3) | | 30/30 | Of the 12 participants out of 600 who reported having undergone screening for diabetes or hypertension, two participants were detected to be screen positive (one person with diabetes and another hypertensive). They were referred to Primary Health Centre (PHC) for diagnosis and were then diagnosed with the disease. Results from in-depth interviews A total of five in-depth interviews were conducted. On coding the scripts, facilitating and limiting factors for program implementation and coverage were identified. These factors, whichever recurred across the interviews, were grouped to identify key areas of focus for overall improvement of program implementation, giving eight themes. The various facilitating factors, limiting factors and themes identified are shown in Figure 1. **Figure 1:** *Framework showing the factors which facilitate and limit the implementation of the population-based screening programHWCs: Health & Wellness Centers, MO: Medical Officer, PBS: Population-Based Screening, NCD: Non-Communicable Diseases, CBAC: Community-Based Assessment Checklist, ASHA: Accredited Social Health Activist, CHO: Community Health Officer, RCH: Reproductive and Child Health, PHC: Primary Health Centre, ANM: Auxiliary Nurse and Midwife* Key facilitating factors that emerged were i) Leadership at all levels of implementation to conduct the enumeration activity or organize the screening camps, and work management skills, ii) Prioritization of NCD activities, iii) Arrangements to ensure human resource and material requirements, iv) Identify CHO as a critical person coordinating the PBS activities, v) Regular incentives for various PBS-related activities to CHO, vi) Positive worker qualities and coordination capabilities, and vii) Engaging community and beneficiaries by informing them of the need and resolving their queries. Key limiting factors for the program implementation were i) Low priority is given to NCD activity relative to other programs, at different levels, from district to ASHA, ii) Lack of adequate equipment and infrastructure to perform screening activities, iii) An inadequate number of health workers from the local community in the cadre of ANM and CHO, iv) Nonpayment of incentives for PBS activities to ASHAs and their need for awareness about the same, v) Lack of ownership over their community, hesitancy to start new initiatives, and vi) Poor response from the community. In addition to these factors, it is imperative to note that lockdown measures implemented for COVID-19 pandemic control efforts led to a temporary cessation of all screening-related activities in the village. Eight themes were identified at the end of the thematic analysis: leadership and management, NCD as a priority, material resources and infrastructure, human resources, training for PBS, work incentives, workers’ qualities, community characteristics, and community engagement. These themes could be considered critical areas of focus for improving the implementation and coverage of the PBS program when a program review is performed at the district and state levels. ## Discussion Our LQAS survey found that the estimated coverage of screening under the population-based NCD screening program was less than the desired target of $50\%$ in district Nuh, Haryana. Using the LQAS technique permitted the classification of villages or towns into good or poor performance for the various program activities and explored multiple factors that could be attributable to their performance. Designing an LQAS survey has made the study operationalizable in a setting without a prior sampling frame, which otherwise would have been difficult through a survey using usual sampling methods. Also, it was found that this technique gives results rapidly, making it a preferable method choice. Studies that assess the feasibility and efficiency of using LQAS as part of routine monitoring systems at the district level could give more evidence about its overall utility. A similar study by Nambiar D et al. assessed the coverage of screening for diabetes and hypertension under the PBS program in four districts of two states using the LQAS technique [13]. However, a lot was defined as the area served by one ANM. It was found that the coverage levels of blood pressure screening were less than the desired threshold in both Delhi ($58.9\%$ in Central and $61.5\%$ in South, both less than $65\%$) and Uttar Pradesh ($30.9\%$ in Shrawasti and $33.4\%$ in Jhansi, both less than $50\%$). Blood sugar screening coverage levels were much less than the desired threshold in Delhi ($47.4\%$ in Central and $53.1\%$ in South, both less than $65\%$) and Uttar Pradesh ($13.7\%$ in Shrawasti and $20.4\%$ in Jhansi, both less than $50\%$) [13]. Although the coverage of screening services was lower than the desired target in both Nambiar et al. and this study, the magnitude of the difference between theirs and this study is significant [13]. This could be attributed to differences in implementation period and health system or community-related factors between states and districts. A health system readiness assessment for rolling out of the universal screening, prevention, and management of common NCDs was done by the National Health Systems Research Centre (NHSRC) reported that there was a low level of readiness, inadequate/no training for ANMs and MOs (during early 2018), screening activities not yet rolled out and cultural barriers for rolling out the program [14]. Although this assessment was conducted a few months earlier to our study, their results are comparable to the findings from the qualitative results of our research. There was no quantitative coverage estimation for various programmatic areas such as population enumeration, CBAC assessment, or screening. The study's strengths were that random sampling methods were used at all sampling levels (village or town, sampling area, road, house, eligible participant), increasing the coverage estimate's generalizability. Also, the random selection at the village level to select sampling areas gave an equal probability of communities with different socio-demographic statuses being chosen for the study. The qualitative method through in-depth interviews helped to supplement the quantitative findings. The study's limitations include that during the data collection period, there was social unrest in the community, which could have affected the validity of the findings [15]. Also, in-depth interviews couldn’t be conducted in the poor-performing lots given the COVID-19 pandemic, which might have affected data saturation. ## Conclusions The screening coverage under the population-based NCD screening program was low in district Nuh, Haryana. This needs to be improved by addressing the identified health system and community-related factors. Making NCD a priority, leadership and management, provision of material and human resources, training for PBS, work incentives, workers’ qualities, and community engagement could be key areas of focus for improving the implementation and coverage of a community-based NCD screening program in India. ## References 1. Ministry of Health and Family Welfare, Government of India. **Operational Guidelines for Prevention, Screening and Control of Common Non-Communicable Diseases: Hypertension, Diabetes and Common Cancers**. (2016) 2. Ministry of Health and Family Welfare, Government of India. **Annual Report 2019-20**. (2021) 3. Ministry of Health and Family Welfare, Government of India. **Towards Universal Health Coverage: Ayushman Bharat Health and Wellness Centres. A Compendium of Health & Wellness Centres Operationalization**. (2020) 4. NITI Aayog, Government of India. **Aspirational Districts Baseline Ranking Map**. (2020) 5. **Office of the Registrar General of India: Provisional population totals: Haryana, census**. (2018) 6. Smith GS. **Development of rapid epidemiologic assessment methods to evaluate health status and delivery of health services**. *Int J Epidemiol* (1989) **18** 0-15 7. Robertson SE, Valadez JJ. **Global review of health care surveys using lot quality assurance sampling (LQAS), 1984-2004**. *Soc Sci Med* (2006) **63** 1648-1660. PMID: 16764978 8. Myatt M, Mai NP, Quynh NQ. **Using lot quality-assurance sampling and area sampling to identify priority areas for trachoma control: Viet Nam**. *Bulletin of the World Health Organization* (2005) **8** 756-763 9. Lemeshow Lemeshow, Stanley; George Stroh, Jr Jr. **Sampling Techniques for Evaluating Health Parameters in Developing Countries**. (1988) 10. World Health Organization. **Monitoring Immunization Services Using the Lot Quality Technique. Global Programme for Vaccines and Immunization**. *Vaccine Research* (1996) 11. Beckworth CA, Anguyo R, Kyakulaga FC, Lwanga SK, Valadez JJ. **Can local staff reliably assess their own programs? A confirmatory test-retest study of Lot Quality Assurance Sampling data collectors in Uganda**. *BMC Health Serv Res* (2016) **16** 396. PMID: 27534743 12. Wani RT. **Socioeconomic status scales-modified Kuppuswamy and Udai Pareekh's scale updated for 2019**. *J Family Med Prim Care* (2019) **8** 1846-1849. PMID: 31334143 13. Nambiar D, Bhaumik S, Pal A, Ved R. **Assessing cardiovascular disease risk factor screening inequalities in India using Lot Quality Assurance Sampling**. *BMC Health Serv Res* (2020) **20** 1077. PMID: 33238995 14. National Health Mission, Ministry of Health and Family Welfare, New Delhi. **Update on ASHA Programme**. (2019) 15. Siddharth Tiwari. **Times of India (Gurgaon News): Thousands in Mewat defy Section 144, join anti-CAA stir**. (2021)
--- title: 'Perceived impact of COVID-19 on routine care of patients with chronic non-communicable diseases: a cross-sectional study' authors: - Desalew Tilahun Beyene - Kenenisa Tegenu Lemma - Samira Awel Sultan - Ismael Ahmed Sinbiro journal: The Pan African Medical Journal year: 2022 pmcid: PMC10038762 doi: 10.11604/pamj.2022.43.212.32769 license: CC BY 4.0 --- # Perceived impact of COVID-19 on routine care of patients with chronic non-communicable diseases: a cross-sectional study ## Abstract ### Introduction patients with chronic non-communicable diseases (chronic liver diseases, chronic respiratory diseases, neurologic diseases, chronic kidney diseases, cardiovascular diseases, diabetes mellitus, and hypertension), primarily poor, rural and neglected populations, have had difficulty accessing health care and have been severely impacted both socially and financially in during the pandemic. As a result, this study was designed to assess the perceived impact of COVID-19 on routine care of chronic non-communicable disease patients in Ethiopia. ### Methods a cross-section survey was conducted among 404 participants from April 1st 2021 to May 30th 2021. Data were collected via interviewer administered questionnaires administered by pre-tested interviewers on socio-demographic characteristics, treatment and clinical features and routine care questionnaires that have been adapted and modified from different literatures. The study consisted of all adult outpatients with at least one chronic non-communicable disease who were followed up. Data were analyzed using the Statistical Package for Social Sciences Version 23. ### Results of the 422 participants, 404 responded for a response rate of $95.7\%$. One out of two (203, $50.2\%$) participants was aged 40 to 50 years. Ninety-one out of hundred (367, $90.8\%$) participants continued to receive routine care face-to-face during COVID-19. One-third (141, $34.9\%$) of study participants had good management of the chronic non-communicable diseases care in the middle of pandemic. A total of 167($41.34\%$) participants thought they were moderately affected changes in healthcare services since the COVID-19 outbreak. Nearly one-third (130, $32.2\%$) of participants were sometimes affected by medication shortages since the start of COVID-19. ### Conclusion this study highlights that most participants continued to receive routine care face-to-face during the COVID-19. About forty-one out of 100 participants perceived that they were moderately affected changes in healthcare services since the outbreak of COVID-19. One-third of participants sometimes perceived that they were affected by medication shortages since the start of COVID-19. ## Introduction The coronavirus (COVID-19) pandemic began in Wuhan, China at the end of 2019 and spread to a dozen countries in early 2020. Since then, the number of cases has continued to rise exponentially throughout the world [1], including in Ethiopia. Today, COVID-19 is a global pandemic with unprecedented health, economic and social consequences for countries around the world [2]. In the past few years, the COVID-19 outbreak has significantly changed the rhythm of human life and overwhelmed the healthcare systems of many countries including developed countries) [1]. Ethiopia experienced its first case of COVID-19 in March 2020. The infection has spread across all regions of the country. In response, the government has taken a number of actions to prevent the spread of this disease. Locking down all schools, declaring social distancing and hand hygiene, restricting large gatherings, limiting travel, preparing health facilities for treatment and quarantining individuals who had known contact and travel history for 5 days were some of the actions the Ethiopian government took [3]. This in turn worsens the health care system, standard of living and cultural norms of people in developing countries such as Ethiopia. Reallocating funds to fight COVID-19 could undermine gains already made to control non-communicable diseases [4]. Routine chronic non-communicable diseases care represents a significant challenge during this pandemic. Continuing routine care for people with chronic non-communicable diseases despite the pandemic is important to prevent long-term impacts on health and mortality [5]. The COVID-19 pandemic has prompted the world to explore innovative ways of continuing outpatient care [3] including case identification, contact tracing, isolation and quarantine, which are actions taken to control the spread of the disease in addition to the preventive measures put in place mainly promoting social distancing and sanitary measures [2]. In developed countries many patients with chronic non-communicable diseases can communicate with their own health care professionals especially during this pandemic through internet-based programs and applications (telemedicine), unlike developing countries such as Ethiopia. Not only that internet is not available to many parts of the country, but even in places that have internet connectivity the people may lack internet literacy. In addition, there is a lack of data on the perceived impact of the pandemic on the routine care of patients with chronic conditions in healthcare settings in the current study area. As a result, this study aimed to assess the impact of the pandemic on the routine care of patients with chronic conditions at the public hospital in the *Jimma area* of Ethiopia. ## Methods Study area and period: the study was conducted in public hospitals in Jimma zone, south-west Ethiopia from April 1st 2021 to May 30th 2021. Jimma zone is located 350 km from Addis Ababa. The zone is divided into 18 districts and a town administration. There are eight hospitals (one referral and teaching hospital, one general hospital and six primary hospitals), 115 health centres and 520 health posts in Jimma zone. An estimated 15 million people from within the country as well as from border zones and South Sudan avail health services from eight public hospitals in Ethiopia, with Jimma Medical Center (JMC) being a major service provider. During the study period, there were a total of 5800 patients on follow-up in the four public hospitals in the South west of Ethiopia i.e. 2000 patients with chronic non-communicable diseases at JMC, 1300 at Agaro Hospital, 1000 Seka hospital and 1500 Shenen Gibe Hospital. Study design, source population and study population: facility based cross-sectional study was used. All patients on chronic follow-up at public hospitals in the Jimma zone, southwest Ethiopia were used as the source population. Eligibility criteria: all Adult outpatients with at least one chronic illness on chronic follow-up at public hospitals in Jimma zone who visited chronic follow-up during the study period were included and severely sick patients in need of immediate medical intervention were excluded. Sample size determination and sampling technique: sample size was determined by using a single population proportion formula by considering a $50\%$ proportion as there is no previous study. The formula for calculating the sample size (n) is as follows: Where, n= Minimum sample size, p= an estimate of the prevalence rate for the population, d= the margin of the sampling error and Zβ/2= standard normal variance (1.96)2 is mostly $5\%$, that is with a $95\%$ confidence level. n = (1.96)2x0.5(1-0.5)/(0.05)2= 384 By considering $10\%$ non-response rate the final sample size is 422. Sampling procedures: Figure 1 shows the sampling procedure. There were eight public hospitals in Jimma zone, and four hospitals from eight were selected using simple random sampling. Health posts and health centers were excluded because there was no chronic follow-up in the health posts and even though health centers had chronic follow-up, they had small number of patient flow as patients don't get adequate services they were referred to nearby hospital. The final sample size for the study was proportionally allocated to each selected hospital. An individual study participant was selected using systematic random sampling techniques from the chronic follow-up registration book of each hospital with a K-value=13. **Figure 1:** *diagramatic representation of sampling procedure; where NF = final sample size (422), Ni = number of patience in each hospital and N = total number of patients in the time of data collection period in four hospitals chronic non-communicable disease (CNCD) = 5800* Data collection tools: a pre-tested, structured interviewer-administered questionnaire was used for data collection which contained three parts including socio-demographic characteristics, clinical and treatment related characteristics and were used to assess the perceived impact of COVID-19 on routine care which was adapted and modified from different literature [3-6]. Perceived impact of COVID-19 was measured by 5-items [7] and each item score was summed up and finally described as described in Table 3. The instrument was translated to the Afaan Oromo and Amharic versions by experts who were fluent in both languages and back translated to English for consistency. The Afaan Oromo and Amharic versions questionnaire were used for the data collection. Data collection procedure: interviewer-administered structured questionnaire and card review were used to collect data. Two Bachelor of Science degree (BSC) holder nurses and one Masters of Science (MSC) nurse were recruited as data collectors and supervisor respectively. Data quality control: both data collectors and supervisor were trained for one day on the objectives of study and data collection techniques. A supervisor checked the completeness and consistency of the questionnaire. The principal investigator evaluated the data before analysis to verify the completeness of the collected data. The pre-test was performed with $5\%$ of the sample size participants prior to actual data collection at the Limmu hospital chronic illness clinic follow up to assess the clarity and reliability of the data collection tool. Limmu hospital was chosen as the pilot study because it is far from the selected hospitals which in turn helps to prevent information dissemination/contamination and second it has similar sociocultural background participants with approximate patient flow and services in comparison with the study hospitals. Data from the pre-test were not included in the actual data. Scale reliability was assessed using the internal consistency of the reliability test. To date, all constructs of routine health care, (RHC) questionnaire reported excellent reliability with an overall alpha value of 0.90. Data processing and analysis: the data were entered and coded in Epi-Data version 3.1. Data were exported to SPSS version 23 for analysis. Descriptive statistics were computed. Ethical consideration: prior to data collection, ethical approval was obtained from the Institutional Review Board (IRB$\frac{005}{2021}$) of Jimma University and administrative permission was also obtained from JMC. During data collection, written informed consent was obtained from the literate participants and oral consent from the illiterate participants. The study participants were briefed on the study objectives and right to withdraw at any point. Data were collected anonymously. Operational definition Chronic non-communicable diseases: (chronic liver diseases, chronic respiratory diseases, neurologic diseases, chronic kidney diseases, cardiovascular diseases, diabetes mellitus, and hypertension) [8]. Perceived impact of COVID-19: was measured by 5-items [7] and each item score was summed up and finally described as described in the Table 3. Variables: socio-demographic variables (age, sex, educational status, marital status, residence, occupational status and household monthly income) and clinical and treatment related characteristics (duration of treatment, presence of respiratory symptoms in the past 14 days, travel history to other areas during COVID-19 outbreak, contact history with a known COVID-19 case, source of medication and presence of comorbidity). Funding: Jimma University covered only the survey cost for this study and there is no funding organization. This funding organization had no role in the design of the study, collection, analysis, and interpretation of data, or in writing the manuscript. Availability of data and materials: due to the lack of consent from the study participants to disclose raw data, these data could not be made available to protect the participants' identities. ## Results Sample characteristics: Table 1 presents the sample characteristics indicating that about half ($50.2\%$) of the participants were in the age range of 40-50 years. More than half ($57.4\%$) of participants were female. Only over one-third ($39.6\%$) of the study participants had no education. More than three-forth ($79.5\%$) of the participants were married and one in two ($52.5\%$) of participants were urban residents. **Table 1** | Variable | category | Frequency | percent | | --- | --- | --- | --- | | Age | 18-39 years old | 99 | 24.5 | | Age | 40-59 years old | 203 | 50.2 | | Age | ≥60 years | 102 | 25.2 | | Sex | Female | 232 | 57.4 | | Sex | Male | 172 | 42.6 | | Educational status | | 160 | 39.6 | | Educational status | Primary or 1-8 grades | 141 | 34.9 | | Educational status | Secondary and higher | 103 | 25.5 | | Marital status | Married | 321 | 79.5 | | Marital status | Single | 49 | 12.1 | | Marital status | Widowed | 32 | 7.9 | | Marital status | Divorced | 2 | 0.5 | | Residence | Urban | 212 | 52.5 | | Residence | Rural | 192 | 47.5 | | Occupational status | Famer | 122 | 30.2 | | Occupational status | Housewife | 119 | 29.5 | | Occupational status | Government employee | 83 | 20.5 | | Occupational status | Merchant | 56 | 13.9 | | Occupational status | Labor worker | 24 | 5.9 | | Household monthly income | <500 Birr | 81 | 20.0 | | Household monthly income | 500-1000 Birr | 53 | 13.1 | | Household monthly income | >1000 Birr | 270 | 66.8 | Clinical and treatment related characteristic: Table 2 summarizes clinical and treatment related characteristics of the participants. More than half ($58.9\%$) of them had been on treatment for less than five years Most ($80.2\%$) participants did not have a contact history with a known COVID-19 case. **Table 2** | Variable | category | Frequency | percent | | --- | --- | --- | --- | | Duration of treatment | <5 Years | 238 | 58.9 | | Duration of treatment | 5-10 Years | 94 | 23.3 | | Duration of treatment | >10 Years | 72 | 17.8 | | Presence of respiratory symptoms in the past 14 days | Yes | 80 | 19.8 | | Presence of respiratory symptoms in the past 14 days | No | 324 | 80.2 | | Travel history to other areas since your last visit | Yes | 34 | 8.4 | | Travel history to other areas since your last visit | No | 370 | 91.6 | | Contact history with a known COVID-19 case | Yes | 46 | 11.4 | | Contact history with a known COVID-19 case | No | 358 | 88.6 | | Source of medication | Free | 119 | 29.5 | | Source of medication | Payments | 285 | 70.5 | | Presence of comorbidity | Yes | 142 | 35.1 | | Presence of comorbidity | No | 262 | 64.9 | Routine Health Care (RHC): Table 3 describes the participants' routine healthcare services. Ninety-one out of hundred ($90.8\%$) participants continued to receive routine care face-to-face during COVID-19. Only more than one-third ($34.9\%$) of the study subjects had poor management of the chronic disease care since the outbreak of COVID-19 while the majority ($41.34\%$) participants thought they were moderately affected by changes in healthcare services since the outbreak of COVID-19. Just one third of participants were sometimes impacted by medication shortages since the start of COVID-19. More than half of participants reported that their health condition was worsened since the outbreak of the COVID-19. Figure 2 demonstrates chronic non-communicable diseases and comorbidities most affected by COVID-19. Over a quarter of the patients had hypertension. Most participants had at least one co-morbidity. ## Discussion COVID-19 is a global public health emergency unprecedented in modern history [9] that has had direct and indirect effects on people with chronic non-communicable diseases. In addition to morbidity and mortality, high rates of community spread and various mitigation efforts, including stay-at-home recommendations, had disrupted lives and created social and economic hardship [10]. It is worthwhile that routine care continues despite the pandemic, to avoid a rise in non-COVID-19 related deaths and morbidity [3]. However, there is a paucity of information on perceived impact of COVID-19 on routine care for people with chronic conditions in the study area. Thus, to our knowledge, this is the first study in the area. Overall, most participants continued to receive routine care face-to face during COVID-19 which is higher than study conducted in United Kingdom [3]. This discrepancy might be due to most participants had no education which deprived them of opportunities to undergo follow-up via phone calls, telemedicine, videos and other social media which in turn imposes them to get routine care face-to-face which is risky to contract COVID-19. Over one-third of study participants had poor management of chronic non-communicable diseases care since the COVID-19 outbreak. This means that the pandemic diverted the attention of world to adapting to new ways of delivering care using telemedicine to reduce face-to-face contact. Adapting new ways of virtual healthcare and digital technologies is imperative to allow HCPs to continue routine appointments and the use of apps can support the self-management of chronic conditions. Nonetheless, most people with non-communicable diseases live in low-middle income countries, where these technologies may not be widely available or practical [3] including Ethiopia. Furthermore, those with comorbidities may depend heavily on regular check-ups or hospital appointments to manage risk factors or experience delays in treatment which may potentially have severe consequences [3]. However, majority participants thought they were moderately affected by changes in healthcare services since COVID-19 outbreak. People with chronic conditions are not only affected by the COVID-19 pandemic in a direct manner, but also in an indirect manner, that is, People with chronic conditions focus more on contracting infection than the existing disease. The COVID-19 pandemic has disrupted all societies, including routine health care systems [11]. Resources at all levels shifted away from chronic disease management and prevention during the outbreak. This leads to serious concerns about the indirect health impact of COVID-19, especially on chronic non-communicable diseases with increased complications and augmented progression due to delayed and diminished access to care and disruption in follow-up at the primary care level [11]. One-third of the participants were sometimes affected due to medication shortages since the start of COVID-19. Because of its high transmissibility and capacity to disrupt international travel and business, the pandemic created travel and business restrictions and was responsible for [12] impact on people with chronic conditions by medication shortages. This might also be because resource allocation at all levels has shifted away from chronic disease management and prevention during the outbreak [11]. More than half of the participants reported that their health had worsened since the start of the COVID-19 pandemic. This means that the pandemic has put patients with chronic conditions in complications and difficulties in routine medical care due to delayed transportation, shortage of medications, and human resources are among the contributing factors for patients with chronic non-communicable diseases to suffer from worsened health conditions during infectious disease pandemic [13]. The study was limited to and only included those who were on follow-up. The study is also limited to descriptive analysis; thus, rigorous analysis was not conducted to control for confounding variables, and the association between outcome and predictors was not done because of its descriptive nature as indicated in the prior study. As this was a single center cross-sectional study without a robust analysis, the generalizability of study is questionable. ## Conclusion In conclusion, people with chronic conditions in low income countries including Ethiopia are disproportionately affected by the pandemic as they continue to receive face-to-face routine care. Thus, healthcare providers (HCP) should provide meticulous care to patients who receive routine care face-to-face to curb the transmission of the pandemic and manage the existing chronic condition. On the top of this, HCP should offer all inclusive patient centered care for all patients not only to treat the pandemic but also to tackle chronic conditions related comorbidity, mortality and complications. ## What is known about this topic Nowadays, COVID-19 is a global pandemic with unprecedented medical, economic and social consequences affecting nations across the world including Ethiopia; In developed countries many patients with chronic non-communicable diseases can communicate with their own health care professionals especially during this pandemic through internet-based programs and applications (telemedicine), unlike for developing countries like such as Ethiopia, Ethiopia; Routine care for chronic disease is an ongoing major challenge during pandemic. ## What this study adds In our study, most participants continued to receive routine care face-to-face during COVID-19; One-third of the patients reported being inconvenienced due to shortage of medications; Care was unaffected for every one in three patients. ## Competing interests The authors declare no competing interests. ## Authors' contributions Desalew Tilahun Beyene, Kenenisa Tegenu Lemma, Samira Awel Sultan and Ismael Ahmed Sinbiro made substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data; took part in drafting the article or revising it critically for important intellectual content; agreed to submit to the current journal; gave final approval of the version to be published; and agreed to be accountable for all aspects of the work. All authors read and approved the final version of the manuscript. ## References 1. Guo D, Han B, Lu Y, Lv C, Fang X, Zhang Z. **Influence of the COVID-19 Pandemic on Quality of Life of Patients with Parkinson´s Disease**. *Parkinsons Dis* (2020.0) **2020** 1216568. PMID: 33062247 2. Haftom M, Petrucka P, Gemechu K, Mamo H, Tsegay T, Amare E. **Knowledge, Attitudes, and Practices Towards COVID-19 Pandemic Among Quarantined Adults in Tigrai Region, Ethiopia**. *Infect Drug Resist* (2020.0) **13** 3727-37. PMID: 33116693 3. Chudasama YV, Gillies CL, Zaccardi F, Coles B, Davies MJ, Seidu S. **Impact of COVID-19 on routine care for chronic non-communicable diseases: A global survey of views from healthcare professionals**. *Diabetes Metab Syndr* (2020.0) **14** 965-7. PMID: 32604016 4. Lim SL, Woo KL, Lim E, Ng F, Chan MY, Gandhi M. **Impact of COVID-19 on health-related quality of life in patients with cardiovascular disease: a multi-ethnic Asian study**. *Health Qual Life Outcomes* (2020.0) **18** 387. PMID: 33317547 5. Schwartz MR, Oppold P. **The Impact of Assistive Technologies on Quality of Life and Psychosocial Well-Being**. *Psycho-Social Perspectives on Mental Health and Well-Being: IGI Global* (2020.0) 161-76 6. Abuduxike G, Aşut Ö, Vaizoğlu SA, Cali S. **Health-seeking behaviors and its determinants: a facility-based cross-sectional study in the Turkish Republic of Northern Cyprus**. *Int J Health Policy Manag* (2020.0) **9** 240-249. PMID: 32613792 7. Sahoo KC, Kanungo S, Mahapatra P, Pati S. **Non-communicable diseases care during COVID-19 pandemic: A mixed-method study in Khurda district of Odisha, India**. *Indian J Med Res* (2021.0) **153** 649-657. PMID: 34380806 8. Organization WH. **WHO global coordination mechanism on the prevention and control of noncommunicable diseases: final report: WHO GCM**. *World Health Organization* 201 9. Ornell F, Halpern SC, Kessler FHP, Narvaez JCdM. **The impact of the COVID-19 pandemic on the mental health of healthcare professionals**. *Cad Saude Publica* (2020.0) **36** e00063520. PMID: 32374807 10. Hacker KA, Briss PA, Richardson L, Wright J, Petersen R. **Peer Reviewed: COVID-19 and Chronic Disease: The Impact Now and in the Future**. *Prev Chronic Dis* (2021.0) **18** E62. PMID: 34138696 11. Danhieux K, Buffel V, Pairon A, Benkheil A, Remmen R, Wouters E. **The impact of COVID-19 on chronic care according to providers: a qualitative study among primary care practices in Belgium**. *BMC Fam Pract* (2020.0) **21** 255. PMID: 33278877 12. Amimo F, Lambert B, Magit A, Hashizume M. **A review of prospective pathways and impacts of COVID-19 on the accessibility, safety, quality, and affordability of essential medicines and vaccines for universal health coverage in Africa**. *Global Health* (2021.0) **17** 42. PMID: 33832487 13. Mohammedhussein M, Hajure M, Ebrahim J, Dule A. **Posttraumatic Stress Symptoms Among Patients with Chronic Medical Disease Amid Covid-19 Pandemic in Southwest Ethiopia**. *International Journal of Biomedical Engineering and Clinical Science* (2021.0) **7**
--- title: 'Challenges and enablers for implementation of WHO ‘Best buys’ interventions targeting risk factors of diabetes and hypertension in South Africa: a mixed methods study' authors: - Jeannine Uwimana Nicol - Lynn Hendricks - Taryn Young journal: The Pan African Medical Journal year: 2022 pmcid: PMC10038766 doi: 10.11604/pamj.2022.43.215.31547 license: CC BY 4.0 --- # Challenges and enablers for implementation of WHO ‘Best buys’ interventions targeting risk factors of diabetes and hypertension in South Africa: a mixed methods study ## Abstract ### Introduction World Health Organization (WHO) recommends the implementation of ‘Best buys’, cost-effective interventions that address risk factors of non-communicable diseases (NCDs). However, country responses to the implementation of these have been slow and undocumented. The aim of this study was to identify and understand challenges and enablers for the implementation of WHO ‘Best buys’ for risk factors of diabetes and hypertension in South Africa (SA). ### Methods a mixed methods research with a sequential dominant status design was used starting with a document review to take stock of WHO ‘Best buys’ in policy in SA. A qualitative study using key informant interviews was then done to identify and understand challenges and enablers for implementation. A triangulation approach for the analysis of both document review and qualitative data was used. ### Results SA has made good progress in including the WHO ‘Best buys’ in the policy. However, several challenges hamper the successful implementation. Most challenges were related to upstream policy implementation processes such as competing interests of actors, lack of balance between economic vs health gains, and lack of funding. Enabling factors included multi-sectoral engagement and collaboration; community ownership and empowerment; building partnerships for co-creation of enabling environments; leveraging on the existing infrastructure of other health programs; contextualization of policies and programs; and political will and leadership. ### Conclusion SA has made good progress in including WHO ‘Best buys’ targeting risk factors of diabetes and hypertension in policy, however, various contextual barriers influence effective implementation. Hence, there is a need to leverage enabling factors to foster the implementation of WHO ‘Best buys’ interventions targeting risk factors of diabetes and hypertension in South Africa. ## Introduction Non-communicable diseases (NCDs) claim about 41 million deaths each year [1] with hypertension and diabetes being major contributors to morbidity and mortality worldwide [2]. The number of people living with diabetes in Sub-Saharan Africa (SSA) is now estimated at 12.1 million people, and this number is expected to rise rapidly, reaching 23.9 million by 2030 [2]. In South Africa (SA), both hypertension and diabetes, are major contributors to morbidity and mortality [3]. In 2013, the SA government committed to reduce by at least $25\%$ the relative premature mortality (under 60 years of age) from NCDs by 2020 through its strategic plan for NCDs [4]. Both population-level interventions addressing modifiable risk factors and integrated health service delivery platforms can contribute to reduction of morbidity and mortality [5]. Some of the modifiable risk factors for cardiovascular disease (CVD) and diabetes include poor diet, insufficient physical activity, and excess body weight. Thus, the risk factor distributions at the population-level can potentially be changed by changing the lifestyle of people, the environments where they live or work, that mitigate the lack of physical activity, smoking, and diet. The United Nation's sustainable development goals (SDGs) included the NCDs with a target of reducing premature mortality by a third in 2030 [6]. But the current trend in decline of global rate of premature deaths from NCDs is not enough to meet this target [7] amidst of existing evidence based interventions such as the WHO ‘Best buys’, which have proven to contribute to the reduction of premature mortality from NCDs if they are well implemented [1]. WHO ‘Best buys’ endorsed in 2017 comprised of 16 practical and cost-effective interventions that work and can be delivered at the primary health level. Critically, these interventions put the emphasis on promoting health and preventing disease, and include interventions such as increasing tobacco taxes; restricting alcohol advertising; reformulating food products with less salt, sugar and fat; vaccinating girls against cervical cancer; treating hypertension and diabetes [1]. Country responses to implementation of WHO ‘Best buys’ have been low and uneven [4,8]. This could be due to various factors including contextual and policy agenda setting process. Breda and colleagues argue that most of the countries particularly in low- and middle-income countries (LMICs) do not have the capacity to implement these global policies and interventions at scale [8]. In addition, contextual factors that have hampered implementation of global policies or interventions such as the WHO ‘Best buys’ include lack of proper adaption of these global policies to fit the specific social, cultural, economic, political, legal, institutional settings and physical environments in which their being implemented [8,9]. Hence, it's important when global policies and, or interventions are implemented at country level to have engagement of stakeholders from all levels of policy making cycle, prevention and management of diseases form the onset of the planning stages, and across all stages of implementation process, to ensure appropriate and effective implementation [8,10-12]. This is paramount for the implementation of health interventions, in particular, given the multi-dimensional determinants of chronic conditions requiring interventions beyond health. Beside lack of multi-sectoral collaboration as a barrier for effective implementation, other factors such conflicting priorities or agenda between different sectors of the same government often leads to disagreements on acceptable interventions and policies [8,13]. Lack of dedicated resources and investment to support coordinated implementation, monitoring and evaluation of NCDs interventions were highlighted in studies conducted in Zambia, Bangladesh and Iran. This study aimed to determine challenges and enablers for implementation of WHO ‘Best buys’ for risk factors of diabetes and hypertension in SA. ## Methods Study design and setting: this is a fully mixed sequential dominant status design [14], whereby a document review was conducted first, followed by a qualitative study. Mixed methods research approach [15,16] was used in order to gain a better understanding on the implementation of the existing policies and programs in South Africa in line with the WHO ‘Best buys’ [1] to address risk factors for diabetes and hypertension, and challenges and opportunities for its implementation. The WHO Global Strategy for Diet, Physical Activity, and Health (DPAS) was used to conceptualize and visualize the kinds of interventions of interest [17]. The DPAS strategy organizes intervention into three types that focus on supporting policies, programs, and the environment programs. These three categories of interventions were aligned with the WHO ‘Best buys’ to inform data collection tools and analyses. The document review focused on identifying and reviewing the existing policy documents (policies, acts, regulations, guidelines, reports, etc.) and programs that addresses NCDs in SA with emphasis on policies and programs targeting risk factors of diabetes and hypertension. A review of relevant reports, journal articles, or policy documents was conducted from November 2019- March 2020. Government reports that contain information regarding the planning, implementation and evaluation of population-level interventions targeting diabetes and hypertension were considered. Various databases were used to run the search and there was no restriction of date and language applied to the search. The document review informed the mapping of stakeholders to be involved in the key informant interviews (KIIs) by identifying experts in supportive policies, programs and enabling environments for NCD prevention in South Africa. The KIIs were conducted to get a deep understanding of challenges and enabling factors for implementation of population WHO ‘Best buys’ in SA. Data collection: for the document review, various databases were searched and these included Pubmed, Ebscohost, Google scholar and Scopus and Cochrane database. Search terms used included diabetes /hypertension/ physical activity/ nutrition/ alcohol consumption/ tobacco smoking/ programs/ interventions/ South Africa. There was no restriction of date and language applied to the search. An excel spreadsheet was designed to extract relevant information on the type of intervention, coverage, target audience, and if there have been any process or impact evaluation. A total of 13 participants who were representative of key stakeholders at national and provincial government departments and agencies involved in NCDs related programs as well as representatives of NGOs and food industry; and representatives of academic and research institutions were purposively selected and interviewed. These interviews took place in April- June 2021 using Microsoft Teams or Zoom due to the COVID-19 restrictions. Interviews were transcribed verbatim, captured, and coded using NVivo [18]. Data analysis and management: a triangulation approach for analysis of both document review and qualitative data was used. The data were compared and contrasted in order to achieve as rich picture of the situation as possible and to increase credibility the study. The data from the document review were analysed narratively. While the qualitative data were analysed thematically using the five stages of the framework approach as described by Pope and Mays [19]. The framework was chosen as it enables the researcher not only to focus on the pre-defined set of questions, but also to consider other themes emerging from the analysis. An analytical grid of key themes was developed in lined with the study aims and familiarisation with the first few transcripts, and then applied to the rest of the transcripts. The WHO DPAS [17] strategy was used as thematic framework for analysis which organizes intervention into three types that focus on supporting policies, programs and the environment programs. These thematic areas were applied to the set of interventions outlined in the WHO ‘Best buys’ [1] by exploring their implementation in South Africa. A constant comparison process was used during analysis, where all relevant data for each category of participants were identified, examined and compared with the rest of the data, to identify dominant themes occurring in all groups. The non-dominant themes were also grouped separately and compared across categories of participants to identify similarities and inconsistencies. Codes and themes were discussed between two authors (Jeannine Uwimana Nicol and Lynn Hendricks) until consensus was reached. To ensure trustworthiness and credibility of qualitative data procedures described by Mays and Pope [19], triangulation of data was observed to enhance trustworthiness of qualitative data [20]. Funding: the review work was supported by the funding from the Collaboration for Evidence-based Healthcare and Public Health in Africa (CEBHA+) project which is funded by the German Federal Ministry of Education and Research (BMBF) as part of the Research Networks for Health Innovation in sub-Saharan Africa Funding Initiative. The funder doesn´t have any role in the review process. Ethics approval: ethics approval for this study was gained from The Health Research Ethics Committee at Stellenbosch University (N$\frac{19}{01}$/001). Consent for participation was provided by participants through a signed consent form. ## Results Twenty-eight policies, legislations, strategic plans and regulations were identified - 8 policies on tobacco use (smoking); 7 policies on harmful consumption of alcohol; 8 policies on unhealthy diet and 5 policies of physical inactivity (PA). Thirteen supportive programs were identified of which 6 targeted unhealthy diet, 3 tobacco smoking and 4 targeted PA. These reviewed documents indicate that all the WHO ‘Best buys’ recommended interventions have been enacted and included in policy in SA (Table 1). **Table 1** | Risk factors | Interventions | Progress | | --- | --- | --- | | Risk factors | Interventions | | | Tobacco use | Tobacco price increases (tax increases) | xxx | | Tobacco use | Smoke -free-indoor workplace and public places | xxx | | Tobacco use | Health information and warnings | xxx | | Tobacco use | Bans on tobacco advertising, promotion and sponsorship | x | | Harmful alcohol use | Alcohol price increase (tax increases) | xxx | | Harmful alcohol use | Restricted access to retailed alcohol (purchase age, restricting locations and hours, government monopoly) | x | | Harmful alcohol use | Bans on alcohol advertising and sponsorship | xxx | | Unhealthy diet and physical inactivity | Food reformulation to reduce salt content (reduced salt intake in food) | xxx | | Unhealthy diet and physical inactivity | Food reformulation to exclude saturated and trans fats | xxx | | Unhealthy diet and physical inactivity | Fiscal measures that increase the price of unhealthy foods or decrease the price of healthy foods | xxx | | Unhealthy diet and physical inactivity | Food labelling restrictions on marketing of unhealthy foods and beverages | x | | Unhealthy diet and physical inactivity | Mass-media campaigns to reduce salt consumption | xx | | Unhealthy diet and physical inactivity | Public awareness through mass media on diet and physical activity | xx | | Unhealthy diet and physical inactivity | Modification of built environment to promote physical activity | xx | | Specific intervention on diseases | Specific intervention on diseases | | | Cardiovascular diseases and diabetes | Counselling and multi-drug therapy for people with a high risk of developing heart attacks and strokes (including those with established CVD) | xxx | | Cardiovascular diseases and diabetes | Treatment of heart attacks with aspirin | Xxx | Challenges for implementation of WHO ‘Best buys’: governmental and industry stakeholders agreed that national policy implementation varied across provinces in South Africa, with some having more organized provincial structures and doing better than others. The predominant themes identified as challenges to the implementation of supportive policies and programs are summarised in Figure 1 and Table 2 provides challenges with illustrative quotes. Through the KIIs, the predominant themes identified as challenges to implementation of WHO ‘Best buys’ included among others lack of multi-sectoral approach; different governance structures and non-uniformity of provincial implementation plan; competing interests and priorities among stakeholders; lack of financial and human resources; lack of monitoring and evaluation systems; lack of community ownership; and inadequate communication strategy to promote behaviour change. **Figure 1:** *challenges for implementation of WHO ‘Best buys’ interventions* TABLE_PLACEHOLDER:Table 2 Lack of financial and human resources: lack funding was identified as an important impediment for both policy formulation and policy implementation. Limited funding allocated to provinces and NGOs at large was identified as the main hindrance to effective implementation of population-based interventions. Some stakeholders working in the research industry expressed that lack of funding to do research relevant to the local health needs within the African context to inform policy has been a challenge. Competing interests and priorities among stakeholders: conflict of interest between the national trade and investment policy vis a vis the national health policy whereby trade, and foreign direct investment have a tendency of promoting the influx of large amounts of processed foods and sugary beverages, alcohol as well as smoking has been perceived by participants as among the challenges that hinder the implementation of both supportive policies and programs targeting diabetes and hypertension in particular the ones that target unhealthy diet, smoking and alcohol consumption. Also, stakeholders perceived that prioritization of policies differed across government departments and between national and provincial government. Sometimes prioritization of policies and programs to be implemented depends on the availability of resources which explains why some provinces have enabling environments to promote physical activities (i.e. cycling lane, outdoor gyms, parks, etc) and some do not have. Thus affect the implementation of supportive policies nationwide. When reflecting on the implementation of the Integrate Food and Nutrition Security Strategy, participants expressed that while the strategy was developed in consultation with various national government departments and agencies, the implementation plan could not be fully funded due to competing budget interests. “ Well, the implementation … okay, you know, we have drafted that [Integrate Food and Nutrition Security Strategy] implementation plan. When we finalized the plan we wanted it approved by cabinet. And they said no, no, where did you think, we are going to get all this money? And they said go back to your respective departments and reprioritise.” ( SH-AAD1) Lack of monitoring evaluation (M and E) system: lack of proper of M and E system for NCDs program has been identified as a challenge for effective policy and program implementation. Also, non-integration of NCDs key data elements within existing electronic M and E system (i.e DHIS, Tier.net) for other health program such as HIV has been problematic in measuring the effect of some policy interventions. “ TB and HIV have the Tear.net system and we have requested multiple times to just put in two additional fields to collect data for hypertension and diabetes, but you know there are so many patients who have co-morbidities and it would make such a difference because Tear.net generates their own reports. We would be able to get so much information based on the Tear.net system, unfortunately we were turned down so many times and even now with the revised Tear.net system, we weren't even offered that opportunity to include those two fields” (SH-ND2) Different governance structures and non-uniformity of implementation plan: governmental and industry stakeholders agreed that national policy implementation varied across provinces in SA, with some having more organized provincial structures and doing better than others. One of the policy makers argue that “What we find at provincial level, is that different provinces have got different set ups, you find provinces that are well organised in terms of the structures, they are well integrated, there are various role players that are participating in those kinds of structures. But you go to the next province there isn't so much in terms of such extensive structures.” Lack of clear roles and responsibilities of actors: there was a perceived connection made by participants between policy implementation success and coordinating mechanisms with provision of clear guidelines on roles and responsibilities. Lack of such clear directives for implementation of these population-level interventions impact on the success of these interventions and at times could lead to tension among actors/implementers. “ One of the challenges in the implementation of the NCD strategy was lack of responsibility for by the NCD cluster overseeing the implementation. There was no time at that stage to develop an agreement on how the programmes would work together to contribute to the set targets. Also, the success of that strategy would have been largely dependent on a very strong coordinating mechanism both within the programmes within the department, but also between the department and other sectors outside the department” (SH-ND4). Balancing behavior change to reduce the risk of NCDs versus determinants of health: behaviour change at individual and community level is required to reduce the risk of NCDs, however, this is especially challenging. For example, the Salt Reduction *Policy is* one of the policies related to healthy diet that has been identified as a policy that engaged extensively the food industry in its formulation. But the policy's implementation process was largely dependent on a communication strategy promoting behaviour change with mass education and campaigns activities at community level. The implementation of the policy has been problematic due to the heavy cost involved in executing the communication strategy/programs. Also, another complexity in promoting behaviour change is the existence of determinants of health such as poverty, access to basic services such as housing that makes a difficult choice for people to opt for behaviour change such as healthy diet. Enablers for implementing WHO ‘Best buys’: the enabling factors for effective implementation of WHO ‘Best buys’ most predominantly identified by participants across all categories included multi-sectoral engagement and collaboration; community ownership and empowerment; building partnerships for co-creation of enabling environments; leveraging on existing infrastructure of other health programs; contextualisation of policies and programs; community driven activism; balancing economic versus health gains; and political will and leadership. Table 3 summarises enabling factors for implementation of WHO ‘Best buys’ with illustrative quotes. **Table 3** | Community ownership and empowerment | “If you are engaged with different communities and share with the communities' evidence about dangers of alcohol consumption…you are able to get by and work with us to implement policies especially awareness about dangers of alcohol misuse. We will strengthen communities to work together to fight against alcohol use.” “… If they are empowered to understand the dangers of alcohol misuse they are the ones who will strengthen the implementation of policy because they understand where the policy comes from and what is the output of the policy and the outcome” | | --- | --- | | Community driven activism for NCDs | “The NGOs are involved…we need them a lot to advocate to various stakeholders, so they do that a lot and because they also work with communities out there, they are able to mobilise communities and they bring in a lot of evidence as well which we might not have” | | Inter-sectoral partnerships and engagement | “WOW! has partnered with various companies including the Heart and Stroke Foundation” to “develop healthy tuckshop guidelines…”and …Kubeka bicycles and gave 3000 bicycles to learners in the Paarl area to help them, not only to be more physically active, but also to live a healthy life> | | Leveraging on existing infrastructure of other health programs | “Bringing it from the perspective of HIV and TB, which has got a platform already and happening, which you can just update to that platform and expand to NCDs. We have seen millions of people screened for HIV and we are missing also screening them for other conditions such as diabetes and hypertension that are equally important.” | | Building partnerships for co-creation of enabling environments | “In terms of physical activity, we started WOW! groups, you might have heard of, they offer a structured program to promote physical activity, not only in the home environment but in the community, and also at work- so it's different populations or different target groups…people meet, they walk, they network and it's just become quite a fun element” | | Creation of economic opportunities while promoting healthy living | “People are producing food now in order for them to sell and be able to get an income from it. So now if there isn't clear markets that has been achieved for these indigenous foods, people will not be so eager to produce them because there will not be any economic gains from it, it will not form part of the rural economist that we are talking about. People will be more eager to be involved in producing crops that are more likely to get them their income that they're looking for”. “…high concentration of alcohol outlets contributes to alcohol misuse but on the other hand tavern owners will see it as an opportunity to have businesses where you make decision about maybe density of alcohol outlets it is not a straightforward thing you have to balance between economy as well as health.” | | Establish mechanism to promote compliance of industries | “At private industries we need to make sure that, because people buy 90% of the food that people buy in South Africa comes from the food industry. So we have a role to play to make sure that we are spending more on nutritious food. And I think what consumers are increasingly asking for these foods anyway, so it is in our interest to make sure that we are providing nutritious food” | | Creating incentives that promote good performance | “on an annual basis we have what we call the WOW annual award ceremony, that now will be in it's 5th year next year and this is what we look at, we look at the whole year's M&E data, quantitative and qualitative. We identify the top performing WOW groups and also the champions who've really performed outstandingly and it's not about the biggest loser, it's not about the group with the highest number of participants, it's about how innovative you are and despite all the odds, that you still achieve and you make change…” | | Political Will and leadership | “There is that politically positive environment, even the district model approach…we should begin to see our political leaders supporting these initiatives …” | Multi-sectoral engagement and collaboration: most participants expressed that one of the key drivers of a successful policy implementation was engagement and collaboration between government departments and with civil society, researchers, and communities. On an inter-departmental government level, engagement and working transversally was found to promote policy implementation as expressed by one of the participants: “I think that is the reason why we are not making a huge, an impact, because really, if you look at it, it cannot be health alone.” Community ownership and empowerment: for communities, a shared sense of ownership enable implementation. Working together with communities also included regular meetings and being responsive to their ongoing and changing needs. Building partnerships for co-creation of enabling environment: creating an enabling environment for making healthier choices includes building sports and recreational facilities as well as safe spaces for children to play. As much as there are indication of enabling environment intervention tailored to risk factors of diabetes and hypertension across the country, most of these interventions focuses on promotion of physical inactivity (i.e. Outdoor Gyms, cycling lanes, public parks) and their coverage vary according to each province, district, etc. Leveraging on existing infrastructure of other health programs: some participants were of the view that existing health programs such as TB and HIV provide an opportunity for integrating interventions geared to NCDs prevention such as diabetes and hypertension as expressed by one participant: “Bringing it from the perspective of HIV and TB, which has got a platform already and happening, which you can just update to that platform and expand to NCDs. Because I could not understand they screen for HIV and not at the same time screen for hypertension and diabetes. It's just so simple during that same encounter, to actually take the blood pressure of people, you wouldn't add to the load.” Civil society and community led activism: using civil society as an inter-mediatory to advocate policy implementation to stakeholders was found to be an enabler, as was reported with the Control for Tobacco Products and Electronic Delivery Systems Bill [21]. ## Discussion South Africa has a comprehensive spectrum of supportive policies and programs targeting risk factors for diabetes and hypertension and NCD prevention at large since the post-apartheid era [22]. These supportive policies and programs targeting risk factors for diabetes and hypertension were designed in line with the WHO ‘Best buys’ [1] with a national coverage (Table 1). This study triangulated findings from a desk review and qualitative data to identify challenges and enabling factors for implementation of WHO ‘Best buys’ related to risk factors of diabetes and hypertension in South Africa. Effective implementation of these interventions have been hindered by various factors such as lack of multi-sectoral approach; different governance structures and non-uniformity of provincial implementation plan; competing interests and priorities among stakeholders; lack of financial and human resources; lack of monitoring and evaluation (M and E) systems including non-integration of NCDs data elements into existing health information systems such as DHIS, HIV electronic; lack of community ownership; and inadequate communication strategy to promote behavior change. These implementation challenges are not unique to South Africa. Lack of multi-sectorial approaches have also been reported by various researchers in LMICs [11,12,23,24]. Conflicts of priorities between government departments on regulations related the economic gains have been reported in other studies [22]. Malawi, Cameroun and Ghana reported how competing interests and priorities have hindered country progress on WHO FCTC and alcohol consumptions as well as unethical sponsoring of sports events by alcohol industry [11,23]. Furthermore, lack of M and E systems for NCDs in general has led to lack of baseline and representative data to assess the effectiveness of population level interventions and their impacts on NCD prevention and control have been reported in other countries [24,25]. Mukanu et al. [ 24] in their study on a review of NCDs policy in Zambia argue that there was no sufficient data to inform some of the activities stipulated in the Zambian NCD strategic plan such as salt reduction and increase in uptake of physical activity. Bourdeaux and colleagues argue that limited funds for NCDs prevention and control is due to the fact most LMICs countries get their funding from NGOs [21]. Moreover, lack of and limited funding for the implementation of NCDs strategic plans has a direct impact on establishment of M and E systems to assess the effectiveness and impact of the policies and programs emanating from these NCDs strategic plans. Although the findings of this study highlight a number of challenges hindering the implementation of WHO ‘Best buys’, there are also enabling factors that could potential influence the effective implementation of these WHO ‘Best buys’. The enabling factors most predominantly identified by participants across all categories include multi-sectoral engagement and collaboration; community ownership and empowerment; building partnerships for co-creation of enabling environments; leveraging on existing infrastructure of other health programs; contextualisation of policies and programs; community driven activism; balancing economic versus health gains; and political will and leadership. Juma et al. [ 11] in their study on NCDs prevention policy process in five African countries (South Africa, Malawi, Kenya, Cameroon and Nigeria) stress the importance of creating a coordination mechanism embedded in inter-sectorality/multi-sectoral approach to enable effective implementation of interventions targeting risk factors for NCDs including diabetes and hypertension. Haldane et al. [ 26] in their review argue that community engagement and empowerment is one the key factors for implementing strategies in health promotion and the prevention and control of chronic diseases such as diabetes and hypertension, and community participation contributes to favourable outcomes at the organizational, community and individual level. There are few studies in South Africa and in LMICS in general that have been reporting on the implementation of WHO ‘Best buys’ interventions targeting risk factors of diabetes and hypertension, and other NCDs [27]. This study highlights challenges and enabling factors related to the implementation of these WHO ‘Best buys’ interventions in South Africa in order to inform local interventions needed for effective implementation. Additionally, the uniqueness of this study lies in the research methods approach used where by multiple sources of data were used and the engagement with various actors involved in policy making and program implementation to provide comprehensive reflection of contextual challenges affecting effective implementation of WHO ‘Best buys’ in South Africa. But also inform possible areas of future research through the identification of enabling factors of implementing these interventions. However, we recognised the limitations of this study which is the lack of national quantitative data to determine process outcomes and impact of these WHO ‘Best buys’ interventions on diabetes and hypertension; and NCDs in general. ## Conclusion In light of the growing burden of diabetes and hypertension, and NCDs in South Africa, there is a need to tackle risk factors for diabetes and hypertension. SA has made a good progress in including WHO ‘Best buys’ targeting these risk factors in policy, however there are various contextual barriers influencing effective implementation of these interventions. It's important to leverage on enabling factors, such as bottom-up approach anchored in multi-sectoral approach, to foster policy implementation. ## What is known about this topic The burden of NCDs in LMICs including South *Africa is* arising particularly the burden related to diabetes and hypertension which poses a threat to already overstretched health systems and human capital development. South Africa has a plethora of policies targeting risk factors of diabetes and hypertension, and has been slow in achieving the SDGs targets related to NCDs. ## What this study adds This study reports on the implementation of WHO ‘Best buys’ interventions targeting risk factors of diabetes and hypertension in South Africa. This study highlights challenges and enabling factors for implementation of WHO ‘Best buys’ interventions targeting risk factors of diabetes and hypertension in South Africa which will inform the revised national strategic plan for NCDs [2020-2025]. ## Competing interests There authors declare no competing interests. ## Authors' contributions Jeannine Uwimana Nicol developed, conceptualised and wrote the paper. She equally conducted the search, screening of eligible documents; and developed a data extraction tool for the document review as well as the interview guides and analysed data. Lynn Hendricks did data extraction, conducted and analysed the qualitative data and contributed to writing of the findings. Taryn Young contributed to protocol development, data interpretation, and critical inputs on the manuscript drafts. All authors have read and approved the final version of the manuscript. ## References 1. **“Best buys” and other recommended interventions for the prevention and control of noncommunicable diseases**. *World Heal Organ* (2017.0) **17** 28 2. **Global Status Report On Noncommunicable Diseases 2014**. (2014.0) 3. Hofman Karen J, Lee R. **Inter-sectoral case study: Successful sodium regulation in South Africa**. (2013.0) 4. **Non-communicable diseases**. *Noncommunicable diseases progress monitor 2020* (2020.0) 5. Pretoria SA. **National Department of Health: Strategic Plan for the Prevention and Control of Non-communicable Diseases 2013-2017**. *Department of Health* (2013.0) 6. **Health in 2015: from MDGs, Millennium Development Goals to SDGs, Sustainable Development Goals** 7. Nishtar S, Niinistö S, Sirisena M, Vázquez T, Skvortsova V, Rubinstein A. **Time to deliver: report of the WHO Independent High-Level Commission on NCDs**. *Lancet* (2018.0) **392** 245-52. PMID: 29866374 8. Breda J, Wickramasinghe K, Peters DH, Rakovac I, Oldenburg B, Mikkelsen B. **One size does not fit all: Implementation of interventions for non-communicable diseases**. *BMJ* (2019.0) **367** 9. Bakhtiari A, Takian A, Majdzadeh R, Haghdoost AA. **Assessment and prioritization of the WHO “best buys” and other recommended interventions for the prevention and control of non-communicable diseases in Iran**. *BMC Public Health* (2020.0) **20** 333. PMID: 32171267 10. Uwimana J, Jackson D, Hausler H, Zarowsky C. **Health system barriers to implementation of collaborative TB and HIV activities including prevention of mother to child transmission in South Africa**. *Trop Med Int Heal* (2012.0) **17** 658-665 11. Juma PA, Mohamed SF, Matanje Mwagomba BL, Ndinda C, Mapa-Tassou C, Oluwasanu M. **Non-communicable disease prevention policy process in five African countries**. *BMC Public Health* (2018.0) **18** 961. PMID: 30168393 12. Shiroya V, Neuhann F, Müller O, Deckert A. **Challenges in policy reforms for non-communicable diseases: the case of diabetes in Kenya**. *Glob Health Action* (2019.0) **12** 1611243. PMID: 31117896 13. Ahaibwe G, Abdool Karim S, Thow AM, Erzse A, Hofman K. **Barriers to, and facilitators of, the adoption of a sugar sweetened beverage tax to prevent non-communicable diseases in Uganda: a policy landscape analysis**. *Glob Health Action* (2021.0) **14** 1892307. PMID: 33874854 14. Leech LA, Onwuegbuzie JA. **A typology of mixed methods research designs**. *Qual Quant* (2007.0) **43** 265-275 15. Greene JC, Caracelli VJ, Graham WF. **Towards a conceptual framework for mixed method evaluation designs**. *Educ Eval Policy Anal* (1989.0) **11** 255-74 16. O'Cathain A, Murphy E, Nicholl J. **Why, and how, mixed methods research is undertaken in health services research in England: a mixed methods study**. *BMC Health Serv Res* (2007.0) **7** 85. PMID: 17570838 17. **Global Strategy on Diet, Physical Activity and Health: a framework to monitor and evaluate implementation** 18. **Unlock insights with the leading qualitative data analysis software** 19. Pope C. **Qualitative research in health care: Analysing qualitative data**. *BMJ* (2000.0) **320** 114-6. PMID: 10625273 20. Roberts P, Priest H. **Reliability and validity in research**. *Nurs Stand* (2006.0) **20** 41-5 21. Boudreaux C, Noble C, Coates MM, Kelley J, Abanda M, Kintu A. **Noncommunicable Disease (NCD) strategic plans in low-and lower-middle income Sub-Saharan Africa: framing and policy response**. *Glob Health Action* (2020.0) **13** 1805165. PMID: 32873212 22. Ndinda C, Ndhlovu TP, Juma P, Asiki G, Kyobutungi C. **The evolution of non-communicable diseases policies in post-apartheid South Africa**. *BMC Public Health* (2018.0) **18** 956. PMID: 30168397 23. Bosu WK. **A comprehensive review of the policy and programmatic response to chronic non-communicable disease in Ghana**. *Ghana Med J* (2012.0) **46** 69-78. PMID: 23661820 24. Mukanu MM, Zulu JM, Mweemba C, Mutale W. **Responding to non-communicable diseases in Zambia: A policy analysis**. *Heal Res Policy Syst* (2017.0) **15** 1-9 25. Biswas T, Pervin S, Tanim MIA, Niessen L, Islam A. **Bangladesh policy on prevention and control of non-communicable diseases: A policy analysis**. *BMC Public Health* (2017.0) **17** 1-11. PMID: 28049454 26. Haldane V, Chuah FLH, Srivastava A, Singh SR, Koh GCH, Seng CK. **Community participation in health services development, implementation, and evaluation: A systematic review of empowerment, health, community, and process outcomes**. *PLoS One* (2019.0) **14** e0216112. PMID: 31075120 27. Allen LN, Pullar J, Wickramasinghe KK, Williams J, Roberts N, Mikkelsen B. **Evaluation of research on interventions aligned to WHO “Best Buys” for NCDs in low-income and lower-middle-income countries: A systematic review from 1990 to 2015**. *BMJ Glob Health* (2018.0) **3** e000535
--- title: 'Locus coeruleus-noradrenergic modulation of trigeminal pain: Implications for trigeminal neuralgia and psychiatric comorbidities' authors: - Basak Donertas-Ayaz - Robert M. Caudle journal: Neurobiology of Pain year: 2023 pmcid: PMC10038791 doi: 10.1016/j.ynpai.2023.100124 license: CC BY 4.0 --- # Locus coeruleus-noradrenergic modulation of trigeminal pain: Implications for trigeminal neuralgia and psychiatric comorbidities ## Highlights •*The locus* coeruleus (LC), the largest source of noradrenaline in the brain, is involved in the sensory and emotional processing of pain.•Chronic pain results in altered functioning of the LC in rodents.•Long-term peripheral nerve injury leads to overactivation of LC neurons. These changes are involved in impaired descending pain modulation and pain-related comorbidities such as depression, anxiety, and sleep disorders.•Evidence regarding the role of the LC in trigeminal neuropathic pain is limited. More studies are needed to explore the role of the LC in trigeminal neuropathic pain. ## Abstract Trigeminal neuralgia is the most common neuropathic pain involving the craniofacial region. Due to the complex pathophysiology, it is therapeutically difficult to manage. Noradrenaline plays an essential role in the modulation of arousal, attention, cognitive function, stress, and pain. The locus coeruleus, the largest source of noradrenaline in the brain, is involved in the sensory and emotional processing of pain. This review summarizes the knowledge about the involvement of noradrenaline in acute and chronic trigeminal pain conditions and how the activity of the locus coeruleus noradrenergic neurons changes in response to acute and chronic pain conditions and how these changes might be involved in pain-related comorbidities including anxiety, depression, and sleep disturbance. ## Introduction Trigeminal neuralgia (TN) is a facial pain condition that is defined by the International Headache Society as “a unilateral disorder characterized by brief electric shock-like pains, abrupt in onset and termination, and limited to the distribution of one or more divisions of the trigeminal nerve that typically are triggered by innocuous stimuli” (Headache Classification Committee of the International Headache Society (IHS), 2018). Although clinically well described, the pathophysiology of TN is not fully understood (Sabalys et al., 2013). Based on the existing evidence, the symptoms may arise from neurovascular compression (classical TN) or underlying disease (secondary TN) or may occur without an apparent cause (idiopathic) Headache Classification Committee of the International Headache Society (IHS), 2018)). The estimated annual incidence of TN was reported to be 4 to 13 per 100,000 people (Shankar Kikkeri et al., 2021). TN is more prevalent in women and adults over the age of 40 (De Toledo et al., 2016, Fallata et al., 2017, Siqueira et al., 2009, Tan et al., 2017, Sathasivam et al., 2017, Jainkittivong et al., 2012). TN imposes substantial health and economic burden on patients, families, and society (Cheng et al., 2017, Tang et al., 2016, Tolle et al., 2006, Allsop et al., 2015). Indeed, TN was reported to reduce quality of life, cause social and occupational impairment, disability (Tolle et al., 2006, Allsop et al., 2015), and psychiatric comorbidities including depression, anxiety, and sleep disorders in patients with TN (Zakrzewska et al., 2017, Devor et al., 2008, Smith, 2013, Wu, 2015, Mačianskytė et al., 2011, Chang et al., 2019). Suboptimal pain management of TN suggests the importance of understanding the detailed mechanisms underlying the pathogenesis to develop novel treatment strategies in TN. Noradrenaline (NA) plays an essential role in the regulation of cognitive function, sleep/wake state, arousal, attention, mood and stress reactions, and pain (Pertovaara, 2013, Glavin, 1985, Borodovitsyna et al., 2017, Mitchell and Weinshenker, 2010). Most pain research on NA focuses on NA inhibition of pain (Pertovaara, 2013, Llorca-Torralba et al., 2016). However, Taylor and Westlund present a convincing argument that in chronic neuropathic pain NA arising from the locus coeruleus (LC) facilitates pain in supraspinal regions (Taylor and Westlund, 2017). Chronic neuropathic pain results in sustained LC neuronal firing throughout the rostral-caudal distribution of LC fibers (Brightwell and Taylor, 2009). They argue that with continuous NA exposure neurons that process nociception become adapted to the inhibitory functions of NA. The net result is that instead of suppressing nociception the NA becomes part of the pro-nociception feedforward mechanisms that lead to enhanced pain. Thus, understanding NA’s transition from an anti-nociception to a pro-nociception regulatory pathway is likely important in the treatment of TN and associated psychological comorbidities. To date, most pain studies have concentrated on noradrenergic modulation of spinal nociceptive transmission and the role of NA has been less extensively studied in trigeminal pain. This review summarizes the noradrenergic modulation of acute and chronic trigeminal pain and then addresses the possible involvement of the noradrenergic system in TN-related comorbidities including anxiety, depression, and sleep disturbance. ## Locus coeruleus- noradrenaline system and a brief overview of the trigeminal pain pathway Seven NA-containing cell groups (A1-A7) provide noradrenergic innervation of the brain and the spinal cord (Dahlstroem and Fuxe, 1964). The A6 cell group, LC, located in dorsolateral pons is the major source of NA in the brain (Benarroch, 2018, Bucci, 2017). Therefore, in this review, the main focus is centered on the LC noradrenergic cell group, and its implications for modulation of trigeminal pain. Ophthalmic, maxillary, and mandibular divisions of the trigeminal nerve (the 5th cranial nerve) carry noxious sensations from the head and face to the trigeminal ganglion (TG) (Maciewicz et al., 1988), see Fig. 1). TG neurons constitute the first-order neurons and nociceptive unmyelinated C and lightly myelinated A-delta fibers coming from the TG are distributed to the trigeminal sensory nuclear complex in the brain stem where they synapse with second-order neurons (Maciewicz et al., 1988). Then, second-order neurons project to the somatosensory and limbic cortices via the thalamus (Maciewicz et al., 1988). The trigeminal sensory nuclear complex consists of the spinal nucleus and main (principal/chief) sensory nucleus (Henssen, 2016). Nociceptive afferents synapse primarily in the spinal nucleus (Maciewicz et al., 1988, Hayashi, 1985, Dessem et al., 2007). The spinal nucleus consists of three subnuclei (subnucleus oralis, subnucleus interpolaris, and subnucleus caudalis) and extends into the upper cervical spinal cord through subnucleus caudalis (also known as medullary dorsal horn) (Maciewicz et al., 1988, Stover et al., 1992, Brown, 1997, Sessle, 2000). The oral nociceptive signal is primarily processed in the principal nucleus, and the subnucleus oralis and interpolaris, while secondarily processed in the subnucleus caudalis, whereas facial nociceptive signals are primarily processed in the subnucleus caudalis (Takemura et al., 2006). Subnucleus oralis is particularly involved in intraoral and perioral nociceptive mechanisms (Dallel et al., 1990). Subnucleus interpolaris contributes to the sensory processing of facial pain (Hayashi, 1985). The subnucleus interpolaris/caudalis transition zone is also involved in deep tissue pain processing (Sugiyo et al., 2005, Wang et al., 2006, Shimizu, 2009, Dubner and Ren, 2004, Ren and Dubner, 2011). In the trigeminal sensory nuclear complex, orofacial nociceptive afferents synapse on second-order wide dynamic range and nociceptive-specific neurons (Maciewicz et al., 1988, Yokota and Matsumoto, 1983, Dubner and Bennett, 1983). These neurons then form the ventral trigeminothalamic tract and synapse with third-order neurons in the ventral posteromedial nucleus of the thalamus (Gauriau and Bernard, 2002, Iwata et al., 1992, Mitchell et al., 2004). From here, the signals are conveyed to the primary and secondary somatosensory cortices (Maciewicz et al., 1988, Sessle and Hu, 1991, Price et al., 2021, Robertson and Kaitz, 1981, Kaitz and Robertson, 1981). Sensory, affective, and cognitive processes modulate nociceptive inputs as they move along the pain pathway at the brainstem and thalamocortical levels (Ossipov et al., 2010).Fig. 1Simplified schematic representation of noxious transmission from face and head to upper brain regions involved in pain modulation. [ 1] Ophthalmic, maxillary, and mandibular branches of the trigeminal nerve carry noxious sensations from the head and face to the trigeminal ganglion (TG). [ 2] TG neurons constitute the first-order neurons and nociceptive unmyelinated C and lightly myelinated A-delta fibers coming from the TG are distributed to [3] the trigeminal sensory nuclear complex (TSNC). The TSNC comprises the spinal nucleus and main sensory nucleus. The spinal nucleus consists of three subnuclei: subnucleus oralis, interpolaris, and caudalis. In the TSNC, orofacial nociceptive afferents synapse on second-order neurons, and [4] these neurons then form the ventral trigeminothalamic tract and synapse with third-order neurons in the ventral posteromedial nucleus (VPM) of the thalamus. [ 5] From the thalamus, nociceptive information is conveyed to the primary and secondary somatosensory cortices. The periaqueductal gray (PAG) and the rostral ventrolateral medulla (RVM) are the two key brain regions that mediate descending pain modulation. The locus coeruleus (LC) receives inputs from the PAG and the RVM and sends inhibitory projections to the TSNC. Adapted from “Discriminative Pain Pathways”, byBioRender.com[2023]. Retrieved fromhttps://app.biorender.com/biorender-templates. Sensory information is mainly encoded by the somatosensory cortex, thalamus, anterior cingulate cortex, insula, and periaqueductal gray while emotional responses are mainly encoded by the amygdala, hippocampus, insula, orbitofrontal cortices, and prefrontal cortex (PFC) (for review see (Bushnell et al., 2013). The periaqueductal gray and the rostral ventrolateral medulla are the two key brain regions that mediate descending pain modulation (Chen and Heinricher, 2019). Major sources of afferents to LC are suggested to arise from nucleus paragigantocellularis in the rostral ventrolateral medulla, and nucleus prepositus hypoglossi in the dorsomedial medulla (Aston-Jones et al., 1986). Nucleus paragigantocellularis is linked to cardiovascular, nociceptive and respiratory functions and provides predominantly excitatory inputs to LC (Ennis and Aston-Jones, 1986, Ennis et al., 1992). Nucleus prepositus hypoglossi is involved in the control of eye movements (Aston-Jones, 1991) and it inhibits LC neurons by γ-Aminobutyric acid (GABA) type A receptors in the LC (Ennis and Aston-Jones, 1988, Ennis and Aston-Jones, 1989). Gu et al. also recently showed that LC received projections from the caudal ventrolateral medulla and LC mediated the antinociceptive responses produced by the caudal ventrolateral medulla in mice (Gu et al., 2023). ## Effects of noradrenaline under non-pathological conditions Neuroanatomical experiments show reciprocal pathways connecting LC to trigeminal sensory nuclei (Couto et al., 2006). Activation of the LC neurons results in NA release from the nerve terminals of LC neurons and NA inhibits the sensory transmission in the trigeminal neurons (Sasa and Takaori, 1973, Sasa et al., 1974, Sasa et al., 1977, Matsutani et al., 2000, Sasa, 1979). NA activates G protein-coupled α and β adrenergic receptors (ARs). ARs have three sub-classes each with three receptor subtypes: α1- (α1A, α1B, α1D), α2- (α2A, α2B, α2C) and β- (β1, β2, β3) (Hieble, 2007, Hieble et al., 1995, Hieble, 1995). Electrophysiological studies investigating the effects of AR agonists in trigeminal sensory pathway showed that several mechanisms mediate the inhibitory actions of NA on primary afferents through activation of ARs. LC-NA mediated actions in TG and subnucleus caudalis are schematized in Fig. 2. NA evokes depolarization of trigeminal subnucleus caudalis neurons via α1-ARs (Han, 2007) while activation of α2-ARs and β-ARs in trigeminal subnucleus caudalis and α2-ARs in TG hyperpolarizes membrane potentials of neurons (Han, 2007, Grudt et al., 1995) by increasing potassium conductance (Grudt et al., 1995) or inhibiting hyperpolarization-activated cation currents (Ih) (Takeda, 2002) and voltage-gated sodium channel currents (Im, 2018). Activation of ARs also modulates inhibitory or excitatory post-synaptic potentials. Activation of primary afferents evokes excitatory postsynaptic potentials in substantia gelatinosa of the trigeminal subnucleus caudalis mediated by glutamate (Travagli and Williams, 1996) while GABA and glycine interneurons inhibit glutamate-induced depolarization of substantia gelatinosa neurons (Travagli and Williams, 1996). Electrophysiological recordings made from neurons of guinea-pig spinal trigeminal nucleus pars caudalis showed that NA inhibits presynaptic glutamate release by activating presynaptic α2-ARs (Travagli and Williams, 1996) and activates GABA/glycine-releasing interneurons, thereby increasing the frequency of inhibitory postsynaptic potentials (Grudt et al., 1995). In addition, activation of α2-ARs reduced N-methyl-D-aspartate (NMDA)-evoked responses in the medullary dorsal horn of rats (Zhang, 1998). These findings demonstrate that the activation of ARs in trigeminal regions inhibits the sensory transmission in the trigeminal neurons via several mechanisms. However, these mechanisms need to be further investigated in acute and chronic trigeminal pain models. Fig. 2Locus coeruleus-noradrenaline mediated actions in trigeminal ganglia (TG) and subnucleus caudalis. Adapted from “Distribution of Norepinephrine Neurotransmitters in the Human Brain”, byBioRender.com[2023]. Retrieved fromhttps://app.biorender.com/biorender-templates. ## Activation of locus coeruleus produces antinociceptive effects on acute trigeminal nociception Acute noxious orofacial stimuli activate the descending noradrenergic pathway. As an example, intracisternal administration of capsaicin or experimental incisor tooth movement increased c-Fos immunoreactivity in LC, indicating a change in neuronal activity (Ter Horst et al., 2001, Magdalena et al., 2004, Bullitt, 1990). In another study, infraorbital nerve stimulation enhanced NA levels in the cat’s spinal superfusate, suggesting the activation of a descending noradrenergic pathway (Tyce and Yaksh, 1981). Activation of LC produces antinociceptive effects mediated by NA. For instance, activation of the LC/subcoeruleus neurons via electrical stimulation decreased both noxious (pinch and heat) and non-noxious stimuli evoked responses of rat subnucleus caudalis neurons (Tsuruoka et al., 2003). Iontophoretically applied NA also inhibited noxious heat evoked activity of sensory trigeminal neurons in rats (Cahusac et al., 1995). The evidence from experimental research showed that α2-AR exerts anti-nociceptive effects on acute trigeminal nociception. For instance, microinjection of the α2-AR agonist clonidine into the medullary dorsal horn reduced NMDA-evoked scratching behavior in the facial region (Wang, 2002). Intracisternal and intraperitoneal injection of clonidine produced antinociceptive effects in carrageenan- and formalin-induced orofacial pain (Nag and Mokha, 2016, Yoon et al., 2015), respectively. ## α2-adrenergic receptor-induced anti-nociception in trigeminal region is modulated by gonadal hormones Animal studies also showed that α2-AR-induced antinociception is modulated by gonadal hormones. Antinociception produced by activation of the α2-AR in the trigeminal region was attenuated by estrogen in female rats and required testosterone in males (Nag and Mokha, 2016, Nag and Mokha, 2004, Nag and Mokha, 2006, Nag and Mokha, 2009). For instance, intracisternal administration of the α2-AR agonist clonidine into the dorsal part of the medullary dorsal horn produced antinociceptive effects only in intact or testosterone-treated orchidectomized male rats and ovariectomized female rats and α2-AR antagonist yohimbine blocked these effects (Nag and Mokha, 2016, Nag and Mokha, 2009). Sex-specific changes in the α2-AR-mediated inhibition may be one of the factors responsible for the higher prevalence of TN in females and may help us to understand gender and age-associated changes in pain modulation. However, there are conflicting results regarding the role of estrogens in the modulation of facial pain. Contrary to above mentioned studies, aromatase knockout mice which are unable to produce estrogen since birth, had increased nociceptive behavior in the orofacial formalin model and daily estradiol treatment reversed the increase (Multon et al., 2005). Another study with ovariectomized rats showed lower nociceptive threshold to mechanical stimulation applied to the whisker pad area and estrogen replacement increased this threshold (Yu, 2011). These conflicting results may be due to differences in experimental models used, phase of estrous cycle studied. Besides, reported results are based on acute pain conditions, sex-related modulation of α2-AR-induced nociception in the trigeminal region should further be investigated in chronic trigeminal pain studies. ## Noradrenergic system might be involved in the central sensitization of medullary dorsal horn neurons Central sensitization refers to increased responsiveness of neurons to non-painful stimuli and is associated with the development and maintenance of chronic pain (Latremoliere and Woolf, 2009). Wang et al., showed that ARs were involved in the central sensitization of medullary dorsal horn neurons. In their study, intrathecal application of an adrenergic antagonist and a sympatholytic compound guanethidine, α-AR antagonist phentolamine, and the α1-AR antagonist prazosin but not the α2-AR antagonist yohimbine attenuated mustard oil-induced trigeminal central sensitization, reflected in increases in mechanoreceptive field size, responses to noxious stimuli, and decreases in activation threshold in nociceptive neurons of subnucleus caudalis (Wang, 2013). ## Effects of noradrenaline under trigeminal neuropathic pain conditions The majority of evidence regarding the effects of LC on neuropathic pain comes from spinal neuropathic pain models. As far as the authors are aware, only one study has directly investigated the involvement of LC in trigeminal neuropathic pain (Kaushal, 2016). In this study, Kaushal et al. demonstrated that the elimination of NA neurons via injection of anti-dopamine β-hydroxylase-saporin into the lateral ventricle and trigeminal brainstem nuclei three weeks after infraorbital nerve injury attenuated mechanical allodynia (Kaushal, 2016). This finding and experimental evidence coming from studies with spinal neuropathic pain (Brightwell and Taylor, 2009, Viisanen and Pertovaara, 2007, Alba-Delgado et al., 2021, Alba-Delgado et al., 2012) suggest that chronic pain may result in altered functioning of pain-modulation circuits including in the LC. Indeed, neuroimaging studies have shown that patients with TN had reduced gray matter volume in various brain regions related to sensory- and cognitive-affective dimensions of pain including the PFC, anterior cingulate cortex, cerebellum, amygdala, periaqueductal gray, insula, thalamus, hypothalamus, putamen, and nucleus accumbens (Zhang, 2018, Tsai et al., 2018, Li et al., 2017, Hayes, 2017). Additionally, patients with trigeminal neuropathy had altered LC functional connectivity with increased connectivity between the rostral ventromedial medulla and decreased connectivity between the ventrolateral periaqueductal gray matter (Mills et al., 2018) which might be related to decreased descending control in the chronic pain patients. ## Long-term peripheral nerve injury leads to hyperactivation of locus coeruleus neurons The LC promotes arousal and LC neurons are most active during wakefulness, and their firing rate decreases in sleep (Aston-Jones and Cohen, 2005). During wakefulness, LC-NA neurons fire spontaneously (tonic), and when salient stimuli are presented the tonic firing changes to phasic bursts of activity (Aston-Jones and Cohen, 2005). Rodent studies with spinal neuropathic pain models showed that the firing rates of LC-NA neurons differ at different stages of neuropathic pain. No study investigated the effects of trigeminal neuropathic pain. The spinal studies showed that in the early stages of neuropathy (seven and 14 days after sciatic nerve injury), the tonic activity of LC neurons was preserved (Viisanen and Pertovaara, 2007, Alba-Delgado et al., 2021, Alba-Delgado et al., 2012, Bravo et al., 2014, Alba-Delgado et al., 2012, Llorca-Torralba et al., 2019) while it turned into irregular tonic activity and exacerbated bilateral phasic responses in the long term (28–30 days after nerve injury) (Alba-Delgado et al., 2021, Llorca-Torralba et al., 2019, Alba-Delgado et al., 2013, Alba-Delgado et al., 2016). Phasic activity induces the release of excitatory neurotransmitter glutamate in the LC (Singewald and Philippu, 1998). Peripheral neuropathy increases the excitatory synaptic transmission to activate noradrenergic neurons, and basal extracellular glutamate concentrations in the LC neurons were increased in rats with spinal neuropathic pain (Rohampour et al., 2017, Suto et al., 2014, Kimura et al., 2015). Local glutamatergic and noradrenergic inputs control nerve injury induced glutamate release in the LC (Hayashida et al., 2018). NA evokes hyperpolarization of LC noradrenergic neurons by activating α2-ARs and reduces their firing rate (Aghajanian and VanderMaelen, 1982, Kawahara et al., 1999), while blocking α2-ARs in the LC potentiates the responses of LC neurons to the excitatory stimuli (Simson and Weiss, 1987). Sciatic nerve injury was shown to increase the expression of α2-AR in the LC 28 days after chronic constriction injury (CCI) with no change seven days after nerve injury (Alba-Delgado et al., 2012, Alba-Delgado et al., 2013). It was shown that blockade of α2-AR and group II metabotropic glutamate receptors (mGluRs) in the LC six weeks after spinal nerve ligation induces glutamate release in the LC to activate the descending noradrenergic pathway, reducing hypersensitivity in rats. Concomitant injection of the AMPA receptor antagonist CNQX into the LC dampened these effects (Hayashida et al., 2018). Furthermore, basal GABA levels in the LC increased after spinal nerve ligation in rats (Yoshizumi, 2012) and presynaptic inhibition of GABAergic inhibitory postsynaptic currents in LC neurons of nerve-injured mice produced analgesic effects through activation mediated by the descending noradrenergic system (Takasu et al., 2008). Kaushal et al. also found that the GABA-synthesizing enzyme glutamic acid decarboxylase (GAD65) immunoreactivity increased in the LC after infraorbital nerve injury and the GABAA receptor antagonist bicuculline injected into the LC alleviated mechanical hypersensitivity when the animals were tested at 10 min and 20 min post-infusion (Kaushal, 2016). The LC also receives inhibitory serotoninergic inputs from the dorsal reticular nucleus (Kim et al., 2004). Serotonin was shown to attenuate sensory stimuli evoked responses in the LC (Segal, 1979) and the glutamate-induced excitation of LC neurons (Aston-Jones et al., 1991). Alba-Delgado et al. proposed that inhibitory input from the dorsal reticular nucleus might block the excitatory input from the nucleus paragigantocellularis in the rostral ventrolateral medulla which maintains the constant tonic LC activity in neuropathic pain (Alba-Delgado et al., 2012). Moreover, the LC receives histaminergic innervation from the tuberomammillary nucleus (Panula et al., 1989) which increases the firing rate of LC noradrenergic neurons by activating histamine H1 and H2 receptors (Korotkova et al., 2005). Histamine injection into LC attenuates mechanical hypersensitivity in rats with spinal nerve ligation (Wei et al., 2014). These studies show that noradrenergic activity in LC changes over time after nerve injury and excitatory and inhibitory inputs control the firing activity of the noradrenergic neurons in LC that are involved in descending noradrenergic pain inhibition. ## Locus coeruleus might be involved in trigeminal neuropathic pain-related comorbidities Pain affects both sensory and affective responses (Singh et al., 2020). Thus, patients with chronic pain are at high risk of developing emotional disturbances. Anxiety and depression are commonly reported psychiatric disorders in patients with TN (Cheng et al., 2017, Melek et al., 2019). Results from the retrospective cohort studies showed that TN increases the risk of developing depressive, anxiety, or sleep disorder (Wu, 2015). Severe pain and treatment failure are also risk factors of depression and anxiety in those patients (Cheng et al., 2017, Chang et al., 2019). Animal models of trigeminal neuropathic pain also induce anxiety-like behaviors in rodents. Trigeminal neuropathic pain produced by CCI of the infraorbital nerve was found to cause anxiety-like behaviors in rats and mice evaluated in the open field, elevated plus-maze or light/dark transition tests approximately two weeks after nerve injury, while no depressive-like behavior was observed in the forced swimming test (McIlwrath, 2020, Gambeta et al., 2018, Chen et al., 2021). Trigeminal inflammatory compression injury, another trigeminal neuropathic pain model, has been shown to induce anxiety-like behavior in mice eight weeks after the nerve injury when evaluated in a light/dark transition test. In contrast, no behavioral change was found one or four weeks after nerve injury (Lyons et al., 2015). Trigeminal injury induced by a chronic mental nerve constriction in mice also increased escape/avoidance behavior to the mechanical stimulation (Montes Angeles et al., 2020). Anxiety-like behavior developed at least two weeks after nerve injury and was reported to occur only in CCI rats who developed allodynia (Gambeta et al., 2018), suggesting that as pain develops, nerve injury-induced activation of ascending and descending modulatory pathways can lead to the development of emotional disturbances. Indeed, a neuroimaging study has shown that along with anxiety-like behavior, the neuronal activity of brain regions involved in sensory and emotional aspects of pain were also changed ten weeks after the induction of trigeminal neuropathic pain in rats (McIlwrath, 2020). Dysregulation of monoamine neurotransmitters might be involved in the development of comorbid anxio-depressive behavior. Indeed, decreased levels of NA and its metabolite, vanillylmandelic acid in cerebrospinal fluid were reported in patients with TN (Strittmatter et al., 1997), however, more studies are needed. ## Altered locus coeruleus function is involved in anxiodepressive symptoms and sleep disturbances in chronic pain Neuropathic pain induces anxio-depressive behaviors in rodents and changes in the activity of LC noradrenergic neurons might contribute to that. It was shown that sciatic nerve injury provokes anxiety and depressive-like behavior in rats with a concomitant increase in firing rate of the LC neurons and increased expression of α2-AR as well as NA transporter, and tyrosine hydroxylase (Alba-Delgado et al., 2013). Similarly, in another study, neuropathic pain induced by streptozotocin-induced diabetes or CCI of the sciatic nerve was shown to cause anxiety-like behavior in rodents (Alba-Delgado et al., 2016, Sieberg et al., 2018). However, burst firing activity of LC neurons and expression of NA transporter, tyrosine hydroxylase, and phosphorylated cAMP-response element-binding protein (CREB) in the LC showed differences between CCI and streptozotocin-treated rats (Alba-Delgado et al., 2016) where the former increased these measures and the later decreased them. These results suggest that although different types of neuropathic pain provoke anxiety-like behavior in rodents, differential effects of neuropathic pain on noradrenergic activity in LC might result from the differences in the etiology of the neuropathic pain. However, mechanisms and brain pathways underlying neuropathic pain-related psychiatric comorbidities remain unclear and should be explored further. LC-NA neurons innervate several brain regions involved in anxiety and depression, including the amygdala, hippocampus, PFC (Cha et al., 2016, Hare and Duman, 2020, Jones and Moore, 1977, Li, 2021). Llorca-Torralba et al. demonstrated that long term neuropathic pain induced depressive like behavior in rats with CCI, reflected by increased immobility and decreased climbing and chemogenetic inhibition of the LC neurons projecting to the rostral anterior cingulate cortex reversed this behavior (Llorca-Torralba et al., 20222022). Furthermore, Camarena-Delgado et al. showed that chemogenetic inactivation of the LC projections to dorsal reticular nucleus induced depressive like behavior in naïve rats, however, it did not modify long-term pain-induced depression in rats with CCI of sciatic nerve (Camarena-Delgado et al., 2022). Llorca-Torralba et al. also demonstrated that the LC-basolateral amygdala (BLA) pathway is involved in the anxiety-like phenotype observed after long-term neuropathic pain, as inhibition of LC neurons projecting to the BLA reversed anxiety in rats with CCI of sciatic nerve while it did not affect the sham treated controls (Llorca-Torralba et al., 2019). Moreover, increasing the firing rate of LC-noradrenergic neurons by photostimulation induced anxiety-like behavior in mice and corticotropin-releasing hormone inputs from the amygdala to the LC mediated this effect (McCall et al., 2015). Optogenetic activation of noradrenergic projections from LC that project to BLA were shown to cause NA release in the BLA and induce anxiety-like behavior mediated by β-ARs (McCall et al., 2017). A neuroimaging study has also demonstrated that patients with TN had decreased gray matter volume in corticolimbic regions, including BLA (Zhang, 2018). Hirschberg et al. also showed that chemogenetic activation of LC-noradrenergic neurons innervating the PFC increased anxiety-like behavior in rats (Hirschberg et al., 2017). Stress may reflect a part of the mechanism underlying these clinical comorbidities as BLA is a key brain region involved in stress and activity of BLA neurons is modulated by NA-mediated stress responses (Sharp, 2017). For instance, stress-induced activation of LC noradrenergic neurons was shown to increase the firing activity in BLA (Giustino et al., 2020). Moreover, footshock stress was shown to increase the spontaneous firing rate of BLA neurons in rats which was reduced after treatment with the systemic β-blocker propranolol and increased by chemogenetic activation of LC noradrenergic neurons (Giustino et al., 2020). Patients with TN had also increased plasma cortisol and adrenocorticotropin levels (Strittmatter et al., 1996), indicating a stress response. Indeed, persistent pain can be considered as a source of stress and stress is a risk factor for many neuropsychiatric disorders, including anxiety and pain. In addition to anxio-depressive behavior, the altered activity in LC impairs the sleep-wake cycle. LC-noradrenergic neurons are also crucial for switching between sleep and wakefulness (Takahashi et al., 2010). Neuropathic pain significantly interferes with sleep and patients with TN had a higher risk for developing sleep disorders (Wu, 2015). For instance, nearly $60\%$ of patients with TN reported experiencing occasional awakenings due to pain (Devor et al., 2008) and they were four times more likely to wake up during sleep than people without trigeminal neuropathy (Benoliel et al., 2009). Results from experimental studies also showed that neuropathic pain causes sleep disturbance with an increase in wakefulness and a decrease in non-rapid eye movement sleep in mice with sciatic nerve injury (Ito, 2013, Koh et al., 2015). Neuropathic pain was also shown to increase the activity of LC- PFC noradrenergic neurons in mice with sciatic nerve ligation (Koh et al., 2015) and chemogenetic activation of these neurons exacerbated spontaneous foot-lifts in rats with tibial nerve injury (Hirschberg et al., 2017) and that may be, at least in part, associated with sleep disturbances under neuropathic pain (Koh et al., 2015). These studies suggest that overactivation of LC induced by neuropathic pain might be involved in emotional symptoms and sleep disturbances induced by chronic pain. ## Locus coeruleus mediates the analgesic action of drugs tested in animal models of pain Antidepressants such as the tricyclic antidepressant amitriptyline, and the serotonin-NA re-uptake inhibitor duloxetine are used to relieve neuropathic pain (Obata, 2017). Several animal studies with spinal neuropathic pain showed that noradrenergic descending inhibitory system mediates the action of antidepressants to relieve pain (Hiroki et al., 2017, Ito et al., 2018, Kremer et al., 2018). Antidepressants are thought to restore the impaired noradrenergic descending inhibitory system in chronic pain states (Hayashida and Obata, 2019) and NA and serotonin increase in the spinal cord plays an important role in the analgesic effect of antidepressants in neuropathic pain. For instance, it was found that intraperitoneal injections of amitriptyline or duloxetine attenuate the spinal nerve ligation-induced hyperalgesia and increase the spinal NA/serotonin levels in rats (Hoshino et al., 2015, Matsuoka et al., 2016). As tricyclic antidepressants and serotonin-NA re-uptake inhibitors modulate the neurotransmission both NA and serotonin, the analgesic effects produced in animal models of neuropathic pain do not solely belong to NA. However, it was found that the acute, systemic administration of antidepressants amitriptyline, duloxetine and mirtazapine which affect both NA and serotonin levels have more potent antinociceptive effects than the serotonin-reuptake inhibitor citalopram in rats with CCI of sciatic nerve (Bomholt et al., 2005). Moreover, compounds with greater NA reuptake inhibitory activity are suggested to be more effective for the treatment of pain than compounds having only serotonin reuptake inhibitory activity (Leventhal et al., 2007), supporting the importance of NA in relieving pain. Animal studies also showed that LC mediates the analgesic effects of various compounds tested in rodent neuropathic pain models. For instance, it was shown that injections of substance P (Muto et al., 2012) or glial cell line-derived neurotrophic factor (Kimura et al., 2015) or morphine (Llorca-Torralba et al., 2018) or histamine (Wei et al., 2014) into LC exert analgesic effects on mechanical allodynia and/or thermal hyperalgesia induced by CCI of the sciatic nerve. Results from the studies with several rodent orofacial pain models also support this notion. It was shown that intraperitoneal injection of carbamazepine, first line-therapy in TN, increases the activity of noradrenergic neurons in the LC of naive rats (Olpe and Jones, 1983). Whisker pad injection of botulinum toxin type A, an alternate therapy in TN, (Morra et al., 2016) reduced the increase in c-Fos expression in LC after formalin-induced orofacial nociception in rats (Matak et al., 2014). Bradykinin injections into the principal sensory trigeminal nucleus and LC produced antinociceptive effect in rats, as assessed by the jaw-opening reflex elicited by the dental pulp electrical stimulation test (Couto et al., 1998) and lesioning of LC with adrenergic neurotoxin N-(2-chloroethyl)-N-ethyl-2-bromobenzylamine (DSP-4) antagonized this effect (Couto et al., 2006). However, it should be noted that as activity of LC-NA neurons changes at acute or chronic pain conditions, future studies are needed to explore the role of LC-NA system in analgesic action of drugs in animal models of acute and chronic orofacial pain conditions. ## Conclusion and perspectives LC inhibition of nociceptive transmission in acute pain and in long-term neuropathic pain increases the tonic activity of LC-NA neurons. These changes may contribute to impaired descending pain modulation and pain-related comorbidities such as depression, anxiety, and sleep disorders. Although, there is limited evidence on the role of the LC in trigeminal neuropathic pain, the literature supports the involvement of the LC in chronification of pain. However, more studies are needed to explore the role of the LC specifically in trigeminal neuropathic pain. The LC also, in part, mediates the analgesic effects of antidepressants that inhibit NA reuptake or drugs exerted analgesic effect in several rodent models of neuropathic pain. This suggests that the LC is an important hub in the sensory and emotional integration of pain. Therapies targeting LC to reverse impairment in descending pain modulation on early stages of neuropathic pain might be beneficial to attenuate or prevent the development of persistent pain and related comorbidities. ## Funding This work was supported by the Facial Pain Research Foundation. ## CRediT authorship contribution statement Basak Donertas-Ayaz: Conceptualization, Investigation, Writing – original draft, Visualization. Robert M. Caudle: Writing – review & editing. ## Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ## References 1. Aghajanian G.K., VanderMaelen C.P.. **alpha 2-adrenoceptor-mediated hyperpolarization of locus coeruleus neurons: intracellular studies in vivo**. *Science* (1982.0) **215** 1394-1396. PMID: 6278591 2. Alba-Delgado C., Borges G., Sánchez-Blázquez P., Ortega J.E., Horrillo I., Mico J.A., Meana J.J., Neto F., Berrocoso E.. **The function of alpha-2-adrenoceptors in the rat locus coeruleus is preserved in the chronic constriction injury model of neuropathic pain**. *Psychopharmacology (Berl.)* (2012.0) **221** 53-65. PMID: 22038538 3. Alba-Delgado C., Mico J.A., Sánchez-Blázquez P., Berrocoso E.. **Analgesic antidepressants promote the responsiveness of locus coeruleus neurons to noxious stimulation: implications for neuropathic pain**. *Pain* (2012.0) **153** 1438-1449. PMID: 22591831 4. Alba-Delgado C., Llorca-Torralba M., Horrillo I., Ortega J.E., Mico J.A., Sánchez-Blázquez P., Meana J.J., Berrocoso E.. **Chronic pain leads to concomitant noradrenergic impairment and mood disorders**. *Biol. Psychiatry* (2013.0) **73** 54-62. PMID: 22854119 5. Alba-Delgado C., Cebada-Aleu A., Mico J.A., Berrocoso E.. **Comorbid anxiety-like behavior and locus coeruleus impairment in diabetic peripheral neuropathy: A comparative study with the chronic constriction injury model**. *Prog. Neuropsychopharmacol. Biol. Psychiatry* (2016.0) **71** 45-56. PMID: 27328428 6. Alba-Delgado C., Mico J.A., Berrocoso E.. **Neuropathic pain increases spontaneous and noxious-evoked activity of locus coeruleus neurons**. *Prog. Neuropsychopharmacol. Biol. Psychiatry* (2021.0) **105** 7. Allsop M.J., Twiddy M., Grant H., Czoski-Murray C., Mon-Williams M., Mushtaq F., Phillips N., Zakrzewska J.M., Pavitt S.. **Diagnosis, medication, and surgical management for patients with trigeminal neuralgia: a qualitative study**. *Acta Neurochir. (Wien)* (2015.0) **157** 1925-1933. PMID: 26329729 8. Aston-Jones G.. **Afferent regulation of locus coeruleus neurons: anatomy, physiology and pharmacology**. *Prog. Brain Res.* (1991.0) **88** 47-75. PMID: 1687622 9. Aston-Jones G., Cohen J.D.. **An integrative theory of locus coeruleus-norepinephrine function: adaptive gain and optimal performance**. *Annu. Rev. Neurosci.* (2005.0) **28** 403-450. PMID: 16022602 10. Aston-Jones G., Ennis M., Pieribone V.A., Nickell W.T., Shipley M.T.. **The brain nucleus locus coeruleus: restricted afferent control of a broad efferent network**. *Science* (1986.0) **234** 734-737. PMID: 3775363 11. Aston-Jones G., Akaoka H., Charlety P., Chouvet G.. **Serotonin selectively attenuates glutamate-evoked activation of noradrenergic locus coeruleus neurons**. *J. Neurosci.* (1991.0) **11** 760-769. PMID: 1672153 12. Benarroch E.E.. **Locus coeruleus**. *Cell Tissue Res.* (2018.0) **373** 221-232. PMID: 28687925 13. Benoliel R., Eliav E., Sharav Y.. **Self-reports of pain-related awakenings in persistent orofacial pain patients**. *J. Orofac. Pain* (2009.0) **23** 330-338. PMID: 19888484 14. Bomholt S.F., Mikkelsen J.D., Blackburn-Munro G.. **Antinociceptive effects of the antidepressants amitriptyline, duloxetine, mirtazapine and citalopram in animal models of acute, persistent and neuropathic pain**. *Neuropharmacology* (2005.0) **48** 252-263. PMID: 15695164 15. Borodovitsyna O., Flamini M., Chandler D.. **Noradrenergic modulation of cognition in health and disease**. *Neural Plast.* (2017.0) **2017** 6031478. PMID: 28596922 16. Bravo L., Torres-Sanchez S., Alba-Delgado C., Mico J.A., Berrocoso E.. **Pain exacerbates chronic mild stress-induced changes in noradrenergic transmission in rats**. *Eur. Neuropsychopharmacol.* (2014.0) **24** 996-1003. PMID: 24491949 17. Brightwell J.J., Taylor B.K.. **Noradrenergic neurons in the locus coeruleus contribute to neuropathic pain**. *Neuroscience* (2009.0) **160** 174-185. PMID: 19223010 18. Brown J.A.. **The trigeminal complex. Anatomy and physiology**. *Neurosurg. Clin. N. Am.* (1997.0) **8** 1-10. PMID: 9018701 19. Bucci D.. **Systematic Morphometry of Catecholamine Nuclei in the Brainstem**. *Front. Neuroanat.* (2017.0) **11** 98. PMID: 29163071 20. Bullitt E.. **Expression of c-fos-like protein as a marker for neuronal activity following noxious stimulation in the rat**. *J Comp Neurol* (1990.0) **296** 517-530. PMID: 2113539 21. Bushnell M.C., Ceko M., Low L.A.. **Cognitive and emotional control of pain and its disruption in chronic pain**. *Nat. Rev. Neurosci.* (2013.0) **14** 502-511. PMID: 23719569 22. Cahusac P.M., Morris R., Hill R.G.. **A pharmacological study of the modulation of neuronal and behavioural nociceptive responses in the rat trigeminal region**. *Brain Res.* (1995.0) **700** 70-82. PMID: 8624730 23. Camarena-Delgado C., Llorca-Torralba M., Suárez-Pereira I., Bravo L., López‐Martín C., Garcia-Partida J.A., Mico J.A., Berrocoso E.. **Nerve injury induces transient locus coeruleus activation over time: role of the locus coeruleus-dorsal reticular nucleus pathway**. *Pain* (2022.0) **163** 943-954. PMID: 35025190 24. Cha J., Greenberg T., Song I., Blair Simpson H., Posner J., Mujica‐Parodi L.R.. **Abnormal hippocampal structure and function in clinical anxiety and comorbid depression**. *Hippocampus* (2016.0) **26** 545-553. PMID: 26743454 25. Chang B., Zhu W., Li S.. **Effects of depression and anxiety on microvascular decompression outcome for trigeminal neuralgia patients**. *World Neurosurg.* (2019.0) **128** e556-e561. PMID: 31071444 26. Chen Q., Heinricher M.M.. **Descending control mechanisms and chronic pain**. *Curr. Rheumatol. Rep.* (2019.0) **21** 13. PMID: 30830471 27. Chen W.-J., Niu J.-Q., Chen Y.-T., Deng W.-J., Xu Y.-Y., Liu J., Luo W.-F., Liu T.. **Unilateral facial injection of Botulinum neurotoxin A attenuates bilateral trigeminal neuropathic pain and anxiety-like behaviors through inhibition of TLR2-mediated neuroinflammation in mice**. *J. Headache Pain* (2021.0) **22** 28. Cheng J., Long J., Hui X., Lei D., Zhang H.. **Effects of microvascular decompression on depression and anxiety in trigeminal neuralgia: a prospective cohort study focused on risk factors and prognosis**. *Clin. Neurol. Neurosurg.* (2017.0) **161** 59-64. PMID: 28858633 29. Couto L.B., Correa F.M., Pela I.R.. **Brain sites involved in the antinociceptive effect of bradykinin in rats**. *Br. J. Pharmacol.* (1998.0) **125** 1578-1584. PMID: 9884087 30. Couto L.B., Moroni C.R., Ferreira C.M.D.R., Elias-Filho D.H., Parada C.A., Pelá I.R., Coimbra N.C.. **Descriptive and functional neuroanatomy of locus coeruleus-noradrenaline-containing neurons involvement in bradykinin-induced antinociception on principal sensory trigeminal nucleus**. *J. Chem. Neuroanat.* (2006.0) **32** 28-45. PMID: 16678997 31. Dahlstroem, A., Fuxe, K., Evidence for the existence of monoamine-containing neurons in the central nervous system. I. Demonstration of monoamines in the cell bodies of brain stem neurons. Acta Physiol. Scand. Suppl., 1964: p. SUPPL 232:1-55. 32. Dallel R., Raboisson P., Woda A., Sessle B.J.. **Properties of nociceptive and non-nociceptive neurons in trigeminal subnucleus oralis of the rat**. *Brain Res.* (1990.0) **521** 95-106. PMID: 2207681 33. De Toledo I.P., Conti Réus J., Fernandes M., Porporatti A.L., Peres M.A., Takaschima A., Linhares M.N., Guerra E., De Luca Canto G.. **Prevalence of trigeminal neuralgia: a systematic review**. *J. Am. Dent. Assoc.* (2016.0) **147** 570-576.e2. PMID: 27017183 34. Dessem D., Moritani M., Ambalavanar R.. **Nociceptive craniofacial muscle primary afferent neurons synapse in both the rostral and caudal brain stem**. *J. Neurophysiol.* (2007.0) **98** 214-223. PMID: 17493918 35. Devor M., Wood I., Sharav Y., Zakrzewska J.M.. **Trigeminal neuralgia during sleep**. *Pain Pract.* (2008.0) **8** 263-268. PMID: 18503619 36. Dubner R., Bennett G.J.. **Spinal and trigeminal mechanisms of nociception**. *Annu. Rev. Neurosci.* (1983.0) **6** 381-418. PMID: 6132587 37. Dubner R., Ren K.. **Brainstem mechanisms of persistent pain following injury**. *J. Orofac. Pain* (2004.0) **18** 299-305. PMID: 15636012 38. Ennis M., Aston-Jones G.. **A potent excitatory input to the nucleus locus coeruleus from the ventrolateral medulla**. *Neurosci. Lett.* (1986.0) **71** 299-305. PMID: 3025783 39. Ennis M., Aston-Jones G.. **Activation of locus coeruleus from nucleus paragigantocellularis: a new excitatory amino acid pathway in brain**. *J. Neurosci.* (1988.0) **8** 3644-3657. PMID: 3193175 40. Ennis M., Aston-Jones G.. **GABA-mediated inhibition of locus coeruleus from the dorsomedial rostral medulla**. *J. Neurosci.* (1989.0) **9** 2973-2981. PMID: 2769374 41. Ennis M., Aston-Jones G., Shiekhattar R.. **Activation of locus coeruleus neurons by nucleus paragigantocellularis or noxious sensory stimulation is mediated by intracoerulear excitatory amino acid neurotransmission**. *Brain Res.* (1992.0) **598** 185-195. PMID: 1336704 42. Fallata A., Salter A., Tyry T., Cutter G.R., Marrie R.A.. **Trigeminal neuralgia commonly precedes the diagnosis of multiple sclerosis**. *Int. J. MS Care* (2017.0) **19** 240-246. PMID: 29070964 43. Gambeta E., Batista M.A., Maschio G.P., Turnes J.d.M., Araya E.I., Chichorro J.G.. **Anxiety- but not depressive-like behaviors are related to facial hyperalgesia in a model of trigeminal neuropathic pain in rats**. *Physiol. Behav.* (2018.0) **191** 131-137. PMID: 29678597 44. Gauriau C., Bernard J.F.. **Pain pathways and parabrachial circuits in the rat**. *Exp. Physiol.* (2002.0) **87** 251-258. PMID: 11856971 45. Giustino T.F., Ramanathan K.R., Totty M.S., Miles O.W., Maren S.. **Locus coeruleus norepinephrine drives stress-induced increases in basolateral amygdala firing and impairs extinction learning**. *J. Neurosci.* (2020.0) **40** 907-916. PMID: 31801809 46. Glavin G.B.. **Stress and brain noradrenaline: a review**. *Neurosci. Biobehav. Rev.* (1985.0) **9** 233-243. PMID: 2861590 47. Grudt T.J., Williams J.T., Travagli R.A.. **Inhibition by 5-hydroxytryptamine and noradrenaline in substantia gelatinosa of guinea-pig spinal trigeminal nucleus**. *J. Physiol.* (1995.0) **485** 113-120. PMID: 7658366 48. Gu X.. **Neurons in the caudal ventrolateral medulla mediate descending pain control**. *Nat. Neurosci.* (2023.0) 49. Han S.K.. **Noradrenaline inhibits substantia gelatinosa neurons in mice trigeminal subnucleus caudalis via alpha(2) and beta adrenoceptors**. *Neurosci. Lett.* (2007.0) **411** 92-97. PMID: 17110030 50. Hare B.D., Duman R.S.. **Prefrontal cortex circuits in depression and anxiety: contribution of discrete neuronal populations and target regions**. *Mol. Psychiatry* (2020.0) **25** 2742-2758. PMID: 32086434 51. Hayashi H.. **Morphology of terminations of small and large myelinated trigeminal primary afferent fibers in the cat**. *J. Comp. Neurol.* (1985.0) **240** 71-89. PMID: 4056105 52. Hayashida K.I., Kimuram M., Eisenach J.C.. **Blockade of alpha2-adrenergic or metabotropic glutamate receptors induces glutamate release in the locus coeruleus to activate descending inhibition in rats with chronic neuropathic hypersensitivity**. *Neurosci. Lett.* (2018.0) **676** 41-45. PMID: 29627342 53. Hayashida K.-I., Obata H.. **Strategies to treat chronic pain and strengthen impaired descending noradrenergic inhibitory system**. *Int. J. Mol. Sci.* (2019.0) **20** 822. PMID: 30769838 54. Hayes D.J.. **Affective circuitry alterations in patients with trigeminal neuralgia**. *Front. Neuroanat.* (2017.0) **11** 73. PMID: 28928638 55. Headache Classification Committee of the International Headache Society (IHS) The International Classification of Headache Disorders, 3rd edition. Cephalalgia, 2018. 38(1): p. 1-211. 56. Henssen D.J.. **New insights in trigeminal anatomy: a double orofacial tract for nociceptive input**. *Front. Neuroanat.* (2016.0) **10** 53. PMID: 27242449 57. Hieble J.P.. **International Union of Pharmacology. X. Recommendation for nomenclature of alpha 1-adrenoceptors: consensus update**. *Pharmacol. Rev.* (1995.0) **47** 267-270. PMID: 7568329 58. Hieble J.P.. **Subclassification and nomenclature of alpha- and beta-adrenoceptors**. *Curr. Top. Med. Chem.* (2007.0) **7** 129-134. PMID: 17266601 59. Hieble J.P., Bondinell W.E., Ruffolo R.R.. **Alpha- and beta-adrenoceptors: from the gene to the clinic. 1. Molecular biology and adrenoceptor subclassification**. *J. Med. Chem.* (1995.0) **38** 3415-3444. PMID: 7658428 60. Hiroki T., Suto T., Saito S., Obata H.. **Repeated administration of amitriptyline in neuropathic pain: modulation of the noradrenergic descending inhibitory system**. *Anesth. Analg.* (2017.0) **125** 1281-1288. PMID: 28787345 61. Hirschberg S., Li Y., Randall A., Kremer E.J., Pickering A.E.. **Functional dichotomy in spinal- vs prefrontal-projecting locus coeruleus modules splits descending noradrenergic analgesia from ascending aversion and anxiety in rats**. *Elife* (2017.0) **6** 62. Hoshino H., Obata H., Saito S.. **Antihyperalgesic effect of duloxetine and amitriptyline in rats after peripheral nerve injury: Influence of descending noradrenergic plasticity**. *Neurosci. Lett.* (2015.0) **602** 62-67. PMID: 26135544 63. Im S.T.. **Dexmedetomidine inhibits voltage-gated sodium channels via alpha2-adrenoceptors in trigeminal ganglion neurons**. *Mediators Inflamm.* (2018.0) **2018** 1782719. PMID: 30245586 64. Ito H.. **Analysis of sleep disorders under pain using an optogenetic tool: possible involvement of the activation of dorsal raphe nucleus-serotonergic neurons**. *Mol. Brain* (2013.0) **6** 59. PMID: 24370235 65. Ito S., Suto T., Saito S., Obata H.. **Repeated administration of duloxetine suppresses neuropathic pain by accumulating effects of noradrenaline in the spinal cord**. *Anesth. Analg.* (2018.0) **126** 298-307. PMID: 28787337 66. Iwata K., Kenshalo D.R., Dubner R., Nahin R.L.. **Diencephalic projections from the superficial and deep laminae of the medullary dorsal horn in the rat**. *J. Comp. Neurol.* (1992.0) **321** 404-420. PMID: 1506477 67. Jainkittivong A., Aneksuk V., Langlais R.P.. **Trigeminal neuralgia: a retrospective study of 188 Thai cases**. *Gerodontology* (2012.0) **29** e611-e617. PMID: 21726275 68. Jones B.E., Moore R.Y.. **Ascending projections of the locus coeruleus in the rat. II. Autoradiographic study**. *Brain Res.* (1977.0) **127** 25-53. PMID: 301051 69. Kaitz S.S., Robertson R.T.. **Thalamic connections with limbic cortex. II. Corticothalamic projections**. *J. Comp. Neurol.* (1981.0) **195** 527-545. PMID: 7204660 70. Kaushal R.. **GABA-A receptor activity in the noradrenergic locus coeruleus drives trigeminal neuropathic pain in the rat; contribution of NAalpha1 receptors in the medial prefrontal cortex**. *Neuroscience* (2016.0) **334** 148-159. PMID: 27520081 71. Kawahara Y., Kawahara H., Westerink B.H.. **Tonic regulation of the activity of noradrenergic neurons in the locus coeruleus of the conscious rat studied by dual-probe microdialysis**. *Brain Res.* (1999.0) **823** 42-48. PMID: 10095010 72. Kim M.-A., Lee H.S., Lee B.Y., Waterhouse B.D.. **Reciprocal connections between subdivisions of the dorsal raphe and the nuclear core of the locus coeruleus in the rat**. *Brain Res.* (2004.0) **1026** 56-67. PMID: 15476697 73. Kimura M., Sakai A., Sakamoto A., Suzuki H.. **Glial cell line-derived neurotrophic factor-mediated enhancement of noradrenergic descending inhibition in the locus coeruleus exerts prolonged analgesia in neuropathic pain**. *Br. J. Pharmacol.* (2015.0) **172** 2469-2478. PMID: 25572945 74. Kimura M., Suto T., Morado-Urbina C.E., Peters C.M., Eisenach J.C., Hayashida K.-I.. **Impaired pain-evoked analgesia after nerve injury in rats reflects altered glutamate regulation in the locus coeruleus**. *Anesthesiology* (2015.0) **123** 899-908. PMID: 26244888 75. Koh K., Hamada A., Hamada Y., Yanase M., Sakaki M., Someya K., Narita M., Kuzumaki N., Ikegami D., Sakai H., Iseki M., Inada E., Narita M.. **Possible involvement of activated locus coeruleus-noradrenergic neurons in pain-related sleep disorders**. *Neurosci. Lett.* (2015.0) **589** 200-206. PMID: 25481765 76. Korotkova T.M., Sergeeva O.A., Ponomarenko A.A., Haas H.L.. **Histamine excites noradrenergic neurons in locus coeruleus in rats**. *Neuropharmacology* (2005.0) **49** 129-134. PMID: 15992588 77. Kremer M., Yalcin I., Goumon Y., Wurtz X., Nexon L., Daniel D., Megat S., Ceredig R.A., Ernst C., Turecki G., Chavant V., Théroux J.-F., Lacaud A., Joganah L.-E., Lelievre V., Massotte D., Lutz P.-E., Gilsbach R., Salvat E., Barrot M.. **A dual noradrenergic mechanism for the relief of neuropathic allodynia by the antidepressant drugs duloxetine and amitriptyline**. *J. Neurosci.* (2018.0) **38** 9934-9954. PMID: 30249798 78. Latremoliere A., Woolf C.J.. **Central sensitization: a generator of pain hypersensitivity by central neural plasticity**. *J. Pain* (2009.0) **10** 895-926. PMID: 19712899 79. Leventhal L., Smith V., Hornby G., Andree T.H., Brandt M.R., Rogers K.E.. **Differential and synergistic effects of selective norepinephrine and serotonin reuptake inhibitors in rodent models of pain**. *J. Pharmacol. Exp. Ther.* (2007.0) **320** 1178-1185. PMID: 17142646 80. Li Y.Y.. **Common and specific alterations of amygdala subregions in major depressive disorder with and without anxiety: a combined structural and resting-state functional MRI study**. *Front. Hum. Neurosci.* (2021.0) **15** 81. Li M., Yan J., Li S., Wang T., Zhan W., Wen H., Ma X., Zhang Y., Tian J., Jiang G.. **Reduced volume of gray matter in patients with trigeminal neuralgia**. *Brain Imaging Behav.* (2017.0) **11** 486-492. PMID: 26899433 82. Llorca-Torralba M.. **Pain and depression comorbidity causes asymmetric plasticity in the locus coeruleus neurons**. *Brain* (2022.0) **145** 154-167. PMID: 34373893 83. Llorca-Torralba M., Borges G., Neto F., Mico J.A., Berrocoso E.. **Noradrenergic Locus Coeruleus pathways in pain modulation**. *Neuroscience* (2016.0) **338** 93-113. PMID: 27267247 84. Llorca-Torralba M., Mico J.A., Berrocoso E.. **Behavioral effects of combined morphine and MK-801 administration to the locus coeruleus of a rat neuropathic pain model**. *Prog. Neuropsychopharmacol. Biol. Psychiatry* (2018.0) **84** 257-266. PMID: 29524514 85. Llorca-Torralba M., Pilar-Cuéllar F., Bravo L., Bruzos-Cidon C., Torrecilla M., Mico J.A., Ugedo L., Garro-Martínez E., Berrocoso E.. **Opioid activity in the locus coeruleus is modulated by chronic neuropathic pain**. *Mol. Neurobiol.* (2019.0) **56** 4135-4150. PMID: 30284123 86. Llorca-Torralba M., Suárez-Pereira I., Bravo L., Camarena-Delgado C., Garcia-Partida J.A., Mico J.A., Berrocoso E.. **Chemogenetic silencing of the locus coeruleus-basolateral amygdala pathway abolishes pain-induced anxiety and enhanced aversive learning in rats**. *Biol. Psychiatry* (2019.0) **85** 1021-1035. PMID: 30987747 87. Lyons D.N.. **Trigeminal Inflammatory Compression (TIC) injury induces chronic facial pain and susceptibility to anxiety-related behaviors**. *Neuroscience* (2015.0) **295** 126-138. PMID: 25818051 88. Mačianskytė D., Janužis G., Kubilius R., Adomaitienė V., Ščiupokas A.. **Associations between chronic pain and depressive symptoms in patients with trigeminal neuralgia**. *Medicina (Kaunas)* (2011.0) **47** 386. PMID: 22112988 89. Maciewicz R., Mason P., Strassman A., Potrebic S.. **Organization of trigeminal nociceptive pathways**. *Semin. Neurol.* (1988.0) **8** 255-264 90. Magdalena C.M., Navarro V.P., Park D.M., Stuani M.B.S., Rocha M.J.A.. **C-fos expression in rat brain nuclei following incisor tooth movement**. *J. Dent. Res.* (2004.0) **83** 50-54. PMID: 14691113 91. Matak I., Rossetto O., Lackovic Z.. **Botulinum toxin type A selectivity for certain types of pain is associated with capsaicin-sensitive neurons**. *Pain* (2014.0) **155** 1516-1526. PMID: 24793910 92. Matsuoka H., Suto T., Saito S., Obata H.. **Amitriptyline, but not pregabalin, reverses the attenuation of noxious stimulus-induced analgesia after nerve injury in rats**. *Anesth. Analg.* (2016.0) **123** 504-510. PMID: 27088996 93. Matsutani K., Tsuruoka M., Shinya A., Furuya R., Kawawa T.. **Stimulation of the locus coeruleus suppresses trigeminal sensorimotor function in the rat**. *Brain Res. Bull.* (2000.0) **53** 827-832. PMID: 11179850 94. McCall J., Al-Hasani R., Siuda E., Hong D., Norris A., Ford C., Bruchas M.. **CRH engagement of the locus coeruleus noradrenergic system mediates stress-induced anxiety**. *Neuron* (2015.0) **87** 605-620. PMID: 26212712 95. McCall J.G., Siuda E.R., Bhatti D.L., Lawson L.A., McElligott Z.A., Stuber G.D., Bruchas M.R.. **Locus coeruleus to basolateral amygdala noradrenergic projections promote anxiety-like behavior**. *Elife* (2017.0) **6** 96. McIlwrath S.L.. **Manganese-enhanced MRI reveals changes within brain anxiety and aversion circuitry in rats with chronic neuropathic pain- and anxiety-like behaviors**. *Neuroimage* (2020.0) **223** 97. Melek L., Smith J., Karamat A., Renton T.. **Comparison of the neuropathic pain symptoms and psychosocial impacts of trigeminal neuralgia and painful posttraumatic trigeminal neuropathy**. *J. Oral Facial Pain Headache* (2019.0) **33** 77-88. PMID: 30703173 98. Mills E.P., Di Pietro F., Alshelh Z., Peck C.C., Murray G.M., Vickers E.R., Henderson L.A.. **Brainstem pain-control circuitry connectivity in chronic neuropathic pain**. *J. Neurosci.* (2018.0) **38** 465-473. PMID: 29175957 99. Mitchell J.L., Silverman M.B., Aicher S.A.. **Rat trigeminal lamina I neurons that project to thalamic or parabrachial nuclei contain the mu-opioid receptor**. *Neuroscience* (2004.0) **128** 571-582. PMID: 15381286 100. Mitchell H.A., Weinshenker D.. **Good night and good luck: norepinephrine in sleep pharmacology**. *Biochem. Pharmacol.* (2010.0) **79** 801-809. PMID: 19833104 101. Montes Angeles C.D., Andrade Gonzalez R.D., Hernandez E.P., García Hernández A.L., Pérez Martínez I.O.. **Sensory, affective, and cognitive effects of trigeminal injury in mice**. *J. Oral Maxillofac. Surg.* (2020.0) **78** 2169-2181. PMID: 32866484 102. Morra M.E., Elgebaly A., Elmaraezy A., Khalil A.M., Altibi A.M.A., Vu T.-H., Mostafa M.R., Huy N.T., Hirayama K.. **Therapeutic efficacy and safety of Botulinum Toxin A Therapy in Trigeminal Neuralgia: a systematic review and meta-analysis of randomized controlled trials**. *J. Headache Pain* (2016.0) **17** 103. Multon S., Pardutz A., Mosen J., Hua M.T., Defays C., Honda S.-I., Harada N., Bohotin C., Franzen R., Schoenen J.. **Lack of estrogen increases pain in the trigeminal formalin model: a behavioural and immunocytochemical study of transgenic ArKO mice**. *Pain* (2005.0) **114** 257-265. PMID: 15733652 104. Muto Y., Sakai A., Sakamoto A., Suzuki H.. **Activation of NK(1) receptors in the locus coeruleus induces analgesia through noradrenergic-mediated descending inhibition in a rat model of neuropathic pain**. *Br. J. Pharmacol.* (2012.0) **166** 1047-1057. PMID: 22188400 105. Nag S., Mokha S.S.. **Estrogen attenuates antinociception produced by stimulation of Kolliker-Fuse nucleus in the rat**. *Eur. J. Neurosci.* (2004.0) **20** 3203-3207. PMID: 15579177 106. Nag S., Mokha S.S.. **Activation of alpha2-adrenoceptors in the trigeminal region produces sex-specific modulation of nociception in the rat**. *Neuroscience* (2006.0) **142** 1255-1262. PMID: 16934408 107. Nag S., Mokha S.S.. **Testosterone is essential for alpha(2)-adrenoceptor-induced antinociception in the trigeminal region of the male rat**. *Neurosci. Lett.* (2009.0) **467** 48-52. PMID: 19819297 108. Nag S., Mokha S.S.. **Activation of the trigeminal alpha2-adrenoceptor produces sex-specific, estrogen dependent thermal antinociception and antihyperalgesia using an operant pain assay in the rat**. *Behav. Brain Res.* (2016.0) **314** 152-158. PMID: 27506651 109. Obata H.. **Analgesic mechanisms of antidepressants for neuropathic pain**. *Int. J. Mol. Sci.* (2017.0) **18** 2483. PMID: 29160850 110. Olpe H.R., Jones R.S.. **The action of anticonvulsant drugs on the firing of locus coeruleus neurons: selective, activating effect of carbamazepine**. *Eur. J. Pharmacol.* (1983.0) **91** 107-110. PMID: 6617734 111. Ossipov M.H., Dussor G.O., Porreca F.. **Central modulation of pain**. *J. Clin. Invest.* (2010.0) **120** 3779-3787. PMID: 21041960 112. Panula P., Pirvola U., Auvinen S., Airaksinen M.S.. **Histamine-immunoreactive nerve fibers in the rat brain**. *Neuroscience* (1989.0) **28** 585-610. PMID: 2710333 113. Pertovaara A.. **The noradrenergic pain regulation system: a potential target for pain therapy**. *Eur. J. Pharmacol.* (2013.0) **716** 2-7. PMID: 23500194 114. Price, S., Daly, D.T., Neuroanatomy, Trigeminal Nucleus, in StatPearls. 2021: Treasure Island (FL). 115. Ren K., Dubner R.. **The role of trigeminal interpolaris-caudalis transition zone in persistent orofacial pain**. *Int. Rev. Neurobiol.* (2011.0) **97** 207-225. PMID: 21708312 116. Robertson R.T., Kaitz S.S.. **Thalamic connections with limbic cortex. I. Thalamocortical projections**. *J. Comp. Neurol.* (1981.0) **195** 501-525. PMID: 7204659 117. Rohampour K., Azizi H., Fathollahi Y., Semnanian S.. **Peripheral nerve injury potentiates excitatory synaptic transmission in locus coeruleus neurons**. *Brain Res. Bull.* (2017.0) **130** 112-117. PMID: 28093336 118. Sabalys G., Juodzbalys G., Wang H.L.. **Aetiology and pathogenesis of trigeminal neuralgia: a comprehensive review**. *J. Oral Maxillofac. Res.* (2013.0) **3** e2. PMID: 24422020 119. Sasa M.. **Microiontophoretic studies on noradrenergic inhibition from locus coeruleus of spinal trigeminal nucleus neurons**. *J. Pharmacol. Exp. Ther.* (1979.0) **210** 311-315. PMID: 480182 120. Sasa M., Munekiyo K., Ikeda H., Takaori S.. **Noradrenaline-mediated inhibition by locus coeruleus of spinal trigeminal neurons**. *Brain Res.* (1974.0) **80** 443-460. PMID: 4153687 121. Sasa M., Igarashi S., Takaori S.. **Influence of the locus coeruleus on interneurons in the spinal trigeminal nucleus**. *Brain Res.* (1977.0) **125** 369-375. PMID: 851884 122. Sasa M., Takaori S.. **Influence of the locus coeruleus on transmission in the spinal trigeminal nucleus neurons**. *Brain Res.* (1973.0) **55** 203-208. PMID: 4351518 123. Sathasivam H.P., Ismail S., Ahmad A.R., Basri N.N., Muhamad H., Mohd. Tahir N.F., Saw C.L., Hj. Kipli N., Lau S.H.. **Trigeminal neuralgia: a retrospective multicentre study of 320 Asian patients**. *Oral Surg. Oral Med. Oral Pathol. Oral Radiol.* (2017.0) **123** 51-57. PMID: 27727102 124. Segal M.. **Serotonergic innervation of the locus coeruleus from the dorsal raphe and its action on responses to noxious stimuli**. *J. Physiol.* (1979.0) **286** 401-415. PMID: 439032 125. Sessle B.J.. **Acute and chronic craniofacial pain: brainstem mechanisms of nociceptive transmission and neuroplasticity, and their clinical correlates**. *Crit. Rev. Oral Biol. Med.* (2000.0) **11** 57-91. PMID: 10682901 126. Sessle B.J., Hu J.W.. **Mechanisms of pain arising from articular tissues**. *Can. J. Physiol. Pharmacol.* (1991.0) **69** 617-626. PMID: 1863912 127. Shankar Kikkeri, N. and S. Nagalli, Trigeminal Neuralgia, in StatPearls. 2021: Treasure Island (FL). 128. Sharp, B.M., Basolateral amygdala and stress-induced hyperexcitability affect motivated behaviors and addiction. Transl. Psychiatry, 2017. 7(8): p. e1194. 129. Shimizu K.. **Differential involvement of trigeminal transition zone and laminated subnucleus caudalis in orofacial deep and cutaneous hyperalgesia: the effects of interleukin-10 and glial inhibitors**. *Mol. Pain* (2009.0) **5** 75. PMID: 20025765 130. Sieberg C.B.. **Neuropathic pain drives anxiety behavior in mice, results consistent with anxiety levels in diabetic neuropathy patients**. *Pain Rep.* (2018.0) **3** 131. Simson P.E., Weiss J.M.. **Alpha-2 receptor blockade increases responsiveness of locus coeruleus neurons to excitatory stimulation**. *J. Neurosci.* (1987.0) **7** 1732-1740. PMID: 2885401 132. Singewald N., Philippu A.. **Release of neurotransmitters in the locus coeruleus**. *Prog. Neurobiol.* (1998.0) **56** 237-267. PMID: 9760703 133. Singh A., Patel D., Li A., Hu L., Zhang Q., Liu Y., Guo X., Robinson E., Martinez E., Doan L., Rudy B., Chen Z.S., Wang J.. **Mapping cortical integration of sensory and affective pain pathways**. *Curr. Biol.* (2020.0) **30** 1703-1715.e5. PMID: 32220320 134. Siqueira S.R., Teixeira M.J., Siqueira J.T.. **Clinical characteristics of patients with trigeminal neuralgia referred to neurosurgery**. *Eur J Dent* (2009.0) **3** 207-212. PMID: 19756195 135. Smith J.G.. **The psychosocial and affective burden of posttraumatic neuropathy following injuries to the trigeminal nerve**. *J. Orofac. Pain* (2013.0) **27** 293-303. PMID: 24171179 136. Stover J.D., Schwab C.A., Matthews M.A.. **Selective deafferentation of convergent inputs to trigeminal subnucleus caudalis: effects on calcitonin gene-related peptide distribution**. *Somatosens. Mot. Res.* (1992.0) **9** 107-130. PMID: 1502861 137. Strittmatter M., Grauer M.T., Fischer C., Hamann G., Hoffmann K.H., Blaes F., Schimrigk K.. **Autonomic nervous system and neuroendocrine changes in patients with idiopathic trigeminal neuralgia**. *Cephalalgia* (1996.0) **16** 476-480. PMID: 8933991 138. Strittmatter M., Grauer M., Isenberg E., Hamann G., Fischer C., Hoffmann K.-H., Biaes F., Schimrigk K.. **Cerebrospinal fluid neuropeptides and monoaminergic transmitters in patients with trigeminal neuralgia**. *Headache* (1997.0) **37** 211-216. PMID: 9150615 139. Sugiyo S., Takemura M., Dubner R., Ren K.e.. **Trigeminal transition zone/rostral ventromedial medulla connections and facilitation of orofacial hyperalgesia after masseter inflammation in rats**. *J. Comp. Neurol.* (2005.0) **493** 510-523. PMID: 16304628 140. Suto T., Severino A.L., Eisenach J.C., Hayashida K.-I.. **Gabapentin increases extracellular glutamatergic level in the locus coeruleus via astroglial glutamate transporter-dependent mechanisms**. *Neuropharmacology* (2014.0) **81** 95-100. PMID: 24495399 141. Takahashi K., Kayama Y., Lin J.S., Sakai K.. **Locus coeruleus neuronal activity during the sleep-waking cycle in mice**. *Neuroscience* (2010.0) **169** 1115-1126. PMID: 20542093 142. Takasu K., Ono H., Tanabe M.. **Gabapentin produces PKA-dependent pre-synaptic inhibition of GABAergic synaptic transmission in LC neurons following partial nerve injury in mice**. *J. Neurochem.* (2008.0) **105** 933-942. PMID: 18182059 143. Takeda M.. **Changes of the excitability of rat trigeminal root ganglion neurons evoked by alpha(2)-adrenoreceptors**. *Neuroscience* (2002.0) **115** 731-741. PMID: 12435412 144. Takemura M., Sugiyo S., Moritani M., Kobayashi M., Yonehara N.. **Mechanisms of orofacial pain control in the central nervous system**. *Arch. Histol. Cytol.* (2006.0) **69** 79-100. PMID: 16819148 145. Tan C.Y., Shahrizaila N., Goh K.J.. **Clinical characteristics, pain, and quality of life experiences of trigeminal neuralgia in a multi-ethnic Asian cohort**. *J. Oral Facial Pain Headache* (2017.0) **31** e15-e20. PMID: 29073667 146. Tang Y.. **Percutaneous trigeminal ganglion radiofrequency thermocoagulation alleviates anxiety and depression disorders in patients with classic trigeminal neuralgia: a cohort study**. *Medicine (Baltimore)* (2016.0) **95** e5379. PMID: 27930513 147. Taylor B.K., Westlund K.N.. **The noradrenergic locus coeruleus as a chronic pain generator**. *J. Neurosci. Res.* (2017.0) **95** 1336-1346. PMID: 27685982 148. Ter Horst G.J., Meijler W.J., Korf J., Kemper RHA. **Trigeminal nociception-induced cerebral Fos expression in the conscious rat**. *Cephalalgia* (2001.0) **21** 963-975. PMID: 11843868 149. Tolle T., Dukes E., Sadosky A.. **Patient burden of trigeminal neuralgia: results from a cross-sectional survey of health state impairment and treatment patterns in six European countries**. *Pain Pract.* (2006.0) **6** 153-160. PMID: 17147591 150. Travagli R.A., Williams J.T.. **Endogenous monoamines inhibit glutamate transmission in the spinal trigeminal nucleus of the guinea-pig**. *J. Physiol.* (1996.0) **491** 177-185. PMID: 9011609 151. Tsai Y.-H., Yuan R., Patel D., Chandrasekaran S., Weng H.-H., Yang J.-T., Lin C.-P., Biswal B.B.. **Altered structure and functional connection in patients with classical trigeminal neuralgia**. *Hum. Brain Mapp.* (2018.0) **39** 609-621. PMID: 29105886 152. Tsuruoka M., Matsutani K., Maeda M., Inoue T.. **Coeruleotrigeminal inhibition of nociceptive processing in the rat trigeminal subnucleus caudalis**. *Brain Res.* (2003.0) **993** 146-153. PMID: 14642840 153. Tyce G.M., Yaksh T.L.. **Monoamine release from cat spinal cord by somatic stimuli: an intrinsic modulatory system**. *J. Physiol.* (1981.0) **314** 513-529. PMID: 7310700 154. Viisanen H., Pertovaara A.. **Influence of peripheral nerve injury on response properties of locus coeruleus neurons and coeruleospinal antinociception in the rat**. *Neuroscience* (2007.0) **146** 1785-1794. PMID: 17445989 155. Wang X.M.. **Effect of antisense knock-down of alpha(2a)- and alpha(2c)-adrenoceptors on the antinociceptive action of clonidine on trigeminal nociception in the rat**. *Pain* (2002.0) **98** 27-35. PMID: 12098614 156. Wang H.. **Central alpha-adrenoceptors contribute to mustard oil-induced central sensitization in the rat medullary dorsal horn**. *Neuroscience* (2013.0) **236** 244-252. PMID: 23333675 157. Wang H.u., Wei F., Dubner R., Ren K.e.. **Selective distribution and function of primary afferent nociceptive inputs from deep muscle tissue to the brainstem trigeminal transition zone**. *J Comp Neurol* (2006.0) **498** 390-402. PMID: 16871539 158. Wei H., Jin C.-Y., Viisanen H., You H.-J., Pertovaara A.. **Histamine in the locus coeruleus promotes descending noradrenergic inhibition of neuropathic hypersensitivity**. *Pharmacol. Res.* (2014.0) **90** 58-66. PMID: 25304183 159. Wu T.H.. **Risk of psychiatric disorders following trigeminal neuralgia: a nationwide population-based retrospective cohort study**. *J. Headache Pain* (2015.0) **16** 64. PMID: 26174508 160. Yokota T., Matsumoto N.. **Somatotopic distribution of trigeminal nociceptive specific neurons within the caudal somatosensory thalamus of cat**. *Neurosci. Lett.* (1983.0) **39** 125-130. PMID: 6633942 161. Yoon S.-Y., Kang S.-Y., Kim H.-W., Kim H.-C., Roh D.-H.. **Clonidine reduces nociceptive responses in mouse orofacial formalin model: potentiation by Sigma-1 receptor antagonist BD1047 without impaired motor coordination**. *Biol. Pharm. Bull.* (2015.0) **38** 1320-1327. PMID: 26328487 162. Yoshizumi M.. **Gabapentin inhibits gamma-amino butyric acid release in the locus coeruleus but not in the spinal dorsal horn after peripheral nerve injury in rats**. *Anesthesiology* (2012.0) **116** 1347-1353. PMID: 22487864 163. Yu L.H.. **Estrogen altered facial mechanical pain threshold and trigeminal P2X3 receptor expression**. *Neuro Endocrinol. Lett.* (2011.0) **32** 811-815. PMID: 22286789 164. Zakrzewska J.M., Wu J., Mon-Williams M., Phillips N., Pavitt S.H.. **Evaluating the impact of trigeminal neuralgia**. *Pain* (2017.0) **158** 1166-1174. PMID: 28114183 165. Zhang K.M.. **alpha2-adrenoceptors modulate NMDA-evoked responses of neurons in superficial and deeper dorsal horn of the medulla**. *J. Neurophysiol.* (1998.0) **80** 2210-2214. PMID: 9772273 166. Zhang Y.. **Dysregulation of pain- and emotion-related networks in trigeminal neuralgia**. *Front. Hum. Neurosci.* (2018.0) **12** 107. PMID: 29662445
--- title: Gastrointestinal conditions related to tooth wear authors: - John P. Howard - Laura J. Howard - Joe Geraghty - A. Johanna Leven - Martin Ashley journal: British Dental Journal year: 2023 pmcid: PMC10038793 doi: 10.1038/s41415-023-5677-0 license: CC BY 4.0 --- # Gastrointestinal conditions related to tooth wear ## Abstract Gastro-oesophageal reflux disease (GORD) is a relatively common condition that occurs in adults and less commonly in children. It develops when the reflux of stomach contents into the oesophagus causes troublesome symptoms and/or complications. Signs and symptoms include heartburn, retrosternal discomfort, epigastric pain and hoarseness, dental erosion, chronic cough, burning mouth syndrome, halitosis and laryngitis. A proportion of patients will, however, have silent reflux. Strongly associated risk factors include family history, age, hiatus hernia, obesity and neurological conditions, such as cerebral palsy. There are different treatment options which may be considered for GORD, consisting of conservative, medical and surgical therapy. Dentists should be aware of the symptoms of GORD and dental signs of intrinsic erosion indicative of possible GORD so that they can question patients about this and, if appropriate, initiate a referral to a general medical practitioner. ## Key points Details the common signs and symptoms of gastro-oesophageal reflux disease (GORD) that dentists should be aware of. The principle causes of GORD are sphincter incompetence, increased gastric pressure and increased gastric volume. There are a number of medical, diet and lifestyle risk factors which contribute to these causes. Management options of GORD consist of conservative, medical and surgical therapy. Dentists should refer patients to their general medical practitioner if undiagnosed GORD is suspected. ## Introduction Gastro-oesophageal reflux disease (GORD) is a relatively prevalent condition worldwide. A 2020 systematic review stated a prevalence rate of $14.2\%$ in adults throughout Europe1,2,3 and $14.53\%$ in the UK.1 GORD is much less common in children, although it is often seen in neurologically impaired children, such as those with cerebral palsy.2 The Royal College of Surgeons' Clinical guidelines for dental erosion defines dental erosion as the irreversible softening and subsequent loss of dental hard tissue due to a chemical process of acid dissolution, but not involving bacterial plaque acid, and not directly associated with mechanical or traumatic factors, or with dental caries.2 Intrinsic acid as a result of GORD is one of the main causes of erosive tooth wear. It can go undiagnosed in patients who may not realise the symptoms of GORD or the potential significance of this. During dental examinations, signs of intrinsic erosive tooth should alert clinicians to a history of possible GORD. Signs of intrinsic acid-related dental erosion are commonly wear on the palatal aspects of maxillary teeth and the occlusal surfaces of mandibular molars, flattened occlusal contours and cupping of cusp tips. Maxillary buccal cervical erosion may also be present if the patient holds gastric contents in their cheeks. Restorations may stand proud and incisal edges may become grooved. As enamel is eroded away, there may be a blueish tinge, and then teeth may appear darker and yellow dentine starts to shine through, often leaving a peripheral ring of enamel.2 Erosive tooth wear tends to leave smoother surfaces compared to other types of wear, although they often occur in combination and a good history will aid diagnosis (Fig. 1, Fig. 2 and Fig. 3 demonstrate these signs, and as such rapid wear had occurred, there has not been any dento-alveolar compensation).Fig. 1 Showing incisal edge wear and buccal wear exposing yellow dentine, with lack of dento-alveolar compensation likely due to the rapid onsetFig. 2 Showing maxillary wear on palatal surfaces, change in occlusal morphology, proud restoration, exposure of dentine with enamel peripheral ring and visualisation of pulp chamberFig. 3 Mandibular wear, showing change in occlusal morphology, hollows and concavities, proud restorations and exposure of dentine These clinical findings, particularly in the absence of significant dietary acids or a disclosed history of an eating disorder, present an opportunity for dentists to ask more questions about a history of GORD. Patients may not be aware of the symptoms of GORD. Dentists are therefore in a unique role as they may be the first healthcare professional to diagnose possible signs of GORD and question the patient. ## Intrinsic acidic sources The Montreal definition states GORD is a condition that develops when the reflux of stomach contents into the oesophagus causes troublesome symptoms and/or complications.4 As dentists, it is not uncommon to be the first healthcare professional to diagnose a systematic disease through observation of its oral manifestation.4,5 The source of intrinsic acid-related to dental erosion is gastric juice.2 Gastric juice is a strongly acidic colourless liquid, with a pH between 1-3. Its immediate secretion is controlled by gastrin release, causing release of the juice that predominately contains hydrochloric acid, the digestive enzyme pepsin and mucus. There are many risk factors for GORD. These can be separated into strong and weak. Strong risk factors include family history, age, hiatus hernia and obesity. The prevalence of GORD is highest in the 35-59 age group, followed by the over-60s, and then the 18-34-year-olds.1 Obese individuals are significantly more at risk of GORD due to the systemic health effects of obesity, as well as the higher intake of food and drinks associated with obesity, such as fatty foods and carbonated drinks.1 Weak risk factors include smoking, stress, high alcohol intake, chocolate/spicy foods, caffeinated drinks, asthma and drug induced (for example, nitrates, calcium channel blockers and non-steroidal anti-inflammatory drugs/aspirin). These may increase GORD due to a combination of a slower rate of digestion, irritation of the oesophagus, higher levels of gastric acid secretion, reduction in lower oesophageal sphincter pressure, and/or a delay in gastric emptying.1 *It is* also important to consider vomiting-related eating disorders as a cause for intrinsic acid wear, such as anorexia nervosa, bulimia nervosa, self-induced vomiting and non-specified eating disorders. If this is suspected, thorough history, explanation and then sign-posting to the patient's general medical practitioner is warranted. ## Signs and symptoms of GORD Systemic symptoms can be either oesophageal, such as heartburn, retrosternal discomfort, epigastric pain and hoarseness,3 or extra-oesophageal, such as chronic cough, burning mouth syndrome, halitosis and laryngitis.1 Dental symptoms may include pain or sensitivity to hot, cold or sweet substances, and in rare cases, dental abscess due to pulpal exposure. Patients may complain of dental signs, such as yellow discolouration of teeth, poor aesthetics due to volume loss, or the feeling fillings have changed or become proud. It is important to note that nearly $25\%$ of adult patients presenting with extensive palatal erosion had pathological GORD diagnosed but did not have any systemic symptoms of reflux. Therefore, in silent reflux, dental erosion may be the only clinical sign present.2,6 Barrett's oesophagus and oesophageal adenocarcinoma are well-known, rare complications of GORD, and any red flag symptoms, such as dysphagia, odynophagia, globus sensation, anorexia, weight loss and bleeding must be treated seriously. In a normal state, gastric contents are prevented from entering the oesophagus by the anti-reflux mechanism. This consists anatomically and functionally of the lower oesophageal sphincter, extrinsic compression from the diaphragm and the acute angle of His (created between the cardia of the stomach and oesophagus) (Fig. 4). GORD occurs when the gastric contents abnormally exit the lower oesophageal sphincter, allowing passage further up the digestive system, causing damage on their journey (Fig. 5).Fig. 4 A closed lower oesophageal sphincter, retaining gastric contents in situFig. 5 An open lower oesophageal sphincter, allowing retrograde passage of gastric contents A summary of principle causes of GORD can be found in Box 1.2 If GORD is suspected, history-taking should include questions about the incidence of belching, heartburn, stomach aches, acidic taste, voice change, hoarseness, chronic cough, vomiting, halitosis, choking and excess salivation to aid onward referral.7,8 ## Management of GORD For many patients, GORD is a chronic relapsing condition. Dentists should refer (with the patient's permission) to the patient's general medical practitioner, where GORD is suspected as an aetiological factor for presenting with erosive tooth wear.2 *Diagnosis is* often clinical; however, upper endoscopy is warranted for red flag symptoms, or no improvement after eight weeks of medical treatment. The role of endoscopy is to confirm diagnosis (erosion/ulcerations or non-erosive reflux disease), exclude atypical causes (eosinophilic esophagitis, candida, herpes simplex) and diagnose complications (Barrett's oesophagus, stricture, adenocarcinoma) (Fig. 6, Fig. 7, Fig. 8).Fig. 6 A tight peptic stricture (caused by long-term acid exposure) at the lower oesophageal sphincter regionFig. 7 Barrett's oesophagus of the lower oesophagus, a premalignant condition caused by long-term acid exposureFig. 8 Grade A oesophagitis caused by excess acid exposure within the lower oesophagus When initial investigations are inconclusive or symptoms persist despite standard management, further investigations may be required in tertiary care. These include assessing the pH in the lower oesophagus (to look at absolute amounts of reflux and concordance with symptoms), high-resolution manometry of the oesophagus (to exclude motility disorders such as achalasia, which can mimic reflux), and also assess whether motility is sufficiently preserved to allow more invasive treatments, such as anti-reflux surgery (for example, a Nissen's fundoplication) or newer treatments (such as magnetic augmentation of the lower oesophageal sphincter). Management options for GORD consists of conservative, medical and surgical therapy. Conservative methods include life-style modification, such as dietary changes, smoking cessation and weight loss management. Dentists are in a good position to give specific dietary advice for GORD-related tooth wear, such as reduced frequency of dietary acid intake (especially at bedtime), avoidance of reflux-provoking foods, and the use of sugar free chewing gum after an acid exposure to increase salivary flow and encourage tooth remineralisation.2,9 Medical methods are most commonly antisecretory drugs, such as protein pump inhibitors (for example, omeprazole) although H2 blockers (for example, ranitidine) and newer therapies, such as K+ competitive acid blockers (for example, vonoprazan), are also used. Antacids to neutralise acid, and alginates which precipitate into a gel on contact with acid, are often used as over-the-counter self-medication by patients. It is important to note that although most antacids and alginates are sugar-free, not all of them are. As alginates are commonly taken last thing at night (to neutralise acid movement when the patient is prone), this may add a cariogenic risk factor, which patients should be informed of. Dentists can provide interventions to help with remineralisation and sensitivity of teeth, such as the use of high fluoride toothpastes and mouthwashes, GC Tooth Mousse, regular application of fluoride varnish and application of resin sealants or dentine bonding agents to provide temporary sensitivity relief.2,9 Lastly, if other therapies are unsuccessful, surgery may be offered. There are a number of options for this, such as fundoplication (either surgical or endoscopic), and other endoscopic procedures. Each of these have general indications and side-effects that a patient will discuss with their surgical team. ## Conclusion Dentists may be the first clinicians to detect signs of GORD and therefore play an important role in screening for GORD. Patients may have silent reflux, not be aware of the symptoms of GORD, or that the frequency with which they are experiencing GORD might be a cause for concern. Dentists therefore can have a positive role in providing patient education and offering referral to the general medical practitioners. Early detection may prevent long-term gastro-oesophageal complications and progression of the tooth wear condition. ## References 1. 1.Nirwan J S, Hasan S S, Babar Z U, Conway B R, Ghori M U. Global Prevalence and Risk Factors of Gastro-oesophageal Reflux Disease (GORD): Systematic Review with Meta-analysis. Sci Rep 2020; 10: 5814. 2. 2.O'Sullivan E, Toor I, Brown L, Watkins S. Clinical guidelines for dental erosion. pp 1-29. London: Faculty of Dental Surgery, the Royal College of Surgeons of England, 2021. 3. 3.El-Serag H B, Sweet S, Winchester C C, Dent J. Update on the epidemiology of gastro-oesophageal reflux disease: a systematic review. Gut 2014; 63: 871-880. 4. 4.Vakil N, van Zanten S V, Kahrilas P, Dent J, Jones R. The Montreal definition and classification of gastroesophageal reflux disease: a global evidence-based consensus. Am J Gastroenterol 2006; 101: 1900-1920. 5. 5.Dundar A, Sengun A. Dental approach to erosive tooth wear in gastroesophageal reflux disease. Afr Health Sci 2014; 14: 481-486. 6. 6.Bartlett D W, Evans D F, Anggiansah A, Smith B G. A study of the association between gastro-oesophageal reflux and palatal dental erosion. Br Dent J 1996; 181: 125-132. 7. 7.Donovan T. Dental erosion. J Esthet Restor Dent 2009; 21: 359-364. 8. 8.Hungin A P S, Molloy-Bland M, Scarpignato C. Revisiting Montreal: New Insights into Symptoms and Their Causes, and Implications for the Future of GERD. Am J Gastroenterol 2019; 114: 414-421. 9. 9.Broliato G A, Volcato D B, Reston E G et al. Esthetic and functional dental rehabilitation in a patient with gastroesophageal reflux. Quintessence Int 2008; 39: 131-137.
--- title: Meal and Physical Activity Detection from Free-living Data for Discovering Disturbance Patterns to Glucose Levels in People with Diabetes authors: - Mohammad Reza Askari - Mudassir Rashid - Xiaoyu Sun - Mert Sevil - Andrew Shahidehpour - Keigo Kawaji - Ali Cinar journal: BioMedInformatics year: 2022 pmcid: PMC10038808 doi: 10.3390/biomedinformatics2020019 license: CC BY 4.0 --- # Meal and Physical Activity Detection from Free-living Data for Discovering Disturbance Patterns to Glucose Levels in People with Diabetes ## Abstract ### Objective: Interpretation of time series data collected in free-living has gained importance in chronic disease management. Some data are collected objectively from sensors and some are estimated and entered by the individual. In type 1 diabetes (T1D), blood glucose concentration (BGC) data measured by continuous glucose monitoring (CGM) systems and insulin doses administered can be used to detect the occurrences of meals and physical activities and generate the personal daily living patterns for use in automated insulin delivery (AID). ### Methods: Two challenges in time-series data collected in daily living are addressed: data quality improvement and detection of unannounced disturbances to BGC. CGM data have missing values for varying periods of time and outliers. People may neglect reporting their meal and physical activity information. In this work, novel methods for preprocessing real-world data collected from people with T1D and detection of meal and exercise events are presented. Four recurrent neural network (RNN) models are investigated to detect the occurrences of meals and physical activities disjointly or concurrently. ### Results: RNNs with long short-term memory (LSTM) with 1D convolution layers and bidirectional LSTM with 1D convolution layers have average accuracy scores of $92.32\%$ and $92.29\%$, and outper-form other RNN models. The F1 scores for each individual range from $96.06\%$ to $91.41\%$ for these two RNNs. ### Conclusions: RNNs with LSTM and 1D convolution layers and bidirectional LSTM with 1D convolution layers provide accurate personalized information about the daily routines of individuals. Significance: Capturing daily behavior patterns enables more accurate future BGC predictions in AID systems and improves BGC regulation. ## Introduction Time series data are widely used in many fields and various data-driven modeling techniques are developed to represent the dynamic characteristics of systems and forecast the future behavior. The growing research in artificial intelligence has provided powerful machine learning (ML) techniques to contribute to data-driven model development. Real-world data provides several challenges to modeling and forecasting, such as missing values and outliers. Such imperfections in data can reduce the accuracy of ML and the models developed. This necessitates data preprocessing for imputation of missing values, down-and up-sampling, and data reconciliation. Data preprocessing is a laborious and time-consuming effort since big data are usually stacked on a large scale [1]. When models are used for forecasting, the accuracy of forecasts improve if the effects of future possible disturbances based on behavior patterns extracted from historical data are incorporated in the forecasts. This paper focuses on these two problems and investigates the benefits of preprocessing the real-world data and the performance of different recurrent neural network (RNN) models for detecting various events that affect blood glucose concentration (BGC) in people with type 1 diabetes (T1D). The behavior patterns detected are used for more accurate predictions of future BGC variations, which can be used for warnings and for increasing the effectiveness of automated insulin delivery (AID) systems. Time series data captured in daily living of people with chronic conditions have many of these challenges to modeling, detection, and forecasting. Focusing on people with T1D, the medical objective is to forecast the BGC of a person with T1D and prevent the excursion of BGC outside a “desired range” (70–180 mg/dL) to reduce the probability of hypo- and hyperglycemia events. In recent years, the number of people with diabetes has grown rapidly around the world, reaching pandemic levels [2,3]. Advances in continuous glucose monitoring (CGM) systems, insulin pump and insulin pen technologies, and in novel insulin formulations enabled many powerful treatment options [4–9]. The current treatment options available to people with T1D range from manual insulin injections to AID. Manual injection (insulin bolus) doses are computed based on the person’s characteristics and the properties of the meal consumed. Current AID systems necessitate manual entry of meal information to give insulin boluses for mitigating the effects of meal on the BGC. Manual adjustment of basal insulin dose, increasing the BGC target level and/or consumption of snacks are the options to mitigate the effects of physical activity. Some people may forget to make these manual entries and a system that can nudge them for providing appropriate information can reduce the extreme excursions in BGC. Commercially available AID systems are hybrid closed-loop systems and they require these manual entries by the user. AID systems, also called artificial pancreas (AP), consist of a CGM, an insulin pump, and a closed-loop control algorithm that manipulates the insulin infusion rate delivered by the pump based on the recent CGM values reported [10–23]. More advanced AID systems that use a multivariable approach [10,24–26] use additional inputs from a wearable devices (such as wristbands) to automatically detect the occurrence of physical activity and incorporate this information to the automated control algorithms for a fully-automated AID system [27]. Most AID systems use model predictive control techniques that predict future BGC values in making their insulin dosing decisions. Knowing the habits of the individual AID user improves the control decisions since the prediction accuracy of the future BGC trajectories can explicitly incorporate the future potential disturbances to the BGC, such as meals and physical activities, that will occur with high likelihood during the future BGC prediction window [24,26]. Consequently, the detection of meal and physical activity events from historical free-living data of a person with T1D will provide useful information for decision-making by both the individual and by the AID system. CGM systems report subcutaneous glucose concentration to infer BGC with a sampling rate of 5 minutes. Self-reported meal and physical activity data are often based on diary entries. Physical activity data can also be captured by wearable devices. The variables reported by wearable devices may have artifacts, noise, missing values, and outliers. The data used in this work includes only CGM values, insulin dosing information, and diary entries of meals and physical activities. Analyzing long-term data of people with T1D indicates that individuals tend to repeat daily habitual behaviors. Figure 1 illustrates the probability of physical activity and meal (indicated as carbohydrate intake) events, either simultaneously or disjointly, for 15 months of CGM, meal, insulin pump, and physical activity self-recorded data of individuals with T1D. Major factors affecting BGC variations usually occur at specific time windows and conditions, and some combinations of events are mutually exclusive. For example, insulin-bolusing and physical activity are less likely to occur simultaneously or during hypoglycemia episodes, since people do not exercise when their BGC is low. People may have different patterns of behavior during the work week versus weekends or holidays. Predicting the probabilities of exercise, meal consumption, and their concurrent occurrence based on historical data using ML can provide important information on the behavior patterns for making medical therapy decisions in diabetes. Motivated by the above considerations, this work develops a framework for predicting the probabilities of meal and physical activity events, including their independent and simultaneous occurrences. A framework is built to handle the inconsistencies and complexities of real-world data, including missing data, outlier removal, feature extraction, and data augmentation. Four different recurrent neural network (RNN) models are developed and evaluated for estimating the probability of events causing large variations in BGC. The advent of deep neural networks (NN) and their advances have had paved the way for processing and analyzing various types of information, namely: time-series, spatial, and time series-spatial data. Long short-term memory (LSTM) NN models are specific sub-categories of recurrent NNs introduced to reduce the computational burden of storing information over extended time intervals [28,29]. LSTMs take advantage of nonlinear dynamic modeling without knowing time-dependency information in the data. Moreover, their multi-step ahead prediction capability makes them an appropriate choice for detecting upcoming events and disturbances that can deteriorate the accuracy of model predictions. The main contributions of this work are the development of NN models capable of estimating the occurrences of meals and physical activities without requiring additional bio-signals from wearable devices, and the integration of convolution layers with LSTM that enable the NN to accurately estimate the output from glucose-insulin input data. The proposed RNN models can be integrated with the control algorithm of an AID system to enhance its performance by readjusting the conservativeness and aggressiveness of the AID system. The remainder of this paper is organized as follows: The next section provides a short description of the data collected from people with T1D. The preprocessing step, including outlier removal, data imputation, and feature extraction is presented in Section 3. Section 4 presents various RNN configurations used in this study. A case study with real-world data and a discussion of the results are presented in Sections 5 and 6, respectively. Finally, Section 6 provides the conclusions. ## Free-living, Self-reported Dataset of People with T1D A total of 300 self-collected T1D datasets were made available for research, and each dataset represents a unique individual. Among all the datasets, 50 T1D datasets include CGM sensor-insulin pump recordings and exercise information such as the time, type, and duration of physical activity recorded from either open/closed-loop insulin pump-sensor data. Meal information is reported as amount of carbohydrates (CHO) consumed in the meal as estimated by the subject. Over or underestimation of CHO in meals is common. The subjects with T1D selected for this study used insulin pump-CGM sensor therapy for up to two years, and some of them have lived with diabetes for more than fifty years Tables 1 and 2 summarize the demographic information of the selected subjects and the definition of the variables collected, respectively. Separate RNN models are developed for each person in order to capture personalized patterns of meal consumption and physical activity. ## Data Preprocessing Using real-world data for developing models usually has numerous challenges: (i) the datasets can be noisy and incomplete; (ii) there may be duplicate CGM samples in some of the datasets; (iii) inconsistencies exist in the sampling rate of CGM and insulin values; (iv) gaps in the time and date can be found due to insulin pump or CGM sensor disconnection. Therefore, the datasets need to be preprocessed before using them for model development. ## Sample Imputation Estimating missing data is an important step before analyzing the data [30]. Missing data is substituted with reasonable estimates (imputation) [31]. In dealing with time-series data such as CGM, observations are sorted according to their chronological order. Therefore, the variable “Time”, described in Table 2, is converted to “Unix time-stamp” and samples are sorted in ascending order of “Unix time-stamp” and gaps without observations are filled with pump-sensor samples labeled as “missing values.” Administered basal insulin is a piecewise constant variable and its amount is calculated by the AID system or by predefined insulin injection scenarios. Applying a simple forward or backward imputation for basal insulin with gaps in duration lasting a maximum of two hours gives reasonable reconstructed values for the missing observations. Gaps lasting more than two hours in missing recordings are imputed with basal insulin values recorded in the previous day at the same time, knowing that insulin injection scenarios usually follow a daily pattern [32]. Variable “Bolus” is a sparse variable (usually nonzero only at times of meals) and its missing samples are imputed with the median imputation approach, considering that bolus injection policy is infrequently altered. Similarly, missing recordings of variables “Nutrition.carbohydrate”, “Smbg”, “Duration”, “Activity.duration”, and “Distance.value” are imputed with the median strategy. A multivariate strategy, which uses CGM, total injected insulin, “Nutrition.carbohydrate”, the “Energy.value”, and “Activity.duration”, is employed to impute missing CGM values. This choice of variables has to do with the dynamic relationship between CGM and the amount of carbohydrate intake, the duration and the intensity of physical activity, and total injected insulin. Estimates of missing CGM samples are obtained by performing probabilistic principal component analysis (PPCA) on the lagged matrices of the CGM data. PPCA is an extension of principal component analysis where the Gaussian conditional distribution of the latent variables is assumed [33]. This formulation of the PPCA facilitates tackling the problem of missing values in the data through the maximum likelihood estimation of the mean and variance of the original data. Before performing PPCA on the feature variables, the lagged array of each feature variable, Xk,j,k∈{CGM,Ins,CHO,EV,AD}, at the jth sampling index is constructed from the past two hours of observations as: [1] Xk,j=[Xk,j,Xk,j−1…Xk,j−24]1×25′k∈{CGM,Ins,CHO,EV,AD}Xj=[X1,j,…,Xk,j…,XM,j]T,X=[X1,…,XN]M×N For an observed set of feature variables Xj, let Tj=[T1,j,…,Tq,j]T be its q-dimensional (q≤M) Gaussian latent transform [34] such that [2] Xi,j=WiTj+μi+ϵi,j where Wi=[Wi,1,…,Wi,q]∈ℝq and μ_=[μ1,…,μM]T∈ℝM represent the ith row of the loading matrix W∈ℝM×q and mean value of the data. ϵi,j∈ℝ is also the measurement noise with the probability distribution [3] p(ϵi,j∣σ2)=N(ϵi,j∣0,σ2). Based on the Gaussian distribution assumption of Tj, and the Gaussian probability distribution of ϵi,j, one can deduce that [4] {p(Tj)=N(Tj∣0,Iq)p(Xi,j∣μi,Wi,σ2)=N(Xi,j∣μi,WiWiT+σ2)p(Xi,j∣Tj,μi,Wi,σ2)=N(Xi,j∣WiTj+μi,σ2). The joint probability distribution p(Xi,j,Tj,μi,Wi,σ2) can be derived from [4] and Bayes’ joint probability rule as [5] p(Xi,j,Tj,μi,Wi,σ2)=1(2πσ2)M2exp(Xi,j−WiTj−μi)2−2σ21(2π)q2exp−TjTTj2 Define the set η={(i,j)∣1≤i≤M,1≤j≤N,Xi,j≠NaN}. The log-likelihood of the joint multivariate Gaussian probability distribution of [5] is calculated over all available observations as [6] ln(p(Xi,j,Tj∣μi,Wi,σ2))=∑i,j∈η∑[ln(p(Xi,j∣Tj,μi,Wi,σ2))+lnp(Tj)]=∑i,j∈η∑−M2ln(2πσ2)−q2ln(2π)−(Xi,j−WiTj−μi)22σ2−TjTTj2 where the log-likelihood [6] is defined for all available observations Xi,j, i,j∈η. By applying the expectation operation with respect to the posterior probability distribution over all latent variables Tj, j∈ηi, where ηi={j∣1≤j≤N,Xi,j≠NaN} [6] becomes [7] E{L}=−∑i,j∈η∑M2ln(σ2)+12E{TjTTj}+12σ2(Xi,j−μi)2−1σ2E{TjT}WiT(Xi,j−μi)+12σ2E{TjTTj}WiWiT Maximizing [7] is feasible by setting all partial derivatives ∂E{L}∂σ2, ∂E{L}∂μi2, and ∂E{L}∂Wi2, $i = 1$,…,M, $j = 1$,…,N to zero [34]. Parameters μi, σ2, and Wi in [8] are updated recursively until they converge to their final values. The final estimation of missing CGM samples is obtained by performing diagonal averaging of the reconstructed lagged matrix X^∈ℝM×N over rows/columns filled with CGM values. Long gaps in CGM recordings might exist in the data, and imputing their values causes problems in accuracy and reliability. Therefore, CGM gaps no more than twenty-five consecutive missing samples (about two hours) are imputed by PPCA. **Algorithm 1** | 1: | procedure Outlierrejectoon(CGM, Smbg, CHO, AD, InsBolus) | procedure Outlierrejectoon(CGM, Smbg, CHO, AD, InsBolus).1 | | --- | --- | --- | | 2: | for i = 1 : N do | ▹ Removing samples outside of the calibration range | | 3: | if CGMk>400mg/dL or CGMk<0mg/dL then | if CGMk>400mg/dL or CGMk<0mg/dL then | | 4: | CGMk←NaN | CGMk←NaN | | 5: | end if | end if | | 6: | end for | end for | | 7: | for i = 2 : N do | for i = 2 : N do | | 8: | ΔCGMk←CGMk−CGMk−1 | ΔCGMk←CGMk−CGMk−1 | | 9: | if ΔCGMk>30mg/dL & all ({CHOk,…,CHOk−9}==0) then | if ΔCGMk>30mg/dL & all ({CHOk,…,CHOk−9}==0) then | | 10: | CGMk←NaN | CGMk←NaN | | 11: | end if | end if | | 12: | if ΔCGMk<30mg/dL & all ({InsBolus,k,…,…,InsBolus,k−6}==0) then | if ΔCGMk<30mg/dL & all ({InsBolus,k,…,…,InsBolus,k−6}==0) then | | 13: | CGMk←NaN | CGMk←NaN | | 14: | end if | end if | | 15: | if ΔCGMk<30mg/dL & all ({ADk,…,ADk−6}==0) then | if ΔCGMk<30mg/dL & all ({ADk,…,ADk−6}==0) then | | 16: | CGMk←NaN | CGMk←NaN | | 17: | end if | end if | | 18: | if Smbgk≠NaN & CGMk≠NaN & abs(Smbgk−CGMk)>18mg/dL then | if Smbgk≠NaN & CGMk≠NaN & abs(Smbgk−CGMk)>18mg/dL then | | 19: | CGMk←NaN | CGMk←NaN | | 20: | end if | end if | | 21: | end for | end for | | 22: | return CGM | return CGM | | 23: | end procedure | end procedure | ## Outlier Removal Signal reconciliation and outlier removal are necessary to avoid misleading interpretation of data and biased results, and to improve the quality of CGM observations. As a simple outlier removal approach for a variable with Gaussian distribution, observations outside ±2.72 standard deviations from the mean, known as Inner Tukey Fences, can be labeled outliers and extreme values [35]. The probability distribution of the CGM data shows a skewed distribution compared to the Gaussian probability distribution. Thus, labeling samples as outliers only based on their probability of occurrence is not the proper way of removing extreme values from the CGM data since it can cause loss of useful CGM information, specifically during hypoglycemia (CGM<70mg/dL) and hyperglycemia (CGM>180mg/dL) events. As another alternative, extreme values, and spikes in the CGM data can be labeled from the prior knowledge and by utilizing other feature variables, namely: “Smbg,” “Nutrition.carbohydrate,” “Bolus,” and “Activity.duration.” Algorithm 1 is proposed to remove outliers from CGM values. Usually, BGC is slightly different from the recordings of CGM signal because of delay between BGC and subcutaneous glucose concentration measured by the CGM device and sensor noise. The noisy signal can deteriorate the performance of data-driven models. Therefore, Algorithm 2, which is based on eigendecomposition of the Hankel matrix of CGM values, is used to reduce the noise in the CGM recordings. ## Feature Extraction Converting raw data into informative feature variables or extracting new features is an essential step of data preprocessing. In this study, four groups of feature variables, including frequency domain, statistical domain, nonlinear domain, and model-based features are calculated and added to each dataset to enhance the prediction power of models. The summarized description of each group of features and the number of past samples required for their calculation are listed in Table 3. Qualitative trend analysis of variables can extract different patterns caused by external factors within a specified time [36,37]. A pairwise multiplication of the sign and magnitude of the first and second derivatives of CGM values indicates carbohydrate intake [38,39], exogenous insulin injection, and physical activity. Therefore, the first and second derivatives of CGM values, calculated by the fourth-order backward difference method are added as feature variables. The sign and magnitude product of the first and second derivatives of CGM, their covariance, Pearson correlation coefficient, and Gaussian kernel similarity are extracted. Statistical feature variables, e.g., mean, standard deviation, variance, skewness, etc., are obtained from the specified time window of CGM values. Similar to the first and second derivatives of CGM values, a set of feature variables, including covariance and correlation coefficients, from pairs of CGM values and derivatives are extracted and augmented to the data. Because of the daily repetition in the trends of CGM and glycemic events, and the longer time window of CGM values, samples collected during the last twenty-four hours are used for frequency-domain feature extraction. Therefore, magnitudes and frequencies of the top three dominant peaks in the power spectrum of CGM values, conveying past long-term variation of the BGC, are included in the set of feature maps. **Algorithm 2** | 1: | procedure cgmdenoising(CGM) | ▹ Smoothing CGM recordings | | --- | --- | --- | | 2: | Qi=[CGMd,…,CGMd+qi−1] | ▹ Qi∈ℝqiisith consecutive CGM recordings | | 3: | qi←|Qi|, pi←floor(qi2), wi←qi−pi+1 | qi←|Qi|, pi←floor(qi2), wi←qi−pi+1 | | 4: | [Ui,Si,Vi]=SVD(Ai) | ▹ Ai∈ℝwi×pi is the Hankel matrix made of Qi | | 5: | S^i←zeros(pi,pi) | S^i←zeros(pi,pi) | | 6: | η←cumsum([s1,…,spi])sum([s1,…,spi]) | ▹ sj>0 are eigenvalues of Si in descending order | | 7: | for j=1: pi do | for j=1: pi do | | 8: | if ηj>0.95 then | if ηj>0.95 then | | 9: | S^i(j,j)←0 | S^i(j,j)←0 | | 10: | else | else | | 11: | S^i(j,j)←Si(j,j) | S^i(j,j)←Si(j,j) | | 12: | end if | end if | | 13: | end for | end for | | 14: | A^i=UiS^iViT | A^i=UiS^iViT | | 15: | Q^i←Diagonalaveraging(A^i) | ▹ Q^i=[CG^Md,…,CG^Md+qi−1] | | 16: | return CG^M | return CG^M | | 17: | end procedure | end procedure | Plasma insulin concentration (PIC) is another feature variable that informs about the carbohydrate intake information and exogenous insulin administration. PIC accounts for the accumulation of subcutaneously injected insulin within the bloodstream, which is gradually consumed by the body to enable the absorption of carbohydrates released from the gastrointestinal track to various cells and tissues. Usually, dynamic physiological models are used to describe and model the glucose and insulin concentration dynamics in diabetes. The main idea of estimating PIC from physiological models stems from predicting the intermediate state variables of physiological models by designing a state observer and utilizing total infused insulin and carbohydrate intake as model inputs, and CGM values as the output of the model [40–42]. In this work, the estimation of PIC and glucose appearance rate are obtained from a physiological model known as Hovorka’s model [43]. Equation [9] presents this nonlinear physiological (compartment) model: [9] dS1(t)dt=Ins(t)−S1(t)tmax,IdS2(t)dt=S1(t)tmax,I−S2(t)tmax,IdI(t)dt=S2(t)tmax,IVI−KeI(t)dx1(t)dt=kb,1I(t)−ka,1x1(t)dx2(t)dt=kb,2I(t)−ka,2x2(t)dx3(t)dt=kb,3I(t)−ka,3x3(t)dQ1(t)dt=Ug(t)−F0,1c(t)−FR(t)−x1(t)Q1(t)+k12Q2(t)+EGP0(1−x3(t))dQ2(t)dt=x1(t)Q1(t)−(k12+x2(t))Q2(t)dGsub(t)dt=1τ(Q1(t)Vg−Gsub(t)) Model [9] is comprised of four sub-models, describing the action of insulin on glucose dynamics, the insulin absorption dynamics, plasma-interstitial tissue glucose concentration dynamics, and the blood glucose dynamics. The state variables of [9], the nominal values of the parameters, and their units are listed in Table 4 [43] Body weight has a significant effect on the variations of the PIC and other state variables as it is used for determining the amount of exogenous insulin to be infused. Although estimating body weight as an augmented state variable of the insulin-CGM model is an effective strategy to cope with the problem of unavailable demographic information, estimating body weight from total amount of daily administered insulin is a more reliable approach. As reported in various studies, total daily injected insulin can have a range 0.4–1.0 units. kg−1.day−1 [44–46]. A fair estimation of body weight can be obtained by calculating the most common amount of injected basal/bolus insulin for each subject and using a conversion factor of 0.5 units.kg−1.day−1 as a rule of thumb to estimate the body weight. The insulin-glucose dynamics [9] in discrete-time format are given by [10] Xk+1′=f′(Xk′,Uk)+Gkωk,ωk≈N(0,Q)Yk′=h′(Xk′)+vk,vk≈N(0,R) where Xk′=[S1,k,S2,k,Ik,x1,k,x2,kx3,k,Q1,k,Q2,k,Gsub,k,tmax,I,k,ke,k,UG,k]∈Rnx denotes the extended state variables and *Uk is* the total injected exogenous insulin. Symbols ωk and νk denote zero-mean Gaussian random process and measurement noises (respectively), representing any other uncertainty and model mismatch that are not taken into account. Further, Q∈Rnx×nx and R∈R represent the positive definite system uncertainty and measurement noise covariance matrices, respectively. Tracking the dynamics of internal state variables of the model [10] is feasible by using a class of Sequential Monte Carlo algorithms known as particle filters. A generic form of the particle filter algorithm proposed by [47] with efficient adaptive Metropolis-Hastings resampling strategy developed in [49] is employed to predict the trajectory of the PIC and other state variables. In order to avoid any misleading state estimations, each state variable is subjected to a constraint to maintain all estimations within meaningful intervals [41]. ## Feature Selection and Dimensionality Reduction Reducing the number of redundant feature variables lowers the computational burden of their extraction and hinders over-parameterized modeling. In this work, a two-step feature selection procedure is used to obtain the optimal subset of feature variables that boost the efficiency of the classifier the most. In the first step, the deviance statistic test is performed to filter out features with low significance (P-value>0.05). In the second step, the training split of all datasets was used in the wrapper feature selection strategy to maximize the accuracy of the classifier in estimating the glycemic events. Sequential floating forward selection (SFFS) approach [50] was applied on a random forest estimator with thirty decision tree classifiers with a maximum depth of six layers to sort out features with the most predictive power in descending order. Consequently, the top twenty feature variables with the highest contribution to the classification accuracy enhancement are used for model development. ## Detection and Classification Methods Detecting the occurrence of events causing large glycemic variations requires solving a supervised classification problem. Hence, all samples required labeling using the information provided in the datasets, specifically using variables “Activity.duration” and “Nutrition.carbohydrate.” In order to determine index sets of each class, let N to be total number of samples and T(k)=ceil(AD(k)/(3×105)) be the sample duration of physical activity at each sampling time k. Define sets of sample indexes as: [11] {Label. Index{1,1}:={i∣k≤i≤k+T(k)−1,$k = 1$,…,N,T(k)≠0,Nutrition.carbohydrate(j)≠0,j=k+1,…,k+T(k)}Label. Index{0,1}:={i∣k≤i≤k+T(k)−1,$k = 1$,…,N,T(k)≠0,Nutrition.carbohydrate(j)=0,j=k+1,…,k+T(k)}Label. Index{1,0}:={k∣1≤k≤N,T(k)=0,Nutrition.carbohydrate(k)≠0}Label. Index{0,0}:={k∣1≤k≤N,T(k)=0,Nutrition.carbohydrate(k)=0} The label indexes defined by [11] corresponds to classes “Meal and Exercise,” “no Meal but Exercise,” “no Exercise but Meal,” “neither Meal nor Exercise,” respectively. Four different configurations of the RNN models are studied to assess the accuracy 249 and performance of each in estimating the joint probability of the carbohydrate intake and physical activity. All four models use 24 past samples of the selected feature variables and Algorithm 3The generic particle filter algorithm [47,48]1:procedure ParticleFilter(xk−1=[xk−11,…,xk−1Ns],wk−1=[wk−11,…,wk−1Ns],zk)2: if k == 0 then▹ Initialization step3: for $i = 1$: Ns do4: w0i←1Ns5: x0i∼N(η,Σ)6: end for7: return x0, w08: else9: for $i = 1$: Ns do10: xki∼π(xk∣xk−1i,zk)▹ Propagate particles11: wki∝wk−1ip(yk∣xki)p(xki∣xk−1i)π(xki∣xk−1i,yk)▹ Update importance sampling weights12: end for13: for $i = 1$: NS do14: wki←wki∑$j = 1$Nwkj▹ Normalize importance sampling weights15: end for16: N^eff←1∑$j = 1$N(wkj)2▹ Calculate the effective sampling size17: if N^eff<Ns then▹ Check for degeneracy issue18: [xk∗,wk∗]=Resample(xk=[xk1,…,xkNs],wk=[wk1,…,wkNs])▹ Resample particles[49]19: xk←xk∗20: wk←wk∗21: end if22: return xk, wk23: end if24:end procedure event estimations are performed one sample backward. Estimating the co-occurrences of the external disturbances should be performed at least one step backward as the effect of disturbance variables needs to be seen first, and then, parameter adjustment and event prediction can be made. Since the imputation of gaps with a high number of consecutive missing values adversely affect the prediction of meal-exercise classes, all remaining samples with missing values after the data imputation step are excluded from parameter optimization. Excluding missing values inside the input tensor can be done either by using a placeholder for missing samples and filtering samples through masking layer or manually removing incomplete samples. The first NN model consists of a masking layer to filter out unimputed samples, followed by a LSTM layer, two dense layers, and a softmax layer to estimate the probability of each class. The LSTM and dense layers are undergo training with dropout and parameter regularization strategies to avoid the drastic growth of hyperparameters. Additionally, the recurrent information stream in the LSTM layer was randomly ignored in the calculation at each run. At each layer of the network, the magnitude of both weights and intercept coefficients was penalized by adding a Li regularizer term to the loss function. The Rectified Linear Unit (ReLu) activation function was chosen as a nonlinear component in all layers. The input variables of the regular LSTM network will have the shape of N×m×L, that denotes the size of samples, the size of lagged samples, and the number of feature variables, respectively. The second model encompasses a series of two 1D convolution layers, each one followed by a max pool layer for downsampling feature maps. The output of the second max pool layer was flattened to achieve time-series extracted feature to feed to to the LSTM layer. A dense layer after LSTM was added to the model and the joint probability of events estimated by calculating the output of the softmax layer. Like the first RNN model, the ReLU activation function was employed in all layers to capture the nonlinearity in the data. L1 regularization method was applied to all hyperparameters of the model. Adding convolution layers with repeated operations to an RNN model paves the way for extracting features for sequence regression or classification problem. This approach has shown a breakthrough in visual time-series prediction from the sequence of images, or videos, for various problems such as activity recognition, textual description, and audio and word sequence prediction [52,53]. Time-distributed convolution layers scan and elicit features from each block of the sequence of the data [54]. Therefore, each sample is reshaped into m×n×L, with $$n = 1$$ blocks at each sample. The third classifier has a 2D convolutional LSTM (ConvLSTM) layer, one dropout layer, two dense, and a softmax layer for probability estimation of each class from the sequences of data. 2-D ConvLSTM structure was designed to capture both temporal and spatial correlation in the data, moving pictures in particular, by employing convolution operation in both input-to-state, and state-to-state transitions [51]. In comparison to a regular LSTM cell, ConvLSTMs perform convolution operation by internal multiplication of inputs and hidden states into kernel filter matrices (Figure 2 c). Similar to previously discussed models, L1 regularization constraint and ReLU activation function are considered in constructing the ConvLSTM model. 2D ConvLSTM import sample of spatiotemporal data in the format of m×s×n×L, where $s = 1$ and $$n = 1$$ stand for the size of the rows and columns of each tensor and $L = 20$ is the number of channels/features on the data [55]. Finally, the last model is comprised of two 1D convolution layers, two max pooling layers, a flatten layer, a bidirectional LSTM (Bi-LSTM) layer, a dense layer, and soft max layer to predict classes. Bi-LSTM units capture the dependency in the sequence of the data in two directions. Hence, as a comparison to a regular LSTM memory unit, Bi-LSTM requires to reversely duplicate the same LSTM unit and employ a merging strategy to calculate the output of the cell [56]. The use of this approach was primarily observed in speech recognition tasks, where instead of real-time interpretation, the whole sequence of the data was analyzed and its superior performance over the regular LSTM was justified [57]. The joint estimation of glycemic events is made one step backward. Therefore, the whole sequence of features are recorded first, and use of an RNN model with Bi-LSTM units for the detection of unannounced disturbances is quite justifiable. The tensor of input data is similar to LSTM with 1D convolutional layers. Figure 2 is the schematic diagram of a regular LSTM, a Bi-LSTM, and a ConvLSTM unit. Figure 3 depicts the structure of the four RNN models to estimate the probability of meal consumption, physical activity, and their concurrent occurrence. The main difference between models (a) and (b) in Figure 3 is convolution and max-pooling layers added before the LSTM layer to extract features map from time series data. Although adding convolutional blocks to an RNN model increases the number of learnable parameters, including weights, biases, and kernel filters, calculating temporal feature maps from input data enhances better discriminates the target classes. ## Case Study Eleven datasets containing CGM sensor-insulin pump, physical activity, and carbohydrate intake information are selected randomly from subject records for a case study. Data imputation and reconciliation, RNN training, and evaluation of results are conducted individually for each subject. Hence, the RNN models are personalized, using only that person’s data. All datasets are preprocessed by the procedure elaborated in the data preprocessing section and feature variables are rescaled to have zero-mean and unit variance. Stratified six-fold cross-validation is applied to $87.5\%$ of samples of each dataset to reduce the variance of predictions. Weight values proportional to the inversion of class sizes are assigned to the corresponding samples to avoid biased predictions caused by imbalanced samples in each class. In order to better assess the performance of each model and to avoid the effects of randomization in the initialization step of back propagation algorithm, each model is trained five times with different random seeds. Hyperparameters of all models are obtained through adaptive moment estimation (Adam) optimization algorithm and $2\%$ of the training sample size was chosen as the size of the training batches. In model training with different random seeds, the number of adjustable parameters, including weights, biases, the size and number of filter kernels, and the learning rate remain constant. One difficulty associated with convolution layers in models (b) and (d) is the optimization of the hyperparameters of the convolutional layers. Usually, RNN models with convolution layers require a relatively high computation time. As a solution, learning rates with small values are preferred for networks with convolutional layers since they lead to a more optimal solution compared to large learning weights which may result in non-optimality and instability. Data preprocessing part of the work was conducted in Matlab 2019a environment and Keras/Keras-gpu 2.3.1 are used to construct and train all RNN models. Keras is a high-class API library with Tensorflow as the backend, all are available on Python environment. We have used two computational resources for data preparations and model training. Table 4 provides the details of hardware resources. ## Discussion of Results Each classifier is evaluated by testing a $12.5\%$ split of all sensor and insulin pump recordings for each subject, corresponding to 3–12 weeks of data for a subject. The average and the standard deviation of performance indexes are reported in Table 6. The lowest performance indexes was achieved by 2D ConvLSTM models. Bi-LSTM with 1D Convolution layer RNN models achieves the highest accuracy for six subjects out of eleven, and LSTM with 1D Convolution RNN for three subjects. Bi-LSTM with 1D Convolution layer RNN models outperformed other models for 4 subjects with weighted F1 scores ranging from $91.41\%$−$96.26\%$. Similarly, LSTM models with 1D Convolution layers achieved highest weighted F1 score for another 4 subjects with score values within $93.65\%$−$96.06\%$. Glycemic events for the rest of 3 subjects showed to be better predicted by regular LSTM models with weighted F1 score between $93.31\%$−$95.18\%$. This indicates that 1D Convolution improves both the accuracy and F1 scores for most of the subjects. Based on the number of adjustable parameters for the four different RNN models used for a specific subject, LSTMs are the most computational demanding blocks in the model. To assess the computational load of developing the various RNN models, we compared the number of learnable parameters (details provided in Supplementary Materials). These values can be highly informative as the number of dropouts in each model and the number of learnable parameters at each epoch (iteration) is invariant. A comparison between 1D conv-LSTM and 1D-Bi-LSTM for one randomly selected subject shows that the number of learnable parameters increase by at least $54\%$, mainly stemming from an extra embedded LSTM in Bidirectional layer (Table S1). While comparing adjustable parameters may not be the most accurate way of determining the computational loads for training the models, they provide a good reference to compare the computational burden of different RNN models. Figure 4 displays a random day selected from the test data to compare the effectiveness of each RNN model in detecting meal and exercise disturbances. Among four possible realizations for the occurrence of events, detecting joint events, Class1,1, is more challenging as it usually shows overlaps with Class0,1 and Class1,0. Another reason for lower detection is the lack of enough information on Class11, knowing that people usually rather having a small snack before and after exercise sessions over having a rescue carbohydrate during physical activity. Furthermore, the AID systems used by subjects record automatically only CGM and insulin infusion values, and meal and physical activity sessions needed to be manually entered to the device, at times an action that may be forgotten by the subject. Meal consumption and physical activity are two prominent disturbances that disrupt BGC regulation, but their opposite effect on BGC makes the prediction of Class1,1 less critical than each of meal intake or only physical activity classes. The confusion matrices of the classification results for one of the subjects (No. 2) are summarized in Table 7. As can be observed from Figure 4 and Table 7, detecting Class0,1 (physical activity) is more challenging in comparison to carbohydrate intake (Class1,0) and Class0,0 (no meal or exercise). One reason for this difficulty is the lack of biosignal information such as 3D accelerometer, blood volume pulse, and heart rate data. Some erroneous detections, such as confusing meals and exercise are dangerous since meals necessitate an insulin bolus while exercise lowers BGC and elimination of insulin infusion and/or increase in target BGC are needed. RNNs with LSTM and 1D convolution layers provide the best overall performance in minimizing such confusions: 2 meals events are classified as exercise ($0.003\%$) and 8 exercise events are classified as meals ($0.125\%$). The proportion of correctly detected exercise and meal events to all actual exercise and meal events for all subjects reveals that series of convolution-max pooling layers could elicit informative feature maps for classification efficiently. Although augmented features such as the first and second derivatives of CGM and PIC enhance the prediction power of the NN models, the secondary feature maps, extracted from all primary features, shows to be a better fit for this classification problem. Besides, repeated 1D kernel filters in convolution layers suit better the time-series nature of the data as opposed to extracting feature maps by utilizing 2D convolution filters on the data. ## Conclusions This work focuses on developing RNN models for detection and classification tasks using time series data containing missing and erroneous values. The first modeling issue arose from the quality of the recorded data in free-living. An outlier rejection algorithm was developed based on multivariable statistical analysis, and signal denoising by decomposition of the Hankel matrix of CGM recordings. A multivariate approach based on PPCA for CGM sample imputation was used to keep the harmony and relationship among the variables. The second issue addressed is the detection of events that affect the behavior of dynamic systems and the classification of these events. Four different RNN models were developed to detect meal and exercise events in the daily lives of individuals with T1D. The results indicate that models with 1D convolution layers can classify events better than regular LSTM RNN and 2D ConvLSTM RNN models, with very low confusion of the events that may cause dangerous situations by prompting erroneous interventions such as giving insulin boluses during exercise. ## References 1. Cai L, Zhu Y. **The challenges of data quality and data quality assessment in the big data era**. *Data science journal* (2015.0) **14** 2. 2.for Disease Control, C.; Prevention.; et al. National Diabetes Statistics Report, 2020; Centers for Disease Control and Prevention, US Department of Health and Human Services: Atlanta, GA, USA, 2020.. *Prevention.; et al. National Diabetes Statistics Report, 2020* (2020.0) 3. Roglic G. **WHO Global report on diabetes: A summary**. *International Journal ofNoncommunicable Diseases* (2016.0) **1** 3 4. Horowitz ME, Kaye WA, Pepper GM, Reynolds KE, Patel SR, Knudson KC, Kale GK, Gutierrez ME, Cotto LA, Horowitz BS. **An analysis of Medtronic MiniMed 670G insulin pump use in clinical practice and the impact on glycemic control, quality of life, and compliance**. *diabetes research and clinical practice* (2021.0) **177** 5. Berget C, Lange S, Messer L, Forlenza GP. **A clinical review of the t:slim X2 insulin pump**. *Expert Opinion on Drug Delivery* (2020.0) **17** 1675-1687. DOI: 10.1080/17425247.2020.1814734 6. Cobry EC, Berget C, Messer LH, Forlenza GP. **Review of the Omnipod**. *Therapeutic Delivery* (2020.0) **11** 507-519. DOI: 10.4155/tde-2020-0055 7. Sangave NA, Aungst TD, Patel DK. **Smart Connected Insulin Pens, Caps, and Attachments: A Review of the Future of Diabetes Technology**. *Diabetes Spectrum* (2019.0) **32** 378-384. DOI: 10.2337/ds18-0069 8. Hoskins M. *New diabetes technology expected* (2022.0) 9. 9.FDA approves first interchangeable biosimilar insulin product for treatment of diabetes. 10. Sevil M, Rashid M, Hajizadeh I, Askari MR, Hobbs N, Brandt R, Park M, Quinn L, Cinar A, Chappel E. *Drug Delivery Devices and Therapeutic Systems* (2021.0) 181-198 11. Boughton CK, Hovorka R. **New closed-loop insulin systems**. *Diabetologia* (2021.0) **64** 1007-1015. PMID: 33550442 12. Brown SA, Kovatchev BP, Raghinaru D, Lum JW, Buckingham BA, Kudva YC, Laffel LM, Levy CJ, Pinsker JE, Wadwa RP. **Six-Month Randomized, Multicenter Trial of Closed-Loop Control in Type 1 Diabetes**. *New England Journal of Medicine* (2019.0) **381** 1707-1717. DOI: 10.1056/NEJMoa1907863. 13. Forlenza GP, Buckingham BA, Brown SA, Bode BW, Levy CJ, Criego AB, Wadwa RP, Cobry EC, Slover RJ, Messer LH. **First Outpatient Evaluation of a Tubeless Automated Insulin Delivery System with Customizable Glucose Targets in Children and Adults with Type 1 Diabetes**. *Diabetes Technology & Therapeutics* (2021.0) **23** 410-424. DOI: 10.1089/dia.2020.0546 14. Ware J, Hovorka R. **Recent advances in closed-loop insulin delivery**. *Metabolism-Clinical and Experimental* (2022.0) **127** 15. Garcia-Tirado J, Lv D, Corbett JP, Colmegna P, Breton MD. **Advanced hybrid artificial pancreas system improves on unannounced meal response - In silico comparison to currently available system**. *Computer Methods and Programs in Biomedicine* (2021.0) **211**. DOI: 10.1016/j.cmpb.2021.106401 16. Haidar A, Legault L, Raffray M, Gouchie-Provencher N, Jacobs PG, El-Fathi A, Rutkowski J, Messier V, Rabasa-Lhoret R. **Comparison Between Closed-Loop Insulin Delivery System (the Artificial Pancreas) and Sensor-Augmented Pump Therapy: A Randomized-Controlled Crossover Trial**. *Diabetes Technology & Therapeutics* (2021.0) **23** 168-174. DOI: 10.1089/dia.2020.0365 17. Paldus B, Lee MH, Morrison D, Zaharieva DP, Jones H, Obeyesekere V, Lu J, Vogrin S, LaGerche A, McAuley SA. **First Randomized Controlled Trial of Hybrid Closed Loop Versus Multiple Daily Injections or Insulin Pump Using Self-Monitoring of Blood Glucose in Free-Living Adults with Type 1 Diabetes Undertaking Exercise**. *Journal of Diabetes Science and Technology* (2021.0) **15** 1399-1401. PMID: 34315271 18. Ekhlaspour L, Forlenza GP, Chernavvsky D, Maahs DM, Wadwa RP, Deboer MD, Messer LH, Town M, Pinnata J, Kruse G. **Closed loop control in adolescents and children during winter sports: Use of the Tandem Control-IQ AP system**. *Pediatric diabetes* (2019.0) **20** 759-768. PMID: 31099946 19. Deshpande S, Pinsker JE, Church MM, Piper M, Andre C, Massa J, Doyle FJ, Eisenberg DM, Dassau E. **Randomized Crossover Comparison of Automated Insulin Delivery Versus Conventional Therapy Using an Unlocked Smartphone with Scheduled Pasta and Rice Meal Challenges in the Outpatient Setting**. *Diabetes technology & therapeutics* (2020.0) **22** 865-874. PMID: 32319791 20. Wilson LM, Jacobs PG, Riddell MC, Zaharieva DP, Castle JR. **Opportunities and challenges in closed-loop systems in type 1 diabetes**. *The Lancet Diabetes & Endocrinology* (2022.0) **10** 6-8. PMID: 34762835 21. Franc S, Benhamou PY, Borot S, Chaillous L, Delemer B, Doron M, Guerci B, Hanaire H, Huneker E, Jeandidier N. **No more hypoglycaemia on days with physical activity and unrestricted diet when using a closed-loop system for 12 weeks: A post hoc secondary analysis of the multicentre, randomized controlled Diabeloop WP7 trial**. *Diabetes, Obesity and Metabolism* (2021.0) **23** 2170-2176 22. Jeyaventhan R, Gallen G, Choudhary P, Hussain S. **A real-world study of user characteristics, safety and efficacy of open-source closed-loop systems and Medtronic 670G**. *Diabetes, Obesity and Metabolism* (2021.0) **23** 1989-1994 23. Jennings P, Hussain S. **Do-it-yourself artificial pancreas systems: a review of the emerging evidence and insights for healthcare professionals**. *Journal of Diabetes Science and Technology* (2020.0) **14** 868-877. PMID: 31847570 24. Hajizadeh I, Askari MR, Kumar R, Zavala VM, Cinar A. **Integrating MPC with Learning-Based and Adaptive Methods to Enhance Safety, Performance and Reliability in Automated Insulin Delivery**. *IFAC-PapersOnLine* (2020.0) **53** 16149-16154 25. Hajizadeh I, Askari MR, Sevil M, Hobbs N, Brandt R, Rashid M, Cinar A, Boubaker O. *Control Theory in Biomedical Engineering* (2020.0) 63-81 26. Askari MR, Hajizadeh I, Rashid M, Hobbs N, Zavala VM, Cinar A. **Adaptive-learning model predictive control for complex physiological systems: Automated insulin delivery in diabetes**. *Annual Reviews in Control* (2020.0) **50** 1-12 27. Garcia-Tirado J, Brown SA, Laichuthai N, Colmegna P, Koravi CL, Ozaslan B, Corbett JP, Barnett CL, Pajewski M, Oliveri MC. **Anticipation of Historical Exercise Patterns by a Novel Artificial Pancreas System Reduces Hypoglycemia During and After Moderate-Intensity Physical Activity in People with Type 1 Diabetes**. *Diabetes Technology & Therapeutics* (2021.0) **23** 277-285. DOI: 10.1089/dia.2020.0516 28. Hochreiter S, Schmidhuber J. **Long short-term memory**. *Neural computation* (1997.0) **9** 1735-1780. PMID: 9377276 29. Yu Y, Si X, Hu C, Zhang J. **A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures**. *Neural Computation* (2019.0) **31** 1235-1270. DOI: 10.1162/neco_a_01199 30. Efron B. **Missing data, imputation, and the bootstrap**. *Journal of the American Statistical Association* (1994.0) **89** 463-475 31. Moritz S, Bartz-Beielstein T. *imputeTS: time series missing value imputation in R. R J* (2017.0) **9** 207 32. Ahola AJ, Mutter S, Forsblom C, Harjutsalo V, Groop PH. **Meal timing, meal frequency, and breakfast skipping in adult individuals with type 1 diabetes-associations with glycaemic control**. *Scientific reports* (2019.0) **9** 1-10. PMID: 30626917 33. Tipping ME, Bishop CM. **Probabilistic principal component analysis**. *Journal of the Royal Statistical Society: Series B (Statistical Methodology)* (1999.0) **61** 611-622 34. Ilin A, Raiko T. **Practical approaches to principal component analysis in the presence of missing values**. *The Journal of Machine Learning Research* (2010.0) **11** 1957-2000 35. Hoaglin DC, John W. *Tukey and Data Analysis. Statistical Science* (2003.0) **18** 311-318 36. Bakshi B, Stephanopoulos G. **Representation of process trends-III. Multiscale extraction of trends from process data**. *Computers & Chemical Engineering* (1994.0) **18** 267-302 37. Cheung JY, Stephanopoulos G. **Representation of process trends-Part I. A formal representation framework**. *Computers & Chemical Engineering* (1990.0) **14** 495-510 38. Samadi S, Turksoy K, Hajizadeh I, Feng J, Sevil M, Cinar A. **Meal Detection and Carbohydrate Estimation Using Continuous Glucose Sensor Data**. *IEEE Journal of Biomedical and Health Informatics* (2017.0) **21** 619-627. PMID: 28278487 39. Samadi S, Rashid M, Turksoy K, Feng J, Hajizadeh I, Hobbs N, Lazaro C, Sevil M, Littlejohn E, Cinar A. **Automatic Detection and Estimation of Unannounced Meals for Multivariable Artificial Pancreas System**. *Diabetes Technology & Therapeutics* (2018.0) **20** 235-246. PMID: 29406789 40. Eberle C, Ament C. **The Unscented Kalman Filter estimates the plasma insulin from glucose measurement**. *Biosystems* (2011.0) **103** 67-72. PMID: 20934485 41. Hajizadeh I, Rashid M, Turksoy K, Samadi S, Feng J, Frantz N, Sevil M, Cengiz E, Cinar A. **Plasma insulin estimation in people with type 1 diabetes mellitus**. *Industrial & Engineering Chemistry Research* (2017.0) **56** 9846-9857 42. Hajizadeh I, Rashid M, Samadi S, Feng J, Sevil M, Hobbs N, Lazaro C, Maloney Z, Brandt R, Yu X. **Adaptive and Personalized Plasma Insulin Concentration Estimation for Artificial Pancreas Systems**. *Journal of Diabetes Science and Technology* (2018.0) **12** 639-649. PMID: 29566547 43. Hovorka R, Canonico V, Chassin LJ, Haueter U, Massi-Benedetti M, Federici MO, Pieber TR, Schaller HC, Schaupp L, Vering T. **Nonlinear model predictive control of glucose concentration in subjects with type 1 diabetes**. *Physiological Measurement* (2004.0) **25** 905-920. PMID: 15382830 44. Ivers NM, Jiang M, Alloo J, Singer A, Ngui D, Casey CG, Catherine HY. **Diabetes Canada 2018 clinical practice guidelines: Key messages for family physicians caring for patients living with type 2 diabetes**. *Canadian Family Physician* (2019.0) **65** 14-24. PMID: 30674509 45. Care F. **Standards of Medical Care in Diabetes 2019**. *Diabetes Care* (2019.0) **42** S124-S138. PMID: 30559237 46. **Type 1 diabetes in adults: diagnosis and management**. *NICE (NG17)* (2015.0) 1-87 47. Arulampalam M, Maskell S, Gordon N, Clapp T. **A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking**. *IEEE Transactions on Signal Processing* (2002.0) **50** 174-188 48. Speekenbrink M. **A tutorial on particle filters**. *Journal of Mathematical Psychology* (2016.0) **73** 140-152 49. Pitt M, Silva R, Giordani P, Kohn R. **Auxiliary particle filtering within adaptive Metropolis-Hastings sampling**. *arXiv preprint arXiv:1006.1914* (2010.0) 50. Pudil P, Novovicova J, Kittler J. **Floating search methods in feature selection**. *Pattern Recognition Letters* (1994.0) **15** 1119-1125 51. Shi X, Chen Z, Wang H, Yeung DY, Wong WK, Woo W. **c. Convolutional LSTM network: A machine learning approach for precipitation nowcasting**. *Advances in neural information processing systems* (2015.0) **28** 802-810 52. Vinyals O, Toshev A, Bengio S, Erhan D. *Show and tell: A neural image caption generator* (2015.0) 3156-3164 53. Donahue J, Anne Hendricks L, Guadarrama S, Rohrbach M, Venugopalan S, Saenko K, Darrell T. *Long-term recurrent convolutional networks for visual recognition and description* (2015.0) 2625-2634 54. Géron A. *Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow: Concepts, tools, and techniques to build intelligent systems* (2019.0) 55. Brownlee J. *Deep Learning for Time Series Forecasting: Predict the Future with MLPs, CNNs and LSTMs in Python* (2018.0) 56. Schuster M, Paliwal K. **Bidirectional recurrent neural networks**. *IEEE Transactions on Signal Processing* (1997.0) **45** 2673-2681 57. Graves A, Schmidhuber J. **Framewise phoneme classification with bidirectional LSTM and other neural network architectures**. *Neural Networks* (2005.0) **18** 602-610. PMID: 16112549
--- title: Intestinal cell type-specific communication networks underlie homeostasis and response to Western diet authors: - Yu-Chen Wang - Yang Cao - Calvin Pan - Zhiqiang Zhou - Lili Yang - Aldons J. Lusis journal: The Journal of Experimental Medicine year: 2023 pmcid: PMC10038833 doi: 10.1084/jem.20221437 license: CC BY 4.0 --- # Intestinal cell type-specific communication networks underlie homeostasis and response to Western diet ## Abstract Intercellular communications across different layers of the intestine remain poorly understood. Here, Wang et al. show the heterogeneity and developmental trajectories of intestinal intraepithelial lymphocytes, lamina propria lymphocytes, and intestinal epithelial cells, and the cell type-specific homeostatic interactomes in response to a Western diet. The small intestine plays a key role in immunity and mediates inflammatory responses to high fat diets. We have used single-cell RNA-sequencing (scRNA-seq) and statistical modeling to examine gaps in our understanding of the dynamic properties of intestinal cells and underlying cellular mechanisms. Our scRNA-seq and flow cytometry studies of different layers of intestinal cells revealed new cell subsets and modeled developmental trajectories of intestinal intraepithelial lymphocytes, lamina propria lymphocytes, conventional dendritic cells, and enterocytes. As compared to chow-fed mice, a high-fat high-sucrose (HFHS) “Western” diet resulted in the accumulation of specific immune cell populations and marked changes to enterocytes nutrient absorption function. Utilizing ligand–receptor analysis, we profiled high-resolution intestine interaction networks across all immune cell and epithelial structural cell types in mice fed chow or HFHS diets. These results revealed novel interactions and communication hubs among intestinal cells, and their potential roles in local as well as systemic inflammation. ## Introduction The small intestine is the largest compartment of the immune system in mammals as well as the primary metabolic organ utilized for nutrient absorption (Chassaing et al., 2014; Mowat and Agace, 2014; Rooks and Garrett, 2016). Accumulating evidence indicates that intestinal immune and epithelial structural cells are involved in the chronic inflammation underlying obesity, atherosclerosis, inflammatory bowel disease, and many other disorders, particularly in individuals consuming high-fat high-carbohydrate “Western” diet (Arnone et al., 2021; Basson et al., 2021; Liu et al., 2021; Mukherjee et al., 2022; Nakanishi et al., 2021; Rescigno, 2014), and nearly one in two adults in the US is projected to be obese by 2030 (Ward et al., 2019). The gastrointestinal tract is secured by a series of protective layers that physically separate the circulatory system from the external milieu, including a mucus layer, epithelium layer, and lamina propria layer (Sundaram et al., 2022). Intestinal intraepithelial lymphocytes (IELs) are the major cell types that reside between the intestinal epithelial cells (IECs) forming the intestinal epithelium layer (Olivares-Villagomez and Van Kaer, 2018). IELs constitute a heterogeneous group of unique T lymphocytes and innate lymphoid cells (ILCs) that are conserved in all vertebrates (Riera Romo et al., 2016). The lamina propria contains a variety of types of myeloid cells and lamina propria lymphocytes (LPLs) and maintains close interactions with both immune cells in the epithelium layer and in the circulatory system (Cader and Kaser, 2013; Welty et al., 2013). The functions and interactions of the intestinal intraepithelial and lamina propria immune cells, as well as IECs, remain poorly understood in mammals, especially their roles in intestinal homeostasis. Previous studies utilizing single-cell RNA-sequencing (scRNA-seq) either provided a basic characterization of the composition of intestinal cells during development (Elmentaite et al., 2021; Elmentaite et al., 2020) or investigated only the IECs (Bomidi et al., 2021; Haber et al., 2017; Wang et al., 2020). However, these studies only identified limited immune cell populations and did not spatially separate the immune cells from different layers of the intestine. Furthermore, those IEC studies were performed alone and lacked analysis of the interactions between those structural cells and immune cells. Thus, the previous studies were unable to determine the full extent of intestinal cell heterogeneity, including bona fide cell lineages, activation states, developmental trajectories, and novel cell types. To address the current knowledge gaps, we have characterized the complete cellular composition and cell type-specific signaling networks of intraepithelial immune cells, lamina propria immune cells, and epithelial structural cells derived from the small intestines of mice maintained on chow and on high-fat high-sucrose (HFHS) diet to study the intestine as a whole. We analyzed ∼75,000 mouse immune and structural cells presented in the epithelium and lamina propria layers of the small intestine, revealing 32 previously identified cell populations and an additional 20 previously uncharacterized populations, including unique subsets of CD8αα+ IEL-T cells, CD8αβ+ IEL-T cells, CD4+ IEL-T cells, CD4+ LPL-T cells, conventional dendritic cells (cDCs), and enterocytes. We demonstrated the accumulations of distinct subsets of those populations in response to a HFHS diet and validated these by flow cytometry using independent cohorts. Utilizing the ligand–receptor analysis, we profiled the homeostatic and inflammatory circuits in the intestine, and established a signalome across intraepithelial immune cells, lamina propria immune cells, and epithelial structural cells in a cell type-specific manner. Based on these data, we identified distinct obesity-associated inflammatory and immunoregulatory pathways enriched in HFHS diet intestine. Our high-resolution, high-dimensional single-cell intestine atlas provides new insights into the functioning, development, regulatory mechanisms, and interactions of all intestinal cell types. ## Single-cell sequencing reveals a high-resolution immune landscape of intestine To investigate the unbiased cellular composition of the small intestine immune system of mouse, the single-cell suspensions of immune cells of two major intestinal compartments, the epithelium layer and lamina propria layer, were isolated from 10 C57BL/6J mice fed a chow diet and 10 mice fed a HFHS diet. Cells were analyzed by flow cytometry or FACS sorted for CD45.2+ live cell population and then profiled utilizing droplet-based 10× Genomics Chromium scRNA-seq (Fig. 1 A). The intestine single-cell atlas obtained 30,318 quality-controlled cells. Cells were clustered based on differential expression of hallmark genes and visualized using a uniform manifold approximation and projection (UMAP) plot (Fig. 1 B and Fig. S1 A). Clustering analysis revealed 17 distinct clusters: CD8αα+ TCRαβ/γδ T cell, CD8αβ+ TCRαβ/γδ T cell, CD4+ TCRαβ T cell, CD4+CD8+ TCRαβ T cell, CD4−CD8− TCRαβ/γδ T cell, cDC, plasmacytoid dendritic cell (pDC), natural killer cell (NK cell), ILC1, ILC2, ILC3, B cell, plasma cell, plasmablast cell, macrophage, eosinophil, and mast cell (Fig. 1 B). Some identified clusters contained cells from both the epithelium layer and the lamina propria layer, and they were clustered as one population in UMAP analysis although they are from completely different pools. It suggests that some IEL-T cells, like CD8αβ+ IEL-T cells and CD4+ IEL-T cells, are transcriptionally similar to their lamina propria counterparts and that they may share a common cell lineage (Fig. 1 C). The CD8αα+ T cell, CD8αβ+ T cell, and CD4−CD8− T cell populations contained both TCRαβ T cells and TCRγδ T cells, indicating high transcriptional similarities between TCRαβ T cells and TCRγδ T cells within these clusters. Although they have different TCR gene expression, they were regarded as the same type of cells in the UMAP due to high similarities in gene expression features. However, there are clear separations between T cell clusters based on their cellular expression patterns of CD8α, CD8β, and CD4 (Fig. S1 B). Furthermore, we use the term “TCRαβ/γδ cell” to indicate scRNA-seq identified cell populations that contain both TCRαβ expressing cells and TCRγδ expressing cells. Frequency analysis of intraepithelial immune cells indicated that B cells and IEL-T cells, especially CD8αα+ TCRαβ/γδ IEL-T cells, are the major cell populations of intraepithelial immune cells (Fig. 1 D). Further analysis revealed dramatically decreased proportions of CD8αα+ TCRαβ/γδ IEL-T cells and CD8αβ+ TCRαβ/γδ IEL-T cells in HFHS diet intestine. In contrast, the proportions of CD4+ TCRαβ IEL-T cells and B cells increased (Fig. 1 D). Frequency analysis of lamina propria immune cells suggested that CD4+ TCRαβ LPL-T cells are the largest LPL-T cell population, and almost all myeloid cells and ILCs were localized within the lamina propria of intestine (Fig. 1 E). Further analysis indicated that the proportions of CD8αα+ TCRαβ/γδ LPL-T cells, CD8αβ+ TCRαβ/γδ LPL-T cells, and eosinophils were reduced in HFHS diet, whereas the proportions of CD4+ TCRαβ LPL-T cells, cDCs, and ILCs (NK cell, ILC1, ILC2 and ILC3) were increased in the HFHS diet (Fig. 1 E). **Figure 1.:** *Single-cell sequencing reveals a high-resolution immune landscape of intestine. (A) Schematic of the experimental pipeline. C57BL/6J mice were fed on a chow diet or a HFHS diet for 8 wk and, following euthanasia, small intestines were isolated. Cells from two major intestinal compartments, the epithelium layer and lamina propria layer, were purified and then subjected to flow cytometry analysis or FACS sorting. Sorted CD45.2+ live cells were analyzed using droplet-based 10× Genomics Chromium scRNA-seq approach. Cells were clustered via differential gene expression for further studies. WT, wild type. (B and C) UMAP plot of intestine intraepithelial immune cells and lamina propria immune cells of 10 mice fed a chow diet and 10 mice fed a HFHS diet. Individual points correspond to single cells. (B) Annotations based on cell type analysis (Fig. S1 A). (C) Annotations indicating the sample original identity. IEL, intraepithelial lymphocytes; LPL, lamina propria lymphocytes. (D and E) Histogram showing the proportion of intraepithelial immune cells (D) and lamina propria immune cells (E) derived from 10 chow diet mice and 10 HFHS diet mice. (F and G) mIHC staining of small intestine ileum sections from chow diet mice and HFHS diet mice. (F) Sections were labeled with anti-CD3, anti-CD8α, anti-CD4, anti-Ki-67, DAPI, and images merge (100×, scale bars, 100 µm). (G) Histogram showing the cell counts of specific cell types of IEL-T cells and LPL-T cells per 1 mm of the intestine along the intestine length. (H–L) Flow cytometry analyses of major intraepithelial immune cell and lamina propria immune cell populations defined by scRNA-seq (gated as FVD−CD45.2+ cells) in an additional cohort of 16 chow diet mice and 16 HFHS diet mice (Fig. S2 D). Pie graphs indicate the proportions of major immune cell populations (H). Histograms indicate the proportions of CD8+ T cells (I), CD4+ TCRαβ T cells (J), CD4−CD8− T cells (K), and CD4+CD8+ TCRαβ T cells (L). CD4+ TCRαβ T cell, CD45.2+TCRαβ+CD8α-CD4+; CD4+CD8+ TCRαβ T cell, CD45.2+TCRαβ+CD8α+CD4+; CD8αα+ TCRαβ T cell, CD45.2+TCRαβ+CD4−CD8α+CD8β-; CD8αβ+ TCRαβ T cell, CD45.2+TCRαβ+CD4−CD8α+CD8β+; CD4−CD8− TCRαβ T cell, CD45.2+TCRαβ+CD4−CD8α-; CD8αα+ TCRγδ T cell, CD45.2+TCRγδ+CD4−CD8α+CD8β-; CD8αβ+ TCRγδ T cell, CD45.2+TCRγδ+CD4−CD8α+CD8β+; CD4−CD8− TCRγδ T cell, CD45.2+TCRγδ+CD4−CD8α-; B cell, CD45.2+CD3-NK1.1-CD19high; plasma cell, CD45.2+ CD3-NK1.1-CD19lowCD38+CD138+; Plasmablast cell, CD45.2+ CD3−NK1.1−CD19lowCD38+CD138−; NK cell/ILC1, CD45.2+CD3−CD19-NK1.1+; cDC, CD45.2+CD11blowCD11chighI-A/I-EhighPDCA-1−; pDC, CD45.2+CD11blowCD11chighI-A/I-EhighPDCA-1+; neutrophil, CD45.2+CD11bhighCD11c−/lowLy-6Ghigh; macrophage, CD45.2+CD11bhighCD11c−/lowLy-6G-F4/80+. FVD, fixable viability dye; ILC, innate lymphoid cell; pDC, plasmacytoid dendritic cell. Representative of 1 (A–E), 4 (F and G), and 4 (H–L) experiments. All data are presented as the mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by two-way ANOVA (G) or by Student’s t test (I–L). Statistics are all two-sided. Source data are available for this figure: SourceData F1.* **Figure S1.:** *Lineage-associated gene signatures of all CD45.2+ intestinal cells and intestine T cell clusters from chow diet and HFHS diet intestine. (A) Dot plot showing the top DEGs for the populations depicted in CD45.2+ intestinal cells (Fig. 1 B). Color saturation indicates the strength of average gene expression, whereas the dot size reflects the percentage of each cell cluster expressing the gene. (B) Violin plots showing the selected genes expression among all intestine T cell clusters in CD45.2+ intestinal cells (Fig. 1 B).* We next used an extensive number of mice in pathology examinations and flow cytometry to verify every observation regarding alternations in cell type proportion from scRNA-seq analysis. Hematoxylin and eosin (H&E) staining was conducted to investigate the morphological change and absolute quantification of immune cells. In response to HFHS diet, intestinal villi exhibited dysmorphology and lamina propria cavities became enlarged (Fig. S2, A and B). The intraepithelial immune cells could be distinguished by their location—between IECs and outside of the lamina propria cavity. The diet resulted in a decrease in intraepithelial immune cells and an increase in infiltration of lamina propria immune cells (Fig. S2, A and C). Multiplex-immunohistochemistry (mIHC) examination showed that IEL-T cells were dominantly CD8+ IEL-T cells (CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells), and the majority of LPL-T cells were CD4+ LPL-T cells (Fig. 1 F). In accordance with our scRNA-seq analysis, CD8+ IEL-T cells and CD8+ LPL-T cells decreased, whereas CD4+ LPL-T cells and CD4+ IEL-T cells increased in HFHS diet intestine (Fig. 1 G). **Figure S2.:** *H&E staining of small intestines and gating strategy of scRNA-seq-defined CD45.2+ intestinal cell populations and viability test in flow cytometry analysis. (A–C) H&E staining of small intestine ileum sections from 16 chow diet mice and 16 HFHS diet mice. (A) Representative images of H&E staining analysis of chow diet and HFHS diet intestine (100×, scale bars, 100 µm, upper panel; 500×, scale bars, 20 µm, lower panel). The yellow shadows in the upper panel indicate the lamina propria area. The white arrows indicate the intraepithelial immune cells in the magnified plots. (B) Histogram showing area calculation of average lamina propria area per intestine villus. (C) Histogram showing the cell counts of intraepithelial immune cells and lamina propria immune cells per 1 mm of the intestine along the intestine length. (D) Representative flow cytometry analysis gating strategy of scRNA-seq-defined intraepithelial immune cell and lamina propria immune cell populations (gated as FVD−CD45.2+ cells) in the intestine (Fig. 1, H–L). CD4+ TCRαβ T cell, CD45.2+TCRαβ+CD8α−CD4+; CD4+CD8+ TCRαβ T cell, CD45.2+TCRαβ+CD8α+CD4+; CD8αα+ TCRαβ T cell, CD45.2+TCRαβ+CD4−CD8α+CD8β−; CD8αβ+ TCRαβ T cell, CD45.2+TCRαβ+CD4−CD8α+CD8β+; CD4−CD8− TCRαβ T cell, CD45.2+TCRαβ+CD4−CD8α−; CD8αα+ TCRγδ T cell, CD45.2+TCRγδ+CD4−CD8α+CD8β−; CD8αβ+ TCRγδ T cell, CD45.2+TCRγδ+CD4−CD8α+CD8β+; CD4−CD8− TCRγδ T cell, CD45.2+TCRγδ+CD4−CD8α-; B cell, CD45.2+CD3−NK1.1−CD19high; plasma cell, CD45.2+ CD3−NK1.1−CD19lowCD38+CD138+; Plasmablast cell, CD45.2+ CD3−NK1.1−CD19lowCD38+CD138−; NK cell/ILC1, CD45.2+CD3−CD19−NK1.1+; cDC, CD45.2+CD11blowCD11chighI-A/I-EhighPDCA-1−; pDC, CD45.2+CD11blowCD11chighI-A/I-EhighPDCA-1+; neutrophil, CD45.2+CD11bhighCD11c-/lowLy-6Ghigh; macrophage, CD45.2+CD11bhighCD11c−/lowLy-6G-F4/80+. FVD, fixable viability dye; NK cell, natural killer cell; ILC, innate lymphoid cell; cDC, conventional dendritic cell; pDC, plasmacytoid dendritic cell. (E–H) C57BL/6J mice were fed on a chow diet or a HFHS diet for 8 wk and, following euthanasia, small intestines were isolated. Intraepithelial and lamina propria immune cells were purified and then subjected to flow cytometry analysis. To study cell viability, cells were stained with FVD followed by Fc blocking and surface marker staining. (E and F) Flow cytometry analyses of CD8αα+ IEL-T cells (E) and CD8αβ+ IEL-T cells (F) in chow diet and HFHS diet fed intestine (n = 8). (G and H) Flow cytometry analyses of CD8αα+ LPL-T cells (G) and CD8αβ+ LPL-T cells (H) in chow diet and HFHS diet fed intestine (n = 4). CD8αα+ T cells, CD45.2+CD3+CD4−CD8α+CD8β− cells; CD8αβ+ T cells, CD45.2+CD3+CD4−CD8α+CD8β+ cells. FVD, fixable viability dye. Representative of 5 (A–C) and 4 (D–H) experiments. All data are presented as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test (B, C, and E–H). Statistics are all two-sided. Source data are available for this figure: SourceData FS2.* To further interrogate the changes and dissect the population of TCRαβ T cells and TCRγδ T cells, an independent cohort of 20 chow diet-fed mice and 20 HFHS diet-fed mice was analyzed by flow cytometry for major immune cell populations (Fig. 1 H and Fig. S2 D). The undefined cell populations here might include monocytes, basophils, mast cells, ILC2s, ILC3s, and pre-cDCs (Sun et al., 2015; Varol et al., 2010). The decreased CD8+ IEL-T cells in the HFHS diet intestine resulted primarily from decreases in the CD8αα+ TCRγδ T cells, CD8αβ+ TCRγδ T cells and CD8αα+ TCRαβ T cells, whereas the decreased CD8+ LPL-T cells resulted primarily from the reduction of CD8αα+ TCRγδ T cells, CD8αβ+ TCRγδ T cells, and CD8αβ+ TCRαβ T cells (Fig. 1 I). Flow cytometry further supported our scRNA-seq analysis, demonstrating the increase of CD4+ IEL-T cells and CD4+ LPL-T cells in HFHS diet intestine (Fig. 1 J). No significant changes were observed in CD4−CD8− TCRαβ/γδ T cells and CD4+CD8+ TCRαβ T cells in both the epithelium layer and lamina propria (Fig. 1, K and L). To determine if the decrease of CD8+ T cells upon feeding the HFHS diet is due to cell death, we used fixable viability dye (FVD) to measure the viability of CD8+ IEL-T cells and CD8+ LPL-T cells in the intestine upon HFHS diet feeding and chow diet feeding. The results show that the viability of CD8αα+ IEL-T and CD8αβ+ IEL-T cells significantly decreased after HFHS diet feeding (Fig. S2, E and F). Furthermore, the viability of CD8αα+ LPL-T and CD8αβ+ LPL-T cells also decreased after HFHS diet feeding (Fig. S2, G and H). These results indicate that the HFHS diet does induce cell death in the intraepithelial immune cells, including both CD8αα+ T cells and CD8αβ+ T cells in the intestine. These results indicated that a HFHS diet is associated with dramatic changes in intestine CD8+ T cells, CD4+ T cells, and cDCs in both the epithelium layer and the lamina propria. These cells likely contain heterogeneous populations or new cell subsets that have not previously been characterized, suggesting the necessity of further subclustering. ## Unique CD8+ IEL-T cell subsets accumulate in HFHS diet intestine Previous studies from mice and humans suggest that CD8αα+ IEL-T cells, including CD8αα+ TCRγδ IEL-T cells and CD8αα+ TCRαβ IEL-T cells, are all natural IELs that possess the potential for self-renewal within the intestine epithelium layer (McDonald et al., 2018; Ruscher et al., 2017), whereas CD8αβ+ IEL-T cells include conventional CD8αβ+ TCRαβ T cells and a novel CD8αβ+ TCRγδ T cell population (Cheroutre et al., 2011; Kadivar et al., 2016; Olivares-Villagomez and Van Kaer, 2018). To further examine whether cellular heterogeneity existed in CD8+ IEL-T cells, we assessed a total of 11,677 pooled cells from all CD8+ IEL-T cells of 10 chow diet mice and 10 HFHS diet mice and UMAP visualization revealed five distinct clusters (Fig. 2 A). Besides the identification of CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells based on the distribution of Cd8a and Cd8b1 gene expression, we found that each of them contained a population expressing high levels of cytotoxicity genes such as Gzma and Gzmb and a separate population expresses high levels of memory T cell signature genes such as Tcf7 and Id3. In addition, a third population was observed in CD8αα+ IEL-T cells which expressed high levels of stem cell-related genes such as Stmn1 and Mki67 (Fig. 2 B). Based on the top variable expressed genes in each subcluster, our analysis identified three populations of CD8αα+ IEL-T cells (effector-like CD8αα+ IEL-T cells, memory-like CD8αα+ IEL-T cells, and proliferating CD8αα+ IEL-T cells) and two populations of CD8αβ+ IEL-T cells (effector-like CD8αβ+ IEL-T cells and memory-like CD8αβ+ IEL-T cells; Fig. 2 C). Assessment of the frequency of each cell cluster indicated that effector-like IEL-T cells are the major subpopulation of both CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells, and revealed a shift in frequency from effector-like IEL-T cells in chow diet intestines to memory-like IEL-T cell in HFHS diet intestines (Fig. 2 D). To further confirm the presence of scRNA-seq-identified cell lineages of CD8αα+ IEL-T cells, we harvested intraepithelial immune cells from an additional cohort of mice (16 chow diet, 16 HFHS diet). Flow cytometry experiments corroborated the results from our scRNA-seq clustering, identifying three distinct CD8αα+ IEL-T cell populations and two CD8αβ+ IEL-T cell populations based on endogenous Ki67 and Granzyme B expression (Fig. 2, E and F; and Fig. S3 A). Furthermore, in support of our scRNA-seq results, they demonstrated that CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells decreased their effector-like cell proportions and increased memory-like cell proportions in HFHS diet intestine (Fig. 2, E and F). **Figure 2.:** *Unique CD8+ IEL-T cell subsets accumulate in HFHS diet intestine. (A and B) UMAP plot of mouse intestine CD8+ IEL-T cells derived from a cohort of 10 chow diet mice and 10 HFHS diet mice. Individual points correspond to single cells. (A) Cluster analysis yields five distinct clusters comprising CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells. (B) Feature plots showing the selected signature genes projection on UMAP. Intensity of gene expression in each cell was indicated in color saturation. (C) Dot plot showing selected top DEGs for the populations depicted in CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells. Color saturation indicates the strength of average gene expression, whereas the dot size reflects the percentage of each cell cluster expressing the gene. (D) Histogram showing the proportions of CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells derived from 10 chow diet mice and 10 HFHS diet mice. (E and F) Flow cytometry analyses of scRNA-seq-defined CD8αα+ IEL-T cell (gated as CD45.2+CD3+CD4−CD8α+CD8β− cells; E) and CD8αβ+ IEL-T cell (gated as CD45.2+CD3+CD4−CD8α+CD8β+ cells; F) populations in an additional cohort of 16 chow diet mice and 16 HFHS diet mice (Fig. S3 A). Effector-like CD8αα+ IEL-T cells, Granzyme B+Ki67−; memory-like CD8αα+ IEL-T cells, Granzyme B−Ki67−; proliferating CD8αα+ IEL-T cells, Ki67+; effector-like CD8αβ+ IEL-T cells, Granzyme B+; memory-like CD8αβ+ IEL-T cells, Granzyme B−. (G) Visualizing the predicted differentiation using a CytoTRACE 3D plot of CD8αα+ IEL-T cells showing the CytoTRACE score component with UMAP coordinates. The color indicates the CytoTRACE score from 1 (red, lowest levels of differentiation) to 0 (blue, highest levels of differentiation). Feature plots showing the selected signature genes projection on 3D plot. Intensity of gene expression in each cell was indicated in color saturation from low (red) to high (blue). (H) Monocle analysis of the CD8αα+ IEL-T cells. The color indicates pseudotime directionality projection on UMAP from the earliest (blue) to the latest (yellow). (I) RNA-velocity analysis of CD8αα+ IEL-T cell subclusters with velocity field projected onto the UMAP plot. Arrows show the local average velocity evaluated on a regular grid and indicate the extrapolated future states of cells. (J) Cell cycle analysis of the CD8αα+ IEL-T cells. Predicted classification of each cell in either G2/M (red), S (blue), or G1 (green) phase was projected on UMAP. (K and L) IPA analysis of upstream transcriptional regulators of the transition from proliferating CD8αα+ IEL-T cells to either effector-like CD8αα+ IEL-T cells (K) or memory-like CD8αα+ IEL-T cells based on DEGs (L). Representative of 1 (A–D and G–L), and 4 (E and F) experiments. All data are presented as the mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by two-way ANOVA (E and F). Statistics are all two-sided.* **Figure S3.:** *Unique T cell subsets accumulate in HFHS diet intestine. (A) Representative flow cytometry analysis gating strategy of scRNA-seq-defined CD8αα+ IEL-T cells (gated as FVD−CD45.2+CD3+CD4−CD8α+CD8β− cells) and CD8αβ+ IEL-T cells (gated as FVD−CD45.2+CD3+CD4−CD8α+CD8β+ cells) for intracellular staining of cytotoxicity molecules and transcription factors (Fig. 2, E and F). FVD, fixable viability dye. (B) Bifurcation heatmap of enriched genes for effector-like CD8αα+ IEL-T cell (left), proliferating CD8αα+ IEL-T cell (middle) and memory-like CD8αα+ IEL-T cell (right). Color indicates increased (red) or decreased (blue) gene expression. (C) IPA of secreted upstream regulators of the transition from proliferating CD8αα+ IEL-T cells to either effector-like CD8αα+ IEL-T cells or memory-like CD8αα+ IEL-T cells. (D) Histogram of flow cytometry analyses of scRNA-seq-defined effector-like and memory-like CD8αα+ IEL-T cells in mice fed with a HFHS diet (n = 4). C57BL/6J mice were fed on a chow diet or a HFHS diet for 8 wk and, following euthanasia, small intestines were isolated. Intraepithelial immune cells were purified and then subjected to flow cytometry analysis. To study cell viability, cells were stained with FVD followed by Fc blocking and surface marker staining. CD8αα+ IEL-T cells, CD45.2+CD3+CD4−CD8α+CD8β− cells; effector-like CD8αα+ IEL-T cells, Granzyme B+Ki67-; memory-like CD8αα+ IEL-T cells, Granzyme B-Ki67−; FVD, fixable viability dye. (E) Representative flow cytometry analysis gating strategy of scRNA-seq-defined CD4+ IEL-T cells and CD4+ LPL-T cells (gated as FVD−CD45.2+CD3+CD8α−CD4+ cells) for intracellular cytokine staining (Fig. 3, E and F). FVD, fixable viability dye. (F) Whole intestinal tissue from ileum to jejunum were collected and analyzed using RT-QPCR. Histogram showing the mRNA expression levels of Il6, Il17a, and Il17f. N = 8. (G) Flow cytometry analyses of the absolute cell count of Th17 LPL-T cells (gated as CD45.2+CD3+CD8α−CD4+IL17A+) from lamina propria immune cells. n = 8. (H) IPA of putative secreted upstream regulators of the transition from memory-like CD4+ T cells to Th1, Th2, Th17, or Treg cells. Representative of 2 (D, F, and G) and 4 (A and E) experiments. All data are presented as the mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test (D). Statistics are all two-sided.* UMAP visualization of CD8αα+ IEL-T cells showed that the identified proliferating CD8αα+ IEL-T cells are transcriptionally close to effector-like CD8αα+ IEL-T cells and memory-like CD8αα+ IEL-T cells, suggesting that proliferating CD8αα+ IEL-T cells represent a committed precursor to both populations (Fig. 2 A). To test this hypothesis, we utilized CytoTRACE analysis, a trajectory reconstruction analysis using gene counts and expression to predict differentiation states in scRNA-seq datasets (Gulati et al., 2020). The 3-dimension (3D) plot suggested that proliferating CD8αα+ IEL-T cells represented the least-differentiated cells and that progressions toward either the effector-like CD8αα+ IEL-T cell or memory-like CD8αα+ IEL-T cell status increased the levels of differentiation (Fig. 2 G). We also utilized Monocle to perform single-cell trajectory analysis of how cells choose between one of several possible end states, placing them along a trajectory corresponding to a biological process (Trapnell et al., 2014). Monocle pseudotime analysis further supported the bifurcating developmental trajectories from proliferating CD8αα+ IEL-T cells to memory-like CD8αα+ IEL-T cell and effector-like CD8αα+ IEL-T cell fates (Fig. 2 H). Furthermore, RNA-velocity analysis was utilized to determine the transcriptional fates of the proliferating CD8αα+ IEL-T cell population. RNA velocity uses the time derivative of the gene expression state, which can be directly estimated by distinguishing between unspliced and spliced mRNAs in common scRNA-seq datasets to predict the future state of individual cells on a timescale (La Manno et al., 2018). Projection of the velocity field arrows onto the UMAP plot extrapolated future states of proliferating CD8αα+ IEL-T cells to both effector-like CD8αα+ IEL-T cell and memory-like CD8αα+ IEL-T cell populations (Fig. 2 I). Cell cycle analysis (Nestorowa et al., 2016) predicted classification of each cell in either the G2/M, S, or G1 phase. Proliferating CD8αα+ IEL-T cells exhibited a highly proliferative status with all cells in either G2/M or S phase, whereas effector-like CD8αα+ IEL-T cells and memory-like CD8αα+ IEL-T cells were mainly in the G1 phase, consistent with the predicted transition (Fig. 2 J). Manually averaged principal curves assigned on CytoTRACE analysis and RNA-velocity, and branch analysis on Monocle trajectory suggest clear bifurcating developmental trajectories from proliferating CD8αα+ IEL-T cells to effector-like CD8αα+ IEL-T cells and memory-like CD8αα+ IEL-T cells (Fig. 2, G–I). Analysis of differentially expressed genes (DEGs) between the effector-like CD8αα+ IEL-T cell and memory-like CD8αα+ IEL-T cell fates from proliferating CD8αα+ IEL-T cells showed a clear bifurcation in gene expression patterns (Fig. S3 B). Ingenuity pathway analysis (IPA) implicated dramatic differences in upstream signaling regulators between the transition to the effector-like CD8αα+ IEL-T cell fate and the transition to the memory-like CD8αα+ IEL-T cell fate, further suggesting a developmental switch to the memory-like CD8αα+ IEL-T cell fate in response to the HFHS diet (Fig. 2, K and L; and Fig. S3 C). To investigate whether CD8αα+ memory-like IEL-T cells and CD8αα+ effector-like IEL-T cells differ in sensitivity to cell death by the HFHS diet, we measured their viability through intracellular staining. The result indicated that there are no significant differences in cell viability between CD8αα+ memory-like IEL-T cells and CD8αα+ effector-like IEL-T cells (Fig. S3 D). This result indicates that development rather than sensitivity might be the major reason for the shift in frequency. These results suggest a previously unidentified shared precursor exists in the CD8αα+ IEL-T cells, and that the accumulation of memory-like CD8αα+ IEL-T cells during HFHS feeding may be due to increased differentiation from proliferating CD8αα+ IEL-T cells. ## Distinct CD4+ T cell subsets accumulate in HFHS diet intestine Previous studies from mice and humans indicate that both CD4+ TCRαβ IEL-T cells and CD4+ TCRαβ LPL-T cells are conventional T cell populations, and characterization of their subsets has not previously been described. To examine whether further cellular heterogeneity existed, we pooled all CD4+ IEL-T cells and CD4+ LPL-T cells from the intestines of 10 HFHS diet mice and 10 chow diet mice and UMAP visualization revealed five distinct populations (Fig. 3 A). These included four classic CD4+ T cell populations: T helper 1 (Th1) cells by expression of Ifng, Il12rb2, and Ccl5; Th2 cells by expression of Il4; Th17 cells by expression of Il22, Il17a, and Il17f; and regulatory T cells (Tregs) by expression of Foxp3 and Il10. In addition, we identified a memory-like CD4+ T cell population with high expression of memory T cell signature genes Tcf7, Id3, and Izumo1r (Fig. 3, B and C). Frequency analysis indicated that memory-like CD4+ IEL-T cells are the dominant CD4+ IEL-T cell subpopulation, and that CD4+ LPL-T cells have higher Th1 cell and Treg cell proportions and lower memory-like T cell proportion compared with CD4+ IEL-T cells (Fig. 3 D). It also suggested a dramatic increase in memory-like CD4+ T cells and Treg cells within epithelium layer and lamina propria of HFHS diet intestine. In contrast, the frequencies of Th1 cells decreased (Fig. 3 D). To further validate the existence and proportional change of identified Th1 cell and Treg cell populations, we utilized both intraepithelial and lamina propria immune cells from an additional cohort of mice (16 chow diet, 16 HFHS diet), and profiled intracellular IFNγ and Foxp3 expression. Flow cytometry analysis further supported the results from scRNA-seq, demonstrating that CD4+ IEL-T cell and CD4+ LPL-T cell decreased their Th1 cell proportions and increased Treg cell proportions in HFHS diet intestine (Fig. 3, E and F; and Fig. S3 E). Due to the huge level of total CD4+ T cell infiltration in the lamina propria upon HFHS diet feeding, we profiled the small intestines from chow diet fed mice and HFHS diet fed mice using RT-QPCR. Expression levels of Ifng indicate that there is a slight increase in Th1 responses at the level of the whole intestinal tissue (Fig. S3 F). Expression of Il17a and Il17f both dramatically increased in the HFHS diet fed mice (Fig. S3 F). Moreover, we used flow cytometry to further verify the changes in Th1 and Th17 responses; the absolute cell counting indicates that there is a slight increase in LPL-Th1 cells and a significant increase in LPL-Th17 cells in the intestine upon HFHS diet feeding (Fig. S3 G). **Figure 3.:** *Distinct CD4+ T cell subsets accumulate in HFHS diet intestine. (A and B) UMAP plot of mouse intestine CD4+ T cells derived from a cohort of 10 chow diet mice and 10 HFHS diet mice, where individual points correspond to single cells. (A) Cluster analysis yields five distinct clusters from CD4+ IEL-T cells and CD4+ LPL-T cells. (B) Feature plots showing the selected signature genes projection on UMAP. Intensity of gene expression in each cell was indicated in color saturation. Th, T helper; Treg cell, regulatory T cell. (C) Dot plot showing selected top DEGs for the populations depicted in CD4+ T cells. Color saturation indicates the strength of average gene expression, whereas the dot size reflects the percentage of each cell cluster expressing the gene. (D) Histogram showing the proportions of CD4+ IEL-T cells and CD4+ LPL-T cells derived from 10 chow diet mice and 10 HFHS diet mice. (E and F) Flow cytometry analyses of scRNA-seq-defined Treg cells (E) and Th1 cells (F) in CD4+ IEL-T cell and CD4+ LPL-T cell populations (gated as CD45.2+CD3+CD8α−CD4+ cells) in an additional cohort of 16 chow diet mice and 16 HFHS diet mice (Fig. S3 E). Treg cells, Foxp3+; Th1 cells, IFNγ+. (G) Visualizing the predicted differentiation using a CytoTRACE 3D plot of CD4+ T cells showing the CytoTRACE score component with UMAP coordinates. The color indicates the levels of CytoTRACE score from 1 (red, lowest levels of differentiation) to 0 (blue, highest levels of differentiation). Feature plots showing the selected signature genes projection on 3D plot. Intensity of gene expression in each cell was indicated in color saturation from low (red) to high (blue). (H) RNA-velocity analysis of CD4+ T cell subclusters with velocity field projected onto the UMAP plot. Arrows show the local average velocity evaluated on a regular grid and indicate the extrapolated future states of cells. (I) Monocle analysis of the CD4+ T cells. The color indicates pseudotime directionality projection on UMAP from the earliest (blue) to the latest (yellow). (J–M) IPA analysis of upstream transcriptional regulators of the transition from memory-like CD4+ T cells to Th1 cells (J), Th2 cells (K), Th17 cells (L), or Treg cells (M) based on DEGs. Representative of 1 (A–D and G–M), and 4 (E and F) experiments. All data are presented as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test (E and F). Statistics are all two-sided.* Although the cellular plasticity of CD4+ T cells in the intestine has been noted (Brucklacher-Waldert et al., 2014), the characterization and ontogeny of those subpopulations are not well understood. UMAP visualization of CD4+ T cell subsets suggested that the memory-like CD4+ T cell population is most transcriptionally related to both Treg and the Th1/Th2/Th17 clusters, whereas Th1, Th2, and Th17 clusters are close to each other (Fig. 3 A). CytoTRACE analysis indicated that memory-like T cell is the least differentiated cell status of CD4+ T cell in the intestine with the highest score, whereas Th1, Th2, Th17, and Treg cells represented highly differentiated cell status with mid to low scores, by which they form four independent columns paralleled with each other in the 3D plot, all related to the memory-like T cells at the top (Fig. 3 G). Projection of the velocity field arrows showed a strong directional flow of the memory-like T cells to both Treg fates and Th1/Th2/Th17 fates (Fig. 3 H). Monocle pseudotime analysis further corroborated the transitions from memory-like CD4+ cells to the Treg cells and Th1/Th2/Th17 cells (Fig. 3 I). Those results suggested that mouse intestine Treg and T helper cell populations might be continuously replenished from the memory-like T cell populations. We then assessed the potential upstream regulators of the DEGs between memory-like CD4+ T cells and Th1, Th2, Th17, as well as Treg cells using IPA. The results suggested that some signaling pathways such as BHLHE40 and CTNNB1 (Wnt/β-catenin) are significant mediators in all the transitions from CD4+ memory-like T cells to Th1, Th2, Th17 cells and to Treg cells, and that some pathways such as RelA/p65 and TGFβ/KLF6 are involved in Th17 and Treg cell transitions (Fig. 3, J–M). Signature regulators were also observed in the transitions from CD4+ memory-like T cells to Th1 (TBX21, STAT4, STAT1), Th2 (STAT5A, STAT5B), Th17 (ID2, EP300, GLI3), and Treg (SMAD3, SMAD4) cells (Fig. 3, J–M; and Fig. S3 H). Thus, these results suggested that the HFHS diet microenvironment-induced shifts in intestine CD4+ T cell populations occurred both within epithelium layer and lamina propria, and that this might be influenced by the transition from CD4+ memory-like T cells toward Th1, Th2, Th17, and Treg cells. ## Distinct myeloid cell subsets accumulate in HFHS diet intestine Classic/conventional dendritic cells (cDCs) are required for peripheral Treg induction in the intestine, mediate tolerance to food antigens, limit reactivity to the gut microbiota and optimize responses to intestinal pathogens (Cabeza-Cabrerizo et al., 2021; Esterházy et al., 2016; Worbs et al., 2017). Previous results indicated increased proportions of Treg cells in CD4+ IEL-T cells and CD4+ LPL-T cells, and increased proportions of memory-like T cells in both CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells in response to the HFHS diet (Fig. 2, D and E; and Fig. 3, D and E). Notably, memory-like T cells are characterized by Xcl1 signature gene expression (Fig. 2 C), and Xcl1 is the most differentially expressed gene in both CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells when two diets were compared (Fig. S4, A and B; and Tables S1 and S2). The receptor for XCL1, XCR1, is known to be exclusively expressed by cDC1 and conserved across species (Cabeza-Cabrerizo et al., 2021; Worbs et al., 2017). To examine whether further cellular heterogeneity existed in cDCs, an additional cohort of 26,848 lamina propria immune cells from 10 HFHS diet mice and 10 chow diet mice were profiled by scRNA-seq. Assessment of pooled cDCs from all samples revealed three distinct populations (Fig. 4 A). cDC1 is discerned by expression of Xcr1, Itgae (CD103), and Clec9a. cDC2B is distinguished by expression of Itgam (CD11b), Sirpa (SIRPα), Cd14, Cx3cr1, Tmem176a, Tmem176b, Clec10a, Clec12a, P2rx7, and lack of Itgae. cDC2A is identified via Itgam, Sirpa, Itgae, and lack of expression of cDC2a markers such as Clec10a and P2rx7 (Fig. 4, B and C). Frequency assessments indicated a decrease in cDC2B proportion in HFHS diet intestine, whereas the frequencies of cDC2A and cDC1 both increased (Fig. 4 D). An additional cohort of mice (12 chow diet, 12 HFHS diet) was profiled by flow cytometry to further confirm the existence and proportional change of identified cDCs (Fig. S2 D). This corroborated the results from the scRNA-seq and identified three distinct cDC populations based on the expression of SIRPα, XCR1, and CD103 (Fig. 4 E). Flow cytometry also demonstrated the increase of cDC2A and cDC1 proportions and decrease of cDC2B proportions in HFHS diet intestine, consistent with scRNA-seq analysis (Fig. 4 F). **Figure S4.:** *Analysis of the macrophage populations and the comparison of key genes related to intestine barrier function and nutrient absorption function between chow diet and HFHS diet intestine. (A) Heatmap showing the top six most upregulated and most downregulated DEGs in CD8αα+ IEL-T cells between chow diet intestine and HFHS diet intestine. (B) Heatmap showing the top six most upregulated and most downregulated DEGs in CD8αβ+ IEL-T cells between chow diet intestine and HFHS diet intestine. (C–G) An independent cohort consisting of 10 chow diet mice and 10 HFHS diet mice were added to the previous cohort (Fig. 1 A) to accumulate more macrophages. (C) UMAP visualization of mouse intestine macrophages derived from lamina propria of 20 chow diet mice and 20 HFHS diet mice is shown, where individual points correspond to single cells. Cluster analysis yields three distinct clusters: Ly6ChiCX3CR1lo monocyte-derived macrophage (blue), CX3CR1hi inflammatory macrophage (red), and CX3CR1hi regulatory macrophage (blue). (D) Feature plots showing the selected signature genes projection on UMAP. Intensity of gene expression in each cell was indicated in color saturation. (E) Histogram showing the proportions of macrophages derived from 20 chow diet mice and 20 HFHS diet mice. Inf. macrophage, CX3CR1hi inflammatory macrophage; Reg. macrophage, CX3CR1hi regulatory macrophage; Mono. Macrophage, Ly6ChiCX3CR1lo monocyte-derived macrophage. (F and G) KEGG pathway gene set enrichment analysis of DEGs between chow diet and HFHS diet mice intestine cells. The top seven most upregulated signaling pathways in response to the HFHS diet were shown. (F) CX3CR1hi inflammatory macrophages. (G) CX3CR1hi regulatory macrophages. (H) Heatmaps of selected genes that are involved in microbiota–host interactions, tight junction, and mucin functions. (I) Violin plots showing the average signature scores calculated for microbiota–host interactions, tight junction, and mucin functions. Each dot represents one cell. (J) Violin plots showing the average signature scores calculated for the absorption of distinct nutrient classes (fat and cholesterol, amino acid, peptide, and carbohydrate). Each dot represents one cell.* **Figure 4.:** *Distinct dendritic cell subsets accumulate in HFHS diet intestine. (A and B) An independent cohort consisting of 10 chow diet mice and 10 HFHS diet mice were added to the previous cohort (Fig. 1 A) to accumulate more cDCs. (A) UMAP visualization of mouse intestine cDCs derived from lamina propria of 20 chow diet mice and 20 HFHS diet mice is shown, where individual points correspond to single cells. Cluster analysis yields three distinct clusters: cDC2B (red), cDC2A (green), and cDC1 (blue). (B) Violin plots showing expression of the selected signature genes in cDC populations. Each dot represents one cell. (C) Dot plot showing selected top DEGs for the populations depicted in cDCs. Color saturation indicates the strength of average gene expression, whereas the dot size reflects the percentage of each cell cluster expressing the gene. (D) Histogram showing the proportions of cDCs derived from 20 chow diet mice and 20 HFHS diet mice. (E and F) Flow cytometry analyses of scRNA-seq-defined mouse cDC populations (gated as CD45.2+CD11blowCD11chighI-A/I-EhighPDCA− cells) in an additional cohort of 12 chow diet mice and 12 HFHS diet mice (E; Fig. S1 A). Histogram showing the proportions of cDC subpopulations (F). cDC1s, XCR1+SIRPα−; cDC2As, SIRPα+CD103+; cDC2Bs, SIRPα+CD103−. (G and H) IPA analysis of transcriptional upstream regulators (G) and secreted upstream regulators (H) of the transition of cDCs from chow diet to HFHS diet based on DEGs. Representative of 1 (A–D, G, and H), and 5 (E and F) experiments. All data are presented as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test (F). Statistics are all two-sided.* Next, we sought to investigate the potential molecular mechanisms regulating the intestine cDC alteration. We assessed the upstream regulators of the DEGs between all chow diet cDCs and HFHS diet cDCs utilizing IPA. The results indicated that upstream transcription factors CITED2 and ZFP36 are at the top of the list (Fig. 4 G). They are involved in TGFβ signaling pathway activation and NF-κB inhibition, and are known for restraining pathogenic proinflammatory gene programs of myeloid cells and promoting the immunoregulatory functions (Makita et al., 2021; Pong Ng et al., 2020). Further analysis of the secreted upstream regulators showed that cytokines and growth factors, such as IL-37, IL1RN, and DKK1, play a critical role in cDC transformation in response to HFHS diet feeding (Fig. 4 H), presumably by negatively regulating the proinflammatory activation of cDCs and facilitating Treg induction (Chu et al., 2021; Dinarello et al., 2016; Li et al., 2015). These data indicate a potential immunoregulatory transformation of cDCs in the intestine upon HFHS diet feeding and identify enhanced signaling pathways such as IL-37, TGFβ (CITED2), and inhibited pathways such as NF-κB (ZFP36), IL-1β (IL1RN), and Wnt (DKK). Collectively, these signaling pathways are likely to play critical roles in the cDC transformation and subset regulation. Furthermore, we also examined whether further cellular heterogeneity existed in macrophages. We assessed pooled macrophages from all lamina propria samples and UMAP visualization revealed three distinct populations (Fig. S4 C). Besides the identification of Ly6C high (Ly6Chi) CX3CR1 low (CX3CR1lo) monocyte-derived macrophages, we found two major heterogeneous populations in the macrophages with high CX3CR1 expression: CX3CR1hi inflammatory macrophages are discerned by expression of Cd86, Cd80, Ccl24, and Cxcl9; CX3CR1hi regulatory macrophages are distinguished by expression of Mrc1 (CD206) and Cd163 (Fig. S4 D). Frequency analysis suggested a slight increase in the percentage of CX3CR1hi inflammatory macrophages and a decrease in the proportion of CX3CR1hi regulatory macrophages in the HFHS diet intestine, while Ly6ChiCX3CR1lo monocyte-derived macrophages remain unchanged (Fig. S4 E). We compared the differentially expressed genes for the populations depicted in the macrophages. The results indicate that CX3CR1hi inflammatory macrophages express high levels of co-stimulatory ligands CD86 and CD80, and chemokine CXCL9, which are critical for T cells activation and recruitment (Table S3). CX3CR1hi regulatory macrophages express high levels of immunoregulatory genes, such as Mrc1 and Clec12a, and also produce growth factors such as IGF-binding proteins (Igfbp4) and growth and differentiation factor 15 (Gdf15), which could further regulate the whole-body metabolism through the circulation (Table S3). Ly6ChiCX3CR1lo monocyte-derived macrophages express high levels of Ly6c2 and activation marker Cd69, indicating their early activation and maturation status (Table S3). Furthermore, we also performed the KEGG pathway enrichment analysis of two major macrophage subpopulations. Analysis of CX3CR1hi inflammatory macrophages in HFHS diet intestine showed a strong upregulation in chemokine, Toll-like receptor and PI3K-Akt signaling pathways, indicating their enhanced proinflammatory status (Fig. S4 F). Analysis of CX3CR1hi regulatory macrophages also showed increased KEGG pathways involved in diseases such as atherosclerosis, diabetic complications, and NAFLD (Fig. S4 G). These results suggest that both CX3CR1hi inflammatory macrophages and CX3CR1hi regulatory macrophages play important roles in the activation and regulation of immune responses in the intestine upon HFHS diet feeding, and CX3CR1hi regulatory macrophages might also be involved in the modulation of systemic metabolic disorders when fed with a HFHS diet. ## Diversity in structural cells and altered nutrient absorption uncovered The intestinal epithelium is composed of a single layer of IECs. Although IECs are non-hematopoietic structural cells, they are an integral component of intestinal immunity and play important roles in the absorption of nutrients, protection of mucosal barrier functions, and maintenance of intestinal homeostasis (Allaire et al., 2018; Peterson and Artis, 2014). However, how IECs are modulated by the intestine microenvironment in response to a HFHS diet is poorly understood. We sorted non-hematopoietic CD45.2− live cells from epithelium layer of a new cohort of 10 chow diet mice and 10 HFHS diet mice, and then profiled them using scRNA-seq (Fig. 5 A). The resulting quality-controlled, mouse intestinal structural cell atlas included 13,664 cells, which were clustered based on differential expression of hallmark genes and visualized using UMAP: enterocytes by expression of Alpi, Apoa1, Apob, Apoa4, Apoc3, Fabp1, Atp5a1, Selenop, and Slc27a4; stem cells by expression of Lgr5, Ascl2, Slc12a2, Axin2, Olfm4, Gkn3, and mKi67; goblet cells by expression of Muc2, Tff3, Clca3, Agr2, Dmbt1, Olfm4, and Dmbt1; Paneth cells by expression of Lyz1, Defa17, Defa22, Defa24, Ang4, and Defa30; enteroendocrine cells (EECs) by expression of Chga, Chgb, Tac1, Tph1, Neurog3, Neurod1, and Cpe; and tuft cells by expression of Dclk1, Trpm5, Gfi1b, Il25, Hck, Sh2d6, Avil, and Trpm5 (Fig. 5, B and C). The enterocytes include heterogeneous subclusters that could be further distinguished by zonation through modulation scores of the top landmark and bottom landmark genes based on their distribution along the intestinal villus axis (Fig. 5, B and D; and Tables S4 and S5; Moor et al., 2018). UMAP visualization of IEC clusters showed that the intestine epithelial stem cells are transcriptionally close to enterocytes (villus bottom) populations, and the enterocytes (villus bottom) appeared to be most transcriptionally related to stem cell and enterocyte (villus top) populations at the same time (Fig. 5 B). CytoTRACE analysis indicated that stem cells are the least differentiated cells and enterocytes (villus top) are highly differentiated, whereas enterocyte (villus bottom) represents a less differentiated cell status in the middle (Fig. 5 E). Projection of the velocity field showed a strong directional flow of stem cell to enterocyte (villus bottom) and then enterocyte (villus top) fates (Fig. 5 F). Monocle analysis indicated a strong developmental trajectory from stem cell towards enterocyte (villus top), with enterocyte (villus bottom) representing an additional transition state (Fig. 5 G). Cell cycle analysis further validated stem cell status through a high proportion of cells in either G2/M or S phase (Fig. 5 H). These data indicated the heterogeneity of enterocytes along the developmental path from the bottom to the top of the villi. **Figure 5.:** *Diversity in structural cells and altered nutrient absorption uncovered. (A) Schematic of the experimental pipeline. C57BL/6J mice were fed on chow diet and HFHS diet for 8 wk before dissection and small intestines were isolated. Cells from epithelium layer were sorted for CD45.2− live cells and analyzed using 10× Genomics Chromium droplet scRNA-seq. Cells were clustered via differential gene expression for further studies. EEC, enteroendocrine cell. (B) UMAP plot of IECs derived from a cohort of 10 chow diet mice and 10 HFHS diet mice, where individual points correspond to single cells. Cluster analysis yields seven distinct clusters comprising IECs. (C and D) Feature plots showing the mean expression of known marker genes for a particular cell type or state projected on UMAP. Average scores in each cell are indicated in color saturation. (C) Average scores of signature genes for cluster annotation of enterocyte, stem cell, goblet cell, Paneth cell, EEC, and tuft cell (genes indicated above each plot). (D) Average scores of signature genes cluster annotation of enterocyte on top of the villus or at the bottom of the villus (genes indicated in Tables S4 and S5). (E) Visualizing the predicted differentiation using CytoTRACE analysis of IECs. The color indicates the levels of CytoTRACE score from 1 (red, lowest levels of differentiation) to 0 (grey, highest levels of differentiation). Feature plots showing the CytoTRACE score projections on UMAP plot. (F) RNA-velocity analysis of IECs with velocity field projected onto the UMAP plot. Arrows show the local average velocity evaluated on a regular grid and indicate the extrapolated future states of cells. (G) Monocle analysis of the IECs. The color indicates pseudotime directionality projection on UMAP from the earliest (blue) to the latest (yellow). (H) Cell cycle analysis of the IECs. Predicted classification of each cell in either G2/M (red), S (blue), or G1 (green) phase was projected on UMAP. (I) Dot plot showing the top DEGs for the enterocyte (villus top) and enterocyte (villus bottom) populations depicted. Color saturation indicates the strength of average gene expression, whereas the dot size reflects the percentage of each cell cluster expressing the gene. Top, enterocyte (villus top); bottom, enterocyte (villus bottom). (J) Histogram showing the proportions of IECs derived from 10 chow diet mice and 10 HFHS diet mice. (K and L) IPA analysis of transcriptional upstream regulators (left) and secreted upstream regulators (right) of the transition from stem cell to enterocyte (villus bottom; K) or enterocyte (villus bottom) to enterocyte (villus top; L) based on DEGs. (M) Heatmaps of selected genes that are critical for the absorption of distinct nutrient classes. (N) Average scores of critical genes expression for the absorption of distinct nutrient classes (fat and cholesterol, amino acid, peptide, and carbohydrate) in each cell from stem cells in the crypt to the enterocytes on the top of the villus. Zonation of the cells along the villus length was profiled by the pseudotime trajectories from Monocle analysis from the earliest (blue) to the latest (yellow; Fig. 5 F). (O) KEGG pathway gene set enrichment analysis of DEGs between chow diet IECs and HFHS diet IECs, indicating the top six-most upregulated (top) and most downregulated signaling pathways (bottom) in response to the HFHS diet.* The top variable expressed genes indicated that enterocytes (villus top) have higher expression of genes that facilitate nutrient transportation, while enterocytes (villus bottom) express higher levels of genes for antimicrobial and immunoregulatory components (Fig. 5 I). We further examined the key functional genes related to intestinal homeostasis and barrier function. Although the enterocyte (villus bottom) proportion decreased slightly (Fig. 5 J), the expression of genes involved in microbiota-host interactions dramatically increased in HFHS diet intestine. The expression of tight junction genes and mucin genes remains unchanged (Fig. S4, H and I). IPA analysis revealed dramatic differences in upstream signaling regulators between the transition from stem cell to the enterocyte (villus bottom) status and the subsequent transition to the enterocyte (villus top) status (Fig. 5, K and L). Notably, transcription factors such as MXD1 and KDM5A promote enterocyte maturation across the development trajectory. Growth factors such as oncostatin M (OSM), pleiotrophin (PTN), and TNFSF8 promote the early development of enterocytes at the bottom of the villus, whereas the WNT (WNT3A, WNT5A) and granulin (GRN) signaling pathways are crucial for the maturation of enterocytes during their migration from the bottom of the villus to the top (Fig. 5, K and L). A fundamental role of IECs is the absorption of nutrients, by which the IEC and whole-body metabolism are tightly interrelated (Zietek et al., 2020). Hence, we examined the transporters for key nutrient families and found that carbohydrate transporters were enriched in the HFHS diet IECs, whereas the expression of the Slc15a1 gene that encodes the main peptide transporter Pept1, the cholesterol transporter Npc1l1 and the expression of genes for the lipoprotein biosynthesis machinery necessary for the assembly of chylomicrons were all decreased (Fig. 5 M and Fig. S4 J). We also noted that the transporters for these key nutrient families have their own expression domains along the pseudo-timeline from stem cells to the enterocyte (villus top). The amino acid and carbohydrate transporters were enriched at the bottom to the middle of the villi, whereas peptide transporters shifted in expression towards the upper villus zones, and the lipid transporters peaked in expression at the villi tips (Fig. 5, G and N). Furthermore, the KEGG pathway enrichment analysis indicated a strong upregulation in carbohydrate absorption and carbon metabolisms in HFHS diet IECs (Fig. 5 O). In sum, single-cell profiling of IECs revealed heterogeneity in enterocytes, distinct transporter zonation and unique metabolic adaptions in response to the HFHS diet. ## Ligand–receptor analysis reveals intestine interactome network In order to investigate how intestine cellular communication networks and context-dependent crosstalk of different cell types enable diverse physiological processes to proceed in the intestine, we performed ligand–receptor analysis on all major intestinal cell populations from chow diet intestine using CellPhoneDB to generate intestine homeostatic interactomes (Efremova et al., 2020). This revealed tens of thousands of potential structural cell-to-immune cell, immune cell-to-immune cell, as well as structural cell-to-structural cell interactions (Tables S6 and S7). All of the interactions were further verified using CellTalkDB, a manually curated database of literature-supported ligand–receptor interactions (Shao et al., 2021). The interactomes indicated that lamina propria myeloid cells are major signal sources. There are strong interactions between myeloid cells and lymphocytes, as well as myeloid cells and structural cells. The strongest interactions were observed within myeloid cells and within structural cells, while moderate levels of interactions were observed within lymphocytes (Fig. 6 A). Connectome web analysis of the chow diet intestine revealed that cDC2B and macrophage are central communication hubs among immune cells. Not only do they have strong interactions between themselves, but also with pDCs, CD4+ LPL-T cells, CD8αα+ IEL-T cells, CD8αβ+ IEL-T cells, and EECs (Fig. 6 B). Notably, EEC is a central communication hub among structural cells and strong interactions were observed with the goblet cells (Fig. 6 B). **Figure 6.:** *Ligand–receptor analysis reveals intestine interactome network. (A) Interaction heatmap plotting the total number of chow diet intestinal cell ligand (y axis) and receptor (x axis) interactions for the specified cell types based on expression of the gene in at least 10% of the cell population. The color represents the number of interactions between cell types: higher number of interactions (red), lower number of interactions (blue). (B) Connectome web analysis of chow diet intestine highly interacting cell types based on ligand–receptor interactions. Vertex (colored node of specific cell type) size is proportional to the number of interactions to and from that cell type, whereas the thickness of the connecting lines is proportional to the number of interactions between two nodes. (C) Dot plots showing expression of ligands (left) and receptors (right) in chow diet mouse intestinal cells. Implicated chemokines are shown in the lower panel. Color saturation in dot indicates the strength of expression in specific cell type, whereas dot size reflects the percentage of cells in each population expressing the gene.* The analysis implicated a variety of uncharacterized as well as validated signaling pathways in intestinal homeostasis. The chow diet intestine interactome suggested that cDC2Bs serve an immunoregulatory role in the chow diet intestine via the production of protein S (Pros1), growth arrest-specific 6 (Gas6), complement component 3 (C3), leukocyte immunoglobulin-like receptor, subfamily B, member 4 (LILRB4; Lilrb4a), CD83 (Cd83), insulin-like growth factor 1 (Igf1), and placenta growth factor (Pgf). The interactions were not only with immune cells like macrophages, pDCs, mast cells and eosinophils, but also cDC2B itself. These immune cells further contribute to the regulation of the IEL-T cells and LPL-T cells via interactions involving Jagged2 (Jag2), and delta-like 4 (Dll4), B7-1 (Cd86), B7-2 (Cd80) and PD-L1 (Cd274). Meanwhile, semaphorin 4D (Sema4d), CCL3, CCL4, and CCL5 signals sent out by intraepithelial and lamina propria lymphocytes could affect macrophages, pDCs, cDC2As, and cDC2Bs in return (Fig. 6 C). As a major signal source of structural cells, EECs play an immunoregulatory role and interact with macrophages, pDCs, cDC2As, and cDC2Bs via vascular endothelial growth factor A (Vegfa) and semaphorin 4A (Sema4a) signals, which are crucial for maintaining T-cell priming function in dendritic cells and promoting the stability and function of Treg cells (Delgoffe et al., 2013). EEC is also the major source of gastric inhibitory polypeptide (Gip) and pancreatic polypeptide (Ppy), which play critical roles in intestinal homeostasis and epithelium integrity (Adriaenssens et al., 2015). As a major signal receiver among structural cells, Goblet cells receive signals via ephrin B1 (Efnb1), ephrin B2 (Efnb2), IGF1, hepatocyte growth factor (Hgf), and epidermal growth factor receptor ligands (Hbegf, Ereg, Areg, and Tgfa) from other structural cells and granulin (Grn) from macrophages, pDCs, and cDCs (Fig. 6 C). ILC2 and ILC3 are the major sources of granulocyte-macrophage colony-stimulating factor (GM-CSF; Csf2), indicating their indispensable role in dendritic cell development and function. Macrophage and cDC2B are the major sources of CXCL16, CXCL10, CXCL9, and platelet factor 4 (Pf4), indicating their roles in the recruitment and maintenance of IELs and LPLs in the intestine. EECs and goblet cells interact with all major immune cell populations including myeloid cells, plasma cells, ILCs, IEL-T cell, and LPL-T cell via expression of CX3CL1, CCL28, and CCL25, suggesting their additional roles in immune cell recruitment and maintenance of overall immune homeostasis (Fig. 6 C). ## Dramatic remodeling of the intestine interactome network in response to the HFHS diet To explore the changes that occurred to the intestine communication networks in response to a HFHS diet, we performed CellPhoneDB ligand–receptor analysis on all HFHS-diet intestinal cells and verified the identified interactions using CellTalkDB. The interactome analyses indicated an increased number of ligand–receptor pairings in myeloid cells in HFHS diet intestine, especially interactions involving cDC1 cells, while there were significantly decreased interactions between lymphocytes. Notably, EECs dramatically decreased their cell–cell interactions in HFHS diet intestine while tuft cells became the major signal source among all structural cells (Fig. 7 A; and Fig. S5, A–C). Connectome web analysis further revealed an enrichment of cDC1, cDC2B, macrophage, and tuft cell communication hubs within HFHS diet intestine (Fig. 7 B). **Figure 7.:** *Dramatic remodeling of the intestine interactome network in response to the HFHS diet. (A) Interaction heatmap plotting the total number of HFHS diet intestinal cell ligand (y axis) and receptor (x axis) interactions for the specified cell types based on expression of the gene in at least 10% of the cell population. The color represents the number of interactions between cell types: higher number of interactions (red), lower number of interactions (blue). (B) Connectome web analysis of HFHS diet intestine highly interacting cell types based on ligand–receptor interactions. Vertex (colored node of specific cell type) size is proportional to the number of interactions to and from that cell type, whereas the thickness of the connecting lines is proportional to the number of interactions between two nodes. (C) Dot plots showing expression of ligands (left) and receptors (right) in HFHS diet mouse intestinal cells. Implicated chemokines are shown in the lower panel. Color saturation in dot indicates the strength of expression in specific cell type, whereas dot size reflects the percentage of cells in each population expressing the gene.* **Figure S5.:** *Comparison between chow diet and HFHS diet intestine interactomes and antibody blockade experiments in vivo. (A) Interaction heatmap showing the alternation of total number of interactions from chow diet intestine to HFHS diet intestine for the specified cell types based on expression of the gene in at least 10% of the cell population. Source cells, y axis; receiver cells, x axis. The color represents the number of interactions changes: higher number of interactions in HFHS intestinal cells (red), lower number of interactions in HFHS intestinal cells (blue). (B) Histogram showing the cell types most upregulated and most downregulated in total receiver counts from chow diet intestine to HFHS diet intestine. Each color represents a cell type. (C) Histogram showing the cell types most upregulated and most downregulated in total source counts from chow diet intestine to HFHS diet intestine. Each color represents a cell type. (D) Significant signaling pathways were ranked based on differences in the overall information flow within the inferred networks between chow diet intestine and HFHS diet intestine. The overall relative information flow of a signaling network is calculated by summarizing all communication probabilities in that network. The top signaling pathways with P value <0.05 were labeled with red (enriched in HFHS diet intestine) and blue (enriched in the chow diet intestine). A paired Wilcoxon test is utilized to determine whether there is significant difference between two datasets. IGF, insulin-like growth factor; GDF, growth differentiation factor; PDGF, platelet-derived growth factor; EGF, epidermal growth factor; EPHA, ephrin-A; EPHB, ephrin-B; APRIL, a proliferation-inducing ligand; SEMA, semaphorin; MPZ, myelin protein zero; GH, growth hormone; CEACAM, carcinoembryonic antigen-related cell adhesion molecules; RELN, reelin; TWEAK, TNF-related weak inducer of apoptosis; NRG, neuregulins; SELL, selection; LIFR, LIF receptor; ANGPT, angiopoietin; OSM, oncostatin M; ALCAM, activated leukocyte cell adhesion molecule; THBS, thrombospondin; JAM, junctional adhesion molecules; LT, lymphotoxin; PARs, protease-activated receptors; SELPLG, selectin P ligand; CLEC, C-type lectin-like receptor; MIF, macrophage migration inhibitory factor. (E–H) Top differentially regulated signaling pathways suggested in g:GOSt functional enrichment analysis of HFHS diet-enriched DEGs for the indicated cell types. (E) g:GOSt analysis of CD4+ IEL-T cell. (F) g:GOSt analysis of CD4+ LPL-T cell. (G) g:GOSt analysis of ILC2. (H) g:GOSt analysis of ILC1. (I–K) C57BL/6J mice received i.p. injection of anti-XCL1 antibodies (25 μg/animal, once per week; n = 4) to block XCL1–XCR1 signal; mice received i.p. injection of isotype antibodies (25 μg/animal, once per week; n = 4) were included as controls. Mice were fed on a HFHS diet for 4 wk and, following euthanasia, small intestines were isolated. Cells from the epithelium layer and lamina propria layer, were purified and then subjected to flow cytometry analysis. WT, wild type. (I) Schematic of the experimental pipeline. (J and K) Histogram showing the proportion of intraepithelial immune cells (J) and lamina propria immune cells (K) derived from mice intestine. (L) C57BL/6J mice received i.p. injection of anti-XCL1 antibodies (25 μg/animal, once per week; n = 4) to block XCL1–XCR1 signal; mice received i.p. injection of isotype antibodies (25 μg/animal, once per week; n = 4) were included as controls. Mice were fed on a HFHS diet for 4 wk and, following euthanasia, small intestines were isolated. Cells from epithelium layer were sorted for CD45.2− live cells and analyzed using RT-QPCR. Histogram showing the mRNA expression levels of genes involved in fat and cholesterol absorption, amino acid transporters, peptide transporters, and carbohydrate transporters in sorted IECs. (M–O) C57BL/6J mice received i.p. injection of anti-CD226 antibodies (25 μg/animal, once per week; n = 4) to block NECTIN-CD226 signal; mice received i.p. injection of isotype antibodies (25 μg/animal, once per week; n = 4) were included as controls. Mice were fed on a HFHS diet for 4 wk and, following euthanasia, small intestines were isolated. Cells from epithelium layer and lamina propria layer, were purified and then subjected to flow cytometry analysis. (M) Schematic of the experimental pipeline. (N and O) Histogram showing the proportion of intraepithelial immune cells (N) and lamina propria immune cells (O) derived from mice intestine. Representative of 2 (I–O) experiments. All data are presented as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by two-way ANOVA (J–L, N, and O). Statistics are all two-sided.* Intestine interactomes analyses showed that cDC1s have increased interactions with macrophages, cDC2As and cDC2Bs, and are the major receivers of the immune-regulatory signal of Wnt4, HGF, and PGF (Albonici et al., 2019; Haseeb et al., 2019; Hübel and Hieronymus, 2015). Further analysis in HFHS diet intestine identified cDC1 as a key regulator expressing immune-regulatory signals such as TGFβ (Tgfb1), ephrin B1, IL-15, annexin A1 (Anxa1), secreted phosphoprotein 1 (Spp1), and LILRB4. Our interaction analyses also indicated that both CD8αα+ T cells and CDαβ+ T cells dramatically increased the production of the chemokine XCL1 and were involved in the XCL1–XCR1-mediated recruitment of cDC1s, and that cDC1s further interact with CD4+ IEL-T cells, CD4+ LPL T cells, and ILC2s via CCL22 (Fig. 7 C; and Fig. S4, A and B). Analysis of the intestine interactome suggested that among myeloid cells, cDC2Bs and eosinophils may contribute to tissue inflammation during HSHS diet via TNF-, IL-1β (Il1b)-, KIT-ligand (Kitl)-, TNFSF13B-, TNFSF3 (Ltb)-, granulin-, semaphorin 4A-, CXCL10-, and CXCL2-mediated regulation of IEL-T cells, LPL-T cells, lamina propria ILCs, and myeloid subsets. The interactome also identified CD4+ IEL-T cells, CD4+ LPL T cells, ILC1s, ILC2s, and ILC3s as major producers of strong pro-inflammatory signals, including Fam3c, TNF, TNFSF3, TNFSF11, TNFSF14, and IL-21 (Fig. 7 C). Together with the increased proportions of CD4+ IEL-T cells, CD4+ LPL T cells and ILCs in HFHS diet intestine (Fig. 1, D and E), these results further supported their roles in the potentiation of inflammatory responses during HFHS diet feeding. The chemokine analysis suggests that CCL3, CCL4, CCL5, and their main receptors (CCR1 and CCR5) play an important role in cDC2A and cDC2B recruitment. cDC2Bs also express high levels of CX3CR1, which receive signals from CX3CL1 produced by IECs (Fig. 7 C). As to the ILCs, NK cells and ILC1s receive signals from CCR5 ligands (CCL3, CCL4, and CCL5) due to the high expression of *Ccr5* gene. ILC1s, ILC2s, and ILC3s have high expression of CCR9, which could receive a signal from CCL25 produced by the IECs (Fig. 7 C). cDC1s and CD4+ LPL-T cells also express high levels of CCR9, indicating that IECs, especially goblet cells and enterocytes, play an important role in immune cell infiltration through the CCL25–CCR9 axis. CD4+ LPL-T cells, cDC1, NK cell, ILC1s, and ILC3s also received signals from CXCR3 ligands (CXCL10, CXCL9, and PF4), which are mainly produced by myeloid cells and are known to play critical roles in lymphocyte infiltration (Fig. 7 C). The analyses also identified tuft cells as an emerging communication hub in HFHS diet intestine and found that they exhibited dramatically increased immune-regulatory signal expression, such as Vegfa, Efnb1, Lilrb4a, Pros1, and C3. Tuft cells are also the major receivers for immune-regulatory signals from immune cells such as IL-4, IL-13, and CD83, and from structural cells, such as ephrin A1 (Efna1), ephrin A4 (Efna4), angiopoietin-2 (ANGPT2; Angpt2), and transthyretin (Ttr; Fig. 7 C and Fig. S5, A–C). Together, these data suggested significant communication shifts within the intestine interactome during HFHS diet and implicate new cell subsets in the potentiation of intestine inflammation and immune regulation in response to HFHS diet. ## A distinct HFHS diet-enriched intestine signalome uncovered To further understand how intestinal cells coordinate their functions, we used CellChat to conduct systems-level analysis through classifying conserved and context-specific signaling pathways (Jin et al., 2021). We projected signaling pathways onto a 2D manifold according to their functional similarity, which heavily weighs the similarities between sender and receiver cell group patterns, and revealed four groups of pathways (Fig. 8 A). Group #1 was dominated by signals produced by both epithelium cells and myeloid cells, such as TGFβ, PECAM, CLEC, and GALECTIN. Group #2 was dominated by co-stimulatory signals from myeloid cells and B cells, such as CD86, CD80, L1CAM, and ICOS. Group #3 mainly contained growth factor signals, such as IGF, GDF (growth differentiation factor), PDGF (platelet-derived growth factor), EGF, and GH (growth hormone). In addition, group #4 mainly consists of signals of immune response activation and regulation, such as IL-1, IL-2, IL-4, IL-6, IL-10, IFN-II, and RANKL. We then compared the information flow for each given signaling pathway by the sum of communication probability across chow and HFHS diet intestine and found that many group #1 pathways, including CD52, ICAM, MHC-I, LAMININ, PECAM, and GALECTIN maintain consistent in chow and HFHS diet intestine. They likely represented core signaling pathways essential for intestine function independent of the diet (Fig. S5 D). In contrast, 33 out of 101 pathways prominently changed information flow upon exposure to the HFHS diet. These consist mainly of signaling pathways from myeloid cells such as CCL, L1CAM, THBS, CLEC, NECTIN, and SEMA4. Presumably, these pathways are involved in inflammatory and intestinal homeostasis (Fig. S5 D). NECTIN pathways showed the most significant increase in information flow among these active pathways in the HFHS diet intestine (Fig. 8 B). Hierarchical plots indicated that the NECTIN signaling network in intestine is highly complex, with cDC2As and cDC1s as the sources of NECTIN ligand in immune cells, and with enterocytes, EECs, goblet cells, and tuft cells as the sources of the NECTIN ligand in structural cells. Meanwhile, the majority of NECTIN signal receivers are LPL-T cells, ILCs, and IEL-T cells, with minor receivers from structural cells (Fig. 8 C). In response to the HFHS diet, EECs lost their expression of NECTIN ligand and responsiveness while tuft cells dramatically increased their NECTIN ligand expression, driving an overall increase in NECTIN communication (Fig. 8 C). The interactions between nectin-2 on APCs and its ligand CD226 on NK cells and T cells have been reported to play an important role in T cells and NK cells activation (Huang et al., 2020; Tahara-Hanaoka et al., 2004; Yeo et al., 2021). Interestingly, CD226, grouped together with signaling pathways involving immune cell activation in group #4 (Fig. 8 A), is also among the most significantly upregulated signaling pathways in HFHS diet intestine (Fig. 8 B and Fig. S5 D). This suggests that the NECTIN–CD226 signaling pathway plays an important role in the activation of LPL-T, ILCs, and IEL-T cells, and the intestinal homeostasis in response to the HFHS diet. **Figure 8.:** *Analysis of signaling pathways and information flow reveals a distinct intestine signalome in response to the HFHS diet. (A) Jointly projecting and clustering signaling pathways from chow diet intestine and HFHS diet intestine onto shared 2D manifold according to functional similarity of the inferred networks. Each dot represents the communication network of one signaling pathway. Dot size is proportional to the communication probability. Different colors represent different groups of signaling pathways. IGF, insulin-like growth factor; GDF, growth differentiation factor; PDGF, platelet-derived growth factor; EGF, epidermal growth factor; EPHA, ephrin-A; EPHB, ephrin-B; APRIL, a proliferation-inducing ligand; SEMA, semaphorin; MPZ, myelin protein zero; GH, growth hormone; CEACAM, carcinoembryonic antigen-related cell adhesion molecules; RELN, reelin; TWEAK, TNF-related weak inducer of apoptosis; NRG, neuregulins; SELL, selection; LIFR, LIF receptor; ANGPT, angiopoietin; OSM, oncostatin M; ALCAM, activated leukocyte cell adhesion molecule; THBS, thrombospondin; JAM, junctional adhesion molecules; LT, lymphotoxin; PARs, protease-activated receptors; SELPLG, selectin P ligand; CLEC, C-type lectin-like receptor; MIF, macrophage migration inhibitory factor. (B) Significant signaling pathways were ranked based on differences in the overall information flow within the inferred networks between chow diet intestine and HFHS diet intestine. The overall information flow of a signaling network is calculated by summarizing all communication probabilities in that network. The top signaling pathways with P value < 1 × 10−5 were listed. Pathways colored red are enriched in HFHS diet intestine, and those colored blue were enriched in the chow diet intestine. A paired Wilcoxon test is utilized to determine whether there is significant difference between two datasets. (C) Hierarchical plot showing the inferred intercellular communication network of NECTIN signaling pathway at chow diet intestine and HFHS diet intestine, respectively. Circle color indicates each cell type and edge width are proportional to the communication probability.* To further explore whether predicted interacting cell types influence transcriptional changes during the HFHS diet challenge, we performed IPA on the DEGs between all chow diet and HFHS diet cell populations (Tables S8 and S9). Our analysis identified ∼269 genes implicated as shared upstream regulators in ≥10 cell types (Table S10). The analysis indicated that many signaling pathways identified utilizing CellPhoneDB and CellChat served as upstream regulators of gene expression in intestinal cells in response to a HFHS diet (Fig. 9 A). Many of the common secreted upstream regulators we identified, such as TNF, IL-1β, IFN-γ, IL-18, CD40L, IL-6, EGF, TGFβ, and IL-21, were previously reported to be associated with obesity and metabolic dysfunction (Table S11). Moreover, our analysis suggested that many previously uncharacterized intestinal immune cells, including CD4+ IEL-T cells, ILC1s, ILC2s, and CD4+ LPL-T cells, expressed high levels of the downstream genes of these key inflammatory pathways associated with obesity, which may contribute to their proliferation and infiltration, in accordance with changes in cell proportion and ligand–receptor interaction analysis (Fig. 7 C and Fig. 9 A). **Figure 9.:** *Analysis of upstream regulators uncovers a distinct intestine signalome in response to the HFHS diet. (A and B) IPA of HFHS immune and non-immune populations showing common upstream regulators. Terms were considered common if implicated in 10 or more cell types from a lineage (Tables S8, S9, and S10). Terms were considered statistically significant if the activation z-score >2. (A) Dot plots showing expression of common secreted upstream regulators from HFHS diet intestinal cells (left) and the putative regulated cell types (right) as suggested by IPA. Color saturation in dot indicates the strength of expression in specific cell type, whereas dot size reflects the percentage of cells in each population expressing the gene (left). The color indicates the implicated cell type, whereas dot size reflects the number of genes downstream of the suggested secreted upstream regulator. Red-highlighted upstream regulators denote those that have been associated with obesity or metabolic dysfunction. Blue-highlighted upstream regulators denote those that have been associated with immune homeostasis or obesity resistance. (B) Dot plot showing common transcriptional signaling upstream regulators. The color indicates the implicated cell type, whereas dot size reflects the number of genes downstream of the suggested signaling upstream regulator. IEL, intraepithelial lymphocyte; LP, lamina propria; Str., structural cells. (C) Top differentially regulated signaling pathways in response to HFHS diet suggested in g:GOSt functional enrichment analysis of common upstream regulators. Upstream regulators were considered common if implicated in 10 or more cell types from a lineage. Terms with −log10 (adjusted P value) >10 were shown.* Our analysis also identified several secreted upstream regulators that have not yet been associated with intestinal inflammation but have been suggested to play a role in the development of metabolic dysfunction (GM-CSF, TNFSF11, IL-2, and Angiotensinogen [Agt]), suggesting that these signals may exert their systemic effects through targeting intestinal immune cells (Table S11). Furthermore, our results also revealed several common upstream regulators associated with immune regulation and resistance to obesity (IL-4, IL-5, IL-33, CD15, HGF, ANGPT2, VEGFA, and neuropilin 1 [Nrp1]), supporting a critical role of intestinal homeostasis in systemic pathogenesis of obesity (Table S11). Many previously uncharacterized structural cells including tuft cells, EECs, stem cells, and enterocytes highly expressed the upstream ligands of many of these immunoregulatory pathways in the obesifying diet, indicating their potential roles in regulating intestinal immune homeostasis (Fig. 9 A). Analysis of non-secreted upstream regulators indicated that EECs, Paneth cells, ILC1s, ILC2s, and CD4+ IEL-T cells expressed the highest levels of the genes downstream of the inflammatory signaling regulators (such as Myc, Nfe2l2, Xbp1, Stat4, and Tcf7l2), indicating those intestinal cells may play an important role contributing to obesity-associated inflammation and metabolic dysfunction (Fig. 9 B). Together, these results further supported the identification of pro-inflammatory signal producers in the ligand-receptor interactome (Fig. S5, E–H). Functional enrichment analysis of the common upstream regulators suggested a number of significant pathways that are transcriptionally regulated in HFHS diet intestine. These involved several pro-inflammatory signaling pathways, including IL-17, JAK-STAT, TNF, PI3K-Akt, AGE-RAGE, and MAPK (Fig. 9 C). Interestingly, signaling pathways in cancer are at the top of the list of most differentially regulated signaling pathways in response to HFHS diet. It is known that dietary fat intake is associated with an elevated risk of developing colorectal cancer (Keum and Giovannucci, 2019). Our results suggest that upstream regulators such as MYC, TP53, FOXO1, H1F1A, and FOS might contribute to an increased risk of cancer development in the intestine in response to an HFHS diet (Table S12). We followed up on the XCR1–XCL1 signaling pathway, which is known to be critical for cDC1 recruitment. To study XCR1–XCL1 signaling pathway in vivo, we utilized anti-XCL1 antibodies i.p. injection to block the XCR1–XCL1 signaling pathway in mice fed on a HFHS diet (Fig. S5 I). The immune cells of the epithelium layer and lamina propria layer were isolated from both the XCL1 signaling blockade group and the control group, and then analyzed by flow cytometry for major immune cell populations (Fig. S5 I). The results show that abrogation of cDC1s increased the CD8αα+ IEL-T cells in the epithelium layer and CD4+ LPL-T cells in the lamina propria layer in the mouse intestine (Fig. S5, J and K). These results indicate that XCL1–XCR1 signaling pathway is involved in the regulation of inflammatory responses and intestinal homeostasis upon HFHS diet feeding. Significant CD4+ LPL-T cell infiltration was observed in the intestine upon HFHS diet feeding (Fig. 1, E–G), and cDC1s were the major receiver and the sender of a variety of immune-regulatory signals, and they appeared to play a central role in the responses to a HFHS diet (Fig. 7, A–C). These results suggest that cDC1s may play a regulatory role in intestinal homeostasis and modulate the CD4+ LPL-T cells and CD8αα+ IEL-T cells activation upon HFHS diet feeding. Furthermore, to study the impact of cDC1 abrogation on IECs upon HFHS diet feeding, we also sorted non-hematopoietic CD45.2− live cells from the epithelium layer of the mice from XCL1–XCR1 signaling blockade group and the control group, and then profiled them using RT-QPCR (Table S13). We examined the gene expression of transporters for key nutrient families, and found that amino acid transporters (Slc7a8, Slc7a9), and peptide transporters (Slc15a1, Slc15a2) were increased in the XCL1–XCR1 signaling blockade group, whereas the expression of the carbohydrate transporters (Slc2a2, Slc2a5) decreased. *The* genes involved in fat and cholesterol absorption (Apob, Apoa1, Npc1l1) remain unchanged (Fig. S5 L). Taken together, these results indicate that cDC1s also play an important role in regulating IEC functions, further supporting their role as a central communication hub in the HFHS diet intestine. We also studied the NECTIN pathway, the most significantly increased pathway in information flow in response to the HFHS diet, in vivo. We utilized anti-CD226 antibodies i.p. injection to block the NECTIN–CD226 signaling pathway in mice fed on a HFHS diet (Fig. S5 M). The immune cells of two major intestinal compartments, the epithelium layer and lamina propria layer, were isolated from both the NECTIN signaling blockade group and the control group, and then analyzed by flow cytometry for major immune cell populations (Fig. S5 M). The results show that there was no significant change in intraepithelial immune cell populations (Fig. S5 N). However, dramatic decreases of CD4+ LPL-T cells, CD4−CD8− LPL-T cells and NK cell/ILC1s were observed in the intestine of mice with the CD226 signaling blockade (Fig. S5 O). In contrast, the proportions of myeloid cells, including cDC2As, cDC2Bs, and macrophages, increased (Fig. S5 O). These results indicate that NECTIN–CD226 signaling pathway is involved in inflammatory responses and intestinal homeostasis. Moreover, the NECTIN pathways showed the most significant increases in information flow among active pathways in the HFHS diet intestine (Fig. 8 B), and the majority of NECTIN signal receivers are T cells and ILCs (Fig. 8 C). Taken together, these results suggest that the NECTIN–CD226 signaling pathway plays an important role in the accumulation of CD4+ LPL-T cells, CD4−CD8− LPL-T cells and ILCs, promoting intestinal inflammation in the HFHS diet feeding mice. Combined, these data suggest that a complex network of inflammatory and immunomodulatory signaling pathways in both immune cells and structural cells regulates the homeostasis in the intestine during HFHS diet-induced obesity. ## Discussion We performed scRNA-seq on a total of 70,830 cells from multiple sorted atlases including intraepithelial CD45+ cells, lamina propria CD45+ cells, and epithelium CD45− cells from mice on chow and HFHS diets. This led to the validation of 32 out of 32 previously identified populations, as well as the identification of an additional 20 distinct cell clusters, including previously uncharacterized populations of CD8αα+ IEL-T cell, CD8αβ+ IEL-T cells, CD4+ IEL-T cells, CD4+ LPL-T cells, cDCs, and enterocytes. Using flow cytometry, we further validated the presence of each immune cell population in the chow diet and HFHS diet intestines using an independent cohort of mice. Finally, we utilized the single-cell ligand–receptor analysis to profile both the chow diet and the HFHS diet intestine interactomes in a cell type-specific way, establishing a signalome across intraepithelial immune cells, lamina propria immune cells and epithelial structural cells in the intestine. Our studies identified unique immune cell populations that accumulate in response to the HFHS diet and the associations between them. For example, Treg cell proportions were dramatically increased in the HFHS diet, as were CD4+ IEL-T cells and CD4+ LPL-T cells, and cDC1s, which are critical for Treg induction. Moreover, memory-like CD8αα+ IEL-T cells as well as memory-like CD8αβ+ IEL-T cells accumulated in HFHS diet intestine, expressed high levels of XCL1, and are likely involved in the recruitment of cDC1s via XCL1–XCR1 signaling axis in response to the diet. This mechanism is very similar to the cDC1 recruitment and immune tolerance induction mechanism discovered in the tumor microenvironment (Böttcher et al., 2018). Although the percentage of Treg cells increased in CD4+ T cell populations (Fig. 1 D), the RT-QPCR analysis and absolute cell counting indicate increased Th1 and Th17 responses in the lamina propria due to increased CD4+ T cell infiltration (Fig. S3, F and G). Analysis of the intestine interactome suggested that CD4+ T cells are one of the major producers of many strong pro-inflammatory signals (Fig. 7 C). Furthermore, IPA analysis indicated that CD4+ T cells expressed high levels of the downstream genes of the key inflammatory pathways associated with obesity (Fig. 9, A and B). These results suggest that CD4+ T cells also play a pro-inflammatory role in response to the HFHS dietary challenge. Notably, cDC1s were the major receiver and the sender of a variety of immune-regulatory signals, and they appeared to play a central role in the responses to a HFHS diet. Moreover, the abrogation of cDC1s significantly increased the CD4+ LPL-T cell and ILCs infiltration in the intestine upon HFHS diet feeding. These results suggested a previously unrecognized role of the CD8+ IEL-T cell–cDC1–CD4+ T cell axis in the balance of intestinal homeostasis (Fig. 10). **Figure 10.:** *Graphical summary of cell–cell interactions in healthy and HFHS diet intestine. In healthy intestine, cDC2Bs and macrophages play as central communication hubs, regulating IEL-T cells and LPL-T cells via Notch ligands, B7-1, B7-2, PD-L1, CXCL9, CXCL10, CXCL16, and PF, and regulating EECs, tuft cells, and goblet cells via GRN and CD83. cDC2Bs are also the major producers of immunoregulatory signals such as PROS, IGF, C3, and GAS6. Among structural cells, EECs play as a central communication hub and interact with macrophages, cDC2Bs, and cDC2As via VEGF, GIP, PPY, and semaphorins. EECs and tuft cells may support goblet cells via production of HBEGF, EREG, AREG, TGFA, GRN, and ephrin-Bs. ILC2 and ILC3 also support cDCs through GM-CSF production. During HFHS diet-induced obesity, EECs decrease their cell–cell interactions while tuft cells become the major signal source among structural cells. Tuft cells may regulate with macrophages and cDC2Bs via VEGF, PROS, C3, and ephrin-Bs, and receive signals from goblet cell and cDC2Bs via CD83, XPR1, KITL, and ephrin-As. Notably, tuft cells, goblet cells, and enterocytes increase their interactions with CD8+ IEL-T cells via NECTIN signaling in HFHS diet intestine. Increased XCL1 production of CD8+ IEL-T cells could further recruit more cDC1s in the lamina propria. As a communication hub in HFHS diet intestine, cDC1 produce immunoregulatory signals such as SPP1, LILRB4, TGFβ, and IL-15, and received HGF and PGF signals from macrophage and cDC2B. cDC1 and cDC2A could also interact with IEL-T cells and LPL-T cells via nection-2 signaling. In contrast, cDC2B produce signals such TNF, IL-1β, TNFSF13B, TNFSF3, semaphorins, GRN, CXCL2, CXCL10, and potentiate inflammation in the HFHS diet intestine. EECs, enteroendocrine cells; GRN, granulin; PROS, protein S; IGF, insulin-like growth factor; GAS6, growth arrest-specific 6; VEGF, vascular endothelial growth factor; GIP, gastric inhibitory polypeptide; PPY, pancreatic polypeptide; HBEGF, heparin binding EGF-like growth factor; EREG, epiregulin; AREG, amphiregulin; XPR1, xenotropic and polytropic retrovirus receptor 1; KITL, KIT ligand; SPP1, secreted phosphoprotein 1; HGF, hepatocyte growth factor; PGF, placental growth factor.* Our studies also spatially reconstituted the stem cell to enterocyte development trajectory in the intestine villus. We found that the transporters of the key nutrient families (amino acid, peptide, lipid, and carbohydrate) have their own expression domains along the developmental timeline zonation from stem cell to villus tip. We also found that IECs in the intestine undergo unique adaptions in response to the HFHS diet, especially the expression of distinct nutrient transporters. Moreover, it has long been known that the essence of immune responses, whether to infection, in autoimmunity, or to cancer, is orchestrated by the principal cell populations of the immune system, that is, lymphocytes and myeloid cells. However, our present results indicate that structural cell types also dictate the immune responses. For example, in response to HFHS diet feeding, EECs reduced expression of the NECTIN signal, which is actively involved in LPL-T cell and IEL-T cell activity in healthy intestine. Furthermore, IPA analysis also showed that EECs expressed the highest levels of genes responsive to the inflammatory intestinal microenvironment. Intestine interactome analyses also indicated that EECs act as a central communication hub among structural cells and largely lose their status to tuft cells in response to the HFHS diet (Fig. 10). These findings highlight the study of IEC populations in regulating the immune homeostasis and systemic metabolism. A limitation of our study is that we did not incorporate the effects of the gut microbiome in our analysis. However, we did observe that the expression of genes involved in microbiota–host interactions was significantly affected by Western diet (Fig. S4, H and I). In summary, our study provides a comprehensive landscape of intestinal immune cell and structural cell populations on a chow diet and in response to a HFHS diet. It also identified cell type-specific transcriptional changes and communication networks that underlie intestinal homeostasis on the two diets. Our study filled a knowledge gap in analyzing the different layers of intestinal cells and interaction networks across immune cells and structural cells. The study revealed many previously uncharacterized intestinal immune cells and their potential roles in inflammatory pathways, and identified many upstream regulators in obesity-associated inflammation, suggesting that they may exert their systemic effects through targeting specific intestinal cells. ## Mice C57BL/6J (B6) mice were purchased from the Jackson Laboratory. 8-wk-old female mice were used for all experiments unless otherwise indicated. All experiments were repeated at least three times unless specifically mentioned. Replicates of each individual experiment are detailed in the figure legends. All animals were maintained at the University of California, Los Angeles (UCLA) animal facilities and all experiments were approved by the Institutional Animal Care and Use Committee of UCLA (ARC-92-169). Mice were maintained on a 12-h light/dark cycle from 6 am to 6 pm at ambient temperature (∼72°F degree) with controlled humidity (∼$45\%$) in specific pathogen-free conditions. Animals were randomly assigned to each treatment group and experiments were performed under standard laboratory procedures of randomization and blinding. For HFHS diet mice models, 8-wk-old mice were fed with either a chow diet (Cat# 2916; Teklad) or a HFHS diet (Cat# D12266B; Research Diets). ## Isolation of mouse intraepithelial immune cells, lamina propria immune cells, and IECs Mice were euthanized using isoflurane (VETONE) and secondary cervical dislocation. The intestine cell isolation protocols were optimized by our lab following the guidelines (Lefrancois and Lycke, 2001; Qiu and Sheridan, 2018; Sydora et al., 1996). The intestines were next maintained in a moistened state, and the Peyer’s patches and the remaining mesentery/fat were excised. The small intestine from ileum to jejunum was collected and gently flushed with a syringe with ice-cold 1× Hanks’ balanced salt solution (HBSS) media (without Ca2+ and Mg2+; Cat# H4641; Sigma-Aldrich) to remove the feces. The intestine was cut open and dissected into 1-cm segments, which were then transferred into a 50 ml tube and gently shaken to let the tissue stretch. They were then washed twice with ice-cold 1 × HBSS and the extra HBSS was removed using a pipet. 20 ml pre-digestion solution (1× HBSS with $5\%$ FBS, 5 mM EDTA, and 1 mM DTT) was added and incubated for 20 min at 37°C with 250 rpm rotation in an incubator. The 1× HBSS containing the epithelial cell layer was collected on ice after the vortex and passed through a 100-μm cell strainer. For intraepithelial immune cell isolation, the 1× HBSS containing the epithelial cell layer was centrifuged at 2,000 rpm for 5 min. The cell pellet was resuspended in 20 ml $40\%$ percoll (Cat# P4937; Sigma-Aldrich) and 20 ml $70\%$ percoll added to the bottom of each tube, spun at 2,000 rpm for 20 min (accelerate at 6 and brake at 2) at room temperature. The middle layer of cells was collected, washed, spun down with 1× PBS, and resuspended in RPMI 1640 culture media (Cat# 10-040-CV; Corning Cellgro; supplemented with $10\%$ FBS (Cat# MT35015CV; Corning), $1\%$ penicillin-streptomycin-glutamine (Cat# 10378016; Gibco), $1\%$ MEM Non-Essential Amino Acids Solution (Cat# 11140050; Gibco), $1\%$ HEPES [Cat# 15630056; Gibco], and $1\%$ sodium pyruvate [Cat# 11360070; Gibco]) for further FACS or flow cytometry analysis. For intraepithelial immune cell scRNA-seq, single cell suspensions were sorted using a FACSAria III flow cytometer to purify live hematopoietic cells (gated as DAPI−CD45.2+ cells). For IEC isolation, the epithelial cell layer in the 1× HBSS was spun at 2,000 rpm for 5 min. The cell pellet was resuspended in RPMI 1640 culture media for further FACS. For epithelial structural cell scRNA-seq, single cell suspensions were sorted using a FACSAria III flow cytometer to purify live non-hematopoietic cells (gated as DAPI−CD45.2− cells). For lamina propria immune cell isolation, the epithelial cell layer was removed and the remaining intestine tissues were washed twice with RPMI 1640 culture media. The extra media was removed and 10 ml RPMI 1640 culture media with 1 mg/ml Collagenase type I (Cat# 17100017; Thermo Fisher Scientific) was added. The sample was incubated for 20 min at 37°C with 250 rpm rotation and vigorously vortexed to allow the tissue to fully dissolve before passing through 100-μm cell strainer. Media containing the lamina propria cells was centrifuged at 2,000 rpm for 5 min and the cell pellet was resuspended in 20 ml $40\%$ percoll. 20 ml $70\%$ percoll was added to the bottom of each tube and spun at 2,000 rpm for 20 min (accelerate at 6 and brake at 2) at room temperature. The middle layer of cells was collected, washed, spun down with 1× PBS, and resuspended in RPMI 1640 culture media for further FACS or flow cytometry analysis. For lamina propria immune cell scRNA-seq, single cell suspensions were sorted using a FACSAria III flow cytometer to purify live hematopoietic cells (gated as DAPI−CD45.2+ cells). ## mIHC staining Small intestines from the ileum to jejunum were collected from experimental mice at the termination of an experiment and kept moistened. They were gently flushed with ice-cold 1× HBSS media to remove the feces with a syringe and cut into 1-cm-long segments. The parts of ileum 2 cm away from the cecum were collected for further mIHC analysis. Intestine sections were stained with the manual Opal 7-Color IHC Kit (Cat# NEL811001KT; Akoya Biosciences) with modification. Intestinal sections were prepared with a formalin-fixed, paraffin-embedded (FFPE) technique. The slides were dewaxed with xylene (3 × 10 min) and rehydrated through a graded series of ethanol solutions: ($100\%$ 1 × 5 min; $95\%$ 1 × 5 min; and $70\%$ 1 × 2 min) and washed in distilled water (1 × 2 min) and TBST (1 × 2 min). Then slides were then placed in a plastic jar with AR buffer and boiled by microwave for 45 s at $100\%$ power and an additional 15 min at $20\%$ power. Then slides were washed in distilled water (1 × 2 min) and TBST (1 × 2 min). The tissue sections were covered with blocking buffer (PerkinElmer antibody diluent buffer, Cat# ARD1001EA) and incubated in a humidified chamber for 10 min at room temperature. Then primary antibodies were applied overnight or at room temperature for 1 h according to antibody sensitivity. After incubation, slides were washed with TBST three times and incubated with secondary antibodies for 30 min at room temperature. Slides were washed three times with TBST and incubated with opal Fluorophore Working Solution (1:50 dilution) for 10 min to amplify the signals. After signal amplification, slides were washed three times with TBST, boiled with a microwave in AR buffer and washed with distilled water and TBST to strip the antibody complex. The steps of blocking, first antibody and secondary antibody were repeated for each antibody. After the incubation of all antibodies, slides were incubated with DAPI (1:2,000 dilution) and mounted. 32 sections from 16 mice in each group were studied by mIHC. An average of the two sections from the same mouse was used for the quantification of the sample. For IHC of intestine tissues, FFPE intestine samples were stained with anti-CD3 (Cat# A0452; Agilent), anti-CD8α (Cat# 14-0808; eBioscience), anti-CD4 (Cat# AB183685; Abcam), and anti-Ki67 (Cat# 12202; Cell Signaling) as primary antibodies. EnVision + System-HRP, Labelled Polymer (goat anti-mouse; Cat# K4001; Agilent), EnVision + System-HRP, Labelled Polymer (goat anti-rabbit; Cat# K4003; Agilent), rabbit anti-rat IgG antibody (Cat# BA-4001; CiteAb) were utilized for secondary antibody staining. All fluorescently labeled slides were scanned on the Vectra Polaris (Akoya Biosciences) at 40× magnification using appropriate exposure times. The data from the multispectral camera were analyzed by the imaging InForm software (Akoya Biosciences) and Phenochart software (Akoya Biosciences). ## H&E staining Small intestine segments were prepared as above (mIHC) and subjected to a FFPE technique. Slides were prepared in 4-μm sections and dewaxed with xylene (Cat# X5-4; Thermo Fisher Scientific; 3 × 10 min), rehydrated through a graded series of ethanol solutions ($100\%$ 3 × 10 min; $95\%$ 2 × 5 min; and $70\%$ 1 × 10 min), and washed in distilled water (1 × 5 min). Samples were stained with hematoxylin (1 × 10 min; Cat# SH26-500D; Thermo Fisher Scientific) and washed in distilled water again (1 × 10 min). Samples were then dipped into acid alcohol (two to five times) before being washed in distilled water (1 × 10 min). They were then dipped into sodium bicarbonate solution (Cat# BP328-1; Thermo Fisher Scientific) 20 times. Samples were then washed in water (1 × 10 min), and dehydrated through a graded series of ethanol solutions with eosin staining in the order: $70\%$ alcohol (20 dips), eosin (1 × 3 min), $95\%$ alcohol (2 × 20 dips), $100\%$ alcohol (3 × 5 min). Lastly, the slides went through xylene (4 × 5 min) and slips were covered. All slides were scanned on the Aperio AT scanning system (Leica) using appropriate exposure times. ## In vivo modeling of antibody blockade in HFHS diet fed mice In some experiments, 8-wk-old male mice received i.p. injection of anti-XCL1 antibody (25 µg per mouse; once per week; Cat# AF486; R&D Systems) to block XCL1 activity 1 d before HFHS diet feeding. IgG isotype control (25 µg per mouse; once per week; Cat# AB-108-C; R&D Systems) was utilized in the control group. In some experiments, 8-wk-old male mice received i.p. injection of anti-CD226 antibody (25 µg per mouse; once per week; Cat# 132010; Biolegend) to block XCL1 activity 1 d before HFHS diet feeding. IgG isotype control (25 µg per mouse; once per week; Cat# 400544; Biolegend) was utilized in the control group. Mice were fed a HFHS diet (Cat# D12266B; Research Diets). Following euthanasia, intraepithelial immune cells, and lamina propria immune cells were isolated for flow cytometry analysis. IECs were isolated and sorted using a FACSAria III flow cytometer (gated as DAPI-CD45.2− cells) for QPCR analysis. ## Real-time quantitative PCR (RT-QPCR) Total RNA was extracted from cells using TRIzol reagent (15596018; Invitrogen; Thermo Fisher Scientific) following the manufacturer’s instructions. iScript cDNA Synthesis Kit (Bio-Rad) was used for reverse transcription. QPCR was performed using the PowerUp SYBR Green Master Mix (Applied Biosystems) and the iCycler Real-time PCR Detection System (Bio-Rad) according to the manufacturer’s instructions. *Housekeeping* gene Ube2d2 was used as an internal control for mice cells. The relative expression of a target gene was calculated using the ΔΔCT method. All sequences of primers used in this study are listed in Table S13. ## Flow cytometry Flow cytometry was used to analyze surface marker and intracellular effector molecule expression in immune cells. Fluorochrome-conjugated monoclonal antibodies specific for mouse CD45.2 (clone 104; Cat# 109830), CD11b (clone M$\frac{1}{70}$; Cat# 101205), F$\frac{4}{80}$ (clone BM8; Cat# 123126), CD11c (clone N418; Cat# 117324), Ly-6G (clone 1A8; Cat# 127612), CD19 (clone 6D5; Cat# 115507), NK1.1 (clone PK136; Cat# 108710), I-A/I-E (clone M$\frac{5}{114.15.2}$; Cat# 107623), TCRβ (clone H57-597; Cat# 109220), TCRγ/δ (clone GL3), CD4 (clone RM4-5; Cat# 100531), CD8α (clone 53-6.7; Cat# 100732), CD8β (clone YTS156.7.7; Cat# 126605), Ki-67 (clone 16A8; Cat# 652403), SIRPα (clone P84; Cat# 144007), CD103 (clone 2E7; Cat# 121405), CD3 (clone 17A2; Cat# 100221), XCR1 (clone ZET; Cat# 148205), CD138 (clone 281-2; Cat# 142505), CD38 (clone 90; Cat# 102717), PDCA-1 (clone 927; Cat# 127015), IFNγ (clone XMG1.2; Cat# 505849), IL17A (clone TC11-18H10.1; Cat# 506915) and Granzyme B (clone QA16A02; Cat# 372204) were purchased from Biolegend. Mouse Fc Block (anti-mouse CD$\frac{16}{32}$; clone 2.4G2; Cat# 553142) was purchased from BD Biosciences. FVD eFluor 506 (Cat# 65-0866-14) was purchased from Thermo Fisher Scientific to exclude dead cells in flow cytometry. Foxp3 (clone FJK-16 s; Cat# 11-5733-82) was purchased from eBiosciences. DAPI (1:1,000 dilution) was utilized to exclude dead cells in FACS. To study cell surface marker expression, cells were stained with FVD followed by Fc blocking and surface marker staining, following standard procedures described previously (Wang et al., 2021). For T cell intracellular cytotoxicity molecule and transcription factor production, intracellular staining of Granzyme B and Ki-67 was performed directly using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Cat# 00-5523-00; Invitrogen) following the manufacturer’s instructions. These cells were co-stained with surface markers to identify CD8αα+ IEL-T cells and CD8αβ+ IEL-T cells (gated as FVD−CD45.2+CD3+CD4−CD8α+CD8β− cells and FVD−CD45.2+CD3+CD4−CD8α+CD8β+ cells in vivo). For T cell intracellular cytokine production, CD4+ T cells were stimulated with PMA (VWR; 50 ng/ml) and Ionomycin (VWR; 500 ng/ml) in the presence of GolgiStop (BD Biosciences; 4 μl per 6 ml culture media) for 4 h. At the end of the culture, cells were collected and intracellular cytokine (i.e., IFNγ and IL17A) staining was performed using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Cat# 00-5523-00; Thermo Fisher Scientific) and following the manufacturer’s instructions. These cells were co-stained with surface markers to identify CD4+ T cells (gated as FVD−CD45.2+CD3+CD8α−CD4+ cells in vivo). Stained cells were analyzed using a MACSQuant Analyzer 10 flow cytometer (Miltenyi Biotec). Data were analyzed using FlowJo software (BD Biosciences). ## FACS (fluorescence-activated cell sorting) Single cell suspensions were harvested from intestines of chow diet-fed and HFHS diet-fed mice and then pooled for further sorting (10 mice were combined for each group). To isolate different cell populations, cells were treated with Fc blocking followed by surface marker staining, following a standard procedure (Wang et al., 2021). DAPI (1:1,000 dilution) was added to single cell suspensions 5 min before sorting to exclude dead cells. Cells were then sorted using a FACSAria III flow cytometer to purify intraepithelial and lamina propria hematopoietic cells (gated as DAPI−CD45.2+ cells) and epithelial structural cells (gated as DAPI−CD45.2− cells). Sorted live cells were counted and immediately utilized for library construction and sequencing. To reduce the potential batch effects in experimental design and handling, 10 mice were sacrificed at the same time for each sample, and all samples were sent to QC and library construction at the same time. ## Library preparation, sequencing, and alignment Cells were stained with trypan blue (Cat# T10282; Thermo Fisher Scientific) and counted using a Cell Countess II automated cell counter (Thermo Fisher Scientific). Libraries were constructed using a Chromium Single Cell 3′ library (10× Genomics) & Gel Bead Kit V2 (10× Genomics, Cat# PN-120237) according to the manufacturer’s instructions. Libraries were sequenced using the NovaSeq 6000 S2 Reagent Kit (Illumina) to a depth around 2 billion reads per library using 2 × 50 read length. Data analysis was performed using a Cellranger Software Suite (10× Genomics). Individual samples were extracted from the sequencer and used as inputs for the Cellranger pipeline to generate the merged digital expression matrix using Cellranger aggregation. The raw data of all samples was output by Cellranger at the same time after the aggregation process. We followed the effective design of single-cell gene expression studies to reduce the potential batch effects (Luecken and Theis, 2019). ## Cell clustering and annotation The merged digital expression matrix generated by Cellranger was analyzed using Seurat (v.4.0.0) following the guidelines. Seurat is an R package developed in 2017 by the Satija lab for single cell RNA sequencing (Butler et al., 2018). We also used the recommended methods for batch integration (Tran et al., 2020). Specifically, cells were first filtered to have at least 100 unique molecular identifiers, at least 100 genes, at most $10\%$ mitochondrial gene expression for intraepithelial and lamina propria immune cells, and at most $50\%$ for IECs. *The* gene counts for each cell were divided by the total gene counts for the cell and multiplied by a scale factor of 10,000, then normalized using the Seurat function NormalizeData through natural-log transformation. Normalization steps were employed to eliminate technical noise or bias so that observed variance in gene expression variance primarily reflects true biological variance. This normalization targets variance from sequencing (library preparation, high dropout event, amplification bias caused by gene length GC content, etc.; Jia et al., 2017). Next, variable genes were found using the Seurat function FindVariableGenes. The ScaleData function was used to regress out the sequencing depth for each cell. *Variable* genes that had been previously identified were used in principal component analysis (PCA) to reduce the dimensions of the data. Following this, 50 principal components were used in UMAP to further reduce the dimensions to 2. The same 50 principal components were also used to group the cells into different clusters by the Seurat function FindClusters. Next, cluster marker genes were found for each cluster using the FindAllMarkers function. Cell types were manually annotated based on the cluster markers. Module scores were calculated using the AddModuleScore function for certain functions and metabolic processes. To calculate the total sample composition based on cell types, the number of cells for each cell type was counted. The counts were then divided by the total number of cells and scaled to $100\%$ for each cell type. For the proportion of cell subclusters, the number of cells for each cell subcluster was counted and then divided by the total number of that cell type before scaling to $100\%$. For potential upstream regulator analyses, differential expression analysis was carried out between cell subclusters. Then IPA was applied to the DEGs to determine the potential upstream regulators driving the differential expression. ## CytoTRACE analysis *The* gene expression matrix for certain cell types was extracted with single cell IDs and the gene names. The databases were uploaded to the CytoTRACE website (https://cytotrace.stanford.edu/) to predict differentiation states. CytoTRACE score of each cell was calculated and integrated back into the scRNA-seq database and plotted on the UMAP afterward. 3D graphics were generated using the CytoTRACE website features. 2D graphics were generated utilizing the FeaturePlot function. ## Pseudotime trajectory Pseudotime trajectories were analyzed using the R package Monocle3 (v.2.18.0; Trapnell et al., 2014). After clustering analysis, the data dimensionality of the intended cell types were reduced using UMAP. Monocle was utilized to learn trajectory through cluster_cells function by the default method. Next, we fit a principal graph within each partition using the learn_graph function and trajectories with numerous branches were reconstructed. After specifying the root nodes of the intended trajectory, we used the order_cells function to calculate where each cell falls in pseudotime of the biological process. Finally, the plot_cells function was utilized to show the trajectory graph in UMAP and to color cells by pseudotime. Heatmaps showing the bifurcation expression patterns were generated using function plot_genes_branched_heatmap with DEGs with the FDR adjusted P value <1e−50. ## RNA-velocity analysis To estimate the RNA velocity of samples, velocyto (La Manno et al., 2018) was used to distinguish unspliced and spliced messenger RNAs in each cell and to recover the directed dynamic information by leveraging mRNA-splicing information. Specifically,. LOOM files with spliced/unspliced expression matrices were generated for each sample. The Velocyto. R-package (v0.6) was utilized for further analysis. After loading. LOOM files information through ReadVelocity function, databases were merged and RunVelocity function was performed to obtain the velocity vectors. Finally, the velocities were projected into a lower-dimensional embedding using the velocity_graph function and visualized on the UMAP embedding in each intended cell cluster using the show.velocity.on.embedding.cor function. All velocyto functions were used with default parameters. ## Cell–cell ligand–receptor interaction CellphoneDB (v.2.1.7), CellTalkDB (v. 0.0.1.6), and CellChat (v.1.1.3) were utilized for ligand–receptor analysis. The raw gene expression counts and cell type annotation for each single cell were analyzed with the CellphoneDB database to determine the potential ligand–receptor pairs and filtered with a threshold of P-value below 0.05. Four groups of major cell types: intraepithelial lymphocytes, lamina propria lymphocytes, lamina propria myeloid cells (accumulated with additional cohort), and structural cells from intestine epithelium were merged and two runs were performed on chow diet intestinal cells and HFHS diet intestinal cells, respectively. The number of interactions between each pair of cell types was plotted based on source and receptor using heatmap and circle interactome. Selected ligand–receptor pairs were plotted in dot plot to indicate expression patterns across all cell types. All ligand–receptor pairs exhibited were manually verified via CellTalkDB, a manually curated database of literature-supported ligand–receptor interactions (http://tcm.zju.edu.cn/celltalkdb/). The same intestinal cell populations were also analyzed with CellChat for systems-level analysis and better interpretation of cell–cell communication. To predict functions for poorly studied pathways, the computeNetSimilarity function was utilized for pathway analysis and netVisual_embedding function was utilized to visualize pathway clustering in 2D plot. To discover major signaling changes in response to diet, the rankNet function was utilized to compare the information flow change of each signaling pathway within the network. To predict and visualize key signaling events between cell populations, netVisual_aggregate function was utilized to visualize the signaling network of NECTIN–CD226 using immune cells and structural cells as receivers, respectively. ## Functional enrichment analysis with g:GOSt g:GOSt analysis was performed on g:Profiler web server (https://biit.cs.ut.ee/gprofiler/gost; Raudvere et al., 2019). Common upstream regulators from HFHS diet intestine were input into the functional enrichment analysis query and then analyzed using g:GOSt functional profiling. Top differentially regulated signaling pathways were shown. Upstream regulators were considered common if implicated in 10 or more cell types from a lineage. Terms were ranked by a normalized enrichment score. ## Statistical analysis GraphPad Prism 6 (GraphPad Software) was used for the graphic representation and statistical analysis of the data. All data were presented as the mean ± SEM. A two-tailed Student’s t test was used for comparison between groups. Multiple comparisons were performed using a two-way ANOVA followed by Sidak’s multiple comparisons test. $P \leq 0.05$ was considered statistically significant. ns, not significant; *$P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$; ****$P \leq 0.0001.$ ## Online supplemental material Fig. S1 shows lineage-associated gene signatures of all CD45.2+ intestinal cells and intestine T cell clusters from the chow diet and HFHS diet intestine. Fig. S2 shows H&E staining of small intestines, gating strategy of scRNA-seq-defined CD45.2+ intestinal cell populations, and viability test in flow cytometry analysis. Fig. S3 shows the flow cytometry gating strategy, signature genes, viability test, and enrichment analysis of the unique T cell subsets accumulate in HFHS diet intestine. Fig. S4 shows the analysis of the macrophage populations and the comparison of key genes related to intestine barrier function and nutrient absorption function between chow diet and HFHS diet intestine. Fig. S5 shows the comparison between chow diet and HFHS diet intestine interactomes and antibody blockade experiments in vivo. Table S1 shows table of the most differentially expressed genes (DEGs) for CD8αα+ IEL-T cells of HFHS diet vs. chow diet. Table S2 shows table of the most DEGs for CD8αβ+ IEL-T cells of HFHS diet vs. chow diet. Table S3 shows the table of the most DEGs for macrophage subsets. Table S4 shows gene lists utilized to create module scores for enterocyte of villus top. Table S5 shows gene lists utilized to create module scores for enterocyte of villus bottom. Table S6 shows count of interacting ligand–receptor pairs in chow diet mice intestine interaction network. Table S7 shows the count of interacting ligand–receptor pairs in HFHS diet mice intestine interaction network. Table S8 shows table of the DEGs between HFHS diet and chow diet intestine for each cell type. Table S9 shows IPA analysis of upstream regulators of each intestine cell type. Table S10 shows the count of cell types shared by the upstream regulators of IPA analyses. Table S11 is a list of upstream regulators implicated in mouse and human studies that have been associated with immune homeostasis and obesity. Table S12 shows the top differentially regulated signaling pathways in response to HFHS diet suggested in g:GOSt functional enrichment analysis of common upstream regulators. Table S13 lists primers used for QPCR. ## Data availability All data associated with this study are presented in the article or supplemental tables. The genomics data generated during this study is available through GEO accession GSE221006. Further information and requests for new reagents generated in this study may be directed to and will be fulfilled by the lead contact, Aldons Lusis. ## References 1. Adriaenssens A., Lam B.Y., Billing L., Skeffington K., Sewing S., Reimann F., Gribble F.. **A transcriptome-led exploration of molecular mechanisms regulating somatostatin-producing D-cells in the gastric epithelium**. *Endocrinology* (2015) **156** 3924-3936. DOI: 10.1210/en.2015-1301 2. Albonici L., Giganti M.G., Modesti A., Manzari V., Bei R.. **Multifaceted role of the placental growth factor (PlGF) in the antitumor immune response and cancer progression**. *Int. J. Mol. Sci.* (2019) **20** 20. DOI: 10.3390/ijms20122970 3. Allaire J.M., Crowley S.M., Law H.T., Chang S.Y., Ko H.J., Vallance B.A.. **The intestinal epithelium: Central coordinator of mucosal immunity**. *Trends Immunol.* (2018) **39** 677-696. DOI: 10.1016/j.it.2018.04.002 4. Arnone D., Vallier M., Hergalant S., Chabot C., Ndiaye N.C., Moulin D., Aignatoaei A.M., Alberto J.M., Louis H., Boulard O.. **Long-term overconsumption of fat and sugar causes a partially reversible pre-inflammatory bowel disease state**. *Front. Nutr.* (2021) **8** 758518. DOI: 10.3389/fnut.2021.758518 5. Basson A.R., Chen C., Sagl F., Trotter A., Bederman I., Gomez-Nguyen A., Sundrud M.S., Ilic S., Cominelli F., Rodriguez-Palacios A.. **Regulation of intestinal inflammation by dietary fats**. *Front. Immunol.* (2021) **11** 604989. DOI: 10.3389/fimmu.2020.604989 6. Bomidi C., Robertson M., Coarfa C., Estes M.K., Blutt S.E.. **Single-cell sequencing of rotavirus-infected intestinal epithelium reveals cell-type specific epithelial repair and tuft cell infection**. *Proc. Natl. Acad. Sci. USA* (2021) **118** 118. DOI: 10.1073/pnas.2112814118 7. Böttcher J.P., Bonavita E., Chakravarty P., Blees H., Cabeza-Cabrerizo M., Sammicheli S., Rogers N.C., Sahai E., Zelenay S., Reis e Sousa C.. **NK cells stimulate recruitment of cDC1 into the tumor microenvironment promoting cancer immune control**. *Cell* (2018) **172** 1022-1037.e14. DOI: 10.1016/j.cell.2018.01.004 8. Brucklacher-Waldert V., Carr E.J., Linterman M.A., Veldhoen M.. **Cellular plasticity of CD4**. *Front. Immunol.* (2014) **5** 488. DOI: 10.3389/fimmu.2014.00488 9. Butler A., Hoffman P., Smibert P., Papalexi E., Satija R.. **Integrating single-cell transcriptomic data across different conditions, technologies, and species**. *Nat. Biotechnol.* (2018) **36** 411-420. DOI: 10.1038/nbt.4096 10. Cabeza-Cabrerizo M., Cardoso A., Minutti C.M., Pereira da Costa M., Reis e Sousa C.. **Dendritic cells revisited**. *Annu. Rev. Immunol.* (2021) **39** 131-166. DOI: 10.1146/annurev-immunol-061020-053707 11. Cader M.Z., Kaser A.. **Recent advances in inflammatory bowel disease: Mucosal immune cells in intestinal inflammation**. *Gut* (2013) **62** 1653-1664. DOI: 10.1136/gutjnl-2012-303955 12. Chassaing B., Kumar M., Baker M.T., Singh V., Vijay-Kumar M.. **Mammalian gut immunity**. *Biomed. J.* (2014) **37** 246-258. DOI: 10.4103/2319-4170.130922 13. Cheroutre H., Lambolez F., Mucida D.. **The light and dark sides of intestinal intraepithelial lymphocytes**. *Nat. Rev. Immunol.* (2011) **11** 445-456. DOI: 10.1038/nri3007 14. Chu H.Y., Chen Z., Wang L., Zhang Z.K., Tan X., Liu S., Zhang B.T., Lu A., Yu Y., Zhang G.. **Dickkopf-1: A promising target for cancer immunotherapy**. *Front. Immunol.* (2021) **12** 658097. DOI: 10.3389/fimmu.2021.658097 15. Delgoffe G.M., Woo S.R., Turnis M.E., Gravano D.M., Guy C., Overacre A.E., Bettini M.L., Vogel P., Finkelstein D., Bonnevier J.. **Stability and function of regulatory T cells is maintained by a neuropilin-1-semaphorin-4a axis**. *Nature* (2013) **501** 252-256. DOI: 10.1038/nature12428 16. Dinarello C.A., Nold-Petry C., Nold M., Fujita M., Li S., Kim S., Bufler P.. **Suppression of innate inflammation and immunity by interleukin-37**. *Eur. J. Immunol.* (2016) **46** 1067-1081. DOI: 10.1002/eji.201545828 17. Efremova M., Vento-Tormo M., Teichmann S.A., Vento-Tormo R.. **CellPhoneDB: Inferring cell-cell communication from combined expression of multi-subunit ligand-receptor complexes**. *Nat. Protoc.* (2020) **15** 1484-1506. DOI: 10.1038/s41596-020-0292-x 18. Elmentaite R., Kumasaka N., Roberts K., Fleming A., Dann E., King H.W., Kleshchevnikov V., Dabrowska M., Pritchard S., Bolt L.. **Cells of the human intestinal tract mapped across space and time**. *Nature* (2021) **597** 250-255. DOI: 10.1038/s41586-021-03852-1 19. Elmentaite R., Ross A.D.B., Roberts K., James K.R., Ortmann D., Gomes T., Nayak K., Tuck L., Pritchard S., Bayraktar O.A.. **Single-cell sequencing of developing human gut reveals transcriptional links to childhood Crohn’s disease**. *Dev. Cell* (2020) **55** 771-783.e5. DOI: 10.1016/j.devcel.2020.11.010 20. Esterházy D., Loschko J., London M., Jove V., Oliveira T.Y., Mucida D.. **Classical dendritic cells are required for dietary antigen-mediated induction of peripheral T**. *Nat. Immunol.* (2016) **17** 545-555. DOI: 10.1038/ni.3408 21. Gulati G.S., Sikandar S.S., Wesche D.J., Manjunath A., Bharadwaj A., Berger M.J., Ilagan F., Kuo A.H., Hsieh R.W., Cai S.. **Single-cell transcriptional diversity is a hallmark of developmental potential**. *Science* (2020) **367** 405-411. DOI: 10.1126/science.aax0249 22. Haber A.L., Biton M., Rogel N., Herbst R.H., Shekhar K., Smillie C., Burgin G., Delorey T.M., Howitt M.R., Katz Y.. **A single-cell survey of the small intestinal epithelium**. *Nature* (2017) **551** 333-339. DOI: 10.1038/nature24489 23. Haseeb M., Pirzada R.H., Ain Q.U., Choi S.. **Wnt signaling in the regulation of immune cell and cancer therapeutics**. *Cells* (2019) **8** 8. DOI: 10.3390/cells8111380 24. Huang Z., Qi G., Miller J.S., Zheng S.G.. **CD226: An emerging role in immunologic diseases**. *Front. Cell Dev. Biol.* (2020) **8** 564. DOI: 10.3389/fcell.2020.00564 25. Hübel J., Hieronymus T.. **HGF/Met-signaling contributes to immune regulation by modulating tolerogenic and motogenic properties of dendritic cells**. *Biomedicines* (2015) **3** 138-148. DOI: 10.3390/biomedicines3010138 26. Jia C., Hu Y., Kelly D., Kim J., Li M., Zhang N.R.. **Accounting for technical noise in differential expression analysis of single-cell RNA sequencing data**. *Nucleic Acids Res.* (2017) **45** 10978-10988. DOI: 10.1093/nar/gkx754 27. Jin S., Guerrero-Juarez C.F., Zhang L., Chang I., Ramos R., Kuan C.H., Myung P., Plikus M.V., Nie Q.. **Inference and analysis of cell-cell communication using CellChat**. *Nat. Commun.* (2021) **12** 1088. DOI: 10.1038/s41467-021-21246-9 28. Kadivar M., Petersson J., Svensson L., Marsal J.. **CD8αβ**. *J. Immunol.* (2016) **197** 4584-4592. DOI: 10.4049/jimmunol.1601146 29. Keum N., Giovannucci E.. **Global burden of colorectal cancer: Emerging trends, risk factors and prevention strategies**. *Nat. Rev. Gastroenterol. Hepatol.* (2019) **16** 713-732. DOI: 10.1038/s41575-019-0189-8 30. La Manno G., Soldatov R., Zeisel A., Braun E., Hochgerner H., Petukhov V., Lidschreiber K., Kastriti M.E., Lönnerberg P., Furlan A.. **RNA velocity of single cells**. *Nature* (2018) **560** 494-498. DOI: 10.1038/s41586-018-0414-6 31. Lefrancois L., Lycke N.. **Isolation of mouse small intestinal intraepithelial lymphocytes, Peyer’s patch, and lamina propria cells**. *Curr. Protoc. Immunol.* (2001) **3** 19. DOI: 10.1002/0471142735.im0319s17 32. Li S., Neff C.P., Barber K., Hong J., Luo Y., Azam T., Palmer B.E., Fujita M., Garlanda C., Mantovani A.. **Extracellular forms of IL-37 inhibit innate inflammation in vitro and in vivo but require the IL-1 family decoy receptor IL-1R8**. *Proc. Natl. Acad. Sci. USA* (2015) **112** 2497-2502. DOI: 10.1073/pnas.1424626112 33. Liu T.C., Kern J.T., Jain U., Sonnek N.M., Xiong S., Simpson K.F., VanDussen K.L., Winkler E.S., Haritunians T., Malique A.. **Western diet induces Paneth cell defects through microbiome alterations and farnesoid X receptor and type I interferon activation**. *Cell Host Microbe* (2021) **29** 988-1001.e6. DOI: 10.1016/j.chom.2021.04.004 34. Luecken M.D., Theis F.J.. **Current best practices in single-cell RNA-seq analysis: A tutorial**. *Mol. Syst. Biol.* (2019) **15**. DOI: 10.15252/msb.20188746 35. Makita S., Takatori H., Nakajima H.. **Post-transcriptional regulation of immune responses and inflammatory diseases by RNA-binding ZFP36 family proteins**. *Front. Immunol.* (2021) **12** 711633. DOI: 10.3389/fimmu.2021.711633 36. McDonald B.D., Jabri B., Bendelac A.. **Diverse developmental pathways of intestinal intraepithelial lymphocytes**. *Nat. Rev. Immunol.* (2018) **18** 514-525. DOI: 10.1038/s41577-018-0013-7 37. Moor A.E., Harnik Y., Ben-Moshe S., Massasa E.E., Rozenberg M., Eilam R., Bahar Halpern K., Itzkovitz S.. **Spatial reconstruction of single enterocytes uncovers broad zonation along the intestinal villus axis**. *Cell* (2018) **175** 1156-1167.e15. DOI: 10.1016/j.cell.2018.08.063 38. Mowat A.M., Agace W.W.. **Regional specialization within the intestinal immune system**. *Nat. Rev. Immunol.* (2014) **14** 667-685. DOI: 10.1038/nri3738 39. Mukherjee P., Chattopadhyay A., Grijalva V., Dorreh N., Lagishetty V., Jacobs J.P., Clifford B.L., Vallim T., Mack J.J., Navab M.. **Oxidized phospholipids cause changes in jejunum mucus that induce dysbiosis and systemic inflammation**. *J. Lipid Res.* (2022) **63** 100153. DOI: 10.1016/j.jlr.2021.100153 40. Nakanishi T., Fukui H., Wang X., Nishiumi S., Yokota H., Makizaki Y., Tanaka Y., Ohno H., Tomita T., Oshima T., Miwa H.. **Effect of a high-fat diet on the small-intestinal environment and mucosal integrity in the gut-liver axis**. *Cells* (2021) **10** 10. DOI: 10.3390/cells10113168 41. Nestorowa S., Hamey F.K., Pijuan Sala B., Diamanti E., Shepherd M., Laurenti E., Wilson N.K., Kent D.G., Göttgens B.. **A single-cell resolution map of mouse hematopoietic stem and progenitor cell differentiation**. *Blood* (2016) **128** e20-e31. DOI: 10.1182/blood-2016-05-716480 42. Olivares-Villagómez D., Van Kaer L.. **Intestinal intraepithelial lymphocytes: Sentinels of the mucosal barrier**. *Trends Immunol.* (2018) **39** 264-275. DOI: 10.1016/j.it.2017.11.003 43. Peterson L.W., Artis D.. **Intestinal epithelial cells: Regulators of barrier function and immune homeostasis**. *Nat. Rev. Immunol.* (2014) **14** 141-153. DOI: 10.1038/nri3608 44. Pong Ng H., Kim G.D., Ricky Chan E., Dunwoodie S.L., Mahabeleshwar G.H.. **CITED2 limits pathogenic inflammatory gene programs in myeloid cells**. *FASEB J.* (2020) **34** 12100-12113. DOI: 10.1096/fj.202000864R 45. Qiu Z., Sheridan B.S.. **Isolating lymphocytes from the mouse small intestinal immune system**. *J. Vis. Exp.* (2018) 57281. PMID: 29553537 46. Raudvere U., Kolberg L., Kuzmin I., Arak T., Adler P., Peterson H., Vilo J.. **g:Profiler: A web server for functional enrichment analysis and conversions of gene lists (2019 update)**. *Nucleic Acids Res.* (2019) **47** W191-W198. DOI: 10.1093/nar/gkz369 47. Rescigno M.. **Intestinal microbiota and its effects on the immune system**. *Cell. Microbiol.* (2014) **16** 1004-1013. DOI: 10.1111/cmi.12301 48. Riera Romo M., Pérez-Martínez D., Castillo Ferrer C.. **Innate immunity in vertebrates: An overview**. *Immunology* (2016) **148** 125-139. DOI: 10.1111/imm.12597 49. Rooks M.G., Garrett W.S.. **Gut microbiota, metabolites and host immunity**. *Nat. Rev. Immunol.* (2016) **16** 341-352. DOI: 10.1038/nri.2016.42 50. Ruscher R., Kummer R.L., Lee Y.J., Jameson S.C., Hogquist K.A.. **CD8αα intraepithelial lymphocytes arise from two main thymic precursors**. *Nat. Immunol.* (2017) **18** 771-779. DOI: 10.1038/ni.3751 51. Shao X., Liao J., Li C., Lu X., Cheng J., Fan X.. **CellTalkDB: A manually curated database of ligand-receptor interactions in humans and mice**. *Brief. Bioinform.* (2021) **22** 22. DOI: 10.1093/bib/bbaa269 52. Sun M., He C., Cong Y., Liu Z.. **Regulatory immune cells in regulation of intestinal inflammatory response to microbiota**. *Mucosal Immunol.* (2015) **8** 969-978. DOI: 10.1038/mi.2015.49 53. Sundaram T.S., Giromini C., Rebucci R., Pistl J., Bhide M., Baldi A.. **Role of omega-3 polyunsaturated fatty acids, citrus pectin, and milk-derived exosomes on intestinal barrier integrity and immunity in animals**. *J. Anim. Sci. Biotechnol.* (2022) **13** 40. DOI: 10.1186/s40104-022-00690-7 54. Sydora B.C., Brossay L., Hagenbaugh A., Kronenberg M., Cheroutre H.. **TAP-independent selection of CD8**. *J. Immunol.* (1996) **156** 4209-4216. DOI: 10.4049/jimmunol.156.11.4209 55. Tahara-Hanaoka S., Shibuya K., Onoda Y., Zhang H., Yamazaki S., Miyamoto A., Honda S., Lanier L.L., Shibuya A.. **Functional characterization of DNAM-1 (CD226) interaction with its ligands PVR (CD155) and nectin-2 (PRR-2/CD112)**. *Int. Immunol.* (2004) **16** 533-538. DOI: 10.1093/intimm/dxh059 56. Tran H.T.N., Ang K.S., Chevrier M., Zhang X., Lee N.Y.S., Goh M., Chen J.. **A benchmark of batch-effect correction methods for single-cell RNA sequencing data**. *Genome Biol.* (2020) **21** 12. DOI: 10.1186/s13059-019-1850-9 57. Trapnell C., Cacchiarelli D., Grimsby J., Pokharel P., Li S., Morse M., Lennon N.J., Livak K.J., Mikkelsen T.S., Rinn J.L.. **The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells**. *Nat. Biotechnol.* (2014) **32** 381-386. DOI: 10.1038/nbt.2859 58. Varol C., Zigmond E., Jung S.. **Securing the immune tightrope: Mononuclear phagocytes in the intestinal lamina propria**. *Nat. Rev. Immunol.* (2010) **10** 415-426. DOI: 10.1038/nri2778 59. Wang Y., Song W., Wang J., Wang T., Xiong X., Qi Z., Fu W., Yang X., Chen Y.G.. **Single-cell transcriptome analysis reveals differential nutrient absorption functions in human intestine**. *J. Exp. Med.* (2020) **217** 217. DOI: 10.1084/jem.20191130 60. Wang Y.C., Wang X., Yu J., Ma F., Li Z., Zhou Y., Zeng S., Ma X., Li Y.R., Neal A.. **Targeting monoamine oxidase A-regulated tumor-associated macrophage polarization for cancer immunotherapy**. *Nat. Commun.* (2021) **12** 3530. DOI: 10.1038/s41467-021-23164-2 61. Ward Z.J., Bleich S.N., Cradock A.L., Barrett J.L., Giles C.M., Flax C., Long M.W., Gortmaker S.L.. **Projected U.S. state-level prevalence of adult obesity and severe obesity**. *N. Engl. J. Med.* (2019) **381** 2440-2450. DOI: 10.1056/NEJMsa1909301 62. Welty N.E., Staley C., Ghilardi N., Sadowsky M.J., Igyártó B.Z., Kaplan D.H.. **Intestinal lamina propria dendritic cells maintain T cell homeostasis but do not affect commensalism**. *J. Exp. Med.* (2013) **210** 2011-2024. DOI: 10.1084/jem.20130728 63. Worbs T., Hammerschmidt S.I., Förster R.. **Dendritic cell migration in health and disease**. *Nat. Rev. Immunol.* (2017) **17** 30-48. DOI: 10.1038/nri.2016.116 64. Yeo J., Ko M., Lee D.H., Park Y., Jin H.S.. **TIGIT/CD226 Axis regulates anti-tumor immunity**. *Pharmaceuticals* (2021) **14** 14. DOI: 10.3390/ph14030200 65. Zietek T., Giesbertz P., Ewers M., Reichart F., Weinmüller M., Urbauer E., Haller D., Demir I.E., Ceyhan G.O., Kessler H., Rath E.. **Organoids to study intestinal nutrient transport, drug uptake and metabolism: Update to the human model and expansion of applications**. *Front. Bioeng. Biotechnol.* (2020) **8** 577656. DOI: 10.3389/fbioe.2020.577656
--- title: Aberrations in the early pregnancy serum metabolic profile in women with prediabetes at two years postpartum authors: - Ella Muhli - Chouaib Benchraka - Mrunalini Lotankar - Noora Houttu - Harri Niinikoski - Leo Lahti - Kirsi Laitinen journal: Metabolomics year: 2023 pmcid: PMC10038958 doi: 10.1007/s11306-023-01994-z license: CC BY 4.0 --- # Aberrations in the early pregnancy serum metabolic profile in women with prediabetes at two years postpartum ## Abstract ### Introduction Aberrations in circulating metabolites have been associated with diabetes and cardiovascular risk. ### Objectives To investigate if early and late pregnancy serum metabolomic profiles differ in women who develop prediabetes by two years postpartum compared to those who remain normoglycemic. ### Methods An NMR metabolomics platform was used to measure 228 serum metabolite variables from women with pre-pregnancy overweight in early and late pregnancy. Co-abundant groups of metabolites were compared between the women who were ($$n = 40$$) or were not ($$n = 138$$) prediabetic at two years postpartum. Random Forests classifiers, based on the metabolic profiles, were used to predict the prediabetes status, and correlations of the metabolites to glycemic traits (fasting glucose and insulin, HOMA2-IR and HbA1c) and hsCRP at postpartum were evaluated. ### Results Women with prediabetes had higher concentrations of small HDL particles, total lipids in small HDL, phospholipids in small HDL and free cholesterol in small HDL in early pregnancy ($$p \leq 0.029$$; adj with pre-pregnancy BMI $$p \leq 0.094$$). The small HDL related metabolites also correlated positively with markers of insulin resistance at postpartum. Similar associations were not detected for metabolites in late pregnancy. A Random Forests classifier based on serum metabolites and clinical variables in early pregnancy displayed an acceptable predictive power for the prediabetes status at postpartum (AUROC 0.668). ### Conclusion Elevated serum concentrations of small HDL particles in early pregnancy associate with prediabetes and insulin resistance at two years postpartum. The serum metabolic profile during pregnancy might be used to identify women at increased risk for type 2 diabetes. ### Supplementary Information The online version contains supplementary material available at 10.1007/s11306-023-01994-z. ## Introduction Maternal metabolism changes during pregnancy to meet the demands of the mother and the feto-placental unit (Lain & Catalano, 2007). Aberrations in these changes are associated with pregnancy complications such as gestational diabetes (GDM) (Kivelä et al., 2021; White et al., 2017), which predisposes the mother herself to subsequent type 2 diabetes and her offspring to obesity in later life (Hod et al., 2015). Prepregnancy overweight is a well-established risk factor for GDM. It has previously been demonstrated, in a cohort of women with overweight, that the serum metabolic profile of women developing GDM differs from those who remain normoglycemic already in early pregnancy (Mokkala et al., 2020). However, thus far, there has been rather little published information on the extent to which the metabolic profile during pregnancy and its potential aberrations influence the onset of diabetes postpartum. Our working hypothesis was that by undertaking comprehensive examination of metabolic profiles, in addition to traditional metabolic markers, it could be possible to elucidate the associations of circulating metabolites during pregnancy to postpartum metabolic disorders and thus to reveal potential targets for interventions. With respect to the traditional metabolic markers, high third trimester glycated hemoglobin (HbA1c) levels at least 36 mmol/mol ($5.4\%$) have been associated with an increased risk of diabetes mellitus in women with GDM from pregnancy up to five years postpartum (Claesson et al., 2017; Varejão et al., 2021). Elevated high-sensitivity C-reactive protein (hsCRP) levels during mid-pregnancy have also been associated with dysglycemia during the first postpartum year (Durnwald et al., 2018; Ozuguz et al., 2011). In non-pregnant populations, certain distinct serum metabolites, such as the levels of branched-chain amino acids (BCAAs), as well as those of phenylalanine, glutamate and several lipids, have been associated with an elevated risk of type 2 diabetes (Long et al., 2020) and cardiovascular events (Ruiz-Canela et al., 2017). The association between the serum metabolic profile during pregnancy and a glucose metabolism disorder at postpartum has previously been described in only one publication (Liu et al., 2021); it was reported that fasting serum levels of BCAAs valine, leucine and isoleucine, acylcarnitine C2 and 3-hydroxybutyrate measured at 28 weeks of gestation were associated with prediabetes or type 2 diabetes 10 to 14 years later. We wanted to investigate the associations of early and late pregnancy serum metabolic profiles to the prediabetes status at two years’ postpartum in an at-risk cohort of women who had overweight before becoming pregnant. We hypothesized that the serum metabolic profiles both in early and late pregnancy would differ between the women with and without prediabetes at two years’ postpartum. The first aim of the study was to investigate the differences in serum metabolites in both early and late pregnancy between the women who later developed prediabetes or remained healthy. We also aimed to investigate if the serum metabolites during pregnancy could be used to predict prediabetes, and furthermore we evaluated the associations of the serum metabolites with glycemic traits at postpartum. ## Participants and study design This study is a follow-up study of women participating in a single-center dietary intervention trial during pregnancy (Pellonperä et al., 2019) (ClinicalTrials.gov: NCT01922791). Briefly, the trial investigated the effect of dietary intervention with fish oil and/or probiotics on maternal and offspring health. The main outcomes were the incidence of GDM and allergy in the offspring. The inclusion criteria were overweight (self-reported prepregnancy BMI ≥ 25 kg/m2) and early pregnancy (< 18 weeks of gestation). The exclusion criteria were GDM diagnosed during the current pregnancy, multifetal pregnancy, and metabolic or inflammatory disease, such as type 1 or type 2 diabetes, celiac disease, or inflammatory bowel disease. A total of 439 women were recruited to the intervention trial. Here, we examined 178 women from whom fasting serum samples in early and/or late pregnancy and fasting blood glucose analyzed for diagnosis of prediabetes at two years postpartum were available. The early pregnancy serum metabolomics analysis was available for 174 of the women and the late pregnancy serum metabolomics analysis for 169 of the women. We excluded the women who used GDM medication (metformin, insulin or both; $$n = 10$$) from the late pregnancy analyses. The women participated in two study visits during pregnancy, in early pregnancy at a mean 13.9 weeks of gestation (SD 2.0 weeks) and in late pregnancy at a mean 35.1 weeks of gestation (SD 0.9 weeks), and in the follow-up visit at two years’ postpartum (mean 2.0 years, SD 0.04 years). The clinical characteristics of the women were inquired by questionnaires. The intakes of energy and macronutrients were calculated from 3-day food diaries filled in near to the study visits using computerized software (AivoDiet 2.0.2.3, Aivo, Turku, Finland), which utilizes the Finnish Food Composition Database Fineli (Fineli). Blood pressure was measured during the study visits with Omron M5-1 (Intelli™ sense, Omron Matsusaka Co., Ltd, Japan). Height was measured to the nearest 0.1 cm with a wall stadiometer at the early pregnancy study visit. Pre-pregnancy BMI was calculated from self-reported weight, obtained from welfare clinic records, and the height measured in early pregnancy. Mean weekly weight gain between early and late pregnancy was calculated from the weights measured at the study visits. A standard 2-hour 75-g OGTT for pregnant women was performed (Working group set up by the Finnish Medical Society Duodecim, the Medical Advisory Board of the Finnish Diabetes Association and the Finnish Gynecological Association, 2013) and diagnosis of GDM was based on at least one value at or above the threshold levels: 0 h ≥ 5.3, 1 h ≥ 10.0 and 2 h ≥ 8.6 mmol/l. Prediabetes at two years postpartum was determined according to the criteria issued by the American Diabetes Association (ADA) as fasting plasma glucose from 5.6 to 6.9 mmol/l (American Diabetes Association, 2011). Of the study participants, 40 had prediabetes; two of them reported having type 2 diabetes. ## Blood sampling and analysis Fasting (9 h minimum) blood samples were drawn from the antecubital vein. The serum was separated and frozen in aliquots at -80 degrees Celsius. The serum metabolic profile was analysed using a high-throughput proton NMR metabolomics platform (Nightingale Health Ltd, Helsinki, Finland) as previously described (Soininen et al., 2015). The platform evaluates 228 metabolites and their ratios, including biomarkers of lipid and glucose metabolism, amino acids, ketone bodies and glycoprotein acetyls (GlycA), a marker of low-grade inflammation. Other sample analyses were assayed in a certified laboratory (Tykslab, the Hospital District of Southwest Finland) immediately after blood sampling. Fasting glucose was measured with an enzymatic method using hexokinase and fasting insulin with an immunoelectrochemiluminometric assay. HbA1c was determined with ion-exchange HPLC. An automated colorimetric immunoassay was used to measure hsCRP. Homeostatic model assessment for insulin resistance (HOMA2-IR) was calculated from fasting glucose and fasting insulin levels (Wallace et al., 2004). ## Statistical analysis The statistical analysis for the clinical characteristics data of the women was made with IBM SPSS Statistics 28.0 for Windows (IBM SPSS, Chicago, IL, USA). The normality of distributions was visually observed from histograms and evaluated using Shapiro-Wilk’s test. Deviations from normality were assumed when Shapiro-Wilk’s test $p \leq 0.05.$ The homogeneity of variances was evaluated with Levene’s test ($p \leq 0.05$ indicating violation of this assumption). Normally distributed continuous variables are summarized with means and standard deviations and non-normally distributed continuous variables with medians and interquartile ranges. Categorical data are presented as frequencies and percentages. Differences in the clinical characteristics were evaluated with the Student’s t-test for normally distributed continuous variables and with the Mann-Whitney U-test for non-normally distributed continuous variables. Pearson chi-square test was used for evaluating differences in categorical variables between the groups. Two-sided p-values < 0.05 were considered significant. Analysis of the metabolomic data was carried out using the R version 4.1.0. In the early pregnancy data, 42 metabolites had missing values; these were imputed using randomly sampled values from the available data of each metabolite independently. Principal Component Analysis (PCA) was done using the calculatePCA function from scater package (McCarthy et al., 2017), after log10 transformation, the data were then scaled to zero mean and unit variance per metabolite, with the functions log10 and rowMeans from the base R package (R Core Team, 2021) and rowSds from the matrixStats package (Bengtsson, 2021). Co-abundant groups of metabolites were computed by using the original metabolomic data. A dissimilarity matrix was calculated using Spearman correlation, with the help of cor function from the stats R package (R Core Team, 2021). Then hierarchical clustering was performed with the hclust from the stats package, using ward. D2 method and a dissimilarity value cut-off of 0.2 (corresponding to a correlation value of 0.8), cutree from stats package. Similar transformation as for the PCA was used to visualize the data for early and late pregnancy as a heatmap, using the Heatmap function from the ComplexHeatmap package (Gu et al., 2016). The dendrogram visualized with the heatmap was based on the same method to compute the co-abundant groups. We analyzed the 211 prevalent metabolites, with a detection threshold of 0.01 and a prevalence of $10\%$ among the samples; calculated using the getPrevalentTaxa function from the mia package (Ernst et al., 2022). In the linear model comparing co-abundant groups of metabolites between the women with and without prediabetes, age, pre-pregnancy BMI and intake of polyunsaturated fatty acids (PUFAs) were chosen as covariates because these differed significantly between the study groups in early pregnancy. The dietary intervention during pregnancy was chosen as a covariate based on prior results (Mokkala et al., 2021). Model 1 included the following covariates: age, intake of PUFAs and the intervention with model 2 including also pre-pregnancy BMI since we wanted to be able to examine metabolites which correlated strongly with BMI. Linear models were carried out using the lm function and p-values adjusted with the Benjamini-Hochberg method, p.adjust from the stats package. A Random Forests classifier was used to predict prediabetes status based on the metabolomic profiles and covariates separately and combined. This was carried out using the ranger package (Wright & Ziegler, 2017) along with cross-validating each model 10 times with the caret R package (Kuhn, 2021). To further investigate the predictive power of each model, parallel models with randomly assigned prediabetes status were trained and cross-validated. The performance of each model was reported in terms of area under the receiver operating characteristic curve (AUROC), using the evalm function from the Mleval (John, 2020). The association of metabolomic data to glycemic traits was analyzed by Spearman correlation and significance was p-value adjusted with the FDR method, using the getExperimentCrossCorrelation from the mia package, where three levels of significance were used: 0.2, 0.05 and 0.001. Correlation values and significance were visualized as a heatmap using the Heatmap function from the ComplexHeatmap package. ## Clinical characteristics The majority i.e. $60\%$ of the women were living with overweight and $40\%$ with obesity. Almost a third of the women developed GDM in this current pregnancy; in most cases, it was treated with diet only (Table 1). The women who developed prediabetes by two years postpartum ($$n = 40$$) had a higher prepregnancy BMI, were older and were more likely to have had GDM during pregnancy compared to those who did not ($$n = 138$$). In addition, their fasting glucose and HbA1c in early pregnancy and fasting glucose, fasting insulin and HOMA2-IR in late pregnancy were higher than those of the women with no prediabetes. There were no significant differences between the groups according to which dietary intervention group they had been assigned in early pregnancy (data not shown). In early pregnancy, the women who later developed prediabetes had greater daily intakes of total fat, monounsaturated fatty acids and PUFAs than those who did not (Online Resource 1). Table 1Clinical characteristics of the study participantsAll womenn = 178No prediabetes at two years postpartumn = 138Prediabetes at two years postpartumn = 40n (%)n (%)n (%)P-valueAge in early pregnancy (years; mean, SD)31.5 (4.6)31.0 (4.7)33.0 (4.3)0.020aPre-pregnancy BMI (kg/m2; median, IQR)29.0 (26.5‒31.5)28.4 (26.2‒31.1)30.5 (27.8‒34.1)0.011bEthnicity1.0c European175 [98]135 [98]40 [100] Asian1 [1]1 [1]0 [0] Other2 [1]2 [1]0 [0]College or university education116 [65]94 [68]22 [55]0.14cSmoking during pregnancy6 [3]4 [3]2 [5]0.62cFamily history of diabetes or metabolic syndrome44 [25]34 [25]10 [25]1.0cPrimiparous82 [46]59 [43]23 [58]0.11cPrior GDM16 [9]13 [9]3 [8]0.77cGDM in the current pregnancy51 [30]27 [20]24 [62]< 0.001cGDM treatment< 0.001c diet only40 [23]25 [19]15 [39] metformin6 [4]2 [2]4 [10] insulin1 [1]0 [0]1 [3] insulin + metformin4 [2]0 [0]4 [10]Weight gain between early and late pregnancy (kg/week; mean, SD)0.42 (0.18)0.43 (0.18)0.36 (0.16)0.054aSystolic blood pressure (mmHg) early pregnancy (median, IQR)117.5 (111.3‒125.0)117.0 (110.5‒125.0)118.8 (113.8‒124.1)0.21b late pregnancy (median, IQR)119.0 (113.1‒128.0)118.8 (112.9‒127.6)120.8 (113.1‒132.0)0.31bDiastolic blood pressure (mmHg) early pregnancy (median, IQR)78.0 (71.5‒83.0)76.5 (71.0‒82.3)79.8 (73.5‒84.8)0.081b late pregnancy (median, IQR)79.3 (73.1‒86.5)79.0 (73.4‒85.6)81.5 (72.0‒91.6)0.39bBreastfeeding duration (months; median, IQR)12.0 (6.0‒15.5)12.1 (6.8‒17.3)12.0 (3.6‒14.6)0.36bFasting glucose (mmol/l) early pregnancy (median, IQR)4.7 (4.5‒5.0)4.7 (4.5‒4.9)4.9 (4.6‒5.2)0.029b late pregnancy (mean, SD)4.6 (0.4)4.5 (0.4)4.8 (0.6)0.020a 2 y postpartum (median, IQR)5.2 (5.0‒5.5)5.1 (4.9‒5.3)5.8 (5.7‒6.2)< 0.001bHbA1c (mmol/mol) early pregnancy (median, IQR)29.0 (28.0‒31.0)29.0 (28.0‒31.0)30.0 (29.0‒32.0)0.003b 2 y postpartum (median, IQR)32.0 (30.3‒34.0)32.0 (30.0‒34.0)35.0 (32.3‒36.0)< 0.001bFasting insulin (mU/l) early pregnancy (median, IQR)10.0 (8.0‒14.0)9.0 (7.0‒14.0)11.0 (8.0‒13.0)0.16b late pregnancy (median, IQR)15.0 (11.0‒20.0)14.0 (11.0‒19.0)16.0 (14.8‒22.0)0.038b 2 y postpartum (median, IQR)11.0 (8.0‒15.0)10.0 (8.0‒14.0)14.0 (11.0‒18.8)< 0.001bHOMA2-IR early pregnancy (median, IQR)1.3 (1.0‒1.8)1.2 (0.9‒1.8)1.4 (1.0‒1.7)0.13b late pregnancy (median, IQR)1.9 (1.4‒2.4)1.8 (1.4‒2.4)2.1 (1.8‒2.8)0.030b 2 y postpartum (median, IQR)1.5 (1.0‒1.9)1.3 (1.0‒1.8)1.9 (1.5‒2.5)< 0.001bhsCRP (mg/l) early pregnancy (median, IQR)5.6 (3.5‒8.8)5.5 (3.5‒8.7)6.3 (3.5‒9.2)0.48b late pregnancy (median, IQR)3.9 (2.2‒6.2)3.9 (2.2‒6.2)3.6 (2.0‒7.7)0.95b 2 y postpartum (median, IQR)1.6 (0.7‒3.4)1.4 (0.7‒3.3)2.3 (1.1‒4.6)0.027bGDM, gestational diabetes; HbA1c, glycated hemoglobin; HOMA2-IR, homeostatic model assessment for insulin resistance; hsCRP, high sensitivity C-reactive protein.aStudent’s t-testbMann-Whitney U-testcChi-square test ## Differences in serum metabolites during pregnancy between the women with and without prediabetes at two years postpartum The levels and/or ratios of many serum metabolites changed from early to late pregnancy as visualized in Online Resource 2 with the majority of the metabolite concentrations and/or ratios displaying an increase. There was no clear clustering of the metabolites in PCA based on the prediabetes status at two years postpartum in either at early or late pregnancy (Fig. 1). Fig. 1Principal Component Analysis of the serum metabolic profiles of the study participants in early ($$n = 174$$) and late pregnancy ($$n = 159$$) and prediabetes status at two years postpartum. Blue = no prediabetes and red = prediabetes Co-abundant groups of serum metabolites in early and late pregnancy were identified using a dissimilarity matrix and hierarchical clustering. Two co-abundant groups of metabolites differed in early pregnancy between the women with and those without prediabetes (Fig. 2). The first group included higher concentrations of small HDL particles, total lipids in small HDL, phospholipids in small HDL and free cholesterol in small HDL in the women with prediabetes compared to the women without prediabetes ($$p \leq 0.029$$, Linear model adjusted for age, intervention and intake of PUFAs). The second group showed a higher phospholipids to total lipids ratio in large HDL particles in the women with prediabetes compared to the women without prediabetes ($$p \leq 0.020$$, Linear model adjusted for age, intervention and intake of PUFAs). When the models were further adjusted for prepregnancy BMI, the associations were attenuated ($$p \leq 0.094$$ and $$p \leq 0.070$$, respectively). Fig. 2Boxplots of early and late pregnancy serum metabolites, which differed significantly ($p \leq 0.05$) between the women with and those without prediabetes at two years postpartum based on a linear model adjusted for age, intervention and intake of polyunsaturated fatty acids. $$n = 174$$ in early pregnancy, $$n = 38$$ with prediabetes and $$n = 136$$ without prediabetes. $$n = 159$$ in late pregnancy, $$n = 29$$ with prediabetes and $$n = 130$$ without prediabetes. The box represents the interquartile range, the line is the median and dots are individual values. ** indicates p-value ≤ 0.01 and *** p-value ≤ 0.001. CAG, co-abundant group of metabolites One co-abundant group of metabolites was higher in late pregnancy in the women with prediabetes as compared to women who did not develop this condition (Fig. 2, $$p \leq 0.014$$, Linear model adjusted for age, intervention and intake of PUFAs). The group included only acetoacetate, and it remained significant even after adjusting for the prepregnancy BMI ($$p \leq 0.020$$). ## Prediction of prediabetes status at two years postpartum with serum metabolites during pregnancy We used AUROC of a Random Forests classifier for predicting the prediabetes status at two years’ postpartum. When based on serum metabolites in early pregnancy, the value was 0.655, while it was 0.438 based on only covariates i.e. age, intervention, prepregnancy BMI and intake of PUFAs in early pregnancy and 0.668 when based on both serum metabolites and covariates in early pregnancy. After 10-fold cross-validation, it was observed that the classifier based on serum metabolites and covariates performed the best (Fig. 3). The following five serum metabolites and covariates were the most important features in the classifier; glycerol, cholesterol esters to total lipids ratio in very large HDL particles, acetoacetate, free cholesterol to total lipids ratio in large HDL and age. Both the classifier based on serum metabolites and the classifier based on serum metabolites and covariates performed significantly better than those classifiers which randomly assigned the prediabetes status. Fig. 3Boxplots of the area under the receiver operating characteristic curve (AUROC) -values of Random Forests models predicting prediabetes status at two years postpartum with serum metabolites, covariates age, intervention, pre-pregnancy BMI and intake of polyunsaturated fatty acids or serum metabolites and covariates. Student’s t-test was used to compare the Random Forests models with true prediabetes status labels to models which assigned the status at random (shuffled labels). $$n = 38$$ with prediabetes and $$n = 136$$ without prediabetes in early pregnancy, $$n = 29$$ with prediabetes and $$n = 130$$ without prediabetes in late pregnancy. Box represents interquartile range, line median and dots individual values. * indicates p-value ≤ 0.05. NS, non-significant The AUROC of a Random Forests classifier predicting the prediabetes status at two years postpartum based on serum metabolites in late pregnancy was 0.64, which was a similar value as that obtained with the classifier based on both serum metabolites and covariates in late pregnancy (AUROC 0.638). The AUROC value of the classifier based on covariates only in late pregnancy was somewhat lower (AUROC 0.53). None of the classifiers differed significantly from the classifiers which assigned the prediabetes status at random (Fig. 3). ## The correlations between serum metabolites during pregnancy and glycemic traits at two years postpartum FDR-adjusted Spearman correlations were used to investigate the associations of serum metabolites in early and late pregnancy to glycemic traits (fasting glucose and insulin, HOMA2-IR and HbA1c) and hsCRP at two years postpartum (Fig. 4). The concentrations of small HDL particles, total lipids in small HDL, phospholipids in small HDL and free cholesterol in small HDL in early pregnancy correlated positively with all of the glycemic traits, in addition to pre-pregnancy BMI, at two years postpartum. There were highly significant (FDR ≤ 0.001) positive correlations between triglycerides in medium size HDL particles in early pregnancy and fasting insulin and HOMA2-IR. The inflammatory marker GlycA correlated positively with all of the glycemic traits, especially with HbA1c at FDR-level ≤ 0.001. Along with many VLDL related variables, the levels of BCAAs (i.e. leucine and isoleucine) and monounsaturated fatty acids correlated positively with fasting insulin and HOMA2-IR. The concentrations of valine, leucine and two ketone bodies i.e. acetoacetate and 3-hydroxybutyrate correlated positively with that of fasting glucose. Glycerol had a highly significant positive correlation with fasting glucose. Highly significant negative correlations were detected between the ratios of n-6 fatty acids to total fatty acids and linoleic acid to total fatty acids and fasting insulin and HOMA2-IR. Fig. 4Heatmaps of FDR-adjusted Spearman correlations between serum metabolites in early or late pregnancy and glycemic traits at two years’ postpartum. $$n = 174$$ in early pregnancy and $$n = 159$$ in late pregnancy. * indicates FDR ≤ 0.2, ** FDR ≤ 0.05 and *** FDR ≤ 0.001. HOMA2-IR, homeostatic model assessment for insulin resistance; HbA1c, glycated hemoglobin; hsCRP, high sensitivity C-reactive protein Of the serum metabolites in late pregnancy, the concentration of isoleucine correlated positively at FDR-level ≤ 0.2 with fasting insulin and HOMA2-IR at two years postpartum (Fig. 4). Similarly, the glucose concentration correlated positively with fasting glucose, citrate with fasting insulin and HOMA2-IR and tyrosine with HbA1c. ## Discussion In this study, we demonstrated that women who developed prediabetes by two years postpartum had higher serum concentrations of small HDL particles, total lipids in small HDL, phospholipids in small HDL and free cholesterol in small HDL in early pregnancy and higher serum concentrations of acetoacetate in late pregnancy as compared to women who did not develop prediabetes. The small HDL related variables also correlated positively with HbA1c and markers of insulin resistance at two years postpartum. We detected elevated serum levels of small HDL particles in early pregnancy in the women who developed prediabetes by two years postpartum, although the association was dependent on the pre-pregnancy BMI. In a large previous study of non-pregnant women, higher levels of small HDL particles and a smaller HDL particle size were associated with incident type 2 diabetes during a follow-up of 13 years (Mora et al., 2010). Similarly, a larger HDL particle size has been associated with a decreased risk of type 2 diabetes in young adults (Ahola-Olli et al., 2019). In addition to type 2 diabetes, elevated concentrations of small HDL particles have been linked with a risk of cardiovascular disease (Kontush, 2015). During early pregnancy, higher concentrations of small HDL particles have been shown to predict GDM (Mokkala et al., 2020). Thus, it appears that high levels of small HDL particles are associated with an increased cardiometabolic risk at different stages of the lifecycle. Therefore it is not unreasonable that we detected similar metabolic features in pregnant women prior to the onset of GDM and prediabetes, as GDM is a known risk factor for type 2 diabetes (Bellamy et al., 2009). Further studies are warranted to clarify if elevated serum levels of small HDL particles during pregnancy predict the onset of type 2 diabetes, especially in women affected by GDM, thus identifying a possible high-risk group in need of targeted screening and interventions to prevent the onset of diabetes. A dietary intervention would represent a feasible approach to exert an impact on metabolism. Indeed, in the same cohort as studied here, dietary supplementation with fish oil and probiotics during pregnancy induced favorable alterations in serum lipid variables, although the alterations were less evident in women with GDM (Mokkala et al., 2021). Since only 40 women developed prediabetes by two years postpartum, we did not investigate the impact of the dietary intervention here. The incidence of prediabetes did not differ between the dietary intervention groups. In addition to the levels of small HDL particles, the levels of two BCAAs leucine and isoleucine, the aromatic amino acid phenylalanine and an inflammatory marker GlycA in early pregnancy correlated positively with markers of insulin resistance at two years postpartum. BCAAs are among the best-established metabolic markers for type 2 diabetes (Long et al., 2020). Higher levels of leucine, isoleucine, phenylalanine and GlycA have been associated with an increased risk for type 2 diabetes in young adults (Ahola-Olli et al., 2019). Liu et al. examined women at 28 weeks of gestation and observed an association between elevated BCAA levels and prediabetes or type 2 diabetes 10 to 14 years later (Liu et al., 2021). Elevated serum levels of leucine, isoleucine and GlycA during pregnancy have also been frequently associated with GDM (Kivelä et al., 2021; Mokkala et al., 2020; White et al., 2017). Recently it has been suggested that higher BCAA and GlycA levels point to a susceptibility to develop type 2 diabetes already decades before the onset of the disease (Bell et al., 2020) and that in fact insulin resistance causally affects BCAA metabolism (Mahendran et al., 2017; Wang et al., 2017). As reviewed recently (White et al., 2021), obesity and insulin resistance per se may increase circulating BCAA levels, which in turn contribute to the development of cardiometabolic diseases. The levels of acetoacetate, a ketone body, were elevated in late pregnancy in the women who later developed prediabetes. It was also the third most important predictor of future prediabetes in early pregnancy in our Random Forests classifier. In addition to this finding, Liu et al. detected an association between another ketone body, 3-hydroxybutyrate measured at 28 weeks of gestation, and postpartum prediabetes or type 2 diabetes (Liu et al., 2021). In a cohort of middle-aged men, elevated fasting levels of acetoacetate were associated with incident type 2 diabetes during a 5-year follow-up and with impaired insulin secretion rather than insulin resistance (Mahendran et al., 2013). Increased levels of acetoacetate have also been detected in women with GDM prior to and at the time of the diagnosis (White et al., 2017). It is evident that the rate of ketogenesis in the liver is regulated by multiple factors; ketogenic substrates include fatty acids and amino acids, especially leucine (Puchalska & Crawford, 2017), which could link elevated acetoacetate levels to increased BCAA metabolism. Based on our results, serum metabolites in early pregnancy could predict future prediabetes when combined in a model with clinical variables. A similar predictive power has been observed in a previous study examining women at 28 weeks of gestation (Liu et al., 2021), but the improvement in the prediction compared to traditional clinical factors alone was minimal. Compared to our Random Forests classifier and the clinical variables based on baseline differences between the prediabetes groups, Liu et al. used a wider range of clinical factors in their model, such as a family history of diabetes, parity and OGTT results during pregnancy. They also included fewer metabolites based on Lasso regression analysis in their model. In our model, glycerol was the most important predictor. In addition to fasting glucose at two years postpartum, it correlated with fasting insulin and HOMA2-IR. Circulating levels of glycerol and fatty acids are elevated by excessive lipolysis in adipose tissue, a feature encountered in individuals with obesity and insulin resistance. Glycerol and fatty acids in turn may promote insulin resistance in skeletal muscle and liver. ( Bódis & Roden, 2018) This study has several strengths. We examined an at-risk group of women for metabolic disorders and had a standardized protocol recording multiple clinical variables from early pregnancy until two years postpartum. All blood samples were collected in the fasting state. The comprehensive collection of background data allowed us to include potential confounding factors into the statistical analyses. Nonetheless, if there had been a larger number of study participants with prediabetes, this could have made it possible to reveal other serum metabolites associated with prediabetes in addition to those detected here, and thus future trials with a larger number of participants are called for in order to verify our findings. In addition, the maternal BMI value may have influenced the associations of metabolites to prediabetes, although we included pre-pregnancy BMI as a covariate in the analyses. Clearly it would be informative to examine the associations also in individuals with normal weight. The fact that the study cohort included women from a high-income European country might somewhat limit the generalization of the results, but the mean age and parity of the study participants correlate well with values currently observed in the Finnish population (Official Statistics of Finland, Perinatal statistics. THL., 2021). ## Conclusion Aberrant serum metabolic profile was detected in early pregnancy in women who developed prediabetes by two years postpartum, namely elevated serum concentrations of small HDL particles, and increased total lipids, phospholipids and free cholesterol in small HDL particles. The association seems to depend on pre-pregnancy BMI. Together with traditional clinical markers, the assessment of the serum metabolic profile in early pregnancy could potentially be used to predict future prediabetes risk. Future studies will be needed to clarify whether the metabolic features detected here reveal an at-risk group of women who would benefit from interventions to prevent type 2 diabetes during and after pregnancy. ## Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 ## References 1. Ahola-Olli AV, Mustelin L, Kalimeri M, Kettunen J, Jokelainen J, Auvinen J, Puukka K, Havulinna AS, Lehtimäki T, Kähönen M, Juonala M, Keinänen-Kiukaanniemi S, Salomaa V, Perola M, Järvelin MR, Ala-Korpela M, Raitakari O, Würtz P. **Circulating metabolites and the risk of type 2 diabetes: A prospective study of 11,896 young adults from four Finnish cohorts**. *Diabetologia* (2019.0) **62** 2298-2309. DOI: 10.1007/s00125-019-05001-w 2. **Diagnosis and classification of diabetes mellitus**. *Diabetes Care* (2011.0) **34** 62-69. DOI: 10.2337/dc11-S062 3. Bell JA, Bull CJ, Gunter MJ, Carslake D, Mahajan A, Smith GD, Timpson NJ, Vincent EE. **Early metabolic features of genetic liability to type 2 diabetes: Cohort study with repeated metabolomics across early life**. *Diabetes Care* (2020.0) **43** 1537-1545. DOI: 10.2337/dc19-2348 4. Bellamy L, Casas JP, Hingorani AD, Williams D. **Type 2 diabetes mellitus after gestational diabetes: A systematic review and meta-analysis**. *Lancet (London England)* (2009.0) **373** 1773-1779. DOI: 10.1016/S0140-6736(09)60731-5 5. Bengtsson, H. (2021). matrixStats: Functions that Apply to Rows and Columns of Matrices (and to Vectors) (R package version 0.61.0). https://cran.r-project.org/package=matrixStats 6. Bódis K, Roden M. **Energy metabolism of white adipose tissue and insulin resistance in humans**. *European Journal of Clinical Investigation* (2018.0) **48** e13017. DOI: 10.1111/eci.13017 7. Claesson R, Ignell C, Shaat N, Berntorp K. **HbA1c as a predictor of diabetes after gestational diabetes mellitus**. *Primary Care Diabetes* (2017.0) **11** 46-51. DOI: 10.1016/j.pcd.2016.09.002 8. Durnwald CP, Downes K, Leite R, Elovitz M, Parry S. **Predicting persistent impaired glucose tolerance in patients with gestational diabetes: The role of high sensitivity CRP and adiponectin**. *Diabetes/Metabolism Research and Reviews* (2018.0) **34** 6-11. DOI: 10.1002/dmrr.2958 9. Ernst, F. G. M., Shetty, S. A., Borman, T., & Lahti, L. (2022). mia: Microbiome analysis. R package version 1.3.13. https://github.com/microbiome/mia 10. Gu ZG, Eils R, Schlesner M. **Complex heatmaps reveal patterns and correlations in multidimensional genomic data**. *Bioinformatics* (2016.0) **32** 2847-2849. DOI: 10.1093/bioinformatics/btw313 11. Hod M, Kapur A, Sacks DA, Hadar E, Agarwal M, Di Renzo GC, Roura C, McIntyre L, Morris HD, Divakar H. **The International Federation of Gynecology and Obstetrics (FIGO) Initiative on gestational diabetes mellitus: A pragmatic guide for diagnosis, management, and care**. *International Journal of Gynaecology and Obstetrics: The Official Organ of the International Federation of Gynaecology and Obstetrics* (2015.0) **131** 173-211. DOI: 10.1016/S0020-7292(15)30033-3 12. John, C. R. (2020). Mleval: Machine Learning Model Evaluation. R package version 0.3. https://cran.r-project.org/package=Mleval 13. Kivelä J, Sormunen-Harju H, Girchenko PV, Huvinen E, Stach-Lempinen B, Kajantie E, Villa PM, Reynolds RM, Hämäläinen EK, Lahti-Pulkkinen M, Murtoniemi KK, Laivuori H, Eriksson JG, Räikkönen K, Koivusalo SB. **Longitudinal metabolic profiling of maternal obesity, gestational diabetes, and hypertensive pregnancy disorders**. *The Journal of Clinical Endocrinology &amp; Metabolism* (2021.0) **106** e4372-e4388. DOI: 10.1210/clinem/dgab475 14. Kontush A. **HDL particle number and size as predictors of cardiovascular disease**. *Frontiers in Pharmacology* (2015.0) **6** 1-6. DOI: 10.3389/fphar.2015.00218 15. Kuhn, M. (2021). caret: Classification and Regression Training. R package version 6.0–88. https://cran.r-project.org/package=caret 16. Lain KY, Catalano PM. **Metabolic changes in pregnancy**. *Clinical Obstetrics and Gynecology* (2007.0) **50** 938-948. DOI: 10.1097/GRF.0b013e31815a5494 17. Liu Y, Kuang A, Bain JR, Muehlbauer MJ, Ilkayeva OR, Lowe LP, Metzger BE, Newgard CB, Scholtens DM, Lowe WL. **Maternal Metabolites Associated with Gestational Diabetes Mellitus and a Postpartum Disorder of Glucose Metabolism**. *The Journal of Clinical Endocrinology &amp; Metabolism* (2021.0) **106** 3283-3294. DOI: 10.1210/clinem/dgab513 18. Long J, Yang Z, Wang L, Han Y, Peng C, Yan C, Yan D. **Metabolite biomarkers of type 2 diabetes mellitus and pre-diabetes: A systematic review and meta-analysis**. *BMC Endocrine Disorders* (2020.0) **20** 1-17. DOI: 10.1186/s12902-020-00653-x 19. Mahendran Y, Jonsson A, Have CT, Allin KH, Witte DR, Jørgensen ME, Grarup N, Pedersen O, Kilpeläinen TO, Hansen T. **Genetic evidence of a causal effect of insulin resistance on branched-chain amino acid levels**. *Diabetologia* (2017.0) **60** 873-878. DOI: 10.1007/s00125-017-4222-6 20. Mahendran Y, Vangipurapu J, Cederberg H, Stančáková A, Pihlajamäki J, Soininen P, Kangas AJ, Paananen J, Civelek M, Saleem NK, Pajukanta P, Lusis AJ, Bonnycastle LL, Morken MA, Collins FS, Mohlke KL, Boehnke M, Ala-Korpela M, Kuusisto J, Laakso M. **Association of ketone body levels with hyperglycemia and type 2 diabetes in 9,398 Finnish men**. *Diabetes* (2013.0) **62** 3618-3626. DOI: 10.2337/db12-1363 21. McCarthy DJ, Campbell KR, Lun ATL, Wills QF. **Scater: Pre-processing, quality control, normalization and visualization of single-cell RNA-seq**. *data in R Bioinformatics* (2017.0) **33** 1179-1186. DOI: 10.1093/bioinformatics/btw777 22. Mokkala K, Vahlberg T, Houttu N, Koivuniemi E, Lahti L, Laitinen K. **Impact of combined consumption of fish oil and probiotics on the serum metabolome in pregnant women with overweight or obesity**. *EBioMedicine* (2021.0) **73** 103655. DOI: 10.1016/j.ebiom.2021.103655 23. Mokkala K, Vahlberg T, Pellonperä O, Houttu N, Koivuniemi E, Laitinen K. **Distinct metabolic profile in early pregnancy of overweight and obese women developing gestational diabetes**. *Journal of Nutrition* (2020.0) **150** 31-37. DOI: 10.1093/jn/nxz220 24. Mora S, Otvos JD, Rosenson RS, Pradhan A, Buring JE, Ridker PM. **Lipoprotein particle size and concentration by nuclear magnetic resonance and incident type 2 diabetes in women**. *Diabetes* (2010.0) **59** 1153-1160. DOI: 10.2337/db09-1114 25. Official Statistics of Finland, Perinatal statistics. THL (2021). Perinatal statistics – parturients, deliveries and newborns 2020. Statistical Report 49/2021. https://thl.fi/en/web/thlfi-en/statistics-and-data/statistics-by-topic/sexual-and-reproductive-health/parturients-deliveries-and-births/perinatal-statistics-parturients-delivers-and-newborns 26. Ozuguz U, Isik S, Berker D, Arduc A, Tutuncu Y, Akbaba G, Gokay F, Guler S. **Gestational diabetes and subclinical inflammation: Evaluation of first year postpartum outcomes**. *Diabetes Research and Clinical Practice* (2011.0) **94** 426-433. DOI: 10.1016/j.diabres.2011.08.024 27. Pellonperä O, Mokkala K, Houttu N, Vahlberg T, Koivuniemi E, Tertti K, Rönnemaa T, Laitinen K. **Efficacy of fish oil and/or probiotic intervention on the incidence of gestational diabetes mellitus in an at-risk group of overweight and obese women: a randomized, placebo-controlled, double-blind clinical trial**. *Diabetes Care* (2019.0) **42** 1009-1017. DOI: 10.2337/dc18-2591 28. Puchalska P, Crawford PA. **Multi-dimensional roles of ketone bodies in fuel metabolism, signaling, and therapeutics**. *Cell Metabolism* (2017.0) **25** 262-284. DOI: 10.1016/j.cmet.2016.12.022 29. R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.r-project.org/ 30. Ruiz-Canela M, Hruby A, Clish CB, Liang L, Martínez-González MA, Hu FB. **Comprehensive metabolomic profiling and incident cardiovascular disease: A systematic review**. *Journal of the American Heart Association* (2017.0) **6** 1-22. DOI: 10.1161/JAHA.117.005705 31. Soininen P, Kangas AJ, Würtz P, Suna T, Ala-Korpela M. **Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics**. *Circulation: Cardiovascular Genetics* (2015.0) **8** 192-206. DOI: 10.1161/CIRCGENETICS.114.000216 32. Varejão AM, Ferreira JL, Dória M, Laranjo M, Araújo MC, Peixinho C, Couto AS, Príncipe RM. **HbA1c as a predictor of postpartum diabetes mellitus after gestational diabetes mellitus**. *Diabetes and Metabolic Syndrome: Clinical Research and Reviews* (2021.0) **15** 4-7. DOI: 10.1016/j.dsx.2021.102269 33. Wallace TM, Levy JC, Matthews DR. **Use and abuse of HOMA modeling**. *Diabetes Care* (2004.0) **27** 1487-1495. DOI: 10.2337/diacare.27.6.1487 34. Wang Q, Holmes MV, Smith GD, Ala-Korpela M. **Genetic support for a causal role of insulin resistance on circulating branched-chain amino acids and inflammation**. *Diabetes Care* (2017.0) **40** 1779-1786. DOI: 10.2337/dc17-1642 35. White PJ, McGarrah RW, Herman MA, Bain JR, Shah SH, Newgard CB. **Insulin action, type 2 diabetes, and branched-chain amino acids: A two-way street**. *Molecular Metabolism* (2021.0) **52** 101261. DOI: 10.1016/j.molmet.2021.101261 36. White SL, Pasupathy D, Sattar N, Nelson SM, Lawlor DA, Briley AL, Seed PT, Welsh P, Poston L. **Metabolic profiling of gestational diabetes in obese women during pregnancy**. *Diabetologia* (2017.0) **60** 1903-1912. DOI: 10.1007/s00125-017-4380-6 37. Working group set up by the Finnish Medical Society Duodecim, the Medical Advisory Board of the Finnish Diabetes Association and the Finnish Gynecological Association. Gestational diabetes. Current Care Guidelines. Helsinki: The Finnish Medical Society Duodecim (2013). Retrieved January 26, 2022, from www.kaypahoito.fi 38. Wright MN, Ziegler A. **Ranger: A fast implementation of Random forests for high dimensional data in C + + and R**. *Journal of Statistical Software* (2017.0) **77** 1-17. DOI: 10.18637/jss.v077.i01
--- title: Prevalence of lower urinary tract symptoms in a cohort of Australian servicewomen and female veterans authors: - Simone D. O’Shea - Rod Pope - Katharine Freire - Robin Orr journal: International Urogynecology Journal year: 2022 pmcid: PMC10038961 doi: 10.1007/s00192-022-05254-x license: CC BY 4.0 --- # Prevalence of lower urinary tract symptoms in a cohort of Australian servicewomen and female veterans ## Abstract ### Introduction and hypothesis Lower urinary tract symptoms (LUTS) are common in the general female population. It was hypothesised that Australian female military personnel and veterans would experience similar types and prevalence of LUTS as the broader Australian female population. ### Methods An online cross-sectional survey was utilised to explore the pelvic health of active servicewomen and veterans in the Australian Defence Force (ADF). For the purposes of this report, only the demographic and LUTS data (excluding urinary tract infections) were extracted and descriptively analysed. ### Results A total of 491 complete survey responses were received and analysed. Respondent characteristics were comparable to those documented in a departmental report regarding ADF servicewomen. No LUTS were reported by $38\%$ of respondents. Regular symptoms of urinary incontinence were experienced by $27\%$ of respondents (stress urinary incontinence = $23\%$, urge urinary incontinence = $16\%$, mixed urinary incontinence = $13\%$), bladder storage issues by 20–$27\%$, and various voiding impairments by 9–$27\%$. In addition, $41\%$ reported regularly experiencing two or more LUTS, and for over two thirds of respondents, LUTS were an ongoing issue. Relationships between age, parity, and symptoms of urinary incontinence were also seen. ### Conclusions Consistent with wider research in Australian female populations, LUTS were commonly experienced during service by ADF female military personnel and veterans. Given the high likelihood of female military personnel experiencing LUTS during their service, and a proportion experiencing ongoing symptoms, tailored monitoring and support for urinary health should be available to enhance occupational health, safety, and performance. ### Supplementary Information The online version contains supplementary material available at 10.1007/s00192-022-05254-x. ## Background Lower urinary tract symptoms (LUTS)—such as incontinence and bladder storage or voiding issues [1]—are more likely to occur in females than males [2–4]. It has been estimated that up to $80\%$ of women experience LUTS. However, the frequency of symptoms and their impact on women’s wellbeing varies from mild to more severe [4]. Prevalence rates also vary across the lifespan between, and within, different types of LUTS [2, 3, 5]. For example, urinary incontinence, or involuntary leakage of urine [1], is one of the most commonly investigated sources of LUTS and reportedly affects between $13\%$–$46\%$ of Australian women [6, 7]. However, urinary incontinence is an umbrella term which encompasses a range of leakage sub-types, such as stress urinary incontinence (SUI, leakage during physical activity), urge urinary incontinence (UUI, leakage associated with urgency) and mixed urinary incontinence (MUI, leakage during physical activity as well as with urgency) [1]. The prevalence rates for these sub-types vary (for example, SUI $16\%$, UUI $7.5\%$, and MUI $18\%$ in one sample of Australian women [8]), with SUI being more common in middle-age women and UUI in older women [8]. Negative physical, emotional, and social ramifications have been reported by women experiencing LUTS, such as reduced physical activity [9], anxiety, depression, and social isolation [10], reduced quality of life [11], and reduced work productivity [12]. Whilst further research is required to determine the relationships between different occupations and LUTS, a recent rapid review and meta-analysis found women engaged in more manual labour or physically demanding occupations had a higher risk of urinary incontinence [12]. In addition, women with LUTS have been found to alter behaviours associated with toileting and fluid intake at work [12] or modify work activities, such as physical training, to manage their symptoms [13]. These findings suggest a bi-directional relationship between LUTS and work. Female military personnel are a growing occupational group. They typically engage in high levels of physical training and load carriage [14], often work in occupational contexts influencing fluid intake and voiding behaviours, and austere environments where sanitation may be problematic [15]. Equipment and workplace culture have also been reported to influence the risk of some types of LUTS and the self-care behaviours women use to prevent or manage urinary symptoms at work [16, 17]. Therefore, servicewomen may have an increased risk of LUTS and impaired occupational performance due to LUTS. A recent narrative review of pelvic floor health in female military populations found that approximately one third of US servicewomen experienced urinary incontinence [18]. However, no prevalence data were found for other LUTS (excluding urinary tract infections) or for women serving in the Australian Defence Force (ADF), which was the setting for the current study. Consideration of the urinary health needs of female military personnel is important for enhancing their health, wellbeing, safety, and occupational performance. To inform potential mitigation and support strategies, it is necessary to understand the prevalence and characteristics of LUTS in Australian female military personnel. Therefore, the aims of this investigation were to determine the types, prevalence, and severity of LUTS experienced by ADF servicewomen and female veterans. ## Method LUTS were explored in this study as part of a larger program of research on the pelvic health of Australian female military personnel and veterans. The research employed a cross-sectional design, using an online, anonymous survey. The survey approach had the advantage of being simple, efficient, and cost-effective to administer to a large sample from the chosen population [19, 20]. The survey was easily accessible to respondents regardless of their location, provided a robust description of the underlying population in relation to the topic area, and allowed for anonymous data collection to protect participant privacy and mitigate the risk of coercion to participate, or not, which is a particular ethical concern in military contexts [21]. Ethics approvals for the study were received from the Human Research Ethics Committees of the Departments of Defence and Veterans’ Affairs [099-19], Charles Sturt University (H19271), and Bond University (TCO1733). Data were collected between October 2019 and June 2020. To be eligible to participate in the survey, individuals needed to identify as biologically female, be aged 18 years or over, and have actively served in the Australian Navy, Army, or Air Force for at least 6 months, in either a part- or full-time capacity. Due to the personal nature of pelvic health, potential participants were recruited through a combination of print (Navy, Army, and Air Force newspapers) and social media (Facebook) advertisements. This non-probability sampling recruitment strategy helped to mitigate the risk of coercion to participate, or not, which might have occurred if survey invitations were promulgated through on-site military establishments. It also enabled women to access information about the survey at a time and place that was comfortable, private, and secure for them. Prospective participants were provided with an information and consent statement on the survey landing page and had to click the ‘proceed’ button at the end of the page to indicate their understanding and consent prior to commencing the survey. Whilst many questionnaires exist for specific pelvic health issues, no previously validated survey instrument was identified to address the breadth of female pelvic health issues to be explored within the study. Therefore, a purpose-built questionnaire was developed by the research team for distribution via the Qualtrics survey platform. The survey questions were developed based on reports of prior research, consideration of the unique research aims and target population, and review of validated questionnaires that have been developed to assess different aspects of female genitourinary health and related issues [20, 22–27]. The online questionnaire utilised a standardised order of questions, which were presented in specific sections (Fig. 1). Skip logic was built into the survey design and allowed respondents to bypass questions that were not relevant to them, based on their responses to preceding questions [28]. The questionnaire included a range of open and closed question styles, including Likert and graphic rating scales, short-answer questions, and open narrative commentary boxes. The latter allowed participants the opportunity to elaborate on their personal experiences in the topic area [28]. The survey took participants on average 30–60 min to complete, depending on the number of questions that were relevant to them and the level of detail they provided. Within this study, only the demographic and prevalence data related to LUTS (excluding urinary tract infections) were analysed and are reported. Fig. 1Sections and order of questioning within Pelvic Health Survey The survey development team consisted of four physiotherapists who had experience with women’s health and/or military populations, and a survey developer. Expert review of the survey instrument to critically assess and assure its functionality, content, language, context and relevance to the population, as well as its acceptability, was provided by: a senior Registered Nurse with extensive health care delivery experience within the military and public health setting; an Obstetrician and Gynaecologist with > 35 years of clinical experience; and an advisory group of active servicewomen ($$n = 10$$) from the Women’s Veterans’ Network Australia, representing each service (Navy, Army, Air Force). The survey was refined to address their feedback and recommendations. Data on the following LUTS were collected within the elements of the survey considered in this report: urinary incontinence symptoms, including SUI, UUI, and MUI; bladder storage symptoms, including urinary urgency (UU) or sudden and pressing need to urinate; urinary frequency or the number of times respondents needed to urinate within an 8-h working shift; and nocturia or increased urine production overnight leading to interrupted sleep. Voiding symptoms, including incomplete bladder emptying, straining to empty the bladder, and weak urinary flow, as well as painful bladder emptying, were also explored. To avoid overinflating reported prevalence rates [5], only LUTS that occurred regularly (more than once per week) were counted when calculating prevalence estimates, and it was assumed that non-responses meant the participant did not experience the symptom. All responses received were exported from the Qualtrics survey platform and imported into SPSS (Version 26) for analysis. Initial analysis was descriptive in nature, first estimating the response rate and demographic characteristics of respondents and then comparing them with previously reported population characteristics. Subsequently, measures of frequency and, where appropriate, indicators of the precision of population estimates were calculated for survey variables reflecting prevalence and severity of types of LUTS. Due to the size of the sample and concerns regarding family-wise error rates associated with numerous statistical tests, sub-group comparisons were limited and descriptive analysis used to identify trends within the data [28]. ## Results In total, 987 survey responses were received. However, the data from 496 survey responses were removed as they were incomplete (demographic data only). Therefore, a total of 491 survey responses remained to inform this study and were subsequently analysed. The respondents were recruited from a population of ADF servicewomen and female veterans estimated to total 48,000, comprising approximately 13,600 active servicewomen [29] and an estimated 34,400 female veterans. In the absence of more precise data, the latter estimate was derived by doubling the number of female veterans who left military service in the 20-year period ending in 2019 (17,200 veterans [30]) to provide a crude estimate of the likely size of the entire Australian female veteran population in 2019. Based on this estimated total population size of servicewomen and female veterans in 2019, the survey sample of 491 respondents comprised approximately $1\%$ of the underlying population. If the survey sample was representative of the underlying population and assuming a $95\%$ confidence level, these figures would provide a margin of error for population estimates derived from the sample of ± $4\%$. However, the assumption of representativeness of the sample needed closer examination and is further considered below. ## Survey participation rate, sample characteristics, and representativeness of the sample The survey participation rate was difficult to estimate. Assuming all ADF servicewomen and female veterans were reached by the survey invitation, as indicated above, the response rate would be approximately $1\%$. Clearly, not all eligible servicewomen and female veterans would have been reached, and so the actual participation rate will be higher but is unascertainable [31]. Nevertheless, if $10\%$ of ADF servicewomen and female veterans were reached by the survey invitation, the participation rate would be approximately $10\%$, which remains relatively low and necessitates further assessment of the representativeness of the survey sample. To assess the representativeness of the sample, the demographic profile of respondents was compared with the profiles of servicewomen obtained from the Defence Census 2019 [29] and Women in the ADF report 2017–2018 [32]. No similar data source for ex-serving Australian female military personnel exists. Therefore, the characteristics of this population are not known beyond their assumed similarity to currently serving women (though they will on average be older—the latter assumed because they were serving women before they were veterans). Thus, the demographic profile of veterans who responded to the survey was also compared to the profile of current servicewomen discussed above, while age differences were expected and assumed in the comparison. The cohorts of respondents from the pelvic health survey were well matched to the servicewomen profiles reported [29, 32], across a range of attributes, once expected age differences for the veteran cohort were considered. Although the median age of our sample of active servicewomen was a few years higher than the median age of servicewomen reported in Defence *Census data* [29], when the proportions of servicewomen aged under and over 50 years in the survey sample and in the underlying population were compared, they were similar (Table 1). The distributions of women across Service arms were also comparable (Table 2), as were the median lengths of service for those currently employed full-time (Defence Census = 5 years, survey respondents = 6 years). However, a smaller proportion of active servicewomen who responded to the survey identified as Reservists ($14\%$, $95\%$ CI 10-$19\%$) compared with Defence *Census data* indicating that $22\%$ of female ADF personnel (2988 of 13,564) were employed in a part-time capacity [29].Table 1Age comparisons between survey respondents and Defence Census 2019 dataComparisonDefence Census [2019]Pelvic Health SurveyActively serving($$n = 299$$)Veterans**($$n = 192$$)Full timeReserveFull time($$n = 257$$)Reserve($$n = 42$$)Full time($$n = 178$$)Reserve($$n = 14$$)Female Age (median years)2841364447.554Age category < 50 years$91\%$$64\%$193 ($89\%$)26 ($67\%$)86 ($55\%$)4 ($36\%$) 50+ years$9\%$$35\%$24 ($11\%$)13 ($33\%$)70 ($45\%$)7 ($64\%$)**Note: Median ages for female veterans (ex-serving women) were expected to be 10–20 years older than median ages for currently serving women and these findings are consistent with that expectation. The age profile for female veterans has not been compared with age data from the Defence Census [2019] because the survey only identified age of veterans at time of completing the survey, not during ServiceTable 2Comparison between survey respondents and Defence Census 2019 data for distributions of women across servicesService armDefence Census [2019]Pelvic Health SurveyAll women($$n = 13$$,574)Full time($$n = 10$$,586)Reserve($$n = 2988$$)All respondents($$n = 491$$)Actively serving($$n = 299$$)Veterans($$n = 192$$)Full time($$n = 257$$)Reserve($$n = 42$$)Full Time($$n = 178$$)Reserve($$n = 14$$)Navy3428 ($25\%$)3026 ($29\%$)402 ($14\%$)105 ($21\%$)58 ($23\%$)4 ($10\%$)38 ($23\%$)5 ($36\%$)Army6161 ($45\%$)4255 ($40\%$)1906 ($64\%$)258 ($53\%$)113 ($44\%$)31 ($76\%$)105 ($59\%$)9 ($64\%$)Air Force3985 ($29\%$)3305 ($31\%$)680 ($23\%$)126 ($26\%$)85 ($33\%$)6 ($14\%$)35 ($20\%$)- A range of occupational groups exists within the ADF. Whilst there was slight variation in the terminology utilised, when comparing the pelvic health survey and Women in Defence report [32], Fig. 2 demonstrates similar patterns in the distributions of roles of women within the ADF.Fig. 2Comparison of ADF occupational categorisations between women participating in pelvic health survey and Women in Defence report 2017–2018. Admin - administration; Comms - communications; IT - information technology. ** Note: slight differences are noted between categories used in Women in Defence [2017-18] report and the survey. Within the pelvic health survey, women were also able to report more than one occupational category To explore the sample for potential nonresponse bias—or, in other words, to appraise whether women without LUTS were potentially less likely to participate in the survey than women with symptoms—we compared the prevalence data for urinary incontinence in this survey with that reported in the literature for Australian women (12.8–$46\%$) [6] as well as for female athletes ($36\%$) [33] and military women (8–$30\%$) [18]. Within the current pelvic health survey, SUI (> 1/week) was found to affect $23\%$ of respondents ($95\%$ CI 19–$27\%$) and UUI (> 1/week) $16\%$ of respondents ($95\%$ CI 13–$20\%$), with $27\%$ of survey respondents ($95\%$ CI 22–$30\%$) reporting at least one type of urinary incontinence and $13\%$ ($95\%$ CI 10–$17\%$) more than one (MUI). Despite the prevalence rates in the current survey falling within or below those reported in the broader literature, the potential for non-response bias remains. Despite being a long questionnaire, strong completion rates for individual questions about prevalence were demonstrated throughout the survey (> $80\%$). Specifically, all questions associated with LUTS prevalence were answered by > $90\%$ of survey participants (range: 90.8–$93.5\%$ question completion rate). Additional characteristics of the survey cohort (Table 3) demonstrate that respondents had a wide range of service experience, were of varied ranks, and had actively participated in field and deployment opportunities. Details are also provided about their health and regular medication use. Table 3Demographic data for respondents, as entire cohort and by current activity statusParticipant attributesAll respondents($$n = 491$$)Activity statusActively serving($$n = 299$$)Veteran($$n = 192$$)Mean (range) Age (years)42 (19–78)38 (19–63)48 (20–78)Service arm Navy105 ($22\%$)62 ($21\%$)43 ($22\%$) Army258 ($53\%$)144 ($48\%$)114 ($59\%$) Air Force126 ($26\%$)91 ($30\%$)35 ($18\%$)Service years < 10 years191 ($39\%$)94 ($31\%$)97 ($51\%$) 10–19 years167 ($34\%$)112 ($38\%$)55 ($29\%$) > 20 years133 ($27\%$)93 ($31\%$)40 ($21\%$)Rank Commissioned Officer172 ($35\%$)138 ($46\%$)34 ($18\%$) NCO/WO*167 ($34\%$)98 ($33\%$)69 ($36\%$) Other rank142 ($29\%$)59 ($20\%$)83 ($43\%$) Cadet/Trainee/Recruit10 ($2\%$)4 ($1\%$)6 ($3\%$)Participated in field activities397 ($81\%$)241 ($81\%$)156 ($81\%$)Experienced deployment342 ($70\%$)236 ($79\%$)106 ($55\%$) Australia166 ($34\%$)112 ($38\%$)54 ($28\%$) Overseas287 ($59\%$)208 ($70\%$)79 ($41\%$)Experienced pregnancy Yes281 ($57\%$)164 ($55\%$)117 ($61\%$) No114 ($23\%$)78 ($26\%$)36 ($19\%$) Unspecified96 ($20\%$)57 ($19\%$)39 ($20\%$)Health during service Nil issues165 ($34\%$)108 ($36\%$)57 ($30\%$) Back/hip pain315 ($64\%$)192 ($64\%$)123 ($64\%$) Other musculoskeletal issues76 ($16\%$)37 ($12\%$)39 ($20\%$) Respiratory conditions96 ($20\%$)47 ($16\%$)49 ($26\%$) BMI > 25122 ($25\%$)79 ($26\%$)43 ($22\%$) Neurological conditions11 ($2\%$)5 ($2\%$)6 ($3\%$) Metabolic conditions6 ($1\%$)3 ($1\%$)3 ($2\%$) Inflammatory conditions27 ($6\%$)10 ($3\%$)17 ($9\%$) Vascular conditions7 ($1\%$)2 ($1\%$)5 ($3\%$) Psychological conditions88 ($18\%$)42 ($14\%$)46 ($24\%$) Eating disorder6 ($1\%$)4 ($1\%$)2 ($1\%$) Food allergy/intolerance82 ($17\%$)54 ($18\%$)28 ($15\%$) Other7 ($1\%$)2 ($1\%$)5 ($3\%$)Medications (regular use in service) No medications101 ($21\%$)67 ($22\%$)34 ($18\%$) Antihistamines72 ($15\%$)42 ($14\%$)30 ($16\%$) Blood pressure12 ($2\%$)7 ($2\%$)5 ($3\%$) Anti-inflammatory141 ($29\%$)86 ($29\%$)55 ($29\%$) Anti-depressants50 ($10\%$)24 ($8\%$)26 ($14\%$) Anti-psychotics3 ($1\%$)1 ($0.3\%$)2 ($1\%$) Sleeping tablets28 ($6\%$)17 ($6\%$)11 ($6\%$) Strong pain medications80 ($16\%$)36 ($12\%$)44 ($23\%$) Antibiotics59 ($12\%$)19 ($6\%$)40 ($21\%$)Smoker (during service)61 ($12\%$)26 ($9\%$)35 ($18\%$)*NCO/WO: non-commissioned officer/warrant officer; BMI: body mass index ## Survey findings Over one third of participants ($$n = 186$$, $38\%$) reported that they did not regularly experience any of the LUTS included within the survey. Table 4 provides a breakdown of the prevalence data for servicewomen and female veterans not experiencing specific LUTS.Table 4Prevalence of ADF servicewomen and female veterans not experiencing LUTS during active serviceNo reported symptomsAll respondents($$n = 491$$)Actively serving($$n = 299$$)Veteran($$n = 192$$)Urinary continence No reported SUI136 ($28\%$)86 ($29\%$)50 ($26\%$) No reported UUI178 ($36\%$)114 ($38\%$)64 ($33\%$) No reported MUI210 ($43\%$)136 ($46\%$)74 ($39\%$)Bladder storage Urinary frequency (< 5/shift)*339 ($69\%$)203 ($68\%$)136 ($71\%$) No urinary urgency138 ($28\%$)90 ($30.1\%$)48 ($25\%$) No nocturia391 ($80\%$)239 ($80\%$)152 ($79\%$)Voiding No emptying concerns360 ($73\%$)223 ($75\%$)137 ($71\%$) No bladder straining445 ($91\%$)273 ($91\%$)172 ($90\%$) No urinary flow concerns425 ($87\%$)261 ($87\%$)164 ($85\%$)Pain No bladder pain emptying430 ($88\%$)273 ($91\%$)157 ($82\%$)SUI, stress urinary incontinence; UUI, urge urinary incontinence; MUI, mixed urinary incontinence. * Normal urinary frequency defined for purposes of prevalence estimates as the need to urinate fewer than five times within an 8-h working shift The prevalence data for LUTS are presented in Table 5 and demonstrate that a variety of LUTS were commonly experienced by women during their military service. Within this cohort of servicewomen and female veterans, $21\%$ of respondents ($$n = 105$$) reported regularly experiencing one LUTS, and $41\%$ ($$n = 200$$) regularly experienced two or more LUTS. A similar pattern of prevalence during service for each symptom, other than UUI, was observed for active servicewomen and veterans. Therefore, combined data are reported in further analyses. Table 5Prevalence of regular LUTS during service in the surveyed cohort of ADF servicewomen and female veteransLUTSAll respondents($$n = 491$$)Actively serving($$n = 299$$)Veteran($$n = 192$$)Urinary incontinence SUI (> 1/week)112 ($23\%$)61 ($20\%$)51 ($27\%$) UUI (> 1/week)78 ($16\%$)32 ($11\%$)46 ($24\%$) MUI (> 1/week)57 ($13\%$)22 ($7\%$)35 ($20\%$)Bladder storage Urinary frequency (5+/shift)*120 ($24\%$)74 ($25\%$)46 ($24\%$) Urinary urgency (> 1/week)131 ($27\%$)76 ($25\%$)55 ($29\%$) Nocturia100 ($20\%$)60 ($20\%$)40 ($21\%$)Voiding Incomplete emptying131 ($27\%$)76 ($25\%$)55 ($29\%$) Bladder straining46 ($9\%$)26 ($9\%$)20 ($10\%$) Weak urinary flow66 ($13\%$)38 ($13\%$)28 ($15\%$)Pain Painful bladder emptying61 ($12\%$)26 ($9\%$)35 ($18\%$)SUI, stress urinary incontinence; UUI, urge urinary incontinence; MUI, mixed urinary incontinence. Urinary frequency defined for purposes of prevalence estimates as the need to urinate five or more times within an 8-h working shift Almost one quarter of respondents ($23\%$, $95\%$ CI 19–$27\%$) reported they experienced SUI more than once per week during service (Fig. 3a). However, when including all women who experienced episodes of SUI even ‘occasionally’ or ‘sometimes’ during activities at work, SUI affected up to $63\%$ of respondents ($95\%$ CI 58–$67\%$). The volume of leakage reported during episodes of SUI was most commonly ‘slightly wet underwear, liner or pad’ ($49\%$) or ‘a dribble’ ($42\%$). However, for one in every ten women who reported SUI, episodes of leakage were reported to require a complete change of their underwear, pad, or clothing. The most common activities contributing to episodes of SUI were those involving an increase in intra-abdominal pressure (e.g., sneezing, coughing) and vertical ground reaction force loading (e.g., running, jumping) (Fig. 3b). Similarly, only $16\%$ ($95\%$ CI 13–$20\%$) and $13\%$ ($95\%$ CI 10–$16\%$) of respondents experienced regular episodes of UUI or MUI, respectively (Fig. 3a), but this increased to $55\%$ ($95\%$ CI 50–$59\%$) and $53\%$ ($95\%$ CI 48–$57\%$), when occasional episodes were included in the prevalence rate. Fig. 3Prevalence of specific lower urinary tract symptom frequency (a) and activities contributing to symptoms of SUI (b) Many servicewomen and female veterans reported that their LUTS were “current and ongoing issues”. LUTS such as SUI, UUI, and UU were reported to be ongoing issues for two thirds of active servicewomen (SUI = $66\%$ $95\%$ CI 60–$71\%$; UUI = $63\%$ $95\%$ CI 57–$68\%$; UU = $69\%$ $95\%$ CI 63–$74\%$), and over three quarters of veterans (SUI = $76\%$ $95\%$ CI 70–$82\%$; UUI = $82\%$ $95\%$ CI 76–$87\%$; UU = $78\%$ $95\%$ CI 72–$83\%$). As expected, due to the older age of the veteran cohort, it was more common for veterans to report they had experienced LUTS for a longer duration, commonly in excess of 10 years (supplementary data). Comparisons of LUTS prevalence rates by service arm, employment rank, age group, and parity groups are presented in Figs. 4, 5, and 6. A similar pattern of prevalence was seen for all urinary symptoms, except UUI (and as a result, MUI), in service comparisons (Fig. 4a), whereas a trend towards higher prevalence rates of UI and UU symptoms were reported by women in Non-Commissioned Officer/Warrant Officer ranks (Fig. 4b). In active servicewomen, symptoms of UUI and UU increased in prevalence with each decade of increasing age, whereas SUI prevalence increased up until mid-life (40–49 years) and then stabilised (Fig. 5). Figure 6 demonstrates a clear relationship between having a history of pregnancy and the experience of symptoms such as SUI, UUI, MUI, and UU.Fig. 4Urinary symptoms for all respondents presented by (a) service arm and (b) employment rank. SUI - stress urinary incontinence more than once per week; UUI – urge urinary incontinence more than once per week; MUI- mixed urinary incontinence more than once per week; UFreq - urinary frequency greater than five times during an eight-hour shift; UU - urinary urgency more than once per week; CO – commissioned officer; NCO/WO – non-commissioned officer/warrant officerFig. 5Prevalence of LUTS in active servicewomen by age category**. SUI - stress urinary incontinence more than once per week; UUI – urge urinary incontinence more than once per week; MUI- mixed urinary incontinence more than once per week; UFreq - urinary frequency greater than five times during an eight-hour shift; UU - urinary urgency more than once per week. ** Age category reporting is not presented for female veterans, as their age during last period of active service was not collected – only their age at time of completing the surveyFig. 6History of pregnancy and prevalence of urinary symptoms. SUI - stress urinary incontinence more than once per week; UUI – urge urinary incontinence more than once per week; MUI- mixed urinary incontinence more than once per week; UFreq - urinary frequency greater than five times during an eight-hour shift; UU - urinary urgency more than once per week; Nulliparous – respondents who reported never being pregnant; Parous – respondents who reported being pregnant one or more times Additional comparisons were targeted to variables that may have previously demonstrated a relationship with LUTS and had reasonable sub-group sample sizes (BMI [34], respiratory conditions [35], and lower back/hip pain [36, 37]). The prevalence of bladder storage, voiding, and pain symptoms were higher in servicewomen and veterans also reporting regular episodes of lower back/hip pain, but no clear trends were seen for the prevalence of LUTS in women associated with higher BMI or respiratory conditions (supplementary data). ## Discussion This was the first known study to explore the prevalence of LUTS in a cohort of Australian female military personnel and veterans. Approximately 1 in every 100 current female personnel and veterans responded, with the profile of respondents similar to the population characteristics of female ADF personnel reported in the Women in the ADF Report 2017–2018 and Defence Census 2019. The survey found regular LUTS were common for female personnel during active military service, and the patterns of LUTS were similar to those reported in the broader female population [6–8]. Australian servicewomen and female veterans commonly reported experiencing one or more LUTS during their military service. Urinary incontinence symptoms of different types affected 13–$23\%$ of these women, bladder storage symptoms 20–$27\%$, voiding symptoms 9–$27\%$, and bladder emptying pain $12\%$ of respondents. Prevalence statistics for LUTS in the general female population are frequently reported for symptoms of urinary incontinence, and the rates identified within this survey among Australian military women and female veterans are consistent with those for Australian women more broadly [6–8]. However, it is important to recognise that prevalence rates are influenced by the definitions used and need to be interpreted in that context. To prevent overinflating prevalence statistics [5], within this study primary calculations of prevalence rates were based on numbers of women reporting consistent and frequent symptoms (more than once/week). However, when all levels of symptom experience were considered, LUTS were reported to occur in between 50–$70\%$ of respondents (depending on symptom type). Whilst these rates do not provide insight to the level of impact on well-being or occupation, they do highlight that it is highly likely that female personnel will experience one or more types of LUTS at some point during their military service and that at least one fifth will experience LUTS of some type consistently. Similar to the general female population and previous studies of female military personnel [13], age and parity status were associated with prevalence rates of LUTS in the military women and female veterans who responded to this survey—particularly symptoms of urinary incontinence. Consistent with findings reported in the broader literature, SUI was more common in younger women, with prevalence peaking at mid-life, whereas UUI rates were shown to steadily increase with age [5, 8]. Recognising the influence that age and parity may have on LUTS highlights that female military personnel may have different genitourinary health risks and needs at varying times within their period of military service. Other research has similarly demonstrated there is variation in the rates of occurrence of different types of LUTS experienced by females across their lifespan [3]. These findings suggest that genitourinary health assessment, monitoring, and support provided to servicewomen should be tailored to specific needs of each servicewoman during their time in service and on separation. Examples of such health care could include screening programmes and education at time of recruitment, post-partum return to work planning, pre-deployment preparation, and support around menopause. Adequate and tailored support may contribute to reducing rates of separation that may be potentially occurring from LUTS that are negatively impacting on occupational performance. Occupational elements altering bladder behaviours at work have been suggested as potential factors influencing the prevalence of female LUTS [12]. Despite being employed in varied occupational roles, all military personnel are required to undertake regular organised and mandated physical and military training. However, the fitness requirements for each service are different, the duration and frequency of physical loading will vary between roles, military personnel may engage in multiple roles throughout their service, and the workplace environment also changes depending on the posting. Therefore, the influence of military occupational factors on the prevalence of female LUTS is likely to be challenging to determine. Within this study, similar patterns of LUTS prevalence were observed for Naval, Army, and Air Force women, suggesting that the Service arm was not a factor influencing LUTS. Contrarily, employment rank may have a relationship with LUTS, as a trend towards higher urinary incontinence and urgency prevalence rates was reported for female Non-Commissioned Officers/Warrant Officers. It is likely that confounding variables, such as age, parity, and length of service, may contribute to this finding, as progression through the ranks is likely to take some time. For the majority of servicewomen and female veterans reporting LUTS, their symptoms were identified to be “current and ongoing”. Whilst active servicewomen with LUTS were more likely to report that they had had their symptoms < 6 years, close to one in five active servicewomen and almost half the female veterans with LUTS reported symptom durations > 10 years. These findings concur with longitudinal studies of women with urinary incontinence, which have found that symptoms can persist and fluctuate significantly, with periods of remission, and reduced or increased symptoms [5]. The presence of persistent LUTS reinforces the potential benefits of targeted support for servicewomen to prevent the cumulative effects of long-term LUTS on occupational performance. However, it must be acknowledged that the survey questions did not explore the level of symptom fluctuation experienced by respondents or remission or cure rates within the cohort. The frequencies of LUTS experienced by servicewomen and female veterans provide insight into the severity of the problem, identifying that a substantial number of servicewomen experience symptoms more than once per week. In addition, other dimensions of symptom characteristics provide insight into severity. For example, the amount of urinary leakage in those experiencing incontinence is another indicator. Consistent with previous research involving women with urinary incontinence [7], most affected respondents reported mild (a “dribble”) to moderate (“slightly wet underwear”) levels of leakage. Despite most experiencing mild to moderate symptoms, the occupational impacts of these symptoms are not fully reflected in the prevalence data. LUTS have been reported to reduce work productivity [12] and reduce female physical exercise participation rates, with those who experience severe symptoms being more likely to cease exercise [38]. Therefore, even mild to moderate symptoms may present challenges for females in military contexts, such as in the field or on deployment, and may influence safety and work performance. For example, servicewomen with urinary incontinence have been found to manage symptoms through fluid restriction and altering their voiding patterns [15], which may increase their risk of developing heat related illness. The prevalence findings from this study highlight that urinary health should be monitored, and personnel offered support to maintain and manage it at work. The findings presented here provide useful understanding of the types, prevalence, and severity of LUTS in Australian servicewomen and female veterans, but the data presented do not provide insights into the impacts of LUTS for female personnel within occupational settings, how servicewomen manage LUTS in military work environments, the influence of symptoms on occupational performance, or the military support available and utilised by servicewomen. Due to the importance of these questions and gaining more in-depth understanding of the impact of pelvic health more broadly (not just limited to LUTS) on the occupational experiences of servicewomen, data from the pelvic health survey associated with work performance are not presented within this report. Whilst the findings of the survey provide useful insights into LUTS for female military personnel within Australia, it is important to acknowledge limitations of the study. Self-report surveys can be impacted by recall bias, which was a particular concern for the cohort of female veterans due to their need to remember historical LUTS, experienced during their last period of active service. The original study criteria excluded female veterans from participating in the survey for this reason. However, feedback was received from the servicewomen consulting on the survey that the female veteran population strongly wanted to participate and share their experiences. Despite this limitation, the similarities in prevalence rates for LUTS reported by servicewomen and female veterans suggest that recall bias may not have had a substantial impact on the findings of the survey. Due to the breadth of the pelvic health survey, completion of the survey was time-consuming, and this may have contributed to the large number of incomplete survey responses needing to be excluded from data analysis. This increased the potential for non-response bias or systematic differences in LUTS between those who did and did not respond to the survey. Consultation with potential participants occurred during survey design to ensure the survey length was acceptable, and data have been presented to allow judgements on the representativeness of the sample to be considered. Finally, generalisation of the survey findings should continue to be done with caution because of the use of non-probability sampling methods [31]. ## Conclusion The survey found that LUTS were commonly experienced by ADF female military personnel during service and that the types of LUTS, their prevalence rates, and the levels of severity of symptoms were similar to those reported in Australian females more broadly. Consistent with wider research, relationships between LUTS and age and parity were also demonstrated in this cohort of servicewomen and female veterans. Given that LUTS may impact on occupational performance and that most ADF servicewomen are likely to experience LUTS at some time during their service, with a proportion experiencing ongoing symptoms, tailored monitoring and support services should be available to servicewomen. These support systems will need to vary in type across the period of service for each individual and in response to variations in work context. ## Supplementary information ESM 1(PDF 203 kb) ## References 1. Haylen BT, De Ridder D, Freeman RM. **An International Urogynecological Association (IUGA)/International Continence Society (ICS) joint report on the terminology for female pelvic floor dysfunction**. *Neurourol Urodyn* (2010.0) **29** 4-20. DOI: 10.1002/nau.20798 2. Maserejian NN, Chen S, Chiu GR. **Incidence of lower urinary tract symptoms in a population-based study of men and women**. *Urology* (2013.0) **82** 560-564. DOI: 10.1016/j.urology.2013.05.009 3. Brady SS, Berry A, Camenga DR, Fitzgerald CM, Gahagan S. **Applying concepts of life course theory and life course epidemiology to the study of bladder health and lower urinary tract symptoms among girls and women**. *Neurourol Urodyn* (2020.0) **39** 1185-1202. DOI: 10.1002/nau.24325 4. 4.Sutcliffe S, Bavendam T, Cain C, et al. The spectrum of bladder health: The relationship between lower urinary tract symptoms and interference with activities. J Women's Health. 2019;28(6):827–841. 5. Minassian VA, Bazi T, Stewart WF. **Clinical epidemiological insights into urinary incontinence**. *Int Urogynecol J* (2017.0) **28** 687-696. DOI: 10.1007/s00192-017-3314-7 6. Botlero R, Urquhart DM, Davis SR, Bell RJ. **Prevalence and incidence of urinary incontinence in women: review of the literature and investigation of methodological issues**. *Int J Urolo: Off J Japanese Urol Assoc* (2008.0) **15** 230-234. DOI: 10.1111/j.1442-2042.2007.01976.x 7. 7.Hawthorne G. Measuring incontinence in Australia. The Department of Psychiatry, The University of Melbourne. 2006; Department of Health and Ageing (DoHA). 8. Botlero R, Davis SR, Urquhart DM, Shortreed S, Bell RJ. **Age-specific prevalence of, and factors associated with, different types of urinary incontinence in community-dwelling Australian women assessed with a validated questionnaire**. *Maturitas* (2009.0) **62** 134-139. DOI: 10.1016/j.maturitas.2008.12.017 9. Nygaard I, Girts T, Fultz NH, Kinchen K, Pohl G, Sternfeld B. **Is urinary incontinence a barrier to exercise in women?**. *Obstet Gynecol* (2005.0) **106** 307-314. DOI: 10.1097/01.AOG.0000168455.39156.0f 10. Coyne KS, Wein AJ, Tubaro A. **The burden of lower urinary tract symptoms: evaluating the effect of LUTS on health-related quality of life, anxiety and depression: EpiLUTS**. *BJU Int* (2009.0) **103** 4-11. DOI: 10.1111/j.1464-410X.2009.08371.x 11. Pizzol D, Demurtas J, Celotto S. **Urinary incontinence and quality of life: a systematic review and meta-analysis**. *Aging Clin Exp Res* (2021.0) **33** 25-35. DOI: 10.1007/s40520-020-01712-y 12. Markland A, Chu H, Epperson CN. **Occupation and lower urinary tract symptoms in women: A rapid review and meta-analysis from the PLUS research consortium**. *Neurourol Urodyn* (2018.0) **37** 2881-2892. DOI: 10.1002/nau.23806 13. Davis G, Sherman R, Wong MF, McClure G, Perez R, Hibbert M. **Urinary incontinence among female soldiers**. *Mil Med* (1999.0) **164** 182-187. DOI: 10.1093/milmed/164.3.182 14. Orr RM, Johnston V, Coyle J, Pope R. **Reported load carriage injuries of the Australian army soldier**. *J Occup Rehabil* (2015.0) **25** 316-322. DOI: 10.1007/s10926-014-9540-7 15. Steele N, Yoder LH. **Military women's urinary patterns, practices, and complications in seployment settings**. *Urol Nurs* (2013.0) **33** 61-78. DOI: 10.7257/1053-816X.2013.33.2.61 16. Criner JA. **Urinary incontinence in vulnerable populations: female soldiers**. *Urol Nurs* (2001.0) **21** 120-124. PMID: 11998280 17. Wilson C, Nelson JP. **Exploring the patterns, practices, and experiences of military women who managed genitourinary symptoms in deployed settings**. *J Obstet Gynecol Neonatal Nurs* (2012.0) **41** 293-302. DOI: 10.1111/j.1552-6909.2012.01347.x 18. O'Shea SD, Pope R, Orr RM, Freire K. **Pelvic floor health in female military personnel: A narrative review**. *J Military Vet Health* (2021.0) **29** 69-78 19. Sedgwick P. **Cross sectional studies: advantages and disadvantages**. *BMJ* (2014.0) **348** g2276. DOI: 10.1136/bmj.g2276 20. Baessler K, O'Neill SM, Maher CF, Battistutta D. **Australian pelvic floor questionnaire: a validated interviewer-administered pelvic floor questionnaire for routine clinic and research**. *Int Urogynecol J Pelvic Floor Dysfunct* (2009.0) **20** 149-158. DOI: 10.1007/s00192-008-0742-4 21. McManus J, Mehta SG, McClinton AR, Lorenzo RAD, Baskin TW. **Informed consent and ethical issues in military medical research**. *Acad Emerg Med* (2005.0) **12** 1120-1126. DOI: 10.1197/j.aem.2005.05.037 22. Shumaker SA, Wyman JF, Uebersax JS, McClish D, Fantl JA. **Health-related quality of life measures for women with urinary incontinence: the Incontinence Impact Questionnaire and the Urogenital Distress Inventory**. *Qual Life Res* (1994.0) **3** 291-306. DOI: 10.1007/BF00451721 23. Avery K, Donovan J, Peters T, Shaw C, Gotoh M, Abrams P. **ICIQ: a brief and robust measure for evaluating the symptoms and impact of urinary incontinence**. *NeurourolUrodyn* (2004.0) **23** 322-330 24. Price N, Jackson SR, Avery K, Brookes ST, Abrams P. **Development and psychometric evaluation of the ICIQ Vaginal Symptoms Questionnaire: the ICIQ-VS**. *BJOG Int J Obstet Gynaecol* (2006.0) **113** 700-712. DOI: 10.1111/j.1471-0528.2006.00938.x 25. Erekson EA, Yip SO, Wedderburn TS. **The Vulvovaginal Symptoms Questionnaire: a questionnaire for measuring vulvovaginal symptoms in postmenopausal women**. *Menopause* (2013.0) **20** 973-979. DOI: 10.1097/GME.0b013e318282600b 26. 26.Hunt AP. Heat Stress Questionnaire (Appendix A). In: Heat strain, hydration status and symptoms of heat illness in surface mine workers. [PhD Thesis, QUT.], Queensland University of Technology; 2011. 27. Chalmers KJ, Catley MJ, Evans SF, Moseley GL. **Clinical assessment of the impact of pelvic pain on women**. *Pain* (2017.0) **158** 498-504. DOI: 10.1097/j.pain.0000000000000789 28. Rea LM, Parker RA. **Designing and Conducting Survey Research: A Comprehensive Guide**. *In* (2014.0) 29. 29.Department of DefenceDefence Census 2019: public report2019AustraliaCommonwealth of Australia. *Defence Census 2019: public report* (2019.0) 30. 30.AIHW. National suicide monitoring of serving and ex-serving Australian Defence Force personnel: 2020 update. In: Australian Government, ed. Australia. 2020. 31. 31.The American Association for Public Opinion Research. Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys.: AAPOR.;2016. 32. 32.Defence People Group. Women in the ADF report 2017 - 2018: A supplement to the Defence Annual Report 2017-2018. 2019. 33. Teixeira RV, Colla C, Sbruzzi G, Mallmann A, Paiva LL. **Prevalence of urinary incontinence in female athletes: a systematic review with meta-analysis**. *Int Urogynecol J* (2018.0) **29** 1717-1725. DOI: 10.1007/s00192-018-3651-1 34. Lai HH, Helmuth ME, Smith AR, Wiseman JB, Gillespie BW, Kirkali Z. **Relationship between central obesity, general obesity, overactive bladder syndrome and urinary incontinence among male and female patients seeking care for their lower urinary tract symptoms**. *Urology* (2019.0) **123** 34-43. DOI: 10.1016/j.urology.2018.09.012 35. Button BM, Holland AE, Sherburn MS, Chase J, Wilson JW, Burge AT. **Prevalence, impact and specialised treatment of urinary incontinence in women with chronic lung disease**. *Physiotherapy* (2019.0) **105** 114-119. DOI: 10.1016/j.physio.2018.07.006 36. Cassidy T, Fortin A, Kaczmer S, Shumaker JTL, Szeto J, Madill SJ. **Relationship between back pain and urinary incontinence in the Canadian population**. *Phys Ther* (2017.0) **97** 449-454. DOI: 10.1093/ptj/pzx020 37. Ludwik C, Mills T, Frawley H. **Prevalence of urinary and faecal incontinence in men and women with chronic low back pain**. *Australian & New Zealand Continence J* (2019.0) **25** 58-63 38. Dakic JG, Hay-Smith J, Kuan-Yin L, Calo M, Cook J, Frawley H. **Effect of pelvic floor symptoms on women’s participation in exercise: A mixed-methods systematic review with meta-analysis**. *J Orthop Sports Phys Ther* (2021.0) **51** 345-361. DOI: 10.2519/jospt.2021.10200
--- title: Carnosine increases insulin-stimulated glucose uptake and reduces methylglyoxal-modified proteins in type-2 diabetic human skeletal muscle cells authors: - Joseph J. Matthews - Mark D. Turner - Livia Santos - Kirsty J. Elliott-Sale - Craig Sale journal: Amino Acids year: 2023 pmcid: PMC10038967 doi: 10.1007/s00726-022-03230-9 license: CC BY 4.0 --- # Carnosine increases insulin-stimulated glucose uptake and reduces methylglyoxal-modified proteins in type-2 diabetic human skeletal muscle cells ## Abstract Type-2 diabetes (T2D) is characterised by a dysregulation of metabolism, including skeletal muscle insulin resistance, mitochondrial dysfunction, and oxidative stress. Reactive species, such as methylglyoxal (MGO) and 4-hydroxynonenal (4-HNE), positively associate with T2D disease severity and can directly interfere with insulin signalling and glucose uptake in skeletal muscle by modifying cellular proteins. The multifunctional dipeptide carnosine, and its rate-limiting precursor β-alanine, have recently been shown to improve glycaemic control in humans and rodents with diabetes. However, the precise mechanisms are unclear and research in human skeletal muscle is limited. Herein, we present novel findings in primary human T2D and lean healthy control (LHC) skeletal muscle cells. Cells were differentiated to myotubes, and treated with 10 mM carnosine, 10 mM β-alanine, or control for 4-days. T2D cells had reduced ATP-linked and maximal respiration compared with LHC cells ($$p \leq 0.016$$ and $$p \leq 0.005$$). Treatment with 10 mM carnosine significantly increased insulin-stimulated glucose uptake in T2D cells ($$p \leq 0.047$$); with no effect in LHC cells. Insulin-stimulation increased MGO-modified proteins in T2D cells by $47\%$; treatment with carnosine attenuated this increase to $9.7\%$ ($$p \leq 0.011$$). There was no effect treatment on cell viability or expression of other proteins. These findings suggest that the beneficial effects of carnosine on glycaemic control may be explained by its scavenging actions in human skeletal muscle. ### Supplementary Information The online version contains supplementary material available at 10.1007/s00726-022-03230-9. ## Introduction Diabetes is a major public health problem; worldwide estimates show that 537 million people aged 20–79 years were living with diabetes in 2021, equivalent to $10.5\%$ of the global adult population (Sun et al. 2022), with type-2 diabetes (T2D) accounting for > $90\%$ of cases. T2D is characterised by a dysregulation of metabolism due to impaired insulin secretion, insulin resistance, or a combination of both (DeFronzo et al. 2015), which develops from impaired fasting glucose or impaired glucose tolerance (also known as prediabetes). Skeletal muscle insulin resistance and mitochondrial dysfunction are hallmarks of T2D, and although the aetiology is multifactorial, emerging research implicates oxidative and carbonyl stress as causative factors (Mey and Haus 2018). Reactive carbonyl species (RCS) [e.g., methylglyoxal (MGO)] and reactive aldehydes [e.g., 4-hydroxynonenal (4-HNE)]; positively associate with disease severity and are elevated in the post-prandial and hyperinsulinaemic state (Mey et al. 2018; Neri et al. 2010). MGO and 4-HNE can directly interfere with insulin signalling in skeletal muscle (Pillon et al. 2012; Riboulet-Chavey et al. 2006) and form adducts with proteins, which modifies their activity and leads to the downstream formation of harmful advanced lipid peroxidation (ALEs) and advanced glycation (AGEs) end-products—further exacerbating T2D and diabetic complications (for a review, see Brings et al. 2017). There is, therefore, a need to develop novel interventions that reduce oxidative and carbonyl stress to help delay or prevent disease development and progression. Carnosine (β-alanyl-L-histidine) is a multifunctional dipeptide with an emerging role in metabolic health and disease (Artioli et al. 2019). It exists in high concentrations in human skeletal muscle (approx. 22 mmol kg−1 dw−1; 5 mmol kg−1 ww−1), which can be increased by up to twofold with prolonged supplementation of its rate-limiting precursor, β-alanine (Matthews et al. 2019; Saunders et al. 2017). Dietary sources include prawns, tuna, mackerel, poultry, and red meats; providing ~ 300–550 mg.day−1 β-alanine (Baguet et al. 2011; Kratz et al. 2017). Importantly, our recent meta-analysis suggested that supplementation with carnosine or β-alanine reduces fasting glucose and glycated haemoglobin (HbA1c) in humans and rodents (Matthews et al. 2021). While this is promising, there are issues with risk of bias and study quality, and the site and mechanism of action remain unclear. The beneficial effects could be due to carnosine forming stable adducts with reactive species, thereby reducing their reactivity and allowing them to be safely metabolised or excreted from the body (Baba et al. 2013; Regazzoni et al. 2016; Szwergold 2005). Consistent with this, evidence from human and rodent studies shows that carnosine supplementation protects against oxidative stress, lipid peroxidation, and AGE and ALE formation (Albrecht et al. 2017; Aldini et al. 2011; Elbarbary et al. 2018; Houjeghani et al. 2018). Mechanistic research from our group showed that C2C12 skeletal muscle cells cultured under glucolipotoxic (GLT) conditions (a model of metabolic stress) had reduced glucose uptake and higher reactive oxygen species (ROS) (Cripps et al. 2017), as well as a suppression of GLUT4 translocation and cellular respiration (ATP-linked and maximal O2 consumption) (Lavilla et al. 2021). In both studies, treatment with 10 mM carnosine led to a partial recovery of glucose uptake and GLUT4 translocation, a near full recovery of cellular respiration, and normalisation of ROS (Cripps et al. 2017; Lavilla et al. 2021). Consistent with its role as a scavenger of reactive species, carnosine prevented $90\%$ and $80\%$ of 4-HNE protein adduction in C2C12 and human skeletal muscle cells (Lavilla et al. 2021). It is also possible that carnosine acts indirectly by activating endogenous anti-carbonylation, e.g., glyoxalase 1 (GLO1) defence systems (Aldini et al. 2021), or by increasing the expression of metabolic proteins involved in regulating mitochondrial health e.g., NAD-dependent deacetylase sirtuins $\frac{1}{3}$ (Sirt1, Sirt3) or peroxisome proliferator-activated receptor gamma coactivator a-alpha (PCG-1α), which have been linked to improvements in insulin signalling and oxidative stress (Jing et al. 2011; Pagel-Langenickel et al. 2008). These actions, coupled with the ability to increase tissue carnosine stores through diet, makes it a promising therapeutic for T2D and prediabetes. Herein we extend our previous work by reporting novel investigations on the effect of carnosine and β-alanine on cellular respiration, glucose uptake, and carbonyl and aldehyde-modified proteins in primary T2D and healthy human skeletal muscle cells. ## Methods Cell culture work was performed in a Class II laminar flow hood under aseptic conditions. For subculture and experiments, cells were kept at 37 °C and $5\%$ CO2 in a humidified incubator, and all reagents and plasticware were purchased from Thermo Fisher Scientific (Loughborough, UK), unless otherwise stated. Growth, differentiation, and treatment media were changed every other day or as part of general sub-culture procedures. ## Human skeletal muscle cell culture Human skeletal myoblasts isolated from lean healthy control (LHC; male, 20y, BMI 21 kg.m2, lot no. 639629) and obese type-2 diabetic (T2D; male, 68y, BMI 33 kg.m2, lot no. 211384) donors were purchased from Lonza Bioscience (Basel, Switzerland). Cells were pre-screened to display ≥ $60\%$ desmin-positive cells at first passage and tested negative for mycoplasma, bacteria, yeast, virus, and fungi. Myoblasts were cultured in skeletal muscle growth medium (PromoCell, Germany): basal medium, 5.5 mM glucose, supplemented with $10\%$ fetal bovine serum (FBS), 50 µg/mL bovine fetuin, 10 µg/mL human insulin, 10 ng/mL human epidermal growth factor, 1 ng/mL human basic fibroblast growth factor, and 0.4 µg/mL dexamethasone. Myoblasts were differentiated to myotubes by changing growth medium to DMEM:F-12 supplemented with $2\%$ horse serum and $0.5\%$ penicillin–streptomycin for 4 to 5 days. Experiments were performed with cells at passages 3–7 due to reduced myogenic potential at higher passages. Myotubes were treated for 4 days with vehicle (control), 10 mM carnosine, or 10 mM β-alanine (Sigma-Aldrich, UK) dissolved in low-glucose DMEM, DMEM:F-12, or GLT DMEM:F-12. GLT media was used for some experiments as an extracellular model of poorly controlled type-2 diabetes (DMEM:F-12 containing 17 mM glucose with added 200 µM palmitic acid and 200 µM oleic acid). DMEM:F12 was supplemented with $2\%$ fatty acid free bovine serum albumin (BSA); stock solutions of 100 mM palmitic acid, dissolved in $100\%$ ethanol, and 100 mM sodium oleate, dissolved in $50\%$ ethanol, were heated to 60 °C and added directly to the BSA (Sigma-Aldrich, UK). The media was then incubated at 37 °C for 1 h to allow fatty acid conjugation, before being sterile filtered through 0.2 µm membrane filters (Lavilla et al. 2021; Marshall et al. 2007). ## Cell viability Cell viability was measured using alamarBlue™. Cells were seeded at 2 × 104 cells/cm2 in 24-well plates, differentiated, treated for 4-days, then washed with Dulbecco’s phosphate buffered saline (DPBS) and incubated with differentiation medium ± $10\%$ (v/v) alamarBlue™ for 1-h at 37 °C and $5\%$ CO2. Samples were transferred to black 96-well plates and fluorescence recorded (excitation 570 nm, emission 585 nm) on a spectrophotometer (CLARIOstar Plus, BMG Labtech, Germany). Results were corrected for background fluorescence by subtracting negative control wells from experimental wells and are expressed as fold change relative to control wells. There was no significant effect of treatment on cell viability (Supplementary Fig. 1). Because of this, the 10 mM treatment concentration was used for subsequent experiments, as this is near to the upper limit of intramuscular skeletal muscle carnosine stores after supplementation with its precursor, β-alanine, in vivo (Saunders et al. 2017). ## Measurement of O2 consumption Oxygen consumption rate (OCR) was measured using a Seahorse XFe24 Analyser and XF Cell Mito Stress Test reagents (Seahorse Bioscience Inc., USA). Myoblasts were seeded at 3 × 104 cells per well in XF24 polystyrene V7 microplates, differentiated, and treated for 4-days. The day before the assay, a Seahorse XF sensor cartridge was hydrated with XF pH 7.4 calibrant and kept in a 37 °C non-CO2 incubator overnight. For the assay, cells were washed with DPBS and treated with bicarbonate-free Seahorse XF DMEM supplemented with 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose, and kept in a 37 °C non-CO2 incubator for 45 min. Carnosine, β-alanine, or vehicle (control) were added to the assay media and pH adjusted to 7.4 ± 0.05. To reduce common variability issues, we allocated 6–7 wells per condition per assay. Metabolic inhibitors were dissolved in assay media and loaded into the sensor cartridge using concentrations that were optimised for each cell type in prior assays. Baseline OCR was measured 4 times for 3 min. Following which, 0.8 µM oligomycin was added to each well to quantify state four (non-phosphorylating) respiration and OCR was measured 3 times for 3 min. Then, 5 µM carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) was added to each well to quantify maximal respiration and OCR was measured 4 times for 3 min each. For the final step, 0.5 µM rotenone/antimycin A was added to each well to quantify non-mitochondrial respiration and OCR was measured 3 times for 3 min each. All measurements were separated by a 2-min wait and 2-min mix. Once completed, assay media was aspirated and protein extracted from each well, which was used to normalise OCR values (pmol.min−1.µg total protein). Included wells were corrected for background OCR by subtracting negative control wells and averaged to provide a single OCR value at each measurement point for each condition. Total area-under-the-curve for O2 consumption was measured as previously described (Narang et al. 2020). ## Measurement of glucose uptake Glucose uptake was measured in myotubes using a bioluminescent assay based on the detection of 2-deoxyglucose-6-phosphate (2DG6P) (Promega, UK). Cells were seeded at 2 × 104 cells/cm2 in 24-well plates, differentiated, treated for 4-days, and serum-starved overnight in low-glucose 5.5 mM DMEM ± treatment. For the assay, cells were washed with DPBS and incubated with glucose-free DMEM ± 100 nM human insulin (Sigma-Aldrich, UK) for 1-h at 37 °C and $5\%$ CO2. The media was then replaced with DPBS containing 200 µM 2DG for 30 min at 37 °C and $5\%$ CO2. For negative control wells 50 µM of cytochalasin B (Sigma-Aldrich, UK), an inhibitor of GLUT-dependent glucose uptake, was added 5 min before the addition of 2DG. Following incubation, 0.4 M HCl + $2\%$ dodecyl trimethyl ammonium bromide (stop buffer) was added to lyse cells, terminate uptake, and eradicate intracellular NADPH, then a high-pH buffer solution (neutralisation buffer) added. For the final step, the 2DG6P detection reagent was added to each well and incubated at room temperature for 1-h. Data were acquired for luminescence using a 0.3–1 s integration on a spectrophotometer (CLARIOstar Plus, BMG Labtech, Germany). Negative control wells were subtracted from experimental wells and relative light units converted to glucose uptake (nmol min−1) using a 2DG standard curve after correcting for assay duration and dilution factor. ## Protein extraction and measurement Myotubes were washed with DPBS and extracted using a cell scraper and centrifuged for 5 min at 300g. Cells were then incubated with ice-cold radioimmunoprecipitation buffer (Pierce®, Thermo Fisher, UK) with a protease inhibitor cocktail (Halt™, Thermo Fisher, UK); agitated for 20 min at 4 °C and centrifuged for 15 min at 4 °C and 13,000g. Protein content was determined via bicinchoninic acid assay with absorbance recorded at 562 nm using a spectrophotometer (CLARIOstar Omega, BMG Labtech, Germany). Negative control wells were subtracted from experimental wells; and data normalised to standard curve values. Protein expression was measured via Western blotting SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel electrophoresis) according to published recommendations (Bass et al. 2017; Mahmood and Yang 2012). Equipment and consumables were from Bio-Rad (Bio-Rad Laboratories Inc., UK), unless otherwise stated. Approx. 20–25 µg total denatured proteins were separated in precast gels with running buffer (25 mM Tris, 190 mM glycine, $0.1\%$ SDS) and transferred to methanol-activated PVDF membranes using a Trans-Blot Turbo system with transfer buffer (25 mM Tris, 190 mM glycine, $20\%$ ethanol; $0.1\%$ SDS was added for proteins larger than 80 kDa). Membranes were blocked ($3\%$ non-fat milk in TBST) at room temperature for 1-h, before incubation with primary antibodies (diluted in $3\%$ BSA in TBST) at 4 °C overnight. The following morning, membranes were incubated with HRP (horseradish-peroxidase)-conjugated secondary antibody (diluted in $5\%$ non-fat milk in TBST) at room temperature for 1 h. To measure multiple proteins of a similar molecular weight, membranes were stripped and re-probed using a mild stripping buffer (200 mM glycine, $0.1\%$ SDS, $1\%$ Tween 20, pH 2.2). Membranes were washed in Tris-buffered saline with Tween 20 (TBST) between each step. Primary antibodies: MGO (ab243074), LDHA (ab84716), Sirt1 (ab110304), Sirt3 (ab45067) (Abcam, Cambridge, MA); GLO1 (sc-133214), PGC-1α (sc-517380) (Santa Cruz Biotechnologies, Santa Cruz, CA); 4-HNE (ab46545). Secondary antibodies (Bio-Rad Laboratories Inc., UK) were selected in a species-specific manner, according to primary antibody instructions. For imaging, each membrane was incubated with Clarity Max enhanced chemiluminescence substrate for 5 min and imaged using the GeneGnome XRQ and GeneSys image acquisition software version 1.7.2. ( SynGene, Cambridge, UK). Semi-quantitative densitometry analysis of blots was performed using Image J (version 1.8.0) and data normalised to β-actin expression. MGO and 4-HNE-modified proteins were measured using the range of detected bands between 10 and 260 and 20 and 100 kDa, an approach used previously in human skeletal muscle tissue (Mey et al. 2018). Data are presented as the fold change relative to control. ## Statistical analysis Results are presented as mean ± standard error of the mean (SEM) ($$n = 3$$ or more independent experiments). Data were analysed for each experiment using a one-way ANOVA with post hoc comparisons using the Tukey (glucose uptake, protein adduction, ATP-linked, and maximal respiration) or the Bonferroni method (cell viability and OCR tAUC), with p values adjusted for multiple comparisons. Differences between absolute basal (LHC vs T2D) and absolute insulin-stimulated (LHS vs T2D) glucose uptake were analysed using two-tailed unpaired t-tests. Analyses were performed using Prism v9.1.0 (GraphPad Software, USA); with p-values < 0.05 considered statistically significant. ## Cellular respiration T2D and T2D-GLT cells had significantly reduced ATP-linked and maximal respiration compared with LHC cells (T2D $$p \leq 0.016$$ and $$p \leq 0.005$$; T2D + GLT $$p \leq 0.0028$$ and $$p \leq 0.003$$; Fig. 1A, B). There were no statistically significant differences between T2D and T2D + GLT. ATP-linked and maximal respiration were not statistically affected by treatment, although there was a small, non-significant increase in all cells treated with carnosine (Fig. 1C, D). Carnosine increased total O2 consumption area-under-the-curve relative to control in T2D cells ($$p \leq 0.035$$, T2D-GLT cells; $$p \leq 0.073$$, Fig. 1E).Fig. 1Skeletal muscle O2 consumption measured in human skeletal myotubes using the Seahorse XFe24 Analyser. A, B ATP-linked and maximal respiration for all cell conditions. C Seahorse Mito Stress Test trace for LHC cells. D Maximal respiration for T2D and T2D-GLT cells. E OCR tAUC for all cell conditions, showing the fold-change relative to each control condition ($$n = 3$$ independent experiments per condition, $$n = 6$$–7 replicates per experiment). * $p \leq 0.05$, **$p \leq 0.01.$ LHC lean healthy control, OCR O2 consumption rate, T2D-GLT type-2 diabetic glucolipotoxic conditions, tAUC total area-under-the-curve ## Glucose uptake Absolute basal and insulin-stimulated glucose uptakes were significantly lower in T2D cells compared with LHC cells (basal: 17.66 ± 1.56 vs. 51.36 ± 2.57 nmol.min−1; $p \leq 0.0001$, insulin-stimulated: 22.11 ± 2.62 vs. 66.89 ± 5.31 nmol.min−1; $$p \leq 0.0003$$). Treatment with carnosine significantly increased the ratio of basal to insulin-stimulated glucose uptake in T2D cells ($$p \leq 0.047$$, Fig. 2A); treatment did not affect LHC cells. Fig. 2Insulin-stimulated glucose uptake and protein adducts in human skeletal myotubes. A Glucose uptake in LHC and T2D; B GLO1 expression in T2D myotubes; C MGO-modified proteins in T2D cells under basal and insulin-stimulated (INS) conditions; D 4-HNE-modified proteins in T2D cells under basal and INS conditions ($$n = 3$$ independent experiments per condition, $$n = 3$$–4 replicates per experiment). * $p \leq 0.05.$ β-al β-alanine, Car carnosine, LHC lean healthy control, OCR O2 consumption rate, T2D type-2 diabetic ## Protein expression and modification Insulin-stimulation increased MGO-modified proteins in T2D cells by $47\%$; treatment with carnosine attenuated this increase to $9.7\%$ ($$p \leq 0.011$$; Fig. 2C). There was no effect of insulin-stimulation or treatment on 4-HNE modified proteins in T2D cells (Fig. 2D); and no effect of treatment on expression of GLO1 (Fig. 2B) or other metabolic proteins (LDHA, PCG-1α, Sirt1, or Sirt3; Supplementary Fig. 2). ## Discussion Our main finding is that treatment of primary T2D skeletal muscle cells with carnosine increases insulin-stimulated glucose uptake. This occurred in parallel with a reduction in insulin-stimulated MGO-modified proteins, which is a possible mechanism of action. To our knowledge, this is the first study to show these effects in T2D skeletal muscle cells. Our data also show that T2D skeletal muscle cells have significantly lower ATP-linked respiration, maximal respiration, and basal and insulin-stimulated glucose uptake when compared with LHC cells. This preservation of the T2D phenotype validates our cell model, which is consistent with previous work in human myotubes (Gaster 2019). The increase in insulin-stimulated glucose uptake occurred independently of changes in cell viability and this finding supports our previous work in mouse C2C12 skeletal muscle cells under metabolic stress (Cripps et al. 2017). Herein, we showed a $47\%$ increase in MGO-modified proteins following insulin-stimulation, which is comparable to the increase seen in T2D muscle following a hyperinsulinaemic–euglycaemic clamp (Mey et al. 2018). Treatment with carnosine attenuated the increase in MGO-modified proteins to $9.7\%$, which could explain the beneficial effects on glucose uptake. In T2D skeletal muscle, insulin-stimulated glycolytic flux leads to MGO formation via the spontaneous oxidation of glycolytic intermediates glyceraldehyde-3-phosphate (G3P) and dihydroxyacetone phosphate (DHAP) (Mey et al. 2018; Phillips and Thornalley 1993). Excessive MGO production affects glucose transport and insulin signalling in a dose-dependent manner by binding directly to insulin receptor substrate (IRS) proteins and altering their structure and function (Riboulet-Chavey et al. 2006). Under normal conditions, MGO and downstream AGE formation is balanced by detoxification through the GLO1 enzyme system (Rabbani and Thornalley 2012). As GLO1 expression is reduced in T2D skeletal muscle (Mey et al. 2018), we explored whether the reduction in MGO-modified proteins might have occurred as a result of a change in GLO1 expression. GLO1 was unaffected by carnosine or β-alanine, which points towards a direct role of carnosine binding to MGO, thereby reducing MGO-mediated protein modification, as shown previously (Hipkiss and Chana 1998). Some studies, however, have suggested that carnosine has only low scavenging activity towards MGO (Colzani et al. 2016; Vistoli et al. 2017). This suggests that carnosine could potentially have reduced MGO-protein modification indirectly by activating endogenous antioxidant and anti-carbonylation defence systems via the nuclear factor erythroid 2-related factor 2 (Nrf2) signalling cascade (for a detailed review, see Aldini et al. 2021) and these mechanisms should be explored in future studies. Carnosine is an efficient scavenger of the reactive aldehyde, 4-HNE (Colzani et al. 2016); and our previous work showed carnosine prevents 4-HNE protein adduction events in human plasma and human skeletal muscle cells (Lavilla et al. 2021). Despite this, we did not see a reduction in basal or insulin-stimulated 4-HNE-modified proteins in the present study. There are several possible explanations for this. Firstly, while plasma 4-HNE increases ~ $33\%$ in the postprandial state in people with T2D (Neri et al. 2010), this might not translate into an increase in intramuscular 4-HNE within the 1-h insulin-stimulation period used in our study. Further, our previous work involved cells being cultured in GLT conditions for 5-days, which suggests that a chronic hyperglycaemic environment is important for generating intracellular 4-HNE. In further support of this assertion, Mey et al. [ 2018] showed that global carbonyl stress was only ~ $11\%$ higher in T2D skeletal muscle under basal conditions and was not affected by a hyperinsulinaemic–euglycaemic clamp. Our data show an increase in total-area-under-the-curve O2 consumption in T2D cells treated with carnosine; T2D-GLT cells followed a similar pattern but did not reach statistical significance. The increase in O2 consumption was from mitochondrial and non-mitochondrial respiration and reflects a general increase in cell respiratory capacity (Mitov et al. 2017). These results should be considered alongside the absence of significant changes in ATP-linked and maximal respiration, which are the markers that best-reflect changes in mitochondrial content and oxidative capacity (Divakaruni et al. 2014). This is supported by the lack of change in protein expression linked to mitochondrial biogenesis. These findings are in contrast to previous work in C2C12 myotubes, where treatment with 800 µM β-alanine induced several markers of mitochondrial biogenesis and increased O2 consumption (Schnuck et al. 2016); and work in rat cardiomyocytes, where treatment with 5 mM β-alanine caused mitochondrial fragmentation and oxidative stress, leading to a reduction in cellular O2 consumption (Shetewy et al. 2016). Our results suggest that treatment with 10 mM β-alanine does not cause these beneficial or deleterious changes in human skeletal muscle. The addition of GLT media to T2D cells had only a negligible effect on cellular respiration. This is in contrast to our previous work that showed a substantial reduction in ATP-linked and maximal respiration in LHC cells under GLT conditions (Lavilla et al. 2021). We take this to show that the extracellular environment has less influence on T2D cellular function due to the inherent defects that persist in vitro, whereas defects can be induced in the LHC cells, and a recovery of function made, owing to their original healthy state. Our study raises several questions that warrant further investigation. We did not see an effect for β-alanine alone on any outcome, which could be due to insufficient time or substrate for carnosine synthesis (DMEM:F-12 culture media contains ~ 203 µM histidine). Because of this, our results suggest that the beneficial effects shown are due to the intact dipeptide carnosine, and not its rate-limiting precursor, although a combined β-alanine and histidine treatment group would be needed to confirm this. We did not directly measure carnosine adducts, or carnosine-binding to MGO, meaning our results are associative, and several other pro-inflammatory reactive species (e.g., malondialdehyde, 3-nitrotyrosine, acrolein, and 4-hydroxy-2-hexanal) could be implicated. Future research should build on this and characterise the effect of carnosine on the insulin signalling cascade by measuring the phosphorylation status of relevant proteins (e.g., Akt and TBC1D4), stress-activated kinase signalling (e.g., AMPK, p38, and CAMKII), and GLUT4 translocation. Data presented herein indicate that carnosine can increase insulin-stimulated glucose uptake and reduce insulin-stimulated MGO-modified proteins in T2D skeletal muscle cells. This expands upon our previous work showing beneficial effects on protein adduction in human cells and glucose uptake in C2C12 cells. These results have implications for the role of carnosine in T2D and prediabetes and suggest that the beneficial effects on glycaemic control in human clinical trials may be explained by its RCS-scavenging actions in human skeletal muscle. ## Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 231 KB) ## References 1. Albrecht T, Schilperoort M, Zhang S, Braun JD, Qiu J, Rodriguez A, Pastene DO, Krämer BK, Köppel H, Baelde H, de Heer E. **Carnosine attenuates the development of both type 2 diabetes and diabetic nephropathy in BTBR ob/ob mice**. *Sci Rep* (2017) **7** 1-6. DOI: 10.1038/srep44492 2. Aldini G, Orioli M, Rossoni G, Savi F, Braidotti P, Vistoli G, Yeum KJ, Negrisoli G, Carini M. **The carbonyl scavenger carnosine ameliorates dyslipidaemia and renal function in Zucker obese rats**. *J Cell Mol Med* (2011) **15** 1339-1354. DOI: 10.1111/j.1582-4934.2010.01101.x 3. Aldini G, de Courten B, Regazzoni L, Gilardoni E, Ferrario G, Baron G, Altomare A, D’Amato A, Vistoli G, Carini M. **Understanding the antioxidant and carbonyl sequestering activity of carnosine: Direct and indirect mechanisms**. *Free Radic Res* (2021) **55** 321-330. DOI: 10.1080/10715762.2020.1856830 4. Artioli GG, Sale C, Jones RL. **Carnosine in health and disease**. *Eur J Sport Sci* (2019) **19** 30-39. DOI: 10.1080/17461391.2018.1444096 5. Baba SP, Hoetker JD, Merchant M, Klein JB, Cai J, Barski OA, Conklin DJ, Bhatnagar A. **Role of aldose reductase in the metabolism and detoxification of carnosine-acrolein conjugates**. *J Biol Chem* (2013) **288** 28163-28179. DOI: 10.1074/jbc.M113.504753 6. Baguet A, Everaert I, De Naeyer H, Reyngoudt H, Stegen S, Beeckman S, Achten E, Vanhee L, Volkaert A, Petrovic M, Taes Y. **Effects of sprint training combined with vegetarian or mixed diet on muscle carnosine content and buffering capacity**. *Eur J Appl Physiol* (2011) **111** 2571-2580. DOI: 10.1007/s00421-011-1877-4 7. Bass JJ, Wilkinson DJ, Rankin D, Phillips BE, Szewczyk NJ, Smith K, Atherton PJ. **An overview of technical considerations for Western blotting applications to physiological research**. *Scand J Med Sci Sports* (2017) **27** 4-25. DOI: 10.1111/sms.12702 8. Brings S, Fleming T, Freichel M, Muckenthaler MU, Herzig S, Nawroth PP. **Dicarbonyls and advanced glycation end-products in the development of diabetic complications and targets for intervention**. *Intl J Mol Sci* (2017) **18** 984. DOI: 10.3390/ijms18050984 9. Colzani M, De Maddis D, Casali G, Carini M, Vistoli G, Aldini G. **Reactivity, selectivity, and reaction mechanisms of aminoguanidine, hydralazine, pyridoxamine, and carnosine as sequestering agents of reactive carbonyl species: a comparative study**. *ChemMedChem* (2016) **11** 1778-1789. DOI: 10.1002/cmdc.201500552 10. Cripps MJ, Hanna K, Lavilla C, Sayers SR, Caton PW, Sims C, De Girolamo L, Sale C, Turner MD. **Carnosine scavenging of glucolipotoxic free radicals enhances insulin secretion and glucose uptake**. *Sci Rep* (2017) **7** 1-7. DOI: 10.1038/s41598-017-13649-w 11. DeFronzo RA, Ferrannini E, Groop L, Henry RR, Herman WH, Holst JJ, Hu FB, Kahn CR, Raz I, Shulman GI, Simonson DC. **Type 2 diabetes mellitus**. *Nat Rev Dis Primers* (2015) **1** 1-22. DOI: 10.1038/nrdp.2015.19 12. Divakaruni AS, Rogers GW, Murphy AN. **Measuring mitochondrial function in permeabilized cells using the seahorse XF analyzer or a Clark-type oxygen electrode**. *Curr Protoc Toxicol* (2014) **60** 25. DOI: 10.1002/0471140856.tx2502s60 13. Elbarbary NS, Ismail EA, El-Naggar AR, Hamouda MH, El-Hamamsy M. **The effect of 12 weeks carnosine supplementation on renal functional integrity and oxidative stress in pediatric patients with diabetic nephropathy: a randomized placebo-controlled trial**. *Pediatr Diabetes* (2018) **19** 470-477. DOI: 10.1111/pedi.12564 14. Gaster M. **The diabetic phenotype is preserved in myotubes established from type 2 diabetic subjects: a critical appraisal**. *APMIS* (2019) **127** 3-26. DOI: 10.1111/apm.12908 15. Hipkiss AR, Chana H. **Carnosine protects proteins against methylglyoxal-mediated modifications**. *Biochem Biophys Res Commun* (1998) **248** 28-32. DOI: 10.1006/bbrc.1998.8806 16. Houjeghani S, Kheirouri S, Faraji E, Jafarabadi MA. **L-Carnosine supplementation attenuated fasting glucose, triglycerides, advanced glycation end products, and tumor necrosis factor–α levels in patients with type 2 diabetes: a double-blind placebo-controlled randomized clinical trial**. *Nutr Res* (2018) **1** 96-106. DOI: 10.1016/j.nutres.2017.11.003 17. Jing E, Emanuelli B, Hirschey MD, Boucher J, Lee KY, Lombard D, Verdin EM, Kahn CR. **Sirtuin-3 (Sirt3) regulates skeletal muscle metabolism and insulin signaling via altered mitochondrial oxidation and reactive oxygen species production**. *Proc Natl Acad Sci* (2011) **108** 14608-14613. DOI: 10.1073/pnas.1111308108 18. Kratz C, de Salles PV, de Andrade Nemezio KM, da Silva RP, Franchini E, Zagatto AM, Gualano B, Artioli GG. **Beta-alanine supplementation enhances judo-related performance in highly-trained athletes**. *J Sci Med Sport* (2017) **20** 403-408. DOI: 10.1016/j.jsams.2016.08.014 19. Lavilla CJ, Billacura MP, Hanna K, Boocock DJ, Coveney C, Miles AK, Foulds GA, Murphy A, Tan A, Jackisch L, Sayers SR, Caton PW, Doig CL, McTernan PG, Colombo SL, Sale C, Turner MD. **Carnosine protects stimulus-secretion coupling through prevention of protein carbonyl adduction events in cells under metabolic stress**. *Free Radical Biol Med* (2021) **1** 65-79. DOI: 10.1016/j.freeradbiomed.2021.08.233 20. Mahmood T, Yang PC. **Western blot: technique, theory, and trouble shooting**. *N Am J Med Sci* (2012) **4** 429. DOI: 10.4103/1947-2714.100998 21. Marshall C, Hitman GA, Cassell PG, Turner MD. **Effect of glucolipotoxicity and rosiglitazone upon insulin secretion**. *Biochem Biophys Res Commun* (2007) **356** 756-762. DOI: 10.1016/j.bbrc.2007.03.047 22. Matthews JJ, Artioli GG, Turner MD, Sale C. **The physiological roles of carnosine and β-alanine in exercising human skeletal muscle**. *Med Sci Sports Exerc* (2019) **51** 2098-2108. DOI: 10.1249/MSS.0000000000002033 23. Matthews JJ, Dolan E, Swinton PA, Santos L, Artioli GG, Turner MD, Elliott-Sale KJ, Sale C. **Effect of carnosine or β-alanine supplementation on markers of glycemic control and insulin resistance in humans and animals: a systematic review and meta-analysis**. *Adv Nutr* (2021) **12** 2216-2231. DOI: 10.1093/advances/nmab087 24. Mey JT, Haus JM. **Dicarbonyl stress and glyoxalase-1 in skeletal muscle: implications for insulin resistance and type 2 diabetes**. *Front Cardiovasc Med* (2018) **10** 117. DOI: 10.3389/fcvm.2018.00117 25. Mey JT, Blackburn BK, Miranda ER, Chaves AB, Briller J, Bonini MG, Haus JM. **Dicarbonyl stress and glyoxalase enzyme system regulation in human skeletal muscle**. *Am J Physiol-Regul Integr Comp Physiol* (2018) **314** R181-R190. DOI: 10.1152/ajpregu.00159.2017 26. Mitov MI, Harris JW, Alstott MC, Zaytseva YY, Evers BM, Butterfield DA. **Temperature induces significant changes in both glycolytic reserve and mitochondrial spare respiratory capacity in colorectal cancer cell lines**. *Exp Cell Res* (2017) **354** 112-121. DOI: 10.1016/j.yexcr.2017.03.046 27. Narang BJ, Atkinson G, Gonzalez JT, Betts JA. **A tool to explore discrete-time data: the time series response analyser**. *Int J Sport Nutr Exerc Metab* (2020) **30** 374-381. DOI: 10.1123/ijsnem.2020-0150 28. Neri S, Calvagno S, Mauceri B, Misseri M, Tsami A, Vecchio C, Mastrosimone G, Di Pino A, Maiorca D, Judica A, Romano G. **Effects of antioxidants on postprandial oxidative stress and endothelial dysfunction in subjects with impaired glucose tolerance and type 2 diabetes**. *Eur J Nutr* (2010) **49** 409-416. DOI: 10.1007/s00394-010-0099-6 29. Pagel-Langenickel I, Bao J, Joseph JJ, Schwartz DR, Mantell BS, Xu X, Raghavachari N, Sack MN. **PGC-1α integrates insulin signaling, mitochondrial regulation, and bioenergetic function in skeletal muscle**. *J Biol Chem* (2008) **283** 22464-22472. DOI: 10.1074/jbc.M800842200 30. Phillips SA, Thornalley PJ. **The formation of methylglyoxal from triose phosphates: investigation using a specific assay for methylglyoxal**. *Eur J Biochem* (1993) **212** 101-105. DOI: 10.1111/j.1432-1033.1993.tb17638.x 31. Pillon NJ, Croze ML, Vella RE, Soulère L, Lagarde M, Soulage CO. **The lipid peroxidation by-product 4-hydroxy-2-nonenal (4-HNE) induces insulin resistance in skeletal muscle through both carbonyl and oxidative stress**. *Endocrinology* (2012) **153** 2099-2111. DOI: 10.1210/en.2011-1957 32. Rabbani N, Thornalley PJ. **Methylglyoxal, glyoxalase 1 and the dicarbonyl proteome**. *Amino Acids* (2012) **42** 1133-1142. DOI: 10.1007/s00726-010-0783-0 33. Regazzoni L, de Courten B, Garzon D, Altomare A, Marinello C, Jakubova M, Vallova S, Krumpolec P, Carini M, Ukropec J, Ukropcova B. **A carnosine intervention study in overweight human volunteers: bioavailability and reactive carbonyl species sequestering effect**. *Sci Rep* (2016) **6** 1-3. DOI: 10.1038/srep27224 34. Riboulet-Chavey A, Pierron A, Durand I, Murdaca J, Giudicelli J, Van Obberghen E. **Methylglyoxal impairs the insulin signaling pathways independently of the formation of intracellular reactive oxygen species**. *Diabetes* (2006) **55** 1289-1299. DOI: 10.2337/db05-0857 35. Saunders B, de Salles PV, De Oliveira LF, da Eira SV, Da Silva RP, Riani L, Franchi M, de Souza GL, Harris RC, Roschel H, Artioli GG. **Twenty-four weeks of β-alanine supplementation on carnosine content, related genes, and exercise**. *Med Sci Sports Exerc* (2017) **49** 896-906. DOI: 10.1249/MSS.0000000000001173 36. Schnuck JK, Sunderland KL, Kuennen MR, Vaughan RA. **Characterization of the metabolic effect of β-alanine on markers of oxidative metabolism and mitochondrial biogenesis in skeletal muscle**. *J Exerc Nutr Biochem* (2016) **20** 34. DOI: 10.20463/jenb.2016.06.20.2.5 37. Shetewy A, Shimada-Takaura K, Warner D, Jong CJ, Mehdi AB, Alexeyev M, Takahashi K, Schaffer SW. **Mitochondrial defects associated with β-alanine toxicity: relevance to hyper-beta-alaninemia**. *Mol Cell Biochem* (2016) **416** 11-22. DOI: 10.1007/s11010-016-2688-z 38. Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB, Stein C, Basit A, Chan JC, Mbanya JC, Pavkov ME. **IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045**. *Diabetes Res Clin Pract* (2022) **1** 109119. DOI: 10.1016/j.diabres.2021.109119 39. Szwergold BS. **Carnosine and anserine act as effective transglycating agents in decomposition of aldose-derived Schiff bases**. *Biochem Biophys Res Commun* (2005) **336** 36-41. DOI: 10.1016/j.bbrc.2005.08.033 40. Vistoli G, Colzani M, Mazzolari A, Gilardoni E, Rivaletto C, Carini M, Aldini G. **Quenching activity of carnosine derivatives towards reactive carbonyl species: Focus on α−(methylglyoxal) and β−(malondialdehyde) dicarbonyls**. *Biochem Biophys Res Commun* (2017) **492** 487-492. DOI: 10.1016/j.bbrc.2017.08.069
--- title: The association of fasting plasma thiol fractions with body fat compartments, biomarker profile, and adipose tissue gene expression authors: - Amany Elshorbagy - Nasser E. Bastani - Sindre Lee-Ødegård - Bente Øvrebø - Nadia Haj-Yasein - Karianne Svendsen - Cheryl Turner - Helga Refsum - Kathrine J. Vinknes - Thomas Olsen journal: Amino Acids year: 2022 pmcid: PMC10038976 doi: 10.1007/s00726-022-03229-2 license: CC BY 4.0 --- # The association of fasting plasma thiol fractions with body fat compartments, biomarker profile, and adipose tissue gene expression ## Abstract People with high plasma total cysteine (tCys) have higher fat mass and higher concentrations of the atherogenic apolipoprotein B (apoB). The disulfide form, cystine, enhanced human adipogenesis and correlated with total fat mass in a Middle-Eastern cohort. In 35 European adults with overweight ($88.6\%$ women) and with dual-X-ray absorptiometry measurements of regional fat, we investigated how cystine compared to other free disulfides in their association with total regional adiposity, plasma lipid and glucose biomarkers, and adipose tissue lipid enzyme mRNA ($$n = 19$$). Most total plasma homocysteine (tHcy) ($78\%$) was protein-bound; $63\%$ of total glutathione (tGSH) was reduced. tCys was $49\%$ protein-bound, $30\%$ mixed-disulfide, $15\%$ cystine, and $6\%$ reduced. Controlling for age and lean mass, cystine and total free cysteine were the fractions most strongly associated with android and total fat: $1\%$ higher cystine predicted $1.97\%$ higher android fat mass ($95\%$ CI 0.64, 3.31) and $1.25\%$ (0.65, 2.98) higher total fat mass (both $$p \leq 0.005$$). A positive association between tCys and apoB (β: $0.64\%$; $95\%$ CI 0.17, $1.12\%$, $$p \leq 0.009$$) was apparently driven by free cysteine and cystine; cystine was also inversely associated with the HDL-associated apolipoprotein A1 (β: −$0.57\%$; $95\%$ CI −0.96, −$0.17\%$, $$p \leq 0.007$$). No independent positive associations with adiposity were noted for tGSH or tHcy fractions. Plasma cystine correlated with CPT1a mRNA (Spearman’s $r = 0.68$, $$p \leq 0.001$$). In conclusion, plasma cystine—but not homocysteine or glutathione disulfides—is associated with android adiposity and an atherogenic plasma apolipoprotein profile. The role of cystine in human adiposity and cardiometabolic risk deserves investigation. ClinicalTrials.gov identifiers: NCT02647970 and NCT03629392. ### Supplementary Information The online version contains supplementary material available at 10.1007/s00726-022-03229-2. ## Introduction In the general population, the highest versus lowest plasma total concentrations of the sulfur amino acid cysteine (tCys) are associated with up to 10 kg higher body total fat mass (Elshorbagy et al. 2008), an association that is independent of various confounders and was replicated in multiple populations of varying age and ethnicity (Elshorbagy et al. 2012b, d, 2013a). Evidence from transgenic and dietary animal models suggests that increased cysteine availability is potentially obesogenic (Hasek et al. 2013; Elshorbagy 2014; Niewiadomski et al. 2016; Wanders et al. 2018). Collectively, these findings have lately led to human dietary interventions designed to reduce dietary sulfur amino acid intake with the aim of improving body adiposity and metabolic health (Olsen et al. 2018, 2020a; Stolt et al. 2021). Plasma tCys includes the sum of free oxidized, free reduced, and protein-bound forms. An oxidized cysteine redox state is associated with ageing, BMI, and cardiovascular disease (CVD) (Oliveira and Laurindo 2018). The epidemiologic association of tCys with adiposity was recently followed up by dissection of which component(s) of the plasma cysteine pool correlates with body fat mass. In 35 healthy adults, tCys in fasting acid-precipitated plasma was on average composed of $62\%$ protein-bound cysteine (bCys), $2\%$ free reduced cysteine, $16\%$ free homogeneous disulfide (cystine), and $20\%$ mixed cysteine disulfides (e.g., cysteine-homocysteine) (Elkafrawy et al. 2021). It was the latter two free oxidized forms that correlated most strongly with fat mass, whereas the major fraction, bCys, did not. Further, ascending physiologic concentrations of cystine enhanced adipogenic differentiation and lipid accumulation in primary culture of human adipose-derived mesenchymal stem cells (Elkafrawy et al. 2021), with implications for fat mass expansion in humans. These findings suggest that it is not the total body cysteine pool, but rather the free disulfide fraction that is associated with obesity, and with induction of PPARG and its target genes during adipogenesis (Elkafrawy et al. 2021). These data warrant replication, and raise several questions. First, it is not known whether other free disulfides in plasma [e.g., homocystine and glutathione (GSH) disulfide (GSSG)] are similarly associated with obesity. Total homocysteine (tHcy) has been thoroughly investigated in relation to BMI, and was found in a systematic review to be modestly higher in individuals with obesity versus normal weight subjects in only 3 out of 8 studies (Wiebe et al. 2018). However, to our knowledge, the association of the different homocysteine and GSH species in plasma has not been investigated. Second, the role of sulfur in adipocytes in rodent models warrants the investigation of circulating tCys, its fractions, and related thiols with human adipose tissue gene expression. Importantly, restricting dietary cysteine (and methionine) availability in rodents alters lipogenic and lipolytic gene expression in adipose tissues, including increased expression of Acaca, Fasn, Scd1, and Cpt1a (Perrone et al. 2010; Hasek et al. 2013) with simultaneous downregulation in liver (Hasek et al. 2013). A more oxidizing environment is present in both visceral and subcutaneous fat from obese compared to lean individuals (Akl et al. 2017), but body fat depots vary in their association with cardiometabolic risk. Whereas upper body fat/android fat is associated with insulin resistance and CVD risk markers in both sexes (Okosun et al. 2015; Vasan et al. 2018), gynoid fat, and in particular leg fat have been found to be either protective (Vasan et al. 2018), or less strongly associated with metabolic risk as android fat (Okosun et al. 2015). Plasma tCys was associated with insulin resistance in children and adults (Elshorbagy et al. 2012d, 2018) and an adverse lipoprotein profile in adults (Elshorbagy et al. 2012b). In line with previously reported associations with overall obesity, we hypothesize that the cystine fraction of tCys is more strongly associated with android than with gynoid fat mass, markers of impaired glucose metabolism, and an adverse lipid profile. The present study aims to replicate and extend our recent findings of the association between cystine and total fat mass (Elkafrawy et al. 2021) by exploring, a—the association of plasma tCys fractions with distinct body fat depots that vary in their association with cardiometabolic risk; and b—if other plasma-free disulfide fractions (of homocysteine and/or GSH) are associated with fat mass in healthy adults. For the thiol fractions consistently associated with body fat depots, we explored associations with plasma markers of lipid and glucose metabolism and adipose tissue mRNA expression of genes involved in lipid metabolism in vivo. ## Participants This cross-sectional study utilizes pooled baseline data from healthy free-living individuals with normal weight ($$n = 15$$) overweight or obesity ($$n = 20$$) (31 women and 4 men) who were recruited between 2016 and 2018 for participation in two dietary interventions trials with sulfur amino acid restriction (Olsen et al. 2018, 2020a) (Study 1 and Study 2, respectively). All study procedures were similar for both studies and details can be found in the referenced publications (Olsen et al. 2018, 2020a). A brief overview follows. Inclusion and exclusion criteria were similar for both studies with some exceptions. In Study 1, 15 normal weight men and women aged 20–40 years were recruited, whereas in Study 2, 20 women with overweight and obesity aged 20–40 years were recruited. Exclusion criteria were nearly identical in both studies and included presence of chronic disease or drug use, smoking, veganism, pregnancy or lactation, and intensive physical activity > 3 times per week. An additional exclusion criterion in (Olsen et al. 2018) was high intake of fatty fish or cod liver oil. Because of the similarities in inclusion/exclusion criteria and to maximize variation in body fat, these data were pooled for the current analysis. All participants gave written informed consent. The study was conducted according to the guidelines in the Declaration of Helsinki, and the Regional Ethics Committee for Medical Research in South East Norway both studies. The original studies were registered with ClinicalTrials.gov, with identifiers NCT02647970 and NCT03629392. ## Assessment of body composition Fasting body composition (lean mass, total fat mass, android fat mass, and gynoid fat mass) measurements were performed with dual-energy X-ray absorptiometry (DXA) using Lunar iDXA (GE Healthcare Lunar, Buckinghamshire, UK) and the software enCORE version 16 including the application CoreScan. System calibration was performed daily and subjects were measured wearing light clothing. ## Anthropometric measurements Weight and height were measured using a calibrated scale with a built-in stadiometer (Seca 285, Birmingham, UK). Waist circumference was measured at the midpoint between the lower margin of the last palpable rib and the top of the iliac crest. Hip circumference was measured with light clothing and at the widest portion of the buttocks. Each measure was repeated three times and the average used for analysis. ## Blood sampling Fasting venous blood samples were collected into 2 EDTA-lined vacuum tubes. To trap the thiols, one tube was pre-treated with 150 mmol/L N-ethylmaleimide at $10\%$ of the tube bringing final concentration of N-ethylmaleimide to 15 mmol/L. Blood was centrifuged immediately for 5 min at 4 °C, followed by precipitation with 5-sulfosalicyclic acid $10\%$, with the resultant supernatant aliquoted and stored at −80 °C until analysis. ## Amino acid assays Determination of plasma concentrations of tCys, tHcy, tGSH and their respective fractions have been reported previously (Antoniades et al. 2006; Olsen et al. 2018, 2020a). Briefly, all analytes were measured using liquid chromatography–tandem mass spectrometry. Coefficients of variation for total aminothiols were 3.4–$6.7\%$. For the total unbound fraction representing the sum of free reduced and disulfide concentrations, the coefficient of variation ranged from 4 to $6\%$. The bound fraction was calculated by subtracting the unbound concentration from the total plasma concentration. ## Plasma clinical biochemistry Measurement of total cholesterol, apolipoprotein-A1 (apoA1), apolipoprotein B (apoB), glucose, and insulin were measured at Department of Medical Biochemistry (Oslo University Hospital Rikshospitalet, Oslo, Norway) by colorimetric and/or enzymatic methods (Cobas c702 analyzer, Roche Diagnostics International Ltd, Rotkreuz, Switzerland). HOMA-IR was calculated as fasting glucose (mmol/l) x fasting insulin (pmol/l)/135. ## Subcutaneous adipose tissue biopsies and quantitative real-time PCR We explored the associations of thiols with steady-state expression of genes involved in lipid metabolism in white adipose tissue. Information on gene expression was available from the 19 participants in Study 2 and included carnitine palmitoyl transferase 1 (CPT1A); sterol regulatory element binding protein 1 (SREBP); stearoyl CoA-desaturase 1 (SCD1); acetyl CoA-carboxylase (ACACA); diacylglycerol acyl transferase 1 (DGAT1); fatty acid synthase (FASN); peroxisome proliferator-activated receptor γ (PPARG) as well as leptin (LEP): *These* genes were selected because they or their circulating markers (i.e., leptin) have been shown to respond to dietary manipulation of cysteine in animals (Perrone et al. 2010; Hasek et al. 2013) and administration of cysteine in in vitro studies (Haj-Yasein et al. 2017; Elkafrawy et al. 2021). Methods for white adipose tissue biopsies and quantitative real-time PCR have been described previously (Olsen et al. 2020a). In brief, subcutaneous white adipose tissue was obtained from the periumbilical region after anaesthesia with a local anaesthetic (5 mL Xylocain 10 mg/mL AstraZeneca, Södertälje, Sweden). After the procedure, biopsies were dissected and snap frozen in liquid nitrogen. Samples were stored at –80 °C until analysis. Prior to analysis, white adipose tissue RNA was isolated using TRIzol (Thermo Fisher Scientific, Waltham, MA, US) and Rneasy lipid Tissue mini kit (Quiagen, Hilden, Germany) according to the manufactured protocol. RNA quantity and quality were checked using a Nanodrop ND-1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). After isolation, 250 ng RNA was reversely transcribed to cDNA using the cDNA Reverse Transcription kit (Applied Biosystems, Waltham, MA, USA). Quantitative real-time PCR was performed with either 2.5 μL diluted cDNA (12.5 ng RNA), a 10 μL Kapa SYBR FAST qPCR Master Mix Universal (KapaBiosystems, Wilmington, MA, US), or 9 μL diluted cDNA (25 ng RNA) and 1 μL predeveloped TaqMan Gene Expression Assays were mixed in 10 μL TaqMan Gene Expression Master Mix (Thermo Fisher Scientific, Waltham, MA, US), on a Bio-Rad CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA, US). Primer sequences can be found in Supplementary file 1. mRNA levels were normalized to TBP and quantified using the ΔΔCt method. ## Statistical analysis All continuous covariates were log-transformed before analysis. Descriptive statistics are presented as geometric means (gM) and the lower and upper limits of the geometric standard deviation (gSD) as determined by gM/gSD and gM × gSD, respectively. Regression models with body fat compartment as the outcome and total aminothiols and their fractions as the predictor were constructed with age and total lean mass as additional covariates as they are considered important confounders of the association between amino acids and body fat compartments. Regression models with metabolic biomarkers as the outcome were constructed with age as an additional covariate. Adding sex to the models did not alter the estimates and was omitted from the models to avoid overadjustment bias. A sensitivity analysis was conducted where we excluded males ($$n = 4$$) in the models with body fat compartments as outcome. Spearman’s rank correlation analyses were performed for aminothiols and their fractions with mRNA transcripts in white adipose tissue. We applied the Benjamini–Hochberg multiple tests under an FDR of 0.05 (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q$$\end{document}Q). The p values from 90 tests on thiols and body composition were ranked (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$i$$\end{document}i), and a q value was calculated for each of the 90 tests (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m$$\end{document}m) using the formula \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(i/m)Q$$\end{document}(i/m)Q. From these calculations, the highest q value under the FDR of 0.05 was $q = 0.048.$ The original p value corresponding to this particular q value was 0.011, and tests with $p \leq 0.011$ were therefore considered significant throughout. All statistical analyses were performed using R version 4.1.0 (R for Statistical Computing, Vienna, Austria). ## Baseline characteristics Baseline characteristics, select biomarkers, and plasma thiols with fractions are presented in Table 1. Descriptive characteristics by study are presented in the Supplementary file 2. Briefly, data were obtained from 35 healthy, free-living subjects (31 women and 4 men) with a gM (gSD limits) age of 29.1 (23.7, 35.8) years, BMI of 26.2 (22.4, 30.7) kg/m2, and total fat mass of 24.3 (16.0, 36.9) kg. Table 1Baseline characteristics ($$n = 35$$)aAge, years29.1 (23.7, 35.8)Women, n (%)31 (88.6)Body adiposity BMI, kg/m226.2 (22.4, 30.7) Waist circumference, cm82.8 (73.8, 92.9) Hip circumference, cm106 (98.5, 114) Total fat mass, kg24.3 (16, 36.9) Android fat mass, kg1.58 (0.8, 3.12) Gynoid fat mass, kg4.92 (3.23, 7.49) Android/total fat mass ratio0.07 (0.05, 0.09) Gynoid/total fat mass ratio0.20 (0.18, 0.23)Plasma biomarkers Glucose, mmol/L4.91 (4.37, 5.51) Insulin, pmol/L44.8 (28.6, 70.1) C-peptide, pmol/L589 [440, 788] HOMA-IR1.63 (1.02, 2.61) Total cholesterol, mmol/L4.29 (3.68, 5.01) Apolipoprotein B, g/L0.74 (0.60, 0.91) Apolipoprotein A1, g/L1.47 (1.21, 1.79) Triglycerides, mmol/L0.87 (0.58, 1.28)Plasma thiols Total cysteine, μmol/L271 [235, 312] Protein-bound cysteine, μmol/L131 [107, 160] Free cysteine, μmol/L138 [114, 166] Reduced cysteine, μmol/L14.8 (10.4, 21) Cystine, μmol/L40.2 (34.1, 47.4) Reduced cysteine/cystine0.37 (0.28, 0.48) Total homocysteine, μmol/L8.31 (6.33, 10.9) Protein-bound homocysteine, μmol/L6.47 (4.79, 8.74) Free homocysteine, μmol/L1.75 (1.17, 2.61) Reduced homocysteine, μmol/L0.19 (0.13, 0.29) Homocystine, μmol/L0.02 (0.02, 0.03) Reduced homocysteine/homocystine10.7 (6.91, 16.6) Total glutathione, μmol/L6.68 (5.07, 8.81) Protein-bound glutathione, μmol/L1.44 (0.51, 4.07) Free glutathione, μmol/L4.68 (3.38, 6.48) Reduced glutathione, μmol/L4.11 (2.82, 5.99) GSSG, μmol/L0.05 (0.03, 0.08) Reduced glutathione/GSSG85.0 (59.3, 122)BMI body mass index, GSSG oxidized glutathioneaContinuous data were log-transformed and presented as geometric mean (gSD limits) *Mean plasma* concentrations of tCys, tHcy, and tGSH were, respectively, 271, 8.31, and 6.68 μmol/L. The three metabolites varied widely in the relative distribution of the different plasma fractions (Fig. 1). For tCys, $49\%$ was protein-bound, $15\%$ was cystine, $6\%$ was free reduced cysteine, and the remainder was mixed cysteine disulfides. The majority of tHcy was protein-bound ($78\%$), and the majority of tGSH was reduced ($63\%$). For both tHcy and tGSH, homogeneous disulfides constituted less than $1\%$.Fig. 1Percentage distribution of plasma total cysteine, total homocysteine, and total glutathione fractions in the study population; $$n = 35$.$ Homogeneous disulfides are, respectively, cystine, homocystine, and glutathione disulfide ## Correlation among thiols A heatmap showing the Spearman correlation coefficients among the thiols is shown in Fig. 2. Total cysteine correlated with tGSH and the reduced and protein-bound GSH fractions ($r = 0.52$ to 0.61). Reduced cysteine correlated positively with the reduced fractions of homocysteine ($r = 0.84$) and GSH ($r = 0.77$). Reduced homocysteine did not correlate with reduced GSH.Fig. 2Illustration of correlation coefficients among thiol fractions. Blue tiles denote positive correlations; red tiles denote inverse correlations with a p value below 0.011. White (blank) tiles are non-significant. Cys cysteine, GSH glutathione, hcy homocysteine, Hcystine homocystine, Red reduced, tCys total cysteine, tGSH total glutathione, tHcy total homocysteine ## Thiol fractions and body fat depots Age- and lean mass-adjusted associations for plasma tCys, tHcy, tGSH and their fractions with total fat mass and different body fat compartments as measured by DXA are illustrated in Fig. 3a–c, and estimates and p values are available in Online Resource 3.Fig. 3Associations for a total cysteine and fractions, b total homocysteine and fractions, and c total glutathione and fractions with body fat compartments. Estimates were obtained from regression models where log-transformed body fat compartment was the dependent variable and log-transformed thiol was the main independent variable. The models were additionally adjusted for age and lean mass. Estimates indicate % change in body fat compartment per % change in the thiol. Note the differing scales on the X-axes. cys cysteine, GSH glutathione, GSSG oxidized glutathione, hcy homocysteine, Hcystine homocystine, tCys total cysteine, tHcy total homocysteine. * indicates $p \leq 0.011$ Plasma tCys concentrations were generally positively correlated with all body fat measures (Fig. 3a; Supplementary file 3). The strongest association was observed for tCys and android fat mass where $1\%$ increase in plasma tCys predicted a $2.62\%$ ($95\%$ CI 1.28, $3.95\%$) increase in android fat mass. With the exception of bCys, all other tCys fractions were consistently and positively associated with total, android, and gynoid fat mass. Of the tCys fractions, the strongest associations were observed for plasma free cysteine (β: $1.82\%$; $95\%$ CI 0.65, $2.98\%$) and cystine (β: $1.97\%$; $95\%$ CI 0.64, $3.31\%$) with android fat mass. People with higher plasma tCys had a significantly higher ratio of android fat to total fat mass (β: $0.96\%$; $95\%$ CI $0.33\%$, $1.59\%$; $p \leq 0.001$), but no difference in gynoid/total fat ratio ($$p \leq 0.78$$). A similar trend for an association with android/total fat ratio was seen for all cysteine fractions, with cystine having the largest estimate (β: $0.72\%$; $95\%$ CI $0.11\%$, $1.34\%$; $$p \leq 0.023$$; Fig. 3a; Supplementary file 3). None of the cysteine fractions were associated with gynoid/total fat ratio (p ≥ 0.55). No clear associations were observed for plasma tHcy with body fat compartments. For tGSH, weak positive associations were observed with total, android, and gynoid fat mass, and android/total fat ratio that were not significant under the adjusted p value. These associations appeared to be driven by protein-bound and reduced GSH (Supplementary file 3, Fig. 3c). When we additionally adjusted the models for tCys, all associations were attenuated and no longer significant (Supplementary file 4). In a sensitivity analysis that included only women ($$n = 31$$), the results were not appreciably different from the main analysis (data not shown). ## tCys, tGSH fractions, and plasma biomarkers Because tCys and its free fractions, and the reduced and protein-bound GSH fractions were consistently and positively associated with body fat compartments, we performed additional analyses investigating their associations with markers of an adverse plasma profile including lipoproteins and markers of glucose metabolism. All estimates were adjusted for age, and can be found in Supplementary file 5. For tCys, a positive association was observed with apoB (β: $0.64\%$; $95\%$ CI 0.17, $1.12\%$, $$p \leq 0.009$$). These observations appeared to be driven by free reduced cysteine (β: $0.26\%$; $95\%$ CI 0.07, $0.46\%$, $$p \leq 0.010$$) and cystine (β: $0.55\%$; $95\%$ CI 0.11, $0.99\%$, $$p \leq 0.016$$), although the observation for cystine with apoB was not significant under the adjusted p value. Cystine was also inversely associated with apoA1 concentrations (β: −$0.57\%$; $95\%$ CI −0.96, −$0.17\%$, $$p \leq 0.007$$). No associations were observed for tCys or its fractions with markers of glucose metabolism (glucose, insulin, HOMA-IR, and C-peptide). Reduced GSH was positively association with apolipoprotein B (β: $0.24\%$; $95\%$ CI 0.06, 0.43). No associations were observed for protein-bound cysteine, protein-bound GSH, or for the reduced cysteine/cystine ratio with markers of glucose or lipid metabolism. ## Correlations with adipose tissue gene expression Because tCys and its free fractions as well as protein-bound and reduced GSH were consistently associated with body fat compartments, we performed univariate correlational analyses between the tCys fractions with significant associations and total fat mass, and lipogenic and lipolytic mRNA transcripts in adipose tissue from 19 women with overweight and obesity. Correlation coefficients and p values are presented in Supplementary file 6. Plasma concentrations of cystine correlated strongly and positively with CPT1A expression (Spearman’s $r = 0.68$, $$p \leq 0.001$$; Fig. 4). Inverse correlations were observed for free reduced cysteine with PPARG expression (r = −0.53, $$p \leq 0.021$$), but this association was not significant under the adjusted p value. A similar, but non-significant association was observed for cystine and PPARG expression (r = −0.40, $$p \leq 0.086$$). No other associations were observed for tCys or other fractions with lipogenic and lipolytic mRNA transcripts. Because expressions of both CPT1A and PPARG are both associated with body fat mass, we adjusted the correlations for total fat mass. This adjustment did not attenuate the observed correlations (cystine and CPT1A, $r = 0.67$, $$p \leq 0.002$$; free reduced cysteine and PPARG, r = −0.56, $$p \leq 0.015$$ and cystine and PPARG, r = −0.46, $$p \leq 0.055$$).Fig. 4Scatter plot, regression line, and unadjusted Spearman correlation coefficient for the association of plasma cystine with CPT1A mRNA transcripts in adipose tissue biopsies. $$n = 19$$ women ## Discussion The plasma low-molecular-weight thiol pool consists of protein-bound, free oxidized disulfide, and free reduced sulfhydryl forms. Plasma tCys correlates positively with fat mass at the population level, an association that was recently shown to be mediated by its free disulfide forms (Elkafrawy et al. 2021). We sought to test the reproducibility of this finding in a cohort of different ethnicity, to investigate whether other free disulfides similarly correlate with fat mass, and to examine the association of the thiol fractions with regional fat distribution. In 35 adults, total unbound cysteine, cystine, and to a lesser extent, reduced cysteine, were the plasma tCys fractions that correlated positively with fat mass, and this association was not shared by Hcy or GSH disulfides. Of the body fat compartments, tCys and its free forms generally correlated more strongly with android than with gynoid fat, and people with higher tCys and cystine had a higher ratio of android fat to total fat mass. Taken together, these findings indicate an association of tCys with an unhealthy body fat distribution, and that the cystine association with adiposity does not reflect a universal association of circulating low-molecular-weight disulfides with obesity. In a relatively large study ($$n = 685$$), plasma cystine increased across BMI categories, but was unrelated to diet quality (Bettermann et al. 2018). Subsequently, we observed that cystine and mixed cysteine disulfides correlate with total fat mass measured by bioimpedance (Elkafrawy et al. 2021). Here, we confirm these findings using DXA, and show that tCys and its free fractions correlate most strongly with android body fat. This finding in a predominantly female young adult cohort is consistent with previous observations that plasma tCys was independently associated with trunk fat/total fat ratio both in 950 children with overweight (Elshorbagy et al. 2012d) and in 610 older adults (Elshorbagy et al. 2013a). Android obesity is a recognized correlate of cardiometabolic risk markers (Vasan et al. 2018) and an independent predictor of future diabetes (Brahimaj et al. 2019). Plasma tCys was also associated with visceral fat area measured by CT (Elshorbagy et al. 2018), and was found recently to predict incident diabetes in older adults (Elshorbagy et al. 2022). Overall, tCys and its free fractions were elevated in individuals with central adiposity in the present study. This finding should be viewed in the wider context of the existing premise that high plasma cystine is a marker of oxidative stress: studies have linked high cystine and low rGSH to multiple adverse outcomes, including CVD pathogenesis (Go and Jones 2011), renal disease progression (Rodrigues et al. 2012), and mortality (Patel et al. 2016). In line with its association with android adiposity, tCys correlated positively with the atherogenic apolipoprotein apoB, and inversely with apoA1, the main antiatherogenic protein component of HDL particles (German and Shapiro 2020). A higher ratio of apoB to apoA1 is a recognized risk factor for cardiovascular disease (Walldius and Jungner 2006). The positive and negative associations of tCys with apoB and apoA1, respectively, were previously observed in our study of 850 European adults (Elshorbagy et al. 2012b). The present study extends these findings by showing that these associations appear to be mediated by the free forms of tCys including free reduced cysteine and cystine. In rats, a high-cystine diet increased circulating concentrations of the atherogenic apolipoprotein B (apoB), with a trend for raising apoB protein and mRNA expression in liver (Sérougne et al. 1995). tCys and its fractions, however, were unrelated to glucose metabolism biomarkers. We have noted from different datasets that over the lower end of the population range of insulinemia (median ~ 45 pmol/L: (Elshorbagy et al. 2017) and the present study), tCys did not correlate with insulin resistance, but that it was associated with insulin resistance in populations with higher insulin (mean 69 and 98 pmol/L (Elshorbagy et al. 2012d, 2018). Overall, the association of tCys with glucose biomarkers appears less consistent than with adiposity, and is more evident when impaired glucose metabolism is present. Given previous findings that cystine influences adipogenic gene expression during adipogenesis in murine (Haj-Yasein et al. 2017) and human (Elkafrawy et al. 2021) cells, we investigated whether, in vivo, the cysteine forms that correlated with fat mass are associated with adipose tissue expression of lipid metabolism genes. The panel of genes in the present study overlapped with genes whose induction was influenced by cystine in vitro in two key genes, PPARG and SCD1. In rodents, changes in cysteine availability due to gene knockouts or dietary modifications have substantial and tissue-specific effects on both Pparg (Elshorbagy et al. 2012a) and Scd1 expression (Elshorbagy et al. 2011, 2013b; Gupta et al. 2013; Elshorbagy 2014), dependent on model and experimental setting. In both the mouse-derived 3T3-L1 cell line (Haj-Yasein et al. 2017), and in human primary preadipocytes, increasing extracellular cystine enhances PPARG and SCD1 expression during adipogenesis, with a six-to-sevenfold difference in their degree of induction in human preadipocytes across cystine concentrations (Elkafrawy et al. 2021). However, in the present study, there was no correlation between plasma cystine and adipose tissue PPARG or SCD1 expression. This suggests that while cystine might influence the degree of induction of PPARG and its target genes in response to adipogenic stimuli in differentiating preadipocytes (Elkafrawy et al. 2021), cystine may be unrelated to their steady-state expression in mature human adipocytes. Individuals with higher plasma cystine in the present cohort had a higher adipose tissue CPT1A mRNA level. CPT1a catalyzes the rate-limiting step of converting long-chain fatty acids into acyl-carnitines, which can then cross the mitochondrial membrane for oxidation in the mitochondria. In rodent studies, limiting dietary cysteine availability increased CPT1a mRNA and protein expression in inguinal adipose tissue (Perrone et al. 2010) (in line with the anti-obesity effects of the diet), but not in liver (Yang et al. 2019). In our short-term sulfur amino acid restriction trial in humans, however, no significant effect on adipose tissue CPT1A was noted despite a nearly fourfold difference in sulfur amino acid intake between the low- and high-intake groups (Olsen et al. 2020a). Although it is a key enzyme in fatty acid oxidation, CPT1A expression in human adipose tissue is positively associated with BMI in cross-sectional analysis (Warfel et al. 2017). We therefore hypothesized that the association of cystine with CPT1A expression in the present study may partly be explained by their respective links to adiposity, but adjusting for fat mass did not attenuate the association. It is therefore presently unclear whether the observed association between plasma cystine and CPT1A expression is physiologically relevant in humans, and this link requires replication and, if confirmed, investigation of possible mechanisms. In addition, CPT1A expression may not correlate with CPT1A protein abundance, although dietary cysteine restriction increased both expression and protein levels (Perrone et al. 2010), suggesting a positive correlation between mRNA and protein levels. No consistent associations were observed for plasma homocysteine or GSH species with fat mass, apart from a positive association of bGSH and rGSH with total fat and android fat. The explanation for this finding is unclear, and conflicts with previous observations that tGSH correlates inversely with percent body fat in adults (Elshorbagy et al. 2022) and children (Elshorbagy et al. 2012d). The finding could be partly related to the current assay method involving addition of N-ethylmalemide into the blood tube, which immediately traps all the thiol species. This not only prevents oxidation of plasma thiols but also reduces the rapid redox changes that occur after blood collection (Mansoor et al. 1992); hence, it is worth verifying this finding in other cohorts where the same method is used. We also postulate that this positive association of rGSH with adiposity may be partly driven by the uniquely strong correlation of GSH species with tCys in the current dataset, since adjusting for tCys weakened the association. In other larger cohorts, we have typically found no association between tCys and tGSH in either adults or children (Elshorbagy et al. 2012d, c). In these cohorts, tGSH showed no or negative associations with adiposity (Elshorbagy et al. 2012d, c), in line with findings from other groups (Bettermann et al. 2018). We note that the GSH/GSSG and reduced cysteine/cystine redox pairs have been reported to be in disequilibrium and differentially affected by various interventions, indicating that the observed attenuation may be brought about by an unknown and unmeasured confounding factor (Jones 2006) or other properties that are unique to this dataset such as the small sample size. Relative concentrations of plasma thiol fractions vary widely with food intake, in a time- and meal-type-dependent fashion (Park et al. 2010; Olsen et al. 2020b). The propensity for plasma reduced thiols to undergo rapid autooxidation also necessitates acid-precipitation following blood withdrawal to preserve the redox forms of the different fractions and enable accurate quantification. Partly owing to these considerations, plasma thiol fractions are generally absent from high-throughput plasma amino acid assays in large cohorts, many of which are based on stored and often non-fasting samples. The present study utilizes overnight-fasted samples that were immediately acid-precipitated and kept cold during centrifugation. This has ensured that the measured thiol fractions are as representative as possible of their steady-state circulating concentrations, and avoided spurious variability due to prandial state and sample treatment—but comes at the expense of the small cohort size. That regional fat estimates were from gold standard DXA measures further enhances reliability of the findings. Most participants were women, which minimizes residual confounding due to sex on the associations seen, but it should be noted that the tCys association with fat mass is marginally stronger in women than men (Elshorbagy et al. 2008). The relation of cysteine forms to adipose tissue gene expression was only tested in a small subset of 19 women, so the null associations for some genes may result from being underpowered for this exploratory analysis, and further studies are needed. In summary, the present study shows that the positive association of plasma cystine with total fat mass is linked to android adiposity and an atherogenic plasma lipoprotein profile; associations that are not shared by other free lower molecular weight disulfides in plasma. Studies are needed to investigate the determinants of plasma tCys and the balance between its free and protein-bound fractions in humans, and whether cystine, given its associations with an unhealthy body fat distribution and lipoprotein profile, predicts cardiometabolic outcomes. ## Supplementary Information Below is the link to the electronic supplementary material. Supplementary file 1 (DOCX 14 KB)Supplementary file 2 (DOCX 18 KB)Supplementary file 3 (DOCX 22 KB)Supplementary file 4 (DOCX 14 KB)Supplementary file 5 (DOCX 19 KB)Supplementary file 6 (DOCX 16 KB) ## References 1. Akl MG, Fawzy E, Deif M. **Perturbed adipose tissue hydrogen peroxide metabolism in centrally obese men: association with insulin resistance**. *PLoS ONE* (2017) **12** e0177268. DOI: 10.1371/journal.pone.0177268 2. Antoniades C, Shirodaria C, Warrick N. **5-Methyltetrahydrofolate rapidly improves endothelial function and decreases superoxide production in human vessels: effects on vascular tetrahydrobiopterin availability and endothelial nitric oxide synthase coupling**. *Circulation* (2006) **114** 1193-1201. DOI: 10.1161/CIRCULATIONAHA.106.612325 3. Bettermann EL, Hartman TJ, Easley KA. **Higher Mediterranean diet quality scores and lower body mass index are associated with a less-oxidized plasma glutathione and cysteine redox status in adults**. *J Nutr* (2018) **148** 245-253. DOI: 10.1093/jn/nxx045 4. Brahimaj A, Rivadeneira F, Muka T. **Novel metabolic indices and incident type 2 diabetes among women and men: the Rotterdam study**. *Diabetologia* (2019) **62** 1581-1590. DOI: 10.1007/s00125-019-4921-2 5. Elkafrawy H, Mehanna R, Ali F. **Extracellular cystine influences human preadipocyte differentiation and correlates with fat mass in healthy adults**. *Amino Acids* (2021) **53** 1623-1634. DOI: 10.1007/s00726-021-03071-y 6. Elshorbagy AK. **Body composition in gene knockouts of sulfur amino acid-metabolizing enzymes**. *Mamm Genome* (2014) **25** 455-463. DOI: 10.1007/s00335-014-9527-x 7. Elshorbagy AK, Nurk E, Gjesdal CG. **Homocysteine, cysteine, and body composition in the Hordaland homocysteine study: does cysteine link amino acid and lipid metabolism?**. *Am J Clin Nutr* (2008) **88** 738-746. DOI: 10.1093/ajcn/88.3.738 8. Elshorbagy AK, Valdivia-Garcia M, Mattocks DAL. **Cysteine supplementation reverses methionine restriction effects on rat adiposity: significance of stearoyl-coenzyme A desaturase**. *J Lipid Res* (2011) **52** 104-112. DOI: 10.1194/jlr.M010215 9. Elshorbagy AK, Church C, Valdivia-Garcia M. **Dietary cystine level affects metabolic rate and glycaemic control in adult mice**. *J Nutr Biochem* (2012) **23** 332-340. DOI: 10.1016/j.jnutbio.2010.12.009 10. Elshorbagy AK, Valdivia-Garcia M, Graham IM. **The association of fasting plasma sulfur-containing compounds with BMI, serum lipids and apolipoproteins**. *Nutr Metab Cardiovasc Dis* (2012) **22** 1031-1038. DOI: 10.1016/j.numecd.2011.01.008 11. Elshorbagy AK, Valdivia-Garcia M, Refsum H, Butte N. **The association of cysteine with obesity, inflammatory cytokines and insulin resistance in Hispanic children and adolescents**. *PLoS ONE* (2012) **7** e44166. DOI: 10.1371/journal.pone.0044166 12. Elshorbagy AK, Nijpels G, Valdivia-Garcia M. **S-adenosylmethionine is associated with fat mass and truncal adiposity in older adults**. *J Nutr* (2013) **143** 1982-1988. DOI: 10.3945/jn.113.179192 13. Elshorbagy AK, Valdivia-Garcia M, Mattocks DAL. **Effect of taurine and N-acetylcysteine on methionine restriction-mediated adiposity resistance**. *Metabolism* (2013) **62** 509-517. DOI: 10.1016/j.metabol.2012.10.005 14. Elshorbagy A, Jernerén F, Basta M. **Amino acid changes during transition to a vegan diet supplemented with fish in healthy humans**. *Eur J Nutr* (2017) **56** 1953-1962. DOI: 10.1007/s00394-016-1237-6 15. Elshorbagy AK, Samocha-Bonet D, Jernerén F. **Food overconsumption in healthy adults triggers early and sustained increases in serum branched-chain amino acids and changes in cysteine linked to fat gain**. *J Nutr* (2018) **148** 1073-1080. DOI: 10.1093/jn/nxy062 16. Elshorbagy AK, Turner C, Bastani N. **The association of serum sulfur amino acids and related metabolites with incident diabetes: a prospective cohort study**. *Eur J Nutr* (2022). DOI: 10.1007/s00394-022-02872-5 17. German CA, Shapiro MD. **Assessing atherosclerotic cardiovascular disease risk with advanced lipid testing: state of the science**. *Eur Cardiol* (2020) **15** e56. DOI: 10.15420/ecr.2019.18 18. Go Y-M, Jones DP. **Cysteine/cystine redox signaling in cardiovascular disease**. *Free Radical Biol Med* (2011) **50** 495-509. DOI: 10.1016/j.freeradbiomed.2010.11.029 19. Gupta S, Melnyk S, Kruger W. **Cystathionine β-synthase-deficient mice thrive on a low-methionine diet**. *FASEB J* (2013). DOI: 10.1096/fj.13-240770 20. Haj-Yasein NN, Berg O, Jernerén F. **Cysteine deprivation prevents induction of peroxisome proliferator-activated receptor gamma-2 and adipose differentiation of 3T3-L1 cells**. *Biochim Biophys Acta Mol Cell Biol Lipids* (2017) **1862** 623-635. DOI: 10.1016/j.bbalip.2017.02.009 21. Hasek BE, Boudreau A, Shin J. **Remodeling the integration of lipid metabolism between liver and adipose tissue by dietary methionine restriction in rats**. *Diabetes* (2013) **62** 3362-3372. DOI: 10.2337/db13-0501 22. Jones DP. **Redefining oxidative stress**. *Antioxid Redox Signal* (2006) **8** 1865-1879. DOI: 10.1089/ars.2006.8.1865 23. Mansoor MA, Svardal AM, Ueland PM. **Determination of the in vivo redox status of cysteine, cysteinylglycine, homocysteine, and glutathione in human plasma**. *Anal Biochem* (1992) **200** 218-229. DOI: 10.1016/0003-2697(92)90456-H 24. Niewiadomski J, Zhou JQ, Roman HB. **Effects of a block in cysteine catabolism on energy balance and fat metabolism in mice**. *Ann N Y Acad Sci* (2016) **1363** 99-115. DOI: 10.1111/nyas.13021 25. Okosun IS, Seale JP, Lyn R. **Commingling effect of gynoid and android fat patterns on cardiometabolic dysregulation in normal weight American adults**. *Nutr Diabetes* (2015) **5** e155. DOI: 10.1038/nutd.2015.5 26. Oliveira PVS, Laurindo FRM. **Implications of plasma thiol redox in disease**. *Clin Sci* (2018) **132** 1257-1280. DOI: 10.1042/CS20180157 27. Olsen T, Øvrebø B, Turner C. **Combining dietary sulfur amino acid restriction with polyunsaturated fatty acid intake in humans: a randomized controlled pilot trial**. *Nutrients* (2018) **10** 1822. DOI: 10.3390/nu10121822 28. Olsen T, Øvrebø B, Haj-Yasein N. **Effects of dietary methionine and cysteine restriction on plasma biomarkers, serum fibroblast growth factor 21, and adipose tissue gene expression in women with overweight or obesity: a double-blind randomized controlled pilot study**. *J Transl Med* (2020) **18** 122. DOI: 10.1186/s12967-020-02288-x 29. Olsen T, Turner C, Øvrebø B. **Postprandial effects of a meal low in sulfur amino acids and high in polyunsaturated fatty acids compared to a meal high in sulfur amino acids and saturated fatty acids on stearoyl CoA-desaturase indices and plasma sulfur amino acids: a pilot study**. *BMC Res Notes* (2020) **13** 1-7. DOI: 10.1186/s13104-020-05222-y 30. Park Y, Ziegler TR, Gletsu-Miller N. **Postprandial cysteine/cystine redox potential in human plasma varies with meal content of sulfur amino acids**. *J Nutr* (2010) **140** 760-765. DOI: 10.3945/jn.109.116764 31. Patel RS, Ghasemzadeh N, Eapen DJ. **Novel biomarker of oxidative stress is associated with risk of death in patients with coronary artery disease**. *Circulation* (2016) **133** 361-369. DOI: 10.1161/CIRCULATIONAHA.115.019790 32. Perrone CE, Mattocks DAL, Jarvis-Morar M. **Methionine restriction effects on mitochondrial biogenesis and aerobic capacity in white adipose tissue, liver, and skeletal muscle of F344 rats**. *Metabolism* (2010) **59** 1000-1011. DOI: 10.1016/j.metabol.2009.10.023 33. Rodrigues SD, Batista GB, Ingberman M. **Plasma cysteine/cystine reduction potential correlates with plasma creatinine levels in chronic kidney disease**. *BPU* (2012) **34** 231-237. DOI: 10.1159/000342627 34. Sérougne C, Felgines C, Férézou J. **Hypercholesterolemia induced by cholesterol- or cystine-enriched diets is characterized by different plasma lipoprotein and apolipoprotein concentrations in rats**. *J Nutr* (1995) **125** 35-41. DOI: 10.1093/jn/125.1.35 35. Stolt E, Olsen T, Elshorbagy A. **Sulfur amino acid restriction, energy metabolism and obesity: a study protocol of an 8-week randomized controlled dietary intervention with whole foods and amino acid supplements**. *J Transl Med* (2021) **19** 153. DOI: 10.1186/s12967-021-02824-3 36. Vasan SK, Osmond C, Canoy D. **Comparison of regional fat measurements by dual-energy X-ray absorptiometry and conventional anthropometry and their association with markers of diabetes and cardiovascular disease risk**. *Int J Obes* (2018) **42** 850-857. DOI: 10.1038/ijo.2017.289 37. Walldius G, Jungner I. **The apoB/apoA-I ratio: a strong, new risk factor for cardiovascular disease and a target for lipid-lowering therapy - a review of the evidence**. *J Intern Med* (2006) **259** 493-519. DOI: 10.1111/j.1365-2796.2006.01643.x 38. Wanders D, Forney LA, Stone KP. **The components of age-dependent effects of dietary methionine restriction on energy balance in rats**. *Obesity* (2018) **26** 740-746. DOI: 10.1002/oby.22146 39. Warfel JD, Vandanmagsar B, Dubuisson OS. **Examination of carnitine palmitoyltransferase 1 abundance in white adipose tissue: implications in obesity research**. *Am J Physiol Regul Integr Comp Physiol* (2017) **312** R816-R820. DOI: 10.1152/ajpregu.00520.2016 40. Wiebe N, Field CJ, Tonelli M. **A systematic review of the vitamin B12, folate and homocysteine triad across body mass index**. *Obes Rev* (2018) **19** 1608-1618. DOI: 10.1111/obr.12724 41. Yang Y, Wang Y, Sun J. **Dietary methionine restriction reduces hepatic steatosis and oxidative stress in high-fat-fed mice by promoting H2S production**. *Food Funct* (2019) **10** 61-77. DOI: 10.1039/c8fo01629a
--- title: Machine learning for post-acute pancreatitis diabetes mellitus prediction and personalized treatment recommendations authors: - Jun Zhang - Yingqi Lv - Jiaying Hou - Chi Zhang - Xuelu Yua - Yifan Wang - Ting Yang - Xianghui Su - Zheng Ye - Ling Li journal: Scientific Reports year: 2023 pmcid: PMC10038980 doi: 10.1038/s41598-023-31947-4 license: CC BY 4.0 --- # Machine learning for post-acute pancreatitis diabetes mellitus prediction and personalized treatment recommendations ## Abstract Post-acute pancreatitis diabetes mellitus (PPDM-A) is the main component of pancreatic exocrine diabetes mellitus. Timely diagnosis of PPDM-A improves patient outcomes and the mitigation of burdens and costs. We aimed to determine risk factors prospectively and predictors of PPDM-A in China, focusing on giving personalized treatment recommendations. Here, we identify and evaluate the best set of predictors of PPDM-A prospectively using retrospective data from 820 patients with acute pancreatitis at four centers by machine learning approaches. We used the L1 regularized logistic regression model to diagnose early PPDM-A via nine clinical variables identified as the best predictors. The model performed well, obtaining the best AUC = 0.819 and F1 = 0.357 in the test set. We interpreted and personalized the model through nomograms and Shapley values. Our model can accurately predict the occurrence of PPDM-A based on just nine clinical pieces of information and allows for early intervention in potential PPDM-A patients through personalized analysis. Future retrospective and prospective studies with multicentre, large sample populations are needed to assess the actual clinical value of the model. ## Introduction Acute pancreatitis (AP) is one of the most common gastrointestinal diseases characterized by acute pancreas inflammation and acinar cell destruction. The global incidence of AP is increasing dramatically worldwide1–3. AP was originally thought to be a self-limiting disease. Most patients recover completely, but only about $20\%$ may develop severe acute pancreatitis (SAP), with a mortality rate of around $30\%$4,5. However, diabetes, as a sequela of AP, has drawn attention in recent years. National Population-Based cohort studies reveal that the risk of PPDM-A is twofold higher than those without AP, and the occurrence of PPDM-A is observed across the spectrum of severity in AP6–8. Its prevalence has tripled in the past decade and is expected to reach 13.6 per 100,000 by 20509. PPDM-A has gained more and more attention in the field of diabetes. Pancreas contains both exocrine and endocrine parts. The exocrine pancreatic disease can lead to diabetes of the exocrine pancreas (DEP), the second most common type of new-onset diabetes in adults (surpassing type 1 diabetes)10. Furthermore, AP is considered the most common cause of DEP, and about $80\%$ of pancreatitis-related DEP is due to AP11. This type of diabetes is characterized by impaired endocrine and exocrine functions, including significant glycemic drift, frequent episodes of hypoglycemia (fragile diabetes)10, and impaired digestion and absorption of nutrients12,13. It seriously threatens human health and places a heavy burden on health care14,15. However, PPDM-A has not drawn sufficient attention and has often been misdiagnosed as T2DM12,16. Diabetes mellitus is highly heterogeneous and requires fine diagnostic staging for precise treatment. Therefore, screening high-risk patients is essential for developing PPDM-A prevention guidelines, delaying islet function damage, avoiding adverse outcomes, and improving the prognosis of PPDM-A. In this study, we used a machine learning model to screen the most important clinical features for predicting PPDM-A and used these clinical features to construct a logistic regression model with L1 regularisation, obtaining good AUC and F1 values. In addition, we interpreted the predictions of the model using nomograms and Shapley values and provided personalized early prevention protocols. This study provides a valuable guide to the occurrence and prevention of PPDM-A. ## Study cohort and baseline characteristics Between 1 October 2016 and 31 October 2021, 3477 admissions for AP were screened in hospital information system (HIS). After using the exclusion criteria described (Supplementary Fig. 1), 820 patients with AP without known diabetes were included in our study. Of these, two-thirds ($$n = 574$$) were randomly assigned to the training set, with the remaining one-third ($$n = 246$$) assigned to a validation cohort. Table 1 shows the baseline characteristics of the patients. The median age was 50 [38, 63] years. The proportion of males was $61.3\%$ ($$n = 503$$). Biliary was the most common cause of AP. 484 ($59\%$) patients presented with mild AP, 280 ($34.1\%$) with moderate AP, and 56 ($6.8\%$) with severe AP; 68 ($8.3\%$) patients had PPDM-A, and they were more likely to be obese ($20.7\%$ vs. $9.2\%$, $$P \leq 0.005$$), presenting with hyperlipidemia and tending to have combined non-alcoholic fatty liver disease (NAFLD) ($75\%$ vs. $45.2\%$, $P \leq 0.001$). Smoking rates were higher in patients with PPDM-A than in those without DM.Table 1Patient demographics and clinical characteristics. CharacteristicsTotal ($$n = 820$$)PPDM-A ($$n = 68$$, $8.3\%$)AP without DM ($$n = 752$$, $91.7\%$)P valueAge, years50 [38, 63]49.5 (37, 56.8)50 [38, 63]0.480Gender, male, n (%)503 (61.3)47 (69.1)456 (60.6)0.169Length of stay, d9 [7,14]11 [7, 15]9 [7, 14]0.206Obesity, n (%)81 (9.9)12 (20.7)69 (9.2)0.005Family history of diabetes, n (%)57 [7]9 (13.2)48 (6.4)0.060Smoking, n (%)206 (25.1)25 (36.8)181 (24.1)0.021Drinking, n (%)196 (23.9)16 (23.5)180 (23.9)0.940NAFLD, n (%)391 (47.7)51 [75]340 (45.2) < 0.001SBP (mmHg)130 [120, 144]132 (120, 146.5)130 [120, 144]0.357DBP (mmHg)80 [72,88]80 [76,87]80 [72,88]0.511Infection (n, %)361 [44]34 [50]327 (43.5)0.300Hypertension, n (%)219 (26.8)24 (35.3)195 [26]0.097Severity, n (%) Mild484 [59]36 (52.9)448 (59.6)0.026 Moderate280 (34.1)22 (32.4)258 (34.3) Severe56 (6.8)10 (14.7)46 (6.1)Amylase (U/L)560 (198.3, 1300)269 (107.5, 985.3)597.5 (216.5, 1300)0.009Admission glucose (mmol/L)6.66 (5.6, 8.3)8.56 (6.95–11.64)6.52 (5.54, 8) < 0.001Ca (mmol/L)2.20 (2.08, 2.31)2.19 (2.04, 2.27)2.20 (2.09, 2.32)0.162ALT (U/L)51.1 (23, 148.9)33.5 (21.1, 71.75)53.45 (23, 155.9)0.047AST (U/L)33 (19.6, 89.45)27.1 (19.9, 54.03)34.8 (19.6, 94.58)0.139ALP (U/L)88.15(65.78, 134.2)87 (62.65, 115.23)88.85 [66, 135]0.259LDH (IU/L)244.45 [188, 343]282.5 (201.69, 384.75)242 (190.7, 340)0.235CK (IU/L)68 [45, 105]69.5 (51.75, 111.98)68 (44.08, 104.25)0.231BUN (mmol/L)4.29 (3.3, 5.6)4.7 (3.9, 6.3)4.2 (3.2, 5.5)0.013Cr (umol/L)65 [52, 77]67.9 (56.25, 84)64.4 [52, 76]0.052UA (umol/L)295 (226, 363.9)332.5 (276, 405.2)289.85 [224, 357]0.003TG (mmol/L)1.24 (0.83, 2.95)3.53 (1.3, 8.77)1.19 (0.8, 2.64) < 0.001TC (mmol/L)4.41 (3.65, 5.54)5.33 (4.19, 7.47)4.35 (3.6, 5.37) < 0.001HDL-C (mmol/L)1.02 (0.75, 1.3)0.995 (0.41, 1.38)1.02 (0.75, 1.29)0.869LDL-C(mmol/L)2.40 (1.88, 3.06)2.55 (1.70, 3.32)2.40 (1.88, 3.05)0.051ANC, n (%)36 (4.4)6 (8.8)30 [4]0.120APFC, n (%)290 (35.4)24 (35.3)266 (35.4)0.990Values are given as median (IQR) or frequencies (percentages).SBP systolic blood pressure, DBP diastolic blood pressure, ALT alanine transaminase, AST aspartate aminotransferase, ALP alkaline phosphatase, LDH lactate dehydrogenase, NAFLD non-alcoholic fatty liver disease, Cr creatinine, BUN blood urea nitrogen, UA uric acid, TG triglyceride, TC total cholesterol, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, APFC acute peripancreatic fluid collection, ANC acute necrotic collection. ## Feature extraction Lasso regression (L1 regularized logistic regression) can be used for feature extraction in classification models. We performed 1000 randomly perturbed lasso regressions to extract the weights of 38 clinical features. By ranking the mean weights of these 38 features and using a threshold of 0.01, we obtained the nine most influential indicators on the classification of the model (Fig. 1), in order of Admission glucose, obesity (BMI > 28 kg/m2), cardiovascular disease (CVD), Age, NAFLD, alanine transaminase (ALT), uric acid (UA), HDL-C < 1.03 mmol/l, Smoking. In addition, indicators with a residual range above 0 included several features, drinking, organ failure, acute peripancreatic fluid collection (APFC), blood urea nitrogen (BUN), creatinine, hypertension, amylase, Ca. The results show that the two most influential factors in PPDM-A are still Admission glucose and obesity. These two indicators are also those associated with type 2 diabetes. It suggests that type 2 diabetes and PPDM-A share common risk factors. Figure 1Core influencing factor screening. The mean values of the weights of the features were ranked by 1000 lasso regressions. Of these, nine features with Feature Importance Score > 0.01 were selected as core genes considered to be associated with PPDM-A. ## Algorithm performance Multiple machine learning algorithms were used to construct the classification models. Following the same approach, we constructed a classification model based on the core nine features. We validated the performance of the model on the training set using fivefold cross-validation (Fig. 2A,B; Supplementary Table 1). Additionally, we performed internal validation on the training set (Fig. 2C,D; Supplementary Table 2). We then tested these models in validation data (Fig. 2E,F; Supplementary Tables 3, 4). The results showed that the best model was obtained with LR L1($C = 1$) at the average level (AUC = 0.819, CA = 0.927, F1 = 0.912, Precision = 0.912, Recall = 0.927; Fig. 2E, Supplementary Table 3). For the prediction of positive events, LR L1($C = 1$) also achieved the best results (AUC = 0.819, CA = 0.927, F1 = 0.357, Precision = 0.625, Recall = 0.250; Fig. 2F, Supplementary Table 4). The previous analysis showed that the prognostic model constructed using the core nine features had the best predictive effect. Figure 2Model performance. Fivefold cross-validation was used to evaluate model performance in the training set. ROC curves and calibration curves were used to compare the strengths and weaknesses of the models. ( A,B) ROC andibration curves of the five machine learning models on the training set using fivefold cross-validation. ( C,D) internal validation on the training set. ( E,F) ROC andibration curves of the five machine learning models in the test set. From Supplementary Table 4, we can find that the model obtained by Logistic Regression (L1 regularization) performs best with AUC = 0.819, F1 = 0.357 in the validation set. ## Assessing the interpretability of model predictions We constructed a nomogram based on LR L1($C = 1$) for nine features. HDL-C < 1.03, CVD, and ALT were predicted to contribute to PPDM-A (Fig. 3A) negatively. To gain insight into the features that contributed most to the model prediction results, we used Sharply Values to assess the importance of core features for model evaluation (Fig. 3B). The factors that most impacted the model predictions included HDL-C < 1.03 mmol/l, BMI > 28 kg/m2, and Admission glucose. HDL-C < 1.03 = TRUE made the most significant contribution to predicting positive events. The contribution of BMI >28 = FALSE in predicting positive events was the opposite of BMI > 28 = TRUE. It suggests that obesity is also a causal factor for disease. Figure 3Model interpretation. We have used two methods of model interpretation. ( A) Nomogram. The trend and magnitude of the effect of the nine core factors on the prediction of positive events can be observed in the figure. Admission glucose, BMI > 28, Age, NAFLD, UA and Smoking are the risk factors for PPDM-A. In contrast, Cardiovascular disease, ALT, HDL-C < 1.03 are negative predicted factors. ( B) Sharpley value was used to explain the effect of the model on prediction. HDL-C < 1.03, BMI > 28 and Admission Glucose were the main factors affecting prediction. BMI > 28, Cardiovascular disease, HDL-C < 1.03 and Smoking are logistic variables, with 0 being FALSE and 1 being TRUE. ## Personalized diagnosis We used the RL L1 ($C = 1$) model constructed with nine features to assess the main influences on the predictions of the six samples using Sharp Value. The results showed that the main contribution to an optimistic prediction for sample 1 came from (BMI > 28 = 0) = FALSE, (HDL-C < 1.03 = 0) = FALSE, with contributions of 0.68, 0.06 in order (Fig. 4A). The probability of an optimistic prediction for sample 1 was 0.83. The main contribution to an optimistic prediction for sample 2 came from (BMI > 28 = 0) = FALSE, (HDL-C < 1.03 = 1) = FALSE, with contributions of 0.62, 0.22 in order (Fig. 4B). The probability that this sample was predicted to be positive was 0.74. For sample 3, (BMI > 28 = 0) = FALSE, (HDL-C < 1.03 = 1) = FALSE contributed a positive predictive likelihood of 0.65, 0.21. Thus BMI > 28 kg/m2 was the leading risk factor for this sample (Fig. 4C). Multiple clinical information in samples 4, 5, and 6 contributed less to the positive prediction (Fig. 4D,E,F). The probability of predicting the occurrence of PPDM-A in each of these samples was less than 0.13.Figure 4Personalized diagnosis. The risk factors for the three positive and three negative predicted samples in the prediction set were studied. ( A) Patient 1 had a BMI > 28 kg/m2 as the main risk factor and a predicted probability of developing diabetes of 0.83. ( B) Patient 2 had a BMI > 28 kg/m2, HDL-C > 1.03 mmol/l as the main factor and a predicted probability of developing diabetes of 0.74. ( C) Patient 3 had a BMI > 28 kg/m2 as the main risk factor and a predicted probability of developing diabetes of 0.89. ( D,E,F) Patient 4, Patient 5, Patient 6 have no factors that make a major contribution to predicting a positive event and all have a predicted probability of developing diabetes of less than 0.13. The contribution of risk factors to this patient can be observed in the graph. Red represents the positive contribution and blue represents the negative contribution. ## Discussion PPDM-A is the most common sequela of pancreatitis11,17,18, and is characterized by poorer glycaemic control, a higher risk of developing cancer, and a higher risk of mortality10,19–21. However, the pathogenesis of diabetes secondary to acute pancreatitis is convoluted, which makes early clinical identification challenging. In addition, there is still no good classification model to predict PPDM-A in advance. Our feature contribution analysis prompted us to try to build a simpler predictive model based on a minimum number of the most influential features. To this end, we could fully evaluate the model’s performance using only nine pieces of information obtained about the patient. This study examined the ability to use nine clinical features to predict PPDM-A, leading to early intervention and effective PPDM-A screening. Our results suggest that clinical features can accurately predict the risk of PPDM-A after the onset of acute pancreatitis, although none of the nine clinical features we included directly reflected islet cell function. The findings reveal that indicators related to pancreatic injury (APFC, PPC, ANC, WON, amylase) affected the predicted outcome during the feature selection process, consistent with previous studies22,23. However, growing evidence compels a reconsideration of the dogma: “β-cell destruction is the only underlying mechanism of diabetes after acute pancreatitis”. Our study reveals that age, BMI, metabolic status, and comorbidities play different roles in individuals and may lead to opposite outcomes. It may be due to the mutual influence of organs on each other in the case of imbalanced glucose metabolism24–26. We assessed the characteristics that had the most profound impact on the model's predictions by Shapley value. We found that admission glucose, obesity (BMI > 28 kg/m2), and HDL-C < 1.03 mmol /l were the three factors that had the most significant impact on the outcome, and this result is consistent with the results of feature extraction. Prior studies also have demonstrated that hyperglycemia during hospitalization of acute critical illness is associated with emergent diabetes and identifying patients for subsequent diabetes screening. In a Scottish retrospective cohort study, $2.3\%$ of patients with an emergency admission to a hospital without previously known diabetes were newly diagnosed with diabetes within 3-years27. In a nationwide national cohort of consecutive patients with acute myocardial infarction without known diabetes, hyperglycemia at admission was significantly associated with subsequent diabetes (odds ratio: 2.56; $95\%$ CI 2.15–3.06)28. Furthermore, changes in lipid metabolism and abnormal distribution of abdominal adipose tissue are significantly associated with PPDM-A. Our work has several clinical applications. Firstly, it can facilitate early intervention in patients at high risk of diabetes. Early intervention for the development of diabetes is not currently studied. However, based on the health management knowledge of type 2 diabetes29–32, we can assume that a combination of diet and exercise can significantly decrease the incidence of diabetes. Due to the low prevalence of PPDM-A, early analysis of the effectiveness of prevention strategies can present some challenges. Our model can identify and recruit people at high risk of developing PPDM-A above $70\%$. Therefore, the current study paves the way for future randomized controlled trials to investigate further the effectiveness of using the model for early prediction of PPDM-A and possible prevention interventions. Another influential application is to help construct effective screening methods for PPDM-A. The prevalence of PPDM-A can already be confirmed considerably by admission glucose, BMI metrics, and HDL-C at the onset of acute pancreatitis. This staged risk assessment model can be used in subsequent studies to construct more rational design protocols for prospective studies of PPDM-A. Finally, by using the Sharpely Value, we can predict the likelihood of PPDM-A occurring in patients and identify key causative factors that can be targeted to give personalized treatment recommendations. ## Limitations Our study has several limitations. Firstly, our model collected HIS data with inherent bias retrospectively from a small number of centers in urban China. Although we have tried to make use of existing knowledge about diabetes and AP in the selection of features in the HIS data, there are additional clinical features that may have been overlooked. These features may have better predictive effects. In addition, our overall data volume was inadequate. Although the sample size requirements for making inferences about the occurrence of PPDM-A using nine clinical information may be reduced, our model obtained low F1 and recall rates in predicting positive events, and these may have led to the under-recording of positive samples. Finally, the population to which the study applies is limited to HIS data information from the Chinese population, and the predictive value of its findings on other populations requires more data accumulation. In conclusion, our work shows that it is possible to make accurate predictions of PPDM-A early in the onset of AP through nine clinical variables. These results may have many implications for the health of patients with PPDM-A. Our predictive model could form the basis for diagnosis and selective screening for PPDM-A and allow for personalized advice to patients on PPDM-A prevention. Future prospective studies, as well as multicentre, multicohort prospective studies, are needed to assess the clinical value of the model. ## Study design and participants This multicentre cross-sectional follow-up study included all consecutive patients with first-episode AP admitted to Zhongda Hospital of Southeast University, Yixing Second People’s Hospital, First Affiliated Hospital of Xinjiang Medical University, and Hunan Provincial People’s Hospital from 1 October 2016 to 31 October 2021. The study was approved by the ethics committee of Zhongda Hospital, affiliated with Southeast University, and performed according to the Declaration of Helsinki and relevant regulations. Informed consent was obtained orally from all participants. Inclusion criteria were as follows:Diagnosis of AP based on international guidelines33;Age ≥ 18 years;Admitted with abdominal pain for < 48 h. Exclusion criteria were as follows:Recurrent AP;Chronic pancreatitis;History of diabetes, HbA1c ≥ $6.5\%$ or hypoglycemic drugs diagnosed before AP attack;History of malignant tumor;Severe heart, liver, kidney and other organ dysfunction;History of immune system diseases or hormone use;Mental illness, unable to cooperate with research;Pregnancy and lactation;Data missing > $10\%$Death during admission. ## Data collection Date of demographic parameters (gender, age), Family history of diabetes, smoking, drinking, Clinical comorbidities (CVD, NAFLD, and hypertension), vital signs (systolic blood pressure, diastolic blood pressure, body mass index (BMI)), laboratory studies (amylase, admission glucose, serum calcium, hepatic and renal functions, lipid profiles), severity and etiology of AP, length of stay, infection condition, local complications and systemic complications of AP were extracted through HIS. ## Definitions and classification BMI was calculated as weight (kg) divided by the square of height (m). According to the Guidelines for Prevention and Control of Overweight and Obesity in Chinese adults, obesity was defined as BMI greater than or equal to 28 kg/m2. The severity of AP was defined as mild, moderately severe, and severe according to the revised Atlanta classification33. Local complications include acute peripancreatic fluid collection(APFC), pancreatic pseudocyst(PPC), acute necrotic collection(ANC), and walled-off necrosis (WON)33. Signs of systemic inflammatory response syndrome (SIRS) defined by presence of two or more criteria: 1. Heart rate > 90 beats/min; 2. Core temperature < 36 °C or > 38 °C; 3. White blood count < 4000 or > 12,000/mm3; 4. Respirations > 20/min or PCO2 < 32 mm Hg. PPDM-A was defined as new onset diabetes more than 90 days after AP with no history of diabetes before the AP episode14,34 and absence of type 1 diabetes–associated autoimmunity. Organ failure is defined as a score of 2 or more for one of these three organ systems using the modified Marshall scoring system35. ## Statistical analysis Logistic regression L1 regularization was used to screen for appropriate features. We extracted features with weights > 0.01 as core classification features. The top 9 features were obtained for the follow-up study. We constructed five common machine learning algorithms using Orange336: logistic regression, neural networks, random forests, catBoost37, and SVM. Fivefold cross-validation was used to evaluate the predictive power of the model. Five metrics, Area under ROC(AUC), Classification accuracy (CA), F1, accuracy, and recall, were used to evaluate the model. AUC is the area under the receiver-operating curve. The larger the AUC, the better the model effect. CA is the proportion of correctly classified samples. F1 is a weighted harmonic mean of precision and recall. F1 can be used to evaluate the model's trade-off between precision and recall metrics. Precision is the proportion of true positives among instances classified as positive. Recall is the proportion of true positives among all positive instances in the data. In assessing the model's effectiveness, we primarily used ROC and Calibration curves in the test set to assess the predictive effectiveness of the model. We analyzed the average performance of the model ground and the performance of the predicted positive events separately. To understand the relationship between individual features and model output, we use Shapley values38, which can be used to evaluate the outcome of complex models and are particularly applicable to artificial neural networks and gradient boosting machines (CatBoost). By averaging over all samples, the Shapley values estimate the contribution of each feature to the overall model prediction. In addition, *Shapley analysis* allows the contribution of each individual risk factor to the diagnosis to be assessed. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31947-4. ## References 1. Lee PJ, Papachristou GI. **New insights into acute pancreatitis**. *Nat. Rev. Gastroenterol. Hepatol.* (2019.0) **16** 479-496. DOI: 10.1038/s41575-019-0158-2 2. Mederos M, Reber H, Girgis MJJ. **Acute pancreatitis: A review**. *JAMA* (2021.0) **325** 382-390. DOI: 10.1001/jama.2020.20317 3. Iannuzzi JP. **Global incidence of acute pancreatitis is increasing over time: A systematic review and meta-analysis**. *Gastroenterology* (2022.0) **162** 122-134. DOI: 10.1053/j.gastro.2021.09.043 4. Schepers NJ. **Impact of characteristics of organ failure and infected necrosis on mortality in necrotising pancreatitis**. *Gut* (2019.0) **68** 1044-1051. DOI: 10.1136/gutjnl-2017-314657 5. Forsmark CE, Vege SS, Wilcox CM. **Acute pancreatitis**. *N. Engl. J. Med.* (2016.0) **375** 1972-1981. DOI: 10.1056/NEJMra1505202 6. Das SL. **Newly diagnosed diabetes mellitus after acute pancreatitis: A systematic review and meta-analysis**. *Gut* (2014.0) **63** 818-831. DOI: 10.1136/gutjnl-2013-305062 7. Shen HN, Yang CC, Chang YH, Lu CL, Li CY. **Risk of diabetes mellitus after first-attack acute pancreatitis: A national population-based study**. *Am. J. Gastroenterol.* (2015.0) **110** 1698-1706. DOI: 10.1038/ajg.2015.356 8. Lee YK, Huang MY, Hsu CY, Su YC. **Bidirectional relationship between diabetes and acute pancreatitis: A population-based cohort study in Taiwan**. *Medicine* (2016.0) **95** e2448. DOI: 10.1097/md.0000000000002448 9. Cho J, Petrov MS. **Pancreatitis, pancreatic cancer, and their metabolic sequelae: Projected burden to 2050**. *Clin. Transl. Gastroenterol.* (2020.0) **11** e00251. DOI: 10.14309/ctg.0000000000000251 10. Woodmansey C. **Incidence, demographics, and clinical characteristics of diabetes of the exocrine pancreas (type 3c): A retrospective cohort study**. *Diabetes Care* (2017.0) **40** 1486-1493. DOI: 10.2337/dc17-0542 11. Petrov MS, Yadav D. **Global epidemiology and holistic prevention of pancreatitis**. *Nat. Rev. Gastroenterol. Hepatol.* (2019.0) **16** 175-184. DOI: 10.1038/s41575-018-0087-5 12. Ewald N. **Prevalence of diabetes mellitus secondary to pancreatic diseases (type 3c)**. *Diabetes Metab. Res. Rev.* (2012.0) **28** 338-342. DOI: 10.1002/dmrr.2260 13. Vujasinovic M. **Pancreatic exocrine insufficiency, diabetes mellitus and serum nutritional markers after acute pancreatitis**. *World J. Gastroenterol.* (2014.0) **20** 18432-18438. DOI: 10.3748/wjg.v20.i48.18432 14. Petrov MS, Basina M. **Diagnosis of endocrine disease: Diagnosing and classifying diabetes in diseases of the exocrine pancreas**. *Eur. J. Endocrinol.* (2021.0) **184** R151-r163. DOI: 10.1530/eje-20-0974 15. Jivanji CJ, Soo DH, Petrov MS. **Towards reducing the risk of new onset diabetes after pancreatitis**. *Minerva Gastroenterol. Dietol.* (2017.0) **63** 270-284. DOI: 10.23736/s1121-421x.16.02365-5 16. Hart PA, Andersen DK, Petrov MS, Goodarzi MO. **Distinguishing diabetes secondary to pancreatic diseases from type 2 diabetes mellitus**. *Curr. Opin. Gastroenterol.* (2021.0) **37** 520-525. DOI: 10.1097/mog.0000000000000754 17. Petrov MS. **Metabolic trifecta after pancreatitis: exocrine pancreatic dysfunction, altered gut microbiota, and new-onset diabetes**. *Clin. Transl. gastroenterol.* (2019.0) **10** e00086. DOI: 10.14309/ctg.0000000000000086 18. Wynne K, Devereaux B, Dornhorst A. **Diabetes of the exocrine pancreas**. *J. Gastroenterol. Hepatol.* (2019.0) **34** 346-354. DOI: 10.1111/jgh.14451 19. Cho J, Scragg R, Petrov MS. **Risk of mortality and hospitalization after post-pancreatitis diabetes mellitus vs. type 2 diabetes mellitus: A population-based matched cohort study**. *Am. J. Gastroenterol.* (2019.0) **114** 804-812. DOI: 10.14309/ajg.0000000000000225 20. Cho J, Scragg R, Petrov MS. **Postpancreatitis diabetes confers higher risk for pancreatic cancer than type 2 diabetes: Results from a nationwide cancer registry**. *Diabetes Care* (2020.0) **43** 2106-2112. DOI: 10.2337/dc20-0207 21. Cho J, Pandol SJ, Petrov MS. **Risk of cause-specific death, its sex and age differences, and life expectancy in post-pancreatitis diabetes mellitus**. *Acta Diabetol.* (2021.0) **58** 797-807. DOI: 10.1007/s00592-021-01683-0 22. Zhi M. **Incidence of new onset diabetes mellitus secondary to acute pancreatitis: A systematic review and meta-analysis**. *Front. Physiol.* (2019.0) **10** 637. DOI: 10.3389/fphys.2019.00637 23. Sheikh S. **Reduced β-cell secretory capacity in pancreatic-insufficient, but not pancreatic-sufficient, cystic fibrosis despite normal glucose tolerance**. *Diabetes* (2017.0) **66** 134-144. DOI: 10.2337/db16-0394 24. Castillo-Armengol J, Fajas L, Lopez-Mejia IC. **Inter-organ communication: A gatekeeper for metabolic health**. *EMBO Rep.* (2019.0) **20** e47903. DOI: 10.15252/embr.201947903 25. Pendharkar SA. **The role of gut-brain axis in regulating glucose metabolism after acute pancreatitis**. *Clin. Transl. Gastroenterol.* (2017.0) **8** e210. DOI: 10.1038/ctg.2016.63 26. Gillies NA, Pendharkar SA, Singh RG, Asrani VM, Petrov MS. **Lipid metabolism in patients with chronic hyperglycemia after an episode of acute pancreatitis**. *Diabetes Metab. Syndr.* (2017.0) **11** S233-s241. DOI: 10.1016/j.dsx.2016.12.037 27. McAllister DA. **Stress hyperglycaemia in hospitalised patients and their 3-year risk of diabetes: A Scottish retrospective cohort study**. *PLoS Med.* (2014.0) **11** e1001708. DOI: 10.1371/journal.pmed.1001708 28. Shore S. **Association between hyperglycemia at admission during hospitalization for acute myocardial infarction and subsequent diabetes: Insights from the veterans administration cardiac care follow-up clinical study**. *Diabetes Care* (2014.0) **37** 409-418. DOI: 10.2337/dc13-1125 29. Johansen MY. **Effect of an intensive lifestyle intervention on glycemic control in patients with type 2 diabetes: A randomized clinical trial**. *JAMA* (2017.0) **318** 637-646. DOI: 10.1001/jama.2017.10169 30. Gregg EW. **Association of an intensive lifestyle intervention with remission of type 2 diabetes**. *JAMA* (2012.0) **308** 2489-2496. DOI: 10.1001/jama.2012.67929 31. Tuomilehto J, Schwarz PE. **Preventing diabetes: Early versus late preventive interventions**. *Diabetes Care* (2016.0) **39** S115-120. DOI: 10.2337/dcS15-3000 32. Guasch-Ferré M, Willett WC. **The Mediterranean diet and health: A comprehensive overview**. *J. Intern. Med.* (2021.0) **290** 549-566. DOI: 10.1111/joim.13333 33. Banks PA. **Classification of acute pancreatitis–2012: Revision of the Atlanta classification and definitions by international consensus**. *Gut* (2013.0) **62** 102-111. DOI: 10.1136/gutjnl-2012-302779 34. **2. Classification and diagnosis of diabetes: Standards of medical care in diabetes-2022**. *Diabetes Care* (2022.0) **45** S17-s38. DOI: 10.2337/dc22-S002 35. Marshall JC. **Multiple organ dysfunction score: A reliable descriptor of a complex clinical outcome**. *Crit. Care Med.* (1995.0) **23** 1638-1652. DOI: 10.1097/00003246-199510000-00007 36. Demšar J. **Orange: Data mining toolbox in Python**. *J. Mach. Learn. Res.* (2013.0) **14** 2349-2353 37. 37.Dorogush, A. V., Ershov, V. & Gulin, A. CatBoost: Gradient boosting with categorical features support (2018). 38. Kalai E, Samet Dov. **On weighted Shapley values**. *Int. J. Game Theory* (1987.0) **16** 205-222. DOI: 10.1007/BF01756292
--- title: New insights into the neuroprotective and beta-secretase1 inhibitor profiles of tirandamycin B isolated from a newly found Streptomyces composti sp. nov. authors: - Thitikorn Duangupama - Jaturong Pratuangdejkul - Sumet Chongruchiroj - Pattama Pittayakhajonwut - Chakapong Intaraudom - Sarin Tadtong - Patcharawee Nunthanavanit - Weerasak Samee - Ya-Wen He - Somboon Tanasupawat - Chitti Thawai journal: Scientific Reports year: 2023 pmcid: PMC10038987 doi: 10.1038/s41598-023-32043-3 license: CC BY 4.0 --- # New insights into the neuroprotective and beta-secretase1 inhibitor profiles of tirandamycin B isolated from a newly found Streptomyces composti sp. nov. ## Abstract Tirandamycin (TAM B) is a tetramic acid antibiotic discovered to be active on a screen designed to find compounds with neuroprotective activity. The producing strain, SBST2-5T, is an actinobacterium that was isolated from wastewater treatment bio–sludge compost collected from Suphanburi province, Thailand. Taxonomic characterization based on a polyphasic approach indicates that strain SBST2-5T is a member of the genus Streptomyces and shows low average nucleotide identity (ANI) ($81.7\%$), average amino-acid identity (AAI) ($78.5\%$), and digital DNA-DNA hybridization (dDDH) ($25.9\%$) values to its closest relative, *Streptomyces thermoviolaceus* NBRC 13905T, values that are significantly below the suggested cut-off values for the species delineation, indicating that strain SBST2-5T could be considered to represent a novel species of the genus Streptomyces. The analysis of secondary metabolites biosynthetic gene clusters (smBGCs) in its genome and chemical investigation led to the isolation of TAM B. Interestingly, TAM B at 20 µg/mL displayed a suppressive effect on beta-secretase 1 (BACE1) with 68.69 ± $8.84\%$ inhibition. Molecular docking simulation reveals the interaction mechanism between TAM B and BACE1 that TAM B was buried in the pocket of BACE-1 by interacting with amino acids Thr231, Asp 228, Gln73, Lys 107 via hydrogen bond and Leu30, Tyr71, Phe108, Ile118 via hydrophobic interaction, indicating that TAM B represents a potential active BACE1 inhibitor. Moreover, TAM B can protect the neuron cells significantly (% neuron viability = 83.10 ± $9.83\%$ and 112.72 ± $6.83\%$) from oxidative stress induced by serum deprivation and Aβ1–42 administration models at 1 ng/mL, respectively, without neurotoxicity on murine P19-derived neuron cells nor cytotoxicity against Vero cells. This study was reportedly the first study to show the neuroprotective and BACE1 inhibitory activities of TAM B. ## Introduction Alzheimer’s disease (AD) is a common neurodegenerative disorder found in elderly people1. This disease is a common form of dementia, including loss of memory and cognitive function impairments2,3. The etiology of AD is unknown. However, several pathological hallmarks, oxidative stress, amyloid-β (Aβ) deposition, tau protein accumulation, and decreased levels of acetylcholine (ACh), show significant roles in the pathophysiology of this disease2. Nowadays, only the acetylcholinesterase inhibitors, rivastigmine, galantamine, donepezil, and a NMDA receptor antagonist, memantine, are approved by the US Food and Drug Administration (USFDA) for the treatment of AD2. Due to the complex nature of AD and these approved medications only delaying the progression of symptoms associated with AD, multi-targeting medication was needed to cure AD1,2. In addition, neuroprotection refers to mechanisms within the nervous system that relatively conserve the neuronal structure and/or function. Therefore, it is one of the strategies for treating neurodegenerative disorders, including AD1. The Aβ hypothesis is one of the Alzheimer’s pathological hallmarks. The Aβ, especially Aβ42, the oligopeptide with 42 amino acids involves in Alzheimer’s disease by generating oxidative stress in neuronal cells and inducing tau protein hyperphosphorylation resulting in toxicity in synapses and mitochondria. In addition, the aggregation of Aβ peptides leads to fibrils formation, which is lastly deposited as senile plaques2. The Aβ is derived from clavation of the amyloid precursor protein (APP) by beta-site APP cleaving enzyme-1 (BACE1, β-secretase) and γ-secretase. Thus, the inhibition of the proteolytic activity of BACE1 is interesting for Alzheimer’s disease treatment2. Natural products from microorganisms are a promising source for the pharmaceutical and biotechnological industries. Actinomycete, especially the genus Streptomyces, is known to be the largest taxon of the phylum Actinomycetota that have proven to be a rich source of secondary metabolites with diverse chemical structures and biological activities4. Streptomyces species are Gram-positive, filamentous actinobacteria that can produce both substrate and aerial mycelia, including spores directly on the aerial mycelium. The types of spore forms, e.g., straight to flexuous, verticillate, spiral, or open-loop, are commonly observed in this genus. All Streptomyces species are found to have LL-diaminopimelic acid in the peptidoglycan5. Many secondary metabolites isolated from Streptomyces spp. have been found to possess neuroprotective activity. For example, mescengricin from *Streptomyces griseoflavus* 2853-SVS4 could reduce the l-glutamate toxicity in chick primary mescencephalic neurons with an EC50 value of 6.0 nM6. Flaviogeranin, a 1,4-naphtoquinone derivative from Streptomyces sp. RAC226 possessed neuroprotective activity by preventing cell death in C6 cells treated with 100 nM of glutamate for 24 h with an EC50 of 8.6 nM7. Neuroprotectins A and B were isolated from Streptomyces sp. Q27107 and exhibited neuroprotective activity against the primary cultured chick telencephalic cells from glutamate- and kainite-induced neurotoxicities8. Lavanduquinocin is a carbazole derivative with an ortho quinone moiety isolated from the fermentation broth of Streptomyces viridochromogenes. This compound could reduce l-glutamate toxicity in neuronal hybridoma N18-RE-105 cells with EC50 values of 4.3 nM9. To date, only N-Methyl-D-Aspartate (NMDA) receptor antagonist drug (memantine), and AChE inhibitor drugs (Donepezil, and Galantamine) are approved for clinical treatment of AD10–14. Thus, an attempt to discover a drug candidate from new Streptomyces species habiting in unexamined environments is one strategy to discover the potential bioactive neuroprotective compounds from nature. Tirandamycins (TAMs), a group of tetramic acid antibiotic with a bicyclic ketal moiety, have been isolated from Streptomyces spp. Tirandamycins (TAMs) are generally known to be an antibacterial agents against Gram-positive and Gram-negative bacteria, including vancomycin-resistant enterococcus (VRE), and Escherichia coli15,16, with low cytotoxicity on human cells17. At the molecular level, TAMs are inhibitors of bacterial ribonucleic acid polymerase (RNAp) (TAM A, IC50 value of 0.8 mM)18. However, TAMs have never been reported to have neuroprotective properties. As part of our continuing work on the screening program for in vitro neuroprotective activity from secondary metabolites produced by Streptomyces species, we found that the crude extract from a culture broth of Streptomyces sp. SBST2-5T exhibited an ability for neuronal protection by showing 110.70 ± $10.11\%$ viability of neurons at the concentration of 1 ng/mL, suggesting the presence of compound(s) with the ability to prevent neuron cell death and led to the discovery of TAM B as the major component involved in this neuroprotective activity. In this study, we have additionally used molecular docking to understand the molecular mechanism of the BACE1 inhibition by TAM B. This is the first report on the mechanism of action of TAM B through anti-BACE1 and neuroprotective activities. The knowledge obtained in this study would be helpful in the rational design of TAM derivatives to be more effective in binding to the BACE1 and be essential in exploring TAM derivatives as lead compounds among natural neuroprotective and anti-BACE1 drugs for the treatment of AD. Besides, the genome-based taxonomic characterization of Streptomyces sp. SBST2-5T, the isolation and identification of the isolated compounds, the evaluation of anti-acetylcholinesterase (anti-AChE), anti-beta-secretase (anti-BACE1), anti-oxidative, neuroprotective, and cytotoxicity activities are also reported. ## Polyphasic taxonomic characterizations of strain SBST2-5T Investigating novel actinomycete from unexamined environments is one strategy to discover promising bioactive compounds from microbial resources. Wastewater treatment is a process used to remove contaminants from wastewater and convert it into an effluent that can be returned to the water cycle. During the wastewater treatment process, the environment inside the system has been found to have a high temperature and less oxygen content. Thus, actinomycetes living in the wastewater treatment process are expected to be different from other terrestrial actinomycetes. An actinobacterium, designed strain SBST2-5T, was isolated from the wastewater treatment bio–sludge compost collected from Suphanburi province, Thailand. Strain SBST2-5T grew easily on the International Streptomyces Project 2 medium (ISP 2). The mature aerial spore mass is grayish white on the culture media tested after 14 days of cultivation at 30 °C (Additional file: Fig. S1). Analysis of whole-cell hydrolysates of strain SBST2-5T revealed the presence of LL-diaminopimelic acid. Galactose, glucose, mannose, and ribose were detected as whole-cell sugars (Additional file: Fig. S2). It contained MK-9(H6) and MK-9(H8) as predominant menaquinones while MK-9(H4) was minor component. Diphosphatidylglycerol, phosphatidylethanolamine, phosphatidylglycerol, phosphatidylinositol, phosphatidylinositol mannoside, three unidentified phospholipids, and ninhydrin-positive lipid were detected (Additional file: Fig. S3). The predominant fatty acids (> $10\%$) were observed to be iso-C15:0 ($27.4\%$), iso-C16:0 ($23.0\%$), and anteiso-C15:0 ($15.7\%$) (Additional file: Table S1). This major fatty acid pattern was also found in *Streptomyces thermoviolaceus* NBRC 13905T with different proportions. Scanning electron microscopy shows that the isolate forms a long straight chain of spores with a hairy surface (Fig. 1). The morphological and chemotaxonomic properties of the isolate are consistent with its classification in the genus Streptomyces19. To identify the taxonomic status at the species level, we first analyzed the 16S rRNA gene sequence (1430 nt) of strain SBST2-5T, and the sequence was submitted to Genbank with the accession number LC430996. The result showed that strain SBST2-5T is a member of the genus Streptomyces, with S. thermoviolaceus NBRC 13905T being the closest neighbor exhibiting the highest 16S rRNA gene sequence similarity value of $98.9\%$ to strain SBST2-5T, slightly higher than the $98.6\%$ proposed threshold for species classification, as recommended by Kim et al.20. Recently genome-based taxonomic approaches such as ANI (< $95\%$)21, AAI (< 95–$96\%$)22, and dDDH (< $70\%$)23,24 have been used as the threshold recommended for differentiating strains into different species; the values in parentheses are cut-off values. Strain SBST2-5T was found to have the ANIb (81.7–$84.9\%$), AAI (78.5–$82.8\%$), and dDDH (25.9–$30.0\%$) values that were significantly lower than the above cut-off values for species demarcation, indicating that the strain may represent a new species of the genus Streptomyces (Additional file: Table S2). In addition, the taxonomic position of strain SBST2-5T in the 16S rRNA gene tree (ML tree) indicated that the strain was clustered in the genus Streptomyces and was placed in a different species node from S. thermoviolaceus NBRC 13905T, the closest relatives (Additional files: Fig. S4). In contrast, the position of strain SBST2-5T in the NJ tree formed a clade together with S. thermoviolaceus NBRC 13905T (Additional file: Fig. S5). The genome-based phylogenetic tree reconstruction by automatic selection of the most closely related type strains by autoMLST revealed that strain SBST2-5T was positioned with *Streptomyces emeiensis* CGMCC 4.3504T25, *Streptomyces griseoflavus* JCM 4479T26, and *Streptomyces ghanaensis* ATCC 14672T27,28. This result emphasized that the position of strain SBST2-5T was separated from the branches carrying S. thermoviolaceus NBRC 13905T (Fig. 2). Moreover, the strain could be distinguished from its closest related type strains, S. thermoviolaceus NBRC 13905T, S. emeiensis CGMCC 4.3504T, S. griseoflavus JCM 4479T, and S. ghanaensis ATCC 14672T by its physiological and biochemical characteristics (Additional file: Table S3). Unlike S. thermoviolaceus NBRC 13905T, strain SBST2-5T could utilize l-arabinose, d-cellobiose, d-galactose, myo-inositol, d-mannose, d-melibiose, d-raffinose, sucrose, d-trehalose, and d-xylitol as sole carbon sources and showed positive results for nitrate reduction, and decomposition of hypoxanthine but the closest related type strain could not. Additionally, strain SBST2-5T tolerates NaCl up to $6\%$, but S. thermoviolaceus NBRC 13905T could only tolerate NaCl up to $1\%$ (w/v). In addition, compared to strain SBST2-5T, S. emeiensis CGMCC 4.3504T was not able to use d-galactose, d-raffinose, sucrose as sole carbon sources and could not grow at 55 °C, while S. griseoflavus JCM 4479T did not use l-arabinose, d-galactose, d-raffinose, sucrose as sole carbon sources. Moreover, the ability to utilize l-arabinose, d-raffinose, sucrose, dl-2-aminobutyric acid, l-cysteine, l-methionine, and the production of α-mannosidase, trypsin were significant phenotypic differences between strain strain SBST2-5T and S. ghanaensis ATCC 14672T. It is evident from taxonomic data that strain SBST2-5T could be judged to represent a novel species of the genus Streptomyces, for which the name Streptomyces composti sp. nov. is proposed. Figure 1Scanning electron micrograph of strain SBST2-5T grown on ISP 2 agar for 14 days at 30 °C. Bar, 2 μm. Figure 2Phylogenomic analysis of strain SBST2-5T and type strains affiliated to the genus Streptomyces based on 120 bacterial conserved single copied gene sets of the members. The bootstrap values on the nodes are displayed by > 50. Bar, 0.05 represents the nucleotide substitution per position. ## Genome analysis for secondary metabolites of strain SBST2-5T Members of the genus Streptomyces are a vital microbial resource because of their potential to synthesize many bioactive natural products with therapeutic applications29. Since Streptomyces spp. have an essential capacity to produce valuable secondary metabolites, exploring new taxa is one strategy that leads to discovering new secondary metabolites from nature. To date, a bioinformatic tool such as antiSMASH30 has been used to describe and/or identify the biosynthesis gene clusters (BGCs) in the genome of the actinobacteria. It also could be used for addressing the natural product potential of Streptomyces spp. The draft genome Streptomyces sp. strain SBST2-5T is composed of 100 contigs with an N50 value of 166 kb. The total genome size is about 6.5 Mb, with genome coverage of 200X. The digital G+C content is $72.2\%$, indicating in the range of the genus Streptomyces19 (Additional file: Table S4). It was deposited in GenBank under the accession number JAATEM000000000. We used antiSMASH version 7.0 beta30 to preliminarily determine and compare the putative biosynthetic gene clusters in the genome of strain SBST2-5T. It was found that the genome of strain SBST2-5T was found to be rich in terpene and NRPS-independent-siderophore clusters. One of them showed $86\%$ homology to the tirandamycin (tam) biosynthetic gene cluster in Streptomyces sp. 307–931 (Additional files: Fig. S6 and Table S5). A detailed analysis of the detected cluster in Streptomyces sp. SBST2-5T showed its difference from the described tam cluster from Streptomyces sp. 307–9. However, the key tirandamycin-forming genes involved in core biosynthetic genes (open reading frame (Orf) 2 and Orf 6), additional biosynthetic genes (Orf3 and Orf12), transport-related gene (Orf13), regulatory genes (Orf11 and Orf15), and other genes (Orf5 and Orf7-9) were observed. In addition, additional biosynthetic genes (Orf1, Orf17, and Orf19) and other genes (Orf4, Orf10, Orf14, Orf16, and Orf18) were observed between the tirandamycin-forming genes in the cluster from the SBST2-5T strain. ( Additional files: Fig. S7 and Table S6). According to the taxogenomic features and the bioinformatic analysis of the gene clusters led to the discovery of TAM and represented the strain as a new TAM producer. It can be concluded that strain SBST2-5T is a potentially prolific source of TAM. ## Compounds isolated from Streptomyces sp. SBST2-5T After purification of the crude extract using several chromatographic steps, compound 1 was isolated from the fermentation broth of Streptomyces sp. SBST2-5T. Compound 1 was identified based on 2D NMR spectroscopic analyses as well as mass spectrometry. The 1H and 13C NMR spectra (Additional files: Figs. S8 and S9) of compound 1 were identical to those of TAM B described in the literature32. The NOESY spectral information (Fig. 3) confirmed that compound 1 had the same relative configuration as that of the previously reported for TAM B32. HRESIMS spectrum (Additional file: Fig. S10) showed a negative ion at m/z 432.1662 [M−H]−, confirming the molecular formula C22H27NO8. Moreover, the optical rotation value of compound 1 ([α]D − 30.19, MeOH; [α]D − 14.64, EtOH) was the same as that of TAM B ([α]D − 14, EtOH)32. Thus, compound 1 was identified as (−)-TAM B (Fig. 4). The complete 1H and 13C NMR spectral data of compound 1 were provided in additional file: Table S7. Many TAM analogs have been found in various strains of Streptomyces spp. such as S. flaveolus33, Streptomyces sp. 307–915, Streptomyces sp. SCSIO166634, Streptomyces sp. SCSIO 4139917, and S. tirandamycinicus35. It was also suggested that TAM is biosynthetically generated from hybrid polyketide synthase (PKS)/ non-ribosomal peptide synthase (NRPS)34. Furthermore, several previous reports indicated that TAM B had potent antibacterial activity against several strains of bacteria17,32,35,36, but there are no reports on its neuroprotective and anti-BACE1 activities. Figure 33D structure of compound 1 with the selected NOESY correlations. Figure 4Chemical structures of compound 1 isolated from Streptomyces sp. SBST2-5T. ## Neuroprotective, anti-acetylcholinesterase (anti-AchE), anti-β-secretase (anti-BACE1), antioxidative, and cytotoxic activities of TAM B Alzheimer’s disease (AD) is one of the major health problems related to age cognitive dysfunction, which involves the decrease of acetylcholine (ACh) neurotransmitters, the accumulation of beta-amyloid (Aβ), and neurofibrillary triangle, and the occurrence of oxidative stress37. A promising way to treat these problems is to protect neurons in the brain. Actinomycetes have been reported as valuable sources of neuroprotective agents6–9. Recently, *Microbispora hainanensis* strain CSR-4 has been reported to be an outstanding actinomycete that produces a new diterpene compound, 2α-hydroxy-8[14], 15-pimaradien-17, 18-dioic acid. This compound displayed acetylcholinesterase (AChE) inhibitory activity with IC50 values of 96.87 ± 2.31 μg/mL38. Microbispora sp. TBRC 6027 is a potential actinomycete that produces several new chromone derivatives revealing neuroprotective activity. These compounds can protect P19-derived neurons from the oxidative stress induced by serum deprivation at very low concentrations (1 ng/mL)39. As part of our continuing work on the search for neuroprotective compound(s) from the actinomycetes, the EtOAc crude extract of Streptomyces sp. SBST2-5T displayed neuroprotective ability against oxidative stress (OS) conditions, with %viability of 109.83 ± $1.24\%$ at 1 ng/mL, implying the presence of compound(s) with the ability to prevent neuronal cell death and led us to isolate the major component, TAM B. TAM B was tested for its effect on the viability of the P19-derived neuron at various concentrations (1–10,000 ng/mL). The result showed that TAM B at all the tested concentrations exhibited no neurotoxicity on the cultured neuron. TAM B at various concentrations of 1–10,000 ng/mL expressed neuron viability range from 95.42 ± $13.74\%$ to 116.53 ± $11.60\%$. Moreover, at a concentration of 1 ng/mL, TAM B promoted neuron viability higher than that of the control (TAM B showed 116.53 ± $11.60\%$ neuron viability) (Additional file: Fig. S11). Thus, TAM B at 1 ng/mL was further evaluated for its neuroprotective ability by serum deprivation and Aβ1-42 administration models. Removal of the serum from the cultured neuron condition can cause neurotoxicity by inducing oxidative stress to the cultured P19-derived neuron2,40,41. In addition, Aβ also induced oxidative damage to the neuron42. TAM B at 1 ng/mL significantly ($p \leq 0.05$ compared with oxidative stress condition) helped to protect the cultured P19-derived neuron from the oxidative stress induced by serum deprivation. 1 nM Quercetin was used as a positive control, which possessed 77.01 ± $14.74\%$ neuron viability, while the neuron cultured in the medium without serum (an oxidative stress condition) showed 36.30 ± $4.97\%$ neuron viability. TAM B at 1 ng/mL exhibited neuron viability at 102.10 ± $9.83\%$ (Additional file: Fig. S12). The neuroprotective ability against Aβ1–42 of TAM B was also evaluated. Aβ1–42 at 4.5 µM (4500 nM) showed toxicity on cultured P19-derived neurons by revealing neuron viability at 74.74 ± $9.05\%$. Co-administration of 4.5 µM Aβ1–42 with 1 ng/mL of TAM B showed great neuroprotective ability against Aβ1–42 by exhibiting a neuron viability of 112.72 ± $6.83\%$ (Additional file: Fig. S13). The in vitro anti-acetylcholinesterase (anti-AchE), anti-β-secretase (anti-BACE1), and anti-oxidant (DPPH radical scavenging assay) activities of TAM B were evaluated for preliminary studying of the mechanism of action of the compounds. In our experiment, it was found that TAM B (250 μg/mL) had very low AChE inhibitory activity (5.36 ± $1.29\%$), while the positive control, galantamine at 1 µg/mL, showed AChE inhibitory activity of 75.98 ± $1.65\%$. Likewise, TAM B displayed low anti-oxidation activity with an IC50 of 166.30 µg/mL. These values suggested that TAM B was not responsible for anti-AChE and DPPH radical scavenging ability because no anti-AChE and anti-oxidation activities were detected. Interestingly, TAM B (20 µg/mL) exhibited BACE1 inhibitory activity of 68.69 ± $8.84\%$. In comparison, quercetin (positive control) at 20 µg/mL showed a value of 67.56 ± $9.40\%$ for BACE1 inhibitory activity, indicating that TAM B had potential anti-BACE1 activity (Additional file: Table S8). In vitro cytotoxicity assays are used to predict the toxicity and assess compounds' safety to various host cells43. It is known that a valuable natural product should have no effect on cellular metabolism and exhibit no toxicity against the host. The effect of TAM B on cell viability was evaluated using the Vero cell line and the human embryonic kidney cell line, HEK293. TAM B at 1000 μg/mL displayed very low cytotoxic activity against Vero cells (%cell viability > $85\%$). In the case of cytotoxicity on HEK293 of TAM B, we found that TAM B at various concentrations of 0–10 μg/mL exhibited HEK293 cell viability range from 64.84 ± $12.64\%$ to 99.88 ± $7.49\%$ (Additional file: Fig. S14). In addition, Yu et al.36 have reported that TAM B shows no cytotoxicity against human hepatic cells. This evidence emphasizes that TAM B is a valuable microbial natural product and shows significant potential to act as a therapeutic drug for AD treatment. According to the anti-oxidation, anti-AchE, anti-BACE1, and neuroprotective activities of TAM B, we suggested that TAM B had a potential for AD treatment through neuroprotective and anti-BACE1 activities. ## Molecular docking and in silico assessment of binding mode between TAM B and human BACE1 Based on the experimental part, TAM B exhibited a potential effect as an inhibitor for human BACE1. Molecular modeling was applied for a virtual understanding of the binding mode and interactions of TAM B in the binding site of human BACE1 using molecular docking and receptor-ligand interactions analyses. To assess the reliability of the docking protocol used in this study, a three-dimensional structure of energy-minimized atabecestat (PubChem CID 68254185) was docked into the binding site of prepared and energy minimized human BACE1 in apo form. The center coordinate and size of search space for docking were generated by covering the active sites of human BACE1 as represented in Fig. 5. The amino acid residues lining in the binding site were shown, including Asp228, Asp32, Tyr71, Gly230, and Ala335, which were identified as key residues for binding of atabecestat, a recently efficacious BACE1 inhibitor that was entered into the EARLY Phase 2b/3 clinical trial for the treatment of preclinical AD patients44. Among 20 docking poses, the best-docked pose of atabecestat presented the lowest negative binding energy score of − 8.0 kcal/mol (Additional files: Table S9 and Fig. S15). The docked conformation with respect to the X-ray crystal conformation of atabecestat in the binding site of human BASE1 (PDB ID: 7CDZ) showed the heavy atom RMSD value of 1.43 Å (Fig. 6) with well-oriented alignment indicating a significantly reliable protocol since the threshold of reliability was 2.0 Å for a good docking45. The structure of energy-minimized TAM B (PubChem CID 54728535) was then docked into the binding site of human BACE1 using the same protocol as atabecestat docking condition. The results of docking analyses showed that all binding energies of 20 docked poses of TAM B were in the range of − 6.8 to − 8.3 kcal/mol (Additional files: Table S9 and Fig. S16). The best docking pose of TAM B in the binding site of human BACE1, showing the binding energy score of − 8.3 kcal/mol was selected. To better understand the binding mode and interaction of TAM B, the predicted structure of tirandamycin-human BACE1 complex was submitted for further energy minimization prior to the analysis of binding interactions within 3.5 Å around TAM B. The interactions of the docked TAM B inside the binding site of BACE1 are shown in Fig. 7. The results showed that the NH2 group on the pyrrolidine-2,4-dione ring of the ligand showed a conventional hydrogen bond with the OH group of Thr231 at distances of 2.20 Å. This NH2 group was also in the vicinity of Asp 228, which formed a carbon-hydrogen bond with the hydrogen in the ring. Hydrogen bonds were observed between the hydroxy methyl group of TAM B and the CO group of Gln73 and Lys107 at distances of 2.89 and 2.46 Å, respectively. Leu30, Tyr71, Phe108, and Ile 118 mediated the alkyl and mixed pi-alkyl hydrophobic interactions. The calculated binding energy (BE) between TAM B and human BACE1 was − 105.66 kcal/mol. Moreover, the interaction energy between sets of atoms across all conformations was calculated using CHARMm. The interaction energy is defined as the sum of the van der Waals (VDW) and electrostatic energy. This nonbonded interaction energy was reported as energy values for each amino acid residue in the binding pocket within 3.5 Å interacting with TAM B (Table 1). The total interaction energy of − 62.82 kcal/mol was found. It comprised total VDW interaction energy and total electrostatic interaction energy of − 25.58 and − 37.24 kcal/mol, respectively. Considering molecular docking and binding mode analysis of tirandamycinB in human BACE1, all results indicated that TAM B could be a potential active human BACE1 inhibitor. Figure 5The three-dimensional structure of human BACE1 representing the search space size of 15 × 15 × 15 Å covering the binding site area for docking protocol (A). The amino acid residues lining in the binding site were depicted, including key residues (Asp32, Tyr71, Asp228, Gly230 and Ala335) for binding of atabecestat (B).Figure 6The reliability of docking protocol represented by the re-docked pose of atabecestat into the binding site of human BACE1. The structures of atabecestat were represented in stick by superposition of docking pose (green-carbon) and co-crystallized structure (cyan-carbon).Figure 7The binding mode of TAM B in the binding site of human BACE1. The binding residues within a radius of 4 Å from the bound TAM were illustrated in 3D graphic (A) and 2D diagram (B), indicating the types of binding interactions and interacting amino acids in the active site of human BACE1.Table 1Interaction energies per-residue within 4 Å of TAM B in the binding site of human BACE1.ResidueTotal interaction energy (kcal/mol)VDW interaction energy (kcal/mol)Electrostatic interaction energy (kcal/mol)LEU30− 1.44− 0.85− 0.59ASP32− 7.71− 1.10− 6.61SER35− 2.04− 0.56− 1.48TYR71− 6.57− 4.55− 2.02GLN73− 6.14− 1.40− 4.73GLY74− 0.53− 0.43− 0.10LYS107− 5.66− 1.88− 3.78PHE108− 1.34− 2.521.18ILE110− 2.15− 1.45− 0.70ILE118− 2.27− 1.14− 1.13ILE226− 1.19− 0.80− 0.39ASP228− 13.41− 1.39− 12.03GLY230− 1.07− 1.550.48THR231− 7.21− 2.56− 4.65ARG235− 2.77− 1.85− 0.92VAL332− 1.32− 1.540.23SUM− 62.82− 25.58− 37.24 ## Preliminary in silico pharmacokinetic ADME of TAM B The preliminary pharmacokinetics ADMET properties of TAM B were predicted using the ADMETlab 2.0 web-based application46. The predicted physicochemical characteristics and ADMET properties were included (Additional file: Fig. S13). Based on Lipinski`s Rule for the central nervous system drugs (RoCNS), TAM B revealed closely CNS drug-likeness properties i.e., molecular weight = 433.170 (MW < 400), LogP = 3.081 (CLogP ≤ 5), a number of H-bond acceptor = 9 (HBA ≤ 7) and a number of H-bond donor = 4 (HBD ≤ 5)47,48. However, the topological polar surface area (TPSA) of compound 1 was 141.610 Å2 which was higher than 90 Å2 as a cut-off for optimal CNS exposure49. TAM B contains a number of rotatable bonds of 6 that was well agreed with the proposed guideline of a rotatable bond count < 8 as an attribute of a successful CNS drug candidate50. The distribution of TAM B was predicted as the probability of being blood–brain barrier (BBB) penetration. Accordingly, TAM B may be further optimized and/or experimentally measured in the BBB permeability to develop as a new lead for CNS active human BACE1 inhibitor in the field of Alzheimer’s disease pharmacotherapy. ## Description of Streptomyces composti sp. nov Streptomyces composti (com.pos’ti. N.L. gen. n. composti of compost). Cells are Gram-stain-positive and aerobic. It grows well on ISP 2, ISP 4, and nutrient agar. Grows moderately on ISP 3, ISP 5, ISP 6, and ISP 7. The growth on Czapek’s sucrose and glucose-asparagine agar is poor. Yellowish-grey series substrate mycelium is observed on ISP 2, ISP 3, ISP 5, ISP 6, and ISP 7. Grayish white to light grey aerial spore masses are formed on ISP 2, ISP 4, ISP 6, ISP 7, and nutrient agar that differentiate into rectiflexibiles spore chains with a hairy surface, and spores are non-motile. The light brown diffusible pigment is observed on ISP 2. The reduction of nitrate, and production of catalase are positive. Negative results are observed for oxidase activity, liquefaction of gelatin, hydrolysis of starch, peptonization of milk, and urease production. Decomposes adenine, hypoxanthine, and l-tyrosine but not cellulose and xanthine. Utilizes l-arabinose, d-galactose, d-glucose, myo-inositol, inulin, d-mannitol, d-mannose, sucrose, d-trehalose, xylitol, and d-xylose; weakly utilizes adonitol, d-cellobiose, d-lactose, d-melezitose, d-melibiose, d-raffinose, d-salicin; but does not utilize dextran, l-rhamnose, d-ribose as sole carbon sources. Utilizes l-arginine, l-asparagine, l-histidine, l-methionine, l-phenylalanine, l-proline, l-serine, l-threonine, and l-valine; weakly utilizes DL-2-aminobutyric acid; but does not utilize l-cysteine and 4-hydroxyproline as sole nitrogen sources. The growth temperature is between 20 and 55 °C. Maximum NaCl for growth is $6\%$ (w/v). The pH range for growth is 5–10. According to the API ZYM system, cells show acid phosphatase, alkaline phosphatase, β-galactosidase, α-glucosidase, and leucine arylamidase activities. Cells produce weakly cystine arylamidase, esterase (C4), N-acetyl-β-glucosaminidase, and valine arylamidase but no activities of cells on α-chymotrypsin, esterase lipase (C8), α-fucosidase, α-galactosidase, β-glucosidase, β-glucuronidase, lipase (C14), α-mannosidase, naphthol-AS-BI-phosphohydrolase, and trypsin are observed. Cell wall peptidoglycan contains LL-diaminopimelic acid. The major menaquinones are MK-9(H6) and MK-9(H8), while MK-9(H4) is minor component. Galactose, glucose, mannose, and ribose are detected as whole-cell sugars. The phospholipid profile contains diphosphatidylglycerol, phosphatidylethanolamine, phosphatidylglycerol, phosphatidylinositol, phosphatidylinositol mannoside, and three unidentified phospholipids. The major fatty acids (> $10\%$) are iso-C15:0, iso-C16:0, anteiso-C15:0. The DNA G+C content of the type strain is $72.2\%$. The type strain, SBST2-5T (= TBRC 9952T = NBRC 113999T), is an actinomycete isolated from the wastewater treatment bio–sludge compost collected from Suphanburi province, Thailand. The GenBank accession number for the 16S rRNA gene sequence of strain SBST2-5T is LC430996. The whole-genome shotgun project has been deposited at GenBank under the accession JAATEM000000000. ## Conclusions In conclusion, we herein reported the identification and genome properties of Streptomyces sp. SBST2-5T. The genome-based taxonomic characterization revealed that Streptomyces sp. SBST2-5T merits classification as a novel species of the genus Streptomyces, for which we propose the name Streptomyces composti sp. nov. The strain represents a hybrid PKS/NRPS gene cluster in its genome involved in the TAM production, and led to the isolation of TAM B as a significant component in its EtOAc extract. To our knowledge, this is the first report of TAM B neuroprotective properties, reducing the toxicity of Aβ1–42, protecting the neuron from oxidative stress induced by serum deprivation and possessing anti-BACE1 activity, which position the compound to be used as multi-target anti-AD agents. An in silico study based on molecular docking simulation was performed to predict the binding mode of the compound into the binding pocket of the BACE1 enzyme and revealed that some amino acids of the BACE1 interact with the compound. Therefore, *It is* concluded that TAM B could be proposed as a potential active BACE1 inhibitor. ## Isolation, cultivation, and preservation of Streptomyces sp. SBST2-5T Strain SBST2-5T was isolated from the wastewater treatment bio–sludge compost collected from Suphanburi province, Thailand (14° 38′ 06ʺ N and 99° 50′ 45ʺ E). The air-dried compost was heated at 100 °C for 1 h. The diluted 1000-fold compost solution was prepared by serial dilution technique with $0.01\%$ sterile sodium dodecyl sulfate (SDS) in distilled water, and an aliquot (0.1 mL) of the sample suspension was taken and spread onto Zhang’s starch soil extract (ZSSE) medium51 supplemented with 50 mg nalidixic acid and 50 mg/L nystatin. After incubation at 30 °C for 14 days, a small light grey colony of strain SBST2–5T was picked and purified on yeast extract–malt extract agar (International Streptomyces Project, ISP 2 medium)52. The purified isolate was preserved at 4 °C on ISP 2 slant and in glycerol $20\%$ (v/v) suspension at – 80 °C or freeze-drying for long-term preservation. Streptomyces sp. SBST2-5T is deposited in Thailand Bioresource Research Center (TBRC) and NITE Biological Resource Center (NBRC) for code numbers TBRC 9952 and NBRC 113999, respectively (Related files: Figs. R1 and R2). ## Morphological, cultural, physiological, and biochemical characteristics Spore morphology was observed by a scanning electron microscope (SEM) using the culture grown at 30 °C for 14 days on ISP 2 agar. Samples for scanning electron microscopy were prepared according to the method of Duangupama et al.53. The arrangement of spore was observed under a scanning electron microscope (model JSM-6610 LV; JEOL). The cultural characteristic was determined by cultivating strain SBST2-5T on various International Streptomyces Project (ISP) media (ISP 2–7)52 at 30 °C for 14 days. The ISCC-NBS color charts were used for determining color designations and names54. Growth at different temperatures (4–60 °C) and different concentrations of NaCl [0–$10\%$ (w/v) at increments of $1\%$] were evaluated on ISP 2 agar after incubation for 14 days. Growth at different pH (4.0–11.0 at an increment of 0.5 pH unit) was determined by cultivation at 30 °C for 14 days in ISP 2 broth using acetate buffer (pH 4.0–5.0), phosphate buffer (pH 6.0–7.0, Tris–HCl (pH 8.0–9.0), glycine–NaOH buffer (pH 10.0) and carbonate bicarbonate buffer (pH 11.0) instead of distilled water. Carbon and nitrogen utilization ($1\%$, w/v), nitrate reduction, gelatin liquefaction, hydrogen sulfide and urease production, the decomposition of adenine, cellulose, hypoxanthine, L-tyrosine, and xanthine were evaluated using the previously described methods55–57. The reference strain, S. thermoviolaceus NBRC 13905T, was cultured under the same conditions for comparative analyses. ## Chemotaxonomic and 16S rRNA gene analyses Dried cells of strain SBST2-5T and S. thermoviolaceus NBRC 13905T, the reference strain, were prepared by cultivation in ISP2 broth on a rotary shaker (200 rpm) at 30 °C for 5 days. The isomer of diaminopimelic acid (DAP) in the peptidoglycan was evaluated using the methods of Hasegawa et al.58. Whole-cell sugar analysis was performed using the TLC method suggested by Komagata and Suzuki59. The polar lipids were extracted and analyzed using the standard protocols of Minnikin et al.60 and Collins and Jones61. To extract the menaquinone in the cells, the method of Collins et al.62 were used. The type of menaquinone was analyzed by high-performance liquid chromatography (HPLC)63. To prepare the cells for fatty acid analysis, strain SBST2-5T and its related type strain were cultivated in ISP 2 broth on a rotary shaker (200 r.p.m.) at 30 °C for five days. The Sherlock Microbial Identification (MIDI system and the ACTIN version 6 database were used for determining the fatty acid components64,65. Genomic DNA was isolated from cells grown in ISP 2 broth according to the method of Tamaoka66. The amplification and sequencing of the 16S rRNA gene were carried out using the method suggested by Nakajima et al.67. For calculating and comparing levels of similarity, the 16S rRNA gene sequence of strain SBST2-5T was analyzed using the EzBioCloud server68. The neighbour-joining (NJ), and maximum-likelihood (ML) trees based on 16S rRNA gene sequences were aligned and reconstructed using MEGA X69. The evolutionary distances among the strains were analyzed using Kimura’ s two-parameter method70. The confidence values of the branches were determined using 1000 replications of the bootstrap resampling method71. ## Genome-based taxonomic characterization and genome analysis for secondary metabolites of strain SBST2-5T Genomic DNA of strain SBST2-5T was sequenced using an Illumina Miseq platform (Illumina, Inc., San Diego, US-CA) with Reagent Kit V3 (600 cycles) using 2 × 250 bp paired-end reads (Chulalongkorn University, Thailand). Sequencing libraries were prepared using the QIAseq FX DNA Library Kit (Qiagen, USA). 100 ng of gDNA was subjected to DNA sequencing library preparation using QIASEQ FX DNA library preparation kit (Qiagen, USA). The quality of the raw reads was checked using FASTQC (Babraham Bioinformatics). Trim Galore (Babraham Bioinformatics) was used to remove the low-quality read and adaptors. SPAdes72 was used for assembling the genome. Genome annotation reports were created by the Prokka software 1.1273. The average nucleotide identity (ANI) result was calculated using the Jspecies74. The average amino acid identity (AAI) value was evaluated using the Kostas Lab ANI calculator75. The digital DNA-DNA hybridization (dDDH) values were calculated using the genome-to-genome distance calculator (GGDC 2.1; blast + method)76 on the TYGS type strain genome server (https://tygs.dsmz.de/)77. The presence of biosynthetic gene clusters in the genome of strain SBST2-5T was analyzed using antiSMASH30. To determine the genes related to tirandamycin production, the genome of strain SBST2-5T was analysed using blastp on the Uniprot database with matrix; blosum62 (https://www.uniprot.org/blast)78. To analyze the taxonomic position of strain SBST2-5T in phylogenomic tree, the genomes affiliated with the genus Streptomyces were downloaded from NCBI, and the genome quality was checked using CheckM v1.0.179. Genomes of $10\%$ contamination were discarded from the study. To construct the phylogenomic tree, an automated multi-locus species tree (autoMLST) pipeline (https://automlst.ziemertlab.com/)80 was used. The final phylogenomic tree was reconstructed using the maximum-likelihood algorithm using MEGA X, with *Micromonospora carbonacea* DSM 43168T (GCF_900091535) as an outgroup. ## Fermentation, extraction, and isolation of a bioactive substance Streptomyces sp. SBST2-5T was grown on ISP2 agar for 7 days. The agar was cut into small pieces and then transferred into 10 × 1 L Erlenmeyer flasks, each containing 250 mL of ISP 2 broth. The seed culture was cultivated at 30 °C on a rotary shaker at 180 rpm for 6 days. Then 25 mL of seed culture was transferred into 80 × 1 L Erlenmeyer flasks, each containing 250 mL of ISP2 medium. The flaks were shaked at 30 °C, 200 rpm, and after 6 days, the whole culture was extracted three times with an equal volume of EtOAc, which was later dried over Na2SO4 and evaporated to yield a reddish-brown gum (1.74 g). The gum was passed through a Sephadex LH20 column (1.5 × 40 cm), eluted with $100\%$ MeOH, and 10 mL fractions were collected for 100 tubes. Each tube was analyzed by HPLC and then concatenated to give two fractions. Both fractions were further purified by a preparative HPLC, using a Sunfire C18 column (diam. 19 × 250 mm, particle size 10 mm) and eluted with a linear gradient system of 5–$70\%$ CH3CN in water over 21 min to furnish compound 1 (tR 19.38 min, 163.3 mg). (−)-Tirandamycin B [1]: pale yellow solid; [α]25D − 30.19 (c 0.34, MeOH) and [α]27D − 14.64 (c 0.16, EtOH); UV (MeOH) λmax (log ε) 202 (4.03), 250 (3.75), 290 (3.82), 339 (4.10); HRESIMS m/z 432.1662 [M − H]− (calcd for C22H26NO8, 432.1664). ## Structure identification of the active compound UV spectrum was performed in MeOH on a Spekol 1200 spectrophotometer from Analytik Jena. Optical rotation was measured with a JASCO P-1030 digital polarimeter. FT-IR spectrum was measured on a Bruker ALPHA spectrometer. NMR spectra were acquired in CDCl3 on a Bruker Avance 500 MHz NMR spectrometer. HRESIMS data was determined on a Bruker MicrOTOF mass spectrometer. ## Antioxidant assay The antioxidant capacity was estimated in terms of radical scavenging activity according to a modified version of Brand-Williams method81. Briefly, 100 µL of tested compound (1–1000 µg/mL dissolved in MeOH) was mixed with 100 µL of freshly prepared DPPH solution (3 × 10−5 M dissolved in MeOH). The reaction mixture was incubated for 30 min. The absorbance was read at 517 nm. Each assay was done in triplicate. Ascorbic acid was used as a positive control. ## Anti-acetylcholinesterase (anti-AChE) assay The AChE inhibitory effect was modified from the spectrophotometric assays described by Panyathip et al.2. In brief, 25 µL of 0.2 U/mL AChE from *Electrophorus electricus* and 25 µL of sample dissolved in 50 mM TRIS–HCl buffer pH 8.0 were added to the 96-well and incubated at 37 °C for 15 min. Then 125 µL of 3 mM 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB) and 25 µL of 1.5 mM acetylthiocholine iodide (ATCI) were added into the 96-wells and incubated at 37 °C for 15 min. The absorbance was measured at 405 nm by a microplate reader (Spectramax, USA). Galanthamine (8 µg/mL in 50 mM TRIS–HCl buffer pH 8.0) was used as the positive control. Each assay was done in 3 independent experiments, and each experiment was run in triplicate. Data were reported as average % inhibition ± SD. ## Beta-secretase 1 (BACE1) inhibition assay The BACE1 inhibitory effect was modified from the assays described by Puksasook et al.1 and Panyathip et al.2. In brief, 20 µL of 0.1 U/mL BACE1 and 20 µL of sample dissolved in 100 mM sodium acetate buffer pH 4.5 were added to the 96-well and incubated at 37 °C for 10 min. Then 50 µL of 750 mM β-secretase substrate in 100 mM sodium acetate buffer pH 4.5 was added and incubated at 37 °C for 20 min. After that, 10 µL of stop buffer (2.5 M sodium acetate buffer pH 4.5) was added to the 96-wells and measured the fluorescence at 380 nm (excitation wavelength) and 510 nm (emission wavelength) by a microplate reader (Spectramax, USA). Quercetin (2 µg/mL) was used as the positive control. Each assay was done in 3 independent experiments, and each experiment was run in triplicate. Data were reported as average % inhibition ± SD. ## Neuroprotective assay by serum deprivation method The assay was performed in a 96-well plate for 3 independent experiments and each experiment was run in triplicate. First, the sample was evaluated for its ability to enhance the viability of the cholinergic P19-derived neuron by XTT reduction assay as described previously2,82,83 at various concentrations (1–10,000 ng/mL). Then, the concentration that enhances the viability of the neuron more than the %neuron viability of the control ($0.5\%$ DMSO in the medium was used as a control representing no effect on the neuron viability) will be selected to further evaluate for neuroprotective activity by serum deprivation method. Quercetin at 1 nM concentration was used as a positive control for the neuroprotective assay82. Data were reported as average %neuron viability ± SD. ## Neuroprotective assay by Aβ1–42 co-administration method The assay was performed in a 96-well plate for three independent experiments, and each experiment was run in triplicate. The concentration of Aβ1–42 at 4.5 µM that gave $70\%$ neuron viability was used as a neurotoxin, and the sample at the concentration that enhanced the viability of the neuron more than the %neuron viability of the control ($0.5\%$DMSO in the medium was used as a control representing no effect on the neuron viability) will be selected to further evaluate for neuroprotective activity by co-administration with 4.5 µM Aβ1–42. Data were reported as average % neuron viability ± SD. ## Cytotoxicity assay The assay was done in a 96-well plate for three independent experiments, and each experiment was run in triplicate. Human embryonic kidney cell line, HEK293 (ATCC CRL-1573), cultured in the conditions recommended by ATCC (Eagle’s minimum essential medium (EMEM) supplemented with $10\%$v/v FBS and $1\%$v/v antibiotic–antimycotic solution). The sample was evaluated for its cytotoxicity on cultured HEK293 cells by XTT reduction assay2 at various concentrations (1–10,000 ng/mL). $0.5\%$ DMSO in the medium was used as a control representing no cytotoxic effect on the cultured cell. Data were reported as average % cell viability ± SD. The green fluorescent protein microplate assay84 (GFPMA) was used for testing cytotoxicity against Vero cells (African green monkey kidney fibroblasts, ATCC CCL-81). Ellipticine was used as a positive control for cytotoxicity against Vero cells. Ellipticine was used as a positive control for cytotoxicity against Vero cells. All experiments were done in triplicate, and data are presented as mean ± standard deviation (mean ± SD). ## Structure preparations, molecular docking and binding mode analyses The X-ray crystal structure of human BACE1 complexed with atabecestat (N-{3-[(4S)-2-amino-4-methyl-4H-1,3-thiazin-4-yl]-4-fluorophenyl}-5-cyanopyridine-2-carboxamide) (PDB ID: 7DCZ) was retrieved from the RCSB Protein Databank (PDB, https://www.rcsb.org/), and used as the target for docking of TAM B44. The three-dimensional structure of TAM B with desired configuration ((3E)-3-[(2E,4E,6R)-1-hydroxy-6-[(1S,2S,4R,6S,7R,8R)-2-(hydroxymethyl)-1,7-dimethyl-5-oxo-3,9,10-trioxatricyclo[4.3.1.02,4]decan-8-yl]-4-methylhepta-2,4-dienylidene]pyrrolidine-2,4-dione) was obtained from PubChem CID 54728535. The preparation of the human BACE1 structure was done using “Prepare Proteins” in BIOVIA Discovery Studio software package (BIOVIA)85. Briefly, the human BACE1 structure (PDB ID: 7DCZ) was cleaned up by removing hetero atoms, including water and atabecestat. The amino acids in missing loop regions of the human BACE1 chain (158-GFPLNQSEVL-167) and A272 were inserted based on the SEQRES data. The amino acid residues in human BACE1 were protonated at pH 7.4 according to the predicted pK values. The protein backbone atoms were restrained with a force constant of 40 kcal/mol. The force field and charges were assigned using CHARMm forcefield86 and Momany-Rone87, respectively. The prepared structure of human BACE1 was submitted for energy minimization with the conjugate gradient algorithms, 10,000 steps, and RMS gradient of 0.001 kcal/(mol × Å). The structure of TAM B was submitted for energy minimization using the conjugate gradient method with 3000 steps, RMS gradient 0.001 kcal/(mol × Å), applying CHARMm forcefield. The energy-minimized structures of a target (human BACE1) and ligand (TAM B) were saved. The plausible binding mode of TAM B in human BACE1 was predicted using the molecular docking technique. Docking of TAM B into the binding site of human BACE1 was performed using Autodock Vina88 interface in VEGA ZZ4989. The input files in pdbqt format for Autodock Vina were prepared using receptor.c and ligand.c scripts in VEGA ZZ platform for protein and ligand, respectively. The center of search space was setup at -40.19, -1.54 and 10.61 for X, Y, Z coordinates based on the binding mode of atabecestat in human BACE1 structure (PDB ID: 7DCZ)48. The search space was set up using 15 × 15 × 15 Å size around the center. Exhaustiveness and binding modes were set at 128 and 20 for ligand docking to generate ten output poses for ligand. The best-docked pose was selected based on energy scoring functions (kcal/mol) and protein–ligand interactions for further analyses. The structure of human BACE1-TAM B complex was submitted for energy minimization with CHARMm forcefield using the conjugate gradient (50,000 steps) to obtain the stable structure with a convergence criterion of 0.001 kcal/(mol × Å) energy RMS Gradient. Harmonic restraint was applied on the protein backbone during the minimization steps with 40 kcal/mol/Å−2. The binding energy (BE in kcal/mol) of the complex was estimated between ligands and target using CHARMm-based energy methods90. The nonbonded interactions (i.e., van der Waals and electrostatic terms) between the TAM B and interacting amino acid residues within 3.5 Å around the ligand were calculated using the CHARMm-based Interaction Energy protocol. The molecular modeling tools for structure preparation, energy minimization, calculations of binding and interaction energies, graphical visualization, and analyses were performed using the BIOVIA Discovery Studio package software (BIOVIA). ## Preliminary in silico ADME prediction The preliminary pharmacokinetic ADME properties of TAM B were predicted using the ADMET lab 2.0 web-based application46. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-32043-3. ## References 1. Puksasook T. **Semisynthesis and biological evaluation of prenylated resveratrol derivatives as multi-targeted agents for Alzheimer's disease**. *J. Nat. Med.* (2017.0) **71** 665-682. DOI: 10.1007/s11418-017-1097-2 2. Panyatip P, Tadtong S, Sousa E, Puthongking P. **BACE1 inhibitor, neuroprotective, and neuritogenic activities of melatonin derivatives**. *Sci. Pharm.* (2020.0) **88** 58. DOI: 10.3390/scipharm88040058 3. Jiaranaikulwanitch J. **Novel multi-functional ascorbic triazole derivatives for amyloidogenic pathway inhibition, anti-inflammation, and neuroprotection**. *Molecules* (2021.0) **26** 1562. DOI: 10.3390/molecules26061562 4. Bérdy J. **Bioactive microbial metabolites**. *J. Antibiot.* (2005.0) **58** 1-26. DOI: 10.1038/ja.2005.1 5. Kämpfer P, Goodfellow M, Kämpfer P, Busse HJ, Trujillo ME, Suzuki K-I, Ludwig W, Whitman WB. **Genus**. *Bergey’s Manual of Systematic Bacteriology* (2012.0) 1455-1767 6. Kim JS, Shin-Ya K, Furihata K, Hayakawa Y, Seto H. **Structure of mescengricin, a novel neuronal cell protecting substance produced by**. *Tetrahedron Lett.* (1997.0) **38** 3431-3434. DOI: 10.1016/S0040-4039(97)00638-2 7. Hayakawa Y. **Flaviogeranin, a new neuroprotective compound from**. *J. Antibiot.* (2010.0) **63** 379-380. DOI: 10.1038/ja.2010.49 8. Kobayashi H. **Neuroprotectins A and B, bicyclohexapeptides protecting chick telencephalic neuronal cells from excitotoxicity. I. Fermentation, isolation, physico-chemical properties and biological activity**. *J. Antibiot.* (2001.0) **54** 1013-1018. DOI: 10.7164/antibiotics.54.1013 9. Shin-Ya K. **A new neuronal cell protecting substance, lavanduquinocin, produced by**. *J. Antibiot.* (1995.0) **48** 574-578. DOI: 10.7164/antibiotics.48.574 10. Whitehouse PJ. **Cholinergic therapy in dementia**. *Acta. Neurol. Scand.* (1993.0) **88** 42-45. DOI: 10.1111/j.1600-0404.1993.tb04254.x 11. Gottwald MD, Rozanski RI. **Rivastigmine, a brain-region selective acetylcholinesterase inhibitor for treating Alzheimer’s disease: Review and current status**. *Expert. Opin. Investig. Drugs.* (1999.0) **8** 1673-1682. DOI: 10.1517/13543784.8.10.1673 12. Ambure P, Roy K. **Advances in quantitative structure-activity relationship models of anti-Alzheimer’s agents**. *Expert. Opin. Drug. Discov.* (2014.0) **9** 697-723. DOI: 10.1517/17460441.2014.909404 13. Cummings JL. **Treatment of Alzheimer’s disease: Current and future therapeutic approaches**. *Rev. Neurol. Dis.* (2004.0) **1** 60-69. PMID: 16400259 14. Standridge JB. **Pharmacotherapeutic approaches to the prevention of Alzheimer’s disease**. *Am. J. Geriatr. Pharmacother.* (2004.0) **2** 119-132. DOI: 10.1016/S1543-5946(04)90017-7 15. Carlson JC, Li S, Burr DA, Sherman DH. **Isolation and characterization of tirandamycins from a marine-derived**. *J. Nat. Prod.* (2009.0) **72** 2076-2079. DOI: 10.1021/np9005597 16. Reusser F. **Tirandamycin, an inhibitor of bacterial ribonucleic acid polymerase**. *Antimicrob. Agents. Chemother.* (1976.0) **10** 618-622. DOI: 10.1128/AAC.10.4.618 17. Cong Z. **Cytotoxic anthracycline and antibacterial tirandamycin analogues from a marine-derived**. *J. Antibiot.* (2019.0) **72** 45-49. DOI: 10.1038/s41429-018-0103-6 18. Zhang X. **Identification of an unexpected shunt pathway product provides new insights into tirandamycin biosynthesis**. *Tetrahedron Lett.* (2016.0) **57** 5919-5923. DOI: 10.1016/j.tetlet.2016.11.080 19. Pridham TG, Hesseltine CW, Benedict RG. **A guide for the classification of streptomycetes according to selected groups; placement of strains in morphological sections**. *Appl. Microbiol.* (1958.0) **6** 52-79. DOI: 10.1128/am.6.1.52-79.1958 20. Kim M, Oh H-S, Park S-C, Chun J. **Towards a taxonomic coherence between average nucleotide identity and 16S rRNA gene sequence similarity for species demarcation of prokaryotes**. *Int. J. Syst. Evol. Microbiol.* (2014.0) **64** 346-351. DOI: 10.1099/ijs.0.059774-0 21. Richter M, Rosselló-Móra R. **Shifting the genomic gold standard for the prokaryotic species definition**. *Proc. Natl. Acad. Sci. USA* (2009.0) **106** 19126-19131. DOI: 10.1073/pnas.0906412106 22. Konstantinidis KT, Rosselló-Móra R, Amann R. **Uncultivated microbes in need of their own taxonomy**. *ISME J.* (2017.0) **11** 2399-2406. DOI: 10.1038/ismej.2017.113 23. Thompson CC. **Microbial genomic taxonomy**. *BMC Genomics* (2013.0) **14** 913. DOI: 10.1186/1471-2164-14-913 24. Auch AF, von Jan M, Klenk HP, Göker M. **Digital DNA–DNA hybridization for microbial species delineation by means of genome-togenome sequence comparison**. *Stand. Genom. Sci.* (2010.0) **2** 117-134. DOI: 10.4056/sigs.531120 25. Sun W, Huang Y, Zhang YQ, Liu ZH. *Int. J. Syst. Bacteriol.* (2007.0) **57** 1635-1639. DOI: 10.1099/ijs.0.64934-0 26. Goodfellow M, Lonsdale C, James AL, MacNamara OC. **Rapid biochemical tests for the characterisation of streptomycetes**. *FEMS. Microbiol. Lett.* (1987.0) **43** 39-44. DOI: 10.1111/j.1574-6968.1987.tb02094.x 27. Goodfellow M. *Antonie Van Leeuwenhoek* (2017.0) **110** 1133-1148. DOI: 10.1007/s10482-017-0886-7 28. Tian Y. *Int. J. Syst. Bacteriol.* (2018.0) **68** 3327-3332. DOI: 10.1099/ijsem.0.002993 29. Lewin GR. **Evolution and ecology of actinobacteria and their bioenergy applications**. *Annu. Rev. Microbiol.* (2016.0) **70** 235-254. DOI: 10.1146/annurev-micro-102215-095748 30. 30.Blin, K. et al. AntiSMASH bacterial version 7 beta. https://antismash.secondarymetabolites.org (2023). 31. Carlson JC. **Identification of the tirandamycin biosynthetic gene cluster from Streptomyces sp 307–9**. *ChemBioChem* (2010.0) **11** 564-572. DOI: 10.1002/cbic.200900658 32. Rinehart KL, MacKellar FA, Grostic MF, Olson EC, Wnuk RJ, Branfman AR. **Tirandamycin. I. Structure assignment**. *J. Am. Chem. Soc.* (1971.0) **93** 4943-4945. DOI: 10.1021/ja00748a067 33. Hagenmaier H, Jaschke KH, Santo L, Scheer M, Zähner H. **Stoffwechselprodukte von Mikroorganismen**. *Arch. Microbiol.* (1976.0) **109** 65-74. DOI: 10.1007/BF00425114 34. Mo X. **Cloning and characterization of the biosynthetic gene cluster of the bacterial RNA polymerase inhibitor tirandamycin from marine-derived**. *Biochem. Biophys. Res. Commun.* (2011.0) **406** 341-347. DOI: 10.1016/j.bbrc.2011.02.040 35. Huang X, Kong F, Zhou S, Huang D, Zheng J, Zhu W. *Front. Microbiol.* (2019.0) **10** 00482. DOI: 10.3389/fmicb.2019.00482 36. Yu Z. **Tirandamycins from**. *Org. Lett.* (2011.0) **13** 2034-2037. DOI: 10.1021/ol200420u 37. Terry AV, Buccafusco JJ. **The cholinergic hypothesis of age and Alzheimer’s disease-related cognitive deficits: Recent challenges and their implications for novel drug development**. *J. Pharmacol. Exp. Ther.* (2003.0) **306** 821-827. DOI: 10.1124/jpet.102.041616 38. Thawai C. **A novel diterpene agent isolated from**. *Sci. Rep.* (2020.0) **10** 11058. DOI: 10.1038/s41598-020-68009-y 39. Kittisrisopit S. **Neuroprotective potential of new chromones isolated from the soil actinomycete**. *Nat. Prod. Res.* (2019.0) **35** 2881-2886. DOI: 10.1080/14786419.2019.1679135 40. Iacovitti L, Stull ND, Johnston K. **Melatonin rescues dopamine neurons from cell death in tissue culture models of oxidative stress**. *Brain Res.* (1997.0) **768** 317-326. DOI: 10.1016/S0006-8993(97)00668-9 41. Atabay C, Cagnoli CM, Kharlamov E, Ikonomovic MD, Manev H. **Removal of serum from primary cultures of cerebellar granule neurons induces oxidative stress and DNA fragmentation: Protection with antioxidants and glutamate receptor antagonists**. *J. Neurosci. Res.* (1996.0) **43** 465-475. DOI: 10.1002/(SICI)1097-4547(19960215)43:4<465::AID-JNR7>3.0.CO;2-D 42. Zhang L. **Neuroprotective effects of salidroside against beta-amyloid-induced oxidative stress in SH-SY5Y human neuroblastoma cells**. *Neurochem. Int.* (2010.0) **57** 547-555. DOI: 10.1016/j.neuint.2010.06.021 43. Menendez JA, Vellon L, Lupu R. **Antitumeral actions of the anti-obesity drug orlistat (XenicalTM) in breast cancer cells: blockade of cell cycle progression, promotion, of apoptotic cell death and PEA3-mediated transcriptional repression of Her2lneu (erbB-2) oncogene**. *Ann. Oncol.* (2005.0) **16** 1253-1267. DOI: 10.1093/annonc/mdi239 44. Koriyama Y. **Discovery of atabecestat (JNJ-54861911): a thiazine-based β-amyloid precursor protein cleaving enzyme 1 inhibitor advanced to the phase 2b/3 EARLY clinical trial**. *J. Med. Chem.* (2021.0) **64** 1873-1888. DOI: 10.1021/acs.jmedchem.0c01917 45. Gohlke H, Hendlich M, Klebe G. **Knowledge-based scoring function to predict protein-ligand interactions**. *J. Mol. Biol.* (2000.0) **295** 337-356. DOI: 10.1006/jmbi.1999.3371 46. Xiong G. **ADMETlab 2.0: An integrated online platform for accurate and comprehensive predictions of ADMET properties**. *Nucleic. Acids. Res.* (2021.0) **49** W5-W14. DOI: 10.1093/nar/gkab255 47. Goldberg JS. **Low molecular weight opioid peptide esters could be developed as a new class of analgesics**. *Perspect. Med. Chem.* (2011.0) **5** 19-26 48. Fernandesa TB, Segrettib MCF, Pollic MC, Parise-Filho R. **Analysis of the applicability and use of Lipinski’s rule for central nervous system drugs**. *Lett. Drug Des. Discov.* (2016.0) **13** 1-8 49. Van de Waterbeemd H, Camenisch G, Folkers G, Chretien JR, Raevsky OA. **Estimation of blood−brain barrier crossing of drugs using molecular size and shape, and H-bonding descriptors**. *J. Drug. Target.* (1998.0) **6** 151-165. DOI: 10.3109/10611869808997889 50. Pajouhesh H, Lenz GR. **Medicinal chemical properties of successful central nervous system drugs**. *NeuroRx* (2005.0) **2** 541-553. DOI: 10.1602/neurorx.2.4.541 51. Zhang J, Zhang L. **Improvement of an isolation medium for actinomycetes**. *Mod. Appl. Sci.* (2011.0) **5** 124-127. DOI: 10.5539/mas.v5n2p124 52. Shirling EB, Gottlieb D. **Methods for characterization of**. *Int. J. Syst. Bacteriol.* (1966.0) **16** 313-340. DOI: 10.1099/00207713-16-3-313 53. Duangupama T. *Int. J. Syst. Bacteriol.* (2022.0) **72** 005320 54. Kelly KL. *Inter-Society Color Council: National Bureau of Standard Color Name Charts Illustrated with Centroid Colors* (1964.0) 55. Arai T. *Culture Media for Actinomycetes* (1975.0) 1-20 56. Williams ST, Cross T, Booth C. **Actinomycetes**. *Methods in Microbiology* (1971.0) 295-334 57. Gordon RE, Barnett DA, Handerhan JE, Pang CHN. *Int. J. Syst. Bacteriol.* (1974.0) **24** 54-63. DOI: 10.1099/00207713-24-1-54 58. Hasegawa T, Takizawa M, Tanida S. **A rapid analysis for chemical grouping of aerobic actinomycetes**. *J. Gen. Appl. Microbiol.* (1983.0) **29** 319-322. DOI: 10.2323/jgam.29.319 59. Komagata K, Suzuki KI. **Lipid and cell-wall analysis in bacterial systematics**. *Methods Microbiol.* (1987.0) **19** 161-207. DOI: 10.1016/S0580-9517(08)70410-0 60. Minnikin DE. **An integrated procedure for the extraction of bacterial isoprenoid quinones and polar lipids**. *J. Microbiol. Methods.* (1984.0) **2** 233-241. DOI: 10.1016/0167-7012(84)90018-6 61. Collins MD, Jones D. **Lipids in the classification and identification of coryneform bacteria containing peptidoglycans based on 2,4-diaminobutyric acid**. *J. Appl. Bacteriol.* (1980.0) **48** 459-470. DOI: 10.1111/j.1365-2672.1980.tb01036.x 62. Collins MD, Pirouz T, Goodfellow M, Minnikin DE. **Distribution of menaquinones in actinomycetes and corynebacteria**. *J. Gen. Microbiol.* (1977.0) **100** 221-230. DOI: 10.1099/00221287-100-2-221 63. Tamaoka J, Katayama-Fujimura Y, Kuraishi H. **Analysis of bacterial menaquinone mixtures by high performance liquid chromatography**. *J. Appl. Bacteriol.* (1983.0) **54** 31-36. DOI: 10.1111/j.1365-2672.1983.tb01297.x 64. Sasser M. *Identification of Bacteria by Gas Chromatography of Cellular Fatty Acids. MIDI Technical Note 101* (1990.0) 65. Kämpfer P, Kroppenstedt RM. **Numerical analysis of fatty acid patterns of coryneform bacteria and related taxa**. *Can. J. Microbiol.* (1996.0) **42** 989-1005. DOI: 10.1139/m96-128 66. Tamaoka J, Goodfellow M, O’Donnell AG. **Determination of DNA base composition**. *Chemical Methods in Prokaryotic Systematics* (1994.0) 463-470 67. Nakajima Y, Kitpreechavanich V, Suzuki K, Kudo T. *Int. J. Syst. Bacteriol.* (1999.0) **49** 1761-1767. DOI: 10.1099/00207713-49-4-1761 68. Yoon SH. **Introducing EzBioCloud: A taxonomically united database of 16S rRNA and whole genome assemblies**. *Int. J. Syst. Evol. Microbiol.* (2017.0) **67** 1613-1617. DOI: 10.1099/ijsem.0.001755 69. Kumar S, Stecher G, Li M, Knyaz C, Tamura K. **MEGA X: Molecular evolutionary genetics analysis across computing platforms**. *Mol. Biol. Evol.* (2018.0) **35** 1547-1549. DOI: 10.1093/molbev/msy096 70. Kimura M. **A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences**. *J. Mol. Evol.* (1980.0) **16** 111-120. DOI: 10.1007/BF01731581 71. Felsenstein J. **Confidence limits on phylogenies: An approach using the bootstrap**. *Evolution* (1985.0) **39** 783-791. DOI: 10.2307/2408678 72. Bankevich A. **SPAdes: A new genome assembly algorithm and its applications to single-cell sequencing**. *J. Comput. Biol.* (2012.0) **19** 455-477. DOI: 10.1089/cmb.2012.0021 73. Seemann T. **Prokka: Rapid prokaryotic genome annotation**. *Bioinformatics* (2014.0) **30** 2068-2069. DOI: 10.1093/bioinformatics/btu153 74. Richter M, Rosselló-Móra R, Glöckner FO, Peplies J. **JSpeciesWS: A web server for prokaryotic species circumscription based on pairwise genome comparison**. *Bioinformatics* (2016.0) **32** 929-931. DOI: 10.1093/bioinformatics/btv681 75. Rodriguez-R LM, Konstantinidis KT. **Bypassing cultivation to identify bacterial species**. *Microb. Mag.* (2014.0) **9** 111-118 76. Meier-Kolthoff JP, Auch AF, Klenk HP, Göker M. **Genome sequence-based species delimitation with confidence intervals and improved distance functions**. *BMC Bioinform.* (2013.0) **14** 60-73. DOI: 10.1186/1471-2105-14-60 77. Meier-Kolthoff JP, Göker M. **TYGS is an automated high throughput platform for state-of-the-art genome-based taxonomy**. *Nat. Commun.* (2019.0) **10** 2182. DOI: 10.1038/s41467-019-10210-3 78. **UniProt: The universal protein knowledgebase in 2023**. *Nucleic. Acids. Res.* (2023.0) **51** D523-D531. DOI: 10.1093/nar/gkac1052 79. Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW. **CheckM: Assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes**. *Genom. Res.* (2015.0) **25** 1043-1055. DOI: 10.1101/gr.186072.114 80. Alanjary M, Steinke K, Ziemert N. **AutoMLST: An automated web server for generating multi-locus species trees highlighting natural product potential**. *Nucleic. Acids. Res.* (2019.0) **47** 276-282. DOI: 10.1093/nar/gkz282 81. Brand-Williams W, Cuvelier ME, Berset C. **Useofafree radical method to evaluate antioxidant activity**. *Leben. Wissens. Technol.* (1995.0) **28** 25-30. DOI: 10.1016/S0023-6438(95)80008-5 82. Tadtong S, Kanlayavattanakul M, Laurith N. **Neuritogenic and neuroprotective activities of fruit residues**. *Nat. Prod. Commun.* (2013.0) **8** 1583-1586. PMID: 24427947 83. Tangsaengvit N, Kitphati W, Tadtong S, Bunyapraphatsara N, Nukoolkarn V. **Neurite outgrowth and neuroprotective effects of Quercetin from Caesalpinia mimosoides Lamk. on cultured P19-derived neurons**. *Evid. Based Complement. Altern. Med.* (2013.0) **2013** 838051. DOI: 10.1155/2013/838051 84. Changsen C, Franzblau SG, Palittapongarnpim P. **Improved green fluorescent protein reporter gene-based microplate screening for antituberculosis compounds by utilizing an acetamidase promoter**. *Antimicrob. Agents. Chemother.* (2003.0) **47** 3682-3687. DOI: 10.1128/AAC.47.12.3682-3687.2003 85. 85.BIOVIA, Dassault Systèmes, Discovery Studio, Client v21.1.0.20298. (Dassault Systèmes, 2021). 86. Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathan S, Karplus M. **CHARMM: A program for macromolecular energy minimization and dynamics calculations**. *J. Comp. Chem.* (1983.0) **4** 187-217. DOI: 10.1002/jcc.540040211 87. Momany FA, Rone RJ. **Validation of the general purpose QUANTA 32/CHARMm force field**. *Comp. Chem.* (1992.0) **13** 888-900. DOI: 10.1002/jcc.540130714 88. Trott O, Olson AJ. **AutoDock Vina, improving the speed and accuracy of docking with a new scoring function, efficient optimization and multithreading Oleg**. *J. Comput. Chem.* (2010.0) **31** 455-461. PMID: 19499576 89. Pedretti A, Mazzolari A, Gervasoni S, Fumagalli L, Vistoli G. **The VEGA suite of programs: An versatile platform for cheminformatics and drug design projects**. *Bioinformatics* (2021.0) **37** 1174-1175. DOI: 10.1093/bioinformatics/btaa774 90. Tirado-Rives J, Jorgensen WL. **Contribution of conformer focusing to the uncertainty in predicting free energies for protein-ligand binding**. *J. Med. Chem.* (2006.0) **49** 5880-5884. DOI: 10.1021/jm060763i
--- title: On the forecastability of food insecurity authors: - Pietro Foini - Michele Tizzoni - Giulia Martini - Daniela Paolotti - Elisa Omodei journal: Scientific Reports year: 2023 pmcid: PMC10038988 doi: 10.1038/s41598-023-29700-y license: CC BY 4.0 --- # On the forecastability of food insecurity ## Abstract Food insecurity, defined as the lack of physical or economic access to safe, nutritious and sufficient food, remains one of the main challenges included in the 2030 Agenda for Sustainable Development. Near real-time data on the food insecurity situation collected by international organizations such as the World Food Programme can be crucial to monitor and forecast time trends of insufficient food consumption levels in countries at risk. Here, using food consumption observations in combination with secondary data on conflict, extreme weather events and economic shocks, we build a forecasting model based on gradient boosted regression trees to create predictions on the evolution of insufficient food consumption trends up to 30 days in to the future in 6 countries (Burkina Faso, Cameroon, Mali, Nigeria, Syria and Yemen). Results show that the number of available historical observations is a key element for the forecasting model performance. Among the 6 countries studied in this work, for those with the longest food insecurity time series, that is Syria and Yemen, the proposed forecasting model allows to forecast the prevalence of people with insufficient food consumption up to 30 days into the future with higher accuracy than a naive approach based on the last measured prevalence only. The framework developed in this work could provide decision makers with a tool to assess how the food insecurity situation will evolve in the near future in countries at risk. Results clearly point to the added value of continuous near real-time data collection at sub-national level. ## Introduction The 2030 Agenda for Sustainable Development, adopted by all United Nations Member States in 2015, calls for urgent action to “end hunger, achieve food security and improved nutrition and promote sustainable agriculture”1. However, in 2019, 650 million people were still undernourished2, with 135 million in 55 countries and territories reported to be acutely food insecure3. These numbers have significantly increased as a consequence of the COVID-19 pandemic, with at least 280 million people reported to be acutely food insecure in 20204. Food insecurity is a complex phenomenon, resulting from the interplay of many environmental, socio-demographic, and political events5. It is usually characterized through its four key pillars: availability, access, utilization, and stability; however, agency and sustainability have also been recently proposed as fundamental dimensions of food insecurity6. Climate variability and extremes weather events are considered among the main drivers, mostly because of their effects on crop production7,8. In recent years, studies have additionally shown that, beyond agricultural yields, extreme temperatures and precipitation conditions also directly negatively affect child nutrition and food security9–12. Conflicts are also deeply intertwined with food insecurity, of which they can be the cause or the consequence13. Several studies have shown that conflict impacts agricultural production, nutritional status, coping and consumption14,15. Food insecurity is also a global public health challenge. Household food insecurity is the leading risk factor of malnutrition, claiming approximately 300, 000 deaths each year. Whether directly or indirectly, due to inadequate food consumption and poor diet quality, it is also accountable for over half of all deaths among children in Sub-Saharan Africa. Among the various determinants of food insecurity in relation to child malnutrition, the main ones are socio-demographic characteristics as well as food prices16. The impact of food insecurity on health cannot be overstated17–19. The current COVID-19 crisis has exacerbated the situation and undermined some of the critical elements influencing food insecurity, e.g. international agricultural supply chains and goods prices impacted by trade restrictions20. In low and middle income countries, social distancing, workplace closures, and restrictions on mobility and trade had cascading effects on economic activity, food prices, and employment21. From a policy perspective, food insecurity will remain a worldwide concern for the next 50 years and beyond. As a 20-year old article in Science stated: “Recently, crop yield has fallen in many areas because of declining investments in research and infrastructure, as well as increasing water scarcity. Climate change and HIV/AIDS are also crucial factors affecting food security in many regions. Although agroecological approaches offer some promise for improving yields, food security in developing countries could be substantially improved by increased investment and policy reforms”22. In the past twenty years, the situation has however significantly worsened and agroecological technologies are not keeping up with the accelerating effects of climate change that, ever increasingly, are affecting countries and populations on much shorter scales (weeks and months rather than years) than in the past. Therefore, an essential step towards achieving hunger reduction is to have access to frequent, up-to-date information on the status of food insecurity in countries facing humanitarian crises, and to estimates of where and when the situation is likely to improve or deteriorate, in order to allow for informed and timely decision-making on resource allocation and on relevant policies and programmes. For this reason, food security assessments are performed on a regular basis. This is done through face-to-face surveys as well as computer-assisted telephone interviews (CATI), which have become very popular in the last few years, and are sometimes complemented by other technologies like interactive voice response and web surveys. The World Food Programme (WFP) is currently monitoring the food security situation in near-real time in a number of countries at the sub-national level, collecting on a daily basis, through remote phone surveys, information on levels of food consumption and food-based coping, as well as other relevant indicators23,24. This unprecedented availability of daily sub-national level data paves the way for new possibilities since it not only allows for a continuous up-to-date picture of the current situation, but could also be used to build predictive models to forecast how the situation will evolve in the future. In this study, we explore the forecastability of insufficient food consumption levels, and show, specifically for Syria and Yemen, that satisfactory predictions up to 30 days into the future can be obtained when enough daily sub-national level historical data is available. Having access to regular real-time estimation of how the situation is likely to evolve in the near future would allow WFP for more informed discussions on need-based humanitarian assistance allocation decisions. Forecasting modeling has been the subject of extensive investigation during the last decade in different fields, from financial markets25,26 to infectious disease epidemiology27–30. However, it is still a relatively new area of research in the context of food security. The Food and Agriculture Organization of the United Nations (FAO) developed a methodology to produce annual country-level estimates of the prevalence of undernourishment, and to project these estimates up to 10 years into the future31,32. Okori and collaborators first proposed to use machine learning models to predict whether a household is in famine or not from household socioeconomic and agricultural production characteristics33,34. The effort of predicting levels of insufficient food consumption has been tackled in the context of Malawi, in a study where the authors built a model trained on 2011 data to estimate the situation in 201335, and more recently in a work proposing a model to nowcast sub-national levels of insufficient food consumption on a global scale36. Both studies propose methods to predict the current situation when primary data is not available, but they do not address the challenge of making projections for the future. The World Bank recently proposed a machine learning approach to forecast transitions into critical states of food insecurity37 and a stochastic model to forecast famine risk38. Concurrently, similar work was carried out by other researchers in the context of Ethiopia39. These studies focus on forecasting month-to-month transitions to different phases of food insecurity, based on the Integrated Food Security Phase Classification (IPC) framework40. In this study, we tackle a different problem: forecasting the daily evolution of the prevalence of people with insufficient food consumption at the sub-national level. This metric, characterizing a given area at a given time, is defined as the prevalence of households, in the specified area and time period, that are identified to have poor or borderline food consumption. Such prevalence is computed from a representative number of household surveys enquiring about one of the core household food insecurity indicators, namely the Food Consumption Score (FCS), which captures households’ dietary diversity and nutrient intake41. The FCS is one of the many indicators developed to monitor food insecurity, each capturing different dimensions of the problem. Other examples include the Household Dietary Diversity Score (HDDS), also focusing on dietary intake, the Coping Strategies Index (CSI) and the Household Hunger Scale (HHS), both focusing on the consequences of constrained access to food42, and the more recent Food Insecurity Experience Scale (FIES), focusing on behaviors and experiences associated with difficulties in accessing food due to resource constraints43. In this study, which constitutes only a first attempt at food insecurity daily forecasting, we choose to focus on the FCS, given its wide adoption by WFP and the consequent data availability. Having access to reliable predictions of the evolution of insufficient food consumption levels over future weeks and months could allow governments and organizations to identify which areas should be monitored more closely and to eventually take timely decisions on resource allocation. Hence, the goal of this study is to develop a forecasting model able to predict, in countries with major food crises, the daily sub-national prevalence of people with insufficient food consumption up to 30 days into the future. The main difference with IPC and the Famine Early Warning Systems Network (FEWS NET)44 is that, while these make use of consensus-based expert opinion, our approach is algorithmic and data-driven, and can hence be applied in an automatic fashion. IPC’s and FEWS NET’s projections are essential for humanitarian action, however they require local expertise and considerable time to be developed. Our main goal is not to replace these efforts, but rather to complement them with an approach that, once improved as more data becomes available, could be used to provide rapidly available forecasts for several places at the time by automatically feeding near real-time data to the proposed algorithms. ## Time trends of insufficient food consumption We study the possibility of forecasting one of the core dimensions of food insecurity by means of a unique data set of daily sub-national time series of the prevalence of people with insufficient food consumption, in six countries in West Africa and the Middle East: Burkina Faso, Cameroon, Mali, Nigeria, Syria and Yemen (see the Methods for a detailed definition of the indicator under investigation). These countries, although having varying socioeconomic and geopolitical characteristics, have all been identified as major food crises where acute food insecurity is driven by conflict, weather extremes and economic shocks3. Among all major food crises, these are the countries for which the largest volume of food insecurity data is available. The length of these time series varies from a minimum of 865 days in Mali 1340 days in Yemen, over the years 2018–2022. Also, geographic coverage varies across countries. In the case of Burkina Faso, the prevalence of insufficient food consumption is available for all administrative units of the country, while only 3 states of Nigeria are included in our dataset (Adamawa, Borno, Yobe), given these are the most at risk areas closely monitored by WFP. Overall, our dataset covers $88\%$ or more of the total population in all countries, with the only exception of Nigeria (see Supplementary Table S4). Since the mode of questionnaire administration can have serious effects on data quality45,46, during the last few years WFP conducted mode experiments in several countries, each demonstrating the feasibility of collecting food security indicators via CATI surveys47,48. However, sampling and selection bias should be assessed and mitigated49–51, hence post-stratification weights were applied by WFP, as detailed in the Methods section. As shown in Fig. 1, in the six countries, time trends of insufficient food consumption display noisy and irregular patterns, underscoring the complex dynamics underlying food insecurity. During the study period, all countries experienced large fluctuations in the prevalence of insufficient food consumption, and such variations were not uniform between sub-national administrative units. In Cameroon, for instance, only a few regions were characterized by a relatively high proportion of food insecure people, generally above \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$50\%$$\end{document}$50\%$, but also exhibiting large fluctuations, such as the rapid decline and subsequent increase observed in the North-West regions. On the other hand, in Syria, the sub-national trends were all similar in terms of relative changes in the affected population, with a general upward trend affecting almost every province beginning in July 2020. In the governorates of Yemen, for which the longest time series are available, the proportion of the population affected by food insecurity varied between $20\%$ and $60\%$ during the years 2018–2022, however, a common national time trend is less recognizable. It should be noted that some of these irregularities could also be partially due to the effects of sampling and selection bias, whose mitigation can rarely be achieved in full. ## Permutation entropy and intrinsic predictability of food insecurity We first quantify the intrinsic predictability of the time series shown in Fig. 1 by means of a permutation entropy analysis. Permutation entropy (PE) is a model-free measure of time series complexity52,53, that is conceptually similar to the Shannon entropy but is based on the frequency distribution of motifs. PE has been extensively used to assess the predictability of time series in different domains including finance and economics54,55, ecology56 and infectious disease epidemiology30. In short, to compute the PE of a time series we translate its real valued sequence \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(x_1, x_2, \dots, x_N)$$\end{document}(x1,x2,⋯,xN) into a frequency distribution of symbols that represent patterns of relations \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_i < x_j$$\end{document}xi<xj, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_i = x_j$$\end{document}xi=xj or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_i > x_j$$\end{document}xi>xj between nearest or distant neighbors, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_i$$\end{document}xi and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x_j$$\end{document}xj. Such frequency distribution is then used to assess the predictability of the time series by computing the Shannon entropy associated with the distribution of permutation patterns in the symbols defined above. In the Methods section we provide a complete formal definition of the PE and its computation. It has been shown that PE can be considered as a measure of intrinsic predictability of a time series and its value is positively associated with forecasting error56. Intuitively, PE quantifies the information that is transmitted from the past to the present state of a time series: a time series that periodically visits the same few symbols among the many possible will have a low entropy and its present state will be easily determined from the past. A random time series that uniformly samples the symbols with equal probability will have a high entropy and its future will not be predictable from past states. In the case of food insecurity, we find that insufficient food consumption trends are not easily predictable based on their past history. As shown in Fig. 2, their predictability, measured as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi = 1-H$$\end{document}χ=1-H, where H is the PE, never reaches values above 0.5 and it is often reduced to 0.1–0.2 within a 10-day horizon. These values are generally much lower than those observed in the case of infectious disease dynamics30 and they are closer to measures of predictability of financial time series57, which are characterized by a high short- and long-term volatility. Confidence intervals around mean predictability values are also narrow, highlighting a consistent lack of recurrent patterns in the insufficient food consumption time series across different time scales, which in turn highlights the presence of intrinsic entropy barriers to their predictability. ## Forecasting food insecurity with secondary information Following from the observation that insufficient food consumption trends are not highly predictable from their own history, we explore whether secondary information can be used to enhance our ability to predict their future evolution. To this end, we revert to information on the key drivers of food insecurity: conflict/physical insecurity, extreme weather events and economic shocks58. We build a set of indicators covering these three domains and develop a forecasting model based on gradient boosted regression trees (XGBoost)59 to make predictions on how the insufficient food consumption trend will evolve up to 30 days into the future. More specifically, in our model we consider as predictors of insufficient food consumption the following indicators (see Methods and the Supplementary Information file for a full description). First, we include daily time series of the prevalence of people using crisis or above crisis food-based coping, which is obtained from another core food insecurity indicator, the reduced Coping Strategy Index (rCSI), by measuring the share of households with rCSI \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ge 19$$\end{document}≥1940,60. Since political unrest can affect food security, we include in our model daily time series of fatalities due to conflict or political violence as reported by the Armed Conflict Location and Event Data Project (ACLED)61. Economic shocks are included into the model by considering monthly variations in the price of cereals and tubers in local currency. The model takes into account the effects of weather events and climate conditions by including time series of rainfall, of its anomaly with respect to long-term averages (over 1 and 3 months), and time series of the Normalized Difference Vegetation Index (NDVI), a standard satellite-based measure of vegetation coverage that is commonly used for drought assessment62, and of NDVI anomaly. Finally, since the food consumption behavior of most of the population in several African and Asian countries is affected by Ramadan, we include a time series that marks the days of the Ramadan period that fall within the time window used to measure people’s food consumption. Figure 3 and Supplementary Fig. S1 show the prediction results of the model for the case of Yemen and Syria, respectively, the countries for which the longest time series of insufficient food consumption prevalence are available. In Yemen (Syria), cross-validated predictions can explain between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$99\%$$\end{document}$99\%$ (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$99\%$$\end{document}$99\%$) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$72\%$$\end{document}$72\%$ (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$47\%$$\end{document}$47\%$) of the variation in insufficient food consumption, with the former being the variation explained by the 1-day into the future forecast, and the latter for the 30-day into the future one (Fig. 3a and Supp. Fig. S1a). This is a significant increase of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2$$\end{document}R2 with respect to a naive prediction based on the last measured value only, which can only explain between \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$99\%$$\end{document}$99\%$ (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$99\%$$\end{document}$99\%$) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$65\%$$\end{document}$65\%$ (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$31\%$$\end{document}$31\%$) of the variation (Fig. 3c and Supp. Fig. S1c), and whose mean squared error (MSE) is larger and with a wider dispersion than the MSE of the proposed model (Fig. 3b, d and Supp. Fig. S1b, d). Specifically, panels c and d show that the proposed model clearly outperforms the naive approach for high values of the prediction horizon, consistently with the expectation that as time progresses the last measured value is not a good guess anymore. The scatterplots in Fig. 3e and Supp. Fig. S1e show the performance of the forecasting models as the predicted insufficient food consumption value against the actual value, for different prediction horizons. As expected, dots get further away from the identity diagonal as the prediction horizon increases up to 30 days, although the general behavior is consistent with a good predictive accuracy. Over short forecasting horizons, typically less than 14 days, a naive approach proves to be a good enough predictor as we do not expect food consumption to suddenly change from one day to the next. However, as we try to forecast further into the future, we see that the forecasting model starts to outperform the naive approach, as shown in Fig. 3f (Supp. Fig. S1f) for the case of two Yemeni (Syrian) provinces over a 30 day horizon. In the case of the remaining four countries, whose available training points are less than half than for Yemen, the model performance is worse than the naive approach across all prediction horizons, both in terms of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2$$\end{document}R2 and MSE (see Supplementary Figs. S2–S5). ## Model performance as a function of data availability Given the relatively poor predictive performance of our models in countries with short time series of insufficient food consumption, we systematically examine how the performance varies as a function of the length of the time series available to train the model and of spatial coverage, indicating the number of sub-national areas. We find that, compared to the naive approach, the performance of our model dramatically increases with the number of available training points, which is given by the product of the two dimensions above: temporal length and spatial coverage (see Fig. 4). Moreover, with a given size of the training set, the proposed model tends to perform better than the naive approach as the forecasting horizon grows, demonstrating that, as expected, the model is better at predicting further into the future than just considering the last available measurement. However, this effect is evident only when a large training set is available (as in the case of Yemen, with more than 20,000 data points), and a small training set reduces the benefit of the model even over longer time horizons. A limitation of this analysis relies on the fact that in our dataset different numbers of training points coincide with different countries, which does not allow to disentangle effects due to the local context from those due to temporal length and spatial coverage. We therefore performed further analyses in order to separately investigate the role played by the number of covered areas and by the time series length. First, we considered the same time series length for all countries by using as starting date for all of them the earliest date available for all countries. In this setting, the difference in the number of training points among the different countries is only due to the different number of areas covered. Results are reported in Supplementary Fig. S6a. Secondly, we considered instead the full time series but we fixed the number of considered areas to the minimum number of areas available for all countries (in this case we had to exclude Nigeria since only three areas are covered there, which would have been too few for the analysis). In this setting, the difference in the number of training points among the different countries is only due to different lengths of the available time series. Results are reported in Supplementary Fig. S6b. In both cases, we re-trained and tested the model using, each time, a different data subset, as described above. Results show that the performance of the model still increases with the dimension under study (the length of the time series and the number of covered areas), confirming our initial results. ## Discussion In this study we addressed the critical challenge of forecasting the daily evolution of a food security indicator, namely the prevalence of people with insufficient food consumption, as measured by WFP. The problem promptly proved difficult given that the analyzed time series exhibit noisy and irregular behavior. This is to be expected since food insecurity in Sub-Saharan Africa and the Middle *East is* a highly dynamic phenomenon, comprising a seasonal component related to agricultural production calendars and religious observances such as Ramadan (during which consumption patterns are completely altered), but also subject to swift changes when external shocks hit, such as the emergence of conflict, extreme weather events or economic shocks63–65. Therefore, forecasting based solely on information on the historical evolution of the target indicator over time would not be successful, as demonstrated through a permutation entropy analysis. Hence, we extended the proposed framework to include historical information on the key drivers of food insecurity and built models that comprise both endogenous (insufficient food consumption itself, as well as food-based coping information) and exogenous factors (conflict-related fatalities, rainfall and vegetation and their anomalies, staple food prices and Ramadan’s occurrence). We showed that the proposed model makes it possible to forecast the prevalence of people with insufficient food consumption up to 30 days into the future with higher accuracy than a naive approach solely based on the last measured prevalence, at least in places where enough training data are available to inform the model. The number of available historical observations proved to be a key element in forecasting success. Even for places with more than 10,000 available training points, which is not an extremely large number but still enough to provide reasonable results in other contexts66, the phenomenon seems to be too complex for the algorithms to learn meaningful patterns. Besides the external shocks related to local socio-economic conditions, it is important to note that all countries under study experienced the global effects of the COVID-19 pandemic since early 2020. The pandemic has significantly impacted food security on a global scale67, disrupting supply chains, limiting access to food due to the adoption of non-pharmaceutical interventions, and increasing the need for food assistance68. In our analysis, we did not include epidemiological variables to model the impact of the COVID-19 pandemic—such as reported cases or deaths—because those epidemiological indicators are likely to be unreliable in the countries under study, due to the lack of adequate surveillance systems69. On the other hand, we expect the effects of the pandemic to be mainly captured by market price trends70, which indeed markedly increased in all countries and in all regions as shown in Fig. S9 of the Supplementary Information. As SARS-CoV-2 continues spreading worldwide, the world economy still suffers from the consequences of the pandemic and its long term effects are hard to predict. Further research will be needed to investigate the complex interplay between the COVID-19 pandemic and food security. Forecasting research within the humanitarian context has only recently started to attract attention from scholars71. In this context, our study represents an initial step towards the application of forecasting approaches to food insecurity at a high spatial and temporal granularity. Our results confirm that nowcasting or one-step-ahead forecasting are feasible, as reported in recent studies36,38, but long-term forecasts are challenging and strongly conditioned by data availability. The methods presented in this study come with limitations, and they could be further improved through several approaches. First, results could be compared with those obtained by other methods such as the autoregressive integrated moving average with exogenous inputs models (ARIMAX). Secondly, more complex forecasting methods could potentially lead to a greater forecasting accuracy, for instance through the use of deep learning techniques72. Additionally, hybrid methods, combining both statistical and ML features, could achieve a better forecasting performance66. Finally, forecasting models could benefit from the inclusion of additional external predictors and in particular from the availability of novel data streams, such as mobile phone data73 or the automated text mining of news74. Another important limitation of this study is the focus on one food insecurity indicator only, namely the prevalence of people with insufficient food consumption obtained from the FCS. As discussed above, a variety of indicators exist, each capturing different dimensions of food insecurity. Future work should therefore aim at developing forecasting models for other indicators too, when enough data is available. Additionally, the spatial resolution of this study was bounded due to data availability for first-level administrative units only. However, when second or third level administrative unit data becomes available, forecasting models at higher spatial resolutions should be developed. Finally, future studies should aim at going beyond the limitation of a 30-day forecasting horizon and propose methods that can forecast up to 2–3 months into the future. In conclusion, our study presents a simple, yet fundamental message for governments and humanitarian organizations on the power of the data they collect: collecting data on a regular basis for long enough periods of time and across enough different geographic areas does not only make it possible to monitor the evolution of a situation in near real-time but also to inform forecasting models that would make it possible to produce estimates of how the situation is likely to evolve in the near future. This means that decision makers would have access in advance to information on areas most at risk of a deterioration in the food security situation, allowing for a more timely response. Predictions should of course be used with caution and considered only as an indication of what may happen in the near future, hence informing preparedness efforts by suggesting a need for further in-depth assessments of the food security situation. ## Target indicator The indicator whose time-evolution we aim to predict is the daily prevalence of people with insufficient food consumption in a given sub-national geographical area. This prevalence is obtained as the weighted share of households in the area that are found to have poor or borderline food consumption according to the Food Consumption Score (FCS)41,75. The FCS is obtained through household surveys by asking how often, during the previous 7 days, a household has consumed food items from different food groups (main staples, pulses, vegetables, fruit, meat and fish, milk, sugar, oil and condiments). Consumption frequencies are then summed up in a weighted fashion, where each food group is weighted according to its nutritional level (with more nutritious foods having higher weights), resulting in the FCS. Thresholds are then applied to label each household as having poor, borderline or acceptable food consumption (as further detailed in Section 3.1 of the Supplementary Information), allowing to eventually compute the prevalence of people in a given area with insufficient (i.e. poor or borderline) food consumption. The time series analyzed in this study were obtained by WFP through daily computer-assisted telephone interviewing (CATI) surveys. Informed consent was obtained from all interviewed subjects, all data collection protocols were approved by the World Food Programme’s Hunger Monitoring Unit, and methods were carried out in accordance with its guidelines and regulations48. Sample sizes were determined by WFP by taking into account modality and adhering to IPC guidelines for a good level of reliability (i.e. as close to 150 households per strata as possible)40. They initially utilize Random-Digit Dialing (RDD) to obtain the most random selection of respondents as possible, while applying some filters to ensure the required geographic and sociodemographic distributions (i.e. not all households reached are actually interviewed, only those matching the specific characteristics needed). This enables WFP, over time, to build a representative sample, and then transition to panel surveys after the initial months of implementation48. In order to compute a statistically representative prevalence of people with insufficient food consumption at sub-national and daily resolution, a rolling window approach is used. That is, for each geographical area, the prevalence of people with insufficient food consumption for a given day is obtained as the weighted share of households with poor or borderline food consumption interviewed during the previous d days, where d varies by country (values are reported in Supplementary Table S2). Missing values in the time series are inferred through linear interpolation. Post-stratification weights are applied by WFP to compute the final share of households with poor or borderline food consumption in a given area and time window, in order to mitigate sampling and modality bias, as detailed in48. This is done through weighting of the data to account for the under-representation of certain demographics. Population weights are applied to compensate for administrative areas that are under- or over-sampled, while demographic weights are introduced to mitigate selection bias and compensate for under-represented households (e.g. low-income or less-educated households). The final weight is given by the product of the two, when both need to be applied. The present study only uses the final aggregated data, that is the estimated percentage of people in a given area with insufficient food consumption. The available data covers six countries over the years 2018–2022. The temporal resolution is daily and the spatial resolution is that of first-level administrative units. The length of the time series and the number of geographical areas covered varies by country and is as follows: Burkina Faso (907 days, 13 areas), Cameroon (977 days, 10 areas), Mali (865 days, 7 areas), Nigeria (1140 days, 3 areas), Syria (1280 days, 12 areas) and Yemen (1340 days, 20 areas). ## Permutation entropy We employ the Permutation Entropy (PE) as a model-free measure of time series predictability30. The main assumption of this approach is to measure the Shannon entropy through the probabilities of encountering trend patterns within the time series. For this reason, the PE first categorizes the continuous time series X in a small set of symbols or alphabet according to their trends. Let x(i), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$$i = 1$,..., N$$\end{document}$i = 1$,...,N, denote sequences of observations from a system X. For a given, but otherwise arbitrary i, m amplitude values \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X_i = \{x(i), x(i+\tau), \dots, x(i+(m-1)\tau)\}$$\end{document}Xi={x(i),x(i+τ),⋯,x(i+(m-1)τ)} are arranged in an ascending order where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau$$\end{document}τ denotes the time delay, and m is the embedding dimension. Each \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X_i$$\end{document}*Xi is* then mapped onto one of the m! possible permutations. The PE of the time series X is given by the Shannon entropy on the permutation orders:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} H = -\sum _\pi p_{\pi } log(p_{\pi }) \end{aligned}$$\end{document}H=-∑πpπlog(pπ)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{\pi }$$\end{document}pπ is the probability of encountering the pattern associated with permutation \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\pi$$\end{document}π. An important convenience of symbolic approaches is that they discount the relative magnitude of the time series76. This is important in our case because different geographical units can differ largely in food insecurity prevalence. The embedding dimension m and the time delay \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau$$\end{document}τ are to be set in order to derive a reliable state space. There exist different procedural approaches in order to deal with this setting decision77,78. In order to find the appropriate embedding dimension for clustering a set of time series, we follow the instructions proposed by Scarpino and Petri30. The time delay is fixed to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tau = 1$$\end{document}τ=1 in order to get results from continuous intervals. Finally, the metric used is the predictability defined as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi = 1 - H$$\end{document}χ=1-H. The closer to 1 the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ is, the more regular and more deterministic the time series is. Contrarily, the smaller \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi$$\end{document}χ is, the more noisy the time series is. As suggested by Scarpino and Petri30, we analyzed the predictability as a function of the length of each time series. Focusing on the predictability over short timescales, we average H over temporal windows by selecting 1000 random points from each time series and calculating H for windows of length 10, 11, 12,..., 100 days. ## Independent variables The following variables were defined to be considered as input features for the forecasting models. The same spatial and temporal coverage of the target indicator is used. ## Prevalence of people using crisis or above crisis food-based coping This prevalence is obtained as the weighted share of households in a given sub-national geographical area that are found to have a reduced Coping Strategy Index (rCSI) greater than or equal to 1960,75. The rCSI is obtained through household surveys by asking if and how often, during the previous 7 days, a household had to adopt the following coping behaviors: relying on less preferred or less expensive food, borrowing food from relatives or friends, limiting portion sizes, restricting adults’ consumption in order for small children to eat and reducing the number of meals eaten in a day. The rCSI is then obtained as a weighted sum of these frequencies, where weights are based on the severity of the strategy, as further detailed in Section 3.2 of the Supplementary Information. The survey data used to build this variable is the same as for the target indicator. A rolling window approach to compute a statistically representative prevalence of people using crisis or above crisis food-based coping at sub-national and daily resolution is also applied, and missing values are interpolated through linear regression. The same post-stratification weighting schemes to mitigate sampling and modality bias are also applied. The present study only uses the final aggregated data, that is the estimated percentage of people in a given area using crisis or above crisis food-based coping. ## Conflict-related fatalities The number of conflict-related fatalities in a given geographical area is obtained from the Armed Conflict Location and Event Data Project (ACLED), a publicly available near-global repository of reported conflict events and related fatalities61. Since each daily value of the target indicator is based on data collected during the previous d days, the number of fatalities associated with the same date and area is also obtained by summing all fatalities reported in the same area during the same d days. Further details are reported in Section 3.3 of the Supplementary Information. ## Market prices Monthly prices of cereals and tubers are obtained from WFP’s publicly available Economic Explorer (https://dataviz.vam.wfp.org/economic_explorer/prices). Cereal and tubers prices for each geographical area and date are obtained by averaging normalized prices (in local currency) across all markets within the area. Further details are reported in Section 3.4 of the Supplementary Information. ## Weather variables In order to measure the performance of the agricultural season, and more specifically whether the rainfall season is drier or wetter than average, and its impact on the vegetation status, for each geographical area and date, we consider the following weather variables, which are defined and computed by WFP as 10-day measurements, for each first-level administrative unit, and made publicly available through its Seasonal Explorer (https://dataviz.vam.wfp.org/seasonal_explorer/rainfall_vegetation/help): the amount of rainfall in mm, its 1-month and 3-month anomalies with respect to the historical average during the same period of the year (expressed in percentage), the normalized difference vegetation index (NDVI), and its anomaly (defined as for rainfall but considering 10-days only since effects of previous rainfall are already integrated by vegetation itself). Further details are reported in Section 3.5 of the Supplementary Information. ## Religious observances Ramadan is a religious observance celebrated by that the majority of the population in the analyzed countries during which food consumption increases. For each date and geographical area we therefore create a variable that takes into account the number of days, within the d days considered to obtain the prevalence of people with insufficient food consumption for the same date and area, that fall within the Ramadan observance period. This variable therefore spans between 1 and n during and after Ramadan, and is otherwise equal to zero during the rest of the year. ## Population The latest population estimate provided by WFP for each geographical area is used as a static variable. ## Geographical area identifiers The total area of each geographical unit, its latitude and longitude, and it waterways size, are also used as static variables. ## Temporal identifiers Temporal information (day, month and year) on the forecasting horizon is also included. ## Preliminary feature selection The proposed independent variables were defined based on expert opinion, and only a minimal preliminary selection was performed to avoid the presence of highly correlated variables within the same category (e.g., fatalities, market prices, weather, etc.). Since the only category that contains more than one variable is weather, this preliminary selection was performed only on the corresponding five variables: rainfall, 1-month rainfall anomaly, 3-month rainfall anomaly, NDVI and NDVI anomaly. We computed the Pearson’s correlation coefficient r between each pair of the above five variables and for each pair where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|r|>0.45$$\end{document}|r|>0.45, one of the two variables was discarded, to avoid collinearity. Given the high information overlap among weather variables, the value 0.45 was chosen as an arbitrary conservative threshold to remove even moderate correlations. As a result of this process, the rainfall 3-month anomaly was removed from all country-specific datasets but Nigeria’s, the NDVI anomaly from Yemen’s and Syria’s, and the NDVI from the remaining four countries (see Supplementary Table S6), leaving rainfall, the 1-month rainfall anomaly and NDVI or its anomaly as the only remaining weather-related features for most countries, given the low level of correlation between them. Finally, the Variance Inflation Factor (VIF) was computed for all remaining variables, including those in the other categories, to test the presence of multicollinearity. The resulting VIF values, reported in Supplementary Figures S15–S20, are all below 3, indicating no significant multicollinearity and hence allowing us to proceed with the obtained set of independent variables. Figure 1Time trends of insufficient food consumption. Each panel displays daily time series of the percentage of people with insufficient food consumption in the first-level administrative units of Burkina Faso, Cameroon, Mali, Nigeria, Syria and Yemen. The six countries are highlighted in the map, and the orange shade indicates the areas that are considered by our analysis. The map was created by the authors using GeoPandas v0.9.079 and the shapefiles provided by OCHA under the CC BY-IGO license on HDX (https://data.humdata.org/).Figure 2Food insecurity is characterized by low predictability. The average predictability \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\chi = 1-H$$\end{document}χ=1-H for daily trends of insufficient food consumption, in the six analyzed countries, is shown as a function of time series length in days. We average H over temporal windows by selecting 1000 random points from each time series and calculating H for windows of length 10, 11, 12,..., 100 days. The solid lines indicate the mean value and the shaded areas mark the interquartile range across all administrative units and starting locations in the time series. Figure 3Forecasting the prevalence of people with insufficient food consumption in Yemen. The forecasting is performed over 5 different monthly splits of all governorates time series, from October 2021 to February 2022. ( a) Box plots of the coefficient of determinations (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2$$\end{document}R2) across the 5 splits for both the proposed and the naive models (in blue and orange, respectively), for each forecasting horizon. ( b) Box plots of the mean squared error (MSE) across the 5 splits for both the proposed and the naive models for each forecasting horizon. ( c) Box plots of the difference between the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2$$\end{document}R2 of the proposed and of the naive model for each split. ( d) Box plots of the difference between the MSE of the naive and of the proposed model for each split. ( e) Predicted vs actual value for each data point in the 5 splits. Colors represent the corresponding forecasting horizon and vary from dark blue (1 day) to yellow (30 days). ( f) Example of forecasting results for December 2021 in Amanat Al Asimah (top) and February 2022 in Abyan (bottom).Figure 4Model performance as a function of the number of available training points. For the six analyzed countries and four different forecasting horizons (1–4 weeks), the figure shows the averaged differences between the MSE of the naive approach and the MSE of the forecasting model across the different splits, as a function of the size of the training set. Error bars correspond to the relative standard deviation. The green area indicates where our model outperforms the naive one, the red area indicates the opposite. ## Forecasting The core of this work revolves around the forecasting effort focusing on predicting the evolution up to 30 days into the future of our insufficient food consumption time series (No investigation into the forecast of the the independent indicators was performed because of the involvement of chain-of-events predictions (e.g. weather or market forecasts) and ethical issues around providing conflict predictions). To this aim, we chose to use the eXtreme Gradient Boosting (XGBoost) algorithm59, a widely used state-of-the-art machine learning technique known for its high performance and flexibility. XGBoost belongs to the category of so-called ensemble learning approaches, which is a branch of machine learning methods that makes use of several models at once to produce a single better output. In the case of XGBoost, the base model is a decision tree, which is considered best-in-class for handling small to medium-sized data. A decision tree is a set of hierarchical choices which eventually lead to a final result, i.e the prediction. Ensemble methods combine several decision trees to produce better predictive performance than utilizing a single decision tree. To create this collection of trees, XGBoost fits consecutive trees by, at every step, trying to solve for errors from the previous tree, using a gradient descent algorithm. The wider context of machine learning approaches used in the time series forecasting field and a more in-depth description of XGBoost can be found in Section 6 of the Supplementary Information. The motivation behind the choice of this algorithm for this study is twofold. First, XGBoost can handle complex and non-linear relationships among the variables, which we expect to have in a complex phenomenon like food insecurity. Secondly, XGBoost has a high degree of flexibility, which makes it the most suitable candidate for a prediction task that is meant to eventually run as an operational tool in near real-time. Specifically, XGBoost can handle missing values in the input variables, which is a feature that makes it possible to automatically run the algorithm on a daily basis, even when a few values might be missing because, for example, of delays in data availability. Since XGBoost does not support a multi-output design, we developed 30 different models, one for each prediction horizon. For each date, the prediction framework is trained to predict levels of insufficient food consumption for a given day into the future based on the information available up to the date under consideration. For further details, see Section 6 of the Supplementary Information. In order to implement our forecasting model based on the usual three stages of training, validation and testing, we adopt a k-fold cross-validation approach in a time-ordered fashion (i.e. the evaluation stage is applied to different historical periods). The validation phase is implemented by splitting each of the n splits of training points of each sub-region into two time order preserving sets: the first $80\%$ samples are used for training and the remaining ones for validation. Validation is performed independently across splits ensuring an unbiased approach. Our validation scheme aims to optimize the prediction framework by acting on two main configurations: model hyper-parameters and feature selection. The aim of this optimization is to find the configuration that returns the best performance as measured on a validation set. See Supplementary Table S8 for a detailed list of the explored hyper-parameters and values, and Supplementary Table S7 for the detailed list of independent variables and time lags considered. For further details, see Section 6 of the Supplementary Information. Finally, in order to assess the goodness of the proposed forecasting model, its performance on the test sets is compared with a naive approach, where the predicted value at any given forecasting horizon is simply given by the last available value in the training and validation set, which represents the last available measured value before the start of the forecasting horizon. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-29700-y. ## References 1. 1.United Nations General Assembly. Transforming Our World: The 2030 Agenda for Sustainable Development. https://sdgs.un.org/2030agenda (2015). 2. 2.FAO IFAD UNICEF WFP and WHO. The State of Food Security and Nutrition in the World 2021. Transforming Food Systems for Food Security, Improved Nutrition and Affordable Healthy Diets for All. https://www.fao.org/publications/sofi/2021/en/ (2021). 3. 3.Food Security Information Network. Global Report on Food Crises. https://www.fsinplatform.org/global-report-food-crises-2020 (2020). 4. 4.WFP Emergency Operations Division. WFP Global Operational Response Plan: Update #3. https://www.wfp.org/publications/wfp-global-operational-response-plan-update-3-november-2021 (2021). 5. 5.Committee on World Food Security. Global Strategic Framework for Food Security and Nutrition. https://www.fao.org/3/me498e/me498e.pdf (2012). 6. Clapp J, Moseley WG, Burlingame B, Termine P. **The case for a six-dimensional food security framework**. *Food Policy* (2021.0) **106** 102164. DOI: 10.1016/j.foodpol.2021.102164 7. Chavez E, Conway G, Ghil M, Sadler M. **An end-to-end assessment of extreme weather impacts on food security**. *Nat. Clim. Change* (2015.0) **5** 997-1001. DOI: 10.1038/nclimate2747 8. Hasegawa T. **Extreme climate events increase risk of global food insecurity and adaptation needs**. *Nat. Food* (2021.0) **2** 587-595. DOI: 10.1038/s43016-021-00335-4 9. Balk D. **Child hunger in the developing world: An analysis of environmental and social correlates**. *Food Policy* (2005.0) **30** 584-611. DOI: 10.1016/j.foodpol.2005.10.007 10. Grace K, Davenport F, Hanson H, Funk C, Shukla S. **Linking climate change and health outcomes: Examining the relationship between temperature, precipitation and birth weight in Africa**. *Glob. Environ. Change* (2015.0) **35** 125-137. DOI: 10.1016/j.gloenvcha.2015.06.010 11. Shively GE. **Infrastructure mitigates the sensitivity of child growth to local agriculture and rainfall in Nepal and Uganda**. *Proc. Natl. Acad. Sci.* (2017.0) **114** 903-908. DOI: 10.1073/pnas.1524482114 12. Randell H, Gray C, Grace K. **Stunted from the start: Early life weather conditions and child undernutrition in Ethiopia**. *Soc. Sci. Med.* (2020.0) **261** 113234. DOI: 10.1016/j.socscimed.2020.113234 13. Dowd C. *Conflict and Hunger* (2020.0) 1-8 14. Martin-Shields CP, Stojetz W. **Food security and conflict: Empirical challenges and future opportunities for research and policy making on food security and conflict**. *World Dev.* (2019.0) **119** 150-164. DOI: 10.1016/j.worlddev.2018.07.011 15. 15.Brück, T. & d’Errico, M. Reprint of: Food Security and Violent Conflict: Introduction to the Special Issue (2019). 16. Drammeh W, Hamid NA, Rohana A. **Determinants of household food insecurity and its association with child malnutrition in sub-Saharan Africa: A review of the literature**. *Curr. Res. Nutr. Food Sci. J.* (2019.0) **7** 610-623. DOI: 10.12944/CRNFSJ.7.3.02 17. Gundersen C, Ziliak JP. **Food insecurity and health outcomes**. *Health Aff.* (2015.0) **34** 1830-1839. DOI: 10.1377/hlthaff.2015.0645 18. Farrell P, Thow AM, Abimbola S, Faruqui N, Negin J. **How food insecurity could lead to obesity in lmics: When not enough is too much: A realist review of how food insecurity could lead to obesity in low-and middle-income countries**. *Health Promot. Int.* (2018.0) **33** 812-826. DOI: 10.1093/heapro/dax026 19. Shayo FK, Lawala PS. **Does food insecurity link to suicidal behaviors among in-school adolescents? Findings from the low-income country of sub-Saharan Africa**. *BMC Psychiatry* (2019.0) **19** 1-8. PMID: 30606141 20. Falkendal T. **Grain export restrictions during COVID-19 risk food insecurity in many low-and middle-income countries**. *Nat. Food* (2021.0) **2** 11-14. DOI: 10.1038/s43016-020-00211-7 21. Mueller V. **Food insecurity and COVID-19 risk in low- and middle-income countries**. *Appl. Econ. Perspect. Policy* (2022.0) **44** 92-109. DOI: 10.1002/aepp.13200 22. Rosegrant MW, Cline SA. **Global food security: Challenges and policies**. *Science* (2003.0) **302** 1917-1919. DOI: 10.1126/science.1092958 23. 23.WFP. Food Security Analysis. https://www.wfp.org/food-security-analysis. Accessed 22 June 2022. 24. 24.WFP. HungerMap LIVE. https://hungermap.wfp.org/. Accessed 22 June 2022. 25. Elliott G, Timmermann A. **Forecasting in economics and finance**. *Ann. Rev. Econ.* (2016.0) **8** 81-110. DOI: 10.1146/annurev-economics-080315-015346 26. Timmermann A. **Forecasting methods in finance**. *Annu. Rev. Financ. Econ.* (2018.0) **10** 449-479. DOI: 10.1146/annurev-financial-110217-022713 27. Tizzoni M. **Real-time numerical forecast of global epidemic spreading: Case study of 2009 a/h1n1pdm**. *BMC Med.* (2012.0) **10** 1-31. DOI: 10.1186/1741-7015-10-165 28. Kraemer MU. **Past and future spread of the arbovirus vectors**. *Nat. Microbiol.* (2019.0) **4** 854-863. DOI: 10.1038/s41564-019-0376-y 29. 29.Perrotta, D., Tizzoni, M. & Paolotti, D. Using participatory web-based surveillance data to improve seasonal influenza forecasting in Italy. In Proceedings of the 26th International Conference on World Wide Web 303–310 (2017). 30. Scarpino S, Petri G. **On the predictability of infectious disease outbreaks**. *Nat. Commun.* (2019.0) **10** 898. DOI: 10.1038/s41467-019-08616-0 31. 31.Wanner, N., Cafiero, C., Troubat, N. & Conforti, P. Refinements to the FAO Methodology for Estimating the Prevalence of Undernourishment Indicator (FAO, 2014). 32. 32.FAO IFAD UNICEF WFP and WHO. The State of Food Security and Nutrition in the World 2020. Transforming Food Systems for Affordable Healthy Diets. https://www.fao.org/publications/sofi/2020/en/ (2020). 33. 33.Mwebaze, E., Okori, W. & Quinn, J. A. Causal structure learning for famine prediction. In 2010 AAAI Spring Symposium Series (2010). 34. 34.Okori, W. & Obua, J. Machine learning classification technique for famine prediction. In Proceedings of the World Congress on Engineering, Vol. 2 4–9 (Citeseer, 2011). 35. Lentz E, Michelson H, Baylis K, Zhou Y. **A data-driven approach improves food insecurity crisis prediction**. *World Dev.* (2019.0) **122** 399-409. DOI: 10.1016/j.worlddev.2019.06.008 36. Martini G. **Machine learning can guide food security efforts when primary data are not available**. *Nat. Food* (2022.0) **3** 716-728. DOI: 10.1038/s43016-022-00587-8 37. 37.Andree, B. P. J., Chamorro, A., Kraay, A., Spencer, P. & Wang, D. Predicting Food Crises (The World Bank, 2020). 38. 38.Wang, D., Andree, B. P. J., Chamorro, A. F. & Girouard Spencer, P. Stochastic Modeling of Food Insecurity (The World Bank, 2020). 39. Westerveld JJ. **Forecasting transitions in the state of food security with machine learning using transferable features**. *Sci. Total Environ.* (2021.0) **786** 147366. DOI: 10.1016/j.scitotenv.2021.147366 40. 40.Partners, I. G. Integrated Food Security Phase Classification: Technical Manual Version 3.0: Evidence and Standards for Better Food Security Decisions (Food and Agriculture Organization of the United Nations, 2019). 41. 41.WFP. Food Consumption Analysis. Calculation and Use of the Food Consumption Score in Food Security Analysis. https://documents.wfp.org/stellent/groups/public/documents/manual_guide_proced/wfp197216.pdf (2008). 42. Vaitla B. **The measurement of household food security: Correlation and latent variable analysis of alternative indicators in a large multi-country dataset**. *Food Policy* (2017.0) **68** 193-205. DOI: 10.1016/j.foodpol.2017.02.006 43. 43.FAO. The Food Insecurity Experience Scale. https://www.fao.org/in-action/voices-of-the-hungry/fies/en/. Accessed 07 Dec 2022. 44. Backer D, Billing T. **Validating famine early warning systems network projections of food security in Africa, 2009–2020**. *Glob. Food Secur.* (2021.0) **29** 100510. DOI: 10.1016/j.gfs.2021.100510 45. Bowling A. **Mode of questionnaire administration can have serious effects on data quality**. *J. Public Health* (2005.0) **27** 281-291. DOI: 10.1093/pubmed/fdi031 46. Gourlay S, Kilic T, Martuscelli A, Wollburg P, Zezza A. **High-frequency phone surveys on COVID-19: Good practices, open questions**. *Food Policy* (2021.0) **105** 102153. DOI: 10.1016/j.foodpol.2021.102153 47. Lamanna C. **Strengths and limitations of computer assisted telephone interviews (CATI) for nutrition data collection in rural Kenya**. *PLoS ONE* (2019.0) **14** e0210050. DOI: 10.1371/journal.pone.0210050 48. 48.WFP. The World Food Programme’s Real-Time Monitoring Systems: Approaches and Methodologies. https://docs.wfp.org/api/documents/WFP-0000135070/download/. Accessed 22 June 2022. 49. Brubaker J, Kilic T, Wollburg P. **Representativeness of individual-level data in COVID-19 phone surveys: Findings from sub-Saharan Africa**. *PLoS ONE* (2021.0) **16** 1-27. DOI: 10.1371/journal.pone.0258877 50. 50.Glazerman, S., Rosenbaum, M., Sandino, R. & Shaughnessy, L. Remote Surveying in a Pandemic: Handbook. https://poverty-action.org/sites/default/files/publications/IPA-Phone-Surveying-in-a-Pandemic-Handbook.pdf (2020). 51. 51.Ambel, A. A., Mcgee, K. R. & Tsegay, A. H. Reducing Bias in Phone Survey Samples: Effectiveness of Reweighting Techniques Using Face-to-Face Surveys as Frames in Four African Countries. Policy Research Working Paper Series 9676 (The World Bank, 2021). 52. Bandt C, Pompe B. **Permutation entropy: A natural complexity measure for time series**. *Phys. Rev. Lett.* (2002.0) **88** 174102. DOI: 10.1103/PhysRevLett.88.174102 53. Garland J, James R, Bradley E. **Model-free quantification of time-series predictability**. *Phys. Rev. E* (2014.0) **90** 052910. DOI: 10.1103/PhysRevE.90.052910 54. Zanin M, Zunino L, Rosso OA, Papo D. **Permutation entropy and its main biomedical and econophysics applications: A review**. *Entropy* (2012.0) **14** 1553-1577. DOI: 10.3390/e14081553 55. Kozak J, Kania K, Juszczuk P. **Permutation entropy as a measure of information gain/loss in the different symbolic descriptions of financial data**. *Entropy* (2020.0) **22** 330. DOI: 10.3390/e22030330 56. Pennekamp F. **The intrinsic predictability of ecological time series and its potential to guide forecasting**. *Ecol. Monogr.* (2019.0) **89** e01359. DOI: 10.1002/ecm.1359 57. Zunino L. **Commodity predictability analysis with a permutation information theory approach**. *Phys. A* (2011.0) **390** 876-890. DOI: 10.1016/j.physa.2010.11.020 58. 58.Food Security Information Network. Global Report on Food Crises. https://www.wfp.org/publications/global-report-food-crises-2021 (2021). 59. 59.Chen, T. & Guestrin, C. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 785–794 (2016). 60. 60.WFP. The Coping Strategies Index: Field Methods Manual. https://documents.wfp.org/stellent/groups/public/documents/manual_guide_proced/wfp211058.pdf (2008). 61. Raleigh C, Linke A, Hegre H, Karlsen J. **Introducing acled: An armed conflict location and event dataset: Special data feature**. *J. Peace Res.* (2010.0) **47** 651-660. DOI: 10.1177/0022343310378914 62. Liu X. **Agricultural drought monitoring: Progress, challenges, and prospects**. *J. Geogr. Sci.* (2016.0) **26** 750-767. DOI: 10.1007/s11442-016-1297-9 63. D’Souza A, Jolliffe D. **Conflict, food price shocks, and food insecurity: The experience of afghan households**. *Food Policy* (2013.0) **42** 32-47. DOI: 10.1016/j.foodpol.2013.06.007 64. 64.Knippenberg, E., Jensen, N. & Constas, M. Resilience, Shocks, and the Dynamics of Food Insecurity: Evidence from Malawi. Technical Report, Working Paper (Cornell University, Ithaca, NY, 2018). 65. Headey D. **Rethinking the global food crisis: The role of trade shocks**. *Food Policy* (2011.0) **36** 136-146. DOI: 10.1016/j.foodpol.2010.10.003 66. Makridakis S, Spiliotis E, Assimakopoulos V. **The m4 competition: 100,000 time series and 61 forecasting methods**. *Int. J. Forecast.* (2020.0) **36** 54-74. DOI: 10.1016/j.ijforecast.2019.04.014 67. Cardwell R, Ghazalian PL. **COVID-19 and international food assistance: Policy proposals to keep food flowing**. *World Dev.* (2020.0) **135** 105059. DOI: 10.1016/j.worlddev.2020.105059 68. Workie E, Mackolil J, Nyika J, Ramadas S. **Deciphering the impact of COVID-19 pandemic on food security, agriculture, and livelihoods: A review of the evidence from developing countries**. *Curr. Res. Environ. Sustain.* (2020.0) **2** 100014. DOI: 10.1016/j.crsust.2020.100014 69. Whittaker C. **Under-reporting of deaths limits our understanding of true burden of COVID-19**. *bmj* (2021.0) **375** n2239. DOI: 10.1136/bmj.n2239 70. Singh S, Nourozi S, Acharya L, Thapa S. **Estimating the potential effects of COVID-19 pandemic on food commodity prices and nutrition security in Nepal**. *J. Nutr. Sci.* (2020.0) **9** e51. DOI: 10.1017/jns.2020.43 71. Altay N, Narayanan A. **Forecasting in humanitarian operations: Literature review and research needs**. *Int. J. Forecast.* (2020.0) **38** 1234-1244. DOI: 10.1016/j.ijforecast.2020.08.001 72. Lim B, Zohren S. **Time-series forecasting with deep learning: A survey**. *Philos. Trans. R. Soc. A* (2021.0) **379** 20200209. DOI: 10.1098/rsta.2020.0209 73. Zufiria PJ. **Identifying seasonal mobility profiles from anonymized and aggregated mobile phone data. Application in food security**. *PLoS ONE* (2018.0) **13** 0195714. DOI: 10.1371/journal.pone.0195714 74. 74.Kwak, H. & An, J. A first look at global news coverage of disasters by using the gdelt dataset. In International Conference on Social Informatics 300–308 (Springer, 2014). 75. 75.The Integrated Food Security Phase Classification (IPC) Global Partners. Technical Manual Version 3.0. Evidence and Standards for Better Food Security and Nutrition Decisions. https://www.ipcinfo.org/fileadmin/user_upload/ipcinfo/docs/IPC_Technical_Manual_3_Final.pdf (2019). 76. Borge-Holthoefer J. **The dynamics of information-driven coordination phenomena: A transfer entropy analysis**. *Sci. Adv.* (2015.0) **2** e1501158. DOI: 10.1126/sciadv.1501158 77. Cao L. **Practical method for determining the minimum embedding dimension of a scalar time series**. *Phys. D* (1997.0) **110** 43-50. DOI: 10.1016/S0167-2789(97)00118-8 78. Kennel MB, Brown R, Abarbanel HDI. **Determining embedding dimension for phase-space reconstruction using a geometrical construction**. *Phys. Rev. A* (1992.0) **45** 3403-3411. DOI: 10.1103/PhysRevA.45.3403 79. 79.Jordahl, K. et al. geopandas/geopandas: v0.9.0. 10.5281/zenodo.4569086 (2021).
--- title: The rhythmic coupling of Egr-1 and Cidea regulates age-related metabolic dysfunction in the liver of male mice authors: - Jing Wu - Dandan Bu - Haiquan Wang - Di Shen - Danyang Chong - Tongyu Zhang - Weiwei Tao - Mengfei Zhao - Yue Zhao - Lei Fang - Peng Li - Bin Xue - Chao-Jun Li journal: Nature Communications year: 2023 pmcid: PMC10038990 doi: 10.1038/s41467-023-36775-8 license: CC BY 4.0 --- # The rhythmic coupling of Egr-1 and Cidea regulates age-related metabolic dysfunction in the liver of male mice ## Abstract The liver lipid metabolism of older individuals canbecome impaired and the circadian rhythm of genes involved in lipid metabolism is also disturbed. Although the link between metabolism and circadian rhythms is already recognized, how these processes are decoupled in liver during aging is still largely unknown. Here, we show that the circadian rhythm for the transcription factor Egr-1 expression is shifted forward with age in male mice. Egr-1 deletion accelerates liver age-related metabolic dysfunction, which associates with increased triglyceride accumulation, disruption of the opposite rhythmic coupling of Egr-1 and Cidea (Cell Death Inducing DFFA Like Effector A) at the transcriptional level and large lipid droplet formation. Importantly, adjustment of the central clock with light via a 4-hour forward shift in 6-month-old mice, leads to recovery the rhythm shift of Egr-1 during aging and largely ameliorated liver metabolic dysfunction. All our collected data suggest that liver Egr-1 might integrate the central and peripheral rhythms and regulate metabolic homeostasis in the liver. Many transcriptomic pathways in the liver show circadian rhythms, which have been reported to be disrupted in aged mice. Here the authors report that the expression of transcription factor Egr-1 decreases and its rhythm is shifted with age in the liver of male mice, and that deletion of Egr-1 results in increased liver fat accumulation. ## Introduction Aging is associated with metabolic disorders such as obesity, hyperglycemia, and hyperlipidemia1 resulting from age-related nutrient sensing deregulation, mitochondrial dysfunction, cellular signaling pathway reprogramming, and insulin resistance2. As the central metabolic organ, the liver exhibits signs of progressive metabolic disorder during aging, such as triglyceride (TG) accumulation enhancement, fatty acid oxidation inhibition, and lipolysis impairment3–5. On the other hand, the circadian rhythms in the superchiasmatic nucleus (SCN) and peripheral tissues also undergo significant changes during aging6. The rhythm of Per1 expression seen in the lungs of young rats (2 months old) is absent in old animals (24–26 months old)7. The expression amplitude of clock genes such as Clk and Pdp1cε is also decreased in the head and body tissues of aged flies (58 days old)8. These declines in peripheral clocks may affect rhythmic changes in hormone release, temperature regulation, and metabolism6. Many metabolic processes in the liver, such as gluconeogenesis, lipogenesis, and bile acid synthesis, also show rhythmic changes regulated by master circadian rhythms, and these metabolic processes exhibit progressive alterations with age6,9,10. It has been reported that $44.8\%$ of genes that are rhythmically expressed in young mice exhibit rhythm disorders in elderly mice; these genes are mainly involved in glycerol metabolism and sterol metabolism2. Deficiencies in circadian genes, such as Bmal1 and Per$\frac{1}{2}$, can accelerate the aging process11. Thus, impairment of metabolic and circadian rhythmic synchronization might be particularly important with age increased in the liver, but how the processes are coordinated still needs to be explored. Early growth response-1 (Egr-1) is a member of the immediate early gene family that recognizes a highly conserved GC-based promoter sequence and then regulates the expression of many target genes12,13. Egr-1 can be activated by a variety of stimuli, including cytokines, growth factors, and hormones, and regulates cell proliferation, cell metabolism, and the hepatic clock circuitry14–16. It has been reported that both the mRNA and protein levels of Egr-1 are significantly decreased in senescent cell lines17,18. RNA sequencing (RNA-seq) data have shown that Egr-1 expression in the livers of 21-month-old mice is significantly decreased19. Previous studies have also shown that Egr-1 can regulate glucose and lipid metabolism and act as a molecular brake that prevents excessive stimulation and regulates fluctuating blood glucose levels under physiologic conditions20,21. After a meal, Egr-1 can be induced by insulin in skeletal muscle cells and inhibit insulin receptor phosphorylation, thus reducing insulin sensitivity20. After long-term fasting, Egr-1 can be activated by glucagon and regulate the expression of gluconeogenesis genes in the liver21. Moreover, Egr-1 acts as an important regulator in lipid metabolism. Egr-1 enhanced by insulin directly inhibits patatin-like phospholipase domain containing 2 (ATGL) transcription and inhibits lipolysis of adipocytes22. Egr-1 also affects the transcription of the key cholesterol synthesis genes Hmgcr, Cyp51, Me1, and Sqle and promotes cholesterol anabolism23. In addition, our previous work has demonstrated that Egr-1 is rhythmically expressed in the livers of young mice and is required for the circadian expression patterns of several core clock genes in the liver; it especially regulates the transcriptional activity of the biorhythm gene Per1. On the other hand, the rhythm of Egr-1 is also regulated by BMAL1/CLOCK heterodimer expression16. Thus, we speculated that Egr-1 may act as a mediator to regulate the age-associated cooperation between circadian rhythms and metabolic patterns. In this work, we demonstrate that the zeitgeber time (ZT) of Egr-1 peak expression is shifted forward with age. Egr-1 deletion accelerates liver age-related TG accumulation by enhancing CD36 expression to facilitate fatty acid uptake and enhancing Cidea (cell death inducing DFFA like effector A) transcriptional expression to form large lipid droplets. The rhythmic coupling of Egr-1 and Cidea can regulate the formation of large lipid droplets in a BMAL1/CLOCK-dependent manner. Aging disrupts the coupling between Egr-1 and Cidea and facilitates large lipid droplet formation, resulting in age-related metabolic dysfunction of the liver. These results indicate that Egr-1 is a key mediator that regulates the age-associated cooperation between circadian rhythm and lipid metabolism in the liver. ## Liver rhythmic lipid metabolism is disrupted with age increased To determine how lipid metabolism changes with age increased, we detected triglyceride (TG) accumulation and the results suggested that TG levels in the liver increased with age (Fig. 1A). Hematoxylin and eosin (H&E) staining and Oil Red O staining also showed that lipid droplet accumulation in the livers of mice increased at 12 months (Fig. 1B), when positive staining for the aging marker β-galactosidase (β-gal) began, and was significantly increased at 21 months (Fig. 1C). This observation indicated a phenomenon of metabolic dysfunction in the liver with age increase. Moreover, to determine how rhythmic metabolism and the hepatic clock change with age increased, mice of different ages (2, 6, and 12 months of age) were sacrificed every 4 h started at zeitgeber time (ZT) 1 over the circadian cycle (ZT1, 5, 9, 13, 17, 21) and hepatic transcriptomic analysis was performed. Heatmaps and Venn diagrams display circadian genes selected using the nonparametric algorithm JTK_cycle24 and genes with p. adjust<0.05 were regarded as circadian genes. Exclusive circadian genes were found and displayed by heatmaps in the 2, 6, and 12-month groups (Fig. 1D–F). Transcriptomics revealed 621 genes exclusively oscillatory in 2-month group, 579 genes exclusively oscillatory in the 6-month group, and 428 genes were only rhythmic in the 12-month group. Among the three groups, only 12 genes kept similar rhythms and fewer metabolism-related genes maintained consistent rhythms (Fig. 1G). GO enrichment analysis of oscillatory genes in a 2-month group and selecting the top 20 biological pathways indicated that 45.1 percent of these processes were related to lipid metabolism (Fig. 1H). However, with age increased, the proportion of lipid metabolism-related pathways gradually decreased to 28.3 percent in a 6-month group (Fig. 1I) and only 3 percent in 12-month group (Fig. 1J). Thus, these results revealed that rhythmic lipid metabolism was disordered with aging. Fig. 1The rhythmic lipid metabolism was disordered with aging. A Liver TG levels of C57BL/6 J mice at 2 months, 6 months, and 12 months (2 months: $$n = 6$$; 6 months: $$n = 5$$; 12 months: $$n = 5$$ biologically independent animals); B H&E and Oil Red O staining of liver tissues of C57BL/6 J mice at 2 months, 6 months, and 12 months; C β-Gal staining of liver tissues of C57BL/6 J mice at 2 months, 6 months, 12 months, and 21 months; D–F Heatmap represents rhythmic genes exclusively in the livers of C57BL/6 J mice at 2 months, 6 months, and 12 months by using high-throughput RNA sequencing. The colors from blue to yellow indicate low to high gene expression levels, respectively; G Venn diagram displays the total number of rhythmic genes (left) and number of rhythmic non-metabolic or metabolic genes (right) in the liver; the black dots mean genes only in the indicated groups; a black line connecting black dots indicates that genes are in the connected group at the same time; H–J Pie charts indicate selected Top 20 biological process by using gene ontology (GO) analysis of genes circadian in 2 months, 6 months, and 12 months groups. The yellow of the outer circle means a lipid-related pathway, blue of the outer circle means a non-lipid-related pathway. Data were represented as mean ± SEM. Exact p values are depicted in the figure. Statistical analysis was performed using one-way ANOVA for A. Source data are provided as a Source Data file. Detailed analysis of the circadian rhythm changes from hepatic transcriptomic and RT-PCR confirmed the rhythmicity change of most circadian genes in 2, 6, and 12 months of age livers. It is intrinsic that although the phase of most circadian genes in 6 months of age livers was moved forward or backward compared with 2 months of age livers, it was similar in 2 and 12 months of age livers, only the amplitude of Bmal1 and Clock and Rorα was significantly changed (Supplementary Fig. 1A–I). However, the related gene number in enriched circadian rhythm-related biological pathways was remarkedly decreased in 12-month-old mice except for the entrainment of the circadian clock and the entrainment of the circadian clock by photoperiod. And some pathways even disappeared like circadian sleep/wake cycle non-REM sleep, positive regulation of circadian sleep/wake cycle, and negative/positive regulation of circadian rhythm, which indicated that sleep/wake problems emerged in old age (Supplementary Fig. 1J). Meanwhile, the regulation of core circadian gene on TG accumulation is totally different between 2- and 12-month-old mice, in which knockdown of Bmal1 or Clock or Rorα could enhance TG accumulation in 2-month-old mice25–27, while decreased TG accumulation in 12-month-old mice (Supplementary Fig. 1K–M), which suggested the “normal” 2-month-old relationship between core clock genes and lipid metabolism was disrupted in the liver of 12-month-old mice. ## The rhythmic phase of Egr-1 in the liver shifts forward with aging In order to figure out the regulators to connect the circadian rhythms and metabolic patterns with age increased, we further analyzed the above transcriptomic data. The heatmap vividly displayed that core clock genes and clock-controlled genes (CCGs) in the circadian rhythm process had different rhythms in different age groups (Fig. 2A). We found that the rhythm of early-response transcription factor Egr-1 in multiple circadian rhythm processes was changed, which have been reported to participate in lipid metabolism22. We then detected the circadian expression of Egr-1 and found that the rhythmicity shifted with age. The mRNA expression of Egr-1 in the livers of 2-month-old mice peaked at ZT5 and declined thereafter, showing the lowest expression at ZT17. Interestingly, with increasing age, the peak of Egr-1 mRNA expression advanced, occurring at ZT1 in 6-month-old mice and at ZT21 in 12-month-old mice (Fig. 2B). The protein expression levels of Egr-1 in the livers of 2-month-old mice peaked at ZT13 and then advanced to ZT9 in 6-month-old mice and to ZT5 in 12-month-old mice (Fig. 2C-F). Moreover, by analyzing gene array data (GSE57809) of live young and old mice19, we found that the mRNA levels of Egr-1 were markedly decreased in old mice (Supplementary Fig. 2A). We also detected the protein levels of Egr-1 at ZT5 and further confirmed their slight decline in the livers of 21-month-old mice (Supplementary Fig. 2B, C). Moreover, we find an increasing tendency in Egr-1 expression in the livers of 6-month-old mice (Supplementary Fig. 2D, E). These results indicate that the circadian rhythm of liver Egr-1 is also altered with aging. Fig. 2The phase of the Egr-1 circadian rhythm in the liver moves forward with aging. A Heatmap represents rhythmic clock and clock-controlled genes exclusively in the livers of C57BL/6 mice at 2 months, 6 months, and 12 months; B mRNA expression of Egr-1 at the indicated time points in the livers of C57BL/6 J mice at 2 months, 6 months, and 12 months (2 month ZT1: $$n = 4$$, ZT5,9,21: $$n = 5$$, ZT13,17:$$n = 6$$; 6 month ZT1,9: $$n = 5$$, ZT5,13: $$n = 4$$, ZT17,21:$$n = 6$$; 12 month: $$n = 5$$ biologically independent animals), black dot means the peak expression of Egr-1 point at each group. C–E Protein expression of Egr-1 at the indicated time points in the livers of C57BL/6 J mice at 2 months, 6 months, and 12 months ($$n = 5$$ biologically independent animals in 2 months; $$n = 3$$ biologically independent animals in 6 months; $$n = 4$$ biologically independent animals in 12 months). F *Quantitative analysis* of the Egr-1 protein levels in C–E, $$n = 3$$ per group. G–I Venn diagrams representing the overlap between WT group lipid-related genes and Egr-1 ChIP-Seq (GSM1037682) genes. J Venn diagrams displayed the overlap among intersecting genes in F–H. Data were represented as mean ± SEM. Exact p values are depicted in the figure. Orange color p value means 6 months versus 2-month group; Purple color p value means 12 months versus 2-month group; Black color p value means 6 months versus 12-month group. Statistical analysis was performed using one-way ANOVA for B and F. Source data are provided as a Source Data file. To examine whether the changed rhythm of liver Egr-1 was relevant to the metabolic dysfunction with age increase, we compared genes in the lipid-related biological process of Fig. 1H–J with the top 2000 genes from a published dataset Egr-1 ChIP-seq (GSM1037682)28, which selected via sorting the score. The results indicated that 27.2 percent of lipid-related circadian genes were regulated by Egr-1 in 2-month group, 33.8 percent of lipid-related circadian genes in 6-month group, and only 21.6 percent of genes in 12-month group (Fig. 2G–I). However, only two of these genes were consistently regulated by Egr-1 with aging (Fig. 2J). These analyses suggested that age-induced changes in Egr-1 rhythms may lead to either targeting different lipid metabolism genes or shifting their rhythm, which in turn results in metabolism dysfunction. ## Egr-1 deletion accelerates liver age-related metabolic dysfunction To determine the causal relationship between Egr-1 circadian rhythm and metabolic process, we generated liver-specific Egr-1-knockout (KO) mice using Alb-Cre transgenic mice. We found that Egr-1 deficiency aggravated aging-related liver lipid metabolic dysfunction. There were no obvious alterations in whole-body weight until 21 months of age, when obvious obesity was found in liver-specific Egr-1 KO mice compared with wild-type (WT) mice (Supplementary Fig. 3A, B). There were significant decreases in liver weight of Egr-1 KO mice aged 2-, 6-, and 12-month-old, although no significant differences were observed in 21-month-old mice (Supplementary Fig. 3C). However, the liver/body weight ratio was already decreased in the group of 2-month-old KO mice (Supplementary Fig. 3D). At the 6th month, the liver TG levels in KO mice were markedly increased, reaching the levels in 12-month-old WT mice; and then TG levels remained at the same magnitude until the 21st month, when they increased again to approximately 90 mM/g tissue (Fig. 3A). H&E staining and oil Red O staining also proved that TGs accumulated and that lipid droplet size increased in hepatocytes after Egr-1 deletion (Fig. 3B, C). Meanwhile, the liver-free fatty acid level in KO mice aged 6-month-old were also significantly increased (Fig. 3D) and the serum-free fatty acid level were decreased (Fig. 3E). Since Egr-1 is a transcription factor, its deletion should alter the expression of downstream genes. We found that Egr-1 deficiency in the 6-month-livers of mice significantly augmented the amplitude of fatty acid uptake genes such as CD36 (Supplementary Fig. 3E), shifted the phase of fatty acid uptake gene FATP (Supplementary Fig. 3F); and suppressed the rhythm of de novo lipogenesis (Supplementary Fig. 3H–L) and TG transport-related genes (Supplementary Fig. 3R–T). The rhythmic expression of genes for fatty acid oxidation were not significantly affected by Egr-1 deletion. Thus, Egr-1 deficiency disrupted the lipid flow (flux) “in and out” balance to enable the accumulation of excessive fatty acids from 6 months onward. Moreover, Sirius Red staining and fibrosis-related gene detection showed that the fatty liver in Egr-1-deleted mice further developed into mild fibrosis (Fig. 3F, G). Moreover, Egr-1 deficiency enhanced hepatocyte aging (Fig. 3H) and even slightly shortened the survival times of the mice (Fig. 3I). According to the above results, we concluded that Egr-1 could regulate metabolic dysfunction in a manner dependent on its transcriptional effects by affecting lipid metabolism. Fig. 3Egr-1 deficiency accelerates liver age-related metabolic dysfunction. A Liver TG levels of WT and Egr-1-LKO mice at 2 months, 6 months, 12 months, and 21 months (WT: 2 months: $$n = 7$$; 6 months: $$n = 7$$; 12 months: $$n = 8$$; 21 months: $$n = 6$$; Egr-1 LKO: 2 months: $$n = 6$$; 6 months: $$n = 5$$; 12 months: $$n = 6$$; 21 months: $$n = 6$$ biologically independent animals). B, C H&E staining and Oil Red O staining of WT and Egr-1-LKO mice at 2 months, 6 months, and 21 months of age. D, E Liver and serum-free fatty acid levels of WT and Egr-1-LKO mice at 6 months ($$n = 6$$ or 7 biologically independent animals in WT group and $$n = 5$$ or 6 biologically independent animals in Egr-1 LKO group). F Sirius Red staining of 21-month-old WT and Egr-1-LKO mice. G mRNA levels of the liver fibrosis marker a-SMA ($$n = 9$$ biologically independent animals). H β-Galactosidase staining indicates the aging process. I Survival curves of WT and Egr-1-LKO mice. Each experiment was repeated three times independently (WT: $$n = 22$$; Egr-1 LKO: $$n = 23$$ biologically independent animals). Data were represented as mean ± SEM. Exact p values are depicted in the figure. Statistical analysis was performed using one-way ANOVA for A and unpaired two-tailed Student’s t-test for D, E, and G. Scale bar: 100 µm. Source data are provided as a Source Data file. ## Transcriptomic analysis of the livers in Egr-1-deleted mice To determine the underlying mechanism of Egr-1-regulated metabolic dysfunction in the liver, the WT, and Egr-1-KO liver samples were collected at Egr-1 highest(H) and lowest(L) zeitgeber time in mice of different ages. Liver samples were obtained at ZT13(H) and ZT17(L) in 2-month group; at ZT9(H) and ZT21(L) in 6-month group; at ZT5(H) and ZT21(L) in 12-month group; then hepatic transcriptomic analysis was performed. Heatmaps displayed significant differential gene expression patterns in each group (Fig. 4A). Venn diagrams showed the selected genes which were directly upregulated or downregulated by Egr-1 at different ages (Fig. 4B and Supplementary Fig. 4A). GO enrichment analysis of these altered genes were accomplished. Interestingly, the processes, which were selected via significantly enriched associated with lipid, mostly have a negative function of liver lipid metabolism in 2-month group, such as negative regulation of lipid storage, negative regulation of lipid localization, and negative regulation of lipid biosynthetic process. However, in 12-month group, that of GO term mainly involved in the positive function of lipid metabolism, including regulation of lipid localization, lipid transport, positive regulation of lipid localization, and so on (Fig. 4C). Given that the shifted phase of Egr-1 circadian rhythm accelerates liver age-related metabolic dysfunction, more precise genes were obtained by overlapping genes that indirect regulated by Egr-1 in 2-month group and genes that were directly regulated by the rhythm of Egr-1 in 12 month group. Further analysis these genes showed that four upregulated genes (Cyp2u1, Enho, Apoa1, Smim22) were related to lipid metabolism and only one downregulated gene Cidea was relevant to lipid metabolism (Fig. 4D and Supplementary Fig. 4B,C).Fig. 4Transcriptomic analysis of the liver in Egr-1-deleted mice with age increased. A Heatmap representation of the genes at Egr-1 highest and lowest zeitgeber time of mice aged 2 months, 6 months, and 12 months. The colors from blue to yellow indicate low to high gene expression levels, respectively. B *Intersect* gene counts statistics of Venn diagrams in Supplementary Fig. 4A. C Selected significantly enriched lipid-related GO terms of 2-month and 12-month intersect genes. The statistical test of data analysis was performed using a hypergeometric test, two-tailed, no adjustment ($p \leq 0.05$). D Venn diagrams representing the overlap between 2m_no_intersect_down and 12_intersect_down group. The lipid metabolism-related genes were labeled with red font. E Euclidean distance of each group difference at different ages. F Volcano plot displayed genes expressed in WT_H and KO_H groups in mice at the age of 6 months. The relative expression changes and significance levels are shown. The statistical test of data analysis was performed using two-tailed, no adjustment ($p \leq 0.05$) from the limma package in the R environment. G Selected significantly enriched GO terms. The x-axis and y-axis represent the enrichment and significance level, respectively, and the size of the circle represents the number of genes associated with the GO term. The statistical test of data analysis was performed using a hypergeometric test, two-tailed, no adjustment ($p \leq 0.05$). H Chord diagram revealing the enrichment levels of genes related to selected GO terms. I Relative FPKM values of Cidea in WT_H and KO_H group. We used four mice per group for the analysis. One-way ANOVA for A and unpaired two-tailed Student’s t-test for D, E, and G. Moreover, Euclidean distances were calculated among all samples at different ages. We found WT_6m_H and KO_6m_H groups showed the biggest difference (Fig. 4E). Phenotype differences were also observed from the 6th month; thus, we compared the transcriptomes between the WT_H and Egr-1-KO_H groups at 6 months, screening 613 genes with significant changes. Among these genes, 309 genes showed significant increases, such as Mup17, Mup19, Cidea, and Gapdh, while 304 genes showed significant decreases, such as the transcription factors Egr-1, Egr-2, Myc, and Atf3 (Fig. 4F). We identified ten biological processes terms, such as the regulation of lipid catabolic process term and the lipid droplet organization term, that were significantly enriched (Fig. 4G). Among the identified enriched genes, we found that Cidea (Cell Death Inducing DFFA Like Effector A) participated in four of ten related metabolic processes, such as lipid localization, TG sequestration, and lipid catabolic process regulation (Fig. 4H). However, Egr-1 was not found in the list of genes related to these ten processes. As an early-primary response transcription factor, Egr-1 functions via its downstream secondary response genes, such as Tnf29. Importantly, we also isolated *Cidea via* overlapping 2-month group and 12-month group in Fig. 4D. Thus, Cidea might also be an Egr-1 target gene. Verification of the relative fragments per kilobase of transcript per million mapped reads (FPKM) values of Cidea showed that the expression of the gene significantly increased with age beginning in the 6th month and that Egr-1 deletion led to an even more significant increase (Fig. 4I). Thus, we hypothesized that Egr-1 might regulate liver metabolic dysfunction through target genes such as Cidea. ## Cidea mediated the Egr-1-induced liver metabolic dysfunction Cidea is a member of the Cide family that promotes lipid turnover and lipid droplet fusion30–32. To verify the RNA-seq results, we detected the expression levels of Cidea in liver samples. We found that both the mRNA (Fig. 5A) and protein levels (Fig. 5B) increased with age, and Egr-1 deficiency augmented the elevations (Fig. 5A, B and Supplementary Fig. 5A). The expression enhancement was further confirmed in isolated primary hepatocytes, indicating that Cidea protein level was indeed augmented after Egr-1 deletion in 6-month-old mice (Fig. 5C and Supplementary Fig. 5B). Oil Red O staining and BODIPY fluorescence staining indicated that the lipid droplet size was increased after Egr-1 deletion in hepatocytes (Fig. 5D, E). Immunostaining also showed that the Cidea protein is highly expressed in Egr-1 LKO hepatocytes (Fig. 5E). However, we did not detect the Cidea localized in larger lipid droplet surface, possibly because the Cidea protein is only transiently associated with lipid droplets. Further examination revealed that the regulatory effect of Egr-1 on lipid droplets was dependent on Cidea. When we knocked down Cidea in primary hepatocytes, the TG level in primary hepatocytes of Egr-1 LKO mice was dramatically decreased; the age-related increase in lipid droplet size induced by Egr-1 LKO was blocked (Fig. 5F, G). Egr-1 overexpression decreased lipid droplet size and TG level, which was reversed by Cidea overexpression (Fig. 5H, I). While dnEgr-1 overexpression, which inhibits Egr-1 transcriptional activity, increased TG level, but blocked with Cidea knocked down (Supplementary Fig. 5D). The accumulation of TG mainly depended on a fatty acid pool, which is $60\%$ from fatty acids uptake and $25\%$ from DNL33. We detected the fatty acid uptake and found that Cidea did not mediate the fatty acid uptake due to Egr-1. When overexpressing Cidea, the decrease of fatty acid uptake by overexpressing Egr-1 was not rescued (Fig. 5J).Fig. 5Egr-1 regulates liver metabolic aging in a Cidea-dependent manner. A mRNA levels of Cidea in samples from RNA-seq ($$n = 3$$ biologically independent animals per group); B Protein levels of Cidea in samples from RNA-seq; C Protein levels of Cidea in primary hepatocytes of WT and Egr-1 LKO mice at 6 months of age; D Oil Red O staining of primary hepatocytes of WT and Egr-1 LKO mice at 6 months of age; E Immunofluorescence staining of primary hepatocytes of WT and Egr-1 LKO mice at 6 months of age; F TG levels in WT and Egr-1 LKO primary hepatocytes from 2-month-old and 6-month-old mice after infection with a ShCidea adenovirus ($$n = 4$$ biologically independent samples); G Oil Red O staining of WT and Egr-1 LKO primary hepatocytes from 2-month-old and 6-month-old mice after infection with a ShCidea adenovirus; H Hepatocyte TG levels after transfection with a Cidea overexpression plasmid or infection with an Egr-1 overexpression adenovirus ($$n = 4$$ biologically independent samples); I Oil Red O staining after transfection with a Cidea overexpression plasmid or infection with an Egr-1 overexpression adenovirus; J the fatty acid uptake ratio after transfection with a Cidea overexpression plasmid or infection with an Egr-1 overexpression adenovirus ($$n = 14$$ biologically independent samples). Data were represented as mean ± SEM. Exact p values are depicted in the figure. Statistical analysis was performed using one-way ANOVA. Source data are provided as a Source Data file. We have found that Egr-1 deficiency in mice liver augmented the expression of fatty acid uptake gene CD36 in 2, 6, and 12 month groups (Supplementary Figs. 3E, 6A, B). Indeed, the CD36 protein level was remarkedly inhibited by overexpressing Egr-1 (Supplementary Fig. 6C, D) and significantly augmented in KO mice (Supplementary Fig. 6E, F). We also predicted five putative Egr-1 binding sites in the Cd36 promoter sequence according to the JASPAR website (Supplementary Fig. 6G). However, we did not find the rhythmic phase of CD36 changed with age increased (Supplementary Fig. 6H). Another possibility that Egr-1 deficiency enhanced hepatic FFA level is that DNL from amino acids might also be augmented. Go enrichment showed that Egr-1 deficiency could significantly facilitate the amino acid transport, such as amino acid transport and regulation of amino acid transmembrane transport (Supplementary Fig. 6I). Further analysis of these genes indicated that most genes were related to glutamine uptake. Kcnj10 acts as a channel protein and is involved in l-glutamate import. Lpcat4 and Ggt1 could enable acyltransferase activity. Slc13a3 and Slc6a19 improve the glutamine transport across the plasma membrane (Supplementary Fig. 6J). Thus, the results indicated that Egr-1 deletion could accelerate liver TGs accumulation by enhancing CD36 expression to facilitate fatty acid uptake, augmenting FFA synthesis from amino acids like glutamine, then enhancing Cidea expression to form large lipid droplet. ## Egr-1/BMAL1/CLOCK complex regulates the rhythm of Cidea Considering the rhythm of CD36 and amino acid-related genes have not significantly changed with age increase, we verified the rhythmicity of Cidea and found that 2-month-old mice also displayed rhythmic Cidea mRNA expression in the liver that peaked at ZT13 and reached a nadir at ZT9 (Fig. 6A). Similar to that of Egr-1, the rhythmicity of Cidea also shifted with age increased, and the peak of Cidea expression in mouse livers was advanced to ZT5 at the 6th month and ZT17 at the 12th month (Fig. 6A). The rhythmic alteration of Cidea from the 6th month was further confirmed by assessment of Cidea protein expression pattern at different ages (Fig. 6B, C). When we isolated primary hepatocytes from the livers of 6-month-old mice and exposed them to horse serum shock, we observed that the amplitude of Cidea expression was dramatically elevated in KO mice (Fig. 6D and Supplementary Fig. 7A). When we overexpressed Egr-1 in wild-type hepatocytes, the amplitude of Cidea expression was considerably reduced (Fig. 6E and Supplementary Fig. 7B). All the above data indicate that Egr-1 can regulate Cidea expression robustness and rhythmicity. Thus, we analyzed the promoter region of Cidea and predicted two GC-rich Egr-1 binding sites and an E-box sequence region. Chromatin immunoprecipitation (ChIP) and qPCR assays demonstrated that Egr-1 could bind with all of these binding sites (Fig. 6F). A luciferase assay showed that Egr-1 binding significantly inhibited the transcription of Cidea (Fig. 6G), which could be rescued by mutation of Egr-1 binding sites. Interestingly, E-box mutation enhanced Cidea transcription to a higher degree than Egr-1 binding site mutation (Fig. 6H), which suggests that the E-box might be more important for Cidea expression than Egr-1 binding sites. Normally, circadian genes can bind with the E-box and regulate downstream gene expression. Thus, we co-overexpressed BMAL1/CLOCK with Egr-1 and found that E-box mutation had a stronger effect than Egr-1 mutation on BMAL1/CLOCK-controlled Cidea expression (Fig. 6H). A coimmunoprecipitation (Co-IP) experiment further confirmed that Egr-1 could form complexes with BMAL1/CLOCK (Fig. 6I). Therefore, we conclude that aging-related Egr-1 rhythm alterations regulate circadian Cidea expression and then affect liver TG accumulation over time. Egr-1 rhythm alteration might also result in the uncoupling of Egr-1 with BMAL1/CLOCK, which is responsible for changes in the robustness and rhythmicity of Cidea expression. Fig. 6Egr-1/BMAL1/CLOCK regulates the robustness and rhythm of Cidea by inhibiting its transcription. A mRNA expression of Cidea at the indicated time points in livers from C57BL/6 J mice at 2 months, 6 months, and 12 months of age (2 months ZT1: $$n = 6$$, ZT5: $$n = 5$$, ZT9,13,17,21:$$n = 4$$; 6 months ZT1: $$n = 4$$, ZT5,13: $$n = 6$$, ZT9:$$n = 5$$, ZT17,21: $$n = 4$$; 12 months: $$n = 4$$ biologically independent animals); B Protein levels of Cidea at the indicated time points in livers from B6 mice at 2 months, 6 months, and 12 months of age that were entrained to a 12-h light/dark cycle. C *Quantitative analysis* of Cidea protein levels in Fig. 6B ($$n = 4$$ biologically independent animals per group). D Protein levels of Cidea in primary hepatocytes isolated from 6-month-old C57BL/6 J mice after deletion of Egr-1. E Protein levels of Cidea in primary hepatocytes isolated from 6-month-old C57BL/6 J mice overexpressing Egr-1. F ChIP assay ($$n = 3$$ or 4 biologically independent samples per group); G Luciferase assay ($$n = 5$$ biologically independent samples per group). H HEK293T cells were co-transfected with Myc-BMAL1, Egr-1, and CLOCK expression plasmids. Co-IP of Egr-1, Myc-BMAL1, and CLOCK was performed. E means Egr-1, B means BMAL1, and C means CLOCK. Each experiment was repeated three times independently. The data represent the mean ± SEM. Exact p values are depicted in the figure. Orange color p value means 6 months versus 2 month group; Purple color p value means 12 months versus 2 month group; Black color p value means 6 months versus 12 month group. Statistical analysis was performed using one-way ANOVA. Source data are provided as a Source Data file. ## Egr-1 phase recovery via light rescues liver metabolic dysfunction The master clock synchronizes peripheral oscillators by imposing light-regulated rest/activity rhythms and feeding/fasting cycles34. We have demonstrated that both light/dark and feeding/fasting cycles can affect the rhythm stability of liver clock oscillators in an Egr-1-dependent manner16. Thus, we hypothesized that restoring the phase of Egr-1 in old mice to that in young mice could delay liver lipid accumulation. To test this hypothesis, we first restricted the feeding time to the daytime. Although the rhythm of Egr-1 expression was reversed (Supplementary Fig. 8A, B), we did not detect a reduction in liver lipid accumulation at 6 months of age (Supplementary Fig. 8C). Then, we advanced the light time forward 4 h according to the phase shift at the 6th month, in which the peak of Egr-1 expression moved forward from ZT13 to ZT9 (Fig. 7A). We found that the body weights (Fig. 7B) and liver weight/body weight ratios (Fig. 7C) of phase-shift mice were significantly recovered to those of young mice. Examination of the robustness of Egr-1 and Cidea expression at ZT5 showed that the Egr-1 downregulation was attenuated at both the protein and mRNA levels (Fig. 7D, E); in addition, the upregulation of Cidea was attenuated at both the protein and mRNA levels in the transfer group at 6 months of age (Fig. 7D, F). Furthermore, quantitative detection showed that age-related TG accumulation was dramatically decreased in the light transfer group (Fig. 7G). H&E staining and Oil Red O staining also confirmed that light transfer could reverse the age-related lipid droplet phenotype (Fig. 7H). Our results indicate that changing the phase of Egr-1 and then *Cidea via* a light shift can ameliorate metabolic dysfunction in the liver. Fig. 7Egr-1 phase restoration via a light shift is able to rescue liver metabolic dysfunction. A Design of experiments in which the phase of illumination was advanced 4 h. B, C Body weights and ratios of liver weight to body weight in the groups at ZT5 (2 months: $$n = 6$$; 6 months: $$n = 7$$; 6 months Trans: $$n = 5$$ biologically independent animals). D, E mRNA levels of Egr-1 and Cidea in the groups at ZT5 (2 months: $$n = 6$$; 6 months: $$n = 7$$; 6 months Trans: $$n = 5$$ biologically independent animals). F Protein levels of Egr-1 and Cidea in the groups at ZT5. G Tissue TG levels in the groups at ZT5 (2 months: $$n = 7$$; 6 months: $$n = 6$$; 6 months Trans: $$n = 5$$ biologically independent animals). H H&E staining and Oil Red O staining in the groups at ZT5. The data represent the mean ± SEM. Statistical analysis was performed using one-way ANOVA. Source data are provided as a Source Data file. All the above observations indicated that the synchronization between circadian rhythm and lipid metabolism is disrupted in old age. We suggest that age-related Egr-1 alteration acts as a master regulator of both circadian rhythms and metabolic patterns in the liver. To make our point clearer, we draw a schematic for the contribution of Egr-1 rhythm to both circadian rhythms and metabolic patterns with age increased (Fig. 8). At a young age, Egr-1 combines with circadian genes BMAL1/CLOCK to form a complex, then regulating the circadian expression of Cidea to maintain the balance of lipid metabolism. With age increased, Egr-1 rhythm alteration might result in the uncoupling of Egr-1 with both circadian genes BMAL1/CLOCK and lipid metabolic genes Cidea, which results in the decoupling of liver circadian and lipid metabolic disorders in ageing mice. Fig. 8Schematic for the contribution of Egr-1 rhythm to the correlation between circadian rhythms and metabolic patterns with age increased. At a young age, Egr-1 combines with circadian genes BMAL1/CLOCK to form a complex, then regulates the circadian expression of Cidea to maintain the balance of lipid metabolism. With age increased, Egr-1 rhythm alteration might result in uncoupling of Egr-1 with both circadian genes BMAL1/CLOCK and lipid metabolic genes Cidea, facilitating the CD36 expression to promote the fatty acid uptake, accelerating the amino acid uptake to form more fatty acid in hepatocytes, thus, leading to the decoupling of liver circadian and the lipid metabolic disorder in ageing mice. However, how the Egr-1 rhythm responds to the master clock remains to be explored. In the work model, the elements of young and old liver were bought from the website (https://www.dreamstime.com/illust ration-non-alcoholic-fatty-liver-disease-comparison-shows-healthy-diseased-image191689868). ## Discussion With age increasing, both metabolic dysfunction and circadian rhythm shifts often gradually emerge. Increasing evidence has suggested that there is a link between metabolism and the circadian rhythm26,35. However, the underlying mechanism that synchronizes metabolism and the circadian rhythm is still largely unknown. Here, we revealed that Egr-1, a regulator of hepatic clock circuitry, could also regulate lipid metabolism by affecting Cidea expression. Both Egr-1 and Cidea showed a rhythmic expression pattern with different phases. The peak of Egr-1 protein expression appeared at daytime near the day/night transition (ZT9-ZT13), while that of Cidea appeared in the middle of the night (ZT17) in 2-month-old mice. High expression of Cidea at nighttime is beneficial for lipid metabolism since mice eat at night. Notably, a high-fat diet (HFD) at nighttime but not daytime can protect the liver from steatosis34. However, the circadian expression of Egr-1 was found shifted forward with age in the current study. The Egr-1 protein expression peak shifted to ZT9 in the 6th month and ZT5 in the 12th month. This peak shift resulted in a phase shift of Cidea from ZT13 to ZT9 in the 6th and 12th months. Meanwhile, the liver showed metabolic dysfunction as a consequence of TG accumulation and lipid droplet formation increase. In our results, we found that Egr-1 deletion results in liver TG accumulation by enhancing CD36 expression to facilitate fatty acid uptake and Cidea expression to form larger lipid droplets. Both the robustness and rhythmicity of Cidea expression were altered under conditions of Egr-1 deficiency. The regulatory effects of Cidea and other lipid metabolism genes on lipid metabolism during the night were therefore impaired, resulting in liver age-related metabolic dysfunction. We have reported that liver Egr-1, as an early growth response factor, may rapidly mediate central signals to maintain proper oscillation of peripheral clocks16. After activation by master circadian signals, Egr-1 binds to the promoter of the *Per1* gene to activate its transcription, which in turn leads to repression of Per 2 and Rev-erbs. The repression of Rev-Erbs may lead to the robust oscillation of Bmal1. Then, BMAL1 binds to and activates Egr-1 transcription. This Egr-1/Per1/BMAL1/Egr-1 feedback loop may help the liver clock keep pace with the master clock and maintain the robustness of circadian oscillation16. We have also reported that Egr-1 can enhance hepatic gluconeogenesis by indirectly regulating gluconeogenic gene expression via activation of C/EBPα transcription21. Herein, we further found that Egr-1 formed a complex with the circadian genes BMAL1/CLOCK to regulate the transcription of lipid metabolism genes such as Cidea. Our observations suggest that Egr-1 might be an important mediator between circadian and metabolic homeostasis in peripheral organs under normal conditions. However, the synchronization between metabolism and circadian rhythm is disrupted during the aging process. Circadian rhythm alterations occur in addition to decreases in Egr-1 expression, which impair both the robustness and rhythmicity of lipid metabolism genes and circadian genes during aging. Thus, metabolic homeostasis and plasticity in aging peripheral organs, such as the liver, are uncoupled from behavioral and physiological rhythms controlled by the master circadian clock that respond to light and feeding cycles, resulting in progressive metabolic dysfunction with aging. Aging is an irresistible natural process, and the circadian rhythms of older people are normally degraded; both losses of amplitude and fragmentation of output rhythms appear in both master and peripheral circadian rhythms36. We have elucidated that the master clock generates circadian rhythms and drives slave oscillators in various peripheral tissues by secreting cyclical neuroendocrine signals and imposing rest/activity rhythms and feeding/fasting cycles37–43, but we still do not know the exact factors that mediate the master clock and peripheral circadian system or the molecules that respond to these factors in peripheral organs. As we discussed above, Egr-1 can regulate both the robustness and rhythmicity of metabolic genes and circadian genes. Egr-1 can not only be transiently activated by many cytokines, growth factors, and hormones but also respond strongly to nutrient signaling under both fasting and feeding conditions. Fasting induces Egr-1 expression in the liver, which promotes hepatic gluconeogenesis by activating C/EBPα transcription21. After a meal, Egr-1 activity can be induced by insulin in skeletal muscle cells and inhibit insulin receptor phosphorylation, thus reducing insulin sensitivity20. It is reasonable to deduce that Egr-1 might be a critical liver responder that is able to integrate the oscillation of the central circadian clock and energy metabolism in peripheral organs. When the expression level and circadian rhythm of liver Egr-1 are altered during aging, metabolic homeostasis, and circadian plasticity are decoupled from the master circadian rhythm, resulting in metabolic aging in the liver. Sleep disorders are associated with an increased risk of metabolic diseases during aging. Epidemiological studies have revealed that shift workers are more predisposed to elevated TG and high-density lipoprotein (HDL)-cholesterol levels and obesity than day workers44. In mammals, peripheral physiological rhythms are regulated by master circadian clocks, which respond to light and feeding cycles34. The timing of feeding can affect the liver biorhythm and metabolic patterns45. However, feeding time restriction did not ameliorate liver lipid accumulation in 6-month-old mice in the current study. Therefore, we attempted to restore the phase of Egr-1 in old mice to that in young mice via a light shift in order to examine the effect on liver metabolic aging. It has been shown that light duration and wavelength can change the central rhythm and peripheral rhythms. For example, a 6-h shift or delay in light time can cause the phase of Per1 to change in central and liver cells46. Filtering light below 480 nm during irradiation in a 12-h light/dark cycle promotes the expression of circadian rhythm genes in central and peripheral tissues and reduces the secretion of melatonin (a marker of aging)47. When we advanced the light time forward 4 h, liver metabolic aging was mostly improved. The robustness of Egr-1 and Cidea expression at ZT5 was also recovered. These results indicate that metabolic aging can be reversed by adjusting the central circadian pace according to the rhythm shift of Egr-1 at that age. Sato et al. reported that there was no difference in the rhythm of circadian genes in young and old mice2. We also compared the circadian transcriptome between 2 months and 12 months of livers. There are 97 genes overlapped in total 1810 rhythmic expressed genes, which included the core clock genes with a similar rhythm between 2-month-old mice and 12-month-old mice. However, most of CCGs oscillations in 2-month-old mice were altered in 12-month-old mice, which mainly enriched in fatty acid metabolic process, lipid localization, steroid metabolic process, and cellular ketone metabolic process by Gene ontology (GO) analysis. Meanwhile, the oscillated CCGs in 12 months were not rhythmically expressed in 2-month-old mice, which enriched positive regulation of cellular protein localization, protein folding, and homeostasis of the number of cells. Thus, although the rhythms of some of the core clock genes were similar, the rhythmicity of downstream cellular and physiological functions that core clock genes regulated, such as metabolic circadian, have been altered with age increase. In summary, we provide evidence that liver Egr-1 can act as a key responder to the master clock to integrate the central and peripheral rhythms. Liver Egr-1 also functions as an important mediator between circadian and metabolic homeostasis in peripheral organs. When the expression level and circadian rhythm of liver Egr-1 are altered during aging, the liver circadian rhythm is decoupled from the central circadian rhythm, and metabolic homeostasis and circadian plasticity are disrupted, resulting in metabolic aging in the liver. More importantly, we found that metabolic aging can be reversed by adjusting the central circadian rhythm according to the rhythm shift of Egr-1 at that age. ## Ethical statement This research complies with all relevant ethical regulations for the boards/committees and institutions that approved the study protocols. All mice were maintained and used in accordance with the Animal Care and Use Committee of the Model Animal Research Center of Nanjing University, Nanjing, China, using approved protocols from the institutional animal care committee (#CS20). ## Animals *We* generated mice with liver-specific KO (LKO) of Egr-1 by crossing Alb-Cre transgenic mice with homozygous floxed Egr-1 mice. Littermates were used as controls. The KO lines (strain 129) were backcrossed for a minimum of six generations to the C57BL/6 J background (Egr-1-loxp mouse background). The mice were housed in a controlled environment with a 12-hour/12-hour light/dark cycle at 20–24 °C and 50–$65\%$ relative humidity and were fed a chow diet ad libitum. Chow diet for reproduction and maintenance of mice were from Xietong Shengwu, China (Reproduction: SFS9112; Maintenance: SWS9102). All the animals used in the study were male mice at 2 months, 6 months, 12 months, and 21 months of age. To analyze the lipid metabolism in Fig. 3, the liver were dissected at Egr-1 highest(H) zeitgeber time in the male mice of different ages. Liver samples were obtained at ZT13(H) in 2-month group; at ZT9(H) in 6-month group; at ZT5(H) in 12-month group, and 21-month group. To analyze gene rhythm expression, the livers of WT and Egr-1 LKO male mice at matched ages were obtained every 6 h starting at ZT1. To analyze the function of Egr-1 in age-related liver lipid accumulation, the light time for 6-month-old male mice was advanced by 4 h for 1 month according to the time at which the peak of Egr-1 expression moved forward from ZT13 to ZT9. The livers of male mice were dissected at ZT5 to analyze lipid metabolism. For restricted feeding, the WT and Egr-1 LKO male mice for 6-month-old mice were fed exclusively during the daytime for 1 month. The livers of male mice were dissected every 6 h starting at ZT1 to analyze the rhythm. Mouse liver samples were obtained by cervical dislocation after anesthesia with or without light, and the mouse carcasses were treated with centralized pollution-free treatment. In the survival experiment, the animals were euthanized when animals in a state of no anesthesia or sedation, were unable to eat or drink, and stood or extremely reluctantly stood up to 24 h. The animals were monitored once a week and once a day if the symptoms described above were present. Moribund animals were euthanized and every animal found dead or euthanized was necropsied. The criteria for euthanasia were based on an independent assessment by the veterinarian according to the AAALAC guidelines, and the animal was represented as dead in the curve only once its condition was deemed unsuitable for continued survival. Animals in the survival curve (WT group: $$n = 22$$; Egr-1 LKO group: $$n = 23$$) were considered as censored deaths. ## Cell culture, adenovirus infection, and plasmid transfection HEK293T (CRL-3216) cell lines were obtained from American Type Culture Collection (ATCC). Primary hepatocytes were isolated and plated at a cell density of 2 × 106 cells per well. The isolated hepatocytes and HEK293T cells were maintained in DMEM with or without Glucose containing $10\%$ fetal bovine serum (FBS). For adenovirus infection, primary hepatocytes were infected with adenovirus for 48 h. Control GFP- and Egr-1-expressing and dnEgr-1-expressing adenoviruses were constructed using an AdEasy adenoviral vector system (Stratagene, San Diego, CA, USA); then, the adenoviruses were purified by CsCl gradient centrifugation and dialyzed in PBS containing $10\%$ glycerol. The ShCidea adenovirus was provided by Professor Li Peng (Tsinghua University, Beijing, China); the information for this adenovirus can be found in a previous study30. For plasmid transfection, primary hepatocytes were transfected with the indicated plasmids using Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol. The Cidea plasmid was provided by Professor Li Peng (Tsinghua University, Beijing, China); the information for this plasmid can be found in a previous study30. HEK293T cells were transfected with Egr-1, BMAL1, and CLOCK plasmids or with a control PCDNA3 plasmid. ## Isolation of mouse primary hepatocytes WT and Egr-1 LKO mice were anesthetized via intraperitoneal injection of sodium pentobarbital (50 mg/kg). Before isolation, the perfusion buffer (#17701-038, GIBCO) with collagenase IV (C5138, Sigma) should be dissolved for 30–60 min at a temperature of 37 °C. The mouse were cleaned with $70\%$ alcohol and made a midline incision to expose the abdominal site. Then, place the catheter into the portal vein. Start the perfusion pump at a rate of 4.5 ml/min with perfusion medium, and immediately cut SVC or IVC to let blood out for about 5 min. The liver was digested by a digest medium for 6–7 min. The liver was removed, and placed in a 100 mm plate filled with a cold 20 ml washing medium. The cells were filtered through 100 µm filters into 50 ml centrifuge tubes and centrifuged at 50×g for 5 min at 4 °C. A pellet containing primary hepatocytes was resuspended in DMEM with glucose containing $10\%$ FBS and then recentrifuged. The final pellet was resuspended in DMEM with glucose containing $10\%$ FBS and plated at a cell density of 2 × 106 cells per six-well. Isolated primary hepatocytes were used immediately and cells can survive for ~4 days. ## mRNA and protein expression analysis and immunoprecipitation Total RNA was extracted using TRIzol (Takara Bio, Kusatsu, Japan) according to the manufacturer’s protocol. The primer sequences utilized in this study are provided in Supplementary Table 1. Q-PCR data were acquired using the Applied Biosystems Viia 7 fast Real-Time PCR system software. For protein analysis, 2 × 106 cells or 20 mg of homogenized tissue was lysed directly in RIPA lysis buffer containing protease inhibitors (cell/tissue weight:RIPA lysis buffer volume, 1:20). Protein (50 µg per sample) was loaded onto gels. Equal amounts of protein were analyzed via western blotting with antibodies against Egr-1 (sc-101033, Santa Cruz Biotechnology, 1:500), Cidea (ab8402, Abcam, 1:1000), CD36 (sc-7309, Santa Cruz Biotechnology, 1:1000), Clock (18094-1-AP, Proteintech, 1:1000), c-Myc (10828-1-AP, Proteintech, 1:1000), β-actin (66009-1-Ig, Proteintech, 1:2000) and a-tubulin (66031-1-Ig, Proteintech, 1:2000). Second antibodies used were: anti-rabbit-HRP (BA1054, Boster, 1:10000) and anti-mouse-HRP(BA1050, Boster, 1:10000). Western blot data were acquired digitally by Image LAB(Bio-Red) software. The quantitative analysis of western blotting bands was observed by ImageJ software $\frac{2.1.0}{1.53}$c. Immunoprecipitation was performed according to a standard protocol. Briefly, antibodies against Myc/IgG were added to form immune complexes with the indicated proteins in the lysates, and these complexes were immunoprecipitated using protein-A/G agarose beads (Santa Cruz Biotechnology). After several washes, the samples were boiled in 2× sample buffer and subjected to western blot analysis with Myc/Clock/Egr-1 antibodies. ## Luciferase assays Primary hepatocytes were seeded into twelve-well dishes at a cell density of 0.8 × 106 cells per well 24 h before transfection. Egr-1, Bmal1, and Clock overexpression constructs, mCidea reporter and mutant plasmids, and internal control Renilla-luciferase plasmids were transfected. The relative luciferase activity levels were determined 48 h following transfection using a Dual-Luciferase Assay System (Promega, Madison, Wisconsin, USA) according to the manufacturer’s protocol. All transfection experiments were performed in triplicate. ## H&E, Sirius Red, and β-gal staining For histological analysis, liver tissue was collected from mice of different ages and fixed in $4\%$ paraformaldehyde (PFA) overnight. The sections were used for H&E staining according to a standard protocol. For nuclear staining, the sections were placed in hematoxylin solution for 1 min and rinsed with distilled water. For Sirius Red staining, paraffin sections were stained by using a Sirius Red staining Kit (Sbjbio, Nanjing, China) according to the manufacturer’s protocol. For β-gal staining, liver tissue was collected from mice of different ages and fixed in $4\%$ PFA overnight. The sections were stained by using a β-gal staining kit (Beyotime, Shanghai, China) according to the manufacturer’s protocol. ## Oil Red O tissue and hepatocyte staining For tissue, livers were collected from mice of different ages and fixed in $4\%$ PFA for 2 h. Then, the samples were placed in $30\%$ sucrose solution at 4 °C overnight to remove the internal moisture, embedded in Tissue Freezing Medium (Leica, UK), and sectioned at 15-μm thickness using a Leica CM1900 Cryostat. The sections were frozen at −70 °C until staining, at which time they were air-dried, rinsed with $60\%$ isopropanol, and stained with freshly prepared Oil Red O working solution ($0.5\%$ Oil Red O:ddH2O = 3:2) for 15 min and rinsed with $60\%$ isopropanol. For hepatocytes, 106 cells were seeded in a 35 mm dish and treated with different experimental compounds and vehicles. The hepatocytes were fixed in $4\%$ PFA for 15 min and then stained with freshly prepared Oil Red O working solution for the appropriate time. ## Hepatocyte BODIPY staining and immunofluorescence staining A suitable number of cells were seeded in a 24-well plate with coverslips. The hepatocytes were fixed in $4\%$ PFA for 20 min, incubated with 1 mg/ml BODIPY dye solution at 37 °C for 15 min, and then sealed with $50\%$ glycerol. For immunofluorescence staining, paraffin sections were deparaffinized, rehydrated, and boiled in citrate buffer (pH 6.0) to retrieve antigens. Afterward, the paraffin sections and frozen sections were permeabilized, blocked, and incubated with the indicated primary antibodies at 4 °C overnight. Subsequently, the sections were incubated with secondary antibodies for 1 h at room temperature. Immunofluorescence microscopy images were acquired using an Olympus SpinSR microscope. Microscopy image analysis was performed using ImageJ software version $\frac{2.1.0}{1.53}$c. ## Metabolic parameters Blood samples were drawn from the retroorbital plexus. After 30 min at room temperature, the blood samples were centrifuged for 10 min at 3000×g to obtain plasma, and the supernatants were collected and stored at −80 °C until analysis. Blood TG, ALT, and AST levels were determined by an automatic biochemical analyzer at Nanjing Drum Tower Hospital, the affiliated hospital of Nanjing University Medical School. The serum-free fatty acid levels were enzymatically detected with a LabAssayTM NEFA (Wako, Japan). Liver tissue TG levels were detected using a tissue/cell TG enzymatic assay kit (Applygen, Beijing, China). Hepatocyte TG levels were detected using a TG assay kit (Jiancheng, Nanjing, China). Liver-free fatty acid levels were detected using a Free Fatty Acid Quantification Colorimetric/Fluorometric Kit (BioVison, Palo Alto, USA). ## ChIP Primary hepatocytes were infected with an Egr-1-expressing adenovirus for 48 h. Chromatin lysates were prepared, precleared with Protein-A/G agarose beads, and immunoprecipitated with antibodies against Egr-1 or control mouse IgG. The beads were extensively washed before reverse-crosslinking. DNA was purified using a PCR purification kit (Qiagen) and subsequently analyzed by real-time PCR. ## RNA extraction, library preparation, and sequencing The livers of C57BL/6 J mice at 2 months, 6 months, and 12 months of age were sacrificed every 4 h starting at ZT1 over the circadian cycle. The livers of WT and Egr-1 LKO mice were selected at zeitgeber time of Egr-1 highest or lowest protein expression in 2 months (H: ZT13; L: ZT17), 6 months (H: ZT9; L: ZT21), and 12 months (H: ZT5; L:ZT21) group. All samples were submitted to Novogene for transcriptome sequencing and analysis. Total RNA was extracted according to the manufacturer’s procedure. The RNA quantity and purity were determined using a NanoPhotometer spectrophotometer and a Qubit2.0 Fluorometer. mRNAs were enriched with poly-T oligo-attached magnetic beads using NEBNext First Strand Synthesis Reaction Buffer (5×) and then fragmented into small pieces. These cleaved RNA fragments were reverse-transcribed to create a cDNA library using Illumina’s NEBNext® UltraTM RNA Library Prep Kit; the average insert size for the paired-end libraries was 300 bp (±50 bp). The library was sequenced on an Illumina HiSeq platform. ## RNA-seq data analysis After sequencing and removal of low-quality reads that contained adapter contamination, low-quality bases, and undetermined bases, the sequenced reads were aligned to the mouse genome using HISAT2. Feature Counts was used to determine the mRNA expression levels by calculating the FPKM values. The expression patterns of all the genes were characterized with the TCseq package. RNA-seq data were analyzed using edgeR software version 3.38.1. ## Differential gene expression analysis The differentially expressed genes (DEGs) were identified with edgeR software. We chose the DEGs for which the absolute value of the log (fold change) was >1.2 and the corresponding false discovery rate (FDR) was <0.05. GO annotation and enrichment of the DEGs were performed with the clusterProfiler package. ## Statistical analysis The data are presented as the mean ± SEM from at least three independent experiments. Statistical analyses were performed using Prism 8 software (GraphPad Software, Inc., San Diego, CA, USA). The data were assessed for normal distributions with the Shapiro-Wilk test. If the normality test was passed, statistical analyses were performed using two-tailed unpaired Student’s t-test (two-group comparison) or one-way ANOVA followed by multiple comparisons with the LSD post hoc test (more than two groups). $P \leq 0.05$ was considered to indicate statistical significance. ## Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. ## Supplementary information Supplementary information Peer Review File Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36775-8. ## Source data Source Data ## Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. ## References 1. Cartee GD, Hepple RT, Bamman MM, Zierath JR. **Exercise promotes healthy aging of skeletal muscle**. *Cell Metab.* (2016.0) **23** 1034-1047. DOI: 10.1016/j.cmet.2016.05.007 2. Sato S. **Circadian reprogramming in the liver identifies metabolic pathways of aging**. *Cell* (2017.0) **170** 664-677.e611. DOI: 10.1016/j.cell.2017.07.042 3. Gong Z, Tas E, Yakar S, Muzumdar R. **Hepatic lipid metabolism and non-alcoholic fatty liver disease in aging**. *Mol. Cell Endocrinol.* (2017.0) **455** 115-130. DOI: 10.1016/j.mce.2016.12.022 4. Van Schaftingen E, Lederer B, Bartrons R, Hers HG. **A kinetic study of pyrophosphate: fructose-6-phosphate phosphotransferase from potato tubers. Application to a microassay of fructose 2,6-bisphosphate**. *Eur. J. Biochem.* (1982.0) **129** 191-195. DOI: 10.1111/j.1432-1033.1982.tb07039.x 5. Anantharaju A, Feller A, Chedid A. **Aging liver. A review**. *Gerontology* (2002.0) **48** 343-353. DOI: 10.1159/000065506 6. Hood S, Amir S. **The aging clock: circadian rhythms and later life**. *J. Clin. Invest.* (2017.0) **127** 437-446. DOI: 10.1172/JCI90328 7. Yamazaki S. **Effects of aging on central and peripheral mammalian clocks**. *Proc. Natl Acad. Sci. USA* (2002.0) **99** 10801-10806. DOI: 10.1073/pnas.152318499 8. Luo W. **Old flies have a robust central oscillator but weaker behavioral rhythms that can be improved by genetic and environmental manipulations**. *Aging Cell* (2012.0) **11** 428-438. DOI: 10.1111/j.1474-9726.2012.00800.x 9. Panda S. **Coordinated transcription of key pathways in the mouse by the circadian clock**. *Cell* (2002.0) **109** 307-320. DOI: 10.1016/S0092-8674(02)00722-5 10. Guan D, Lazar MA. **Interconnections between circadian clocks and metabolism**. *J. Clin. Invest.* (2021.0) **131** e148278. DOI: 10.1172/JCI148278 11. Froy O. **Circadian rhythms, aging, and life span in mammals**. *Physiology* (2011.0) **26** 225-235. DOI: 10.1152/physiol.00012.2011 12. Silverman ES, Collins T. **Pathways of Egr-1-mediated gene transcription in vascular biology**. *Am. J. Pathol.* (1999.0) **154** 665-670. DOI: 10.1016/S0002-9440(10)65312-6 13. Hamilton TB, Borel F, Romaniuk PJ. **Comparison of the DNA binding characteristics of the related zinc finger proteins WT1 and EGR1**. *Biochemistry* (1998.0) **37** 2051-2058. DOI: 10.1021/bi9717993 14. Chavrier P. **A gene encoding a protein with zinc fingers is activated during G0/G1 transition in cultured cells**. *EMBO J.* (1988.0) **7** 29-35. DOI: 10.1002/j.1460-2075.1988.tb02780.x 15. Huang RP, Liu C, Fan Y, Mercola D, Adamson ED. **Egr-1 negatively regulates human tumor cell growth via the DNA-binding domain**. *Cancer Res.* (1995.0) **55** 5054-5062. PMID: 7585551 16. Tao W. **EGR1 regulates hepatic clock gene amplitude by activating Per1 transcription**. *Sci. Rep.* (2015.0) **5** 15212. DOI: 10.1038/srep15212 17. Meyyappan M, Wheaton K, Riabowol KT. **Decreased expression and activity of the immediate-early growth response (Egr-1) gene product during cellular senescence**. *J. Cell Physiol.* (1999.0) **179** 29-39. DOI: 10.1002/(SICI)1097-4652(199904)179:1<29::AID-JCP4>3.0.CO;2-D 18. Penner MR. **Age-related changes in Egr1 transcription and DNA methylation within the hippocampus**. *Hippocampus* (2016.0) **26** 1008-1020. DOI: 10.1002/hipo.22583 19. Bochkis IM, Przybylski D, Chen J, Regev A. **Changes in nucleosome occupancy associated with metabolic alterations in aged mammalian liver**. *Cell Rep.* (2014.0) **9** 996-1006. DOI: 10.1016/j.celrep.2014.09.048 20. Wu J. **Early growth response-1 negative feedback regulates skeletal muscle postprandial insulin sensitivity via activating Ptp1b transcription**. *FASEB J.* (2018.0) **32** 4370-4379. DOI: 10.1096/fj.201701340R 21. Shen N. **The constitutive activation of Egr-1/C/EBPa mediates the development of type 2 diabetes mellitus by enhancing hepatic gluconeogenesis**. *Am. J. Pathol.* (2015.0) **185** 513-523. DOI: 10.1016/j.ajpath.2014.10.016 22. Chakrabarti P. **Insulin inhibits lipolysis in adipocytes via the evolutionarily conserved mTORC1-Egr1-ATGL-mediated pathway**. *Mol. Cell Biol.* (2013.0) **33** 3659-3666. DOI: 10.1128/MCB.01584-12 23. Gokey NG, Lopez-Anido C, Gillian-Daniel AL, Svaren J. **Early growth response 1 (Egr1) regulates cholesterol biosynthetic gene expression**. *J. Biol. Chem.* (2011.0) **286** 29501-29510. DOI: 10.1074/jbc.M111.263509 24. Hughes ME, Hogenesch JB, Kornacker K. **JTK_CYCLE: an efficient nonparametric algorithm for detecting rhythmic components in genome-scale data sets**. *J. Biol. Rhythms* (2010.0) **25** 372-380. DOI: 10.1177/0748730410379711 25. Kim K. **RORalpha controls hepatic lipid homeostasis via negative regulation of PPARgamma transcriptional network**. *Nat. Commun.* (2017.0) **8** 162. DOI: 10.1038/s41467-017-00215-1 26. Rudic RD. **BMAL1 and CLOCK, two essential components of the circadian clock, are involved in glucose homeostasis**. *PLoS Biol.* (2004.0) **2** e377. DOI: 10.1371/journal.pbio.0020377 27. 27.Shigeki, S. et al. Deficient of a Clock Gene Brain and Muscle Arnt-Like Protein-1 (BMAL1) Induces Dyslipidemia and Ectopic Fat Formation. PLoS ONE.6, e25231 (2011). 28. Jakobsen JS. **Temporal mapping of CEBPA and CEBPB binding during liver regeneration reveals dynamic occupancy and specific regulatory codes for homeostatic and cell cycle gene batteries**. *Genome Res.* (2013.0) **23** 592-603. DOI: 10.1101/gr.146399.112 29. Yu X. **Egr-1 decreases adipocyte insulin sensitivity by tilting PI3K/Akt and MAPK signal balance in mice**. *EMBO J.* (2011.0) **30** 3754-3765. DOI: 10.1038/emboj.2011.277 30. Zhou L. **Cidea promotes hepatic steatosis by sensing dietary fatty acids**. *Hepatology* (2012.0) **56** 95-107. DOI: 10.1002/hep.25611 31. Xu W. **Differential roles of cell death-inducing DNA fragmentation factor-alpha-like effector (CIDE) proteins in promoting lipid droplet fusion and growth in subpopulations of hepatocytes**. *J. Biol. Chem.* (2016.0) **291** 4282-4293. DOI: 10.1074/jbc.M115.701094 32. Zhou Z. **Cidea-deficient mice have lean phenotype and are resistant to obesity**. *Nat. Genet.* (2003.0) **35** 49-56. DOI: 10.1038/ng1225 33. Roden M. **Mechanisms of disease: hepatic steatosis in type 2 diabetes–pathogenesis and clinical relevance**. *Nat. Clin. Pr. Endocrinol. Metab.* (2006.0) **2** 335-348. DOI: 10.1038/ncpendmet0190 34. Schibler U, Ripperger J, Brown SA. **Peripheral circadian oscillators in mammals: time and food**. *J. Biol. Rhythms* (2003.0) **18** 250-260. DOI: 10.1177/0748730403018003007 35. Liu J. **CLOCK and BMAL1 regulate muscle insulin sensitivity via SIRT1 in male mice**. *Endocrinology* (2016.0) **157** 2259-2269. DOI: 10.1210/en.2015-2027 36. Hofman MASD. **Living by the clock: the circadian pacemaker in older people**. *Ageing Res. Rev.* (2006.0) **5** 33-51. DOI: 10.1016/j.arr.2005.07.001 37. Maury E, Ramsey KM, Bass J. **Circadian rhythms and metabolic syndrome: from experimental genetics to human disease**. *Circ. Res.* (2010.0) **106** 447-462. DOI: 10.1161/CIRCRESAHA.109.208355 38. Wijnen H, Young MW. **Interplay of circadian clocks and metabolic rhythms**. *Annu. Rev. Genet.* (2006.0) **40** 409-448. DOI: 10.1146/annurev.genet.40.110405.090603 39. Liu Z. **SIRT7 couples light-driven body temperature cues to hepatic circadian phase coherence and gluconeogenesis**. *Nat. Metab.* (2019.0) **1** 1141-1156. DOI: 10.1038/s42255-019-0136-6 40. Wijnen H, Young MW. **Interplay of circadian clocks and metabolic rhythms**. *Annu. Rev. Genet.* (2006.0) **40** 409-448. DOI: 10.1146/annurev.genet.40.110405.090603 41. Korenčič A, Košir R, Bordyugov G, Lehmann R, Rozman D, Herzel H. **Timing of circadian genes in mammalian tissues**. *Sci. Rep.* (2014.0) **4** 5782. DOI: 10.1038/srep05782 42. Maury E, Ramsey KM, Bass J. **Circadian rhythms and metabolic syndrome: from experimental genetics to human disease**. *Circ. Res.* (2010.0) **106** 447-462. DOI: 10.1161/CIRCRESAHA.109.208355 43. Schibler U, Ripperger J, Brown SA. **Peripheral circadian oscillators in mammals: time and food**. *J. Biol. Rhythms* (2003.0) **18** 250-260. DOI: 10.1177/0748730403018003007 44. Karlsson B, Knutsson A, Lindahl B. **Is there an association between shift work and having a metabolic syndrome? Results from a population based study of 27,485 people**. *Occup. Environ. Med.* (2001.0) **58** 747-752. DOI: 10.1136/oem.58.11.747 45. Chaix A, Lin T, Le HD, Chang MW, Panda S. **Time-restricted feeding prevents obesity and metabolic syndrome in mice lacking a circadian clock**. *Cell Metab.* (2019.0) **29** 303-319.e304. DOI: 10.1016/j.cmet.2018.08.004 46. Davidson AJ, Yamazaki S, Arble DM, Menaker M, Block GD. **Resetting of central and peripheral circadian oscillators in aged rats**. *Neurobiol. Aging* (2008.0) **29** 471-477. DOI: 10.1016/j.neurobiolaging.2006.10.018 47. Casper RF, Rahman S. **Spectral modulation of light wavelengths using optical filters: effect on melatonin secretion**. *Fertil. Steril.* (2014.0) **102** 336-338. DOI: 10.1016/j.fertnstert.2014.06.006
--- title: Calpain activity is negatively regulated by a KCTD7–Cullin-3 complex via non-degradative ubiquitination authors: - Jaiprakash Sharma - Shalaka Mulherkar - Uan-I Chen - Yan Xiong - Lakshya Bajaj - Byoung-Kyu Cho - Young Ah Goo - Hon-Chiu Eastwood Leung - Kimberley F. Tolias - Marco Sardiello journal: Cell Discovery year: 2023 pmcid: PMC10038992 doi: 10.1038/s41421-023-00533-3 license: CC BY 4.0 --- # Calpain activity is negatively regulated by a KCTD7–Cullin-3 complex via non-degradative ubiquitination ## Abstract Calpains are a class of non-lysosomal cysteine proteases that exert their regulatory functions via limited proteolysis of their substrates. Similar to the lysosomal and proteasomal systems, calpain dysregulation is implicated in the pathogenesis of neurodegenerative disease and cancer. Despite intensive efforts placed on the identification of mechanisms that regulate calpains, however, calpain protein modifications that regulate calpain activity are incompletely understood. Here we show that calpains are regulated by KCTD7, a cytosolic protein of previously uncharacterized function whose pathogenic mutations result in epilepsy, progressive ataxia, and severe neurocognitive deterioration. We show that KCTD7 works in complex with Cullin-3 and Rbx1 to execute atypical, non-degradative ubiquitination of calpains at specific sites (K398 of calpain 1, and K280 and K674 of calpain 2). Experiments based on single-lysine mutants of ubiquitin determined that KCTD7 mediates ubiquitination of calpain 1 via K6-, K27-, K29-, and K63-linked chains, whereas it uses K6-mediated ubiquitination to modify calpain 2. Loss of KCTD7-mediated ubiquitination of calpains led to calpain hyperactivation, aberrant cleavage of downstream targets, and caspase-3 activation. CRISPR/Cas9-mediated knockout of Kctd7 in mice phenotypically recapitulated human KCTD7 deficiency and resulted in calpain hyperactivation, behavioral impairments, and neurodegeneration. These phenotypes were largely prevented by pharmacological inhibition of calpains, thus demonstrating a major role of calpain dysregulation in KCTD7-associated disease. Finally, we determined that Cullin-3–KCTD7 mediates ubiquitination of all ubiquitous calpains. These results unveil a novel mechanism and potential target to restrain calpain activity in human disease and shed light on the molecular pathogenesis of KCTD7-associated disease. ## Introduction Calpains are a unique class of cytosolic calcium-dependent proteases that regulate processes as different as cell proliferation, differentiation and migration, apoptosis, and membrane fusion by targeting specific protein substrates1–8. Loss-of-function mutations in several calpain genes cause human disease9–14. Notably, calpain hyperactivity also plays a pathogenic role in metabolic and degenerative diseases such as type-2 diabetes, Duchenne muscular dystrophy, and Parkinson’s and Alzheimer’s diseases via improper processing of key regulatory proteins2,8,15–18. Owing to the involvement of calpain dysregulation in human disease, intensive effort has been focused on the identification of mechanisms that regulate calpain activity. A well-studied modulator of calpain activity is cytosolic calcium19,20. The levels of calcium required to maximally activate calpains, however, do not exist within living cells (except under certain pathological contexts)2,6,21, indicating that other pathways regulate calpain activity in normal conditions. In fact, autolysis, phosphorylation, and binding of phospholipids have also been proposed to regulate calpains2,22. In addition, while calpains are primarily cytosolic, their dynamic association with the cytoskeleton, cell membrane, components of the secretory pathway, and nuclear membrane suggests tight spatiotemporal regulation23–25. KCTD7 (potassium channel tetramerization domain containing 7) is a protein with unknown function that is defective in progressive myoclonic epilepsy-3 (EPM3), neuronal ceroid lipofuscinosis 14 (CLN14), and opsoclonus-myoclonus syndrome (OMS)26–28. Pathogenic KCTD7 mutations cause epilepsy, progressive ataxia, and severe neurocognitive deterioration27–29. The molecular function of KCTD7 has remained uncharacterized, as has the pathogenic mechanism linking KCTD7 loss of function to disease30,31. KCTD7 has been shown to interact with the Cullin-3 ubiquitin ligase complex27,32,33. Other members of the KCTD protein family have also been shown to interact with cullin-ring-ligases (CRLs) and function as adapters for certain substrates34–39. The interaction between KCTD and cullin proteins is mediated by the Bric-a-brack, Tram-track, Broad complex (BTB) domain, a relatively conserved N-terminal domain of KCTD proteins which facilitates homo- or heterodimerization and protein–protein interaction39–41. By using proteomics and cell biology approaches, here we show that the KCTD7–Cullin-3 ubiquitin ligase regulates calpain activity by non-degradative ubiquitination at specific protein sites. Loss of calpain ubiquitination leads to calpain hyperactivation, aberrant cleavage of downstream targets, and cell death associated with caspase-3 activation. The analysis of Kctd7 knock-out mice shows that loss of KCTD7 leads to higher calpain activity and results in neurodegeneration and behavioral impairment, which are prevented by calpain inhibition. Molecular analyses also show that KCTD7-mediated non-degradative ubiquitination is a common feature shared by the calpain protein family. These results identify an unanticipated regulatory mechanism of calpains and clarify the molecular pathogenesis of loss of KCTD7 function as mediated by calpain hyperactivity. ## KCTD7 is a component of the Cullin-3 E3 ligase complex and interacts with calpains To identify candidate KCTD7 targets, we performed tandem affinity purification (TAP) followed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) in HEK293 cell line stably expressing human KCTD7 fused to streptavidin- and calmodulin-binding peptides separated by an HA tag (SBP-HA-CBP-KCTD7) (Supplementary Fig. S1a–d). LC-MS/MS analysis of purified proteins showed that KCTD7 interacts with Cullin-3 as previously reported27,32 and also identified several additional interactors (Supplementary Table S1). Immunoprecipitation (IP) of Cullin-3-myc followed by immunoblotting of KCTD7-Flag confirmed the interaction between the two proteins, which was significantly decreased by the introduction of KCTD7 disease-causing mutations (Fig. 1a, b). We also tested KCTD7 deletion constructs for their ability to interact with full-length Cullin-3. The results showed that truncated KCTD7 constructs (N-terminal: amino acids 1–149, BTB-domain only: amino acids 51–149, or C-terminal end: amino acids 150–289) were unable to bind to Cullin-3, indicating that full-length KCTD7 is required for this interaction (Fig. 1c). Cullin-3-associated RING-H2 finger protein Rbx1 (RING Box Protein-1) is the recruiter of ubiquitin-conjugating enzymes (E2s) that catalyzes the transfer of ubiquitin onto substrates42–45. Co-immunoprecipitation (Co-IP) experiments confirmed the formation of a Cullin-3–Rbx1–KCTD7 complex (Fig. 1d). Among the proteins that co-purified with KCTD7 were CAPN2 (calpain 2) and CAPNS1, the catalytic and regulatory subunits of m-calpain, respectively. CAPNS1 is also part of μ-calpain, a heterodimer of CAPN1 (calpain 1) and CAPNS12. CAPNS1 is required for the stability of both calpain catalytic subunits in vivo, and probably functions as an intramolecular chaperone46,47. Reciprocal co-IP experiments confirmed the interaction between KCTD7 and calpain subunits (Fig. 1e). To identify the protein domains that mediate calpain–KCTD7 interaction, we first performed co-IP experiments using full-length calpain 1 with KCTD7 constructs encoding the full-length protein, its N-terminal domain, or its C-terminal domain. The results showed that full-length KCTD7 is required for its interaction with calpain 1 (Supplementary Fig. S1e). We then co-expressed full-length KCTD7 with constructs carrying combinations of the four calpain 1 domains (domain I to IV) and performed co-IP. The results showed that domain III of calpain 1 is necessary and sufficient for calpain 1 interaction with KCTD7 (Supplementary Fig. S1f). Confocal microscopic analysis of HeLa cells that co-express KCTD7 with either calpain subunits showed that both proteins broadly co-distribute with KCTD7 in the cytosol (Fig. 1f). Similarly, subcellular fractionation of HeLa cells by differential velocity centrifugation showed that endogenous KCTD7 and calpain subunits are all highly enriched in the cytosol (Supplementary Fig. S1g).Fig. 1KCTD7 is a component of the Cullin-3 E3 ligase complex and interacts with calpains.a Co-IP analysis of Flag-tagged KCTD7 and myc-tagged Cullin-3 in HEK293 cells. b Co-IP analysis of KCTD7 mutant constructs and Cullin-3-Myc in HEK293 cells. c Co-IP analysis of KCTD7 deletion constructs and Cullin-3-Myc in HEK293 cells. d Co-IP analyses of Flag-tagged KCTD7, myc-tagged Cullin-3, and HA-tagged Rbx1 in HEK293 cells. e Co-IP analyses of GFP-tagged KCTD7 and Flag-tagged calpain 1, calpain 2, and CAPNS1 in HEK293 cells. f Confocal microscopy of HeLa cells showing KCTD7-GFP co-distribute with calpain 1-Flag, calpain 2-Flag, and CAPNS1-Flag. Trace outline is used for line-scan (white dashed line) analysis of relative fluorescence intensity of indicated protein signals. Scale bar, 10 μm. g Co-IP analyses of myc-tagged Cullin-3 with Flag-tagged calpain 1, calpain 2 and CAPNS1 in HEK293 cells. h Co-IP analyses of Flag-tagged KCTD7 mutant constructs with endogenous calpain 1, calpain 2, and CAPNS1 in HEK293 cells. Co-IP also revealed interactions between Cullin-3 and calpain subunits (Fig. 1g). Calpain–KCTD7 interactions were hindered by certain disease-associated KCTD7 point mutations (Fig. 1h). Finally, knock-down of calpain subunits did not affect KCTD7–Cullin-3 interaction (Supplementary Fig. S1h–j). Together, these data indicate that KCTD7 is a component of the Cullin-3 E3 ligase complex, where it functions as an adapter for calpain subunits. ## KCTD7–Cullin-3 ubiquitin ligase regulates calpain activity To test whether calpains are targeted by the KCTD7–Cullin-3 complex for proteasomal degradation, we first checked whether modulation of the levels of the KCTD7–Cullin-3 complex leads to changes in proteins levels of calpain subunits. Transient expression of Cullin-3 and KCTD7 either alone or in combination in HEK293 cells did not lead to any changes in the levels of calpain subunits (Fig. 2a, b), nor did siRNA- or shRNA-mediated knockdown of endogenous KCTD7 or CUL3 (Fig. 2c–j and Supplementary Fig. S1k–l). Similarly, CRISPR-mediated knock-out of KCTD7 in HeLa cells did not result in the accumulation of any calpain subunits (Supplementary Fig. S1m, n).Fig. 2KCTD7–Cullin-3 ubiquitin ligase regulates calpain activity.a, b Cullin-3 and KCTD7 were expressed in HEK293 cells alone or in combination for 36 h. Lysates were probed with antibodies as indicated, $$n = 4$.$ c–f KCTD7 was knocked down using shRNAs (c, d) or siRNA (e, f) in HEK293 cells for 48 h. Cell lysates were probed with antibodies for calpain subunit and KCTD7 as indicated. GAPDH served as a loading control, $$n = 4$.$ g–j Cullin-3 was knocked down by shRNAs (g, h) or siRNA (i, j) in HEK293 cells for 48 h. Cell lysates were probed with antibodies for calpain subunit and Cullin-3 as indicated. GAPDH served as a loading control, $$n = 4$.$ k, l Calpain 1 was expressed alone or in combination with KCTD7 in HEK293 cells and α-spectrin cleavage was evaluated by immunoblotting, $$n = 5$.$ m, n, HEK293 cells were treated with ionomycin/CaCl2 alone or in combination with KCTD7 expression and α-spectrin cleavage was evaluated by immunoblotting, $$n = 3$.$ Data represent means ± SEM; ∗$P \leq 0.05$, ∗∗$P \leq 0.01$, ns not significant. We therefore sought to determine whether KCTD7–Cullin-3 affects the activity of calpains. Overexpression of calpain 1 resulted in greater rates of cleavage of the cytoskeletal protein α-spectrin, a known substrate of calpain 1 and calpain 2; this increase was reduced by co-expressing KCTD7 (Fig. 2k, l). Expression of KCTD7 also reduced the activity of endogenous calpains induced by ionomycin and CaCl2 treatment (Fig. 2m, n and Supplementary Fig. S1o, p). These data indicate that the KCTD7–Cullin-3 complex regulates the activities of calpain 1 and calpain 2 without affecting their protein levels. ## KCTD7 regulates calpain activity via atypical ubiquitination An in vivo ubiquitination assay showed that calpain 1, calpain 2, and CAPNS1 all undergo ubiquitination (Fig. 3a). Inhibition of proteasomal degradation by MG132, however, did not lead to the accumulation of any of them (Supplementary Fig. S2a–d) nor resulted in their increased ubiquitination (Supplementary Fig. S2e–g), thereby suggesting that these ubiquitination events are non-proteolytic in nature. Fig. 3KCTD7–Cullin-3 ubiquitin ligase regulates calpain activity via atypical ubiquitination.a In vivo ubiquitination assay for calpain 1, calpain 2, and CAPNS1 performed by coexpressing either protein with HA-Ub in HEK293 cells for 36 h. b In vivo ubiquitination assay in HEK293 cells for calpain 1, calpain 2, and CAPNS1 using the indicated ubiquitin mutants. HA-Ub K0 has all lysine residues mutated. c In vivo ubiquitination assay for calpain 1, calpain 2, and CAPNS1 performed by expressing Cullin-3 and KCTD7 alone or in combination in HEK293 cells for 36 h. d In vivo ubiquitination assay for calpain 1, calpain 2, and CAPNS1 performed upon simultaneous expression of KCTD7 and knock-down of Cullin-3 in HEK293 cells for 48 h. e In vivo ubiquitination assay in HEK293 cells for calpain 1, calpain 2, and CAPNS1 performed by expressing the indicated ubiquitin mutants alone or in combination with KCTD7. We next investigated which type of ubiquitin-chain linkage is preferred for modification of calpain 1, calpain 2, and CAPNS1. The ubiquitin molecule contains seven lysine residues (K6, K11, K27, K29, K33, K48, and K63), any of which could mediate ubiquitin chain elongation (polyubiquitination). We executed an in vivo ubiquitination assay by expressing calpain subunits with either wild-type HA-ubiquitin or ubiquitin mutants in which all but one of their seven lysine residues were substituted with arginine. The results showed that calpain 1 is modified using most of the ubiquitin lysine residues, calpain 2 is preferably modified via K27 and K29, and CAPNS1 is preferably modified via K6, K27, K29, and K63 (Fig. 3b). A time-course analysis upon treatment with the calpain inducers, ionomycin and CaCl2, showed differential ubiquitination dynamics, indicating activity-dependent changes in ubiquitination of calpain subunits (Supplementary Fig. S2h). Additional testing showed that KCTD7 overexpression dramatically increased ubiquitination of calpain 1, calpain 2, and CAPNS1, and co-expression of Cullin-3 further increased ubiquitination of these substrates (Fig. 3c). In vitro ubiquitination assays confirmed that the Cullin-3–Rbx1–KCTD7 complex mediates ubiquitination of calpain subunits (Supplementary Fig. S2i). Knock-down of Cullin-3 reduced KCTD7-mediated ubiquitination of calpain 1, calpain 2, and CAPNS1, demonstrating that Cullin-3 is required for ubiquitination of these proteins (Fig. 3d). Using the ubiquitin mutants, we showed that KCTD7 mediates ubiquitination of calpain 1 and CAPNS1 via the same lysine residues established above (K6, K27, K29 and K63; Fig. 3e), whereas it uses K6-mediated ubiquitination to modify calpain 2 (Fig. 3e). Finally, several disease-causing mutants of KCTD7 were unable to ubiquitinate calpain 1 (Fig. 4a, b).Fig. 4Calpain mutant resistant to KCTD7-mediated ubiquitination is hyperactive.a In vivo ubiquitination assay for calpain1 performed by expressing WT KCTD7 or pathogenic mutants in HEK293 cells for 36 h. b In vitro ubiquitination assay for calpain 1. c In vivo ubiquitination assay in HEK293 cells for calpain 1 K–R mutants. d In vivo ubiquitination assay for calpain 1 K–R mutants performed with or without KCTD7 in HEK293 cells. e, f Co-IP analysis of Myc-tagged CAPNS1 and Flag-tagged WT calpain 1 or calpain 1-K398R in calpain 1 knock-out HEK293 cells, $$n = 5$.$ g Confocal microscopy of HeLa cells showing distribution of WT calpain 1 or calpain 1-K398R. Na+/K+-ATPase was used as plasma membrane marker. Scale bar, 10 μm. h–j WT calpain 1 or K398R mutant were expressed in calpain 1 knock out Hela cells and cleavage of α-spectrin cleavage and caspase 3 were evaluated by immunoblotting, $$n = 5$.$ Data represent means ± SEM; ∗$P \leq 0.05$, ∗∗$P \leq 0.01$, ns not significant. Together, these data establish that KCTD7–Cullin-3 ubiquitin ligase regulates calpain activity via atypical ubiquitination using distinct polyubiquitination chains. ## KCTD7–Cullin-3 ubiquitinates calpain 1 at K398 and calpain 2 at K280 and K674 Protein sequence analysis of calpains conducted with UbiNet48 identified several potential ubiquitination sites (Supplementary Table S2). Using site-directed mutagenesis, we generated several calpain 1 lysine-to-arginine mutants. In vivo ubiquitination assays showed that the highly conserved lysine at position 398 (Supplementary Fig. S2j) is required for KCTD7-mediated ubiquitination of calpain 1 (Fig. 4c, d), which was confirmed by mass spectrometry (Supplementary Fig. S2k, l). Interestingly, K398 resides in protein domain III, which mediates interaction with KCTD7 as shown above. To identify the functional consequences of calpain 1 K398R mutation, we expressed either wild-type (WT) or K398R calpain 1 in CAPN1 knock-out cells that we generated by using CRISPR/Cas9-mediated genome editing (Supplementary Fig. S2m). To check whether calpain 1 ubiquitination alters its calcium sensitivity, we performed in vitro assays by incubating immunopurified WT and ubiquitination-deficient (K398R) calpain 1 for 1 h at different calcium concentrations. We used calpain autolysis as an indicator of calpain activation. Calpain 1-K398R showed identical calcium sensitivity to WT calpain 1 (Supplementary Fig. S2n). Co-IP with CAPNS1 showed increased interaction of calpain 1-K398R compared to WT calpain 1 (Fig. 4e, f). Microscopic analysis of transiently transfected HeLa cells showed partial mislocalization of calpain 1-K398R to the plasma membrane (Fig. 4g). Immunoblotting analysis showed significantly higher α-spectrin cleavage upon calpain 1-K398R expression compared to the WT construct, indicating calpain hyperactivity (Fig. 4h, i). Cleaved caspase-3, a marker of apoptosis and an indicator of increased calpain activity49–51, was also increased upon expression of the calpain 1-K398R mutant compared to the WT construct (Fig. 4h, j). Together, these data demonstrate that Cullin-3–KCTD7 ubiquitin ligase controls calpain 1 activity through ubiquitination at K398 and that loss of ubiquitination affects calpain localization and increases the stability of calpain 1-CAPNS1 complex, thereby resulting in higher calpain activity. Additional mutagenesis experiments identified K280 and K674 as the sites of Cullin-3–KCTD7-mediated ubiquitination of calpain 2 (Supplementary Fig. S3a, b). Of note, neither site corresponds to K398 of calpain 1 (Supplementary Fig. S3c). Thus, KCTD7–Cullin-3 controls calpain 1 and calpain 2 using distinct polyubiquitination chains at different protein sites. ## KCTD7-mediated ubiquitination is a general feature of the calpain protein family Next, we investigated whether Cullin-3–KCTD7 ubiquitination is a common feature of the calpain protein family and focused on the ubiquitously expressed family members9–11,14,52–56. Exogenous expression of all ubiquitous calpains showed changes in α-spectrin cleavage for a subset of calpains only, suggesting substrate specificity (Fig. 5a, b). In vivo ubiquitination assay showed that all calpains undergo ubiquitination (Fig. 5c). Similar to calpain 1 and 2, proteasomal inhibition by MG-132 or lactacystin did not lead to the accumulation of any of the calpains, indicating that their ubiquitination is not a degradation signal (Supplementary Fig. 4a–h). Co-IP assays showed that all calpains interact with Cullin-3 (Fig. 5d) and KCTD7 (Fig. 5e). Overexpression of KCTD7, but not of Cullin-3, increased the ubiquitination levels of ubiquitous calpains (Fig. 5f, g). Together, these data establish Cullin-3–KCTD7 ubiquitin ligase as a general modifier of the calpain protein family and indicate that the levels of the adapter KCTD7 are a rate-limiting factor for the regulation of calpain activities. Fig. 5Ubiquitination-mediated regulation is common in calpain family.a, b Ubiquitous expressing calpains were expressed in HEK293 cells and α-spectrin cleavage was evaluated by immunoblotting, $$n = 3$.$ c In vivo ubiquitination assay for calpains performed by coexpressing with HA-Ub in HEK293 cells for 36 h. d Co-IP analysis of Flag-tagged calpains and myc-tagged Cullin-3 in HEK293 cells. e Co-IP analysis of Flag-tagged calpains and myc-tagged KCTD7 in HEK293 cells. f In vivo ubiquitination assay in HEK293 cells for calpains performed by expressing alone or in combination with Cullin-3. g In vivo ubiquitination assay in HEK293 cells for calpains performed by expressing alone or in combination with KCTD7. ## KCTD7 regulates calpain activity in vivo To study the regulatory role of KCTD7 in vivo, we generated a Kctd7 knock-out (Kctd7 KO) mouse line (Kctd7–/–) by deleting exon 2 using CRISPR/Cas9 genome editing (Supplementary Fig. S5a–d). Kctd7–/– mice were viable and fertile and did not show any gross abnormalities, but they had a median lifespan of only eight months (Supplementary Fig. S5e). Immunohistochemical analysis of brain tissue showed that, in WT mice, CAPNS1 displayed a dual subcellular localization in the cytosol and at lysosomes. In Kctd7–/– mice, CAPNS1 lost its lysosomal localization and relocated to the plasma membrane (Fig. 6a, b and Supplementary Fig. S5f, h), which is a known indicator of increased calpain activity23–25. Immunoblot analysis of brain samples from Kctd7–/– mice showed a significantly higher autolysis of calpain 1 and CAPNS1 compared to WT mice (Fig. 6c–e and Supplementary Fig. S5g), another indicator of increased calpain activity57–59. To confirm increased calpain activity, we analyzed the interaction of calpains with their substrates and their processing status. Co-IP analyses showed that loss of KCTD7 led to a dramatic increase in the interaction of the ubiquitous calpain substrate, the cytoskeletal protein α-spectrin, with both calpain 1 and CAPNS1 (Fig. 6f–h); α-spectrin cleavage was also markedly increased (Fig. 6f, i). It should be noticed that this increased activity is associated with an overall small fraction of total calpain 1 and CAPNS1 proteins undergoing autolysis. Increased cleavage of α-spectrin was also observed in primary Kctd7–/– astrocytes (Supplementary Fig. S5i, j). To check whether calpain activity was increased in neurons, we analyzed the status of the neuron-specific calpain substrate p3560. Immunoblot analysis showed that p35-to-p25 conversion was significantly higher in multiple brain regions from Kctd7–/– mice than in their WT counterparts (Fig. 6j–l and Supplementary Fig. S5k–p), thus confirming increased neuronal activity of calpains in the absence of KCTD7.Fig. 6KCTD7 regulates calpain activity in vivo.a, b Confocal microscopy analysis of cerebellar tissue from 5-week-old WT and Kctd7–/– mice showing CAPNS1 co-localization with the lysosomal marker LAMP1 in Purkinje cells. Trace outline is used for line-scan (white dashed line) analysis of relative fluorescence intensity of CAPNS1 and LAMP1 signals. Scale bar, 20 μm. c–e Cerebellar tissues from 5-week-old WT and Kctd7–/– mice were lysed in RIPA buffer and immunoblotted with the indicated antibodies, $$n = 4$.$ f–l, Cerebellar tissues from 5-week-old WT and Kctd7–/– mice were lysed in NP-40 lysis buffer and calpain 1 and CAPNS1 were immunoprecipitated and probed for interaction with α-spectrin by immunoblotting (f–h). Lysates from same tissue samples were also immunoblotted for α-spectrin (f–i) and p35 (j–l) to evaluate calpain-mediated cleavage, $$n = 4$.$ Data represent means ± SEM; ∗$P \leq 0.05$, ∗∗$P \leq 0.01$, ns not significant. As KCTD7 deficiency has been linked to a subtype of neuronal ceroid lipofuscinosis (NCL)27, we checked whether other NCL subtypes are characterized by higher calpain activity. An analysis of the cleavage of α-spectrin and p35 in the brain of mouse models of CLN3, CLN6, and CLN8 diseases showed no obvious differences compared to their WT littermates (Supplementary Fig. S6a–l), indicating specific involvement of KCTD7 in the regulation of calpains. ## KCTD7 deficiency leads to motor incoordination and cerebellar degeneration Mutations in the KCTD7 gene cause a spectrum of progressive neurodegenerative phenotypes characterized by ataxia and psychomotor decline/motor incoordination preceded in some (but not all) cases by intractable myoclonic seizures after several months of normal development26–28. Progressive decline finally results in severe motor and mental retardation and early death. Additional variable features include dysarthria, truncal ataxia, loss of fine finger movements, and microcephaly26–28. Brain imaging shows global cortical and cerebellar atrophy and thinning of the corpus callosum. Differently from other similar progressive neurodegenerative diseases, however, there is an absence of retinal degeneration27,33,61. Prior work has shown that Kctd7 expresses ubiquitously, with significantly higher levels in Purkinje cells of the cerebellum32. Thus, to identify pathological phenotypes in Kctd7–/– mice, we focused primarily on cerebellar tissue and function. Confocal imaging and immunoblot analysis of cerebellar tissue of Kctd7–/– mice at 5 weeks and 6 months of age revealed a severe loss of Purkinje cells associated with caspase-3 activation (an indicator of cell death) at both time points (Fig. 7a, b and Supplementary Fig. S7a–e). Neurodegeneration was accompanied by neuroinflammation, as evidenced by a marked increase in GFAP immunoreactivity (a marker of neuroinflammation) compared to the WT littermates (Fig. 7c–e and Supplementary Fig. S8a, b). In contrast, there was no increase in astrogliosis in the cortex or hippocampus (Supplementary Fig. 8c–f).Fig. 7KCTD7 deficiency leads to behavioral impairment and neurodegeneration.a–c Confocal microscopy analysis of the cerebellum from 5-week-old WT and Kctd7–/– mice. α-calbindin antibody and α-GFAP antibody were used to label Purkinje cells (a) and astrocytes (c), respectively. Quantification of Purkinje cell numbers is reported in b, $$n = 5$$ per genotype. Scale bars, 200 μm. d, e *Immunoblot analysis* of cerebellar tissue from 5-week-old WT and Kctd7–/– mice using α-GFAP antibody. f Rotarod test measuring the latency to fall for 8-week-old WT ($$n = 10$$) and Kctd7–/– ($$n = 16$$) mice. g Inverted pole test measuring the time used to climb down from the pole for 8-week-old WT ($$n = 10$$) and Kctd7–/– ($$n = 19$$) mice. Data represent means ± SEM; ∗$P \leq 0.05$, ∗∗$P \leq 0.01$, ∗∗∗$P \leq 0.001$, ns not significant. We assessed motor performance of Kctd7–/– mice at 8 weeks of age using the accelerating rotating rod (rotarod) test and found that the performance of Kctd7–/– mice was significantly impaired compared to their WT littermates (Fig. 7f). In a vertical pole test, Kctd7–/– mice placed at the top of the pole took significantly longer than their WT littermates to climb down (Fig. 7g). We did not observe any significant differences in muscle strength as measured by the wire suspension test, grid suspension test (Supplementary Fig. S8g, h), or forepaw grip strength test (Supplementary Fig. S8i), confirming that the observed motor incoordination was not due to muscle weakness. ## Calpain inhibition in Kctd7–/– mice ameliorates behavioral impairments and neurodegeneration To test whether aberrant calpain activity drives neurodegeneration in Kctd7–/– mice, we treated them with either the calpain inhibitor E-64 (6.4 mg/kg, i.p. daily) or vehicle for 5 weeks starting at weaning (Fig. 8a)62. E-64 is an irreversible, potent, and highly selective cysteine protease inhibitor. The trans-epoxysuccinyl group (active moiety) of E-64 forms an irreversible thiolester bond with the thiol group of the cysteine residue at the active site of the protein63. E-64 is suitable for in vivo studies because it permeates tissues and cells and has low toxicity63. Calpain inhibition through E64 has been shown to be beneficial in neurodegenerative disease62,64. E-64 significantly inhibited calpain activity in vivo as observed by a marked decrease in α-spectrin cleavage (Fig. 8b, c). Neuropathological analyses revealed that calpain inhibition resulted in a significant prevention of Purkinje cell loss (Fig. 8d, e) and in a reduction of astrocytosis (Fig. 8f–h) in Kctd7–/– mice. Behavioral testing showed improved performance of Kctd7–/– mice upon calpain inhibition in both the rotarod and pole tests compared to vehicle-treated mice (Fig. 8i, j). Together, these data indicate that calpain hyperactivity is a driving factor to neuropathology caused by KCTD7 deficiency. Fig. 8Calpain inhibition in Kctd7–/– mice ameliorates behavioral impairments and neurodegeneration.a Schematic diagram of drug treatment, tissue preparation and behavioral tests. b, c *Immunoblot analysis* of cerebellar tissue from 8-week-old WT and Kctd7–/– mice treated with E-64 or vehicle using α-spectrin antibody, $$n = 3$$ for WT (untreated and treated), $$n = 4$$ for Kctd7–/– (untreated and treated). d–f Confocal microscopy analysis of the cerebellum from 8-week-old WT and Kctd7–/– mice treated with E-64 or vehicle. α-calbindin antibody and α-GFAP antibody were used to label Purkinje cells (d) and astrocytes (f), respectively. Scale bars, 200 μm. Quantitation of Purkinje cell numbers is reported in e, $$n = 5$.$ g, h *Immunoblot analysis* of cerebellar tissue from 8-week-old WT and Kctd7–/– mice treated with E-64 or vehicle using α-GFAP antibody, $$n = 3$$ for WT (untreated and treated), $$n = 4$$ for Kctd7–/– (untreated and treated). i Rotarod test measuring the latency to fall for 8-week-old WT and Kctd7–/– mice treated with E-64 or vehicle. $$n = 10$$ for vehicle-treated mice (WT and Kctd7–/–). $$n = 12$$ for E-64-treated mice (WT and Kctd7–/–). j Inverted pole test measuring the time used to climb down from the pole for 8-week-old WT and Kctd7–/– mice treated with E-64 or vehicle. $$n = 10$$ for vehicle-treated mice (WT and Kctd7–/–). $$n = 12$$ for E-64-treated mice (WT and Kctd7–/–); ∗$P \leq 0.05$, ∗∗$P \leq 0.01$, ∗∗∗$P \leq 0.001$, ns not significant. ## Discussion This study establishes that KCTD7 is a key regulator of calpains, a class of non-lysosomal calcium-activated cysteine proteases that exert their regulatory functions via limited proteolysis of their substrates2. We found that KCTD7 forms a complex with Cullin-3 to work as a substrate adapter that mediates calpain ubiquitination. Our results show that such ubiquitination is non-degradative and prevents calpain autolysis, thereby resulting in modulation of calpain function (Fig. 9). Interestingly, other members of the KCTD protein family have been shown to interact with Cullin-3, but they target their protein substrates to the ubiquitin-proteasome pathway for degradation65. Previous work has shown that several KCTD proteins participate in apoptosis and cell proliferation, differentiation, and metabolism, and loss-of-function mutations in at least 4 KCTD genes cause human disease29,66–72. At the molecular level, KCTD proteins have been implicated in the regulation of transcription, DNA replication, amino acid signaling to mTORC1, and regulation of Rho GTPases in brain development among other processes65. Thus, there are significant divergences among the mechanisms of action of KCTD proteins as well as the processes they regulate, with KCTD7 being the first example of pathway regulation via non-degradative ubiquitination. Fig. 9Schematic model of calpain regulation by the Cullin-3–RBX1–KCTD7 complex. Shown is a comparison between WT conditions and deficiency of KCTD7 leading to calpain dysregulation and activation of a pathogenic cascade. We observed that loss of KCTD7 in mice leads to higher calpain activity in the brain, which is associated with motor impairment, severe loss of Purkinje cells, neuroinflammation, and premature death. Pharmacological inhibition of calpains ameliorated behavioral phenotypes and neuropathology in Kctd7–/– mice, thus demonstrating that calpain dysregulation is a key driver in the pathogenesis of disease. Previous work established that KCTD7 regulates retinal neurovascular patterning and function73 and also plays a role in the maintenance of autophagic homeostasis33 — the latter being a characteristic shared by various NCL proteins74–78. Our data, however, show that calpain hyperactivity is a specific consequence of KCTD7 loss of function among several NCL proteins investigated by using mouse models of disease. Abnormal calpain activity has been shown to contribute to various metabolic and degenerative diseases, especially late-onset neurodegenerative disorders2,8,15,16,79–81. Work performed in animal models of frontotemporal dementia, Parkinson’s disease, Huntington’s disease, and spinocerebellar ataxia type 3 has shown that calpain-mediated fragmentation of the proteins implicated in disease (TDP-43, α-synuclein, huntingtin, and ataxin-3) either increases their tendency to aggregate or directly leads to neurotoxicity82–85. The emerging consensus is that interventions based on the modulation of calpain activity could be a viable avenue to treat diseases with calpain-associated toxicity; indeed, pharmacological calpain inhibitors are being tested in clinical trials for several human diseases86,87. Classical studies have shown that local calcium concentration and subcellular localization are important factors in the regulation of calpain activity23,24,88,89. A more recent study has shown that the membrane protein Ttm50, a subunit of the TIM23 complex in the inner mitochondrial membrane, facilitates calpain activation by mediating calpain localization at the Golgi/ER25,90. Ttm50 binding increases calpain sensitivity to calcium, possibly either by promoting a calpain conformational change, or by increasing the binding affinity for calcium. Thus, Ttm50 acts as a calpain anchor by localizing calpain to the calcium stores at the Golgi/ER. Similarly, in KCTD7-mediated control of calpain activity, subcellular localization of calpains appears to play an important role in their activation. Ubiquitination-deficient calpain 1-K398R indeed shows altered localization to the plasma membrane. This change does not seem to be coupled with calcium sensitivity but is rather associated with significantly stronger binding to CAPNS1, which is required for calpain 1 stability and activation90. Our results uncover a new mechanism of calpain regulation that could be harnessed to restrain abnormal activity of calpains for therapeutic applications by leveraging the cellular components that execute such control. Using mutants of ubiquitin where only single lysine residues were available for linkage, we determined that the KCTD7–Cullin-3 ligase complex promotes the preferential incorporation of Ub-K6, Ub-K27, Ub-K29, and Ub-K63 for ubiquitination of calpain 1 and CAPNS1, whereas it uses Ub-K6 for calpain 2. Different polyubiquitin chain linkages direct substrates towards different pathways91. The functions of these lysine residues are less characterized compared to K48, which instead has a well-defined role in proteasomal-mediated protein degradation92–94. Recent work has demonstrated that atypical ubiquitination can regulate protein activity, localization, and affinity to binding partners95. Of note, previous work has shown that loss of ubiquitin K6 alters the ubiquitin-proteasome system (UPS) and results in Ca2+ elevation, hyperactivation of calpains, and consequent cleavage of calpain substrates96. Thus, our finding that ubiquitin K6 is used for ubiquitination-mediated regulation of the calpains provides a mechanistic understanding of these results and links functionally calpain activity to the UPS. We also identified K398 of calpain 1 and K280 and K674 of calpain 2 as the specific sites that undergo Cullin-3–KCTD7-mediated ubiquitination. Our experiments show that ubiquitination-resistant calpain 1-K398R is hyperactive and activates caspase-3, thus demonstrating the role of ubiquitination in the regulation of calpain activity and bridging the link with downstream disease-associated pathways. In summary, this study establishes a new paradigm for calpain regulation based on KCTD7-mediated atypical ubiquitination. The identification of the function of KCTD7 as a calpain adapter sheds light on the molecular pathogenesis of KCTD7-associated disease and suggests novel therapeutic avenues to mitigate the pathological effects of calpain hyperactivity in neurodegenerative disease and cancer. ## Antibodies and reagents Antibodies used in this study are reported in Supplementary Table S3. Reagents and chemicals include: *Ionomycin calcium* salt (Sigma # I3909), Protein G-Agarose (Sigma #11243233001), Streptavidin Sepharose High Performance bead (Sigma #17-5113-01), Calmodulin Sepharose 4B (Sigma #17-0529-01), Dynabeads™ Protein G for Immunoprecipitation (Thermo Fisher Scientific #10004D), jetPRIME®, DNA and siRNA transfection reagent, Polyplus-transfection® reagent (VWR #89129-922), TRIzol™ Reagent (Thermo Fisher Scientific #15596018), Quantitect reverse transcriptase (Qiagen #205313), Xpert Protease inhibitor cocktail (GenDEPOT #P3100-100), Xpert Phosphatase inhibitor cocktail (GenDEPOT #P3200-020), Penicillin-Streptomycin (GenDEPOT #CA005-100), 10× PBS Buffer (GenDEPOT #P2100-100), DMEM, High Glucose with L-Glutamine (GenDEPOT #CM002-050), Opti-MEM™ Reduced Serum Medium (Thermo Fisher Scientific #31985070), FBS Opti-Gold, Us Origin (GenDEPOT #F0900-050), iTaq Universal SYBR® Green Supermix (Bio-Rad #1725124), Immun-Blot PVDF Membrane (Bio-Rad #1620177), TGX™ FastCast™ Acrylamide Kit (Bio-Rad #161-0173, #161-0175), Precision Plus Protein™ Dual Color Standards (Bio-Rad #1610394), Blotto, non-fat dry milk (Santa-Cruz # sc-2325), SuperSignal™ West Dura Extended Duration Substrate (Thermo Fisher Scientific #34076), VECTASHIELD HardSet Antifade Mounting Medium with DAPI (Vectorlabs #H-1500). ## Plasmid constructs and siRNA cDNAs generated by retrotranscription of RNAs from HeLa and HEK293 cells using QuantiTect Reverse Transcription kit were used to PCR-amplify the coding sequences of KCTD7 and CAPNS1, which were then inserted into p3×Flag-CMV-14 vector (pcDNA) by using the In-Fusion cloning kit (Clontech). Oligonucleotides used for In-Fusion cloning are reported in Supplementary Table S4. The following plasmids were obtained from AddGene: Tandem affinity purification plasmid pIRESpuro-GLUE (pGLUE) empty backbone (#15100, deposited by Randall Moon); pcDNA3-myc-CUL3 (#19893, deposited by Yue Xiong); p3×Flag-CAPN1 (#60941, deposited by Yi Zhang), p3×Flag-CAPN2 (#60942, deposited by Yi Zhang); Flag-CAPN1 and Flag-CAPN2 (#60941 and #60942, deposited by Yi Zhang), HA-ubiquitin WT, HA-ubiquitin K0, HA-ubiquitin K33, HA-ubiquitin K48, HA-ubiquitin K63 (#17608, #17603, #17607, #17605, and #17606, deposited by Ted Dawson), HA-ubiquitin K6, HA-ubiquitin K11, HA-ubiquitin K27, HA-ubiquitin K29 (#22900, #22901, #22902, and #22903, deposited by Sandra Weller). siRNAs were obtained from Santa-Cruz; KCTD7 (#sc-89656), Cullin-3 (#sc-35130), Calpain 1 (#sc-29885), and Calpain 2 (#sc-41459), Calpain reg. (# sc-29887). ON-TARGETplus Non-targeting Pool of siRNAs was purchased from Dharmacon (#D-001810-10-05). shRNAs targeting Cullin-3 or KCTD7 were obtained from Cell-Based Assay Screening Service (C-BASS) at Baylor College of Medicine. ## Cell culture, lysate preparation, and western blot analysis HeLa or HEK293 cells (obtained and certified from ATCC) were cultured in DMEM containing $10\%$ FBS and antibiotics (Penicillin/Streptomycin) and were free from mycoplasma contamination. siRNAs or shRNAs were transfected using jetPRIME transfection reagent and incubated for 48–72 h prior to western blot analysis. Plasmids were transfected using jetPRIME transfection reagent and left to express from 30 to 72 h, depending on the downstream application. Unless otherwise mentioned, cells were lysed with RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 % NP-40, $0.5\%$ sodium deoxycholate, $0.1\%$ SDS) supplemented with protease and phosphatase inhibitors on ice for 30 min with vortexing. Lysates were cleared by centrifugation (13,000 rpm/15 min, 4 °C) followed by protein quantification via BCA assay (Pierce). Sample buffer with reducing agent was added to each lysate followed by 5 min incubation at 95 °C. Samples were spun down and run on Tris-Glycine gel, transferred to a PVDF membrane and blocked for 1 h with blocking buffer ($5\%$ w/v, dried skimmed milk in Tris-buffered saline, pH 7.4, and $0.2\%$ Tween 20, TBST) prior to overnight primary antibody incubation. Detection was carried out with SuperSignal™ West Dura Extended Duration Substrate reagent. Images were detected with ImageQuant LAS 4000 (GE Healthcare) and quantified by *Fiji analysis* software. ## Tandem-affinity purification and mass spectrometry HEK293 cells (2 × 108 cells) constitutively expressing SBP-HA-CBP-tagged KCTD7 were selected and maintained in DMEM containing 1 µg/mL puromycin. A stable cell line expressing low levels of KCTD7 was used for the tandem-affinity purification procedure. Briefly, the cells were lysed with lysis buffer ($10\%$ glycerol, 50 mM HEPES-KOH, pH 8.0, 100 mM KCl, 2 mM EDTA, $0.1\%$ NP-40, 2 mM DTT, 10 mM NaF, 0.25 mM NaOVO3, and protease inhibitors). The lysates were cleared by centrifugation at 13,000× g for 15 min and then incubated at 4 °C with 100 μL of packed streptavidin resin. The beads were washed 5 times, and protein complexes were then eluted from the streptavidin resin in calmodulin-binding buffer supplemented with 2 mM biotin. The second round of affinity purification was performed with 100 μL of calmodulin resin. After several washes, the protein complexes on beads were directly digested with sequencing-grade trypsin (Sigma). The peptide supernatant solution was removed to a new Eppendorf tube. The beads were further extracted with 100 μL of 1:2 ($5\%$ formic acid: acetonitrile) solution with shaking at 37 °C for 15 min. The supernatant was pooled with the previous supernatant. The solutions were Speedvac dried. The digests were resuspended in $0.1\%$ formic acid and $5\%$ acetonitrile solution. The concentration of digest was measured using NanoOrange. The resulting peptide mixture was then analyzed by nano liquid chromatography-tandem mass spectrometry (nanoLC-MS/MS) using data-dependent acquisition on cHiPLC nano liquid chromatography system (Eksigent) and TripleTOF 5600 mass spectrometer (ABSCIEX). Acquired spectra were searched against a FASTA file containing the human NCBI sequences by using the ProteinPilot version 4.5 software. ## Immunofluorescence and Immunohistochemistry Cells were grown on glass coverslips in 24-well plates and were transiently transfected with appropriate plasmids for 24–36 h. Post transfection, coverslips were washed in PBS and fixed with $4\%$ paraformaldehyde for 15 min. Cells were washed 3 times for 5 min each with PBS and permeabilized with $0.1\%$ TritonX-100 in PBS for 5 min. After permeabilization, cells were blocked with $10\%$ goat serum for 1 h. Primary antibody incubation was carried out overnight at 4 °C. After 3–4 washes with PBS (5 min each), coverslips were incubated with fluorophore-conjugated secondary antibody for 1 h. Coverslips were finally washed 3–5 times with PBS and mounted on glass slides with Vectashield DAPI (Vector Laboratories). Images were acquired through a 63× oil immersion objective Zeiss 880 confocal laser microscope (Zeiss, Oberkochen, Germany). For mouse tissue samples, free floating or slide-mounted sections were permeabilized and blocked in PBS containing $0.3\%$ Triton X-100 and $10\%$ normal goat serum. Samples were incubated with the primary antibody overnight in blocking buffer and were washed three times with PBS the next day prior to adding fluorescent-conjugated secondary antibody (either Alexa 488 or Alexa 568, 1:500 concentration) and incubating for 2 h at room temperature. Tissues were then washed three times (10 min each) and slides were coverslipped using Vectashield DAPI and imaged using a Zeiss 880 confocal laser microscope. ## Immunoprecipitation Cells were transfected with the desired plasmids as described. 36–48 h later, cells were collected and lysis was performed on ice for 30 min with brief vortexing using lysis buffer ($1\%$ NP-40, 150 mM NaCl, 50 mM Tris, pH 8.0, $10\%$ glycerol, 1 mM EDTA and protease inhibitors). Cell debris were removed by centrifugation (13,000 rpm/15 min, 4 °C) and pre-cleared with un-conjugated beads. Primary antibodies (2–5 µg) were added to the lysates and incubated with rotation overnight at 4 °C. The next day, Dynabeads were added to the lysate and incubated for 2 h at 4 °C with rocking. Beads were then washed four times with 500 µL of lysis buffer before being eluted in Laemmli buffer at 95 °C for 10 min. A similar procedure was also followed for coimmunoprecipitation using mouse tissue samples. ## Subcellular fractionation Subcellular fractionation of HeLa cells was performed as described previously25. Briefly, four 15-cm plates of HeLa cells were washed with ice-cold PBS with protease inhibitors. Cells were then homogenized in 5 mL of buffer A (0.3 M sucrose, 1 mM EDTA, 1 mM MgSO4, 10 mM MES-KOH, pH 6.5) containing protease and phosphatase inhibitors by 5 freeze-thaw cycles using liquid nitrogen and a 37 °C waterbath. The cell homogenates were centrifuged at 1000× g for 10 min to remove unbroken cells and nuclei. The supernatant was next centrifuged at 8000× g and 35,000× g for 30 min each, which sediments the mitochondria. A subsequent ultracentrifugation (150,000× g for 120 min) resulted in separation of Golgi/ER and cytosol. Equal amounts of proteins for each fraction were processed for western blotting analysis. ## In vitro ubiquitination assay Flag-tagged calpain subunits and KCTD7 protein were immunopurified from transiently transfected HEK293 cells using Flag antibodies and peptides. Purified proteins were incubated with 12 μg of ubiquitin (R&D Systems; U-100H), 120 ng UBE1 (R&D Systems; E-304-050), 300 ng UbcH5b/UBE2D2 (R&D Systems; E2-622-100), 150 ng His6-CUL3/NEDD8/RBX1 Complex Protein (R&D Systems; E3-436-025), and 10 mM MgATP Solution (R&D Systems; B-20) in E3 Ligase Reaction Buffer (R&D Systems; B-71) for 1 h at 37 °C with gentle shaking. The reactions were quenched with addition of the SDS-PAGE sample buffer and boiling and were analyzed by SDS-PAGE. ## Generation of Kctd7–/– mice Kctd7–/– mice were generated at Mouse Embryonic Stem Cell Core and Genetically Engineered Mouse Core at Baylor College of Medicine. Two gRNAs were designed to cut both ends of the exon 2 of the *Kctd7* gene (cgacctgatagcccttaaatggg and tgtgtgcgtgaggcccgaatggg). Established methods were used to co-microinject 100 ng/µL Cas9 mRNA and 20 ng/µL validated sgRNA (each) into the cytoplasm of 100 C57BL/6NJ embryos. Following micro-injection, zygotes were transferred into pseudopregnant FVB females. Both C57BL/6 J and FVB mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Potential founder mice were genotyped using tail tissue DNA and deletion was confirmed using Sanger sequencing. Genotyping was done using the primers indicated in Supplementary Table S4. The top ten potential off-target sites for each sgRNA were sequenced to confirm no off-target effects. Founder mice were backcrossed at least three times prior to experimentation to eliminate other potential off-target mutations. All animals were housed in a Level 3, AALAS-certified facility on a 12h-light cycle. Husbandry, housing, euthanasia, and experimental guidelines were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Baylor College of Medicine. ## Rotarod Eight- to nine-week-old mice were tested. After 30 min habituation in the test room, motor coordination was measured using an accelerating rotating rod (Ugo Basile Biological Research Apparatus, Varese, Italy). Mice were tested for 4 consecutive days, 4 trials each, with an interval of 30 min between trials to rest. Each trial lasted for a maximum of 5 min and the rod accelerated from 4 to 40 rpm. “ Latency to fall” was recorded either when the mouse fell from the rod or when the mouse had ridden the rotating rod for two revolutions without regaining control. Behavioral scores were subjected to statistical analysis by two-way ANOVA with Bonferroni’s post-hoc analysis. ## Vertical pole Animals were habituated to the experimental room for 30–60 min prior to testing. Mice were placed on the top of a rough-surfaced vertical pole (45 cm high, 1.1 cm in diameter). The time for each mouse to turn downward and to reach the floor was recorded. Each trial lasted for a maximum of 120 s and each mouse performed three trials with 5 min inter-trial intervals. ## Grip strength Mice were habituated in the test room for 30 min. Each mouse was allowed to grab the bar of a digital grip strength meter (Columbus Instruments, Columbus, OH) with both forepaws while being held by the tail and then pulled away from the meter with a constant slow force until the forepaws released. The grip (in kg of force) was recorded, and the procedure repeated twice for a total of three pulls, which were averaged for the final result. Data is shown as means ± SD. ## Wire suspension The mice were habituated to the experimental room for 30–60 min prior to testing. The mice were forced to grasp a 3‐mm wire and hang from it on their forepaws. The ability of the mice to grasp the wire was scored and the time for which they held the wire (maximum 30 s) was registered. ## Inverted grid The mice were habituated to the experimental room for 30–60 min prior to testing. A grid screen measuring 20 cm × 25 cm with a mesh density of 9 squares/cm2 was elevated 45 cm above a cage with bedding. Each subject was placed head oriented downward in the middle of the grid screen. When it was determined that the subject had proper grip on the screen, it was inverted 180°. The hang time (duration a subject held on to the screen without falling) was recorded, up to a cutoff time of 60 s. Any subject that was able to climb onto the top of the screen was assigned a time of 60 s. ## Mass spectrometry analysis Trypsin (800 ng) was added to immunoprecipitated proteins on the beads for 6 h at 37 °C. The initial digested samples were centrifuged for 2 min at 5000× g and the supernatants were collected into fresh tubes. Beads were washed twice with 100 mM ammonium bicarbonate and the supernatants were pooled. The resulting samples were reduced with 20 mM dithiothreitol at 37 °C for 1 h, and cysteine was alkylated with 80 mM iodoacetamide for 45 min in dark. Samples were treated with 600 ng of trypsin to overnight incubation at 37 °C. The resulting peptides were desalted using solid-phase extraction on a C18 Spin column and eluted with $0.1\%$ FA in $80\%$ ACN. Peptides were analyzed by LC-MS/MS using a nanoElute coupled to a timsTOF Pro2 Mass Spectrometer (Bruker Daltonics). Samples were loaded on a capillary C18 column (15 cm length, 75 μm inner diameter, 1.9 μm particle size, 120 Å pore size; Bruker Daltonics). The flow rate was kept at 300 nL/min. Solvent A was $0.1\%$ FA in water, and Solvent B was $0.1\%$ FA in ACN. The peptides were separated on a 100 min analytical gradient from $2\%$ ACN/$0.1\%$ FA to $35\%$ ACN/$0.1\%$ FA for a total of 120 min gradient. The timsTOF Pro2 was operated in the PASEF mode. MS and MS/MS spectra were acquired from 100 to 1700 m/z. The inverse reduced ion mobility 1/K0 was set to 0.60−1.60 V·s/cm2 over a ramp time of 100 ms. Data-dependent acquisition was performed using 10 PASEF MS/MS scans per cycle with a near $100\%$ duty cycle. The resulting protein tandem MS data was queried for protein identification against the SwissProt human database (released on April, 2021) using MaxQuant v2.1.0.0. The following modifications were set as search parameters: peptide mass tolerance at 20 ppm, trypsin digestion cleavage after K or R (except when followed by P), 2 allowed missed cleavage sites, carbamidomethylated cysteine (static modification), and oxidized methionine, deaminated asparagine/glutamine, protein N-term acetylation, and Diglycyl lysine (variable modification). Search results were validated with peptide and protein FDR both at 0.01. High confident ubiquitination sites were verified with high localization probability (> 0.95) and manual confirmation of MS/MS spectrum. The mass spectrometry proteomics data have been deposited to the ProteomeXchange *Consortium via* the PRIDE partner repository with the dataset identifier PXD038842. ## Supplementary information Supplementary Information The online version contains supplementary material available at 10.1038/s41421-023-00533-3. ## References 1. Sorimachi H, Ishiura S, Suzuki K. **Structure and physiological function of calpains**. *Biochem. J.* (1997) **328** 721-732. DOI: 10.1042/bj3280721 2. Goll DE, Thompson VF, Li H, Wei W, Cong J. **The calpain system**. *Physiol. Rev.* (2003) **83** 731-801. DOI: 10.1152/physrev.00029.2002 3. Huttenlocher A. **Regulation of cell migration by the calcium-dependent protease calpain**. *J. Biol. Chem.* (1997) **272** 32719-32722. DOI: 10.1074/jbc.272.52.32719 4. Palecek SP, Huttenlocher A, Horwitz AF, Lauffenburger DA. **Physical and biochemical regulation of integrin release during rear detachment of migrating cells**. *J. Cell Sci.* (1998) **111** 929-940. DOI: 10.1242/jcs.111.7.929 5. Croce K. **Inhibition of calpain blocks platelet secretion, aggregation, and spreading**. *J. Biol. Chem.* (1999) **274** 36321-36327. DOI: 10.1074/jbc.274.51.36321 6. Franco SJ, Huttenlocher A. **Regulating cell migration: calpains make the cut**. *J. Cell Sci* (2005) **118** 3829-3838. DOI: 10.1242/jcs.02562 7. Momeni HR. **Role of calpain in apoptosis**. *Cell J.* (2011) **13** 65-72. PMID: 23507938 8. Camins A, Verdaguer E, Folch J, Pallàs M. **Involvement of calpain activation in neurodegenerative processes**. *CNS Drug Rev.* (2006) **12** 135-148. DOI: 10.1111/j.1527-3458.2006.00135.x 9. Richard I. **Mutations in the proteolytic enzyme calpain 3 cause limb-girdle muscular dystrophy type 2A**. *Cell* (1995) **81** 27-40. DOI: 10.1016/0092-8674(95)90368-2 10. Gan-Or Z. **Mutations in CAPN1 cause autosomal-recessive hereditary spastic paraplegia**. *Am. J. Hum. Genet.* (2016) **98** 1038-1046. DOI: 10.1016/j.ajhg.2016.04.002 11. Wang Y. **Defects in the CAPN1 gene result in alterations in cerebellar development and cerebellar ataxia in mice and humans**. *Cell Rep* (2016) **16** 79-91. DOI: 10.1016/j.celrep.2016.05.044 12. Sleiman PMA. **GWAS identifies four novel eosinophilic esophagitis loci**. *Nat. Commun.* (2014) **5** 5593. DOI: 10.1038/ncomms6593 13. Zha C. **Biallelic variants in the small optic lobe calpain CAPN15 are associated with congenital eye anomalies, deafness and other neurodevelopmental deficits**. *Hum. Mol. Genet.* (2020) **29** 3054-3063. DOI: 10.1093/hmg/ddaa198 14. Horikawa Y. **Genetic variation in the gene encoding calpain-10 is associated with type 2 diabetes mellitus**. *Nat. Genet.* (2000) **26** 163-175. DOI: 10.1038/79876 15. Liu J, Liu MC, Wang KKW. **Calpain in the CNS: from synaptic function to neurotoxicity**. *Sci. Signal.* (2008) **1** re1. DOI: 10.1126/stke.114re1 16. Nixon RA. **The calpains in aging and aging-related diseases**. *Ageing Res. Rev.* (2003) **2** 407-418. DOI: 10.1016/S1568-1637(03)00029-1 17. Adamec E, Mohan P, Vonsattel JP, Nixon RA. **Calpain activation in neurodegenerative diseases: confocal immunofluorescence study with antibodies specifically recognizing the active form of calpain 2**. *Acta Neuropathol.* (2002) **104** 92-104. DOI: 10.1007/s00401-002-0528-6 18. Saito K, Elce JS, Hamos JE, Nixon RA. **Widespread activation of calcium-activated neutral proteinase (calpain) in the brain in Alzheimer disease: a potential molecular basis for neuronal degeneration**. *Proc. Natl. Acad. Sci. USA* (1993) **90** 2628-2632. DOI: 10.1073/pnas.90.7.2628 19. Ishiura S, Murofushi H, Suzuki K, Imahori K. **Studies of a calcium-activated neutral protease from chicken skeletal muscle. I. Purification and characterization**. *J. Biochem.* (1978) **84** 225-230. DOI: 10.1093/oxfordjournals.jbchem.a132111 20. Ohno S. **Evolutionary origin of a calcium-dependent protease by fusion of genes for a thiol protease and a calcium-binding protein?**. *Nature* (1984) **312** 566-570. DOI: 10.1038/312566a0 21. Wingrave JM. **Early induction of secondary injury factors causing activation of calpain and mitochondria-mediated neuronal apoptosis following spinal cord injury in rats**. *J. Neurosci. Res.* (2003) **73** 95-104. DOI: 10.1002/jnr.10607 22. Ono Y, Sorimachi H. **Calpains: an elaborate proteolytic system**. *Biochim. Biophys. Acta* (2012) **1824** 224-236. DOI: 10.1016/j.bbapap.2011.08.005 23. Hood JL, Brooks WH, Roszman TL. **Differential compartmentalization of the calpain/calpastatin network with the endoplasmic reticulum and Golgi apparatus**. *J. Biol. Chem.* (2004) **279** 43126-43135. DOI: 10.1074/jbc.M408100200 24. Leloup L. **m-Calpain activation is regulated by its membrane localization and by its binding to phosphatidylinositol 4,5-bisphosphate**. *J. Biol. Chem.* (2010) **285** 33549-33566. DOI: 10.1074/jbc.M110.123604 25. Metwally E, Zhao G, Wang Q, Zhang YQ. **Ttm50 facilitates calpain activation by anchoring it to calcium stores and increasing its sensitivity to calcium**. *Cell Res.* (2021) **31** 433-449. DOI: 10.1038/s41422-020-0388-4 26. Kousi M. **Novel mutations consolidate KCTD7 as a progressive myoclonus epilepsy gene**. *J. Med. Genet.* (2012) **49** 391-399. DOI: 10.1136/jmedgenet-2012-100859 27. Staropoli JF. **A homozygous mutation in KCTD7 links neuronal ceroid lipofuscinosis to the ubiquitin-proteasome system**. *Am. J. Hum. Genet.* (2012) **91** 202-208. DOI: 10.1016/j.ajhg.2012.05.023 28. Blumkin L. **A compound heterozygous missense mutation and a large deletion in the KCTD7 gene presenting as an opsoclonus-myoclonus ataxia-like syndrome**. *J. Neurol.* (2012) **259** 2590-2598. DOI: 10.1007/s00415-012-6545-z 29. Van Bogaert P. **Mutation of a potassium channel-related gene in progressive myoclonic epilepsy**. *Ann. Neurol.* (2007) **61** 579-586. DOI: 10.1002/ana.21121 30. Mole SE. **Clinical challenges and future therapeutic approaches for neuronal ceroid lipofuscinosis**. *Lancet Neurol.* (2019) **18** 107-116. DOI: 10.1016/S1474-4422(18)30368-5 31. Cotman SL, Mole SE, Kohan R. **Future perspectives: moving towards NCL treatments**. *Biochim. Biophys. Acta* (2015) **1852** 2336-2338. DOI: 10.1016/j.bbadis.2015.04.001 32. Azizieh R. **Progressive myoclonic epilepsy-associated gene KCTD7 is a regulator of potassium conductance in neurons**. *Mol. Neurobiol.* (2011) **44** 111-121. DOI: 10.1007/s12035-011-8194-0 33. Metz KA. **KCTD7 deficiency defines a distinct neurodegenerative disorder with a conserved autophagy-lysosome defect**. *Ann. Neurol.* (2018) **84** 766-780. DOI: 10.1002/ana.25351 34. Pintard L, Willems A, Peter M. **Cullin-based ubiquitin ligases: Cul3-BTB complexes join the family**. *EMBO J.* (2004) **23** 1681-1687. DOI: 10.1038/sj.emboj.7600186 35. Krek W. **BTB proteins as henchmen of Cul3-based ubiquitin ligases**. *Nat. Cell Biol.* (2003) **5** 950-951. DOI: 10.1038/ncb1103-950 36. Skoblov M. **Protein partners of KCTD proteins provide insights about their functional roles in cell differentiation and vertebrate development**. *Bioessays* (2013) **35** 586-596. DOI: 10.1002/bies.201300002 37. Smaldone G. **Cullin 3 recognition is not a universal property among KCTD proteins**. *PLoS ONE* (2015) **10** e0126808. DOI: 10.1371/journal.pone.0126808 38. Schwenk J. **Native GABA(B) receptors are heteromultimers with a family of auxiliary subunits**. *Nature* (2010) **465** 231-235. DOI: 10.1038/nature08964 39. Liu Z, Xiang Y, Sun G. **The KCTD family of proteins: structure, function, disease relevance**. *Cell Biosci.* (2013) **3** 45. DOI: 10.1186/2045-3701-3-45 40. Ahmad KF, Engel CK, Privé GG. **Crystal structure of the BTB domain from PLZF**. *Proc. Natl. Acad. Sci. USA* (1998) **95** 12123-12128. DOI: 10.1073/pnas.95.21.12123 41. Stogios PJ, Downs GS, Jauhal JJS, Nandra SK, Privé GG. **Sequence and structural analysis of BTB domain proteins**. *Genome Biol.* (2005) **6** R82. DOI: 10.1186/gb-2005-6-10-r82 42. Tan P. **Recruitment of a ROC1-CUL1 ubiquitin ligase by Skp1 and HOS to catalyze the ubiquitination of I kappa B alpha**. *Mol. Cell* (1999) **3** 527-533. DOI: 10.1016/S1097-2765(00)80481-5 43. Ohta T, Michel JJ, Schottelius AJ, Xiong Y. **ROC1, a homolog of APC11, represents a family of cullin partners with an associated ubiquitin ligase activity**. *Mol. Cell* (1999) **3** 535-541. DOI: 10.1016/S1097-2765(00)80482-7 44. Kamura T. **Rbx1, a component of the VHL tumor suppressor complex and SCF ubiquitin ligase**. *Science* (1999) **284** 657-661. DOI: 10.1126/science.284.5414.657 45. Zheng N. **Structure of the Cul1-Rbx1-Skp1-F boxSkp2 SCF ubiquitin ligase complex**. *Nature* (2002) **416** 703-709. DOI: 10.1038/416703a 46. Arthur JS, Elce JS, Hegadorn C, Williams K, Greer PA. **Disruption of the murine calpain small subunit gene, Capn4: calpain is essential for embryonic development but not for cell growth and division**. *Mol. Cell. Biol.* (2000) **20** 4474-4481. DOI: 10.1128/MCB.20.12.4474-4481.2000 47. Zimmerman UJ, Boring L, Pak JH, Mukerjee N, Wang KK. **The calpain small subunit gene is essential: its inactivation results in embryonic lethality**. *IUBMB Life* (2000) **50** 63-68. DOI: 10.1080/15216540050176610 48. Nguyen V-N, Huang K-Y, Weng JT-Y, Lai KR, Lee T-Y. **UbiNet: an online resource for exploring the functional associations and regulatory networks of protein ubiquitylation**. *Database (Oxford)* (2016) **2016** baw054. DOI: 10.1093/database/baw054 49. Nelson WB, Smuder AJ, Hudson MB, Talbert EE, Powers SK. **Cross-talk between the calpain and caspase-3 proteolytic systems in the diaphragm during prolonged mechanical ventilation**. *Crit. Care Med.* (2012) **40** 1857-1863. DOI: 10.1097/CCM.0b013e318246bb5d 50. Ray SK, Matzelle DD, Wilford GG, Hogan EL, Banik NL. **Inhibition of calpain-mediated apoptosis by E-64 d-reduced immediate early gene (IEG) expression and reactive astrogliosis in the lesion and penumbra following spinal cord injury in rats**. *Brain Res.* (2001) **916** 115-126. DOI: 10.1016/S0006-8993(01)02874-8 51. Ray SK, Hogan EL, Banik NL. **Calpain in the pathophysiology of spinal cord injury: neuroprotection with calpain inhibitors**. *Brain Res. Brain Res. Rev.* (2003) **42** 169-185. DOI: 10.1016/S0165-0173(03)00152-8 52. Ono Y, Saido TC, Sorimachi H. **Calpain research for drug discovery: challenges and potential**. *Nat. Rev. Drug Discov.* (2016) **15** 854-876. DOI: 10.1038/nrd.2016.212 53. Forman OP, De Risio L, Mellersh CS. **Missense mutation in CAPN1 is associated with spinocerebellar ataxia in the Parson Russell Terrier dog breed**. *PLoS ONE* (2013) **8** e64627. DOI: 10.1371/journal.pone.0064627 54. Richard I. **Loss of calpain 3 proteolytic activity leads to muscular dystrophy and to apoptosis-associated IkappaBalpha/nuclear factor kappaB pathway perturbation in mice**. *J. Cell Biol.* (2000) **151** 1583-1590. DOI: 10.1083/jcb.151.7.1583 55. Mahajan VB. **Calpain-5 mutations cause autoimmune uveitis, retinal neovascularization, and photoreceptor degeneration**. *PLoS Genet.* (2012) **8** e1003001. DOI: 10.1371/journal.pgen.1003001 56. Bochner R. **Calpain 12 function revealed through the study of an atypical case of autosomal recessive congenital ichthyosis**. *J. Invest. Dermatol.* (2017) **137** 385-393. DOI: 10.1016/j.jid.2016.07.043 57. Baki A, Tompa P, Alexa A, Molnár O, Friedrich P. **Autolysis parallels activation of mu-calpain**. *Biochem. J.* (1996) **318** 897-901. DOI: 10.1042/bj3180897 58. Suzuki K, Sorimachi H. **A novel aspect of calpain activation**. *FEBS Lett* (1998) **433** 1-4. DOI: 10.1016/S0014-5793(98)00856-4 59. Inomata M, Imahori K, Kawashima S. **Autolytic activation of calcium-activated neutral protease**. *Biochem. Biophys. Res. Commun.* (1986) **138** 638-643. DOI: 10.1016/S0006-291X(86)80544-7 60. Lee MS. **Neurotoxicity induces cleavage of p35 to p25 by calpain**. *Nature* (2000) **405** 360-364. DOI: 10.1038/35012636 61. Moen MN. **Pathogenic variants in KCTD7 perturb neuronal K+ fluxes and glutamine transport**. *Brain* (2016) **139** 3109-3120. DOI: 10.1093/brain/aww244 62. Trinchese F. **Inhibition of calpains improves memory and synaptic transmission in a mouse model of Alzheimer disease**. *J. Clin. Invest.* (2008) **118** 2796-2807. DOI: 10.1172/JCI34254 63. Siklos M, BenAissa M, Thatcher GRJ. **Cysteine proteases as therapeutic targets: does selectivity matter? A systematic review of calpain and cathepsin inhibitors**. *Acta Pharm. Sin. B* (2015) **5** 506-519. DOI: 10.1016/j.apsb.2015.08.001 64. Zhang Z. **Therapeutic efficacy of E-64-d, a selective calpain inhibitor, in experimental acute spinal cord injury**. *Biomed. Res. Int.* (2015) **2015** 134242. PMID: 26240815 65. Teng X. **KCTD: a new gene family involved in neurodevelopmental and neuropsychiatric disorders**. *CNS Neurosci. Ther.* (2019) **25** 887-902. DOI: 10.1111/cns.13156 66. Golzio C. **KCTD13 is a major driver of mirrored neuroanatomical phenotypes of the 16p11.2 copy number variant**. *Nature* (2012) **485** 363-367. DOI: 10.1038/nature11091 67. Mencacci NE. **A missense mutation in KCTD17 causes autosomal dominant myoclonus-dystonia**. *Am. J. Hum. Genet.* (2015) **96** 938-947. DOI: 10.1016/j.ajhg.2015.04.008 68. **Integrated genomic and molecular characterization of cervical cancer**. *Nature* (2017) **543** 378-384. DOI: 10.1038/nature21386 69. Greif PA. **Identification of recurring tumor-specific somatic mutations in acute myeloid leukemia by transcriptome sequencing**. *Leukemia* (2011) **25** 821-827. DOI: 10.1038/leu.2011.19 70. Cathomas F. **Altered emotionality and neuronal excitability in mice lacking KCTD12, an auxiliary subunit of GABAB receptors associated with mood disorders**. *Transl. Psychiatry* (2015) **5** e510. DOI: 10.1038/tp.2015.8 71. De Smaele E. **Identification and characterization of KCASH2 and KCASH3, 2 novel Cullin3 adaptors suppressing histone deacetylase and Hedgehog activity in medulloblastoma**. *Neoplasia* (2011) **13** 374-385. DOI: 10.1593/neo.101630 72. Di Marcotullio L. **REN(KCTD11) is a suppressor of Hedgehog signaling and is deleted in human medulloblastoma**. *Proc. Natl. Acad. Sci. USA* (2004) **101** 10833-10838. DOI: 10.1073/pnas.0400690101 73. Alevy J, Burger CA, Albrecht NE, Jiang D, Samuel MA. **Progressive myoclonic epilepsy-associated gene Kctd7 regulates retinal neurovascular patterning and function**. *Neurochem. Int.* (2019) **129** 104486. DOI: 10.1016/j.neuint.2019.104486 74. Cao Y. **Autophagy is disrupted in a knock-in mouse model of juvenile neuronal ceroid lipofuscinosis**. *J. Biol. Chem.* (2006) **281** 20483-20493. DOI: 10.1074/jbc.M602180200 75. Palmieri M. **mTORC1-independent TFEB activation via Akt inhibition promotes cellular clearance in neurodegenerative storage diseases**. *Nat. Commun.* (2017) **8** 14338. DOI: 10.1038/ncomms14338 76. Leinonen H. **Retinal degeneration in A mouse model of CLN5 disease is associated with compromised autophagy**. *Sci. Rep.* (2017) **7** 1597. DOI: 10.1038/s41598-017-01716-1 77. Thelen M. **Disruption of the autophagy-lysosome pathway is involved in neuropathology of the nclf mouse model of neuronal ceroid lipofuscinosis**. *PLoS ONE* (2012) **7** e35493. DOI: 10.1371/journal.pone.0035493 78. Brandenstein L, Schweizer M, Sedlacik J, Fiehler J, Storch S. **Lysosomal dysfunction and impaired autophagy in a novel mouse model deficient for the lysosomal membrane protein Cln7**. *Hum. Mol. Genet.* (2016) **25** 777-791. DOI: 10.1093/hmg/ddv615 79. Mattson MP, Magnus T. **Ageing and neuronal vulnerability**. *Nat. Rev. Neurosci.* (2006) **7** 278-294. DOI: 10.1038/nrn1886 80. Liu S, Yin F, Zhang J, Qian Y. **The role of calpains in traumatic brain injury**. *Brain Inj.* (2014) **28** 133-137. DOI: 10.3109/02699052.2013.860479 81. Moretti D, Del Bello B, Allavena G, Maellaro E. **Calpains and cancer: friends or enemies?**. *Arch. Biochem. Biophys.* (2014) **564** 26-36. DOI: 10.1016/j.abb.2014.09.018 82. Yamashita T. **A role for calpain-dependent cleavage of TDP-43 in amyotrophic lateral sclerosis pathology**. *Nat. Commun.* (2012) **3** 1307. DOI: 10.1038/ncomms2303 83. Hübener J. **Calpain-mediated ataxin-3 cleavage in the molecular pathogenesis of spinocerebellar ataxia type 3 (SCA3)**. *Hum. Mol. Genet.* (2013) **22** 508-518. DOI: 10.1093/hmg/dds449 84. Gafni J, Ellerby LM. **Calpain activation in Huntington’s disease**. *J. Neurosci.* (2002) **22** 4842-4849. DOI: 10.1523/JNEUROSCI.22-12-04842.2002 85. Dufty BM. **Calpain-cleavage of alpha-synuclein: connecting proteolytic processing to disease-linked aggregation**. *Am. J. Pathol.* (2007) **170** 1725-1738. DOI: 10.2353/ajpath.2007.061232 86. Lon H-K. **Pharmacokinetics, safety, tolerability, and pharmacodynamics of alicapistat, a selective inhibitor of human calpains 1 and 2 for the treatment of alzheimer disease: an overview of phase 1 studies**. *Clin. Pharmacol. Drug Dev.* (2019) **8** 290-303. DOI: 10.1002/cpdd.598 87. Ennes-Vidal V, Menna-Barreto RFS, Branquinha MH, Dos Santos ALS, D’Avila-Levy CM. **Why calpain inhibitors are interesting leading compounds to search for new therapeutic options to treat leishmaniasis?**. *Parasitology* (2017) **144** 117-123. DOI: 10.1017/S003118201600189X 88. Strobl S. **The crystal structure of calcium-free human m-calpain suggests an electrostatic switch mechanism for activation by calcium**. *Proc. Natl. Acad. Sci. USA* (2000) **97** 588-592. DOI: 10.1073/pnas.97.2.588 89. Guroff G. **A neutral, calcium-activated proteinase from the soluble fraction of rat brain**. *J. Biol. Chem.* (1964) **239** 149-155. DOI: 10.1016/S0021-9258(18)51762-2 90. Metwally E, Zhao G, Zhang YQ. **The calcium-dependent protease calpain in neuronal remodeling and neurodegeneration**. *Trends Neurosci.* (2021) **44** 741-752. DOI: 10.1016/j.tins.2021.07.003 91. Swatek KN, Komander D. **Ubiquitin modifications**. *Cell Res* (2016) **26** 399-422. DOI: 10.1038/cr.2016.39 92. Yau R, Rape M. **The increasing complexity of the ubiquitin code**. *Nat. Cell Biol.* (2016) **18** 579-586. DOI: 10.1038/ncb3358 93. Husnjak K, Dikic I. **Ubiquitin-binding proteins: decoders of ubiquitin-mediated cellular functions**. *Annu. Rev. Biochem.* (2012) **81** 291-322. DOI: 10.1146/annurev-biochem-051810-094654 94. Komander D, Rape M. **The ubiquitin code**. *Annu. Rev. Biochem.* (2012) **81** 203-229. DOI: 10.1146/annurev-biochem-060310-170328 95. Kulathu Y, Komander D. **Atypical ubiquitylation - the unexplored world of polyubiquitin beyond Lys48 and Lys63 linkages**. *Nat. Rev. Mol. Cell Biol.* (2012) **13** 508-523. DOI: 10.1038/nrm3394 96. Liu K. **Altered ubiquitin causes perturbed calcium homeostasis, hyperactivation of calpain, dysregulated differentiation, and cataract**. *Proc. Natl. Acad. Sci. USA* (2015) **112** 1071-1076. DOI: 10.1073/pnas.1404059112
--- title: Development of a comfort suit-type soft-wearable robot with flexible artificial muscles for walking assistance authors: - Jiaoli Piao - Minseo Kim - Jeesoo Kim - Changhwan Kim - Seunghee Han - Inryeol Back - Je-sung Koh - Sumin Koo journal: Scientific Reports year: 2023 pmcid: PMC10038994 doi: 10.1038/s41598-023-32117-2 license: CC BY 4.0 --- # Development of a comfort suit-type soft-wearable robot with flexible artificial muscles for walking assistance ## Abstract Anchoring components are added to wearable robots to ensure a stable interaction between the suits and the human body and to minimize the displacement of the suits. However, these components can apply pressure to the body and can cause user dissatisfaction, which can decrease willingness to use the suits. Therefore, this study aims to develop a suit-type soft-wearable robot platform for walking assistance by providing comfortable garment pressure to ensure user satisfaction. The first prototype of a wearable robot suit was developed with anchoring components on the shoulders, waist, and thighs based on previous research results. Wear tests were conducted to measure garment pressure depending on posture using pressure sensors, and satisfaction surveys were conducted. The second prototype design was then developed, and performance tests with flexible artificial muscles and a satisfaction survey were conducted. Regarding the first prototype, the participants felt more than normal pressure in the shoulders and relatively less pressure in the thighs and calves. Thus, compared to the first design, the second design ensured a decreased garment pressure and resulted in an improvement of overall user satisfaction. These results can help provide guidance in the development of wearable robots by taking pressure comfort and user satisfaction into consideration. ## Introduction A wearable robot is a body-worn robotic system that offers high load, high mobility, and posture persistence capabilities through functions such as posture control, situational recognition, and posture-signal generation in response to various environments1,2. The global wearable robot-exoskeleton market is expected to grow with a compound annual growth (CAGR) of $43.6\%$, from $952.5 million in 2022 to $11,995.7 billion by 2029, with North America holding the largest market share and Asia’s market growing rapidly3. A wearable robot can assist human movements by driving actuators based on the movement intentions of the wearer1,2. Wearable robots were developed to make human life easier and to perform dangerous tasks efficiently; however, their rigidity and weight limit natural body movements. Furthermore, research has been lacking on wearable robots that take comfort into consideration from the early stages of the development process. This has led to the development of soft and flexible suit-type wearable robots, wherein anchoring methods are utilized to stabilize the wearable robots on the body4,5. The anchoring methods can be classified based on the methods and materials used: elastic forces of stretchy fabrics6, stitching techniques7, webbing straps8, and hook and loop fasteners9. However, anchoring components can create uncomfortable garment pressure, limit natural movement, and even cause blood circulation disorders and injuries, which can lead to a discontinuance of use of the wearable robots. Therefore, developing wearable robots that take pressure and user satisfaction into consideration is recommended. Garment pressure is the pressure generated by the contact between the garment and the human body, and it is attributed to the garment’s weight, the tightening or pulling of the garment material, and the deformation of stretchy fabrics to fit the shape of the human body. An appropriate level of garment pressure can provide positive attributes for the human body (e.g., support and recovery), whereas excessive pressure can hinder movement. Furthermore, garment pressure can cause increased venous blood flow, metabolite removal, and muscle oxygenation because of the reduced cross-sectional area of blood vessels10–13. It is also necessary to consider the fabric elongation recovery rate and dimensional change rate after washing because these factors can change the garment pressure. Compression can be categorized as mild (20 mmHg), medium (20–40 mmHg), strong (40–60 mmHg), and very strong (over 60 mmHg)14,15. A suitable pressure for comfort is less than 29.4 mmHg, and the discomfort threshold is approximately 44.10–73.50 mmHg14,16. If the compression pressure exceeds 50 mmHg, it may cause ambulatory venous hypertension reduction and vein occlusion while walking14,15. Therefore, the comfort zone of garment pressure is considered to be between 20 and 30 mmHg. Although various types of walking-assistive wearable robots are being developed, there is a lack of research on wearable robots that takes garment pressure and user satisfaction into consideration. Therefore, we attempted to develop suit-type soft wearable robot platforms for daily life walking assistance that are comfortable in terms of garment pressure and that can ensure wearer satisfaction. We expect that the results will help in the development of wearable robots that take pressure comfort and user wear satisfaction into account. ## Platform design and prototype development Different fabrics have unique features that can help a soft-wearable robot suit function better while providing comfort to the wearers. Stretchy fabric can provide stretchiness for comfort and can fit various sizes, while a mesh fabric can provide better ventilation; therefore, these fabrics were applied to areas such as the back of the knees that tend to become sweaty. Meanwhile, the nonstretchy fabric can assist in limiting the pulling power to minimize the dislocation of important components such as actuators, anchoring lines and other suitable details. Compression is achieved by the elongation of fibers and threads, and the elasticity of the material significantly affects the changes in pressure on the body18. Therefore, fabrics with excellent elongation-recovery rates and washing-dimension-change rates, which affect garment pressure, were used19, as shown in Fig. 1a. Fabric A is a stretchy fabric that has the highest tensile strength (wale 363.3 N, course 268.8 N). Fabric B is a wicking fabric located at the side and back of the pants, which are the areas prone sweatiness, and thus, the fabric can provide comfort. Fabric C is a nonstretchy fabric with high tensile strength; it was used at the waist to provide strong anchoring. Fabric D is 7 mm thick and has a high tensile strength of $256\%$ and 2317 N; therefore, it was used in the first design in the thigh and calf areas, where durability is required for walking. Fabric E is a band material with a sufficient elongation-recovery rate and high tensile strength, and it was used in the second design (Fig. 2a).Figure 1First design. ( a) Results of performance tests of applied fabrics. ( b) First design and specifications. ( c) Developed prototype based on the first design. Figure 2Second design. ( a) Second design and specification, (b) modified aspects of the second design from the first design. ( c) Developed second prototype. Two flexible supports of 10 cm (length) × 3 cm (width) × 0.1 cm (thickness) pieces were printed with thermoplastic polyurethane (TPU) with three 0.5-cm holes using a 3D printer (Shindo 3D WOX 2X). The supports were inserted into the waist belt, and the front harness was connected to the waist belt. The Boa dials were positioned in a straight line to enhance stability when the nitinol wires contracted and pulled at the suit for walking assistance, as shown in Fig. 1b,c. For the routing lines, the linear actuator consisted of nitinol wires with Teflon tubes, which crossed each other above the knees and past the sides of the knees at 10°. The wearers were able to adjust the linear actuator’s length with the Boa dial and the buckles to fit the actuator to their legs, and the harness was detachable. The first design was modified based on the garment pressure measurement and survey results, and the second design was developed as shown in Fig. 2. The two designs were developed to be size-adjustable for practical use and commercialization. Notably, customizations based on each person’s body size will highly increase the cost. Therefore, to achieve the size-adjustable function, the two prototypes used Boa dials, buckles and slides, hooks and loops, webbing bands, stretchy fabrics, and elastic bands to change the lengths and widths of the suits as desired. The horizontal webbing band and the buckle crossing the chest helped adjust the width of the torso. The waist band could be adjusted using the hooks and loops, and all anchoring parts were adjustable using the slides. Adjustable elastic bands were placed on the thighs, where the large muscles change body size more frequently than in other areas, to anchor the devices while ensuring comfort. The lengths of the routing lines of the actuator could be adjusted with the Boa dial, and the webbing strap on the pants hemlines were length-adjustable with the webbing straps and slides. The stretchy fabric allowed adjustments for the pants overall in terms of the width and the length, and the nonstretchy fabric was used for the back of the torso back to provide strong support. ## Textile-based flexible artificial muscle design for ankle assistance Shape memory alloys (SMAs) exhibit hysteresis behavior because of a nonlinear thermoelastic material response20. SMAs have the drawback of low controllability and inaccurate positioning21; however, differences in Young’s modulus of the austenite and martensite phases can provide more comfortability in user movements (i.e., when the SMA is extended while the user swings their legs). Therefore, hysteresis behavior in the SMA would be helpful for both actuating and releasing states. A high modulus in the actuation state provides high force assistance for the ankle, and a low modulus in the releasing state reduces the disturbance in stepping on the ground. The cooling rate of the SMA is determined by thermal diffusion and convection. Many studies have been reported to reduce cooling time using water, fans, and lower diameter SMA22–24. We attempted to find the optimal fabrics integrated with SMA to reduce cooling time and adopted a small diameter (0.006 mil) to enable assisted walking (Fig. 3a,b).Figure 3Characterization of the SMA actuator and conceptual actuator design of a suit-type soft wearable robot. ( a) Schematic diagram of the experimental setup for tensile testing. ( b) Measured 0.006-mil SMA blocked force as functions of strain and input current. ( c) Overview of the suit-type soft wearable robot. ( d) The cooling and heating of the SMA actuators. ( e) The torque inducing the plantar flexion of the ankle. We designed a textile-based flexible artificial muscle actuation system to provide parallel assistance to the plantar flexor muscles (soleus muscle) during normal gait, as shown in Fig. 3c. Figure 3c shows routing lines and anchor points on the robot for driving torque at the ankle joint. Webbing straps (gray) are placed and fixed on the waist, knee, and calf joint to effectively transmit the contraction force generated by SMA actuators to the ankle. Each end of the SMA wire is anchored to the webbing strap at body parts that can endure high force and stress. The SMA actuators require a simple electronics system and have their own unique characteristics of shape memory effects (SMEs)25. Compared to other soft actuators, the SMA actuator can produce high force, and compared to electromagnetic motors, it has advantages such as silent actuation, flexibility, and a low profile. Therefore, SMA is suitable as a potential artificial muscle for use in a suit-type soft wearable robot for daily life applications. The large reversible deformation of the SMA is caused by changes in the crystalline arrangement, and it depends on the temperature and state of stress attributed to the native capability of the SMA. The SMA actuator can be stimulated via Joule heating, as shown in Fig. 3d. The anchor point on the waist was used to hold the wire so that it did not fall off or slip during actuation. The actuation displacement occurred behind the Achilles tendon. The strap on the heel was pulled by the actuator, and the SMA wire’s contracting force acted in parallel with the soleus muscle, which generated a driving torque on the ankle joint. The torque induced the plantar flexion of the ankle, as shown in Fig. 3e. The electric power conditions for actuation were 71.0 W (1.6 A and 44.4 V) in all experiments. Notably, exposed electrical wires and/or SMA actuators can cause injury such as electric shock and skin burn to the user; however, our actuator was intrinsically safe and unable to cause these injuries because the actuation time was short (approximately 200 ms), which means the applied heat energy is very small, and the encapsulated structure with the Teflon tube insulates the heat of the SMA wire. Therefore, the user is guaranteed safety from skin burns even after cyclic actuation. The biomechanical analysis in previous work26 indicated that the required actuation frequency for walking assistance was approximately 1 Hz. We conducted a study on design parameters (diameter of SMA, heating time, and cooling time) to satisfy the actuation speed of 1 Hz. The user’s intention can be detected by feedback information from sensors attached on the user’s feet, which includes a force-sensitive resistor (FSR), inertial measurement unit (IMU), and electromyography (EMG) sensor. ## Garment pressure measurement The following experiments were conducted with participants: garment pressure measurement and satisfaction survey, which were approved by the institutional review board (IRB) at Konkuk University. All methods, including experimental protocols, were carried out in accordance with relevant guidelines and regulations of the IRB. All participants provided informed consent before the experiment. The garment pressures of the first and second designs were measured while the participants wore them. Furthermore, the differences in the designs with the presence and absence of a harness were compared. A weight of 2 kg was added, and the suit was tested to simulate the actuator performance on the ankle. Compression area, posture change, and wearing duration can affect garment pressure14–17, and therefore, these factors were considered when measuring pressure. The pressures were measured at the shoulders, waist, front and back thighs, front lower thigh, and front and back calves27,28 (Fig. 4a) when the participants stood, walked, and bent their knees29,30, as shown in Fig. 4b. The second design was changed in comparison to the first design, and stitches were added; the front lower thigh was only measured for the first design. Figure 4Garment pressure measurement process. ( a) The garment pressure measuring points, with a total of eight different points. ( b) Postures for measuring the garment pressure. ( c) Overview of the garment pressure measuring system. The existing research on measuring garment pressure showed that researchers had experimented with different ranges of height and weight with one garment (within the normal body mass index (BMI) range). Thus, we referred to existing research methods and checked if they were in the normal range of BMI. After our convenience sampling, male volunteers in their 20 s and 30 s with an average BMI participated in the experiment. The mean height, weight, and BMI of the 12 participants were 174.3 cm (standard deviation (SD) = 4.24 and range (R) = 171–182), 68.12 kg (SD = 4.48, $R = 60$–73), and 22.65, respectively, which were in the normal range of 19–24. We conducted these measurements at least 2 h after the participants had eaten30. The participants were asked to adjust the anchoring parts so that the parts felt simultaneously comfortable and supportive as they would in real life. The researchers also checked if the anchoring parts fit and were sufficiently tight for appropriate functioning, if the important parts and SMA wire lengths were placed next to the target body location, and if the prototype was worn properly. The garment pressure was measured using the air-pack method equipment (AMI 3037-2-2B, Sanko Tsusho), as shown in Fig. 4c. Each participant wore a short-sleeved T-shirt ($100\%$ cotton) so that the amount of clothing on the upper body was the same for all participants. Then, measurements were started after checking the stability of the participants while standing or sitting for 30 min29. The test environmental conditions were a temperature of 25 ± 1℃, humidity of 65 ± $10\%$ RH, and airflow of 0.5 m/s29, and points were marked at the measurement locations and attached sensors when measuring. The garment pressures were measured 2–3 times for each movement, and participants had 10-min rests between each posture30. Each part was measured for a total of 60 s. Based on previous research31, 10 s were eliminated from the beginning and from the end of the collected data for noise removal, such that the average value for a duration of 40 s was analyzed. ## Satisfaction evaluation A face-to-face interview survey was conducted with the 12 participants26. A total of 43 questions were asked, including 9 demographic questions (gender, age, height, weight, residence area, occupation, health status, garment-wearing habits) and 28 questions about subjective garment pressure (shoulder, waist, front thigh, back thigh, front lower thigh, front calf, back calf, 2-kg weight)30,31. The 16 questions on design satisfaction consisted of questions on durability, functionality, design, materials, and purchase and use intentions26,33–35 and were based on a 5-point Likert scale, but also included an open-ended question for suggestions. The Cronbach’s alpha value was 0.99 and showed high reliability (> 0.80), and internal consistency was confirmed. Descriptive analyses, such as frequency and percentage, and t tests were performed using SPSS 25.0. ## Performance test of the wearable robot This experiment was performed with an ankle-articulated mannequin wearing our robots, as shown in Fig. 5a, to evaluate the robot performance. In this experiment, the ankle range of motion and actuation control performance were evaluated. For this purpose, wire-type SMA actuators with a 0.006-in diameter were embedded to generate an assistive force for ankle motion. Power and signal input were controlled by a microcontroller unit (Uno R3; Arduino). The actuators were powered by the input current of the power supply, and ON–OFF switch control was performed by the controller unit and metal–oxide–semiconductor field-effect transistor (MOSFET) logic circuit. SMA actuators 600 mm long were wired at both ends, taking into consideration the $4\%$ strain of the wire-type SMA (Flexinol® actuator wire; Dynalloy).Figure 5Testing setup. ( a) Overview of the testing setup for cycling actuation and performance testing of the wearable robot. The scale bar in (a) represents 200 mm. ( b) The sagittal plane ankle angle was measured by the motion capture system. The scale bar in (b) represents 100 mm. ( c) Distributed forces on the garment caused by actuating force. Commercially available SMA wire (Flexinol® by Dynalloy, Inc.) was used in all experiments. The strain of the SMA wire is typically 3–$5\%$ of the i–s original length. In addition, we studied the mechanical and electrical characteristics of the SMA actuator from previous work26 and found the stable strain range of the SMA wire. The length of the SMA wire can be adjusted using the Boa dial, depending on the user, and therefore, the desired stress and force to a soleus muscle can be fit according to the user. The actuation stroke of the SMA wire is determined by its original length. Based on the analysis of ankle motions during normal walking26, we designed the SMA wire to have an actuation displacement of 2–3 cm, which is 60 cm by wiring, near the knee joint. A variation in the actuation stroke from the inherent differences in height and/or body proportion may cause a slight difference in the actuation force and stroke; however, it can be compensated for by compliance with the SMA actuator. If there is a large dimensional difference in users, the actuation stroke can be tailored by changing the wired points. In the cyclic actuation test, a spring that could generate an antagonistic force was installed to recover the angle of the ankle after actuation, as shown in Fig. 5b. We measured the angular change over time while applying electrical power input to the SMA wire after adjusting the position of the foot parallel to the ground. The sagittal plane ankle angle was measured using the positions of the three reflectors tracked by the motion capture system (Primex 13; Optitrack). The forces are represented by red and orange arrows in Fig. 5c. Red indicates the force direction from the additional mass near the distal ankle joint, and orange indicates the reaction forces near the anchor. Light red lines depict the routing path of the shape-memory alloy actuator. Finally, the white circles indicate anchor points at the shoulder, waist, thigh, and ankle joint. The SMA wire connected from the ankle to the waist generates a contraction force when it is actuated via Joule heating, which results in tugging on the garment. Body parts with larger bones, such as the shoulder, hip and plantar aspects of bilateral feet36, support most of the carrying mass and body. The actuating force from the SMA actuator is supported by the anchored waist, as shown in Fig. 5c. The waist textile and shoulder strap (harness) are physically connected by hooks, which consequently supports the substantial loads at the shoulder. Therefore, forces generated by the SMA wire for the ankle joint are distributed throughout the anchor on the entire body, which helps prevent stress and slippage on the local areas of the body during actuation. Based on this analysis, we measured the garment pressure on the various anchor points and body parts (white dots in Fig. 5c) by suspending 2 kg of constant load (corresponding to the maximum actuation force) on the distal ankle. Pressures on the various anchor points and body parts increase ($80\%$ increment at the shoulder and waist, which are the main supporting body parts). These experiments indicate that actuation forces were properly distributed across the entire body by our routing design, and a meaningful correlation was found between the actuation force and the measured garment pressure. The weight for the garment pressure measurement was determined by the force generated by the SMA. Based on the results in previous work26, the maximum actuating force was set to 21 N. Therefore, the experiment was conducted by suspending a weight of 2 kg at the distal ankle joint to identify the correlation between garment pressure and actuating forces. In this experiment, SMA wires with a diameter of 0.006 inches were used to assist walking. The diameter of the SMA is a key design parameter that determines various aspects of actuator performance, such as actuation force and frequency. An SMA with a larger diameter can provide a high actuation force, but it has a low actuation frequency. The SMA wire’s smaller diameter provides a low actuation force but has a high frequency. We adopted the SMA with a 0.006-inch diameter to achieve a frequency sufficient for the walking gait, which was suitable for fast actuation at approximately 1 Hz of the walking cycle. ## First design: garment pressure measurement The garment pressures were lowest when participants were standing, increased when participants were walking, and were highest when the participants bent their knees, as shown in Fig. 6a. There were more garment shape changes when participants walked or bent over than when they were standing. The garment pressures were the highest at the front thigh (40.49 mmHg), front calf, and back thigh when participants bent their knees. Figure 6Garment pressure. ( a) Results of the measured garment pressure and satisfaction-measured garment pressure by posture. ( b) Measured garment pressure before and after the 2-kg weight was added. ( c) Garment pressure with and without the harness. On average, the front calf had the highest pressure (37.51 mmHg), followed by the front lower thigh, back thigh, back calf, front thigh, shoulder, and waist. The garment pressure at all measured body areas increased when the 2-kg weight was added to simulate the actuator’s activation and to pull on the pants, as seen in Fig. 6b. The pressure at the shoulder increased the most ($101.40\%$), followed by the waist ($80.91\%$), back thigh ($37.12\%$), back calf ($31.29\%$), front calf ($21.99\%$), front thigh ($10.24\%$), and front lower thigh ($9.05\%$). After the removal of the harness, the garment pressure increased at the waist (+ $36.23\%$), back thigh (+ $21.96\%$), and back calf (+ $9.35\%$); however, the garment pressure decreased at the front thigh (− $9.65\%$), front lower thigh (− $0.14\%$), and front calf (− $14.26\%$), as shown in Fig. 6c. ## First design: satisfaction evaluation Participants evaluated their satisfaction with the first design in terms of wearability, function, design, material, and purchase intentions (Fig. 7). Overall, the participants were satisfied with the first design except for the overall design and size/fit ($M = 2.91$ for each). The most satisfactory aspect was wearability with respect to noninterference with movement ($M = 4.18$, SD = 0.75), followed by static posture assistive function and durability. In the design, the anchoring methods were the most satisfactory aspects, and the materials were also satisfactory ($M = 3.09$–3.55). In the open-ended question, participants wanted to increase the crotch length and to add layers to the crotch parts so they could wear the garment as daily wear. Furthermore, the purchase intention was higher than the wear intention, which may be because of dissatisfaction with the overall design, size/fit, and garment pressure. For the second design, the participants suggested an increase in crotch length and modified two-in-one pants. Figure 7First design: satisfaction evaluation. ## Second design: garment pressure measurement The second design and prototype were developed based on the results for the previous design (Fig. 2). First, all measured garment pressures for the second design were lower than those for the first design. The area with the greatest decrease was the back thigh (33.24 → 13.55 mmHg; − $59.24\%$), followed by the front calf (37.51 → 17.07 mmHg; − $54.49\%$), front thigh (28.90 → 17.63 mmHg; − $39.00\%$), shoulder (20.78 → 12.76 mmHg; − $38.59\%$), back calf (29.95 → 23.79 mmHg; − $20.57\%$), and waist (11.26 → 11.16 mmHg; − $0.89\%$). These decreases indicate that the design changes decreased the garment pressure (0.89–$59.24\%$) and that the garment pressures for all areas were lowered; no part showed an increased level of garment pressure. Second, garment pressures for the second design ranged from 20.27 to 29.75 mmHg when weight was added (Fig. 8b). These values were lower than those for the first design (20.37–45.76 mmHg). The shoulder showed the greatest increase ($48.98\%$), followed by the back thigh ($47.03\%$), waist ($44.94\%$), front calf ($42.62\%$), back calf ($19.95\%$), and front thigh ($19.35\%$).Figure 8Results of measured garment pressure and satisfaction. ( a) Measured garment pressure by posture. ( b) Measured garment pressure before and after the 2-kg weight was added. ( c). Garment pressure with and without the harness. Third, the harness of the second design decreased the garment pressure for all measured body areas (Fig. 8c). The greatest changes after the removal of the harness were at the waist (+ $57.44\%$), back thigh (+ $28.93\%$), back calf ($20.26\%$), front calf ($13.06\%$), and front thigh ($6.47\%$). ## Second design: satisfaction evaluation A comparison of the levels of satisfaction between the first and second designs revealed that satisfaction increased with regard to all aspects for the second design (Fig. 9). After the t test, the second design showed a significantly increased level of satisfaction with wearability (t = − 3.225, $p \leq 0.01$), function (t = − 7.147, $p \leq 0.001$), design (t = − 3.830, $p \leq 0.01$), and material (t = − 3.601, $p \leq 0.001$) compared to these measures for the first design. The unsatisfactory aspects of the first design, such as overall design and size/fit, were improved to satisfactory for the second design. Both purchase and wear intentions increased for the second design compared to those for the first design (t = − 3.225, $p \leq 0.001$).Figure 9Differences in satisfaction between the first and second designs. ## Performance test results Figure 10a shows the result of cyclic actuation with an electric input of 1.6 A for both suit designs. A single actuation cycle has a heating time of 200 ms and a cooling time of 800 ms. In this experiment with the testbed (Fig. 10b), the maximum rotation angles are 11.8° in the first design and 15.8° in the second design. Figure 10Results of the cyclic actuation and performance tests for the suit-type soft wearable robots. ( a) Cyclic actuation with an electric input of 1.6 A for the first and second designs. ( b) Average rotation angles of the first and second designs. The results for the first design indicated that a high actuation force generated by the SMA caused the deformation of the garment to create slippage. Thus, the anchoring parts on the waist and heel were subsequently used to reduce slippage during actuation. The performance of the second design was improved by using stiffer band fabrics (Fabric E) and increasing the width of the 3D-printed waist support to reduce the deformation of the garment. ## First design: garment pressure measurement and satisfaction evaluation First, the waist was at the suggested level of approximately or less than 20 mmHg14,15; however, the other areas were higher than that. The shoulder was under comfort garment pressure for a few scenarios, such as for bending the knee before adding the 2-kg weight and with a harness. We recommend modifying the first design to lower the garment pressure, especially at the front calf and the lower front and back thighs, which had garment pressures higher than the comfort range of 20–30 mmHg14. Second, because the shoulder area holds the weight, it is necessary to have shoulder components. It is advisable to decrease the garment pressure, especially on the front calf, back thigh, and shoulder. The front lower thigh showed a low garment pressure because the fixing force of the belt decreased because of the added weight. Third, the magnitude of the increases in pressure was greater than that of the decreases in pressure. Thus, the harness helped decrease the garment pressure at the waist, back thigh, and calf. This may be because the actuator was on the back calf area. This can help decrease the garment pressure on the back side of the suits. However, it is also advisable to decrease the garment pressure in the front areas. ## Design modifications: second design and prototype First, the front calf and the front and back thighs showed the top three highest garment pressures, with the X-shaped anchoring band being fixed in the pants seams. Second, the shoulder showed the greatest increase in garment pressure when weight was added, followed by the waist. Therefore, we modified the shoulder and waist areas by changing the waist band fabric and adding a wider and rounded harness to better support the shoulders. Third, in the open-ended questions, participants suggested two-in-one pants to cover the fitted leggings so they could be worn on a daily basis. Thus, a short pants design was added in the second design for a better design. Fourth, the Teflon tube started from the Boa dials and passed through the connection hole of the shorts, and it was fixed to the surface of the long pants. Notably, the equipment that weighs the most, such as batteries and drivers, requires the wearer to have a relatively large surface area to relieve pressure. Therefore, the pocket for inserting electronic components, such as batteries, was located between the waist and the hip to relieve weight pressure without restricting movement37. ## Second design: garment pressure measurement and satisfaction evaluation In this design, we focused primarily on modifying areas other than the waist because the waist had an acceptable pressure level in the first design; thus, the waist showed the least change in the second design. All garment pressures (11.16–23.79 mmHg) were lower than the range of the comfort zone (20–30 mmHg), with the back calf (23.79 mmHg) having the highest pressure, and where the actuator was placed with high stability (Fig. 8a). Thus, the second design showed an improvement over the first design with respect to garment pressure, and it could be further modified to lower the back calf garment pressure while maintaining the stability of actuators in a future design. For the first design, the garment pressure decreased in the shoulder and waist but increased at the front thigh, front calf, and back thigh. Thus, compared to that in the first design, the garment pressure on the shoulder and waist areas in the second design was effectively distributed to the front thigh, front calf, and back thighs. In contrast, in the first design the harness increased the garment pressures on the waist, back thigh, and back calf. This supports the conclusion that the harness of the second design helped decrease the garment pressure at the waist, back thigh, and back calf. It is thus advisable to employ a harness to distribute the garment pressure for the suit-type soft wearable robot platform. Overall, the satisfaction with the second design was higher than that with the first design, including purchase intention. ## Design guidelines The following design guidelines are suggested. First, it is better to design garments to be like those worn in daily life, such as in the style of two-in-one pants. Second, for movement comfort and a better fit, it is recommended to make size-adjustable components such as Boa dials, buckles and slides, hooks and loops, webbing bands, stretchy fabrics, and elastic bands. Third, to make the garment more comfortable to wear while moving around, side slits can be added to the pants, and pocket pouches with electrical circuits can be placed on the waist back. Fourth, a wider and rounder harness on the shoulder can enhance stability and comfort. Fifth, to reinforce the strength and durability, a two-layered harness and a waist band can be used. Next, for ease of wearing and taking off, a front concealed zipper is suggested. Finally, it is recommended to use stretchy fabric overall, with mesh fabric for zones that are prone to sweatiness, and to use non-stretchy fabrics for the front waist and harness to withstand pulling power. ## Conclusion In this study, two versions of suit-type walking-assistive wearable robot platforms were developed, and garment pressure and satisfaction after wearing were evaluated. The second design decreased the garment pressure and increased user satisfaction, and it was therefore recommended for the suit-type platform. The garment pressures for the second design were all lower than those for the first design. The garment pressure when standing was the lowest, and the overall garment pressures increased progressively when participants walked and bent their knees. This shows that movements with more extreme angles increase garment pressure. When weight was added, the garment pressure increased less in the second design and remained lower than that in the first design. The garment pressures applied to the shoulder and waist areas were better distributed to the front thigh, front calf, and back thighs in the second design than in the first design. The garment pressure of the second design increased for all measured areas when the harness was removed. The greatest changes were at the waist, back thigh, and back calf. It is therefore suggested to use a harness to distribute the garment pressure for the suit-type soft wearable robot. User satisfaction increased greatly for the second design compared to that for the first design (0.01 < $p \leq 0.001$). The function increased the most, the previously unsatisfactory aspects of overall design and size/fit became satisfactory, and the intentions concerning purchase and wear improved for the second design compared to those for the first design. Finally, the second design showed larger maximum rotation angles than those of the first design. The high actuation force generated by the SMA caused the deformation of the garment to create slippage, while the anchoring components of the second design reduced slippage during actuation. This research provides strategies for designing suit-type walking-assistive wearable robot platforms to provide comfortable garment pressure and to satisfy wearers. The modified second design showed decreased garment pressure and higher satisfaction than the first design. We hope that the developed designs and guidelines will help in the development of wearable robots and that continuous modification will have a positive effect on the commercialization of wearable robots by increasing wearer satisfaction. All garment pressures were lower than the suggested level for comfort except at the back calf, where the actuator is located. Future studies can focus on how to decrease the garment pressure on the actuator locations as well. It is recommended different sizes of suits be developed for diverse body sizes to see if there are any differences in results with size-adjustable suits. Furthermore, different types of suits should be designed with different actuator types for diverse users to fit adequately and to accommodate different needs related to gender, age, and targeted body areas. This study is expected to help in the development of comfortable-to-wear and satisfactory suit-type soft wearable robots by providing design strategies and development processes for wearable robot developers. ## References 1. Walsh C. **Human-in-the-loop development of soft wearable robots**. *Nat. Rev. Mater* (2018.0) **3** 78-80. DOI: 10.1038/s41578-018-0011-1 2. Low KH, Liu X, Yu H. **Development of NTU wearable exoskeleton system for assistive technologies**. *Mech. Auto.* (2005.0) **2005** 1099-1106 3. 3.Straits Research. Global exoskeleton wearable robots market analysis till 2029. straitsresearch.com. https://straitsresearch.com/report/exoskeleton-wearable-robots-market/ (2022). 4. Park SJ, Park CH. **Suit-type wearable robot powered by shape-memory-alloy-based fabric muscle**. *Sci. Rep.* (2019.0) **9** 9157. DOI: 10.1038/s41598-019-45722-x 5. Park SJ, Choi K, Rodrigue H, Park CH. **Fabric muscle with a cooling acceleration structure for upper limb assistance soft exosuits**. *Sci. Rep.* (2022.0) **12** 11398. DOI: 10.1038/s41598-022-15682-w 6. Cornwall W. **In pursuit of the perfect power suit**. *Science* (2015.0) **350** 270-273. DOI: 10.1126/science.350.6258.270 7. 7.Mengüç, Y., Park, Y.L., Martinez-Villalpando, E., Aubin, P., Zisook, M., Stirling, L., Wood, R.J. & Walsh, C.J., Soft wearable motion sensing suit for lower limb biomechanics measurements. Int. Conf. Rob. Auto. 5309–5316 (2015). 8. 8.Anderson, A., Waldie, J. & Newman, D. Modeling and design of a biosuit donning system. Int. Conf. Environ. Syst.6213 (2010). 9. 9.Awad, L.N., Bae, J., O’donnell, K., De Rossi, S.M., Hendron, K., Sloot, L.H., Kudzia, P., Allen, S., Holt, K.G., Ellis, T.D. & Walsh, C.J. A soft robotic exosuit improves walking in patients after stroke. Sci. Trans. Med.9, eaai9084 (2017). 10. Bringard A, Denis R, Belluye N, Perrey S. **Effects of compression tights on calf muscle oxygenation and venous pooling during quiet resting in supine and standing positions**. *J. Sports Med. Phys. Fitness* (2006.0) **46** 548. PMID: 17119519 11. Mayrovitz HN, Sims N. **Effects of ankle-to-knee external pressures on skin blood perfusion under and distal to compression**. *Adv. Skin Wound Care* (2003.0) **16** 198-202. DOI: 10.1097/00129334-200307000-00013 12. Stanton JR, Freis ED, Wilkins RW. **The acceleration of linear flow in the deep veins of the lower extremity of man by local compression**. *J. Clin. Invest.* (1949.0) **28** 553-558. DOI: 10.1172/JCI102104 13. Troynikov O, Ashayeri E, Burton M, Subic A, Alam F, Marteau S. **Factors influencing the effectiveness of compression garments used in sports**. *Proc. Eng.* (2010.0) **2** 2823-2829. DOI: 10.1016/j.proeng.2010.04.073 14. Xiong Y, Tao X. **Compression garments for medical therapy and sports**. *Polymers* (2018.0) **10** 2823-2829. DOI: 10.3390/polym10060663 15. Partsch H, Clark M, Mosti G, Steinlechner E, Schuren JAN, Abel M, Benigni JP, Coleridge-Smith P, Cornu-Thénard A, Flour M, Hutchinson J. **Classification of compression bandages: Practical aspects**. *Dermatol. Surg.* (2008.0) **34** 60-609. PMID: 18053049 16. Denton MJ. **Fit, stretch, and comfort**. *Textiles* (1972.0) **3** 12-17 17. Tanaka S, Midorikawa T, Tokura H. **Effects of pressure exerted on the skin by elastic cord on the core temperature, body weight loss and salivary secretion rate at 35 C**. *Eur. J. Appl. Physiol.* (2006.0) **96** 471-476. DOI: 10.1007/s00421-005-0099-z 18. 18.McLoughlin, J. & Sabir, T. High-performance apparel: materials, development, and applications. (Woodhead Publishing, 2017). 19. Park S, Kim S, Sim K, Piao J, Han R, Kim S, Koo S. **Development of suits for upper-body movement-assistive wearable robots for industrial workers**. *Text. Res. J.* (2022.0) **92** 3261-3276. DOI: 10.1177/00405175211041717 20. 20.Reynolds, D.R. & Kloucek, P. Numerical modeling of nonlinear thermodynamics in SMA Wires. (Lawrence Livermore National Lab, 2004). 21. Jani JM, Leary M, Subic A, Gibson MA. **A review of shape memory alloy research, applications and opportunities**. *Mater. Des.* (2014.0) **56** 1078-1113. DOI: 10.1016/j.matdes.2013.11.084 22. 22.Jeong, J., Yasir, I.B., Han, J., Park, C.H., Bok, S.K. & Kyung, K.U. Design of shape memory alloy-based soft wearable robot for assisting wrist motion. Appl. Sci.9(19), 4025 (2019). 23. Park SJ, Park CH. **Suit-type wearable robot powered by shape-memory-alloy-based fabric muscle**. *Sci. Rep.* (2019.0) **9** 1-8. PMID: 30626917 24. Kim D, Kim B, Shin B, Shin D, Lee CK, Chung JS, Seo J, Kim YT, Sung G, Seo W, Kim S. **Actuating compact wearable augmented reality devices by multifunctional artificial muscle**. *Nat. Comm.* (2022.0) **13** 1-13 25. Tanaka K, Kobayashi S, Sato Y. **Thermomechanics of transformation pseudoelasticity and shape memory effect in alloys**. *Int. J. Plast.* (1986.0) **2** 59-72. DOI: 10.1016/0749-6419(86)90016-1 26. Kim C, Kim G, Lee Y, Lee G, Han S, Kang D, Koo SH, Koh JS. **Shape memory alloy actuator-embedded smart clothes for ankle assistance**. *Smart Mater. Struct.* (2020.0) **29** 5. DOI: 10.1088/1361-665X/ab78b5 27. Zuo K, Wang Y, Li H, Wu C, Wen H, Zhu W. **A study on the shaping efficacy grading model of underbust corsets based on three-dimensional body scanning**. *Text. Res. J.* (2022.0) **92** 3680-3692. DOI: 10.1177/00405175221091577 28. Liu K, Wang J, Hong Y. **Wearing comfort analysis from aspect of numerical garment pressure using 3D virtual-reality and data mining technology**. *Int. J. Clothing Sci. Technol.* (2017.0) **29** 166-179. DOI: 10.1108/IJCST-03-2016-0017 29. Mengna G, Kuzmichev VE. **Pressure and comfort perception in the system “female body–dress”**. *AUTEX Res. J.* (2013.0) **13** 71-78. DOI: 10.2478/v10304-012-0032-6 30. Cho SH. **Analysis of clothing pressure for commercially customized burn patient's medical compression garments for men in their 20s**. *J. Kor. Fashion Costume Des. Assoc.* (2019.0) **21** 55-67 31. Song KH, Kim JH, Park SH. **Development and application of measurement system for clothing pressure**. *Kor J. Human Eco.* (2002.0) **11** 307-319 32. Kim NY, Hong K. **A screening test of extensibility and the prediction of clothing pressure for commercial compression T-shirts**. *J. Kor. Soc. Clothing Tex.* (2017.0) **41** 1010-1021. DOI: 10.5850/JKSCT.2017.41.6.1010 33. 33.Bae, S.H., Kim, B.G., Kim, M.S. & Chung, Y.C. Practical multiplexing techniques for next-generation data center network. In 22nd Int. Conf. Transparent Opt. Networks (ICTON) 1–4 (2020). 34. Chae UR, Kim K, Choi J, Hyun DJ, Yun J, Lee GH, Hyun YG, Lee J, Chung M. **Systematic usability evaluation on two harnesses for a wearable chairless exoskeleton**. *Int. J. Ind. Ergo.* (2021.0) **84** 103162. DOI: 10.1016/j.ergon.2021.103162 35. 35.Meyer, J.T., Schrade, S.O., Lambercy, O. & Gassert, R. User-centered design and evaluation of physical interfaces for an exoskeleton for paraplegic users. In 16th Int. Conf. Rehab. Robotics (ICORR), 1159–1166 (2019). 36. Wehner M, Quinlivan B, Aubin PM, Martinez-Villalpando E, Baumann M, Stirling L, Holt K, Wood R, Walsh C. **A lightweight soft exosuit for gait assistance**. *IEEE Int. Conf. Rob. Auto.* (2013.0) **1** 3362-3369 37. Gemperle F, Kasabach C, Stivoric J, Bauer M, Martin R. **Design for wearability**. *Int. Symp. Wear. Comp.* (1998.0) **1** 116-122
--- title: Distribution and association of interpregnancy weight change with subsequent pregnancy outcomes in Asian women authors: - Chee Wai Ku - Tuck Seng Cheng - Chee Onn Ku - Kathy Xinzhuo Zhou - Yin Bun Cheung - Keith M. Godfrey - Wee Meng Han - Fabian Yap - Jerry Kok Yen Chan - See Ling Loy journal: Scientific Reports year: 2023 pmcid: PMC10039003 doi: 10.1038/s41598-023-31954-5 license: CC BY 4.0 --- # Distribution and association of interpregnancy weight change with subsequent pregnancy outcomes in Asian women ## Abstract The extent of interpregnancy weight change and its association with subsequent pregnancy outcomes among Asians remain unclear. We examined changes in maternal body mass index (BMI) between the first two deliveries and outcomes in the second delivery. Medical records of women with their first two consecutive deliveries between 2015 and 2020 at KK Women’s and Children’s Hospital, Singapore were retrieved. Gestational-age-adjusted BMI was determined by standardising to 12 weeks gestation and interpregnancy BMI change was calculated as the difference between both pregnancies. Pregnancy outcomes were analysed using modified Poisson regression models. Of 6264 included women with a median interpregnancy interval of 1.44 years, $40.7\%$ had a stable BMI change within ± 1 kg/m2, $10.3\%$ lost > 1 kg/m2, $34.3\%$ gained 1–3 kg/m2 and $14.8\%$ gained ≥ 3 kg/m2. Compared to women with stable BMI change, those with > 1 kg/m2 loss had higher risk of low birthweight (adjusted risk ratio [RR] 1.36; $95\%$ confidence interval 1.02–1.80), while those with 1–3 kg/m2 gain had higher risks of large-for-gestational-age birth (1.16; 1.03–1.31), gestational diabetes (1.25; 1.06–1.49) and emergency Caesarean delivery (1.16; 1.03–1.31); these risks were higher in those with ≥ 3 kg/m2 gain. Our study strengthens the case for interpregnancy weight management to improve subsequent pregnancy outcomes. ## Introduction The rates of overweight and obesity continue to increase worldwide1. In women, pregnancy is a life stage that can alter their weight trajectory due to the risk of weight gain during or between pregnancies2,3. Higher parity has been associated with higher pre-pregnancy body mass index (BMI) and subsequent development of obesity4,5. On average, women gain approximately 1 kg/m2 between consecutive pregnancies, with greater interpregnancy BMI gain observed in those with a higher BMI before pregnancy6. In women who are overweight or obese, or underweight, the risks of adverse perinatal outcomes are well documented7,8. However, the extent to which interpregnancy weight change influences the risks of subsequent maternal and neonatal outcomes remain poorly understood9, and most studies have been focused on Western populations10. It is essential to personalize weight management planning for Asian women as they have higher health risks at lower BMI thresholds than Caucasian women11, and they have unique sociocultural factors which may influence weight management behaviours before, during, and after pregnancy10. The interpregnancy period represents a unique phase of the reproductive life-course. A recent systematic review and meta-analysis with data pooled from 11 Western countries showed that women with interpregnancy weight gain had increased risks of gestational diabetes, hypertensive disorders, large-for-gestational-age birth, and Caesarean delivery, while those with interpregnancy weight loss had increased risks for preterm delivery and small-for-gestational-age birth10. In the present study, our aims were to (i) describe the distribution of weight changes in BMI between first and second pregnancies among Singaporean women and (ii) examine whether similar associations between interpregnancy BMI changes and pregnancy outcomes would be observed in Asian women, compared to those reported in the aforementioned meta-analysis among Caucasians10. ## Material and methods Secondary routine healthcare data was retrieved from women with their first two consecutive deliveries from January 2015 to September 2020 at the KK Women’s and Children’s Hospital (KKH), Singapore. KKH houses the largest public maternity unit in Singapore and manages one-third of all live births in this country with approximately 12,000 deliveries every year, across a wide sociodemographic spectrum. We retrospectively extracted electronic medical records of women who had singleton births at ≥ 24 weeks gestation in the first and second pregnancies. Only women aged ≥ 21 years and conceived naturally in the first and second pregnancies were included. Women with missing information about BMI (at first and/ or second pregnancies) and interpregnancy interval were excluded. Ethics approval was obtained from the Centralised Institutional Review Board of SingHealth (reference $\frac{2020}{2018}$). Informed consent was waived due to the retrospective nature of the study by the Centralised Institutional Review Board of SingHealth. All methods were performed in accordance with the relevant guidelines and regulations. ## Interpregnancy BMI change and interval Maternal weight in kilograms and height in centimetres were routinely measured at the first antenatal appointment of the first and second deliveries. BMI, calculated as weight (in kilograms) divided by height (in metres) squared, at the first antenatal visit during the first and second pregnancies, was used to determine the interpregnancy BMI change. Given that gestational age at the first antenatal visit varied, the BMI measures were standardised separately in the first and second deliveries, by using linear regression with BMI at the first antenatal visit as dependent variable and gestational age centred at 12 weeks (i.e. the mean gestational age at the first antenatal visits where we had more data available to ensure a more accurate prediction of BMI) as the independent variable, calculating the residuals, and adding the residual values to the regression predicted mean BMI at 12 weeks. This is in keeping with data showing that weight in early pregnancy is a valid method for estimating pre-pregnancy weight12. We repeated the analysis by adjusting for maternal age and ethnicity in the multivariable linear regression model. Since a strong correlation (r > 0.95) was noted between both versions of predicted BMI at 12 weeks, the one that was derived using the simpler method without adjustment was used for all study analyses. The difference between gestational-age-adjusted BMI at both visits was then calculated as the change in BMI from the first to second deliveries and further categorized as BMI stable − 1 to < 1 kg/m2, BMI loss > 1 kg/m2, moderate BMI gain 1 to < 3 kg/m2 and excess BMI gain ≥ 3 kg/m2. The gestational-age-adjusted BMI at 12 weeks was used to represent the pre-pregnancy BMI in both pregnancies and was categorized using cut-offs for Asian populations: underweight (< 18.5 kg/m2), normal weight (18.5–22.9 kg/m2), overweight (23–27.49 kg/m2) and obese (≥ 27.5 kg/m2)11. The interpregnancy interval was calculated based on the period between the first delivery date and the second delivery conception date, which was derived by subtracting gestational age at delivery for the second birth from the duration between delivery dates of two consecutive births13. ## Pregnancy outcomes Neonatal outcomes included preterm delivery (< 37 completed gestation weeks), low birthweight (< 2.5 kg), high birthweight (≥ 4 kg), small-for-gestational-age (SGA) and large-for-gestational-age (LGA). SGA and LGA were defined as birthweight for sex and gestational age below the 10th centile and above the 90th centile, respectively, using the algorithm reported by Mikolaiczyk et al.14 based on a reference sample of healthy livebirths from the Growing Up in Singapore Towards healthy Outcomes (GUSTO) cohort, which is the largest pregnancy cohort study involving approximately 1000 mother–child pairs in Singapore15. Maternal outcomes included gestational diabetes mellitus (GDM) as diagnosed by a risk-based, 2-point oral glucose tolerance test (OGTT) between 2015 and 201716, and a universal 3-point OGTT between 2018 and 202017, elective and emergency Caesarean deliveries. Gestational hypertensive disorders were not included in the analysis due to incomplete information recorded in the electronic medical database. ## Statistical analysis The differences in characteristics between excluded and included women were compared using chi-square tests for categorical variables and independent t-tests for continuous variables. The associations of interpregnancy BMI change with subsequent pregnancy outcomes in the second pregnancy were examined using modified Poisson regression models to estimate risk ratios (RRs) and $95\%$ confidence intervals (CIs)18. The change in interpregnancy BMI was included as a categorical exposure (BMI stable, loss, moderate gain, or excessive gain), with stable BMI used as the reference group, as conventionally used in other studies10. The models were adjusted for maternal age (continuous), ethnicity (categorical), gestational-age-adjusted BMI at 12 weeks in the first pregnancy (continuous), interpregnancy interval (continuous) and pregnancy outcomes in the first pregnancy (categorical). As the effect of interpregnancy change on pregnancy outcomes may differ by maternal BMI at the beginning of the first pregnancy, we performed post-hoc analysis to examine whether there was any effect modification by weight status < 23 versus ≥ 23 kg/m2 at 12 weeks during the first pregnancy on any observed association. These models included categorical interpregnancy BMI change, weight status (effect modifier), the interaction terms between categorical interpregnancy BMI change weight status (3 degrees of freedom), and potential confounders as the independent variables. The results were stratified by weight status. Sensitivity analyses were performed using a similar modified Poisson regression to analyse the associations of the crude (unstandardised for gestational age) change in interpregnancy BMI with pregnancy outcomes, with confounders adjustment. These analyses were restricted to those with measures before or at 12 weeks gestation for both pregnancies. Statistical analyses were performed using Stata 16 (Stata, College Station, TX, USA). ## Women’s characteristics This study initially enrolled 7095 women with singleton first and second pregnancies. Of these, we excluded 831 women without BMI measured in one of the pregnancies ($$n = 772$$) or in both pregnancies ($$n = 59$$), leaving 6264 women in the final sample. Compared to excluded women, the included women tended to be older by only 0.5 years on average (28.4 vs. 27.9 years, $$p \leq 0.015$$) (see Supplementary Table S1 online). All other background variables were similar between the women included and excluded from the analysis (each $p \leq 0.05$). Of 6264 included women, $40.7\%$ had a stable interpregnancy BMI (− 1 to < 1 kg/m2), $10.3\%$ had BMI loss (> 1 kg/m2), $34.3\%$ had moderate BMI gain (1 to < 3 kg/m2) and $14.8\%$ had excess BMI gain (≥ 3 kg/m2) (Table 1). Women of younger age, Malay ethnicity and with higher BMI in the first pregnancy tended to experience excess BMI gain between their first two pregnancies. Table 1Characteristics of participants according to their interpregnancy BMI change status ($$n = 6264$$).CharacteristicsTotalBMI change status between first two pregnanciesStableLossModerate gainExcess gain(− 1 to < 1 kg/m2)(> 1 kg/m2)(1 to < 3 kg/m2)(≥ 3 kg/m2)$$n = 6264$$$n = 2548$; $40.7\%$$$n = 643$$; $10.3\%$$$n = 2146$$; $34.3\%$$$n = 927$$; $14.8\%$Maternal age in the first pregnancy, years28.36 ± 4.3128.58 ± 4.3228.61 ± 4.6428.49 ± 4.1727.27 ± 4.18Ethnicity Chinese2600 (41.5)1192 (46.8)275 (42.8)916 (42.7)217 (23.4) Malay1902 (30.4)704 (27.6)182 (28.3)620 (28.9)396 (42.7) Indian666 (10.6)231 (9.1)79 (12.3)233 (10.9)123 (13.3) Others1096 (17.5)421 (16.5)107 (16.6)377 (17.6)191 (20.6)BMI at 12-week gestation in the first pregnancy, kg/m223.76 ± 4.9723.19 ± 5.0425.36 ± 5.4423.35 ± 4.4825.17 ± 5.02BMI categories at 12-week gestation in the first pregnancy Underweight (< 18.5 kg/m2)585 (9.3)319 (12.6)17 (2.7)197 (9.2)52 (5.6) Normal weight (18.5–22.9 kg/m2)2719 (43.4)1209 (47.4)242 (37.6)974 (45.4)294 (31.7) Overweight (23–27.4 kg/m2)1785 (28.5)607 (23.8)208 (32.3)656 (30.6)314 (33.9) Obesity (≥ 27.5 kg/m2)1175 (18.8)413 (16.2)176 (27.4)319 (14.8)267 (28.8)BMI at 12-week gestation in the second pregnancy, kg/m224.92 ± 5.4023.33 ± 5.0323.35 ± 5.2725.23 ± 4.5829.65 ± 5.32BMI categories at 12-week gestation in the second pregnancy Underweight (< 18.5 kg/m2)416 (6.6)286 (11.2)81 (12.6)48 (2.2)1 (0.1) Normal weight (18.5–22.9 kg/m2)2262 (36.1)1210 (47.5)302 (47.0)694 (32.3)56 (6.0) Overweight (23–27.4 kg/m2)1972 (31.5)626 (24.6)145 (22.6)887 (41.3)314 (33.9) Obesity (≥ 27.5 kg/m2)1614 (25.8)426 (16.7)115 (17.8)517 (24.2)556 (60.0)Interpregnancy BMI change, kg/m20.97 (− 0.04 to 2.21)0.19 (− 0.25 to 0.59)− 1.72 (− 2.40 to -1.31)1.84 (1.40–2.32)4.02 (3.45–4.95)Interpregnancy interval, years1.44 (0.89–2.19)1.39 (0.88–2.04)1.34 (0.84–1.97)1.49 (0.89–2.27)1.61 (0.97–2.61)Data are presented as number (percentage) for categorical variables, and as mean ± standard deviation or median (25th–75th percentiles) for continuous variables. BMI, body mass index. ## Distribution of interpregnancy BMI change Overall, BMI tended to change (increase or decrease) among women who gave the second birth in the first two years after the first delivery and was stable at that level among women who gave the second birth later, regardless of the initial weight status (Fig. 1). Women who were overweight and obese in their first pregnancy tended to experience interpregnancy BMI loss or gain as compared to those who were underweight and normal weight, who tended to be BMI stable ($p \leq 0.001$) (Fig. 2a). In particular, those women who were overweight or obese had higher BMI loss than women with a normal BMI (median BMI loss 1.9 vs. 1.5 kg/m2, $p \leq 0.001$) (see Supplemental Fig. S1 online).Figure 1Cross-sectional trends of BMI change over interpregnancy interval. BMI categories of women were measured at 12-week gestation in the first pregnancy ($$n = 6264$$). BMI, body mass index. Figure 2Interpregnancy BMI change status and BMI categories in the second pregnancy by BMI categories in the first pregnancy. Bar chart showing the distribution of (a) body mass index (BMI) change status between first two pregnancies and (b) BMI categories at 12-week gestation in the second pregnancy, by BMI categories of women at 12-week gestation in the first pregnancy ($$n = 6264$$). BMI categories were classified based on the cut-offs for Asian populations. BMI, body mass index. In total, $24.5\%$ of women gained weight between pregnancies and progressed to a higher BMI category; while $5.4\%$ of women lost weight and dropped to a lower BMI category. Although at least $90\%$ of women who were overweight or obese in the first pregnancy remained at least overweight in the second pregnancy, nearly two-thirds of women with normal weight and half of women who were underweight remained in the same weight status in the first and second pregnancies (Fig. 2b). Similar distributions of interpregnancy BMI change status and weight status in the second pregnancy were observed across weight status in the first pregnancy based on the WHO conventional cut-offs (see Supplementary Table S2 online). ## Interpregnancy BMI change and subsequent pregnancy outcomes Compared to women with a stable BMI from the first to the second pregnancy, those with BMI loss had a higher risk of low birthweight delivery (RR 1.36; $95\%$ CI 1.02–1.80). Women with moderate BMI gain had higher risks of LGA birth (1.16; 1.03–1.31), GDM (1.25; 1.06–1.49) and emergency Caesarean delivery (1.16; 1.03–1.31) in the second pregnancy; these risks were higher in those with excess BMI gain (Table 2). Similar findings were obtained in a sensitivity analysis using crude interpregnancy BMI change (see Supplementary Table S3 online). In women with BMI ≥ 23 kg/m2, BMI loss was associated with increased risk of low birthweight (1.64; 1.09–2.47) and SGA deliveries (1.54; 1.02–2.34) (Table 3). In women with BMI < 23 kg/m2, moderate (1.31; 1.07–1.59) and excess BMI gains (1.35; 1.04–1.77) were associated with an increased risk of emergency Caesarean. Table 2Association between interpregnancy BMI change status and outcomes of second pregnancy. Outcomes of second pregnancyBMI change status between first two pregnanciesStableLossModerate gainExcess gain(-1 to < 1 kg/m2)(> 1 kg/m2)(1 to < 3 kg/m2)(≥ 3 kg/m2)n (%)n (%)RR ($95\%$ CI)n (%)RR ($95\%$ CI)n (%)RR ($95\%$ CI)Offspring birth weight Normal 2.5 to < 4 kg2341 (92.2)574 (89.7)1.001965 (92.0)1.00839 (90.8)1.00 Low < 2.5 kg163 (6.4)55 (8.6)1.36 (1.02, 1.80)136 (6.4)1.00 (0.81, 1.25)57 (6.2)0.97 (0.72, 1.31) High ≥ 4 kg35 (1.4)11 (1.7)0.88 (0.46, 1.69)34 (1.6)1.09 (0.68, 1.74)28 (3.0)1.62 (0.97, 2.71)Offspring birth size AGA 10–90 percentile1932 (76.2)488 (76.2)1.001583 (74.1)1.00643 (69.7)1.00 SGA < 10 percentile206 (8.1)54 (8.4)1.15 (0.87, 1.52)161 (7.5)0.94 (0.78, 1.14)61 (6.6)0.84 (0.64, 1.11) LGA > 90 percentile399 (15.7)98 (15.3)0.87 (0.72, 1.05)391 (18.3)1.16 (1.03, 1.31)219 (23.7)1.40 (1.21, 1.61)Preterm delivery < 37 weeks No2385 (93.6)592 (92.1)1.002012 (93.8)1.00863 (93.1)1.00 Yes163 (6.4)51 (7.9)1.08 (0.80, 1.45)134 (6.2)0.96 (0.78, 1.20)64 (6.9)1.07 (0.81, 1.41)Gestational diabetes No2352 (92.3)580 (90.2)1.001956 (91.1)1.00806 (86.9)1.00 Yes196 (7.7)63 (9.8)1.04 (0.81, 1.34)190 (8.9)1.25 (1.06, 1.49)121 (13.1)1.63 (1.35, 1.97)Mode of delivery Vaginal delivery1892 (76.7)541 (74.4)1.001573 (74.5)1.00718 (74.9)1.00 Elective caesarean345 (14.0)102 (14.0)1.01 (0.89, 1.15)301 (14.3)1.02 (0.95, 1.10)123 (12.8)1.04 (0.95, 1.15) Emergency caesarean229 (9.3)84 (11.6)1.09 (0.93, 1.28)238 (11.3)1.16 (1.03, 1.31)118 (12.3)1.14 (1.00, 1.32)Risk ratios are adjusted for maternal age and BMI at 12-week gestation in the first pregnancy, ethnicity, interpregnancy interval and respective pregnancy outcomes in the first pregnancy. BMI stable (− 1 to < 1 kg/m2) serves as the reference group. BMI, body mass index; RR, risk ratio; CI, confidence interval; AGA, appropriate for gestational age; SGA, small-for-gestational-age; LGA, large-for-gestational-age. Table 3Association between interpregnancy BMI change status and outcomes of second pregnancy, by weight status at 12-week gestation in the first pregnancy. Outcomes of second pregnancyBMI < 23 kg/m2BMI ≥ 23 kg/m2P-interactionLossModerate gainExcess gainLossModerate gainExcess gain(> 1 kg/m2)(1 to < 3 kg/m2)(≥ 3 kg/m2)(> 1 kg/m2)(1 to < 3 kg/m2)(≥ 3 kg/m2)RR ($95\%$ CI)RR ($95\%$ CI)RR ($95\%$ CI)RR ($95\%$ CI)RR ($95\%$ CI)RR ($95\%$ CI)Offspring birth weight Low < 2.5 kg (vs. Normal 2.5 to < 4 kg)1.22 (0.79, 1.90)0.84 (0.63, 1.11)1.07 (0.72, 1.59)1.64 (1.09, 2.47)1.43 (0.99, 2.07)1.06 (0.66 2.71)0.076 High ≥ 4 kg (vs. Normal 2.5 to < 4 kg)0.52 (0.07, 4.04)1.30 (0.56, 3.02)3.00 (1.17, 7.68)0.93 (0.46, 1.86)1.01 (0.58, 1.76)1.34 (0.74, 2.43)0.428Offspring birth size SGA < 10 percentile (vs. AGA 10–90 percentile)0.96 (0.63, 1.45)0.84 (0.67, 1.06)0.88 (0.62, 1.25)1.54 (1.02, 2.34)1.29 (0.91, 1.83)0.90 (0.58, 1.42)0.078 LGA > 90 percentile (vs. AGA 10–90 percentile)0.73 (0.49, 1.09)1.16 (0.96, 1.40)1.67 (1.30, 2.15)0.88 (0.71, 1.10)1.13 (0.97, 1.31)1.23 (1.03, 1.46)0.092Preterm delivery < 37 weeks Yes (vs. No)1.03 (0.65, 1.65)0.82 (0.60, 1.12)1.24 (0.81, 1.88)1.09 (0.74, 1.61)1.13 (0.83, 1.53)1.02 (0.71, 1.48)0.327Gestational diabetes Yes (vs. No)1.22 (0.76, 1.95)1.11 (0.83, 1.48)1.77 (1.20, 2.61)0.99 (0.74, 1.31)1.27 (1.03, 1.58)1.50 (1.20, 1.88)0.541Mode of delivery Elective caesarean (vs. vaginal delivery)1.10 (0.89, 1.36)1.09 (0.97, 1.22)1.09 (0.92, 1.28)0.97 (0.84, 1.13)0.97 (0.88, 1.07)1.02 (0.91 1.16)0.348 Emergency caesarean (vs. vaginal delivery)1.34 (0.96, 1.88)1.31 (1.07, 1.59)1.35 (1.04, 1.77)0.98 (0.82, 1.17)1.06 (0.91, 1.24)1.05 (0.89, 1.23)0.133Risk ratios are adjusted for maternal age and BMI at 12-week gestation in the first pregnancy, ethnicity, interpregnancy interval and respective pregnancy outcomes in the first pregnancy. BMI stable (− 1 to < 1 kg/m2) serves as the reference group. BMI, body mass index; RR, risk ratio; CI, confidence interval; AGA, appropriate for gestational age; SGA, small-for-gestational-age; LGA, large-for-gestational-age. ## Discussion In this cohort that included 6264 women, about a quarter increased while $5\%$ lowered their BMI category between their first and second pregnancies. Approximately half gained ≥ 1 kg/m2, of which one-third had excess gain of ≥ 3 kg/m2; only $10\%$ lost > 1 kg/m2 between pregnancies. Overall, BMI tended to change among women who birthed their second child in the first two years after the first delivery, and was stable among women who birthed their second child later, regardless of the initial weight status. Interpregnancy BMI gain was associated with increased risks of LGA, GDM and emergency Caesarean delivery in the second pregnancy. Conversely, an increased risk of low birthweight was observed in women with BMI loss between their first two pregnancies. When the results were further stratified by BMI in the first pregnancy, a higher risk of emergency Caesarean delivery was evident in women with a BMI < 23 kg/m2 experiencing interpregnancy BMI gain, while higher risks of low birthweight and SGA were evident in women with a BMI ≥ 23 kg/m2 experiencing interpregnancy BMI loss. The interpregnancy period is a valuable opportunity to address pregnancy complications and optimise health for the next pregnancy and the rest of the life-course. Despite recommendations to return to pre-pregnancy weight between 6 and 12 months postpartum, with the goal of a normal BMI19, about half the women in our study increased their BMI during the first two years post-delivery instead. A study conducted among Caucasian women also showed similar findings, where almost $20\%$ of normal-weight women became overweight or obese in their next pregnancy, whereas more than $90\%$ of overweight or obese women maintained their status in the next pregnancy20. This highlights the urgent need to implement intervention strategies that include targeted lifestyle modifications to prevent increased BMI during the interpregnancy period. Interpregnancy BMI gain and the associated increased risks of subsequent LGA, GDM and emergency Caesarean delivery are consistent with previous studies9,10,21. These adverse complications could be the result of reduced insulin sensitivity due to interpregnancy weight gain accompanied by body fat rather than muscle gain, which is common among Asians20,22–25. The increased risk of emergency Caesarean delivery in women with an initial BMI < 23 kg/m2 is consistent with a recent meta-analysis23, suggesting an increased susceptibility of lean women to subsequent delivery complications in response to weight gain between pregnancies. However, the indications for emergency Caesarean delivery were unclear in our data and should be further examined in future studies. Similarly, interpregnancy BMI gain has been associated with increased risks of hypertensive disorders9,26 and stillbirth10, but we were unable to analyse these outcomes due to incomplete outcome data. In view of multiple adverse pregnancy outcomes, long-term obesity, and related health risks in women and their offspring, our study, together with many others13,27–32, call for nationwide efforts to break the vicious cycle of interpregnancy weight gain and poor metabolic health. We found that offspring of women with BMI loss between their first two pregnancies had a higher risk of low birthweight. This is supported by a study on interpregnancy weight change among women in three consecutive pregnancies, showing that BMI loss was associated with an increased risk of low placental weight and SGA births33. Another study also showed that a decrease in BMI > 1 kg/m2 between the first two consecutive births was associated with a higher risk of low birthweight (< 2.5 kg)34. This phenomenon could be explained by insulin sensitivity induced by weight loss, resulting in less glucose crossing the placenta, which contributed to an increased risk of small fetal size23. A meta-analysis showed that interpregnancy weight loss and SGA was only apparent in women with initial BMI < 25 kg/m2, but not among those with BMI ≥ 25 kg/m210. However, our study observed that women with BMI ≥ 23 kg/m2 in the first pregnancy who lost weight during the interpregnancy interval had a higher risk of low birthweight and SGA. Although not reaching statistical significance, women of BMI < 23 kg/m2 in the first pregnancy with an interpregnancy BMI loss also had a higher risk of low birthweight, albeit with a smaller effect size. These findings should be interpreted with caution as they may be attributed to the greater weight loss among women who were overweight or obese within the interpregnancy interval of 1–2 years, compared with women with a normal BMI (BMI loss 1.9 vs. 1.5 kg/m2, $p \leq 0.001$) (see Supplemental Fig. S1 online). In addition, unlike other studies that showed a reduction in the risk of adverse pregnancy outcomes among overweight and obese women who lost weight10,20,21,23, our study did not find a significant reduction in risk among women with BMI ≥ 23 kg/m2 who lost weight. Despite the current emphasis on BMI, it represents a crude measure of adiposity and an imperfect assessment of metabolic health35. This was highlighted by a recent study that showed that metabolic health status, rather than BMI, played a greater role in fecundability36. Therefore, interpregnancy BMI loss may not truly reflect the metabolic health status of our study participants, which confounds the positive effects of weight loss in overweight and obese women. Furthermore, changes in body composition and fat distribution between pregnancies, and gestational weight gain (GWG) during pregnancy in overweight or obese women can impact subsequent pregnancy outcomes37. The lack of metabolic health, GWG and other data in our study precludes making recommendations for the amount of weight loss to improve pregnancy outcomes, and further studies including this information are needed to make such recommendations. Despite the higher risk of low birthweight and SGA in women with BMI ≥ 23 kg/m2 who lost weight during the interpregnancy interval, it is important to balance this with the benefits of achieving a normal BMI, especially in women living with obesity, given the potential for other adverse perinatal outcomes, such as GDM, hypertensive disorders of pregnancy, macrosomia, birth trauma, and stillbirth7,8. Based on the trend of interpregnancy BMI change, the first two years post-delivery likely represents the best window of opportunity to intervene to return to pre-pregnancy BMI, regardless of initial weight status. Effective lifestyle interventions that aim to limit postpartum weight retention during this window are crucial to improving perinatal outcomes. Such interventions should ideally be engaging, grounded by behaviour change theories, and integrate components of both diet and physical activity38. An electronic health intervention for postpartum women with excessive GWG resulted in restrained eating, along with decreased uncontrolled eating and energy intake39. However, other behaviours such as emotional eating, physical activity, and sedentary time remain unchanged39. To improve the success of lifestyle interventions, it is essential to identify additional facilitators and barriers faced by these women. Although these were identified among overweight and obese women trying to conceive40, it remains unclear whether such findings are applicable to women of normal weight. This is the first study to investigate the distribution and outcomes of interpregnancy weight change in Asian women, with a substantial sample size of women from the three largest ethnicities in Singapore (Chinese, Malay, and Indian) where the findings may be generalizable to other Asian populations. However, the study employed statistical modelling to predict the maternal BMI at 12 weeks and used it as the pre-pregnancy BMI. This might result in misclassification of weight status and interpregnancy weight change categories. In addition, since BMI is an imperfect measurement of metabolic health35, future studies should investigate how other markers of metabolic health, such as insulin resistance, lipid profile and body composition, are associated with adverse perinatal outcomes. The GDM screening policy underwent a transition during the study period, from a risk-based 2-point OGTT between 2015 and 2017 to a universal 3-point OGTT from 2018 to 2020, thus, the incidence of GDM may be underestimated in the earlier years41,42. We did not account for the association of GWG with adverse perinatal outcomes, including fetal growth, preterm delivery, GDM, and Caesarean delivery37. Hence, we are unable to determine whether the association of interpregnancy BMI with adverse perinatal outcomes would be mediated by GWG, which should be a focus for future studies. We did not evaluate other adverse pregnancy outcomes such as intervening miscarriage, as ascertainment of this outcome is known to be incomplete, while our study was underpowered to examine low prevalence outcomes such as stillbirth. We did not account for the socioeconomic status and lifestyle habits of the women in the analysis due to the lack of data from medical records. Finally, long-term outcomes of these women and their offspring were not available to provide insights on their long-term health. ## Conclusion This study has shown that a large proportion of women increase their BMI, and a small proportion decrease their BMI between their first two pregnancies. An increase and a decrease in BMI between pregnancies are associated with a higher risk of adverse outcomes in the second pregnancy. These findings highlight the importance of interpregnancy weight management to achieve better pregnancy outcomes subsequently. However, the recommended magnitude of weight loss beyond their pre-pregnancy weight remains unclear, especially for those who are overweight or obese, where a loss > 1 kg/m2 was associated with SGA and low birthweight. Future studies should examine the role of interpregnancy weight management interventions among Asian women, and to examine the role of metabolic health in adverse pregnancy outcomes with measurement of GWG, body composition and metabolic biomarkers. This will shed light on possible aetiologies of low birthweight/SGA and weight loss and guide personalized interventions and BMI targets for women with lean BMI and those who are overweight or obese. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31954-5. ## References 1. **Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128·9 million children, adolescents, and adults**. *Lancet* (2017) **390** 2627-2642. DOI: 10.1016/s0140-6736(17)32129-3 2. Gunderson EP. **Childbearing and obesity in women: Weight before, during, and after pregnancy**. *Obstet. Gynecol. Clin. North Am.* (2009) **36** 317-332. DOI: 10.1016/j.ogc.2009.04.001 3. Nehring I, Schmoll S, Beyerlein A, Hauner H, von Kries R. **Gestational weight gain and long-term postpartum weight retention: A meta-analysis**. *Am. J. Clin. Nutr.* (2011) **94** 1225-1231. DOI: 10.3945/ajcn.111.015289 4. Bastian LA, West NA, Corcoran C, Munger RG. **Number of children and the risk of obesity in older women**. *Prev. Med.* (2005) **40** 99-104. DOI: 10.1016/j.ypmed.2004.05.007 5. Hill B. **Is parity a risk factor for excessive weight gain during pregnancy and postpartum weight retention? A systematic review and meta-analysis**. *Obes. Rev.* (2017) **18** 755-764. DOI: 10.1111/obr.12538 6. Sumithran P. **How common is substantial weight gain after pregnancy?**. *Obes. Res. Clin. Pract.* (2018) **12** 139-145. DOI: 10.1016/j.orcp.2017.10.007 7. Dsouza R, Horyn I, Pavalagantharajah S, Zaffar N, Jacob CE. **Maternal body mass index and pregnancy outcomes: a systematic review and metaanalysis**. *Am. J. Obstet. Gynecol. MFM* (2019) **1** 100041. DOI: 10.1016/j.ajogmf.2019.100041 8. Liu P. **Association between perinatal outcomes and maternal pre-pregnancy body mass index**. *Obes. Rev.* (2016) **17** 1091-1102. DOI: 10.1111/obr.12455 9. Teulings N, Masconi KL, Ozanne SE, Aiken CE, Wood AM. **Effect of interpregnancy weight change on perinatal outcomes: Systematic review and meta-analysis**. *BMC Pregnancy Childbirth* (2019) **19** 386. DOI: 10.1186/s12884-019-2566-2 10. Nagpal TS. **Does prepregnancy weight change have an effect on subsequent pregnancy health outcomes? A systematic review and meta-analysis**. *Obes. Rev.* (2022) **23** 13324. DOI: 10.1111/obr.13324 11. **Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies**. *Lancet* (2004) **363** 157-163. DOI: 10.1016/s0140-6736(03)15268-3 12. Inskip H. **Measured weight in early pregnancy is a valid method for estimating pre-pregnancy weight**. *J. Dev. Orig. Health Dis.* (2021) **12** 561-569. DOI: 10.1017/s2040174420000926 13. Ziauddeen N. **Interpregnancy weight gain and childhood obesity: Analysis of a UK population-based cohort**. *Int. J. Obes. (Lond.)* (2022) **46** 211-219. DOI: 10.1038/s41366-021-00979-z 14. Mikolajczyk RT. **A global reference for fetal-weight and birthweight percentiles**. *Lancet* (2011) **377** 1855-1861. DOI: 10.1016/s0140-6736(11)60364-4 15. Soh SE. **Cohort profile: Growing up in Singapore towards healthy outcomes (GUSTO) birth cohort study**. *Int. J. Epidemiol.* (2014) **43** 1401-1409. DOI: 10.1093/ije/dyt125 16. Alberti KG, Zimmet PZ. **Definition, diagnosis and classification of diabetes mellitus and its complications Part 1: Diagnosis and classification of diabetes mellitus provisional report of a WHO consultation**. *Diabet. Med.* (1998) **15** 539-553. DOI: 10.1002/(sici)1096-9136(199807)15:7<539::Aid-dia668>3.0.Co;2-s 17. **Diagnostic criteria and classification of hyperglycaemia first detected in pregnancy: A World Health Organization Guideline**. *Diabetes Res. Clin. Pract.* (2014) **103** 341-363. DOI: 10.1016/j.diabres.2013.10.012 18. Zou G. **A modified Poisson regression approach to prospective studies with binary data**. *Am. J. Epidemiol.* (2004) **159** 702-706. DOI: 10.1093/aje/kwh090 19. Louis JM, Bryant A, Ramos D, Stuebe A, Blackwell SC. **Interpregnancy Care**. *Am. J. Obstet. Gynecol.* (2019) **220** B2-b18. DOI: 10.1016/j.ajog.2018.11.1098 20. McBain RD, Dekker GA, Clifton VL, Mol BW, Grzeskowiak LE. **Impact of inter-pregnancy BMI change on perinatal outcomes: A retrospective cohort study**. *Eur. J. Obstet. Gynecol. Reprod. Biol.* (2016) **205** 98-104. DOI: 10.1016/j.ejogrb.2016.07.487 21. Martínez-Hortelano JA. **Interpregnancy weight change and gestational diabetes mellitus: A systematic review and meta-analysis**. *Obesity (Silver Spring)* (2021) **29** 454-464. DOI: 10.1002/oby.23083 22. Gunderson EP. **Childbearing may increase visceral adipose tissue independent of overall increase in body fat**. *Obesity (Silver Spring)* (2008) **16** 1078-1084. DOI: 10.1038/oby.2008.40 23. Oteng-Ntim E. **Interpregnancy weight change and adverse pregnancy outcomes: A systematic review and meta-analysis**. *BMJ Open* (2018) **8** e018778. DOI: 10.1136/bmjopen-2017-018778 24. Retnakaran R, Hanley AJ, Connelly PW, Sermer M, Zinman B. **Ethnicity modifies the effect of obesity on insulin resistance in pregnancy: A comparison of Asian, South Asian, and Caucasian women**. *J. Clin. Endocrinol. Metab.* (2006) **91** 93-97. DOI: 10.1210/jc.2005-1253 25. Sorbye LM, Skjaerven R, Klungsoyr K, Morken NH. **Gestational diabetes mellitus and interpregnancy weight change: A population-based cohort study**. *PLoS Med.* (2017) **14** e1002367. DOI: 10.1371/journal.pmed.1002367 26. Kawakita T, Downs SK, Franco S, Ghofranian A, Thomas A. **Interpregnancy body mass index change and risk of hypertensive disorders in pregnancy**. *J. Matern. Fetal Neonatal Med.* (2022) **35** 3223-3228. DOI: 10.1080/14767058.2020.1817371 27. Adane AA, Dobson A, Tooth L, Mishra GD. **Maternal preconception weight trajectories are associated with offsprings' childhood obesity**. *Int. J. Obes. (Lond.)* (2018) **42** 1265-1274. DOI: 10.1038/s41366-018-0078-1 28. Alwan NA, Grove G, Taylor E, Ziauddeen N. **Maternal weight change between successive pregnancies: An opportunity for lifecourse obesity prevention**. *Proc. Nutr. Soc.* (2020) **79** 272-282. DOI: 10.1017/s0029665120007065 29. Catalano PM, Ehrenberg HM. **The short- and long-term implications of maternal obesity on the mother and her offspring**. *BJOG* (2006) **113** 1126-1133. DOI: 10.1111/j.1471-0528.2006.00989.x 30. Derraik JG, Ayyavoo A, Hofman PL, Biggs JB, Cutfield WS. **Increasing maternal prepregnancy body mass index is associated with reduced insulin sensitivity and increased blood pressure in their children**. *Clin. Endocrinol. (Oxf.)* (2015) **83** 352-356. DOI: 10.1111/cen.12665 31. Endres LK. **Postpartum weight retention risk factors and relationship to obesity at 1 year**. *Obstet. Gynecol.* (2015) **125** 144-152. DOI: 10.1097/aog.0000000000000565 32. Wahabi HA, Fayed AA, Tharkar S, Esmaeil SA, Bakhsh H. **Postpartum weight retention and cardiometabolic risk among Saudi women: A follow-up study of RAHMA subcohort**. *Biomed. Res. Int.* (2019) **2019** 2957429. DOI: 10.1155/2019/2957429 33. Wallace JM, Bhattacharya S, Horgan GW. **Weight change across the start of three consecutive pregnancies and the risk of maternal morbidity and SGA birth at the second and third pregnancy**. *PLoS One* (2017) **12** e0179589. DOI: 10.1371/journal.pone.0179589 34. Bogaerts A. **Interpregnancy weight change and risk for adverse perinatal outcome**. *Obstet. Gynecol.* (2013) **122** 999-1009. DOI: 10.1097/AOG.0b013e3182a7f63e 35. Cirulli ET. **Profound perturbation of the metabolome in obesity is associated with health risk**. *Cell Metab.* (2019) **29** 488-500. DOI: 10.1016/j.cmet.2018.09.022 36. Loy SL. **Metabolic health status and fecundability in a Singapore preconception cohort study**. *Am. J. Obstet. Gynecol.* (2022) **226** e711-714. DOI: 10.1016/j.ajog.2021.11.1374 37. Champion ML, Harper LM. **Gestational weight gain: Update on outcomes and interventions**. *Curr. Diabetes Rep.* (2020) **20** 11. DOI: 10.1007/s11892-020-1296-1 38. van der Pligt P. **Systematic review of lifestyle interventions to limit postpartum weight retention: Implications for future opportunities to prevent maternal overweight and obesity following childbirth**. *Obes. Rev.* (2013) **14** 792-805. DOI: 10.1111/obr.12053 39. Bijlholt M, Ameye L, Van Uytsel H, Devlieger R, Bogaerts A. **The INTER-ACT E-Health supported lifestyle intervention improves postpartum food intake and eating behavior, but not physical activity and sedentary behaviour—A randomized controlled trial**. *Nutrients* (2021) **13** 1287. DOI: 10.3390/nu13041287 40. Ku CW. **Developing a lifestyle intervention program for overweight or obese preconception, pregnant and postpartum women using qualitative methods**. *Sci. Rep.* (2022) **12** 2511. DOI: 10.1038/s41598-022-06564-2 41. Chi C. **Impact of adopting the 2013 World Health Organization criteria for diagnosis of gestational diabetes in a multi-ethnic Asian cohort: A prospective study**. *BMC Pregnancy Childbirth* (2018) **18** 69. DOI: 10.1186/s12884-018-1707-3 42. Chong YS. **Ethnic differences translate to inadequacy of high-risk screening for gestational diabetes mellitus in an Asian population: A cohort study**. *BMC Pregnancy Childbirth* (2014) **14** 345. DOI: 10.1186/1471-2393-14-345
--- title: The effect of dapagliflozin on uric acid excretion and serum uric acid level in advanced CKD authors: - Yukimasa Iwata - Shoki Notsu - Yushi Kawamura - Waka Mitani - Shinjiro Tamai - Madoka Morimoto - Masafumi Yamato journal: Scientific Reports year: 2023 pmcid: PMC10039024 doi: 10.1038/s41598-023-32072-y license: CC BY 4.0 --- # The effect of dapagliflozin on uric acid excretion and serum uric acid level in advanced CKD ## Abstract Sodium–glucose cotransporter 2 inhibitors (SGLT2i) exhibit renoprotective effect in patients with chronic kidney disease (CKD) and reduce serum uric acid (UA) in patients with diabetes mellitus. However, it is not clarified whether SGLT2i reduce serum UA levels in patients with advanced CKD. This study aimed to investigate the impact of SGLT2i on change in serum UA levels in patients with advanced CKD. Data of 121 Japanese patients with CKD who were newly administered 10 mg dapagliflozin in our department between August 2021 and August 2022 were analyzed. Changes in UA and fractional excretion of UA (FEUA) were analyzed using multiple regression analysis. Of 75 patients, 21 ($28.0\%$) patients, 24 ($32.0\%$) patients, 29 ($38.7\%$) patients, and 1 ($1.3\%$) patient were categorized as having CKD stage 3a, 3b, 4, and 5, respectively. The median age was 67 years, and $72.0\%$ were male. 23 ($30.7\%$) of patients had diabetes mellitus. The median estimated glomerular filtration rate, serum UA, and FEUA were 35.7 mL/min/1.73 m2, 6.4 mg/dL, and $6.76\%$, respectively, at the time of dapagliflozin administration. After administration, serum UA decreased to 5.6 mg/dL and FEUA increased to $9.22\%$. Dapagliflozin increases FEUA and reduces serum UA levels in patients with advanced CKD. ## Introduction Chronic kidney disease (CKD) is a global concern, affecting more than 10 million patients in Japan and 700 million worldwide1,2. Renin angiotensin aldosterone system inhibitors have been shown to reduce urinary protein levels and delay CKD progression3,4. Recently, it has been shown that sodium–glucose cotransporter 2 inhibitors (SGLT2i) have a renoprotective effect5–7. Although the mechanism of their renoprotective effect is believed to involve the reduction of intra-glomerular pressure by tubular glomerular feedback8, SGLT2i have pleiotropic effects and their mechanism of renal protection has not been fully revealed9. Serum uric acid (UA) and renal outcomes are not fully understood; however, it has been reported that allopurinol reduces serum UA levels and improves renal outcomes10,11. On the other hand, allopurinol did not improve renal outcomes12,13. One of the pleiotropic effects of SGLT2i is reduction of serum UA in patients with type 2 diabetes mellitus (DM) whose estimated glomerular filtration rate (eGFR) was maintained at > 60 ml/min/1.73 m214–17; however, the effect of lowering serum UA levels by SGLT2i in patients with CKD whose renal function reduced to ˂60 ml/min/1.73 m2 is unclear. Dapagliflozin 10 mg is the only drug approved for the treatment of CKD with or without DM in Japan as SGLT2i and we analyzed patients who were administered dapagliflozin 10 mg. Approximately $70\%$ of serum UA is excreted into the urine and the rest is thought to be excreted into the intestinal tract. The renal transport system such as urate transporter 1 (URAT1), adenosine triphosphate-binding cassette transporter G2 (ABCG2), glucose transporter 9 (GLUT9) isoform2 plays an important role in the regulation of serum UA level. SGLT2i increase expression of GLUT9 and ABCG218,19. Thus, this study aimed to investigate the effect of SGLT2i in lowering serum UA levels in patients with CKD. In addition, this study aimed to examine the effect of SGLT2i in the change in urinary UA excretion. ## Patients This study was conducted in accordance with the Declaration of Helsinki guidelines and approved by institutional review board of Sakai City Medical Center (No. 21-261). All patients were provided with the option to opt out of the study. The need for informed consent was waived using our hospital’s opt-out method. This was a single-center retrospective cohort study. A total of 121 Japanese patients with CKD who were newly administered dapagliflozin 10 mg in our department from August 2021 to August 2022 and were followed-up for at least 2 weeks were enrolled in the study. Thirty-four patients were excluded from the analysis owing to changes from other SGLT2i ($$n = 3$$) or dapagliflozin dose ($$n = 2$$). In addition, patients with changes in the dose of diuretics ($$n = 3$$) and antihyperuricemic agents ($$n = 3$$), and those whose serum UA level was not measured before or after dapagliflozin 10 mg administration ($$n = 17$$) were also excluded. Twelve patients with CKD stages 1 and 2 were excluded from the study. In study 1, 75 patients were retrospectively studied (Fig. 1).Figure 1The flow chart of the study population. Furthermore, patients whose fractional excretion of uric acid (FEUA) was not measured before or after dapagliflozin 10 mg administration ($$n = 40$$) were excluded from the study. A total of 35 patients were included in Study 2. ## Data collection and definition We collected the data of 121 patients from the electronic medical charts of Sakai City Medical Center, including demographics (sex and age), medications, comorbidities (history of heart failure and DM), and clinical and laboratory variables (blood pressure, pulse rate, serum creatinine, blood urea nitrogen, UA, hemoglobin, albumin, FEUA, and proteinuria). FEUA was calculated as (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$100 \times urine\; UA \times serum\;creatinine \div serum\;UA \div urine\;creatinine$$\end{document}100×urineUA×serumcreatinine÷serumUA÷urinecreatinine). eGFR was estimated using the formula for Japanese people \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left({eGFR \left[{{\text{mL}}/{\text{min}}/1.73\;{\text{m}}^{2} } \right] = 194 \times serum creatinine - 1.094 \times age - 0.287 \times 0.739 \left[{for females} \right]} \right)$$\end{document}eGFRmL/min/1.73m2=194×serumcreatinine-1.094×age-0.287×0.739forfemales20. Patients were divided into four groups according to their renal function and CKD category (G3–G5). ## Outcomes This study’s primary and secondary outcomes were, respectively, changes in serum UA and FEUA after administration of dapagliflozin 10 mg. ## Statistical analyses Continuous variables are presented as medians (interquartile range [IQR]) and were compared using the Kruskal–Wallis test. Categorical variables are presented as numbers and percentages and were compared using the Fisher’s exact test. Changes in UA and FEUA were compared using the Wilcoxon signed-rank test and Kruskal–Wallis test. Multiple regression analysis was used to estimate the independent factors associated with changes in UA and FEUA. Spearman’s rank correlation coefficient was used to evaluate an association between change in FEUA and serum UA. All statistical analyses were performed using EZR software (Saitama Medical Center, Jichi Medical University, Saitama, Japan)21. ## Baseline characteristics At administration of dapagliflozin 10 mg, 21 ($28.0\%$) patients, 24 ($32.0\%$) patients, 29 ($38.7\%$) patients, and 1 ($1.3\%$) patient were categorized as having CKD stages 3a, 3b, 4, and 5, respectively (Table 1). Patient characteristics and baseline laboratory data for each group are shown in Table 1.Table1Patient characteristics and laboratory data at starting dapagliflozin 10 mg by CKD stage. All patients $$n = 75$$CKD stage 3a $$n = 21$$CKD stage 3b $$n = 24$$CKD stage 4 $$n = 29$$CKD stage 5 $$n = 1$$p valueMale (%)54 (72.0)15 (71.4)21 (87.5)17 (58.6)1 (100.0)0.059Age67 [57–77]62 [53–67]71 [60–77]73 [62–79]38 [38–38]0.041Comorbid condition Heart failure (%)3 (4.0)0 (0.0)0 (0.0)3 (10.3)0 (0.0)0.175 *Diabetic mellitus* (%)23 (30.7)6 (28.6)8 (33.3)9 (31.0)0 (0.0)0.904Clinical and laboratory variables Systolic blood pressure (mmHg)133 [122–140]130 [118–138]137 [125–145]135 [122–140]116 [116–116]0.184 Diastolic blood pressure (mmHg)75 [68–83]82 [72–85]76 [68–81]71 [66–77]74 [74–74]0.254 Pulse rate (bpm)79 [70–84]82 [73–86]75 [69–87]78 [69–89]76 [67–88]0. 856 Creatinine (mg/dL)1.49 [1.23–1.95]1.12 [1.01–1.23]1.47 [1.33–1.58]2.01 [1.91–2.24]4.11 [4.11–4.11] < 0.001 Estimated glomerular filtration rate (ml/min/1.73m2)35.7 [26.2–46.3]51.3 [47.6–55.0]38.4 [34.8–42.4]24.7 [20.67–27.2]14.6 [14.6–14.6] < 0.001 *Blood urea* nitrogen (mg/dL)24 [18–32]16 [16–18]23 [20–27]33 [27–41]43 [43–43] < 0.001 Uric acid (mg/dL)6.4 [5.6–7.0]6.1 [5.5–7.2]6.5 [5.9–6.8]6.5 [5.8–7.0]6.8 [6.8–6.8]0.837 Hemoglobin (g/dL)12.6 [11.1–13.9]13.7 [12.8–14.4]12. 5 [10.9–14.5]11.9 [10.6–12.7]12.9 [12.9–12.9]0.005 Albumin (g/dL)3.9 [3.7–4.1]4.1 [3.8–4.3]3.9 [3.8–4.0]3.9 [3.6–4.0]3.8 [3.8–3.8]0.119 Fractional excretion uric acid (%)6.76 [4.80–8.77]6.79 [4.97–8.83]6.27 [4.07–6.84]7.42 [5.03–9.62]0.578 Urine protein-to-creatinine ratio (g/gCr)0.62 [0.09–1.63]0.13 [0.05–0.83]0.38 [0.07–1.07]1.13 [0.34–2.48]2.77 [2.77–2.77]0.017Medication Renin-angiotensin system inhibitor (%)53 (70.7)12 (57.1)17 (70.8)23 (79.3)1 (100.0)0.346 Diuretics (%)16 (21.3)3 (14.3)7 (29.2)6 (20.7)0 (0.0)0.62 Anti hyperuricemic agents (%)38 (50.7)6 (28.6)14 (58.3)17 (58.6)1 (100.0)0.095 The median age was 67 years, and $72.0\%$ of the patients were men. 23 ($30.7\%$) of patients had diabetes mellitus and all patients were Japanese. The median eGFR was 35.7 mL/min/1.73 m2 and median serum UA level was 6.4 mg/dL. There was no significant difference in the use of antihyperuricemic agents and serum UA levels between CKD stages. ## Study 1: change in serum UA levels by CKD stage Figure 2a shows a comparison of serum UA levels before and after dapagliflozin administration. A significant decrease in serum UA level (before administration; 6.4 [5.6–7.0] mg/dL vs. after administration; 5.6 [4.7–6.5] mg/dL, $p \leq 0.001$) was observed before and after dapagliflozin administration. Rate of UA change was − 0.12 [− 0.20 to − 0.04]% by SGLT2i administration. Figure 2Change in serum UA and FEUA at various CKD stages. Figure 2b–e shows the group differences in serum UA changes by CKD stages. Multiple regression analysis did not reveal a relationship between eGFR, change in eGFR and rate of UA change (Table 2).Table 2Indicators of rate of change in serum UA level. Univariate analysisMultivariate analysisβp valueβp valueEstimated glomerular filtration rate (ml/min/1.73m2)− 0.1720.139− 0.1810.117Change of estimated glomerular filtration rate (ml/min/1.73m2)− 0.1870.109− 0.1870.153Uric acid (mg/dL)− 0.1420.224− 0.1420.306Male− 0.0170.886− 0.0170.957DM0.2210.0560.2220.081 ## Study 2: change in FEUA by CKD stage Compared with FEUA before SGLT2i administration, a significant increase in FEUA was observed after dapagliflozin administration ($p \leq 0.001$) (Fig. 2f). After dapagliflozin administration, FEUA was changed 6.76 [4.95–7.38]% to 9.28 [8.43–11.73]%, 6.27 [4.07–6.84]% to 8.15 [6.19–8.52]% and 7.42 [5.81–9.62]% to 9.59 [8.80–11.63]% in CKD stage 3a, 3b and 4, respectively (Fig. 2g–i). Furthermore, multiple regression analysis did not reveal a significant association between eGFR and rate of FEUA change after dapagliflozin administration (Table 3).Table 3Indicators of rate of change in FEUA.Univariate analysisMultivariate analysisβp valueβp valueEstimated glomerular filtration rate (ml/min/1.73m2)0.2520.1455.4120.199Uric acid (mg/dL)− 0.3470.041− 0.3340.059DM− 0.1060.5450.0790.671 Rate of FEUA change did not correlate with rate of serum UA change (Fig. 3).Figure 3Correlation between change in serum UA and FEUA. ## Discussion Our study showed that there was an increase in FEUA and a decrease in serum UA levels in patients with advanced CKD after dapagliflozin administration compared to the values before administration. Serum UA levels have been shown to be associated with CKD progression22,23. In addition, allopurinol has been demonstrated to lower serum UA levels and slow CKD progression10,11. However, there are some reports that allopurinol did not improve renal outcomes12,13. Control of serum UA is important for patients with CKD; however, many patients fail to reach the target serum UA level24. Post-hoc analysis of CANVAS trial revealed that renal function of patients who received canagliflozin was better than that of patients who received placebo, even in patients with urine albumin-creatinine ratio ˂30 mg/gCr25. The renoprotective effect of SGLT2i may not be due to lowering of urinary protein alone, as one of the pleiotropic effects of SGLT2i is reduction of serum UA levels, which may lead to cardiac and renal protection9. Similar to our result, previous studies have shown that SGLT2i increase FEUA and decrease serum UA levels in patients with type 2 DM14–17. Ohashi et al. reported that SGLT2i increased FEUA (before administration; 5.98 ± $2.59\%$ vs. after administration; 7.71 ± $3.22\%$) and reduce serum UA levels (before administration; 6.13 ± 1.36 mg/dL vs. after administration; 5.20 ± 1.11 mg/dL) in patients with type 2 DM whose eGFR was 63.25 ± 24.66 mL/min/1.73 m217. Our results in patients with CKD stage 1 and 2 showed that there was a decrease in serum UA levels and an increase in FEUA after dapagliflozin administration same as previous studies (Supplementary Table 1 and Supplementary Fig. 1)14–17. Furthermore, our study revealed that levels of serum UA decreased (before administration: 6.2 [5.5–6.9] mg/dL vs. after administration: 5.4 [4.7–6.1] mg/dL) and FEUA increased (before administration: 6.70 [4.22–8.37]% vs. after administration: 8.77 [6.44–8.37]%) in patients with CKD without DM (Supplementary Table 2 and Supplementary Fig. 2). The impact of SGLT2i on UA may not depend on renal function or DM. Typically, serum UA is higher in male than in female. Our study showed that SGLT2i reduce serum UA both in male and female patients (Supplementary Fig. 3). The mechanism underlying SGLT2i’s lowering of serum UA levels has not been clearly elucidated. URAT1 is already known as uric acid transporter expressed in proximal tubule26. Nokikov et al. reported that URAT1 and GLUT9 are required for the increase in FEUA in response to canagliflozin27. Chino et al. reported that increased urinary UA excretion may be attributed to glycosuria caused by SGLT2 inhibitors on GLUT9 isoform 2 or any other transporter(s) at the proximal tubule, and may inhibit uric acid reabsorption mediated by GLUT9 isoform 2 at the collecting duct of the renal tubule18. Although SGLT2i increased FEUA and decreased serum UA in our study, change in FEUA and change in serum UA was not associated with eGFR. Furthermore, we could not find a significant association between change in FEUA and change in serum UA. Mechanisms other than urinary excretion such as intestinal excretion of UA may be involved in lowering effects of SGLT2i on serum UA levels in patients with advanced CKD. ABCG2 is a high-capacity urate secretion transporter28. In animal studies, intestinal urate transport by ABCG2 compensated for abnormal urate handling in the setting of declining renal function19. Lu et al. reported that empagliflozin promoted ABCG2 expression in the kidneys and ileum of diabetic mice and attenuated hyperuricemia29. Although these reports suggest that SGLT2i increase urinary UA excretion and promote intestinal extrusion of uric acid through UA transporters, no obvious evidence has yet been found. Further studies are needed to demonstrate the underlying mechanism of SGLT2i’s reduction of serum UA levels. ## Limitations The present study had some limitations. First, this was a single-center retrospective cohort study with a relatively small sample size. Second, we did not investigate the dietary characteristics of each patient. Third, because we did not assess difficult endpoints such as ESRD and CKD progression, we did not show whether reduction of UA by SGLT2i may improve renal outcomes. ## Conclusion Our results suggest that dapagliflozin increases FEUA and decreases serum UA levels in patients with advanced CKD. However, the mechanism of the effect of dapagliflozin on lowering serum UA level has not been fully elucidated, and it is not clear whether this effect can improve renal outcomes. Therefore, further studies are needed to identify the mechanism and impact of SGLT2i on renal outcomes by lowering of serum UA levels. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-32072-y. ## References 1. Imai E. **Prevalence of chronic kidney disease (CKD) in the Japanese general population predicted by the MDRD equation modified by a Japanese coefficient**. *Clin. Exp. Nephrol.* (2007) **11** 156-163. DOI: 10.1007/s10157-007-0463-x 2. **Global, regional, and national burden of chronic kidney disease, 1990–2017: A systematic analysis for the global burden of disease study 2017**. *Lancet* (2020) **395** 709-733. DOI: 10.1016/S0140-6736(20)30045-3 3. Lewis EJ. **Renoprotective effect of the angiotensin-receptor antagonist irbesartan in patients with nephropathy due to type 2 diabetes**. *N. Engl. J. Med.* (2001) **345** 851-860. DOI: 10.1056/NEJMoa011303 4. Brenner BM. **Effects of losartan on renal and cardiovascular outcomes in patients with type 2 diabetes and nephropathy**. *N. Engl. J. Med.* (2001) **345** 861-869. DOI: 10.1056/NEJMoa011161 5. Wanner C. **Empagliflozin and progression of kidney disease in type 2 diabetes**. *N. Engl. J. Med.* (2016) **375** 323-334. DOI: 10.1056/NEJMoa1515920 6. Perkovic V. **Canagliflozin and renal outcomes in type 2 diabetes and nephropathy**. *N. Engl. J. Med.* (2019) **380** 2295-2306. DOI: 10.1056/NEJMoa1811744 7. Heerspink HJL. **Dapagliflozin in patients with chronic kidney disease**. *N. Engl. J. Med.* (2020) **383** 1436-1446. DOI: 10.1056/NEJMoa2024816 8. Leoncini G. **SGLT2is and renal protection: From biological mechanisms to real-world clinical benefits**. *Int. J. Mol. Sci.* (2021) **22** 4441. DOI: 10.3390/ijms22094441 9. Patel DK, Strong J. **The pleiotropic effects of sodium–glucose cotransporter-2 inhibitors: Beyond the glycemic benefit**. *Diabetes Ther.* (2019) **10** 1771-1792. DOI: 10.1007/s13300-019-00686-z 10. Goicoechea M. **Allopurinol and progression of CKD and cardiovascular events: Long-term follow-up of a randomized clinical trial**. *Am. J. Kidney Dis.* (2015) **65** 543-549. DOI: 10.1053/j.ajkd.2014.11.016 11. Siu YP, Leung KT, Tong MKH, Kwan TH. **Use of allopurinol in slowing the progression of renal disease through its ability to lower serum uric acid level**. *Am. J. Kidney Dis.* (2006) **47** 51-59. DOI: 10.1053/j.ajkd.2005.10.006 12. Badve SV. **Effects of allopurinol on the progression of chronic kidney disease**. *N. Engl. J. Med.* (2020) **382** 2504-2513. DOI: 10.1056/NEJMoa1915833 13. Doria A. **Serum urate lowering with allopurinol and kidney function in type 1 diabetes**. *N. Engl. J. Med.* (2020) **382** 2493-2503. DOI: 10.1056/NEJMoa1916624 14. Ferreira JP. **Empagliflozin and uric acid metabolism in diabetes: A post hoc analysis of the EMPA-REG outcome trial**. *Diabetes Obes. Metab.* (2022) **24** 135-141. DOI: 10.1111/dom.14559 15. Davies MJ, Trujillo A, Vijapurkar U, Damaraju CV, Meininger G. **Effect of canagliflozin on serum uric acid in patients with type 2 diabetes mellitus**. *Diabetes Obes Metab.* (2015) **17** 426-429. DOI: 10.1111/dom.12439 16. Sujik DLS. **SGLT2 inhibition and uric acid excretion in patients with type 2 diabetes and normal kidney function**. *Clin. J. Am. Soc. Nephrol.* (2022) **17** 663-671. DOI: 10.2215/CJN.11480821 17. Ohashi N. **Sodium–Glucose cotransporter-2 inhibitor immediately decreases serum uric acid levels in type 2 diabetic patients**. *Med. Sci. Monit.* (2020) **26** e926086. DOI: 10.12659/MSM.926086 18. Chino Y. **SGLT2 inhibitor lowers serum uric acid through alteration of uric acid transport activity in renal tubule by increased glycosuria**. *Biopharm. Drug Dispos.* (2014) **35** 391-404. DOI: 10.1002/bdd.1909 19. Yano H, Tamura Y, Kobayashi K, Tanemoto M, Uchida S. **Uric acid transporter ABCG2 is increased in the intestine of the 5/6 nephrectomy rat model of chronic kidney disease**. *Clin. Exp. Nephrol.* (2014) **18** 50-55. DOI: 10.1007/s10157-013-0806-8 20. Matsuo S. **Revised equations for estimated GFR from serum creatinine in Japan**. *Am. J. Kidney Dis.* (2009) **53** 982-992. DOI: 10.1053/j.ajkd.2008.12.034 21. Kanda Y. **Investigation of the freely available easy-to-use software ‘EZR’ for medical statistics**. *Bone Marrow Transpl.* (2013) **48** 452-458. DOI: 10.1038/bmt.2012.244 22. Iseki K. **Significance of hyperuricemia on the early detection of renal failure in a cohort of screened subjects**. *Hypertens Res.* (2001) **24** 691-697. DOI: 10.1291/hypres.24.691 23. Bellomo G. **Association of uric acid with change in kidney function in healthy normotensive individuals**. *Am. J. Kidney Dis.* (2010) **56** 264-272. DOI: 10.1053/j.ajkd.2010.01.019 24. Koto R, Nakajima A, Horiuchi H, Yamanaka H. **Real-world treatment of gout and asymptomatic hyperuricemia: A cross-sectional study of Japanese health insurance claims data**. *Mod. Rheumatol.* (2021) **31** 261-269. DOI: 10.1080/14397595.2020.1784556 25. Neuen BL. **Effect of canagliflozin on renal and cardiovascular outcomes across different levels of albuminuria: Data from the CANVAS program**. *J. Am. Soc. Nephrol.* (2019) **30** 2229-2242. DOI: 10.1681/ASN.2019010064 26. Enomoto A. **Molecular identification of a renal urate anion exchanger that regulates blood urate levels**. *Nature* (2002) **417** 447-452. DOI: 10.1038/nature742 27. Nokikov A. **SGLT2 inhibition and renal urate excretion: Role of luminal glucose, GLUT9, and URAT1**. *Am. J. Physiol. Ren. Physiol.* (2019) **316** F173-F185. DOI: 10.1152/ajprenal.00462.2018 28. Matsuo H. **Common defects of ABCG2, a high-capacity urate exporter, cause gout: A function-based genetic analysis in a Japanese population**. *Sci. Transl. Med.* (2009) **1** 5ra11. DOI: 10.1126/scitranslmed.3000237 29. Lu YH. **Empagliflozin attenuates hyperuricemia by upregulation of ABCG2 via AMPK/AKT/CREB signaling pathway in type 2 diabetic mice**. *Int. J. Biol. Sci.* (2020) **16** 529-542. DOI: 10.7150/ijbs.33007
--- title: The m6A reader PRRC2A is essential for meiosis I completion during spermatogenesis authors: - Xinshui Tan - Caihong Zheng - Yinghua Zhuang - Pengpeng Jin - Fengchao Wang journal: Nature Communications year: 2023 pmcid: PMC10039029 doi: 10.1038/s41467-023-37252-y license: CC BY 4.0 --- # The m6A reader PRRC2A is essential for meiosis I completion during spermatogenesis ## Abstract N6-methyladenosine (m6A) and its reader proteins YTHDC1, YTHDC2, and YTHDF2 have been shown to exert essential functions during spermatogenesis. However, much remains unknown about m6A regulation mechanisms and the functions of specific readers during the meiotic cell cycle. Here, we show that the m6A reader Proline rich coiled-coil 2A (PRRC2A) is essential for male fertility. Germ cell-specific knockout of Prrc2a causes XY asynapsis and impaired meiotic sex chromosome inactivation in late-prophase spermatocytes. Moreover, PRRC2A-null spermatocytes exhibit delayed metaphase entry, chromosome misalignment, and spindle disorganization at metaphase I and are finally arrested at this stage. Sequencing data reveal that PRRC2A decreases the RNA abundance or improves the translation efficiency of targeting transcripts. Specifically, PRRC2A recognizes spermatogonia-specific transcripts and downregulates their RNA abundance to maintain the spermatocyte expression pattern during the meiosis prophase. *For* genes involved in meiotic cell division, PRRC2A improves the translation efficiency of their transcripts. Further, co-immunoprecipitation data show that PRRC2A interacts with several proteins regulating mRNA metabolism or translation (YBX1, YBX2, PABPC1, FXR1, and EIF4G3). Our study reveals post-transcriptional functions of PRRC2A and demonstrates its critical role in the completion of meiosis I in spermatogenesis. Modification of RNA with m6A has been shown to be important during spermatogenesis. Here they identify post-transcriptional functions of PRRC2A, showing it promotes transcriptome transition from spermatogonia to spermatocytes and the translation of genes related to cell division. ## Introduction N6-methyladenosine (m6A), the most prevalent modification of eukaryotic mRNA, post-transcriptionally regulates many physiological and pathological processes1. The dynamics of m6A is coordinated by the interplay between the m6A “writer” and “eraser” proteins. In mammalian cells, the methyltransferase complex comprising METTL3 and METTL14 acts as the writer to catalyze the m6A methylation onto RNAs2,3. Conversely, erasers FTO4 and ALKBH55 remove these m6A modifications. Many reader proteins have been discovered recently, including YTH family proteins6–13, HNRNP family proteins14–16, IGF2BPs17, eIF318, FMRP19, SND120, and PRRC2A21. Reader proteins generally act as functional mediators to regulate the splicing, transportation, stability/degradation, storage, and translation status of m6A-modified RNAs22. In a previous study, PRRC2A was demonstrated as an m6A reader and regulates oligodendroglial specification in brain development by promoting the stability of Olig2 mRNA21. During spermatogenesis, male germ cells develop from spermatogonia to spermatocytes, and then to spermatids23. In prophase I, germ cells undergo pairing and recombination of homologous chromosomes. Based on the appearance of chromosomes, prophase I is divided into four substages named leptotene, zygotene, pachytene, and diplotene24. Notably, in pachytene and diplotene spermatocytes, transcription on X and Y chromosomes is silenced, which is called meiotic sex chromosome inactivation (MSCI)24. Subsequently, homologous chromosomes and sister chromosomes separate during meiosis I and meiosis II, respectively23. Similar to somatic cells, the centrosome functions in regulating spindle and chromosome behaviors during the metaphase of male meiosis25,26. Proper chromosome alignment is dependent on normal kinetochore-microtubule attachment and spindle organization27. When the kinetochore-microtubule attachment or chromosome alignment is disrupted, the spindle assembly checkpoint (SAC) was activated and this prevents the metaphase-to-anaphase transition to avoid the formation of aneuploidy gametes during metaphase of meiosis28. A previous study showed that m6A-modified transcripts exist in all types of male germ cells11,29. Correspondingly, m6A writers, METTL3 and METTL14, were shown to regulate the maintenance of spermatogonial stem cells, spermatogonia differentiation, meiosis initiation, and spermiogenesis29,30. And the ablation of the m6A eraser, ALKBH5, leads to the depletion of pachytene spermatocytes and spermatids in testes5. FTO, another eraser, modulates the cell cycle in the mouse GC-1 spermatogonial cell line31. m6A is required for the proper development at all stages of spermatogenesis. As the functional mediators of m6A, several readers have been reported to regulate spermatogenesis. Specifically, YTHDC2 was shown to regulate meiosis prophase by modulating mRNA abundance and translation efficiency7,8,11,32,33, but recent studies suggested that the functions of YTHDC2 in spermatogenesis are independent of m6A recognition34,35. YTHDC1 is required for the survival and development of spermatogonia36. YTHDF2 facilitates cell proliferation and cell adhesion of GC-1 cells by modulating mRNA stability, while ablation of YTHDF2 results in male subfertility likely by affecting spermatid development37,38. However, there are no reports about how m6A-modified RNAs are regulated during meiotic metaphase or about whether an m6A reader is involved. Here, we show that PRRC2A is highly expressed in testes and is essential for male fertility. PRRC2A-null spermatocytes show defective XY synapsis and MSCI at late prophase. PRRC2A deficiency leads to delayed metaphase entry, chromosome misalignment, spindle abnormalities, and finally arrests at metaphase I. Combining RNA-seq, Ribo-seq, and PRRC2A RIP-seq, we found that PRRC2A recognizes and downregulates spermatogonia-specific transcripts during meiosis prophase to promote the transcriptome transition from spermatogonia to spermatocytes. PRRC2A binds transcripts of genes involved in meiotic cell division and enhances their translation efficiency. Further, we found that PRRC2A interacts with YBX1, YBX2, PABPC1, FXR1, and EIF4G3, and potentially regulates mRNA metabolism and translation together with these cofactors. Our study reveals the post-transcriptional functions of PRRC2A and its essential role in the completion of meiosis I during spermatogenesis. ## PRRC2A is essential for male fertility We found that Prrc2a was highly expressed in the adult testes (Fig. 1a). Immunoblotting showed that the PRRC2A protein was broadly expressed during the first wave of spermatogenesis, and we observed a rapid increase in the PRRC2A level starting from postnatal day 14 and continuing to day 22 (P14-22) (Fig. 1b), the period when early pachytene spermatocytes appear and develop through meiosis I and meiosis II to produce early round spermatids39. RNA in situ hybridization showed that PRRC2A mRNA was expressed at low levels in spermatogonia, spermatocytes from preleptotene stage to early/mid-pachytene stage, and step 7-10 spermatids, and was highly expressed in spermatocytes from late-pachytene stage to metaphase stage and step 1–6 spermatids (Supplementary Fig. 1a). Similarly, immunostaining showed that the PRRC2A protein was localized in the cytoplasm of all types of germ cells except elongated spermatids, and had stronger signals and a granular distribution in both late-pachytene spermatocytes and round spermatids (Fig. 1c). In the testis, a type of germ cell-specific RNA granule known as chromatoid body (marked by DDX4 and MIWI) exists in late pachytene spermatocytes up until the round spermatid stage40. Co-staining of PRRC2A with DDX4 and MIWI confirmed that PRRC2A colocalizes with chromatoid bodies (Fig. 1c, Supplementary Fig. 1b).Fig. 1PRRC2A is highly expressed in testes and is essential for spermatogenesis.a qPCR analysis of Prrc2a mRNA levels in various organs of adult mice. Two-sided student’s t-test. Error bars, $$n = 3$$ mice, mean ± SEM. Source data are provided as a Source Data file. b WB analysis of PRRC2A protein levels in mice testes of indicated ages. c Immunostaining of PRRC2A and DDX4 in P60 testis sections. The right panels show enlarged images of indicated areas. Spg spermatogonia, L leptotene spermatocyte, eP early pachytene spermatocyte, lP late-pachytene spermatocyte, rS round spermatid. Arrowheads indicate chromatoid bodies. Scale bar, 20 μm. d Morphology of representative testes from P60 control and Prrc2a-cko mice. e Ratios of testis weight to body weight of P60 control and Prrc2a-cko mice ($$n = 7$$). Two-sided student’s t-test. Error bars, mean ± SEM. $p \leq 0.0001.$ *Source data* are provided as a Source Data file. f H&E staining in testis and epididymis sections of P60 control and Prrc2a-cko mice. Stages of the seminiferous epithelial cycle are indicated. Arrows indicate round spermatids with abnormal nuclear morphology. Arrowheads indicate detached round spermatids. Scale bar, 50 μm (up) and 100 μm (down). g TUNEL staining in P60 control and Prrc2a-cko testis sections. Arrowheads indicate germ cells with TUNEL signals. M, metaphase spermatocyte; rS, round spermatid. Scale bar, 50 μm. h Numbers of TUNEL-positive cells per tubule of the indicated stage in P60 control and Prrc2a-cko testes ($$n = 3$$). Two-sided student’s t-test. Error bars, mean ± SEM. $$p \leq 0.0010$$ (for I–III), 0.0539 (for IV–VII), 0.0103 (for VIII–X), 0.0161 (for XI–XII). Source data are provided as a Source Data file. i Numbers of TUNEL-positive cells per tubule in control and Prrc2a-cko testes ($$n = 3$$) of indicated ages. Two-sided student’s t-test. Error bars, mean ± SEM. $$p \leq 0.4448$$ (for P18), 0.0149 (for P20), 0.0308 (for P22), 0.0015 (for P24), 0.0071 (for P28), 0.0016 (for P60). Source data are provided as a Source Data file. To investigate the function of PRRC2A in spermatogenesis, we generated a mouse line with germ cell-specific knockout of Prrc2a (Supplementary Fig. 2a). Briefly, CRISPR-Cas9 technology was used to place loxP sites on both sides of the region from exon 2 to exon 5 in the mouse Prrc2a locus to generate Prrc2afl/fl mice. Then, Stra8-Cre transgenic mice were crossed with Prrc2afl/fl mice to generate Prrc2a-cko (Stra8-Cre; Prrc2afl/fl or Stra8-Cre; Prrc2afl/Δ) mice, from which Prrc2a was knocked out starting from differentiated spermatogonia and persisting through the later stages of the male germline (Supplementary Fig. 2a, b). After confirming the knockout of Prrc2a by qPCR and immunoblotting in P60 Prrc2a-cko testes (Supplementary Fig. 2c, d), we found that Prrc2a-cko adult male mice were infertile with a $37\%$ reduction in testis weight compared to littermate controls (Fig. 1d, e). They mated with female mice (which formed normal vaginal plugs), but no offspring were produced (Supplementary Fig. 2e). In Prrc2a-cko testes, the number of round spermatids decreased dramatically, and no elongating spermatids or elongated spermatids were detected (Fig. 1f). There were round spermatids that appeared with abnormal nuclear morphology and detached from the seminiferous epithelium (Fig. 1f). In the Prrc2a-cko epididymis, lumens did not contain sperms or had defective round spermatids (Fig. 1f, Supplementary Fig. 3b). We next performed co-staining of the acrosome marker PNA (peanut agglutinin) and DDX4 in testes from P18 to P24 mice to investigate the development of round spermatids (Supplementary Fig. 3a). At P18, there were no spermatids in the control or Prrc2a-cko testis. From P20 to P24, round spermatids were generated in large quantities and developed to steps 7–9 in control testes. In contrast, only a few round spermatids were produced in Prrc2a-cko testes, and their development did not advance beyond steps 4–6. Additionally, we found many multinucleated germ cells which appeared like incompletely divided spermatocytes or as unseparated round spermatids in P60 Prrc2a-cko testes and epididymides (Supplementary Fig. 3b). Most multinucleated cells contained two nuclei, and a few contained three or more nuclei, phenotypes indicative of impaired meiotic cell division41 (Supplementary Fig. 3c). We then performed TUNEL staining and found increased apoptosis in Prrc2a-cko testes, specifically noting that most of the apoptotic cells were metaphase spermatocytes or round spermatids (Fig. 1g, Supplementary Fig. 4a). Further, we staged seminiferous tubules based on the morphology and composition of germ cells and pattern of γH2AX staining (Supplementary Fig. 3d) and found that the number of apoptotic cells was greatly increased in stage I–III tubules (which contain newly produced round spermatids) (Fig. 1h). These results established that many round spermatids undergo apoptosis immediately after being generated in the Prrc2a-cko testes. Further, we examined apoptosis in P18 to P28 testes: compared with control testis, the number of apoptotic cells in Prrc2a-cko testes was similar at P18 but then increased significantly from P20 to P28 and kept at P60 (Fig. 1i). Meiotic cell division occurs around P18 to P20 during the first wave of spermatogenesis, indicating defects in Prrc2a-cko metaphase spermatocytes. ## PRRC2A modulates XY synapsis and MSCI To investigate apoptosis in metaphase I spermatocytes, we first examined DNA double-strand breaks (DSBs) and homologous chromosome synapsis by staining γH2AX and SYCP3 on chromosome spreads of meiotic prophase spermatocytes (Fig. 2a). PRRC2A-null prophase spermatocytes develop normally through the leptotene, zygotene, pachytene, diplotene stages, and the proportion of each type of spermatocytes was similar to controls (Supplementary Fig. 4b). γH2AX signals in PRRC2A-null spermatocytes were distributed around all chromosomes at the leptotene and zygotene stages, then disappeared from autosomes but remained in XY body at the pachytene and diplotene stages, outcomes similar to the control spermatocytes (Fig. 2a). SYCP1 and SYCP3 are essential components of the synaptonemal complex, we examined their expression on chromosomes and found autosomes in PRRC2A-null spermatocytes synapsed at the pachytene stage and separated at the diplotene stage as controls (Fig. 2a, Supplementary Fig. 4c). However, there were dramatically more pachytene spermatocytes with XY asynapsis in Prrc2a-cko testes (for early pachytene, control: $5.4\%$ vs Prrc2a-cko: $22.0\%$; for late-pachytene, control: $5.9\%$ vs Prrc2a-cko: $16.4\%$) (Fig. 2b, c). During the process of meiotic DSB repair, DMC1 and MLH1 are required for strand invasion and crossover formation, respectively24. Immunostaining showed that the number of DMC1 foci did not change after PRRC2A deletion (Fig. 2d, e). The number of MLH1 foci in PRRC2A-null spermatocytes ($$n = 22$.52$) was similar to the number in controls ($$n = 22$.59$). But in spermatocytes with asynapsed XY, the number was decreased ($$n = 21$.56$), corresponding to the XY asynapsis (Fig. 2f, g).Fig. 2PRRC2A deficiency leads to XY asynapsis and impaired MSCI.a, b Immunostaining of γH2AX and SYCP3 on chromosome spreads of control and PRRC2A-null spermatocytes. b Represent PRRC2A-null spermatocytes with XY asynapsis. Scale bar, 10 μm. c Percentage of early or late spermatocytes with XY asynapsis in total spermatocytes from P60 control and Prrc2a-cko testes. More than 500 chromosome spreads of spermatocytes from 4 mice were counted in each group of control and Prrc2a-cko. Two-sided student’s t-test. Error bars, mean ± SEM. $$p \leq 0.0028$$ (for eP), 0.0011 (for lP). Source data are provided as a Source Data file. d, e Immunostaining of DMC1 and SYCP3 on chromosome spreads and quantification of DMC1 foci distributed on chromosomes in control and Prrc2a-cko early-zygotene (Control, $$n = 31$$; Prrc2a-cko, $$n = 32$$), late-zygotene (Control, $$n = 38$$; Prrc2a-cko, $$n = 34$$), and early pachytene (Control, $$n = 57$$; Prrc2a-cko, $$n = 48$$) spermatocytes. Two-sided student’s t-test. Error bars, mean ± SEM. $$p \leq 0.9496$$ (for eZ), 0.1636 (for lZ), 0.4683 (for eP). Scale bar, 10 μm. Source data are provided as a Source Data file. f, g Immunostaining of MLH1 and SYCP3 on chromosome spreads and quantification of MLH1 foci distributed on chromosomes in control ($$n = 111$$ cells from three mice) and Prrc2a-cko pachytene spermatocytes with synapsed XY-synapsis (CKO XY-synapsis, $$n = 166$$ cells from three mice) or asynapsed XY (CKO XY-asynapsis, $$n = 41$$ cells from three mice). Two-sided student’s t-test. Error bars, mean ± SEM. ns $$p \leq 0.7998$$, *$$p \leq 0.0147$$, **$$p \leq 0.0065.$$ Scale bar, 10 μm. Source data are provided as a Source Data file. h Immunostaining of POL II and SYCP3 on chromosome spreads of control and PRRC2A-null spermatocytes. Circles indicate XY bodies. Scale bar, 10 μm. i Quantification of signal intensity of POL II within XY body of control and PRRC2A-null spermatocytes. eP early pachytene spermatocyte ($$n = 39$$, 39); lP late-pachytene spermatocyte ($$n = 31$$, 26); D diplotene spermatocyte ($$n = 46$$, 28). Two-sided student’s t-test. Error bars, mean ± SEM. ** $$p \leq 0.0021$$, ***$p \leq 0.0001.$ *Source data* are provided as a Source Data file. j XY regions of control and Prrc2a-cko pachytene spermatocytes immunostained with silencing factors (MDC1 or ATR) and SYCP3. Scale bar, 2 μm. Note that XY asynapsis and impaired MSCI frequently co-occur in defective spermatocytes, such as those from Brdt-/-42 and Raptor-/-43 mice, we further examined the expression pattern of POL II (Fig. 2h). Compared to control spermatocytes, there were increased POL II signals in the XY region of PRRC2A-null pachytene and diplotene spermatocytes, clearly suggesting the disruption of MSCI (Fig. 2i). Moreover, consistent with the notion that defective MSCI causes mid-pachytene arrest and apoptosis44,45, we found PRRC2A deficiency caused noticeably more apoptosis in mid-pachytene spermatocytes compared with control (Supplementary Fig. 4a). We next examined the expression of two MSCI factors, MDC146 and ATR47, and found that their expression pattern in the XY region was not affected in PRRC2A-null spermatocytes whether the sex chromosomes were synapsed or not (Fig. 2j). The impaired MSCI may be caused by the downstream epigenetic abnormalities in the XY region, such as H3K9me348. Surprisingly, fluorescence-activated cell sorting (FACS) analysis of cellular composition showed the normal number of early prophase spermatocytes but noticeably fewer late prophase spermatocytes in Prrc2a-cko P20 testes, although the proportion of late prophase spermatocytes became normal in adult Prrc2a-cko testes (Supplementary Fig. 4d). These results indicate a delayed progression from zygonema to pachynema and this delay may result in increased apoptosis starting from the early pachytene stage (Supplementary Fig. 4a). ## PRRC2A deficiency causes delayed entry and arrest at meiotic metaphase To characterize meiotic metaphase, we performed immunostaining against the metaphase cell marker phospho-Histone H3 (Thr3) (pH3) (Fig. 3a). In Prrc2a-cko adult (P60) testes, the total number of pH3-positive (pH3+) spermatocytes did not differ from control testes (Fig. 3b), but the number of tubules containing pH3+ spermatocytes was significantly increased (Fig. 3c). We then quantified the number of pH3+ metaphase spermatocytes in tubules at different stages to examine their distribution (Fig. 3d). In control testes, over $95\%$ metaphase spermatocytes appeared in the XI–XII stage tubules. However, in Prrc2a-cko testes, a large number of metaphase spermatocytes were also detected in stage I–III tubules, and even some in stage IV–X tubules (Fig. 3a, d). Previous studies have shown that cell division is delayed by the SAC when chromosomes cannot attach to or align on the spindle properly, and these defective cells will undergo apoptosis if defects are not repaired in time28. Interestingly, the distribution pattern of pH3+ spermatocytes was similar to the distribution pattern of apoptotic cells in Prrc2a-cko testes (Fig. 1h), indicating that PRRC2A deficiency potentially leads to impairment of meiotic cell division and accounts for the increased apoptosis in stage I–X tubules. Fig. 3PRRC2A deficiency leads to chromosome misalignment and spindle disorganization.a Immunostaining of pH3 and γH2AX in P60 control and Prrc2a-cko testis sections. Seminiferous tubule stages are indicated. Arrowheads indicate metaphase spermatocytes. Scale bar, 50 μm. b The number of pH3+ spermatocytes per seminiferous tubule in control and Prrc2a-cko testes of indicated ages ($$n = 4$$ mice for P18, P20, P60; $$n = 3$$ mice for P22, P24, P28, P35). Two-sided student’s t-test. Error bars, mean ± SEM. $$p \leq 0.7785$$ (for P18), 0.0149 (for P20), 0.0334 (for P22), 0.0231 (for P24), 0.0250 (for P28), 0.6855 (for P35), 0.4705 (for P60). Source data are provided as a Source Data file. c The ratio of seminiferous tubules with pH3+ spermatocytes to total tubules in control and Prrc2a-cko testes of indicated ages ($$n = 4$$ mice for P18, P20, P60; $$n = 3$$ mice for P22, P24, P28, P35). Two-sided student’s t-test. Error bars, mean ± SEM. $$p \leq 0.7580$$ (for P18), 0.0014 (for P20), 0.1047 (for P22), 0.1941 (for P24), 0.0092 (for P28), 0.0456 (for P35), 0.0054 (for P60). Source data are provided as a Source Data file. d The number of pH3+ spermatocytes per seminiferous tubules of the indicated stage in P60 control and Prrc2a-cko testes. Two-sided student’s t-test. Error bars, $$n = 4$$ mice, mean ± SEM. $$p \leq 0.0014$$ (for I–III), 0.0591 (for IV–VII), 0.0242 (for VIII–X), 0.0176 (for XI–XII). Source data are provided as a Source Data file. e H&E staining in P60 control and Prrc2a-cko testes sections. Arrowheads indicate misaligned chromosomes in metaphase I spermatocytes. Scale bar, 20 μm. f Representative metaphase I spermatocytes in P60 control and Prrc2a-cko testis sections immunostained with α-tubulin and γ-tubulin. White arrowheads indicate misaligned chromosomes, yellow arrowheads indicate scattered γ-tubulin foci outside the spindle pole, yellow arrows indicate disintegrated MTOC at the spindle pole, white arrows indicate spindle poles. Scale bar, 5 μm. g Percentage of metaphase I spermatocytes with chromosomes misalignment ($$n = 4$$ biological independent mice), scattered γ-tubulin foci, disintegrated MTOC, and disorganized spindle ($$n = 3$$ biological independent mice) in P60 control and Prrc2a-cko testes. Two-sided student’s t-test. Error bars, mean ± SEM. $$p \leq 0.0007$$, 0.0018, 0.0423, 0.0057. Source data are provided as a Source Data file. h Fluorescence intensity of γ-tubulin at the spindle pole relative to that in control ($$n = 97$$, 100 cells). Two-sided student’s t-test. Error bars, mean ± SEM. *** $p \leq 0.0001.$ *Source data* are provided as a Source Data file. i, j The average length and width of the alignment and spindles in control and Prrc2a-cko metaphase I spermatocytes immunostained with α-tubulin and γ-tubulin ($$n = 61$$ cells for control, $$n = 64$$ cells for Prrc2a-cko). Two-sided student’s t-test. Error bars, mean ± SEM. The value of each data is 10.21 ± 0.16, 10.07 ± 0.13, 6.07 ± 0.15, 7.55 ± 0.15, 9.05 ± 0.32, 10.25 ± 0.22, 7.25 ± 0.16, 7.38 ± 0.15 from left to right. $$p \leq 0.5051$$ (ns), <0.0001 (***), 0.0018 (**), 0.5759 (ns). Scale bar, 5 μm. Source data are provided as a Source Data file. Moreover, the decreased number of metaphase spermatocytes in XI–XII stages tubules of Prrc2a-cko adult testes also indicated delayed metaphase entry (Fig. 3d). We further examined the production of metaphase spermatocytes in juvenile testes (Fig. 3b, c). At P18, few metaphase spermatocytes were observed in both control and Prrc2a-cko testes. For control mice, a large number of metaphase spermatocytes were produced starting from P20 (Fig. 3b). But we observed decreased numbers of metaphase spermatocytes in Prrc2a-cko testes from P20 to P28, showing a slower increasing trend than controls (Fig. 3b). In control testes, around $10\%$ of the seminiferous tubules contained metaphase spermatocytes at P20, P22, and P24 (Fig. 3c). However, <$8\%$ of tubules contained metaphase spermatocytes in Prrc2a-cko testes from P20 to P24 (Fig. 3c). These results suggest PRRC2A-null spermatocytes exhibited delayed metaphase entry. Moreover, from P28 to adult, the percentage of tubules contained metaphase spermatocytes in Prrc2a-cko testes (over $10\%$) was more than the percentage in controls (around $8\%$), also suggesting the metaphase arrest after PRRC2A deficiency. ## PRRC2A facilitates chromosome alignment and spindle assembly The defective developmental progress of metaphase spermatocytes promoted us to investigate the influence of PRRC2A on critical processes during meiotic cell division. We observed chromosome misalignment in PPRC2A-null metaphase I spermatocytes (Prrc2a-cko: $26.75\%$ vs control: $4.75\%$) (Fig. 3e, g). To further characterize the spindle morphology and chromosome behaviors, we performed the co-staining of α-tubulin and γ-tubulin in P60 testes sections (Fig. 3f). In PRRC2A-null spermatocytes, the misaligned chromosomes localized outside of the spindle (Fig. 3f2). γ-tubulin, a ubiquitous component of microtubule organizing centers (MTOCs), accumulates at spindle poles in control spermatocytes. In contrast, γ-tubulin foci in PRRC2A-null spermatocytes were scattered outside the spindle pole (Prrc2a-cko: $28.84\%$ vs control: $0.94\%$) (Fig. 3f3, g) or were disintegrated at the spindle pole, phenotypes indicating the disintegrated MTOC (Prrc2a-cko: $11.09\%$ vs control: $2.22\%$) (Fig. 3f4, g). Moreover, the fluorescence intensity of γ-tubulin at spindle poles in PRRC2A-null spermatocytes was reduced to $67\%$ of that in control spermatocytes, suggesting defects in MTOC formation (Fig. 3h). Given that MTOCs are known to function in spindle assembly49, the spindle morphology possibly was affected by PRRC2A deficiency. As expected, we observed disorganized spindles in some PRRC2A-null metaphase I spermatocytes (Prrc2a-cko: $10.87\%$ vs control: $0.25\%$), and the spindle was asymmetric with more than two spindle poles (Fig. 3f5, g). We next measured the length and width of alignment and spindles and observed no disruption of normal alignment length and spindle length in PRRC2A-null spermatocytes, but the alignment width and spindle width were significantly increased in PRRC2A-null spermatocytes (Fig. 3i, j). Additionally, the intersection angle between the spindle polarity axis and the equatorial plate did not differ from the control cells (Supplementary Fig. 4e, f). Thus, PRRC2A is required for proper chromosome alignment and spindle morphology, and PRRC2A deficiency most likely activated the spindle assembly checkpoint and caused metaphase I arrest. Past studies suggested the lack of kinetochore-microtubule attachment or impaired centrosome assembly leads to defective alignment and spindle27,28,49. So, we performed the co-staining of CREST (a marker of kinetochore) and α-tubulin and found normal kinetochore-microtubule attachment in PRRC2A-null spermatocytes, even in the case of chromosome misalignment and spindle disorganization (Fig. 4a). To examine centrosome assembly, we examined the expression of CEP192 (a protein involved in centrosome assembly) and found that the signal intensity of CEP192 at the spindle pole was significantly reduced—or even disappeared—in PRRC2A-null spermatocytes (Fig. 4b, e). Previous studies have shown that CEP192 is required for spindle formation in oocytes50 and CEP192 deficiency in Hela cells leads to malformed spindles, abnormal chromosomal alignment, and formation of multinucleated cells51,52, findings similar to our results (Supplementary Fig. 3b, c). The immunoblotting also showed the decreased protein abundance of CEP192 and CEP152 (a centrosome protein required for the organization of MTOC in meiosis of mouse oocytes50) in PRRC2A-null spermatocytes (Figs. 4c, d and 7e), indicating the centrosome defects. Fig. 4PRRC2A deficiency leads to impaired centrosome and downregulation of MPF.a Representative metaphase I spermatocytes in P60 control and Prrc2a-cko testes sections immunostained with CREST and α-tubulin. Scale bar, 5 μm. b Representative metaphase I spermatocytes in P60 control and Prrc2a-cko testes sections immunostained with CEP192, α-tubulin. Arrows indicate spindle poles. Scale bar, 5 μm. c WB test of indicated proteins in control and PRRC2A-null spermatocytes from mice. d Quantification of the relative level of indicated proteins in control and PRRC2A-null spermatocytes. Two-sided student’s t-test. Error bars, $$n = 3$$ mice, mean ± SEM. ns $$p \leq 0.1088$$ (for CCNB1), 0.1735 (for CCNA2), **$$p \leq 0.0022$$, ***$p \leq 0.0001.$ *Source data* are provided as a Source Data file. e Quantification of fluorescence intensity of CEP192 foci at spindle pole ($$n = 69$$, 79 cells). Two-sided student’s t-test. Error bars, mean ± SEM. *** $p \leq 0.0001.$ *Source data* are provided as a Source Data file. Additionally, we examined the metaphase-promoting factor (MPF), which is required for prophase-to-metaphase transition in both mitosis and meiosis53. MPF is composed of cyclin-dependent kinase 1 (CDK1) and cyclin B1 (CCNB1). We found decreased protein abundance of CDK1 which was consistent with the observation of metaphase entry delay, although the level of CCNB1 was normal (Fig. 4c, d). We also examined the expression of mitotic marker CCNA254 and found its protein abundance was not affected in PRRC2A-null spermatocytes (Fig. 4c, d). ## PRRC2A reduces abundance and promotes translation of its targets The majority of cytoplasmic m6A readers have been shown to regulate m6A-modified transcripts by affecting their mRNA stability and/or their translation efficiency1. To explore the mechanism(s) through which PRRC2A regulates spermatogenesis, we purified Hoechst 33342 stained prophase spermatocytes (from leptotene to diplotene) from adult Prrc2a-cko and control testes using FACS according to previously reported method55(Supplementary Fig. 4d), then assessed their transcriptome and translatome There were 2033 differentially transcribed genes (DTG, up: 1140 and down: 893) and 686 differential translation efficiency genes (DTEG, up: 287 and down: 399) (Fig. 5a, Supplementary Data 1). Among them, the mRNA abundance and translation efficiency of 476 genes were both differentially changed (DTG&DTEG), 1557 genes were exclusive DTGs, and 210 genes were exclusive DTEGs (Fig. 5a). Gene ontology (GO) analysis indicated that the downregulated DTGs were enriched for annotated functions related to spermatogenesis and cilium movement, but no enrichment for spermatogenesis-related terms was found among the upregulated DTGs (Fig. 5b). Conversely, terms related to male production were enriched among upregulated DTEGs, and terms enriched among downregulated DTEGs were not related to spermatogenesis (Fig. 5c). Thus, transcriptome and translational efficiency profile were altered by PRRC2A deficiency and showed the opposite change patterns. Fig. 5PRRC2A decreases the mRNA abundance and improves the translation efficiency of its targets.a Scatter plots of Ribo-seq vs RNA-seq in control and PRRC2A-null spermatocytes. Genes were classified according to their regulation with the parameters indicated criteria. Gray dots (no sig) indicate genes with no significant change. Green dots (exclusive DTG) indicate genes with significant changes in RNA abundance (fold change (FC) > 1.5, p-adjusted < 0.05) and with no change in translation efficiency. Blue dots (exclusive DTEG) indicate genes with significant changes (FC > 1.5, p-value < 0.05) in translation efficiency and with no change in RNA abundance. Red dots (DTG&DTEG) indicate genes with significant changes in RNA abundance and translation efficiency. b, c Top GO terms in biological process categories of down and upregulated DTGs (b) and DTEGs (c). d The pie chart shows the distribution region of PRRC2A-binding peaks on transcripts in one of two repeats. e Distribution of PRRC2A-binding peaks along with transcripts. f Consensus binding motif of PRRC2A identified by HOMER ($$p \leq 1$$e−73). g Scatter plots of Ribo-seq vs RNA-seq in control and PRRC2A-null spermatocytes. Genes were classified into groups according to indicated criteria. Gray dots (non-target) indicate genes with no PRRC2A-binding sites. Orange dots indicate genes with PRRC2A-binding sites. Red dots indicate genes with overlapped PRRC2A-binding and m6A-modified sites. h, i Cumulative distribution of RNA abundance (left two panels) and translational efficiency (right two panels) changes between control and PRRC2A-null spermatocytes. Top two panels show non-targets (blue), PRRC2A-RIP targets (red) (H). Bottom two panels show targets with PRRC2A-binding m6A-not-modified sites (blue) and targets with one (green), two (orange), and more than three (red) PRRC2A-binding m6A-modified sites (I). p-values were calculated using two-sided Wilcoxon test. Additionally, the majority of X and Y chromosome-linked genes were upregulated in PRRC2A-null spermatocytes (168 up vs. 8 down) (Supplementary Fig. 6a), a trend corresponding to the impaired MSCI we observed in PRRC2A-null spermatocytes (Fig. 2h, i). We also confirmed the upregulation of several representative genes by qPCR (Supplementary Fig. 6b). To identify the targets of PRRC2A, we inserted a 3×flag tag after the Prrc2a CDS region to generate Prrc2a-flag transgenic mice by CRISPR-Cas9 (Supplementary Fig. 5a, b) and tested whether these mice were suitable for the subsequent PRRC2A-RIP experiment. We found that the PRRC2A-Flag proteins were expressed in transgenic testes with similar abundance to controls and were efficiently pulled down using the anti-Flag antibody (Supplementary Fig. 5c). The immunofluorescence co-staining of Flag, PRRC2A, and DDX4 showed that PRRC2A-Flag proteins had the same expression pattern as PRRC2A (Fig. 1c, Supplementary Fig. 5d). These results supported the use of this transgenic mouse line for subsequent immunoprecipitation (IP) experiments. We then performed PRRC2A RIP-seq using P20 Prrc2a-flag mouse testes to identify candidate binding target RNAs of PRRC2A. A total of 9442 peaks from 5,981 genes were identified in two biological replicates (Supplementary Data 2), most of which ($94.31\%$ and $96.42\%$ for each replicate) were for mRNAs (Fig. 5d and Supplementary Fig. 6c). PRRC2A-binding peaks were enriched in the CDS and 3'UTR region of mRNAs and were highly enriched near the stop codon (Fig. 5d, e), results consistent with the previously reported distribution patterns for m6A modification generally56,57 and for PRRC2A-binding peaks specifically21. A motif analysis identified the known consensus m6A motif GGACU56,57 among PRRC2A-binding peaks (Fig. 5f). Further, we overlapped our PRRC2A-binding peak dataset with m6A-modified peaks from a previously reported P20 MeRIP-seq dataset11 and found that 3366 peaks (from 2263 transcripts) of 9442 ($35.65\%$) PRRC2A-binding peaks carried m6A modifications (Supplementary Fig. 6d, Supplementary Data 3). These findings support that PRRC2A functions as an m6A reader in testes. We then followed PRRC2A-binding transcripts and reported m6A-modified sites in the RNA-seq and Ribo-seq data and 4081 PRRC2A-binding targets (1954 of them contain m6A modifications) were obtained (Fig. 5g). PRRC2A targets showed more increased RNA abundance compared with non-targets in PRRC2A-null spermatocytes (Fig. 5h). Conversely, translational efficiency (TE) of PRRC2A targets became lower than non-targets in PRRC2A-null spermatocytes (Fig. 5h). Remarkably, as the number of PRRC2A-binding m6A-modified peaks increased, the RNA abundance increased and the TE decreased correspondingly, and consequently, transcripts with more than three binding peaks exhibited the highest RNA abundance and the lowest TE (Fig. 5i). Therefore, PRRC2A reduces the abundance of its targets and improves their translation efficiency during the male meiotic prophase. Moreover, PRRC2A-binding methylated transcripts showed higher RNA abundance and lower TE than PRRC2A-binding targets without m6A modifications, this result indicated that PRRC2A exerts functions dependently on m6A and confirmed its role as an m6A reader (Fig. 5i). ## PRRC2A promotes transcriptome transition from spermatogonia to spermatocytes To better understand the expression changes in PRRC2A-null spermatocytes, we compared our RNA-seq and Ribo-seq data with published transcriptomes of sorted spermatogenic cell types58. Genes with increased RNA and ribosome-protected fragment (RPF) abundance were preferentially expressed in spermatogonia, whereas the downregulated genes were preferentially expressed in spermatocytes and spermatids (Fig. 6a). Correspondingly, we found that spermatogonia-specific genes were upregulated with increased RNA and RPF abundance in PRRC2A-null spermatocytes (Fig. 6b, c), including genes involved in spermatogonial stem cell maintenance (Id4, Gfra1, Ret, Nanos3, Plzf, Sall4, Foxo1, Etv5) and spermatogonial differentiation (Stra8, Kit, Dmrt1, Sox3) (Fig. 6e). However, spermatocyte-specific genes (such as Ccdc36 and Aurkc) and spermatid-specific genes (including several genes essential for spermiogenesis, such as Prm$\frac{1}{2}$/3 and Tnp$\frac{1}{2}$) were downregulated in PRRC2A-null spermatocytes (Fig. 6b, c, e). *The* gene set enrichment analysis (GSEA) of RNA-seq and Ribo-seq data using cell-type-specific gene sets (Supplementary Data 4) yielded similar results (Fig. 6d). Interestingly, we noticed genes required for meiotic recombination (such as Spo11 and Hormad1) were not affected by PRRC2A deficiency (Fig. 6e) which explained why PRRC2A-null spermatocytes were able to progress through meiotic prophase, although with delayed progress in juvenile testes and increased apoptosis starting from pachytene stage in adult testes. These results suggested PRRC2A is required for the transition of expression profile from spermatogonia to spermatocyte and the upregulation of genes required for spermiogenesis. PRRC2A deficiency caused partial loss of spermatocyte identity, and consequently leads to multiple meiotic defects and failed spermiogenesis. Fig. 6PRRC2A deficiency causes defective transcriptome transition from spermatogonia to spermatocytes.a Heatmap showing the changes of RNA abundance, TE, and RPF for genes with differentially RNA abundance (p-adjusted < 0.05 as cutoff) in PRRC2A-null spermatocytes versus control spermatocytes. Genes with PRRC2A-binding m6A-modified sites and their expression levels in different cell types were correspondingly shown. b, c Box plots show the RNA and RPF abundance of cell-type-specific genes from RNA-seq and Ribo-seq data (Prrc2a-cko versus control, two-sided Wilcox test) ($$n = 186$$, 395, 267, 125 genes). The box indicates the lower ($25\%$) and upper ($75\%$) quantile and the white line indicates the median. Whiskers extend from $2.5\%$ to $97.5\%$ percentile, non-outlier data points. d GSEA analysis of cell-type-specific gene sets in RNA-seq and Ribo-seq data (Prrc2a-cko versus control). e A heatmap showing the fold change of RNA and RPF abundance for representative cell-type-specific genes from RNA-seq and Ribo-seq data (Prrc2a-cko versus control). f Integrative Genomics Viewer shows the distribution of m6A-modified peaks and PRRC2A-binding peaks along with indicated transcripts in PRRC2A RIP-seq and MeRIP-seq11 data. Blue peaks and red peaks represent reads in the input and IP groups, respectively. Red boxes show the area containing both m6A-modified peaks and PRRC2A-binding peaks. g Relative abundance of indicated mRNAs in control and PRRC2A-null spermatocytes detected by qPCR. Two-sided student’s t-test. Error bars, $$n = 4$$ samples, mean ± SEM. ns $$p \leq 0.2311$$, ***$p \leq 0.0001.$ *Source data* are provided as a Source Data file. Further, the majority of PRRC2A-binding methylated targets were upregulated and overlapped more with spermatogonia-specific genes than other cell-type-specific genes (Fig. 5h, 6a, Supplementary Fig. 6e). We obtained 59 upregulated target transcripts from spermatogonia-specific gene sets, among them, Plzf, Sall4, Foxo1, and Sox3 were four representative genes known to regulate spermatogonial stem cell maintenance or differentiation (Fig. 6e, Supplementary Fig. 6e). Their transcripts contained overlapped PRRC2A-binding peaks and m6A modification sites (Fig. 6f), and their increased RNA abundance was verified in PRRC2A-null spermatocytes by qPCR (Fig. 6g). Thus, PRRC2A recognizes spermatogonia-specific transcripts and downregulates their expression during meiotic prophase to promote the transition from mitosis to meiosis. Moreover, although alterations of RPF abundance in PRRC2A-null spermatocytes showed a pattern similar to that of RNA abundance, the TE profile had an opposite alteration pattern (Fig. 6a), possibly suggesting a type of translational balance mechanism in spermatocytes to compensate for the disruption of the transcriptional landscape (but failed). ## PRRC2A promotes translation of genes involved in meiotic cell division To investigate the potential regulatory impacts of PRRC2A during meiotic metaphase, we examined the expression of genes involved in meiotic cell division by GSEA (using the GO database) in RNA-seq and Ribo-seq data. We found that gene sets annotated with “meiotic cell cycle”, “meiotic nuclear division”, and “spindle” showed increased RNA abundance but decreased RPF in PRRC2A-null spermatocytes (Fig. 7a), which indicated that PRRC2A deficiency led to reduced translation efficiency of these genes. Previous studies reported that Cep19250, and Wnk159 regulate cell division, and their deficiency leads to impaired spindle assembly and chromosome alignment during meiosis or mitosis, phenotypes similar to the defects we observed in PRRC2A-null metaphase spermatocytes (Fig. 3f). RNA immunoprecipitation data showed that their transcripts bore overlapped PRRC2A-binding peaks and m6A modification sites (Fig. 7b), and we also verified that using PRRC2A RIP-qPCR and MeRIP-qPCR (Fig. 7c, d). We found that protein levels of CEP192 and WNK1 were reduced in PRRC2A-null spermatocytes (Fig. 7e, f), but their RNA abundance was not affected or slightly increased after PRRC2A deficiency (Fig. 7g), which indicates reduced translational efficiency of these transcripts in PRRC2A-null spermatocytes. Additionally, transcripts of Dazl contained no m6A modification or PRRC2A-binding peak (Fig. 7b–d) and were used as the control. Its mRNA level and protein level were not changed in PRRC2A-null spermatocytes (Fig. 7e–g). Therefore, PRRC2A recognizes transcripts of genes involved in meiotic cell division and promotes the translation, thereby facilitating the progression of male meiotic metaphase. Fig. 7PRRC2A promotes the expression of genes involved in meiotic cell division.a GSEA analysis of indicated gene sets in RNA-seq and Ribo-seq data (Prrc2a-cko versus control). b Integrative Genomics Viewer showed the distribution of m6A-modified peaks and PRRC2A-binding peaks along with indicated transcripts in PRRC2A RIP-seq and m6A-seq11 data. Blue peaks and red peaks represent reads in the input and IP groups, respectively. Red boxes show the area containing both m6A-modified peaks and PRRC2A-binding peaks. c, d MeRIP-qPCR and PRRC2A RIP-qPCR analysis of indicated transcripts in P20 testes. Two-sided student’s t-test. Error bars, $$n = 3$$ biological repeats, mean ± SEM. ns $p \leq 0.05$, **$p \leq 0.01.$ *Source data* are provided as a Source Data file. e WB test of indicated protein in control and PRRC2A-null spermatocytes. f, g Quantification of the relative protein (f) and RNA (g) level of indicated genes in control and PRRC2A-null spermatocytes. Two-sided student’s t-test. Error bars, $$n = 3$$ biological repeats, mean ± SEM. ns $p \leq 0.05$, *$p \leq 0.05$, ***$p \leq 0.001.$ *Source data* are provided as a Source Data file. h Testis lysates were subjected to IP with anti-Flag or IgG control antibodies. IP groups were treated with RNase inhibitor (RNasin) or RNaseA respectively. Indicated proteins were detected by WB. i WB test of indicated protein in control and PRRC2A-null spermatocytes. j Testis lysates were subjected to IP with anti-Flag or IgG control antibodies. Indicated proteins were detected by WB. CB chromatoid body, SG stress granule, PB processing body. To explore the mechanism of PRRC2A-mediated regulation of mRNA metabolism and translation, we further performed co-IP coupled with mass spectrometry (IP-MS) for PRRC2A and identified a series of interacting proteins (Supplementary Fig. 7a, b). GO analysis of the candidate interacting proteins showed enrichment for functional annotations related to mRNA metabolism and cell cycle (Supplementary Fig. 7c). Among them, Y-Box Binding Protein 1 and 2 (YBX1 and YBX2), and fragile-X mental retardation autosomal 1 (FXR1) were verified to bind PRRC2A by co-IP (Fig. 7h, Supplementary Fig. 7b), and previous studies showed that they could promote mRNA degradation60–62. We also found that PRRC2A interacted with poly(A) tail binding protein C1 (PABPC1) and eukaryotic translation initiation factor 4 gamma 3 (EIF4G3) (Fig. 7h, Supplementary Fig. 7b) which were reported to facilitate translation initiation during spermatogenesis63,64. A recent study showed that FXR1 also promotes translation together with EIF4G3 and PABPC1 to drive spermiogenesis65. EIF4G3 has also been shown to be required for male fertility, and its deficiency causes failed entry to the meiotic metaphase during spermatogenesis, defects also observed in PRRC2A-null spermatocytes64. We found the protein abundance of EIF4G3 was decreased in PRRC2A-null spermatocytes, while the abundance of its reported downstream, HSPA264, was normal (Fig. 7i). Other cofactors of PRRC2A were not affected by PRRC2A deficiency (Fig. 7i). Collectively, PRRC2A potentially recruits different transcripts and corresponding cofactors to promote mRNA decay and/or translation of its targets during meiotic prophase. Additionally, we found the interaction between PRRC2A with YBX1 and YBX2 can be disrupted by RNase treatment, while the interaction of PRRC2A with other cofactors resisted this treatment. So, we believed that PRRC2A indirectly binds YBX1 and YBX2 and this interaction is dependent on the existence of RNA (Fig. 7h). Notably, both YBX2 and PABPC1 are components of the chromatoid body66, and we also successfully verified the interactions of PRRC2A with two marker proteins of the chromatoid body, DDX4 and MIWI (Fig. 7j). Given that the chromatoid body was reported to harbor components of nonsense-mediated mRNA decay (NMD) pathway66 and supports NMD of long 3' UTR mRNAs67, we speculated that PRRC2A potentially recruits transcripts into the chromatoid body and mediates their decay. Further, several previous studies have reported that cytosolic m6A readers co-localize with stress granules and/or with processing bodies6,7,17,21,68,69. In testes, we found no evidence that PRRC2A interacts with a stress granule marker (G3BP1), or with markers of processing bodies (PATL1, DCP1A) (Fig. 7j). However, PABPC1, EIF4G3, and FXR1 are key components of the stress granule70, suggesting PRRC2A may partly interact with stress granules. Previous reports showed that stress granules can interact with P bodies or undergo autophagy to mediate RNA degradation in stress conditions such as heat stress, oxidative stress, or viral infections71. Therefore, PRRC2A may promote mRNA degradation with stress granules in male germ cells under stress. ## Discussion Past studies reported that N6-Methyladenosine (m6A) exists in male germ cells29. m6A writers such as METTL3 and METTL1429,30, m6A erasers like ALKBH5 and FTO5,31 are required for spermatogenesis, indicating an important role of m6A in spermatogenesis. As the functional mediator of m6A, multiple m6A readers have been functionally linked to the development of spermatogonia (YTHDF237 and YTHDC136), meiotic initiation, pachytene progression (YTHDC27,8,11,32,33), and spermiogenesis (YTHDF238). But recent studies showed that YTHDC2 regulates spermatogenesis independent of m6A recognition34,35. These previous studies have collectively showcased the diverse regulatory influences of m6A in spermatogenesis. In the present study, we demonstrate that m6A reader PRRC2A is essential for the completion of meiosis I. Detailly, we generated a transgenic mouse line with conditional knockout of Prrc2a in male germ cells and found that PRRC2A deficiency leads to male sterility. PRRC2A-null spermatocytes exhibit delayed developmental progress in juvenile testes and show defective XY synapsis and MSCI at late prophase in adult testes. Combining the results of histological analysis and immunostaining, we found that PRRC2A deficiency leads to delayed metaphase entry. PRRC2A-null spermatocytes exhibit chromosome misalignment and spindle disorganization, and consequently, are probably delayed by the SAC and undergo apoptosis. Some escape from SAC may produce round spermatids with abnormal nuclear morphology or multinucleated cells. These defective round spermatids cannot develop to the advanced stage and undergo apoptosis soon. Notably, the first wave of spermatogenesis is not the same as adult spermatogenesis in the Prrc2a-cko testes. Although the delayed entry and defective completion of metaphase occur during both the first wave of spermatogenesis (Figs. 1i, and 3b, c) and in adult spermatogenesis (Figs. 1h and 3d), the delayed progression from zygonema to pachynema only occur during the first wave of spermatogenesis (Supplementary Fig. 4d). The progression during the meiotic prophase of adult spermatogenesis was normal in Prrc2a-cko testes (Fig. 2a, Supplementary Fig. 4b). In mammalian cells, cytosolic m6A readers generally affect mRNA metabolism and translation efficiency. YTHDF26 and YTHDF310 mediate RNA decay, while IGF2BPs17, FMRP19, SND120, and PRRC2A21 stabilize RNA targets. YTHDF113, YTHDF39,10, IGF2BPs17, eIF318 were reported to modulate mRNA translation. Here, we showed that PRRC2A is a cytosolic m6A reader in male germ cells that mediated the decay of its mRNA targets or improves the translation efficiency, which is different from the previous study21. During meiotic prophase, PRRC2A binds and downregulates spermatogonia-specific transcripts to promote the transcriptome transition from spermatogonia to spermatocytes. Noted that the development of differentiated spermatogonia and early prophase spermatocytes is not affected by PRRC2A deficiency, this may be because PRRC2A does not play a central role at these stages. The expression of PRRC2A is low during these stages (Fig. 1b, c, S1) and some other proteins could also promote the transition from mitosis to meiosis, such as YTHDC27,8,11,32,34,35. Further, Ccna2, the typical gene whose overexpression results in impaired meiosis initiation54, is expressed normally in PRRC2A-null spermatocytes (Fig. 4c, d). So, the development of differentiated spermatogonia and early prophase spermatocytes may be less affected and could still develop well, although PRRC2A deficiency causes the abnormal upregulation of spermatogonia-specific genes. During the late prophase, PRRC2A is highly expressed (Fig. 1b, c, S1) and exerts critical function, so multi-aspect defects occur in PRRC2A-null spermatocytes. PRRC2A also enhances the translation of genes involved in meiotic cell division to facilitate the progression of meiotic metaphase. We found that PRRC2A interacts with multiple cofactors which are reported to degrade mRNA and/or promote translation initiation, including YBX1, YBX2, FXR1, PABPC1, and EIF4G3. Noted that mouse PRRC2A is a large protein with 2157 aa (isoform 2, NCBI ID: NP_001185973.1) or 2158 aa (isoform 1, NCBI ID: NP_064411.2). However, besides the GRE domain (domain enriched with glycine, arginine, and glutamic acid; <648 aa) used to bind m6A modification21, the function of its BAT-N domain and other large regions is basically unknown. Additionally, PRRC2A was localized in RNA granules (chromatoid body) (Fig. 1c), this indicated the possible phase separation functions of PRRC2A. Thus, we speculate that PRRC2A potentially acts as a scaffold to recruit different transcript targets and corresponding cofactors in different developmental stages and locations, so that performs diverse regulations to its targets. Additionally, considering that many PRRC2A-binding transcripts do not contain m6A modification (2127 of 4081 PRRC2A-binding targets) (Fig. 5g), the large unknown regions of PRRC2A may also exert functions independent of m6A recognition to regulate mRNA targets directly. Interestingly, many cytosolic m6A readers were reported to localize in RNA granules. In several human cell lines, YTHDF1, YTHDF2, and YTHDF3 are localized in stress granules under stress conditions, and YTHDF2 also exists with processing bodies6,68,72. IGF2BPs protect mRNAs from degradation in the processing body or store mRNAs in stress granules17. In spermatocytes, YTHDC2 was revealed to exist in RNA germ granules7. Further, PRRC2A is also co-localized with YTHDF2 in the RNA granule in HT-22 cells21. Here, we showed that PRRC2A is localized in the chromatoid body, a type of RNA granule specific for male germ cells. Chromatoid bodies contained several components of the NMD pathway66 and are thought to mediate mRNA decay67. We speculate that PRRC2A may recruit target mRNAs to chromatoid bodies and promote the degradation. Additionally, PABPC1, FXR1, and EIF4G3 are also components of the stress granule and are revealed to interact with PRRC2A, results suggesting PRRC2A may partly interact with stress granules. This indicates that PRRC2A probably regulates target transcripts with other types of RNA granules under different conditions. Here, our present study demonstrates that PRRC2A is required for the completion of male meiosis I by improving the degradation or translational efficiency of its mRNA targets. This research provides an important reference and theoretical basis for follow-up research and clinical practice in other processes in addition to spermatogenesis. Notably, although PRRC2A is expressed at low levels in other organs of adult mice, such as spleen and lung, and brain, PRRC2A may still play a role in special developmental stages or specific cells in these tissues, such as Pdgfrα- or NG2-positive cells in the embryonic brain21. Moreover, epidemiological studies reported that PRRC2A is associated with cancer73,74, neurological diseases75, autoimmune diseases76, diabetes77, and obesity78. Altogether, PRRC2A is widely involved in the regulation of various diseases and may be a potential clinical treatment target. More studies are needed to reveal the role of PRRC2A in other physiological and pathological processes. ## Mice All animal experiments were approved by the Chinese Ministry of Health national guidelines and performed following institutional regulations of Institutional Animal Care and Use Committee at the National Institute of Biological Sciences, Beijing. All mice in this study were C57BL6 strains. Mice were maintained under specific pathogen-free conditions of a 12 h light/dark cycle at controlled temperature (20-25 °C) and humidity (50-$70\%$) and were provided with food and water ad libitum in the Animal Care Facility at National Institute of Biological Sciences, Beijing. Transgenic mice were generated using CRISPR/Cas9 technology. The sgRNAs were prepared using MEGAshortscript T7 Transcription kit (Ambion) according to the manufacturer’s instructions. For Prrc2af/f mice, DNA fragments containing exons 2–5 of the Prrc2a gene flanked by two loxP sites and two homology arms were used as donor templates. For Prrc2a-flag mice, DNA fragments containing the last coding exon (exon30) of the Prrc2a gene followed by a 3×flag tag and two homology arms were used as donor templates. After the co-incubation of Cas9 protein (NEB) and sgRNA, the Cas9-sgRNA complex and donor templates were injected into C57BL/6 zygotes. Injected zygotes were transferred into pseudo-pregnant CD1 female mice. The resulting founder mice were genotyped and mated with C57BL6 mice. For germ cell-specific knockout of Prrc2a, Stra8-Cre mice (Jackson Laboratory) were used. Briefly, Prrc2af/f male mice were crossed with Stra8-Cre; Prrc2af/f or Stra8-Cre; Prrc2af/Δ female mice, and descendants with Stra8-Cre were Prrc2a-cko mice. Only male mice were used in the current study. Sequence of gRNA and primers used for genotyping were listed in Supplementary Table 1. ## Histological analysis Testes and epididymis were dissected and fixed in Davidson’s Fluid (Formaldehyde: Ethanol: Glacial acetic acid: H2O = 6:3:1:10) overnight at 4 °C. Samples were dehydrated through an ethanol series ($70\%$, $80\%$, $90\%$, $100\%$ ethanol in H2O) and embedded in paraffin. The 5 µm sections were cut using a microtome (Leica RM2245) and mounted on adhesion microscope slides (CITOTEST). After deparaffinization and hydration, sections were stained with hematoxylin and eosin (H&E) following standard protocols. Images were acquired with the Olympus VS120 microscope. ## Chromosome spreads Adult testes were dissected, and tunica albuginea was removed. 1.5–2 cm seminiferous tubules were incubated in 400 μL hypotonic extraction buffer (30 mM Tris-HCl pH 8.5, 50 mM sucrose, 17 mM citric acid, 5 mM EDTA) for 60 min at room temperature and minced in 100 mM sucrose/H2O. 20 μL cell suspensions were dropped onto adhesion microscope slides (CITOTEST) spread with 20 μL fixation buffer ($1\%$ PFA and $0.15\%$ Triton X-100) from a high place, fixed for 3 h at room temperature, and air-dried for 1 h in room temperature. Slides were washed three times in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and 1.8 mM KH2PO4) before immunohistochemical experiments. ## Immunostaining and TUNEL staining Testes were dissected and fixed in $4\%$ paraformaldehyde overnight at 4 °C. For paraffin sections, samples were treated as in histological analysis, and 5 µm sections were used for immunostaining. For frozen sections, samples were dehydrated in $30\%$ sucrose/PBS and embedded in O.C.T compound (Tissue-Tek). 10 µm slices were cut using a cryostat (Leica CM1950) and dried overnight at 42 °C. After washing three times with PBS, paraffin or frozen sections were subjected to antigen retrieval with sodium citrate buffer (10 mM sodium citrate, $0.05\%$ Tween-20, pH 6.0) or Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA, $0.05\%$ Tween-20, pH 9.0). Sections were blocked in ADB ($1\%$ normal donkey serum, $0.3\%$ BSA, $0.05\%$ Triton X-100) for 1 h at room temperature and incubated with primary antibodies in ADB overnight at 4 °C. After washing three times with PBST (PBS with $0.1\%$ Tween-20), slides were incubated with secondary antibodies in ADB for 1 h at room temperature and washed another three times followed by staining with 1 μg/mL 4',6-Diamidino-2-phenylindole (DAPI) (Invitrogen, D3571). For staining of chromosome spreads, slides were blocked with ADB and incubated with primary antibodies, secondary antibodies, and DAPI (Invitrogen, D3571) as above. For detecting the acrosome, 1 μg/mL PNA (Sigma, L7381) was used to stain sections for 20 min at room temperature. For TUNEL staining, assays were performed using In Situ Cell Death Detection Kit (Roche, 11684795910) following the manufacturer’s instructions. Images were acquired using the confocal microscope Zeiss LSM800 (Zen 2.3 (blue edition) software) or Nikon A1-R (NIS-Elements AR 5.21.00 software) and microscope Olympus VS120 (OLYMUPUS VS-ASW 2.9 software). ## Western blot Testes were lysed in RIPA buffer (Sigma, R0278) with 1: 10 protease inhibitor cocktail (Roche, 04693116001). The lysate was centrifuged at 13,000 rpm 4 °C for 20 min. The protein concentration of supernatants was measured with Quick StartTM Bradford 1 x Dye Reagent (BIO-RAD, 500-0205). For isolated germ cells, they were lysed with 1x loading buffer ($2\%$ SDS, $10\%$ Glycerol, 50 mM Tris-HCL, $1\%$ β-mercaptoethanol, and $0.05\%$ bromophenol blue dye, PH6.8) directly. The equal quality of proteins of each sample was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene difluoride (PVDF) membranes (Millipore, IPVH00010). Membranes were blocked with $5\%$ skim milk in TBST (20 mM Tris, 150 mM NaCl, pH 7.6 with $0.05\%$ Tween-20) and incubated with the primary antibody in $5\%$ skim milk/TBST. After three times of wash with TBST, membranes were incubated with HRP-conjugated secondary antibody in $5\%$ skim milk/TBST for 1 h at room temperature followed by another three times of washing with TBST. ECL reagents (BIO-RAD, 170-5060 or NCM Biotech, P10300B) were added onto membranes, and signals were detected by XBT X-ray film (Carestream, 6535876). All uncropped and unprocessed scans of western blots were supplied in Supplementary Fig. 8 in the Supplementary Information. ## Antibody Antibodies used in the immunostaining and WB were listed in Supplementary Table 2. ## RNA purification and quantitative real-time PCR Tissues or cells were homogenized in TRIZOL (Invitrogen, 15596026) and chloroform was added to extract twice. After centrifugation, the upper aqueous phase was transferred into isopropanol followed by another centrifugation to collect RNA pellets. Glycogen (Thermo Scientific, R0551) was used when precipitating a small amount of RNA. Pellet was washed two or three times using $75\%$ ethanol, air-dried, and solubilized in nuclease-free water. Reverse transcription was performed using PrimeScript® RT reagent kit with gDNA Eraser (TaKaRa, RR047A) according to the manufacturer’s instructions. Quantitative real-time PCR (qPCR) was performed using SYBR Green master mix (TaKaRa, DRR420A) and Bio-Rad CFX96 Real-Time System (Bio-Rad CFX Manager 3.1 software). Relative mRNA expression was measured using the Delta-Delta CT method, and Rpl6 was used for normalization. Primers used were listed in Supplementary Table 1. ## Immunoprecipitation and Mass spectrometry For immunoprecipitation, four P20 testis was homogenized in 1 mL ice-cold lysis buffer (150 mM NaCl, 10 mM HEPES pH 7.6, 2 mM EDTA, $0.5\%$ NP-40, 0.5 mM 1,4-dithiothreitol (DTT), 1:10 protease inhibitors cocktail (Roche, 04693116001), 100 U/mL Recombinant RNasin® Ribonuclease Inhibitor (Promega, N2515)) and incubated on ice for 20 min to lyse the tissue. For RNase treatment, RNasin and DTT were replaced by 10 mg/mL RNase A (Tiangen). The lysate was centrifuged at 13,000 ×g for 20 min at 4 °C and supernatants were collected. 100 μL supernatants were saved as input and mixed with 25 μL 5 x loading buffer ($10\%$ SDS, $50\%$ Glycerol, 250 mM Tris-HCL, $5\%$ β-mercaptoethanol and $0.1\%$ bromophenol blue dye, pH 6.8). 30 μL Protein G magnetic beads (Invitrogen, 10004D) conjugated with 1.5 μg mouse antibody to Flag (Sigma, F1804) or mouse IgG isotype control (CST, 5415 S) were added into 300 μL supernatants and incubated overnight at 4 °C. Supernatants were discarded and beads were washed 8 times with 1 mL ice-cold NT2 buffer (200 mM NaCl, 50 mM HEPES pH 7.6, 2 mM EDTA, $0.05\%$ NP-40) at 4 °C. For RNasin treatment, 0.5 mM DTT and 40 U/mL RNase inhibitor were added. For the RNase A treatment group, after 3 times washes, beads were treated with NT2 buffer supplemented with 10 mg/mL RNase A for 20 min at room temperature and washed with NT2 buffer another 5 times. Then, beads were boiled in 1x loading buffer ($2\%$ SDS, $10\%$ Glycerol, 50 mM Tris-HCL, $1\%$ β-mercaptoethanol, and $0.05\%$ bromophenol blue dye, pH6.8) for 10 min at 95 °C. After centrifugation, supernatants were subjected to Western blot detection. For the identification of interaction proteins by MS, IP was performed as above with some modifications. In brief, eight P20 Prrc2a-flag and wild-type testes were used and supernatants of lysate were incubated with 60 μL Protein G magnetic beads (Invitrogen, 10004D) conjugated 3 μg mouse antibody to Flag. After 8 times washes, beads were eluted using 0.2 mg/mL Flag peptide (Sigma, F4799) in NT2 buffer for 1 h at 4 °C. Samples were separated on SDS-PAGE followed by silver staining (Sigma, PROTSIL1). The stained proteins were destained and in-gel digested with trypsin (10 ng mL-1 trypsin, 50 mM ammonium bicarbonate, pH 8.0) overnight at 37 °C. Peptides were extracted with $5\%$ formic acid/$50\%$ acetonitrile and $0.1\%$ formic acid/$75\%$ acetonitrile sequentially. The extracted peptides were separated by an analytical capillary column (50 μm × 10 cm) and sprayed into an LTQ ORBITRAP Velos mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) equipped with a nano-ESI ion source. Identified peptides were searched in the IPI (International Protein Index) Mouse protein database on the Mascot server (Matrix Science Ltd, UK). ## Isolation of late meiotic spermatocytes The procedure was modified from the previous study55. Testes were dissected and tunica albuginea was removed in PBS. Seminiferous tubules of one control testis or two Prrc2a-cko testes were dispersed with tweezers and incubated in 10 mL DMEM medium containing 1 mg/mL Collagenase IV (YEASEN, C3125030) and 0.1 mg/mL DNase I (YEASEN, D2122070) for 25 min at 34 °C with rotation. Tubule fragments were collected by gravity settlement and washed twice with 10 mL PBS. Then, tubules were digested in 10 mL $0.05\%$ Trypsin/EDTA (Gibco, 25300062) containing 0.1 mg/mL DNase I for 8 min at 34 °C and were gently pipetted up and down to disperse germ cells. 1 mL FBS was added and the cell suspension was passed through nylon mesh with 40 μm pore size (FALCON, 352340). After centrifuging of 300 × g for 5 min, germ cells were washed by 1 mL DMEM containing $10\%$ FBS and resuspended by 4 mL DMEM containing $10\%$ FBS. Then, 4 μL 10 mg/mL Hoechst 33342 (Sigma, B2261) was added, and cells were stained for 60 min at 34 °C with rotation. Before sorting, 8 μL 1 mg/mL propidium iodide (Invitrogen, P3566) was added, and the cell suspension was filtered by 40-μm nylon mesh another time. Cell suspensions were sorted by BD FACSAria Fusion-II with the 70 µm nozzle using BD FACSDiva software (version 8.0.3). 355 nm laser was used to excite Hoechst 33342 and fluorescence was recorded with a $\frac{450}{40}$ nm band-pass filter (Hoechst blue) and a 635 nm long filter (Hoechst red). The gating strategy referred to the previous study55. Spermatocytes were collected in DMEM containing $10\%$ FBS for subsequent experiments. Flowjo software (vX.0.7) was used for data analysis. ## RNA-seq Purified mRNAs from isolated spermatocytes of adult control and Prrc2a-cko testes and were used to construct libraries using NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB, E7645L) in the sequencing center at National Institute of Biological Sciences, Beijing. The libraries were sequenced on the Illumina HiSeq 2500 platform using the single-end 75 bp sequencing strategy. Raw sequencing reads were trimmed to remove low-quality bases by Trim Galore (version 0.6.4)79 in single-end mode. Then the trimmed reads were aligned to the mouse reference genome assembly build GRCm38_68 (mm10) using STAR (version 2.7.3a)80 with default parameters. The mapped reads were annotated to ensemble gene exons (Mus_musculus.GRCm38.99.gtf) and counted for each gene by featureCounts (version 2.0.0)81, the expression level of each gene was quantified as reads per kilobase of transcript per million reads mapped (RPKM). Differentially expressed genes (DEGs) between Prcc2a knockout and control were identified based on read counts by using DEseq2 (version 1.30.1)82 with the following cutoffs: fold change > 1.5, adjusted p-value < 0.05, mean RPKM > 1, each RPKM > 0. Heatmap plots were plotted on http://www.bioinformatics.com.cn, a free online platform for data analysis and visualization. ## Ribo-seq Cells were treated with 100 µg/mL cycloheximide (CST, 2112) to block translational elongation during the whole process of spermatocyte isolation using FACS. Isolated spermatocytes were then frozen with liquid nitrogen and stored at −80 °C for subsequent experiments. For ribosome footprints (RFs) recovery, samples were dissolved in 400 µL of lysis buffer and the ribosomal profiling technique was carried out as reported previously83, with a few modifications as described below. The extracts were incubated on ice for 10 min and were triturated ten times through a 26-G needle. The lysate was centrifuged at 20,000 × g for 10 min at 4 °C, and the supernatant was collected. Then, 10 µL of RNase I (NEB) and 6 µL of DNase I (NEB) were added to 400 µL of lysate, which was then incubated for 45 min at room temperature. Nuclease digestion was stopped by adding 10 µL of SUPERase·In RNase inhibitor (Ambion). Next, 100 μL of digested RFs were added to the equilibrated size exclusion columns (illustra MicroSpin S-400 HR Columns; GE Healthcare; catalog no. 27-5140-01) and centrifuged at 600 g for 2 min. Then, 10 μL $10\%$ (wt/vol) SDS was added to the elution, and RFs with a size >17 nt were isolated according to the RNA Clean and Concentrator-25 kit (Zymo Research; R1017). rRNA was removed using the method reported previously84. Briefly, short (50–80 bases) antisense DNA probes complementary to rRNA sequences were added to the solution containing RFs, then RNase H (NEB) and DNase I (NEB) were added to digest rRNA and residual DNA probes. Finally, RFs were further purified using magnet beads (Vazyme). Ribo-seq libraries were constructed using NEBNext® Multiple Small RNA Library Prep Set for Illumina® (catalog no. E7300S, E7300L) following the manufacturer’s instructions and were sequenced using Illumina HiSeqTM X10. We followed the pre-processing procedure of the ribosome profiling data as described previously85. Briefly, the Trim Galore (version 0.6.4)79 was used to trim the 3' adapter in the raw reads. Low-quality reads with Phred quality score >20 were removed. Next, the trimmed reads were aligned to abundant sequences (including rRNA, tRNA, and mtRNA) by using Bowtie2 (version 2.3.5.1)86 with no mismatch allowed, and mapped reads were discarded. Then the unmapped reads were aligned to the mouse reference genome assembly build GRCm38_68 (mm10) using STAR (version 2.7.3a)80 as did in mRNA processing, and two mismatches were allowed in this step. To reduce the technical noise of ribosome profiling, in the quantification of ribosome-protected mRNA fragments (RPFs), the following filters were processed. Firstly, the Ribo-seq reads with length between 26 and 34 nt were selected. Besides, the multiple aligned reads were discarded, and only the uniquely mapped reads to the coding regions were retained. Third, reads aligned to the first 15 and last 5 codons were excluded. After filtrations, the quantification of RPFs was processed by featureCounts (version 2.0.0)81. The counts matrix of RPFs were combined with expression matrix, and implemented into Xtail package (version 1.1.5)85 to calculate the differential translation efficiencies. The threshold for differential translational efficiency was fold change > 1.5, p-value < 0.05. Heatmap plots were plotted on http://www.bioinformatics.com.cn, a free online platform for data analysis and visualization. ## Functional enrichment analysis DTGs and DTEGs were performed for functional enrichment analysis by using the online tool Metascape (http://metascape.org)87. Gene Ontology biological processes (BP) pathways were selected as ontology sources. Terms with p-value < 0.01 were retained as significant enrichment. Gene set enrichment analysis for the cell-type-specific gene sets and gene sets annotated with GO terms “meiotic cell cycle”, “meiotic nuclear division”, and “spindle” was performed using GSEA software (version 4.1.0)88 with 1000 gene set permutations ## PRRC2A RIP-seq The procedure was modified from the previous study21. Thirty-two testes from P20 Prrc2a-flag mice were dissected and homogenized in 8 mL ice-cold lysis buffer (150 mM NaCl, 10 mM HEPES pH 7.6, 2 mM EDTA, $0.5\%$ NP-40, 0.5 mM 1,4-dithiothreitol (DTT), 1:10 protease inhibitors cocktail (Roche, 04693116001), 100 U/mL Recombinant RNasin® Ribonuclease Inhibitor (Promega, N2515)) and incubated on ice for 20 min to lyse the tissue. The lysate was centrifuged at 13,000 × g for 20 min at 4 °C supernatants were collected, then repeat centrifugation twice. Protein concentration was measured using Quick StartTM Bradford 1x Dye Reagent (BIO-RAD, 500-0205) and the sample was diluted to make the concentration not >5 mg/mL. 200 μL sample was used as input and RNA was purified using the TRIZOL method. The remaining sample was incubated with 6 μg mouse IgG isotype control (CST, 5415 S) for 1 h at 4 °C and 75 μL protein G magnetic beads (Invitrogen, 10004D) for 1 h at 4 °C to preclear. The beads were removed and 250 μL protein G magnetic beads conjugated with 20 μg mouse anti-Flag antibody (Sigma, F1804) were added into supernatants and incubated overnight at 4 °C with rotation. Supernatants were discarded and beads were washed 8 times with 1 mL ice-cold NT2 buffer (200 mM NaCl, 50 mM HEPES pH 7.6, 2 mM EDTA, $0.05\%$ NP-40). Next, beads were washed twice with 1 mL ice-cold DNase buffer (10 mM Tris-HCL, 2.5 mM MgCl2, 0.5 mM CaCl2, $0.05\%$ NP-40, pH 7.6) at 4 °C and incubated in 500 μL DNase buffer containing 100 U/mL DNase I (NEB, M0303S) for 10 min at 37 °C. After that, beads were washed twice with 1 mL ice-cold MNase buffer (50 mM Tris-HCL, 5 mM CaCl2, 100 μg/mL BSA, pH 7.9) and digested in MNase buffer containing 1 U/mL MNase (NEB, M0247S) for 10 min at 37 °C Then, supernatants were discarded and beads were washed twice with 1 mL 1×PNK + EGTA buffer (50 mM Tris-HCl pH 7.5, 20 mM EGTA, $0.5\%$ NP-40, pH 8.0) immediately at 4 °C and twice with 1 mL NT2 buffer 4 °C. Collected beads were digested with 4 mg/mL proteinase K (Roche, 03115828001) in 200 μL 1×PK buffer for 40 min at 55 °C. RNA was extracted from supernatants using the TRIZOL method. Input and RIP samples were subjected to rRNA removal using rRNA Depletion Kit (NEB, E6310S) according to the manufacturer’s instructions. The resulting RNA samples were examined by Agilent 2100 Bioanalyzer and used to generate the library using NEBNext® Ultra™ II DNA Library Prep Kit for Illumina® (NEB, E7645L). Sequencing was performed on HiSeq X Ten System with 150 bp paired-end-sequencing reactions. Biological replicates were performed and samples for each biological replicate were collected from different animals. ## PRRC2A RIP-seq and m6A MeRIP-seq analysis The raw sequencing reads were trimmed to remove the adapter sequences and low-quality bases by Trim Galore (version 0.6.4)79. Reads that were >35 bps were retained and aligned to the mouse genome (mm10) by using Bowtie2 (version 2.3.5.1)86. Then uniquely aligned reads with mapping quality score ≥ 20 were kept for downstream peak-calling. Peak-calling was processed by MACS2 (version 2.1.2)89 for each biological replicate by comparing with input sample with parameters: “–keep-dup all –nomodel -q 0.05”. The peaks derived from two replicates were merged and the overlapped peaks were retained as high-confidence binding regions of PRRC2A or m6A modification regions. The m6A MeRIP-seq data was referred to the published dataset (GEO: GSE102346)11. The overlapped peaks between PRRC2A RIP-seq and MeRIP-seq were obtained using BEDTools’ intersect (version 2.28.0)90. The coverage profiles of RIP-seq were counted and visualized by igvtools (version 2.9.4)91. Peak with counts >1 and counts ratio (IP versus Input) >1 were used for the next analysis. ## Motif Identification We selected all peaks ranked by q-value to investigate the motifs enriched in these regions by using HOMER (version 4.11.1)92. To obtain high-quality motifs, repeat sequences were masked, motif length was restricted to 5–10, and the finding region was limited to 100 nt relative from the peak center. Furthermore, background sequences for each peak were generated on mRNA sequences by BEDTools’ shuffleBed (version 2.28.0)90 command to remove the background signal of selected genes. ## PRRC2A RIP-qPCR RNA immunoprecipitation of PRRC2A was performed as procedures in PRRC2A RIP-seq with some modifications. Four P20 testes from P20 wild-type mice were used and homogenized. Tissue lysate was precleared with 2 μg mouse IgG isotype control (CST, 5415 S) and 25 μL protein G magnetic beads (Invitrogen, 10004D). The precleared lysate was incubated with 50 μL protein G magnetic beads conjugated mouse IgG isotype control (IgG group) or mouse antibody of Flag (Sigma, F1804)(Flag group). Then, beads were washed with NT2 buffer, treated with DNase I (NEB, M0303S) and MNase (NEB, M0247S), and digested with proteinase K (Roche, 03115828001). Supernatants were collected and RNA was extracted using the TRIZOL method. RNA samples were reverse transcribed using PrimeScript ® RT reagent kit with gDNA Eraser (TaKaRa, RR047A) and detected by qPCR using SYBR Green master mix (TaKaRa, DRR420A). Three biological replicates were performed and samples for each biological replicate were collected from different animals. The recovery rate of specific transcripts was calculated by comparing the relative mRNA abundance of target genes in the IP group with the input group, and the level of Gapdh was used to normalize IP and IgG groups. Primers used were listed in Supplementary Table 1. ## MeRIP-qPCR RNA samples were extracted from four P20 wild-type testes and purified using the MiRNeasy mini kit (Qiagen, 217004) with DNase set (Qiagen, 79254). Purified RNAs (50 μg) were broken into ~300 nt fragments using RNA Fragmentation Reagents (Invitrogen, AM8740) by 30 s incubation at 94 °C. Fragmented RNAs were collected and purified by ethanol precipitation. Next, the RNA sample was incubated with 40 μL protein A magnetic beads (Invitrogen, 10002D) conjugated with 2 μg Rabbit IgG isotype control (CST, 2729 S) or rabbit antibody to m6A (Synaptic Systems, 202003) in IPP buffer (150 mM NaCl, $0.1\%$ NP-40, and 10 mM Tris-HCl, pH 7.4, 1 mM 1,4-dithiothreitol (DTT), 40 U/mL Recombinant RNasin® Ribonuclease Inhibitor (Promega, N2515)) for 4 h at 4 °C. Then, beads were washed three times with IPP buffer and digested by Proteinase K (Roche, 03115828001). Bound RNA was extracted from supernatants using the TRIZOL method and reverse transcribed using PrimeScript ® RT reagent kit with gDNA Eraser (TaKaRa, RR047A). Relative mRNA abundance of target genes was detected by qPCR using SYBR Green master mix (TaKaRa, DRR420A). Three biological replicates were performed and samples for each biological replicate were collected from different animals. The recovery rate of specific transcripts was calculated by comparing the relative mRNA abundance of target genes in the IP group with the input group, and the level of Gapdh was used to normalize IP and IgG groups. Primers used were listed in Supplementary Table 1. ## Phenotype characterizing and quantification analysis For the staging of seminiferous tubules, due to the lack of late-stage spermatids in Prrc2a-cko testes, stages of seminiferous tubules were determined based on the expression pattern of γH2AX and the morphology and composition of germ cells (Supplementary Fig. 3d). In adult Prrc2a-cko testes, tubules with early pachytene spermatocytes (cells close to the base of the lumen and with nuclear dense punctate γH2AX signal) and round spermatids were determined as stage I–III; tubules with mid-pachytene spermatocytes (cells relatively far from the base of the lumen and with nuclear dense punctate γH2AX signal) and round spermatids (little or no) were determined as stage IV–VII; tubules with leptotene spermatocytes (cells located at the base of the lumen and with nuclear γH2AX signal), late-pachytene spermatocytes (cells far from the base of the lumen and had large nuclei with dense punctate γH2AX signal) and no spermatid were determined as stage VIII–X; tubules with zygotene spermatocytes (cells located at the base of the lumen and with nuclear dense but not completely concentrated γH2AX signal) and diplotene spermatocytes (cells far from the base of the lumen and had large nuclei with dense punctate γH2AX signal) were determined as stage XI–XII. Only round and characteristic tubules were considered for followed analysis. Germ cells far from the base of the lumen and with condensed chromosomes aligning or aligned on the equatorial plate were determined as metaphase spermatocytes. For phenotype characterization using histology and immunofluorescence experiments, three biological replicates were used and representative images were shown. For quantification of staining of TUNEL and pH3, at least 3 testes from different mice were used. For quantification of the immunofluorescence intensity of POL II in the XY region, intensity in the XY region and the whole cell was measured and exported as 8-bit intensity values. The intensity of the area except for the XY body was calculated and used to normalize the intensity in the XY region. For quantification of MLH1 and DMC1 foci in chromosome spreads of prophase spermatocytes, foci along chromosomes were counted. For quantification of the immunofluorescence intensity of γ-tubulin and CEP192, the intensity at the spindle pole was measured and exported as 8-bit intensity values. The width and length of the alignment and spindle and the intersection angle between the spindle polarity axis and equatorial plate were measured in P60 testes sections stained with α-tubulin and γ-tubulin. Fiji software93 was used to measure the fluorescence intensity, in images acquired with the confocal microscope. Scatter plots and bar plots were created and analyzed with GraphPad Prism 6 (GraphPad Software). All data are presented as the means ± SEM. Significance difference was tested with Two-sided student’ s t-test (ns $p \leq 0.05$; *$p \leq 0.05$; **$p \leq 0.01$; ***$p \leq 0.001$). ## Statistics and reproducibility For all histology, immunofluorescence, western blot, immunoprecipitation, and qPCR experiments, we performed at least three independent biological replicates. All uncropped and unprocessed scans of the western blots were supplied in Supplementary Fig. 8 in the Supplementary Information. ## Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. ## Supplementary information Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1–4 Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-37252-y. ## Source data Source Data ## Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available. ## References 1. Jiang X. **The role of m6A modification in the biological functions and diseases**. *Signal Transduct. Target. Ther.* (2021.0) **6** 74. DOI: 10.1038/s41392-020-00450-x 2. Bokar JA, Rath-Shambaugh ME, Ludwiczak R, Narayan P, Rottman F. **Characterization and partial purification of mRNA N6-adenosine methyltransferase from HeLa cell nuclei. Internal mRNA methylation requires a multisubunit complex**. *J. Biol. Chem.* (1994.0) **269** 17697-17704. DOI: 10.1016/S0021-9258(17)32497-3 3. Liu J. **A METTL3-METTL14 complex mediates mammalian nuclear RNA N6-adenosine methylation**. *Nat. Chem. Biol.* (2014.0) **10** 93-95. DOI: 10.1038/nchembio.1432 4. Jia G. **N6-methyladenosine in nuclear RNA is a major substrate of the obesity-associated FTO**. *Nat. Chem. Biol.* (2011.0) **7** 885-887. DOI: 10.1038/nchembio.687 5. Zheng G. **ALKBH5 is a mammalian RNA demethylase that impacts RNA metabolism and mouse fertility**. *Mol. Cell* (2013.0) **49** 18-29. DOI: 10.1016/j.molcel.2012.10.015 6. Wang X. **N6-methyladenosine-dependent regulation of messenger RNA stability**. *Nature* (2014.0) **505** 117-120. DOI: 10.1038/nature12730 7. 7.Bailey, A. S. et al. The conserved RNA helicase YTHDC2 regulates the transition from proliferation to differentiation in the germline. eLife10.7554/eLife.26116 (2017). 8. Hsu PJ. **Ythdc2 is an N(6)-methyladenosine binding protein that regulates mammalian spermatogenesis**. *Cell Res.* (2017.0) **27** 1115-1127. DOI: 10.1038/cr.2017.99 9. Li A. **Cytoplasmic m(6)A reader YTHDF3 promotes mRNA translation**. *Cell Res.* (2017.0) **27** 444-447. DOI: 10.1038/cr.2017.10 10. Shi H. **YTHDF3 facilitates translation and decay of N(6)-methyladenosine-modified RNA**. *Cell Res.* (2017.0) **27** 315-328. DOI: 10.1038/cr.2017.15 11. Wojtas MN. **Regulation of m(6)A transcripts by the 3'–>5' RNA helicase YTHDC2 is essential for a successful meiotic program in the mammalian germline**. *Mol. Cell* (2017.0) **68** 374-387.e312. DOI: 10.1016/j.molcel.2017.09.021 12. Xiao W. **Nuclear m(6)A reader YTHDC1 regulates mRNA splicing**. *Mol. Cell* (2016.0) **61** 507-519. DOI: 10.1016/j.molcel.2016.01.012 13. Wang X. **N(6)-methyladenosine modulates messenger RNA translation efficiency**. *Cell* (2015.0) **161** 1388-1399. DOI: 10.1016/j.cell.2015.05.014 14. Liu N. **N6-methyladenosine alters RNA structure to regulate binding of a low-complexity protein**. *Nucleic Acids Res.* (2017.0) **45** 6051-6063. DOI: 10.1093/nar/gkx141 15. Alarcon CR. **HNRNPA2B1 is a mediator of m(6)A-dependent nuclear RNA processing events**. *Cell* (2015.0) **162** 1299-1308. DOI: 10.1016/j.cell.2015.08.011 16. Liu N. **N(6)-methyladenosine-dependent RNA structural switches regulate RNA-protein interactions**. *Nature* (2015.0) **518** 560-564. DOI: 10.1038/nature14234 17. Huang H. **Recognition of RNA N(6)-methyladenosine by IGF2BP proteins enhances mRNA stability and translation**. *Nat. Cell Biol.* (2018.0) **20** 285-295. DOI: 10.1038/s41556-018-0045-z 18. Meyer KD. **5’ UTR m(6)A promotes cap-independent translation**. *Cell* (2015.0) **163** 999-1010. DOI: 10.1016/j.cell.2015.10.012 19. Zhang F. **Fragile X mental retardation protein modulates the stability of its m6A-marked messenger RNA targets**. *Hum. Mol. Genet* (2018.0) **27** 3936-3950. PMID: 30107516 20. 20.Baquero-Perez, B. et al. The Tudor SND1 protein is an m(6)A RNA reader essential for replication of Kaposi’s sarcoma-associated herpesvirus. eLife10.7554/eLife.47261 (2019). 21. Wu R. **A novel m(6)A reader Prrc2a controls oligodendroglial specification and myelination**. *Cell Res.* (2019.0) **29** 23-41. DOI: 10.1038/s41422-018-0113-8 22. Shi H, Wei J, He C. **Where, when, and how: context-dependent functions of RNA methylation writers, readers, and erasers**. *Mol. Cell* (2019.0) **74** 640-650. DOI: 10.1016/j.molcel.2019.04.025 23. Handel MA, Schimenti JC. **Genetics of mammalian meiosis: regulation, dynamics and impact on fertility**. *Nat. Rev. Genet* (2010.0) **11** 124-136. DOI: 10.1038/nrg2723 24. Baudat F, Imai Y, de Massy B. **Meiotic recombination in mammals: localization and regulation**. *Nat. Rev. Genet.* (2013.0) **14** 794-806. DOI: 10.1038/nrg3573 25. Wellard SR. **Overlapping roles for PLK1 and Aurora A during meiotic centrosome biogenesis in mouse spermatocytes**. *EMBO Rep.* (2021.0) **22** e51023. DOI: 10.15252/embr.202154106 26. Alfaro E. **PLK1 regulates centrosome migration and spindle dynamics in male mouse meiosis**. *EMBO Rep.* (2021.0) **22** e51030. DOI: 10.15252/embr.202051030 27. 27.Maiato, H., Gomes, A. M., Sousa, F. & Barisic, M. Mechanisms of chromosome congression during mitosis. Biology10.3390/biology6010013 (2017). 28. Marston AL, Wassmann K. **Multiple duties for spindle assembly checkpoint kinases in meiosis**. *Front. Cell Dev. Biol.* (2017.0) **5** 109. DOI: 10.3389/fcell.2017.00109 29. Lin Z. **Mettl3-/Mettl14-mediated mRNA N(6)-methyladenosine modulates murine spermatogenesis**. *Cell Res.* (2017.0) **27** 1216-1230. DOI: 10.1038/cr.2017.117 30. Xu K. **Mettl3-mediated m(6)A regulates spermatogonial differentiation and meiosis initiation**. *Cell Res.* (2017.0) **27** 1100-1114. DOI: 10.1038/cr.2017.100 31. Huang T. **FTO knockout causes chromosome instability and G2/M arrest in mouse GC-1 cells**. *Front. Genet.* (2018.0) **9** 732. DOI: 10.3389/fgene.2018.00732 32. 32.Jain, D. et al. ketu mutant mice uncover an essential meiotic function for the ancient RNA helicase YTHDC2. eLife10.7554/eLife.30919 (2018). 33. Liu R. **YTHDC2 is essential for pachytene progression and prevents aberrant microtubule-driven telomere clustering in male meiosis**. *Cell Rep.* (2021.0) **37** 110110. DOI: 10.1016/j.celrep.2021.110110 34. Saito Y. **YTHDC2 control of gametogenesis requires helicase activity but not m(6)A binding**. *Genes Dev.* (2022.0) **36** 180-194. DOI: 10.1101/gad.349190.121 35. Li L. **The XRN1-regulated RNA helicase activity of YTHDC2 ensures mouse fertility independently of m(6)A recognition**. *Mol. Cell* (2022.0) **82** 1678-1690.e1612. DOI: 10.1016/j.molcel.2022.02.034 36. Kasowitz SD. **Nuclear m6A reader YTHDC1 regulates alternative polyadenylation and splicing during mouse oocyte development**. *PLoS Genet.* (2018.0) **14** e1007412. DOI: 10.1371/journal.pgen.1007412 37. Huang T. **YTHDF2 promotes spermagonial adhesion through modulating MMPs decay via m(6)A/mRNA pathway**. *Cell Death Dis.* (2020.0) **11** 37. DOI: 10.1038/s41419-020-2235-4 38. Lasman L. **Context-dependent functional compensation between Ythdf m(6)A reader proteins**. *Genes Dev.* (2020.0) **34** 1373-1391. DOI: 10.1101/gad.340695.120 39. Meikar O, Da Ros M, Korhonen H, Kotaja N. **Chromatoid body and small RNAs in male germ cells**. *Reproduction* (2011.0) **142** 195-209. DOI: 10.1530/REP-11-0057 40. Lehtiniemi T, Kotaja N. **Germ granule-mediated RNA regulation in male germ cells**. *Reproduction* (2018.0) **155** R77-R91. DOI: 10.1530/REP-17-0356 41. O’Donnell L. **Mechanisms of spermiogenesis and spermiation and how they are disturbed**. *Spermatogenesis* (2014.0) **4** e979623. DOI: 10.4161/21565562.2014.979623 42. Manterola M. **BRDT is an essential epigenetic regulator for proper chromatin organization, silencing of sex chromosomes and crossover formation in male meiosis**. *PLoS Genet.* (2018.0) **14** e1007209. DOI: 10.1371/journal.pgen.1007209 43. Xiong M. **Conditional ablation of Raptor in the male germline causes infertility due to meiotic arrest and impaired inactivation of sex chromosomes**. *FASEB J.* (2017.0) **31** 3934-3949. DOI: 10.1096/fj.201700251R 44. Royo H. **Evidence that meiotic sex chromosome inactivation is essential for male fertility**. *Curr. Biol.* (2010.0) **20** 2117-2123. DOI: 10.1016/j.cub.2010.11.010 45. Turner JM. **Meiotic silencing in mammals**. *Ann. Rev. Genet.* (2015.0) **49** 395-412. DOI: 10.1146/annurev-genet-112414-055145 46. Ichijima Y. **MDC1 directs chromosome-wide silencing of the sex chromosomes in male germ cells**. *Genes Dev.* (2011.0) **25** 959-971. DOI: 10.1101/gad.2030811 47. Royo H. **ATR acts stage specifically to regulate multiple aspects of mammalian meiotic silencing**. *Genes Dev.* (2013.0) **27** 1484-1494. DOI: 10.1101/gad.219477.113 48. Hirota T. **SETDB1 links the meiotic DNA damage response to sex chromosome silencing in mice**. *Dev. Cell* (2018.0) **47** 645-659 e646. DOI: 10.1016/j.devcel.2018.10.004 49. Petry S. **Mechanisms of mitotic spindle assembly**. *Ann. Rev. Biochem.* (2016.0) **85** 659-683. DOI: 10.1146/annurev-biochem-060815-014528 50. Lee IW. **Distinct roles of Cep192 and Cep152 in acentriolar MTOCs and spindle formation during mouse oocyte maturation**. *FASEB J.* (2018.0) **32** 625-638. DOI: 10.1096/fj.201700559RR 51. Gomez-Ferreria MA. **Human Cep192 is required for mitotic centrosome and spindle assembly**. *Curr. Biol:* (2007.0) **17** 1960-1966. DOI: 10.1016/j.cub.2007.10.019 52. Joukov V, Walter JC, De Nicolo A. **The Cep192-organized aurora A-Plk1 cascade is essential for centrosome cycle and bipolar spindle assembly**. *Mol. Cell* (2014.0) **55** 578-591. DOI: 10.1016/j.molcel.2014.06.016 53. Jones KT. **Turning it on and off: M-phase promoting factor during meiotic maturation and fertilization**. *Mol. Hum. Reprod.* (2004.0) **10** 1-5. DOI: 10.1093/molehr/gah009 54. Soh YQS. **Meioc maintains an extended meiotic prophase I in mice**. *PLoS Genet.* (2017.0) **13** e1006704. DOI: 10.1371/journal.pgen.1006704 55. Gaysinskaya V, Soh IY, van der Heijden GW, Bortvin A. **Optimized flow cytometry isolation of murine spermatocytes**. *Cytometry A* (2014.0) **85** 556-565. DOI: 10.1002/cyto.a.22463 56. Dominissini D. **Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq**. *Nature* (2012.0) **485** 201-206. DOI: 10.1038/nature11112 57. Meyer KD. **Comprehensive analysis of mRNA methylation reveals enrichment in 3’ UTRs and near stop codons**. *Cell* (2012.0) **149** 1635-1646. DOI: 10.1016/j.cell.2012.05.003 58. Soumillon M. **Cellular source and mechanisms of high transcriptome complexity in the mammalian testis**. *Cell Rep.* (2013.0) **3** 2179-2190. DOI: 10.1016/j.celrep.2013.05.031 59. Tu SW, Bugde A, Luby-Phelps K, Cobb MH. **WNK1 is required for mitosis and abscission**. *Proc. Natl Acad. Sci. USA* (2011.0) **108** 1385-1390. DOI: 10.1073/pnas.1018567108 60. Park OH. **Identification and molecular characterization of cellular factors required for glucocorticoid receptor-mediated mRNA decay**. *Genes Dev.* (2016.0) **30** 2093-2105. DOI: 10.1101/gad.286484.116 61. Medvedev S, Yang J, Hecht NB, Schultz RM. **CDC2A (CDK1)-mediated phosphorylation of MSY2 triggers maternal mRNA degradation during mouse oocyte maturation**. *Dev. Biol.* (2008.0) **321** 205-215. DOI: 10.1016/j.ydbio.2008.06.016 62. Herman AB. **FXR1 is an IL-19-responsive RNA-binding protein that destabilizes pro-inflammatory transcripts in vascular smooth muscle cells**. *Cell Rep.* (2018.0) **24** 1176-1189. DOI: 10.1016/j.celrep.2018.07.002 63. Kimura M, Ishida K, Kashiwabara S, Baba T. **Characterization of two cytoplasmic poly(A)-binding proteins, PABPC1 and PABPC2, in mouse spermatogenic cells**. *Biol. Rep.* (2009.0) **80** 545-554. DOI: 10.1095/biolreprod.108.072553 64. Sun F, Palmer K, Handel MA. **Mutation of Eif4g3, encoding a eukaryotic translation initiation factor, causes male infertility and meiotic arrest of mouse spermatocytes**. *Dev.* (2010.0) **137** 1699-1707. DOI: 10.1242/dev.043125 65. Kang JY. **LLPS of FXR1 drives spermiogenesis by activating translation of stored mRNAs**. *Science* (2022.0) **377** eabj6647. DOI: 10.1126/science.abj6647 66. Meikar O. **An atlas of chromatoid body components**. *RNA* (2014.0) **20** 483-495. DOI: 10.1261/rna.043729.113 67. Fanourgakis G, Lesche M, Akpinar M, Dahl A, Jessberger R. **Chromatoid body protein TDRD6 supports long 3’ UTR triggered nonsense mediated mRNA decay**. *PLoS Genet.* (2016.0) **12** e1005857. DOI: 10.1371/journal.pgen.1005857 68. Anders M. **Dynamic m(6)A methylation facilitates mRNA triaging to stress granules**. *Life Sci. Alliance* (2018.0) **1** e201800113. DOI: 10.26508/lsa.201800113 69. Fu Y, Zhuang X. **m(6)A-binding YTHDF proteins promote stress granule formation**. *Nat. Chem. Biol.* (2020.0) **16** 955-963. DOI: 10.1038/s41589-020-0524-y 70. Youn JY. **High-density proximity mapping reveals the subcellular organization of mRNA-associated granules and bodies**. *Mol. Cell* (2018.0) **69** 517-532.e511. DOI: 10.1016/j.molcel.2017.12.020 71. Protter DSW, Parker R. **Principles and properties of stress granules**. *Trends Cell Biol.* (2016.0) **26** 668-679. DOI: 10.1016/j.tcb.2016.05.004 72. Ries RJ. **m(6)A enhances the phase separation potential of mRNA**. *Nature* (2019.0) **571** 424-428. DOI: 10.1038/s41586-019-1374-1 73. Jin G. **Low-frequency coding variants at 6p21.33 and 20q11.21 are associated with lung cancer risk in Chinese populations**. *Am. J. Hum. Genet.* (2015.0) **96** 832-840. DOI: 10.1016/j.ajhg.2015.03.009 74. Pan Q. **Association of MHC class-III gene polymorphisms with ER-positive breast cancer in Chinese Han population**. *Genet. Mol. Res.* (2012.0) **11** 4299-4306. DOI: 10.4238/2012.September.17.1 75. Zhang J. **Common genetic variants in PRRC2A are associated with both neuromyelitis optica spectrum disorder and multiple sclerosis in Han Chinese population**. *J. Neurol.* (2021.0) **268** 506-515. DOI: 10.1007/s00415-020-10184-z 76. Singal DP, Li J, Zhu Y. **HLA class III region and susceptibility to rheumatoid arthritis**. *Clin. Exp. Rheumatol.* (2000.0) **18** 485-491. PMID: 10949724 77. Hashimoto M. **Genetic contribution of the BAT2 gene microsatellite polymorphism to the age-at-onset of insulin-dependent diabetes mellitus**. *Hum. Genet.* (1999.0) **105** 197-199. DOI: 10.1007/s004399900100 78. Thorleifsson G. **Genome-wide association yields new sequence variants at seven loci that associate with measures of obesity**. *Nat. Genet.* (2009.0) **41** 18-24. DOI: 10.1038/ng.274 79. Martin M. **Cutadapt removes adapter sequences from high-throughput sequencing reads**. *EMBnet J.* (2011.0) **17** 10-12. DOI: 10.14806/ej.17.1.200 80. Dobin A. **STAR: ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013.0) **29** 15-21. DOI: 10.1093/bioinformatics/bts635 81. Liao Y, Smyth GK, Shi W. **featureCounts: an efficient general purpose program for assigning sequence reads to genomic features**. *Bioinformatics* (2014.0) **30** 923-930. DOI: 10.1093/bioinformatics/btt656 82. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol.* (2014.0) **15** 550. DOI: 10.1186/s13059-014-0550-8 83. Ingolia NT, Brar GA, Rouskin S, McGeachy AM, Weissman JS. **The ribosome profiling strategy for monitoring translation in vivo by deep sequencing of ribosome-protected mRNA fragments**. *Nat. Protoc.* (2012.0) **7** 1534-1550. DOI: 10.1038/nprot.2012.086 84. Morlan JD, Qu K, Sinicropi DV. **Selective depletion of rRNA enables whole transcriptome profiling of archival fixed tissue**. *PLoS One* (2012.0) **7** e42882. DOI: 10.1371/journal.pone.0042882 85. Xiao Z, Zou Q, Liu Y, Yang X. **Genome-wide assessment of differential translations with ribosome profiling data**. *Nat. Commun.* (2016.0) **7** 11194. DOI: 10.1038/ncomms11194 86. Langmead B, Salzberg SL. **Fast gapped-read alignment with Bowtie 2**. *Nat. Methods* (2012.0) **9** 357-359. DOI: 10.1038/nmeth.1923 87. Zhou Y. **Metascape provides a biologist-oriented resource for the analysis of systems-level datasets**. *Nat. Commun.* (2019.0) **10** 1523. DOI: 10.1038/s41467-019-09234-6 88. Subramanian A. **Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles**. *Proc. Natl Acad. Sci. USA* (2005.0) **102** 15545-15550. DOI: 10.1073/pnas.0506580102 89. Feng J, Liu T, Qin B, Zhang Y, Liu XS. **Identifying ChIP-seq enrichment using MACS**. *Nat. Protoc.* (2012.0) **7** 1728-1740. DOI: 10.1038/nprot.2012.101 90. Quinlan AR, Hall IM. **BEDTools: a flexible suite of utilities for comparing genomic features**. *Bioinformatics* (2010.0) **26** 841-842. DOI: 10.1093/bioinformatics/btq033 91. Robinson JT. **Integrative genomics viewer**. *Nat. Biotechnol.* (2011.0) **29** 24-26. DOI: 10.1038/nbt.1754 92. Heinz S. **Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities**. *Mol. Cell* (2010.0) **38** 576-589. DOI: 10.1016/j.molcel.2010.05.004 93. Schindelin J. **Fiji: an open-source platform for biological-image analysis**. *Nat. Methods* (2012.0) **9** 676-682. DOI: 10.1038/nmeth.2019
--- title: GPAT3 regulates the synthesis of lipid intermediate LPA and exacerbates Kupffer cell inflammation mediated by the ERK signaling pathway authors: - Guoqiang Fan - Yanfei Li - Yibo Zong - Xiaoyi Suo - Yimin Jia - Mingming Gao - Xiaojing Yang journal: Cell Death & Disease year: 2023 pmcid: PMC10039030 doi: 10.1038/s41419-023-05741-z license: CC BY 4.0 --- # GPAT3 regulates the synthesis of lipid intermediate LPA and exacerbates Kupffer cell inflammation mediated by the ERK signaling pathway ## Abstract In the process of inflammatory activation, macrophages exhibit lipid metabolism disorders and accumulate lipid droplets. Kupffer cells (KCs) are the resident hepatic macrophage with critical defense functions in the pathogenesis of several types of liver disease. How dysregulated lipid metabolism contributes to perturbed KCs functions remains elusive. Here we report that glycerol-3-phosphate acyltransferase 3 (GPAT3) plays a key role in KCs inflammation response. Our findings indicate that lipopolysaccharide (LPS)-mediated inflammatory activation markedly increased lipid droplets (LDs) accumulation in KCs. This increase could be attributed to significantly up-regulated GPAT3. The loss of GPAT3 function obviously reduced KCs inflammation reaction both in vivo and in vitro, and was accompanied by improved mitochondrial function and decreased production of lysophosphatidic acid (LPA), in turn inhibiting extracellular regulated protein kinases (ERK) signaling pathway. Overall, this study highlights the role of GPAT3 in inflammatory activation of KCs and could thus be a potential therapeutic target for the treatment of inflammation-related liver disease. ## Introduction Inflammatory liver injury results in the development of many liver diseases, such as hepatitis, hepatic fibrosis, alcohol-related or non-alcoholic disorders, and hepatoma. Kupffer cells (KCs) are unique resident macrophages of the liver and important defense cells that eliminate bacteria and toxins and play a leading role in the occurrence of endotoxic liver injury by releasing various inflammatory mediators [1–3]. Activated KCs drive the inflammatory response to liver injury by secreting several mediators that regulate inflammation and homeostasis [4]. Previously studies have reported that targeted regulation of KCs reduces the incidence of liver disease, such as steatosis and liver injury [5, 6]. As hepatic disease induced by inflammatory liver injury has become a serious global health threat, the potential regulatory factors and molecular mechanisms of KC inflammation have been emerging as a subject of interest. Recent studies have shown that the immune system and lipid metabolism are in close coordination and cooperation [7–10]. Lipids are required by all cells, ensuring the energy and essential fatty acids necessary for cells, and maintaining basic biochemical and biophysical properties. Upon activation of the inflammatory response, macrophages rapidly induce changes to lipid metabolic and energetic homeostasis [11–13]. These perturbations of lipid metabolism in macrophages change cellular functions. For instance, previous studies have shown lipid-overloaded macrophages in adipose tissue stimulate the release of pro-inflammatory cytokines [14]. Considerable evidences have emerged suggesting that some lipid metabolism-related factors may regulate macrophage inflammation response [15–17]. Therefore, understanding the directional interactions between cellular dysregulated lipid metabolism and KCs functions in inflammation situations is very important. To explore the relationship between KCs inflammation and lipid metabolism and clarify how changes in lipid metabolism are integrated with the signaling pathways that specify macrophage functions, we analyzed the response of KCs to LPS stimulation using transcriptome and lipidomics techniques. We identified the glycerol-3-phosphate acyltransferase 3 (GPAT3) gene related to lipid metabolism, which is highly expressed in activated KCs. We discovered a hitherto unrecognized function of GPAT3 as a regulator of Kupffer cells function that promotes the inflammatory response and mitochondrial dysfunction, and this action is dependent on LPA-mediated ERK signaling pathway. These findings demonstrate the novel function of GPAT3 manipulating KCs inflammation response. These results will contribute to therapy for inflammation-related liver disease. ## High GPAT3 expression and lipid reprogramming in inflammatory Kupffer cells To understand the functional significance of lipid in inflammatory KCs, we determined the transcriptomic and lipidomics profiles of KCs under LPS treatment. A Gene Ontology (GO) enrichment analysis of differentially expressed genes (DEGs) was performed to explore multiple aspects of LPS-stimulated KCs, including organismal systems, metabolism, human diseases, genetic information processing, environmental information processing, and cellular processes (Fig. 1A). Notably, metabolism is one of the main categories in the GO items and lipid metabolism was the top-ranking affected metabolic pathway, highlighting the importance of lipids among the metabolic changes triggered by LPS stimulation (Fig. 1A). We found that the expression of genes responsible for lipid metabolism changed markedly in inflammatory KCs, including genes involved in triacylglycerol (TG) synthesis and utilization (Fig. 1B). We also determined that GPAT3 was the most significantly upregulated, as shown in the volcano plot (Fig. 1C). Furthermore, we confirmed by quantitative polymerase chain reaction (qPCR) and western blot that GPAT3 increased significantly after LPS stimulation (Fig. 1D). Immunofluorescence staining further revealed that LPS treatment increased GPAT3 expression in KCs of mice livers (Fig. 1E). In addition, GPAT3 mRNA expression was significantly increased in liver of patients with nonalcoholic steatohepatitis (NASH) by GEO database analysis (Database were obtained from GEO database: http://www.ncbi.nlm.nih.gov/geo/, Data set number is GSE63067) (Fig. 1F). We also found that GPAT3 mRNA expression increased significantly after the KC inflammatory response was induced by TNF-α (Figs. 1G and S1A). GPAT3 is localized in the ER membrane, there are three other subtypes GPAT1, GPAT2 and GPAT4 [18, 19]. GPATs is the rate-limiting enzyme in the de novo pathway of glycerolipid synthesis. Our results found that the mRNA expression of GPAT1, GPAT2 and GPAT4 did not change significantly after LPS treatment in KCs, only GPAT1 decreased after LPS treatment with 12 h (Fig. S1B). These results suggested that GPAT3 may play an important role in the inflammatory process of KCs. Fig. 1Inflammatory activation results in the high GPAT3 expression. A GO analysis of differentially expressed genes (DEGs) in the normal and LPS-stimulated KCs (stimulated with 1 μg/ml LPS for 24 h). DEGs were classified under six categories as indicated. Red arrow indicates lipid metabolism as the top-ranking affected metabolism pathway ($$n = 3$$). B Heatmap showing the expression of lipid metabolism pathway genes, as measured using transcriptomics, in the normal and LPS-stimulated KCs ($$n = 3$$). C Volcanic map showing the GPAT3 was the most significant difference among lipid metabolism genes after LPS stimulated KCs for 24 h ($$n = 3$$). D qPCR and Western blot analysis of GPAT3 expression in normal and LPS (100 ng/ml)-stimulated KCs at different time points ($$n = 3$$). E Immunofluorescence staining for GPAT3 expression in KCs from the LPS-treated liver mice tissues. F$\frac{4}{80}$ was used as a KCs marker. DAPI was used to visualize nuclei. Scale bars represent 50 μm ($$n = 3$$). F The mRNA expression of GPAT3 in human liver from the healthy and NASH (7 healthy human and 9 NASH patients, Database were obtained from GEO database: http://www.ncbi.nlm.nih.gov/geo/, Data set number is GSE63067). G GPAT3 mRNA expression in KCs after stimulated with TNF-α (100 ng/ml) for 24 h ($$n = 3$$). H LPA concentrations in the supernatant of KCs after stimulated with LPS for 24 h ($$n = 6$$). I *Lipidomic analysis* showing the levels of different LPC species levels in KCs ($$n = 6$$). Data represents mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ The lipidomics analysis revealed broad, remarkable changes in lipid composition as a result of activation of the inflammatory response, and the principal component analysis further revealed significant differences in lipids between the LPS and control groups (Fig. S1C). GPAT3 catalyzes the conversion of glycerol-3-phosphate to LPA [19]. The enzyme-linked immunosorbent assay (ELISA) results show that the secretion of LPA significantly increased after treating the KCs with LPS (Fig. 1H). Moreover, the content of lysophosphatidylcholine (LPC; LPA can be generated by catalyzing LPC) also increased significantly according to the lipidomics analysis (Fig. 1I). Plasma and hepatic TG levels increased significantly in LPS-treated mice (Fig. S1D). KCs, as the first barrier against pathogens in the liver microenvironment, also accumulated TG and LDs under LPS stimulation (Fig. S1E-H), consistent with the previous report that inflammatory macrophages increase the accumulation of lipid droplets [17, 20–22]. Taken together, our data revealed that inflammatory KCs induced high GPAT3 expression and lipids reprogramming. ## Blocking GPAT3 inhibits LPS-induced Kupffer cells inflammation We transfected GPAT3 siRNA (si-GPAT3) into KCs to determine the role of GPAT3 in inflammatory KCs. As expected, the expression of GPAT3 mRNA and protein was significantly inhibited after transfection with si-GPAT3 (Fig. S2A, B). We next queried the effects of the loss of GPAT3 function on the LPS-induced inflammatory response. As a result, transfected GPAT3 siRNA in LPS-stimulated KCs was accompanied by a significant reduction in inflammatory capacity, as measured by interleukin (IL)-1β, NOD-, LRR- and pyrin domain-containing protein 3 (NLRP3), IL-6, IL-1α; cyclooxygenase (COX)2; and TNF-α expression (Fig. 2A–C and Fig. S2C, D). In addition, transcriptome analysis was performed on inflammatory KCs transfected with si-GPAT3; 42 genes were significantly upregulated, and 74 genes were significantly downregulated in the si-GPAT3 group compared to the si-N.C. group (Fig. S2E). Notably, we observed reductions in the expression of genes involved in the inflammatory response, including IL-1α, IL-6, and IL-1β and reductions in mRNAs encoding the chemokine (C-X-C motif) ligand 10 (Cxcl10), chemokine (C-C motif) ligand 5 (Ccl5) and Ccl2 in inflammatory KCs as a result of inhibiting GPAT3 (Fig. 2D). Gene expression of the anti-inflammatory factor IL-10 increased significantly after inhibiting GPAT3 (Fig. 2D). The Gene Ontology (GO) enrichment analysis further demonstrated that the loss of GPAT3 function significantly affected the immune response of LPS-activated KCs (Fig. 2E). FSG67 (2-(nonylsulfonamido) benzoic acid, 10-4577-Focus Biomolecules, Plymouth Meeting, PA, USA) is a GPAT inhibitor with a broad-spectrum inhibitory effect on GPAT activity [23]. The FSG67 treatment significantly inhibited the protein expression of GPAT3 (Fig. 2F). Therefore, we also used FSG67 for further study. The expression of IL-1β and IL-6 decreased significantly under the FSG67 treatment in KCs (Fig. 2G, H). Overall, these results suggest that blocking GPAT3 promoted the strong anti-inflammatory ability of activated KCs. Fig. 2Blocking GPAT3 decreases inflammatory response in LPS-stimulated Kupffer cells. A The mRNA expression of IL-1α, IL-6, IL-1β and NLRP3 in si-N.C. or si-GPAT3 KCs with or without LPS (100 ng/ml, 12 h) ($$n = 3$$). B, C The protein levels of IL-1β, NLRP3, COX2, IL-1α and IL-6 in si-N.C. or si-GPAT3 KCs with or without LPS (100 ng/ml, 12 h) ($$n = 3$$). D Heatmap showing the inflammatory cytokines genes expression in LPS-stimulated KCs with or without si-GPAT3 as measured by transcriptomics ($$n = 3$$). E Scatter plot of GO enrichment differential genes of si-N.C. vs si-GPAT3 ($$n = 3$$). F The expression of GPAT3 in KCs treated with FSG67 (150 μM, 1 h) was analyzed by Western blot and qPCR ($$n = 3$$). G, H The levels of IL-1β and IL-6 in KCs were pretreated with 150 μM FSG67 for 6 h and then were treated with 100 ng/mL LPS for 6 h ($$n = 3$$). Data represents mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ ## GPAT3 deletion improves LPS-induced Kupffer cells mitochondrial dysfunction Mitochondrial β oxidation involves the conversion of long-chain FA to acylcarnitine by the carnitine palmitoyltransferase 1a (Cpt1a) [24]. Cpt1a decreased in inflammatory KCs according to the transcriptome and qPCR analyses (Fig. 3A, B). Cpt1a converts long-chain FAs into acylcarnitine in the outer mitochondrial membrane, which are transported to the mitochondrial intima and finally into the matrix for FA oxidation [25]. Decreased Cpt1a was associated with a significant drop in acylcarnitine in the LPS group (Fig. 3C). We reasoned that a reduced KC inflammatory response associated with the loss of GPAT3 may be associated with changes in mitochondrial function. Therefore, we assessed mitochondrial function. The mitochondrial membrane potential (JC-1) (Fig. 3D) and mitochondrial mass (Fig. 3E) increased, respectively, in cells that lacked GPAT3 function under the LPS treatment. Mitochondrial reactive oxygen species (ROS) content (MitoSOX) was decreased in the loss of GPAT3 function after LPS treatment (Fig. 3F). Consistent with impaired mitochondrial function in inflammatory macrophages [26–28], transmission electron microscopy showed that the mitochondrial cristae in LPS-stimulated KCs were looser than resting cells, but this parameter improved by inhibiting GPAT3 (Fig. 3G). In addition, Cpt1b increased significantly in inflammatory KCs after transfection with GPAT3 siRNA (Fig. S2F). Although Cpt1b is mainly expressed in the myocardium and skeletal muscle, it is still detected in KCs. GPAT3 is located primarily in the endoplasmic reticulum (ER) [29] and is thought to be responsible for TG synthesis in different cells [30, 31]. We speculate that the ER may be affected by the loss of GPAT3 function. However, a lack of transcriptional responses within the unfolded-protein response genes Herpud1, Edem2, Bax, Eif2ak3, Bcap31, Atf6, Bak1, Serinc3, and Xbp1 indicated that GPAT3 deficiency may do not cause ER stress in LPS-activated KCs (Fig. S2G).Fig. 3The loss of GPAT3 function improve mitochondrial function in inflammatory Kupffer cells. A Heatmap showing the Cpt1a, Cpt1c, Cpt1b and *Cpt2* genes expression in the normal and LPS-stimulated KCs (stimulated with 1 μg/ml LPS for 24 h) ($$n = 3$$). B qPCR analysis of Cpt1a expression in normal and LPS (100 ng/ml)-stimulated KCs at different time points ($$n = 3$$). C *Lipidomics data* showing the AcCa (13:0) content with or without LPS (1 μg/ml 24 h) ($$n = 6$$). D The ratio of PE to FITC indicates the level of mitochondrial membrane potential. JC-1 fluorescence and F mitochondrial reactive oxygen species (MitoSOX) levels in si-N.C. or si-GPAT3 KCs with or without LPS (1 μg/ml, 18 h) (MFI, mean fluorescence intensity) ($$n = 3$$). E Mitotracker fluorescence visualized by fluorescence microscopy (scale bars represent 50 μm) ($$n = 3$$). G Electron microscopy of KCs and quantification of mitochondrial ultrastructural abnormalities at 18 h post stimulation with LPS without or with si-GPAT3. White arrows, mitochondria (scale bars represent 400 nm) ($$n = 3$$). Data represents mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ ## GPAT3 knockout ameliorates LPS-induced hepatic injury and primary KCs inflammation in mice To further verify the function of GPAT3, GPAT3 knockout mice were used. Plasma alanine aminotransferase (ALT), aspartate aminotransferase (AST) and Lactate dehydrogenase (LDH) were significantly decreased in GPAT3-/- mice compared with GPAT3+/+ mice under LPS stimulation (Fig. 4B). In addition, deletion of GPAT3 inhibits LPS-induced plasma IL-1α and TNFα contents (Fig. 4C). Histologically, HE staining revealed smaller lipid droplets and less ballooning degeneration in GPAT3-/- mice compared with GPAT3+/+ mice under LPS stimulation (Fig. 4D). These results suggest that GPAT3 KO ameliorates LPS-induced liver injury in mice. Furthermore, in primary KCs, inflammation response was significantly increased by LPS treated in GPAT3+/+ mice, but deletion of GPAT3 significantly inhibits the rise of inflammation (Fig, 4E). The JC-1was significantly increased in primary KCs that lacked GPAT3 function under the LPS treatment (Fig. 4F).Fig. 4GPAT3 deletion prevents LPS-induced hepatic injury and primary KCs inflammatory response. A Experimental design of LPS-induced inflammation model using GPAT3+/+ and GPAT3-/- mice. B Plasma ALT, AST and LDH concentration of mice ($$n = 6$$). C Plasma IL-1β and IL-6 production of mice ($$n = 6$$). D Representative images of HE staining (Scale bars represent 200 μm) in liver of mice ($$n = 3$$). E The mRNA levels of IL-1α, IL-6 and MCP-1 in primary KCs ($$n = 3$$). F Mitochondrial membrane potential in primary KCs ($$n = 3$$). Data represents mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ ## GPAT3 enhances the Kupffer cells inflammation depending on the synthesis of LPA GPAT3 is a key rate-limiting enzyme in the first step of triglyceride synthesis, the si-GPAT3 treatment reduced TG accumulation in LPS-activated KCs (Fig. 5A). Bodipy $\frac{493}{503}$ staining showed that LD accumulation induced by LPS decreased in the si-GPAT3 treatment group compared to the si-N.C. group (Fig. 5B). Strikingly, the intermediate product of triglyceride synthesis, LPA, was decreased in the KC supernatant that lacked GPAT3 activity under the LPS treatment (Fig. 5C). This may imply that LPS-induced inflammation and abnormal mitochondrial function are related to GPAT3-mediated LPA. Next, we explored whether exogenous LPA promoted inflammatory cytokine production in KCs. Interestingly, exogenous LPA increased IL-1β, NLRP3, IL-1α, IL-6, and COX2 expression in KCs (Fig. 5D–F). In addition, i.p. injected LPA also had significant effects on inflammation in vivo. Although there was no change in body weight between the control and LPA-treated mice (Fig. S3A), LPA aggravated liver injury which showed a significant increase in plasma LDH level of mice (Fig. 5H). Moreover, the LPA treatment significantly increased the plasma levels of the inflammatory products, such as IL-6 and IL-1β (Fig. 5I). HE staining showed that inflammatory cells infiltrated in the LPA group compared to the control group (Fig. 5J). Additionally, LPA promoted the expression of proinflammatory cytokines, including IL-1α, IL-6, IL-1β, TNF-α, and NLRP3 in the liver of mice (Fig. S3B). Also, LPA promoted the accumulation of CD68 content according to the IHC staining analysis (Fig. 5K) and the increase in CD68 expression is of interest, as CD68 is strongly implicated in activated MI-type KCs [32]. Thus, our results indicate that inflammatory KCs increased the GPAT3 level to enhance the synthesis of LPA, which augmented the inflammatory activity signal. Fig. 5Exogenous LPA enhanced inflammatory response in vitro and vivo. A Effect of si-GPAT3 or si-N.C. on TG content in KCs stimulated with LPS (1 μg/ml) for 24 h ($$n = 3$$). B Bodipy $\frac{493}{503}$ fluorescence visualized by fluorescence microscopy (scale bars represent 50 μm) ($$n = 3$$). C LPA concentrations in the supernatant of KCs after LPS (100 ng/ml 12 h) stimulated and with or without si-GPAT3 ($$n = 3$$). D The mRNA expression of IL-1β, NLRP3, IL-1α and IL-6 in KCs treated with LPA (30 μM 12 h) ($$n = 3$$). E, F The effects of exogenous LPA (30 μM 12 h) on the protein expression of inflammatory factors IL-1β, NLRP3, COX2, IL-1α and IL-6 in KCs ($$n = 3$$). G Experimental design of mice treated with LPA. H, I The contents of LDH, IL-6 and IL-1βin plasma of LPA treated with mice ($$n = 6$$). J, K Representative images and quantification of HE staining (Scale bars represent 100 μm) and CD68 IHC (Scale bars represent 200 μm) after LPA treatment in liver of mice ($$n = 3$$). Data represents mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ ## Mitochondrial function and the PKCθ/ERK/c-Jun pathway is involved in the LPA-induced inflammatory response As previously mentioned, the mitochondrial membrane potential (Fig. 3D) increased in mice lacking GPAT3 function under the LPS treatment. Here, exogenously adding LPA reduced the mitochondrial membrane potential of the KCs (Fig. 6A). This further demonstrates that the mitochondrial function affected by deleting GPAT3 was due to reduced LPA synthesis. In addition, a previous study reported that LPA promotes ROS production by activating protein kinase C (PKC) in PC-3 human prostate cancer cells [33], and LPA induces inflammation through the mitogen-activated protein kinase (MAPK) pathway [34, 35]. The expression of PKCθ, P-ERK$\frac{1}{2}$, and P-c-Jun increased significantly under the LPA treatment (Fig. 6B). In contrast, the PKCθ, P-ERK$\frac{1}{2}$, and P-c-Jun pathways were suppressed in cells that lacked GPAT3 function under the LPS treatment (Fig. 6C). Moreover, we pretreated KCs with the ERK inhibitor U0126 to investigate whether the LPA-enhanced P-ERK$\frac{1}{2}$ pathway is required for the inflammatory response. The results showed that U0126 abolished the LPA-induced KC inflammatory response, such as NLRP3 and COX2 expression (Fig. 6D). Moreover, U0126 abolished LPA-induced P-c-Jun activation, but the U0126 treatment did not affect the LPA-induced increase in PKCθ (Fig. 6D). Collectively, these findings demonstrate that GPAT3 promotes the inflammatory response through LPA-mediated activation of PKCθ, P-ERK$\frac{1}{2}$ and P-c-Jun signaling. Fig. 6GPAT3 promotes inflammation response through the LPA-mediated PKCθ, ERK$\frac{1}{2}$ and c-Jun pathway. A Mitochondrial membrane potential (JC-1) in KCs treated with 30 μM LPA for 12 h ($$n = 3$$). B Western blot images and quantification of PKCθ, P-ERK$\frac{1}{2}$ and P-c-Jun protein expression in KCs treated with LPA (30 μM 12 h) ($$n = 3$$). C Western blot images and quantification of GPAT3, PKCθ, P-ERK$\frac{1}{2}$ and P-c-Jun protein expression in LPS-stimulated KCs transfected with GPAT3 siRNA ($$n = 3$$). D Western blot images and quantification of inflammatory cytokines and ERK$\frac{1}{2}$ pathway expression in control or LPA treatment KCs and pretreatment U0126 (10 μM 2 h, ERK inhibitor) ($$n = 3$$). Data represents mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ ## Discussion In the present study, we identified GPAT3 as a key lipid metabolic gene involved in KCs inflammation progression. Functional experiments in vitro and in vivo demonstrated the regulatory role of GPAT3 depletion in LPS-induced KCs inflammation response. Mechanistically, GPAT3 inhibited mitochondrial function and increased the production of LPA, which promoted P-ERK signaling (Fig. 7). Therefore, our findings revealed a functional associated GPAT3 in KCs inflammation progression and identified it as a potential therapeutic target for liver injury treatment. Fig. 7Graphical abstract. LPS induced Kupffer cells lipid metabolism remodeling and high GPAT3 expression. GPAT3 plays a critical role in the pro-inflammatory response and mitochondrial function by regulating the synthesis of LPA and up-regulate P-ERK signaling pathway. Accumulating reports have shown that activated macrophages change lipid composition, and that targeted regulation of fatty acid synthesis may affect the macrophage inflammatory response [17, 36, 37]. KCs are the main macrophages in the liver and are the first barrier involved in the hepatic immune response [38, 39]. Although hepatic inflammatory injury is a huge public health burden worldwide, how dysregulated lipid metabolism contributes to perturbed KC function remains unclear. Previous studies have reported that TLRs lead to the accumulation of TG in activated macrophages through a variety of pathways [40] and that different proinflammatory stimuli lead to reshaping of the lipid in macrophages in a signal-specific manner [11, 16, 17]. Our data show that the LPS treatment resulted in lipid reprogramming in KCs, meanwhile, activated KCs showed an accumulation of LDs. Transcriptome analysis showed that lipid metabolism was the most affected subset of the entire metabolism category in response to LPS stimulation. GPAT3 is localized in the ER membrane, catalyzes glycerol-3-phosphate (G3P) to produce LPA, and is the rate-limiting enzyme for the first reaction in the TG synthetic pathway [29, 41, 42]. GPAT3 plays an important role in intestinal and hepatic lipid homeostasis, dietary lipid absorption, and the production of intestinal hormones [43, 44]. Moreover, loss of function of GPAT3 alleviates insulin resistance and hepatic steatosis in seipin-/-mice (a mouse model of severe congenital generalized lipodystrophy) [45]. Although these studies suggest that GPAT3 plays an important role in regulating lipid homeostasis, the role of GPAT3 in the inflammation of macrophage remains unclear. In this study, we demonstrated that GPAT3 was the most significantly upregulated gene among 52 upregulated lipid metabolic genes detected in the transcriptome of KCs. Our functional study using knockdown and inhibitor FSG67 confirmed the pro-inflammatory effect of GPAT3 in vitro; that is, deleting GPAT3 significantly inhibited LPS-induced KCs inflammation. Furthermore, knockout GPAT3 also had significant effects on inflammation in vivo, and LPS-induced inflammation diminished significantly in GPAT3 KO mice. Whatever, more detailed analyses on the role of GPAT3 in liver inflammation may require studies using KC type-specific GPAT3-deficient mice. Inflammatory-activated macrophages convert FA to TG and store it in LDs while reducing mitochondrial oxidation [22]. LPS treatment of macrophages decreases Cpt1a expression [40, 46]. Consistent with this finding, our results show that Cpt1a expression decreased significantly in LPS-stimulated KCs. Inflammatory KCs had decreased acylcarnitine contents, which was correlated with decreased Cpt1a expression. Cellular acylcarnitine level is a signature of altered mitochondrial function [47, 48], suggesting that the function of GPAT3 is related to mitochondria in inflammatory KCs. Our findings indicate that mitochondrial function improved, including the mitochondrial membrane potential, MitoSOX, and mitochondrial mass in cells that lacked GPAT3 function under the LPS treatment. In support of this notion, we observed mitochondrial structure by transmission electron microscopy (TEM) and found that the mitochondrial cristae in LPS-stimulated KCs were looser, but this parameter improved by inhibiting GPAT3. Like these results, the GPAT inhibitor FSG67 enhances palmitate oxidation of hypothalamic neurons and increases ATP contents [49]. Previous studies have shown that acute and chronic inflammation increases LPA concentration in the mouse brain [34]. Our results show that LPS-stimulated KCs increased secretion of LPA, while LPA was produced by catalyzing glycerol-3-phosphate with GPAT3, suggesting that the function of GPAT3 in inflammatory KCs may depend on the production of LPA. To test this hypothesis, we examined LPA levels in the supernatant of inflammatory KCs with loss of GPAT3 function. As expected, LPA content decreased significantly. LPA increases ROS production and induces the expression of inflammatory cytokines [33, 50]. In our study, the results showed that exogenous LPA promoted the inflammatory response in vitro and an i.p. injection of LPA also had significant effects on inflammation in vivo. An i.p. injection of LPA aggravated liver injury, which was manifested as a significant increase in the plasma LDH level, HE-stained inflammatory cell infiltration, and upregulation of the KC marker CD68. Furthermore, we explored the mechanism of GPAT3 dependence on LPA and found that exogenous LPA promoted the PKCθ, P-ERK$\frac{1}{2}$ and P-c-Jun signaling pathways. Meanwhile, this pathway was suppressed in KCs that lacked GPAT3 function under the LPS treatment. These results are consistent with previous reports that LPA induces inflammation through the PKC/MAPK pathway [33–35]. In conclusion, our current study delineates a previously undiscovered function of GPAT3 in the inflammatory response, and mitochondrial dysfunction. Inflamed KCs trigger the upregulation of GPAT3 closely followed by reprogrammed lipid metabolism properties, and the effect of GPAT3 on the inflammatory response depended on LPA production. LPA is a key link between the elevated GPAT3 levels and a KC lipid metabolic disorder that drives the development of inflammation in KCs. Increasing evidence supports the notion that different inflammatory signals reprogram lipid metabolism in macrophages, making lipids an excellent target for inflammatory therapy [51–53]. Our discovery suggests that targeting GPAT3 may be an effective therapeutic strategy for regulating the inflammatory response of KCs and improving inflammatory liver disease. ## Animals and treatments GPAT3 knockout (KO) mice with a C57BL/6 J background were generated using CRISPR/Cas9 system, and the sgRNA targeting sites were designed on exon 2 [45]. At 8 weeks of age, GPAT3 KO and WT mice were intraperitoneally injected with 5 mg/kg LPS (L2880-Sigma, St. Louis, MO, USA). After 12 h, the plasma and liver were collected and the primary KCs were isolated from liver for subsequent analyses. C57BL/6 J male mice (age 6–8 weeks) were purchased from Yangzhou University Comparative Medical Center. All mice were housed at 22 ± 1 °C, under a 12 h light/12 h dark cycle and fed at the Animal Experiment Center of Nanjing Agricultural University. The mice were allowed to adapt to their environment for one week. All mice had free access to water and food. In the first experiment, mice were intraperitoneally injected with 5 mg/kg LPS, plasma and liver were collected 12 h post-LPS injection to detect the content of TG and GPAT3. In the second experiment, mice were intraperitoneally injected with 3 mg/kg LPA (L7260-Sigma) or an equivalent volume of solvent ($1\%$ BSA, control) once a day for the LPA experiments, after 5 days, the plasma and liver were collected for subsequent analyses. All animal experiments were approved by the Animal Ethics Committee of Nanjing Agricultural University, China. Euthanasia and sampling procedures complied with the “Guidelines on the Ethical Treatment of Experimental Animals” [2006] No. 398 published by the Ministry of Science and Technology, China, and with the “Regulations Regarding the Management and Treatment of Experimental Animals” [2008] No. 45, published by the Jiangsu Provincial People’s Government. ## Cell culture and transfection Kupffer cells (KCs) were obtained from the BeNa Culture Collection (BNCC340733, Beijing, China), and were cultured in Roswell Park Memorial Institute 1640 medium (cat no. 350-000-CL, Wisent, Nanjing, China) containing $10\%$ fetal bovine serum and $1\%$ penicillin/streptomycin (Gibco, Grand Island, NY, USA) at 37 °C in a $5\%$ CO2 atmosphere. Specific GPAT3 small-interfering RNA (siRNA) was synthesized by GenePharma (Shanghai, China) for GPAT3 knockdown, and the sequences of the GPAT3 siRNAs were: sense (CAAGGAGUCAGCUCUUAAATT), antisense (UUUAAGAGCUGACUCCUUGTT). GPAT3 siRNA was transfected into KCs using the JetPRIME® transfection reagent (Polyplus Transfection, Beijing, China). Scrambled siRNA was used as the negative control (si-N.C.). ## Lipidomics Liquid chromatography-mass spectrometry (LC-MS) and the data analysis were performed by BioNovoGene Co., Ltd. (Suzhou, China). Briefly, 107 cells were collected, quickly frozen in liquid nitrogen, and the lipids were extracted with chloroform/methanol ($\frac{2}{1}$, v/v). LC-MS was carried out using an Acquity UPLC® BEH C18 (100 × 2.1 mm, 1.7 µm, Waters, Milford, MA, USA) column on a Thermo Ultimate 3000, and then using the Thermo Q Exactive Focus mass spectrometer. Data were analyzed with LipidSearch software. ## Transcriptomics Total RNA of KCs was isolated and purified using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). The amount and purity of the RNA in each sample were quantified using the NanoDrop ND-1000 (NanoDrop Technologies, Wilmington, DE, USA). The mRNA library was constructed and sequenced by BioNovoGene Co., Ltd. and LC-Bio Technology CO., Ltd. (Hangzhou, China). The differentially expressed mRNAs with fold changes >2 or fold changes <0.5 were selected with a p-value <0.05 using the R packages edgeR or DESeq2, followed by GO enrichment analyses of the differentially expressed mRNAs. ## Isolation of primary Kupffer cells Primary KCs isolation was performed as previously described [54]. Briefly, the liver was perfused with 10 mL of phosphate-buffered saline and then digested with $0.1\%$ type IV collagenase. Following digestion, the liver homogenate was filtered through a 75 μm stainless steel wire mesh to remove undigested tissue. The cell suspension was centrifuged at 50 g (Eppendorf 5810 R, Germany) for 5 min at 4 °C. The top suspension was separated with $60\%$ Percoll and then centrifuged at 2500 g for 25 min. The darker layer in the middle-comprised KCs. ## Total RNA isolation and quantitative polymerase chain reaction (PCR) Total RNA was isolated using TRIzol reagent (Tsingke, Beijing, China). A 1 μg portion of RNA was reverse transcribed to cDNA using the Reverse Transcription Master Kit (Vazyme, Nanjing, China) according to the manufacturer’s instructions. Two microliters of diluted cDNA (1:20, v/v) were used for qPCR with the Mx3000P Real-Time Polymerase Chain Reaction (PCR) System (Stratagene Inc., La Jolla, CA, USA). GAPDH was chosen as the reference gene. All primers were synthesized by Tsingke (Beijing, China). The qPCR primer sequences are listed in Table S1. ## Total protein extraction and western blot analysis KCs were lysed in RIPA buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, $1\%$ NP40, $0.5\%$ Na-deoxycholate, and $0.1\%$ SDS) containing the complete EDTA-free and PhosSTOP protease inhibitor cocktail (Bimake, Houston, TX, USA). The protein concentration was determined following the manufacturer’s protocol for the BCA Protein Assay kit (TransGen Biotech, Beijing, China). A total of 30–50 μg of protein was used for $10\%$ sodium dodecyl sulfate-polyacrylamide gel electrophoresis, which was transferred to a nitrocellulose membrane. The antibodies used for the western blot analysis are listed in Table S2. Images were captured using the Tannon-5200 (Shanghai, China) and band density was analyzed using Image J software. GAPDH was used as a loading control for these specific proteins. ## Flow cytometry KCs were incubated with 3 μM BODIPY (D3922, ThermoFisher, Waltham, MA, USA), JC-1 (C2005, 1:1,000; Beyotime, Beijing, China) or MitoSOX™ Red (M36008, ThermoFisher, 1:1,000) for 30 min, respectively. Data were acquired by flow cytometry on the BD FACSVerse (BD Biosciences, Brea, CA, USA) and analyzed with the BD FACSuite. ## Fluorescence microscopy KCs were stained with 3 μM BODIPY™ $\frac{493}{503}$ for 20 min and then fixed in $4\%$ paraformaldehyde for 30 min. KCs were stained with 200 nmol/L MitoTrackerTM Green FM (Molecular Probes, Invitrogen, Sunnyvale, CA, USA) for 30 min. Subsequently, these cells were stained with DAPI for 5 min and observed by fluorescent microscopy. Image J software was used to analyze the mean fluorescence intensity (MFI) of each image and MFI = sum of fluorescence intensity in the region /Area of the region. Liver sections were dewaxed and antigen repaired with citrate buffer solution. Each section was soaked with in Tris-buffered saline containing $0.3\%$ Triton X-100 for 1 h, blocked with $10\%$ goat serum, and incubated with the primary antibody GPAT3 (20603-1-AP, Proteintech, Wuhang, China) and F$\frac{4}{80}$ (11-4801-82, Invitrogen, USA) overnight at 4 °C and then with the secondary antibody. DAPI was used as a marker for cell nuclei. Images for immunofluorescence staining were captured using a fluorescence microscope. ## Detection of TG content TG content was measured using the Tissue/Cell Triglyceride (TG) Assay Kit (E1013, Applygen, Beijing, China) according to the manufacturer’s instructions. ## Enzyme-linked immunosorbent assay IL-1α, IL-1β, IL-6 and TNF-α (Jiangsu Meimian Industry Co., Ltd., China) and LPA (MyBioSource, San Diego, CA, USA) levels were determined using ELISA kits according to the manufacturer’s protocols. ## Plasma biochemical measurements Plasma TG, ALT, AST, and LDH levels were measured using an automatic biochemical analyzer (7020, Hitachi, Tokyo, Japan). ## Hematoxylin and eosin staining Fresh livers were fixed in $4\%$ paraformaldehyde and then paraffin-embedded. The sections were soaked and stained in *Harris alum* hematoxylin for about 5 min, and then washed in alcohol containing $0.5\%$ hydrochloric acid for 10 s. After washing, the samples were soaked and dyed in eosin for 30 s, then dehydrated, transparent and sealed with neutral balsam. Finally, the morphology of the liver was examined under a microscope. ## Immunohistochemistry Liver tissues were fixed in $4\%$ paraformaldehyde for 24 h, paraffin-embedded, and sectioned at 5 mm. The tissue was dewaxed in xylene and antigen repair was performed by boiling the sections in citric acid buffer for 15 min, cooling for 20 min, and soaking in $3\%$ hydrogen peroxide for 15 min. The sections were blocked in $5\%$ goat serum before incubation with primary antibodies CD68 (BA3638, BOSTER, Wuhang, China) overnight at 4 °C. All sections were incubated with secondary antibody for 30 min before developing the color using the 3,3′-diaminobenzidine tetrahydrochloride substrate. ## Transmission electron microscopy Fresh KCs were prepared and fixed in $0.25\%$ glutaraldehyde, post-fixed in $1\%$ osmium tetroxide, and embedded in resin. Ultrathin sections were cut and stained with uranyl acetate and lead citrate. The mitochondrial ultrastructure was determined with a model H-7650 transmission electron microscope (Hitachi H-7650, Hitachi Technologies, Tokyo, Japan). For quantification of mitochondrial morphological abnormality, we used the method previously reported [55]. Briefly, we assessed four parameters: electron density (light-dark); cristae swelling (tight-swollen); vacuole number (zero, one or two, three or more); and membrane damage (intact-degenerated), and we scored each of these to a scale of 1 (normal) to 3 (abnormal). Individual mitochondria were categorized as normal (score = 0–5); moderate (score = 6–9); or severe (score >10). ## Statistical analysis All statistical analyses were performed using Prism 8 software (GraphPad Software Inc., La Jolla, CA, USA) and the results are presented as mean ± SEM. Differences were detected using either Student’s t-test (two-group comparison) or one-way analysis of variance (more than two groups). A p-value <0.05 was considered significant for all analyses. Further details on statistical analysis are listed in the figure legends. ## Supplementary information Figure S1 Figure S2 Figure S3 Supplementary Figure legends Supplementary tables Reproducibility checklist Original western blots The online version contains supplementary material available at 10.1038/s41419-023-05741-z. ## References 1. Suzuki S, Toledo-Pereyra LH. **Interleukin 1 and tumor necrosis factor production as the initial stimulants of liver ischemia and reperfusion injury**. *J Surg Res* (1994) **57** 253-8. DOI: 10.1006/jsre.1994.1140 2. Li P, He K, Li J, Liu Z, Gong J. **The role of Kupffer cells in hepatic diseases**. *Mol Immunol* (2017) **85** 222-9. DOI: 10.1016/j.molimm.2017.02.018 3. Van der Heide D, Weiskirchen R, Bansal R. **Therapeutic Targeting of Hepatic Macrophages for the Treatment of Liver Diseases**. *Front Immunol* (2019) **10** 2852. DOI: 10.3389/fimmu.2019.02852 4. Smedsrod B, De Bleser PJ, Braet F, Lovisetti P, Vanderkerken K, Wisse E. **Cell biology of liver endothelial and Kupffer cells**. *Gut* (1994) **35** 1509-16. DOI: 10.1136/gut.35.11.1509 5. Tosello-Trampont AC, Landes SG, Nguyen V, Novobrantseva TI, Hahn YS. **Kuppfer cells trigger nonalcoholic steatohepatitis development in diet-induced mouse model through tumor necrosis factor-alpha production**. *J Biol Chem* (2012) **287** 40161-72. DOI: 10.1074/jbc.M112.417014 6. Stienstra R, Saudale F, Duval C, Keshtkar S, Groener JE, van Rooijen N. **Kupffer cells promote hepatic steatosis via interleukin-1beta-dependent suppression of peroxisome proliferator-activated receptor alpha activity**. *Hepatology* (2010) **51** 511-22. DOI: 10.1002/hep.23337 7. Cildir G, Akincilar SC, Tergaonkar V. **Chronic adipose tissue inflammation: all immune cells on the stage**. *Trends Mol Med* (2013) **19** 487-500. DOI: 10.1016/j.molmed.2013.05.001 8. Qian X, Yang Z, Mao E, Chen E. **Regulation of fatty acid synthesis in immune cells**. *Scand J Immunol* (2018) **88** e12713. DOI: 10.1111/sji.12713 9. Hubler MJ, Kennedy AJ. **Role of lipids in the metabolism and activation of immune cells**. *J Nutr Biochem* (2016) **34** 1-7. DOI: 10.1016/j.jnutbio.2015.11.002 10. Den Brok MH, Raaijmakers TK, Collado-Camps E, Adema GJ. **Lipid Droplets as Immune Modulators in Myeloid Cells**. *Trends Immunol* (2018) **39** 380-92. DOI: 10.1016/j.it.2018.01.012 11. Blanc M, Hsieh WY, Robertson KA, Watterson S, Shui G, Lacaze P. **Host defense against viral infection involves interferon mediated down-regulation of sterol biosynthesis**. *PLoS Biol* (2011) **9** e1000598. DOI: 10.1371/journal.pbio.1000598 12. Liu SY, Aliyari R, Chikere K, Li G, Marsden MD, Smith JK. **Interferon-inducible cholesterol-25-hydroxylase broadly inhibits viral entry by production of 25-hydroxycholesterol**. *Immunity* (2013) **38** 92-105. DOI: 10.1016/j.immuni.2012.11.005 13. York AG, Williams KJ, Argus JP, Zhou QD, Brar G, Vergnes L. **Limiting Cholesterol Biosynthetic Flux Spontaneously Engages Type I IFN Signaling**. *Cell* (2015) **163** 1716-29. DOI: 10.1016/j.cell.2015.11.045 14. Xiu F, Diao L, Qi P, Catapano M, Jeschke MG. **Palmitate differentially regulates the polarization of differentiating and differentiated macrophages**. *Immunology* (2016) **147** 82-96. DOI: 10.1111/imm.12543 15. Chen Y, Yang M, Huang W, Chen W, Zhao Y, Schulte ML. **Mitochondrial Metabolic Reprogramming by CD36 Signaling Drives Macrophage Inflammatory Responses**. *Circ Res* (2019) **125** 1087-102. DOI: 10.1161/CIRCRESAHA.119.315833 16. Oishi Y, Spann NJ, Link VM, Muse ED, Strid T, Edillor C. **SREBP1 Contributes to Resolution of Pro-inflammatory TLR4 Signaling by Reprogramming Fatty Acid Metabolism**. *Cell Metab* (2017) **25** 412-27. DOI: 10.1016/j.cmet.2016.11.009 17. Hsieh WY, Zhou QD, York AG, Williams KJ, Scumpia PO, Kronenberger EB. **Toll-Like Receptors Induce Signal-Specific Reprogramming of the Macrophage Lipidome**. *Cell Metab* (2020) **32** 128-43. DOI: 10.1016/j.cmet.2020.05.003 18. Yu J, Loh K, Song ZY, Yang HQ, Zhang Y, Lin S. **Update on glycerol-3-phosphate acyltransferases: the roles in the development of insulin resistance**. *Nutr Diabetes* (2018) **8** 34. DOI: 10.1038/s41387-018-0045-x 19. Cao J, Li JL, Li D, Tobin JF, Gimeno RE. **Molecular identification of microsomal acyl-CoA:glycerol-3-phosphate acyltransferase, a key enzyme in de novo triacylglycerol synthesis**. *Proc Natl Acad Sci* (2006) **103** 19695-700. DOI: 10.1073/pnas.0609140103 20. Jha AK, Huang SC, Sergushichev A, Lampropoulou V, Ivanova Y, Loginicheva E. **Network integration of parallel metabolic and transcriptional data reveals metabolic modules that regulate macrophage polarization**. *Immunity* (2015) **42** 419-30. DOI: 10.1016/j.immuni.2015.02.005 21. Huang YL, Morales-Rosado J, Ray J, Myers TG, Kho T, Lu M. **Toll-like receptor agonists promote prolonged triglyceride storage in macrophages**. *J Biol Chem* (2014) **289** 3001-12. DOI: 10.1074/jbc.M113.524587 22. Castoldi A, Monteiro LB, van Teijlingen Bakker N, Sanin DE, Rana N, Corrado M. **Triacylglycerol synthesis enhances macrophage inflammatory function**. *Nat Commun* (2020) **11** 4107. DOI: 10.1038/s41467-020-17881-3 23. Kuhajda FP, Aja S, Tu Y, Han WF, Medghalchi SM, El Meskini R. **Pharmacological glycerol-3-phosphate acyltransferase inhibition decreases food intake and adiposity and increases insulin sensitivity in diet-induced obesity**. *Am J Physiol Regul Integr Comp Physiol* (2011) **301** R116-30. DOI: 10.1152/ajpregu.00147.2011 24. Thiam AR, Farese RV, Walther TC. **The biophysics and cell biology of lipid droplets**. *Nat Rev Mol Cell Biol* (2013) **14** 775-86. DOI: 10.1038/nrm3699 25. Wang J, Xiang H, Lu Y, Wu T, Ji G. **The role and therapeutic implication of CPTs in fatty acid oxidation and cancers progression**. *Am J Cancer Res* (2021) **11** 2477-94. PMID: 34249411 26. Cameron AM, Castoldi A, Sanin DE, Flachsmann LJ, Field CS, Puleston DJ. **Inflammatory macrophage dependence on NAD(+) salvage is a consequence of reactive oxygen species-mediated DNA damage**. *Nat Immunol* (2019) **20** 420-32. DOI: 10.1038/s41590-019-0336-y 27. Cogliati S, Frezza C, Soriano ME, Varanita T, Quintana-Cabrera R, Corrado M. **Mitochondrial cristae shape determines respiratory chain supercomplexes assembly and respiratory efficiency**. *Cell* (2013) **155** 160-71. DOI: 10.1016/j.cell.2013.08.032 28. Buck MD, O’Sullivan D, Klein Geltink RI, Curtis JD, Chang CH, Sanin DE. **Mitochondrial Dynamics Controls T Cell Fate through Metabolic Programming**. *Cell* (2016) **166** 63-76. DOI: 10.1016/j.cell.2016.05.035 29. Wendel AA, Lewin TM, Coleman RA. **Glycerol-3-phosphate acyltransferases: rate limiting enzymes of triacylglycerol biosynthesis**. *Biochim Biophys Acta* (2009) **1791** 501-6. DOI: 10.1016/j.bbalip.2008.10.010 30. Shan D, Li JL, Wu L, Li D, Hurov J, Tobin JF. **GPAT3 and GPAT4 are regulated by insulin-stimulated phosphorylation and play distinct roles in adipogenesis**. *J Lipid Res* (2010) **51** 1971-81. DOI: 10.1194/jlr.M006304 31. Quiroga IY, Pellon-Maison M, Gonzalez MC, Coleman RA, Gonzalez-Baro MR. **Triacylglycerol synthesis directed by glycerol-3-phosphate acyltransferases -3 and -4 is required for lipid droplet formation and the modulation of the inflammatory response during macrophage to foam cell transition**. *Atherosclerosis* (2021) **316** 1-7. DOI: 10.1016/j.atherosclerosis.2020.11.022 32. Ikarashi M, Nakashima H, Kinoshita M, Sato A, Nakashima M, Miyazaki H. **Distinct development and functions of resident and recruited liver Kupffer cells/macrophages**. *J Leukoc Biol* (2013) **94** 1325-36. DOI: 10.1189/jlb.0313144 33. Lin CC, Lin CE, Lin YC, Ju TK, Huang YL, Lee MS. **Lysophosphatidic acid induces reactive oxygen species generation by activating protein kinase C in PC-3 human prostate cancer cells**. *Biochem Biophys Res Commun* (2013) **440** 564-9. DOI: 10.1016/j.bbrc.2013.09.104 34. Plastira I, Bernhart E, Joshi L, Koyani CN, Strohmaier H, Reicher H. **MAPK signaling determines lysophosphatidic acid (LPA)-induced inflammation in microglia**. *J Neuroinflammation* (2020) **17** 127. DOI: 10.1186/s12974-020-01809-1 35. Wang H, Tu S, Yang S, Shen P, Huang Y, Ba X. **Berberine Modulates LPA Function to Inhibit the Proliferation and Inflammation of FLS-RA via p38/ERK MAPK Pathway Mediated by LPA1**. *Evid Based Complement Altern Med* (2019) **2019** 2580207. DOI: 10.1155/2019/2580207 36. Yan J, Horng T. **Lipid Metabolism in Regulation of Macrophage Functions**. *Trends Cell Biol* (2020) **30** 979-89. DOI: 10.1016/j.tcb.2020.09.006 37. Hu J, Wang H, Li X, Liu Y, Mi Y, Kong H. **Fibrinogen-like protein 2 aggravates nonalcoholic steatohepatitis via interaction with TLR4, eliciting inflammation in macrophages and inducing hepatic lipid metabolism disorder**. *Theranostics* (2020) **10** 9702-20. DOI: 10.7150/thno.44297 38. Slevin E, Baiocchi L, Wu N, Ekser B, Sato K, Lin E. **Kupffer Cells: Inflammation Pathways and Cell-Cell Interactions in Alcohol-Associated Liver Disease**. *Am J Pathol* (2020) **190** 2185-93. DOI: 10.1016/j.ajpath.2020.08.014 39. Dixon LJ, Barnes M, Tang H, Pritchard MT, Nagy LE. **Kupffer cells in the liver**. *Compr Physiol* (2013) **3** 785-97. DOI: 10.1002/cphy.c120026 40. Feingold KR, Shigenaga JK, Kazemi MR, McDonald CM, Patzek SM, Cross AS. **Mechanisms of triglyceride accumulation in activated macrophages**. *J Leukoc Biol* (2012) **92** 829-39. DOI: 10.1189/jlb.1111537 41. Alves-Bezerra M, Gondim KC. **Triacylglycerol biosynthesis occurs via the glycerol-3-phosphate pathway in the insect Rhodnius prolixus**. *Biochim Biophys Acta* (2012) **1821** 1462-71. DOI: 10.1016/j.bbalip.2012.08.002 42. Alves-Bezerra M, Ramos IB, De Paula IF, Maya-Monteiro CM, Klett EL, Coleman RA. **Deficiency of glycerol-3-phosphate acyltransferase 1 decreases triacylglycerol storage and induces fatty acid oxidation in insect fat body**. *Biochim Biophys Acta Mol Cell Biol Lipids* (2017) **1862** 324-36. DOI: 10.1016/j.bbalip.2016.12.004 43. Khatun I, Clark RW, Vera NB, Kou K, Erion DM, Coskran T. **Characterization of a Novel Intestinal Glycerol-3-phosphate Acyltransferase Pathway and Its Role in Lipid Homeostasis**. *J Biol Chem* (2016) **291** 2602-15. DOI: 10.1074/jbc.M115.683359 44. Cao G, Konrad RJ, Li SD, Hammond C. **Glycerolipid acyltransferases in triglyceride metabolism and energy homeostasis-potential as drug targets**. *Endocr Metab Immune Disord Drug Targets* (2012) **12** 197-206. DOI: 10.2174/187153012800493459 45. Gao M, Liu L, Wang X, Mak HY, Liu G, Yang H. **GPAT3 deficiency alleviates insulin resistance and hepatic steatosis in a mouse model of severe congenital generalized lipodystrophy**. *Hum Mol Genet* (2020) **29** 432-43. DOI: 10.1093/hmg/ddz300 46. Meiser J, Kramer L, Sapcariu SC, Battello N, Ghelfi J, D’Herouel AF. **Pro-inflammatory Macrophages Sustain Pyruvate Oxidation through Pyruvate Dehydrogenase for the Synthesis of Itaconate and to Enable Cytokine Expression**. *J Biol Chem* (2016) **291** 3932-46. DOI: 10.1074/jbc.M115.676817 47. Gao X, Lee K, Reid MA, Sanderson SM, Qiu C, Li S. **Serine Availability Influences Mitochondrial Dynamics and Function through Lipid Metabolism**. *Cell Rep* (2018) **22** 3507-20. DOI: 10.1016/j.celrep.2018.03.017 48. Smith SA, Ogawa SA, Chau L, Whelan KA, Hamilton KE, Chen J. **Mitochondrial dysfunction in inflammatory bowel disease alters intestinal epithelial metabolism of hepatic acylcarnitines**. *J Clin Invest* (2021) **131** e133371. DOI: 10.1172/JCI133371 49. McFadden JW, Aja S, Li Q, Bandaru VV, Kim EK, Haughey NJ. **Increasing fatty acid oxidation remodels the hypothalamic neurometabolome to mitigate stress and inflammation**. *PLoS One* (2014) **9** e115642. DOI: 10.1371/journal.pone.0115642 50. Chang CL, Lin ME, Hsu HY, Yao CL, Hwang SM, Pan CY. **Lysophosphatidic acid-induced interleukin-1 beta expression is mediated through Gi/Rho and the generation of reactive oxygen species in macrophages**. *J Biomed Sci* (2008) **15** 357-63. DOI: 10.1007/s11373-007-9223-x 51. Batista-Gonzalez A, Vidal R, Criollo A, Carreno LJ. **New Insights on the Role of Lipid Metabolism in the Metabolic Reprogramming of Macrophages**. *Front Immunol* (2019) **10** 2993. DOI: 10.3389/fimmu.2019.02993 52. Shimizu T. **Lipid mediators in health and disease: enzymes and receptors as therapeutic targets for the regulation of immunity and inflammation**. *Annu Rev Pharm Toxicol* (2009) **49** 123-50. DOI: 10.1146/annurev.pharmtox.011008.145616 53. Lawrence T, Willoughby DA, Gilroy DW. **Anti-inflammatory lipid mediators and insights into the resolution of inflammation**. *Nat Rev Immunol* (2002) **2** 787-95. DOI: 10.1038/nri915 54. Fan G, Li Y, Chen J, Zong Y, Yang X. **DHA/AA alleviates LPS-induced Kupffer cells pyroptosis via GPR120 interaction with NLRP3 to inhibit inflammasome complexes assembly**. *Cell Death Dis* (2021) **12** 73. DOI: 10.1038/s41419-020-03347-3 55. Chakrabarti L, Zahra R, Jackson SM, Kazemi-Esfarjani P, Sopher BL, Mason AG. **Mitochondrial dysfunction in NnaD mutant flies and Purkinje cell degeneration mice reveals a role for Nna proteins in neuronal bioenergetics**. *Neuron* (2010) **66** 835-47. DOI: 10.1016/j.neuron.2010.05.024
--- title: 'Health-promoting behavior to enhance perceived meaning and control of life in chronic disease patients with role limitations and depressive symptoms: a network approach' authors: - Je-Yeon Yun - Young Ho Yun journal: Scientific Reports year: 2023 pmcid: PMC10039031 doi: 10.1038/s41598-023-31867-3 license: CC BY 4.0 --- # Health-promoting behavior to enhance perceived meaning and control of life in chronic disease patients with role limitations and depressive symptoms: a network approach ## Abstract The association between health-related role limitations in the mental and physical subdomains and clinical status (i.e., chronic disease and comorbid depressive symptoms) is mediated by health-promoting behaviors. To enhance health-promoting behaviors in adults with chronic disease, it is necessary to identify item-level associations among targets of health-related monitoring and management. Therefore, the current study used a network approach to examine associations among health-related role limitations, depressive symptoms, existential well-being, socioeconomic position, and health-promoting behavior in adults with chronic disease. A total of 535 adults (mean ± SD age = 62.9 ± 11.9 years; males, $$n = 231$$, females, $$n = 304$$) who were regularly visiting an outpatient clinic for chronic disease treatment participated in this cross-sectional study. Data on participant demographics, chronic disease diagnoses, socioeconomic status, health-related role limitations (12-item short form survey scores), depressive symptoms (patient health questionnaire-9 scores), existential well-being (scores for four items of the McGill quality of life questionnaire-Revised), and health-promoting behavior (Healthy Habits Questionnaire scores) were acquired. “ Undirected regularized partial correlations” and “directional joint probability distributions” among these variables were calculated using a mixed graphical model (MGM) and directed acyclic graph (DAG). In the MGM, the most influential nodes were emotional well-being, feelings of failure, and health-related limitations affecting usual role and physical activities. According to both the MGM and DAG, the relationship between emotional well-being and feelings of failure mediated the relationships of health-related role limitations with concentration difficulty and suicidal ideation. A positive mindset was dependent on the probability distributions of suicidal ideation, controllability of life, and positive self-image. Both the meaning of life and a positive mindset had direct associations with proactive living. Specifically, proactive living was associated with a balanced diet, regular exercise, volunteering in the community, and nurturing intimacy in social interactions. The meaning and controllability of life in individuals with chronic diseases could mediate the relationships of health-promoting behavior with health-related limitations related to usual role activities, physical activities, and depressive symptoms. Thus, interventions targeting health-promoting behaviors should aim to enhance the meaning and controllability of life (as it pertains to limitations in usual role and physical activities), as well as promote proactive screening and timely psychiatric treatment of depressive symptoms including feelings of failure, concentration difficulties, and suicidal ideation. ## Importance of health-promoting behavior in the treatment of chronic disease patients Chronic disease can be defined as a medical condition lasting ≥ 1 year that requires ongoing medical attention and/or limits activities of daily living1. Cardiovascular diseases (such as heart attacks and stroke), cancers, chronic respiratory diseases (such as chronic obstructive pulmonary disease and asthma), diabetes, and chronic renal diseases account for > $80\%$ of all chronic disease-related deaths that occur before the age of 70 years1,2. Additionally, chronic diseases such as hypertension, dyslipidemia (high low-density lipoprotein cholesterol), and diabetes are major risk factors for cardiovascular diseases1. Regarding health-related role limitations, chronic arthritis is a leading cause of work disability in the United States and a common cause of chronic pain1. Chronic arthritis affects about one in four adults in the United States and is more prevalent among those diagnosed with diabetes and/or cardiovascular disease compared with adults without chronic diseases3. Osteoporosis is a major risk factor for all-cause mortality in adults aged ≥ 60 years4, and is associated with substantial health-related role limitations even in the absence of frailty fractures5. Globally, 30–$60\%$ of adults have multiple chronic conditions that can negatively affect health, function, and quality of life6,7. For instance, hypertension, diabetes, dyslipidemia, chronic pulmonary disease, chronic renal disease, chronic arthritis, and osteoporosis are frequently comorbid and associated with higher risks of all-cause mortality and health-related role limitations. The likelihood of chronic diseases might depend on the combined effects of genetic, physiological, environmental, and behavioral factors2. For instance, lower socioeconomic status is associated with a higher risk of developing chronic diseases such as cardiovascular disease8,9. Moreover, in adults with multiple chronic diseases, lower educational achievement and monthly household income are associated with worse clinical status at follow-up assessments10. Accordingly, comorbid depressive symptoms and poor health-related behaviors, such as non-adherence to pharmacotherapy, reportedly mediated the relationship between lower socioeconomic status and uncontrolled blood pressure in middle-aged and older adults with hypertension11. To improve patient outcomes and potentially prevent chronic diseases, health-promoting behaviors such as maintaining a healthy diet, engaging in regular exercise and physical activity, getting enough sleep, quitting smoking, limiting alcohol intake, going for regular health screenings, making time for leisure activities, talking with friends about personal concerns and feelings, and participating in social activities within community- or faith-based organizations are crucial12,13. However, the vast majority of individuals newly diagnosed with a chronic disease do not adopt health-promoting behaviors and thus fail to achieve long-term improvements in health14. ## Health-related role limitations and health-promoting behaviors in chronic disease patients Health-related role limitations are a major element of disease-related burden15. Chronic non-communicable diseases (i.e., diseases that are typically caused by unhealthy behaviors rather than being spread through infection) are an important cause of health-related disability and role limitation in developed countries16. The 12-item Short Form Survey (SF-12)17 can be used to measure limitations in physical health (i.e., general health, physical activities, and usual role activities [in association with physical health problems or bodily pain]) and mental health (i.e., vitality, emotional well-being, and usual role or social activities [in association with emotional problems]). Health-related role limitations might be associated with decreased adherence to health-promoting behaviors in patients with chronic disease. Among adults with hypertension, dyslipidemia, and/or diabetes, those with fewer health-related role limitations in physical and mental health subdomains engaged in more health-promoting behaviors, specifically adequate fruit/vegetable intake and abstaining from smoking (which influenced mental health), and adequate physical activity (which influenced physical health, regardless of clinical status)18. Also, ego resilience had a stronger mediating effect on the association between emotional outlook and positive emotion at baseline versus after a health-promoting intervention aimed at increasing physical activity among adults in the workplace19. In contrast, in adults with hypertension, greater health-related role limitations in the mental health subdomain were associated with difficulty adhering to a healthy diet and engaging in regular exercise, as well as current smoking status20. Furthermore, the presence of psychological distress might decrease adherence to recommended preventive care, such as influenza vaccinations and annual dental check-ups21. Currently, the mechanisms underlying the association between health-related functional status and behavioral outcomes (i.e., health-promoting behaviors) are not well understood. ## Comorbid depressive symptoms and health-promoting behaviors in chronic disease patients Chronic diseases such as hypertension22, diabetes mellitus23, pulmonary hypertension24, chronic kidney disease25, and rheumatoid arthritis26 are associated with increased rates of psychological distress, depressive symptoms, anxiety, and cognitive disturbance27–29. Furthermore, more severe comorbid depressive symptoms are associated with a higher number of chronic disease diagnosese30. Disease-related physical limitations and psychological distress, clinical deterioration, maladaptive health-related behaviors, and depressive symptoms are often mutually reinforcing31,32. Furthermore, comorbid depressive symptoms may affect health-related behaviors such as self-care behavior, physical activity, sleep time, eating habits, and compliance with treatment31,33,34, and were associated with adverse health and social outcomes in patients with chronic disease27. For example, depression negatively affects blood sugar control in diabetics because of lowered compliance to behaviors such as following a specific diet, taking medications on time, assessing metabolic parameters, and maintaining a consistent sleep cycle35. Furthermore, adults with comorbid depression are less likely to quit smoking after being diagnosed with acute coronary syndrome36. To date, few studies have examined the associations between health-promoting behaviors and depressive symptoms in patients with chronic disease37–39. ## Existential well-being and health-promoting behaviors in chronic disease patients Existential well-being depends on an individual’s perspective on the meaning and purpose of life, satisfaction regarding their own life, and feelings regarding death and suffering40. The meaningful existence subdomain40–42 of the McGill quality of life questionnaire-revised (MQOL-R)43,44 measures existential well-being in terms of the meaning/purpose of life, progress/fulfillment of life goals, feeling of having control over one’s own life, and positive self-image. Existential well-being and health-promoting behaviors might have a bidirectional relationship. For instance, middle-aged adults with a greater sense of purpose in life are more likely to be physically active and less likely to experience sleep problems compared with those with less sense of purpose45. Furthermore, health-promoting behaviors, such as regular physical activity, at baseline were associated with a stronger purpose of life 4 years later46. Greater existential well-being has been linked with a lower rate of depression and better-preserved health. In patients with chronic renal disease undergoing peritoneal dialysis, fewer health-related role limitations in the physical and mental subdomains, as well as milder depressive symptoms, were correlated with a stronger sense of meaning and purpose in life47. Conversely, a diagnosis of stroke and/or depression at baseline appeared to contribute to a weaker sense of purpose 4 years later46. Existential well-being, i.e., a sense of control in life, is associated with emotional well-being (positive affect and optimism), less severe depressive symptoms, and health-promoting behaviors including physical activity, good sleep hygiene, and social involvement48. However, few studies have examined the item-level associations among health-related role limitations, depressive symptoms, facets of existential well-being, and diverse health-promoting behaviors in people with chronic diseases. ## Study aim and hypothesis Physical and mental limitations (i.e., limitations in physical activities [due to health problems], role activities [due to physical health problems or bodily pain], vitality, emotional well-being, and usual or social activities [due to emotional problems])49–51, depressive symptoms52,53, existential well-being (meaning/purpose of life, progress/fulfillment of life goals, having control over one’s own life, and positive self-image)40–42, and socioeconomic status (educational attainment, employment, and monthly household income)52–54 are all associated with health-promoting behaviors (maintaining a healthy diet, engaging in regular exercise and physical activity, getting enough sleep, abstaining from smoking, limiting alcohol intake, going for regular health screenings, making time for leisure activities, talking with friends about concerns and feelings, and participating in social activities within community- or faith-based organizations)12,13 in adults with chronic diseases. Moreover, health-related role limitations in the mental and physical subdomains are associated with chronic disease and comorbid depressive symptoms in adults with chronic disease, where these relationships are mediated by health-promoting behaviors51. Adults with chronic diseases who had more severe depressive symptoms experienced greater enhancement of self-efficacy after an intervention targeting health-promoting behaviors55. It can be difficult to capture the multiple mechanisms underlying health-promoting behaviors in adults with chronic disease, which include health status (health-related role limitations), cognitions (existential well-being), emotions (depressive symptoms), and behaviors (i.e., health-promoting behaviors)56–58, using correlation and regression analyses. Furthermore, it can be difficult to determine the variance in item-level responses among participants with similar summary scores59,60. To identify factors that should be targeted by interventions aimed at enhancing health-promoting behaviors in adults with chronic diseases, the relationships among assessment items must be examined61,62. To identify item-level associations among health-related role limitations, depressive symptoms, existential well-being, socioeconomic position, and health-promoting behaviors in adults with chronic disease, we used a network approach38,59,63,64. A mixed graphical model (MGM)65 comprised of continuous, categorical, and/or ordinal variables can be used to “regularize” partial correlations. In addition, directed acyclic graphs (DAGs)63 displaying probability distributions among variables, and the magnitude and direction of their relationships, can reveal conditional dependence. In the current study, we tested our hypothesis that there may be associations among health-related role limitations in physical and mental subdomains, vitality, emotional well-being, depressive symptoms, existential well-being, socioeconomic position, and health-promoting behaviors using a network approach. We also examined regularized partial correlations (MGMs) and “regularized directional independent associations” (DAGs) among assessment items. We expected to identify factors mediating the relationships of health-promoting behaviors with health-related role limitations, depressive symptoms, existential well-being, and socioeconomic status. ## Participants and data collection In this cross-sectional study, we recruited chronic disease patients who had been diagnosed at Seoul National University Hospital between October 2016 and February 2017. The inclusion criteria were as follows: aged ≥ 19 years, diagnosed with ≥ 1 chronic disease (hypertension4,66, hyperlipidemia66, diabetes4,66, chronic pulmonary disease67,68, chronic renal disease69,70, chronic arthritis67,71, or osteoporosis4,5), visited the outpatient clinic at Seoul National University Hospital between October 2016 and February 2017 for treatment of chronic diseases, fluent in Korean, and willing to participate in this study. A total of 535 adults regularly visited the outpatient clinic for treatment of chronic diseases and completed self-report questionnaires between October 5, 2016 and February 28, 2017. All participants received sufficient information about the study and provided written informed consent. The survey was anonymous and confidential. This study was approved by the clinical research ethics committee of Seoul National University Hospital College of Medicine (IRB number 1601-075-734). All procedures were performed in accordance with relevant guidelines and regulations. ## Measures Demographic information (age, sex, marital status, residential area) and chronic disease diagnoses (hypertension, hyperlipidemia, diabetes mellitus, chronic pulmonary disease, chronic renal disease, chronic arthritis, and osteoporosis) were confirmed by a physician at the outpatient clinic. Information about socioeconomic status (educational attainment, monthly household income, employment status), health-related role limitations (assessed by the 12-item Short Form Survey [SF-12])72, depressive symptoms (assessed by the Patient Health Questionnaire-9 [PHQ-9]73,74, existential well-being (assessed by four items of the MQOL-R)43,44, and health-promoting behavior (assessed by the Healthy Habits Questionnaire [HHQ]75) was acquired; no clinical evaluations were conducted by psychiatrists. ## Health-related role limitations: SF-12 The SF-12 was developed to measure health-related quality of life17. It consists of 12 items distributed among eight subdomains of physical and mental health49: including general health (item 1), vitality (item 10), emotional well-being (items 9 and 11), limitations in physical activities (moderate activities and climbing several stairs) due to health problems (items 2 and 3), limitations in usual role activities due to physical health problems (items 4 and 5), limitations in usual role activities due to bodily pain (item 8), limitations in usual role activities due to emotional problems (items 6 and 7), and limitations in social activities due to physical or emotional problems (item 12)49,50. The 5- and 3-point Likert scale scores for the items were averaged within each subdomain49,50 to be used as nodes in the MGM (Fig. 1) and DAG (Fig. 2).Figure 1(A) Mixed graphical model comprised of health-related role limitation, depressive symptoms, existential well-being, and health behaviors in chronic disease. The “node predictability value” (variance in a given node’s value explained by the nodes with which it is connected) is indicated by the shadowed parts of rings surrounding each node. ( B) Betweenness centrality values (the proportion of the shortest paths in the network containing a given node) of health-related role limitation, depressive symptoms, existential well-being, and health behaviors calculated from (A). The x-axis represents z-scores. Figure 2Directed acyclic graph of health-related role limitation, depressive symptoms, existential well-being, and health behaviors in chronic disease. The magnitude of association between two items is displayed by the thickness of an edge; the thicker the edge, the stronger the association between the two items connected. ## Depressive symptoms: PHQ-9 The PHQ-973,74 evaluates depressive symptoms including hopelessness, anhedonia, sleep disturbance, fatigue, changes in appetite, guilt, concentration difficulties, psychomotor agitation/retardation, and suicidal ideation via nine self-report items. The total scores on the PHQ-9 reflect the severity of depression, which is classified as no depression (score of 0–4), mild depressive symptoms (score of 5–9), moderate depressive symptoms (score of 10–19), or severe depressive symptoms (score of 20–27)76. The 4-point Likert scale scores (none, 3–4 days, 8–10 days, or 12–14 days) for the nine depressive symptoms were used as nodes in the MGM (Fig. 1) and DAG (Fig. 2). ## Existential well-being: MQOL-R The MQOL-R43,44 comprises 14 items and is used to measure the level of subjective well-being throughout the lifespan in terms of physical symptoms (3 items), psychological symptoms (4 items), social support (3 items), and meaningful existence (4 items)40–42. In the current study, the 10-point Likert scale scores for four items related to existential well-being (meaningfulness and purposefulness of life, progress/fulfillment of life goals, having control over one’s own life, and positive self-image as a person) were used as nodes in the MGM (Fig. 1) and DAG (Fig. 2). ## Health-promoting behavior: HHQ The HHQ75 examines the practice of 12 important health-promoting behaviors: positive mindset, regular exercise, balanced diet, proactive attitude, regular medical checkups, volunteering in the community, religious and existential activities, nonsmoking, controlled drinking, work-life balance, nurturing intimacy in social interactions, and consistently taking prescribed medications. Five-point Likert scale scores (no intention, intention of performing within the next six months, intention of performing within the next month, have been practicing for < 6 months, have been practicing for ≥ 6 months) were translated into binary data (have been practicing for any duration = 1, not practicing yet = 0) to be used as nodes in the MGM (Fig. 1) and DAG (Fig. 2). ## Statistical analysis We used network analyses to elucidate the associations among the scores for the eight subdomains of the SF-12, nine PHQ-9 depressive symptoms (including suicidal ideation), four (MQOL-R) existential well-being items, 12 HHQ health-related habits, and three variables related to socioeconomic status (educational attainment, monthly household income, and employment status). “ Undirected” and “directional conditional dependencies”77 among the variables were estimated using MGMs (undirected networks comprising nodes showing conditional dependencies or regularized partial correlations) and a DAG (a Bayesian network showing the conditional probability distributions and directional dependencies of nodes with parent nodes)78,79, respectively. In these networks, variables served as nodes ($$n = 36$$) and edges between nodes represented conditional dependence59. ## MGM This study included categorical, ordinal, and continuous data80, and we estimated “undirected conditional dependencies”77 between variables using MGMs with the R package mgm65. MGMs represent edge weights by generating node-wise regression coefficients81. To prevent potentially spurious associations, mgm employs the least absolute shrinkage and selection operator (LASSO) approach82. LASSO reduces all edge weights so that they approach zero and sets small weights to exactly zero82. To derive MGM networks, we optimized the edge weights during LASSO regularization (controlled by parameter λ) using a pairwise model (interaction order $k = 2$) and the extended Bayesian information criterion (tuning parameter γ = 0)59. Two variables were considered independent if they were not connected when conditioned on other variables83. The thickness of each edge in the network represents the strength of the association, with thicker edges representing stronger associations84. To identify the principal nodes within the MGM networks, we calculated the betweenness centrality (proportion of paths in the network that contain a given node)85; four items (nodes) within the top $12\%$ were defined as hubs38. Similar to R2 regression coefficients, “node predictability values” were estimated, i.e., values reflecting how accurately a node could be predicted by the other nodes it shared an edge with59,62,86. Node predictability values were visualized using pie charts62. The MGM network structure was visualized using the Fruchterman − Reingold algorithm, which was performed using the R package qgraph84. ## DAG The DAG is a Bayesian approach for modeling networks with edges that are directed and noncircular, with the goal of discerning relationships among nodes based on cross-sectional data63. Using the R package bnlearn, we applied the ‘hill-climbing’ greedy search algorithm to our dataset87. The DAG added and removed edges (connecting variables or nodes), and reversed their direction, until the goodness of fit satisfied the BIC88. An iterative bootstrapping procedure with 10,000 iterations was used to determine whether an edge existed between two symptom nodes88,89. In the second step, we generated an averaged network by retaining edges that were consistently present in the 10,000 bootstrapped networks. The cut-off point for consistently present nodes was set using a statistical method with high sensitivity and specificity88,90. Finally, a BIC value was computed for each edge, with higher values (depicted by thicker edges) signifying greater importance within the network structure88. ## Sample characteristics A total of 535 adults [mean ± SD age = 62.9 ± 11.9 years; males, $$n = 231$$; females, $$n = 304$$) participated in the current study. Table 1 provides the demographic, chronic disease diagnosis, socioeconomic status, health-related role limitation (SF-12 scores), depressive symptom (PHQ-9 scores), existential well-being (scores on four MQOL-R items), and health-promoting behavior (HHQ scores) data of the participants. The mean number of diagnosed chronic diseases per participant was 2.1 ± 0.9 and the most common diagnoses were hypertension ($$n = 284$$, $53.1\%$), hyperlipidemia ($$n = 279$$, $52.1\%$), and diabetes ($$n = 249$$, $46.5\%$). Regarding educational attainment, 230 ($43.0\%$) participants had an education level of college or higher level, and another 183 ($34.2\%$) were high school graduates. A total of 197 ($36.8\%$) participants were employed ($$n = 96$$; $17.9\%$) or self-employed ($$n = 101$$, $18.9\%$), and another 338 ($63.2\%$) were unemployed or retired. Only 196 ($36.6\%$) of the participants had a monthly household income ≥ 3000 US dollars (third quintile of the monthly average income of households with one or more family members in the Republic of Korea during 2016 and 2017 [https://kosis.kr/]).Table 1Sample characteristics. Item contentsUnit of responseValuesAbbreviationsDemographics Age [mean ± SD]Years62.9 ± 11.9– Sex [N(%)]Male/female231 ($43.2\%$)/304 ($56.8\%$)– *Marital status* [N(%)]Married/(single, divorced, or widowed)412 ($77.0\%$)/123 ($23.0\%$)– *Residential area* [N(%)]Urban/rural504 ($94.2\%$)/31 ($5.8\%$)–Chronic diseases Hypertension [N(%)]*Current diagnosis* of chronic disease284 ($53.1\%$)– Hyperlipidemia [N(%)]279 ($52.1\%$)– Diabetes [N(%)]249 ($46.5\%$)– Chronic pulmonary disease [N(%)]96 ($17.9\%$)– Chronic renal disease [N(%)]74 ($13.8\%$)– *Chronic arthritis* [N(%)]73 ($13.6\%$)– Osteoporosis [N(%)]67 ($12.5\%$)–Socioeconomic position Educational attainment [N]College or higher/high school/middle school/elementary school/none$\frac{230}{183}$/$\frac{67}{52}$/3Edu *Employment status* [N(%)]Employed or self-employed/unemployed or retired197 ($36.8\%$)/338 ($63.2\%$)Employ Monthly household income [N] < $\frac{800}{800}$–$\frac{1600}{1600}$–$\frac{2400}{2400}$–3200/≥ 3200 US dollars$\frac{72}{80}$/$\frac{99}{88}$/196IncomeHealth-related role limitation: 12-item Short Form Survey (SF-12) General health status [mean ± SD]Excellent[100]/very good[75]/good[50]/fair[25]/poor[0]47.0 ± 21.3SF01 Vitality [mean ± SD]All the time[100]/most of the time[75]/some of the time[50]/a little of the time[25]/none of the time[0]58.8 ± 31.7SF02Emotional well-being [mean ± SD]67.4 ± 24.4SF03 Limitations in social activities due to physical or emotional problems [mean ± SD]82.7 ± 26.0SF04 Limitations in usual role activities due to physical health problems [mean ± SD]All the time[0]/most of the time[25]/some of the time[50]/a little of the time[75]/none of the time[100]69.7 ± 29.6SF05 Limitations in usual role activities due to emotional problems [mean ± SD]77.1 ± 26.8SF06 Limitations in usual role activities due to bodily pain [mean ± SD]Not at all[100]/a little bit[75]/moderately[50]/quite a bit[25]/extremely[0]75.1 ± 27.0SF07 Limitations in physical activities due to health problems [mean ± SD]Yes limited a lot[0]/yes limited a little[50]/no not limited at all[100]66.8 ± 33.1SF08Depressive symptoms: patient health questionnaire-9 (PHQ-9) Anhedonia [mean ± SD]Not at all[0]/2-6 days [1]/7–12 days [2]/nearly everyday [3] out of the last 2 weeks0.6 ± 0.9PHQ01 Hopelessness [mean ± SD]0.5 ± 0.9PHQ02 Sleep disturbance [mean ± SD]0.7 ± 1.0PHQ03 Fatigue [mean ± SD]1.0 ± 1.0PHQ04 Appetite change [mean ± SD]0.5 ± 0.9PHQ05 Feelings of failure [mean ± SD]0.4 ± 0.8PHQ06 Concentration difficulties [mean ± SD]0.4 ± 0.7PHQ07 Psychomotor change [mean ± SD]0.2 ± 0.6PHQ08 Suicidal ideation [mean ± SD]0.2 ± 0.5PHQ09Existential well-being: McGill quality of life questionnaire-revised (MQOL-R) Meaning and purpose of personal existence [mean ± SD]10-Point Likert scale: utterly meaningless and without purpose[0]-to–very purposeful and meaningful[10]7.6 ± 2.2MQOL01 Achieving life goals [mean ± SD]10-Point Likert scale: made no progress whatsoever[0]-to-progressed to Complete fulfillment[10]7.3 ± 2.2MQOL02 Having control over one’s own life [mean ± SD]10-Point Likert scale: no control over my life[0]-to-complete control over my life[10]7.8 ± 2.1MQOL03 Feeling good about oneself as a person [mean ± SD]10-Point Likert scale: completely disagree[0]-to-completeky agree[10]8.0 ± 2.0MQOL04Health-related behaviors: healthy habits questionnaire (HHQ) Positive mindsetNo intension of practice[1]/intension of practice within 6 months[2]/intension of practice within 1 month[3]/have been practicing less than 6 months[4]/have been practicing more than 6 months[5]4.4 ± 1.0HB01 Regular exewrcise4.1 ± 1.2HB02 Healthy eating4.3 ± 1.1HB03 Proactive living4.4 ± 1.0HB04 Regular medical check-up4.6 ± 0.9HB05 Volunteering in community3.5 ± 1.5HB06 Religious activity3.6 ± 1.8HB07 Non-smoking4.7 + 1.0HB08 Controlled drinking4.5 ± 1.1HB09 Work-life balance4.4 ± 1.0HB10 Foster intimate interpersonal relationships4.5 ± 0.9HB11 Compliance to prescribed medication4.8 ± 0.6HB12 According to the SF-12 scores, the general health, physical functioning and role limitations due to physical health, and role limitation due to pain was fair, good, and very good, respectively. Vitality and emotional well-being were also good, and role limitations due to emotional problems and social functioning were very good. The severity of depressive symptoms, as evaluated using the PHQ-9, was mild (mean total PHQ-9 score = 4.4 ± 5.0). Regarding existential well-being, the median MQOL-R satisfaction (meaning/purpose of life, progress/fulfillment of life goals, feeling of having control over one’s own life, and self-image) score was 8 out of 10. Finally, the mean performance of health behavior (regardless of duration) score was 9.4 ± 2.8. The most commonly practiced health-promoting behaviors were consistently taking prescribed medications ($$n = 506$$, $94.6\%$) and abstaining from smoking ($$n = 485$$, $90.7\%$). The least frequently practiced health-promoting behaviors were volunteering in the community ($$n = 279$$, $52.1\%$) and religious and existential activities ($$n = 325$$, $60.7\%$). ## Overview The MGM network for physical and mental quality of life (8 SF-12 domains), depressive symptoms (nine PHQ-9 items including suicidal ideation), existential well-being (four MQOL-R items), health-promoting behavior (12 HHQ items), and socioeconomic status (educational attainment, monthly household income, and employment status) is presented in Fig. 1A and Table 2. Although 636 edges were possible, only 42 ($6.6\%$) were evident in the final MGM. According to the betweenness centrality data displayed in Fig. 1B, four items (nodes) within the top $12\%$ of all 36 nodes comprising the MGM functioned as hubs in the MGM: emotional well-being (calm/peaceful and downhearted/sad; z-score = 3.25), limitations in usual role activities due to emotional problems (z-score = 1.92), physical health problems (z-score = 1.70), and feeling like a failure (z-score = 2.11). Furthermore, the degree to which each node could be predicted by the other nodes with which it shared an edge was estimated; the results are shown in Fig. 1A. In the DAG analysis (Fig. 2, Table 3), the strength of the association between two items is denoted by the thickness of the edge, as stated above91. Table 3 lists the edge weights (from strongest to weakest). The directions of the edge weights indicate the increase or decrease in a given score that would be expected if the arc were removed from the DAG92. If two items are strongly related, the edge weight will be negative and its absolute value will be large91.Table 2Weight adjacency matrix from the estimated mixed graphical model (MGM).ItemsPHQ01PHQ02PHQ03PHQ04PHQ05PHQ06PHQ07PHQ08PHQ09HB01HB02HB03HB04HB05HB06HB07HB08HB09HB10HB11HB12SF01SF04SF05SF07SF06SF03SF02SF08MQOL01MQOL02MQOL03MQOL04eduincomeemployPHQ010.00PHQ020.430.00PHQ030.000.000.00PHQ040.080.200.300.00PHQ050.000.000.140.230.00PHQ060.090.200.000.000.000.00PHQ070.000.130.100.000.000.220.00PHQ080.170.000.000.000.080.000.390.00PHQ090.000.270.000.000.000.220.000.150.00HB010.000.000.000.000.000.000.000.000.000.00HB020.000.000.000.000.000.000.000.000.000.000.00HB030.000.000.000.000.000.000.000.000.000.000.040.00HB040.000.000.000.000.000.000.000.000.000.090.000.050.00HB050.000.000.000.000.000.000.000.000.000.000.000.000.000.00HB060.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00HB070.000.000.000.000.000.000.000.000.000.000.000.000.000.000.030.00HB080.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00HB090.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.120.00HB100.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00HB110.000.000.000.000.000.000.000.000.000.020.000.000.020.000.000.000.000.000.030.00HB120.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00SF010.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00SF040.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00SF050.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.300.00SF070.000.000.000.000.000.060.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.300.00SF060.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00SF030.000.090.000.000.000.200.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.180.000.00SF020.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.300.00SF080.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00MQOL010.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00MQOL020.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.020.000.000.050.00MQOL030.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.030.00MQOL040.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.020.040.050.00edu0.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.00income0.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.030.000.000.000.000.000.000.000.000.000.000.080.00employ0.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.050.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.000.090.00Edge weights indicate the strengths of conditional dependence relations between two different items connected by an undirected edge, which can be understood as partial correlations. With employment of least absolute shrinkage and selection operator (LASSO), all edge weights were shrunk toward zero and small edge weight values were set to exactly zero. Significant values are in bold. Table 3Estimated edge weights in directed acyclic network. RankAssociation directionEdge weightRankAssociation directionEdge weightRankAssociation directionEdge weightFromToFromToFromTo1SF03SF02 − 277.9431SF05SF06 − 15.4961SF06SF08 − 5.152HB01HB04 − 127.5232PHQ07PHQ03 − 13.7562HB10HB05 − 5.113HB08HB09 − 122.2533MQOL01HB04 − 13.5763SF04Income − 4.774SF05SF07 − 114.5434SF04SF06 − 13.5264HB07HB08 − 4.595SF04SF05 − 104.3635SF03SF08 − 13.0765SF02PHQ04 − 4.316SF07SF03 − 103.3536SF05SF08 − 12.3366PHQ01PHQ06 − 4.187PHQ01PHQ02 − 97.1537MQOL04SF01 − 12.0367SF04SF02 − 3.858MQOL01MQOL02 − 67.4438PHQ06PHQ07 − 11.6068SF07SF02 − 3.809MQOL03MQOL04 − 65.4839HB04HB11 − 11.3069PHQ09PHQ07 − 3.7410PHQ04PHQ03 − 63.4240HB06HB07 − 11.2870SF02PHQ06 − 3.7311SF03PHQ01 − 63.1341HB10HB03 − 10.7571SF07SF08 − 3.5612HB11HB10 − 55.9642HB03HB05 − 10.6372SF06PHQ04 − 3.4413MQOL03MQOL02 − 54.8643HB04HB02 − 10.5473PHQ04HB09 − 3.1514SF07MQOL01 − 43.4744HB11HB09 − 9.7674HB01HB02 − 2.7815EduIncome − 37.3545PHQ03PHQ05 − 8.7975SF02SF01 − 2.6116SF03SF06 − 32.7146HB01HB11 − 8.5476MQOL01HB05 − 2.4517PHQ04PHQ05 − 29.3147HB02Employ − 8.4077IncomeHB10 − 2.4118HB04HB03 − 28.2048EmployHB07 − 7.9478EmployHB09 − 2.4019HB04HB06 − 28.0549HB11HB07 − 7.8079HB10HB12 − 2.1420SF03PHQ02 − 25.9850HB02HB10 − 7.1080PHQ02PHQ07 − 1.9721PHQ07MQOL03 − 25.3551SF03MQOL02 − 7.0481EduPHQ01 − 1.9322PHQ02PHQ09 − 24.2352PHQ01PHQ08 − 6.8082MQOL04HB01 − 1.7423PHQ06PHQ09 − 22.3253HB05HB12 − 6.7983MQOL03HB01 − 1.7124PHQ02PHQ06 − 20.6954HB11HB12 − 6.6784PHQ01PHQ07 − 1.6025PHQ02PHQ04 − 20.2355HB01HB05 − 6.4785EmployHB10 − 1.5826PHQ07PHQ08 − 19.8556HB10HB08 − 6.4786SF03SF01 − 1.3927MQOL02MQOL04 − 18.7257HB09HB12 − 6.2287HB03HB08 − 1.2128SF03PHQ06 − 18.3758PHQ09HB01 − 6.0088PHQ01PHQ04 − 0.9329MQOL01MQOL04 − 17.3559MQOL03HB11 − 5.5589HB01HB03 − 0.9230IncomeEmploy − 17.1060PHQ09PHQ08 − 5.4490SF05PHQ04 − 0.18Directional weights with negative numbers and larger absolute values indicate stronger directional associations between two different items connected by directional edges in the directed acyclic network. ## Perceived meaning, achievement of life goals, control of life and health-promoting behavior Perceived meaning, achievement of life goals and control of one’s life were directly influenced by emotional problem-related limitations in usual role activities, emotional well-being (calm/peaceful and not sad), and concentration difficulties, respectively (Fig. 2, Table 3). There was a partial correlation between perceived achievement of life goals (existential well-being) and emotional well-being (calm/peaceful and not sad; $r = 0.02$), which contributed to the high predictive value ($61.8\%$) for the achievement of life goals item (Fig. 1A, Table 2). Positive mindset, as the hub node for health-promoting behavior in the DAG, was influenced by the joint probability distributions of suicidal ideation, perceived control of one’s life, and positive self-image. Positive mindset and perceived meaning of life (existential well-being) were both parent nodes of proactive living (health-promoting behavior). Furthermore, positive mindset, proactive living, and perceived controllability of one’s life were directly associated with the fostering of intimate interpersonal relationships (health-promoting behavior) (Fig. 2, Table 3). The types of health-promoting behavior with the highest node predictability values in the MGM were proactive living ($43.3\%$; via partial correlations with positive mindset [$r = 0.09$], balanced diet [$r = 0.05$], and nurturing intimacy in social interactions [$r = 0.02$]), positive mindset ($40.0\%$; via partial correlations with proactive living [$r = 0.09$] and nurturing intimacy in social interactions [$r = 0.02$]), and nonsmoking ($38.4\%$; via a partial correlation with controlled drinking [$r = 0.12$]) (Fig. 1A, Table 2). In the DAG, positive mindset had a direction influence on proactive living, nurturing intimacy in social interactions, regular medical checkups, regular exercise, and eating a balanced diet. Proactive living, which was influenced by the joint probability distributions of positive mindset and the meaning and purpose of life, also had direct associations with a balanced diet, volunteering in the community, nurturing intimacy in social interactions, and regular exercise (Fig. 2, Table 3). ## Associations between health-related role limitations and depressive symptoms Among the physical and mental health subdomains, node predictability values in the MGM were highest for emotional well-being (calm/peaceful and not sad; $80.3\%$) and vitality ($71.4\%$), and lowest for general health ($34.8\%$) and limitations in physical activities due to health problems ($39.5\%$) (Fig. 1A, Table 2). In the DAG, limitations in physical activities due to health problems emerged was the most pivotal node, and had direct associations with limitations in usual role activities due to physical health problems and pain, monthly household income, and vitality. In turn, limitations in physical activities due to health problems had a direct influence on role limitations due to emotional problems and pain, limitations in social activities due to physical or emotional problems, and fatigue (Fig. 2, Table 3). Among the health-related role limitation items, the strongest partial correlations in the MGM were between limitations in usual role activities due to physical health problems and limitations in physical activities due to health problems; limitations in usual role activities due to physical health problems and those due to emotional problems; and emotional well-being (calm/peaceful and not sad) and vitality ($r = 0.30$ for all three relationships; Fig. 1A, Table 2). Health-related role limitations and depressive symptoms were connected in the MGM through the regularized partial correlation between emotional well-being and feeling like a failure ($r = 0.20$; Fig. 1A, Table 2). In the DAG, feeling like a failure was affected by the joint probability distributions of emotional well-being, vitality, hopelessness, and anhedonia. The emergence of depressive symptoms in the DAG was related to anhedonia through the joint probability distributions of emotional well-being and educational attainment. Anhedonia had direct associations with other depressive symptoms (feelings of hopelessness), psychomotor retardation/agitation, concentration difficulties, fatigue, and feeling like a failure (Fig. 2, Table 3). The predictability of depressive symptoms in the MGM included both low values, i.e., $27.7\%$ (psychomotor agitation/retardation) and $29.9\%$ (changed appetite and eating behaviors), and relatively high values, i.e., $45.0\%$ (fatigue) and $52.0\%$ (hopelessness). The strongest partial correlations among the depressive symptoms in the MGM were between hopelessness and anhedonia ($r = 0.43$), and psychomotor agitation/retardation and concentration difficulty ($r = 0.39$). Notably, the node predictability value of suicidal ideation was $38.5\%$ due to its partial correlations with hopelessness ($r = 0.27$), feeling like a failure ($r = 0.22$), and psychomotor agitation/retardation ($r = 0.15$) (Fig. 1A, Table 2). Likewise, in the DAG, suicidal ideation was influenced by the joint probability distributions of feeling like a failure and hopelessness (Fig. 2, Table 3). ## The relationship between emotional well-being and the feeling of failure in individuals with chronic disease mediates the relationships of limitations in usual role activities due to emotional problems with concentration difficulties and suicidal ideation In the current study, the most influential nodes in the MGM (indicated by the betweenness centrality values) in terms of the regularized partial correlations among items were emotional well-being (calm/peaceful and downhearted/sad), feeling like a failure, limitations in usual role activities due to emotional problems (low mood and anxiety), and limitations in physical activities due to health problems. Likewise, the DAG in the current study revealed direct associations of limitations in physical activities due to health problems with limitations in usual role activities due to physical health problems and/or bodily pain, monthly household income, and vitality. The prevalence of chronic disease is a significant predictor of depressive symptoms93. Physical symptoms such as pain, dyspnea, gait and balance problems, and frailty, in addition to low ability to accept illness and long illness duration, may contribute to reduced physical activity, impaired activities of daily living, limited role functioning in occupational and social settings, depressive symptoms, anxiety, and lower life satisfaction94–99. Medication for chronic disease combined with antidepressants22, behavioral inerventions100, cognitive-behavioral therapy101, and face-to-face social support102 could be helpful for improving the physical health status and decreasing the depressive symptoms of patients with chronic disease. Our findings emphasize the importance of preserving the usual role activities and social functioning of patients with chronic disease. In both the MGM and DAG, regularized partial associations (MGM) and the joint probability distribution of “directional conditional dependencies” (DAG) between emotional well-being (calm/peaceful and downhearted/sad) and feeling like a failure mediated the relationships of limitations in usual role activities due to emotional problems with concentration difficulties and suicidal ideation. In the MGM, the predictability of suicidal ideation was $38.5\%$ due to partial correlations with hopelessness, feeling like a failure, and psychomotor change. Previous studies have also suggested that factors potentially exacerbating suicidal ideation include hopelessness (pessimism about the future), concentration difficulties, low motivation levels, and distress related to serious medical illnesses103–105. In our DAG, suicidal ideation was dependent on the joint probability distributions of feeling like a failure and hopelessness. Additionally, suicidal ideation had direct associations with psychomotor changes, concentration difficulties, and a positive mindset. Combining health education106 (delivered using electronic mobile devices) with motivational interviewing107 could improve depressive symptoms and self-efficacy. ## Associations between the meaning/controllability of life and health-promoting behaviors The current findings emphasize the importance of existential well-being to health-promoting behaviors. In the DAG, a positive mindset and proactive living mediated the relationships of existential well-being with health-promoting behaviors. Specifically, the health-promoting behavior of maintaining a positive mindset was influenced by the joint probability distributions of suicidal ideation, positive self-image, and feelings of control over one’s life. Perceived loss of control over one’s life was significantly associated with perceived unaffordability of long-term life projects108. In a life crisis, such as a new cancer diagnosis, people desire agency in daily life and interpersonal connections109. Therefore, improving the knowledge, skills, attitudes, and self-awareness necessary for good health behaviors could reduce anxiety and depressive symptoms in adults with chronic disease110. In this study, both a positive mindset and strong sense of purpose in life had direct associations with other health-promoting behaviors of proactive living. Proactive living had direct associations with a balanced diet, regular exercise, volunteering in the community, and nurturing intimacy in social interactions. The importance of the meaning and purpose of life has been emphasized with respect to treatment adherence in people with newly diagnosed cancer111. Good health behaviors, such as adherence to prescribed medication, are difficult to achieve and maintain, especially in patients with chronic disease112. Conversely, for people faced with a life crisis, loss of meaning is strongly connected with suicidal ideation104. As the search for and presence of meaning in life are highly correlated in patients with chronic disease or pain, finding meaning in life must be a focus during the management of chronic disease patients113. People with physical illness and pain might find meaning and hope through personal activities and interpersonal relationships111,113. Regular assessments of interpersonal communication and social connections109, combined with cognitive-behavioral therapy114 and behavioral modification115, could enhance emotional well-being and life meaning116 in patients with chronic disease. ## Limitations The current study had several limitations. First, as we used a cross-sectional design, establishing causality in the associations between the different variables was not possible. Second, we used self-report measures to examine health-promoting behavior. Real-time tracking of daily behaviors using wearable devices, such as smart watches, might provide more detailed and accurate measurements of health-promoting behavior. Third, we did not perform subgroup analyses or construct MGMs and DAGs for individual physical illnesses. However, the seven chronic diseases examined in this study (hypertension, diabetes, dyslipidemia, chronic pulmonary disease, chronic renal disease, chronic arthritis, and osteoporosis) are frequently comorbid, and 30–$60\%$ of adults have multiple chronic conditions6,7. Future studies with more participants are needed to compare network-level associations among different chronic diseases. ## Conclusions We applied network analysis to identify factors that could be targeted by interventions to enhance health-promoting behaviors in adults with chronic diseases. The meaning and controllability of life of individuals with chronic disease mediated the associations of health-related limitations in usual role activities, physical activities, and depressive symptoms. Interventions targeting health-promoting behaviors, the meaning and controllability of life, limitations of usual role activities, and physical activities, as well as proactive screening and timely psychiatric consultations to address depressive symptoms (such as feelings of failure, concentration difficulties, and suicidal ideation) are needed. ## References 1. 1.(NCCDPHP), N. C. F. C. D. P. A. H. P. About Chronic Diseases. https://www.cdc.gov/chronicdisease/about/index.htm (2022). 2. 2.(WHO), W. H. O. Noncommunicable Diseases. https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases (2020). 3. Barbour KE, Helmick CG, Boring M, Brady TJ. **Vital signs: Prevalence of doctor-diagnosed arthritis and arthritis-attributable activity limitation—United States, 2013–2015**. *Morb. Mortal Wkly. Rep.* (2017.0) **66** 246-253. DOI: 10.15585/mmwr.mm6609e1 4. Gutzwiller JP. **Osteoporosis, diabetes, and hypertension are major risk factors for mortality in older adults: An intermediate report on a prospective survey of 1467 community-dwelling elderly healthy pensioners in Switzerland**. *BMC Geriatr.* (2018.0) **18** 115. DOI: 10.1186/s12877-018-0809-0 5. Hopman WM. **Longitudinal assessment of health-related quality of life in osteoporosis: Data from the population-based Canadian Multicentre Osteoporosis Study**. *Osteoporos Int.* (2019.0) **30** 1635-1644. DOI: 10.1007/s00198-019-05000-y 6. Vetrano DL. **Frailty and multimorbidity: A systematic review and meta-analysis**. *J. Gerontol. A Biol. Sci. Med. Sci.* (2019.0) **74** 659-666. DOI: 10.1093/gerona/gly110 7. Griffith LE. **Multimorbidity frameworks impact prevalence and relationships with patient-important outcomes**. *J. Am. Geriatr. Soc.* (2019.0) **67** 1632-1640. DOI: 10.1111/jgs.15921 8. Lee M, Park S, Lee KS. **Relationship between morbidity and health behavior in chronic diseases**. *J. Clin. Med.* (2020.0) **9** 010121. DOI: 10.3390/jcm9010121 9. Vadini F. **Homocysteine and education but not lipoprotein (a) predict estimated 10-year risk of cardiovascular disease in blood donors: A community based cross-sectional study**. *BMC Cardiovasc. Disord.* (2019.0) **19** 177. DOI: 10.1186/s12872-019-1157-5 10. Mira R, Newton T, Sabbah W. **Inequalities in the progress of multiple chronic conditions: A systematic review of longitudinal studies**. *PLoS ONE* (2022.0) **17** e0263357. DOI: 10.1371/journal.pone.0263357 11. Qin W, Xu L. **Pathways linking relative deprivation to blood pressure control: the mediating role of depression and medication adherence among Chinese middle-aged and older hypertensive patients**. *BMC Geriatr.* (2023.0) **23** 57. DOI: 10.1186/s12877-023-03769-6 12. 12.(NCCDPHP), N. C. F. C. D. P. A. H. P. How You Can Prevent Chronic Diseases. https://www.cdc.gov/chronicdisease/about/prevent/index.htm (2023). 13. 13.(NCCDPHP), N. C. F. C. D. P. A. H. P. Promoting Health for Adults. https://www.cdc.gov/chronicdisease/resources/publications/factsheets/promoting-health-for-adults.htm (2022). 14. Newsom JT. **Health behavior change following chronic illness in middle and later life**. *J. Gerontol. B Psychol. Sci. Soc. Sci.* (2012.0) **67** 279-288. DOI: 10.1093/geronb/gbr103 15. Murray CJ. **Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: A systematic analysis for the Global Burden of Disease Study 2010**. *Lancet* (2012.0) **380** 2197-2223. DOI: 10.1016/s0140-6736(12)61689-4 16. 16.World Health OrganizationWorld Report on Disability 20112011World Health Organization. *World Report on Disability 2011* (2011.0) 17. Clause-Verdreau AC, Audureau É, Leplège A, Coste J. **Contrasted trends in health-related quality of life across gender, age categories and work status in France, 1995–2016: Repeated population-based cross-sectional surveys using the SF-36**. *J. Epidemiol. Community Health* (2019.0) **73** 65-72. DOI: 10.1136/jech-2018-210941 18. González-Chica DA. **Are we reducing the risk of cardiovascular disease and improving the quality of life through preventive health care? Results of a population-based study in South Australia**. *Prev. Med.* (2017.0) **99** 164-170. DOI: 10.1016/j.ypmed.2017.02.007 19. Seaton CL, Bottorff JL, Jones-Bricker M, Lamont S. **The role of positive emotion and ego-resilience in determining men's physical activity following a workplace health intervention**. *Am. J. Mens Health* (2018.0) **12** 1916-1928. DOI: 10.1177/1557988318803744 20. Trivedi RB, Ayotte B, Edelman D, Bosworth HB. **The association of emotional well-being and marital status with treatment adherence among patients with hypertension**. *J. Behav. Med.* (2008.0) **31** 489-497. DOI: 10.1007/s10865-008-9173-4 21. Thorpe JM, Kalinowski CT, Patterson ME, Sleath BL. **Psychological distress as a barrier to preventive care in community-dwelling elderly in the United States**. *Med. Care* (2006.0) **44** 187-191. DOI: 10.1097/01.mlr.0000196965.54871.d5 22. Wang L. **Effects of combination treatment in hypertensive patients with depression: A systematic review and meta-analysis of 27 randomized controlled trials**. *Ther. Clin. Risk Manag.* (2022.0) **18** 197-211. DOI: 10.2147/tcrm.S347622 23. Liu X. **A systematic review and meta-analysis of the prevalence and risk factors of depression in type 2 diabetes patients in China**. *Front. Med.* (2022.0) **9** 759499. DOI: 10.3389/fmed.2022.759499 24. Rawlings GH, Beail N, Armstrong I, Thompson AR. **Self-help cognitive behavioural therapy for anxiety in pulmonary hypertension: Pilot randomised controlled trial**. *ERJ Open Res.* (2022.0) **8** 1. DOI: 10.1183/23120541.00526-2021 25. de Donato A, Buonincontri V, Borriello G, Martinelli G, Mone P. **The dopamine system: Insights between kidney and brain**. *Kidney Blood Press. Res.* (2022.0). DOI: 10.1159/000522132 26. Hill J. **The prevalence of comorbidity in rheumatoid arthritis: A systematic review and meta-analysis**. *Br. J. Community Nurs.* (2022.0) **27** 232-241. DOI: 10.12968/bjcn.2022.27.5.232 27. Kayser J, Hu RX, Rosenscruggs D, Li L, Xiang X. **A systematic review of the impact of select mindfulness interventions on psychological outcomes among older adults with chronic health conditions**. *Clin. Gerontol.* (2022.0). DOI: 10.1080/07317115.2022.2076636 28. Araya R. **Effect of a digital intervention on depressive symptoms in patients with comorbid hypertension or diabetes in Brazil and Peru: Two randomized clinical trials**. *JAMA* (2021.0) **325** 1852-1862. DOI: 10.1001/jama.2021.4348 29. Dingwall KM. **Effectiveness of wellbeing intervention for chronic kidney disease (WICKD): Results of a randomised controlled trial**. *BMC Nephrol.* (2021.0) **22** 136. DOI: 10.1186/s12882-021-02344-8 30. Gabriel A. **Evaluating depressive symptoms among low-socioeconomic-status African American women aged 40 to 75 years with uncontrolled hypertension: A secondary analysis of a randomized clinical trial**. *JAMA Psychiatry* (2021.0) **78** 426-432. DOI: 10.1001/jamapsychiatry.2020.4622 31. Shapiro MS. **Special psychosocial issues in diabetes management: Diabetes distress, disordered eating, and depression**. *Prim. Care* (2022.0) **49** 363-374. DOI: 10.1016/j.pop.2021.11.007 32. Køpfli ML. **Effect of telemonitoring on quality of life for patients with chronic obstructive pulmonary disease—A randomized controlled trial**. *Chron. Respir. Dis.* (2023.0) **20** 14799731231157771. DOI: 10.1177/14799731231157771 33. Uphoff E. **Behavioural activation therapy for depression in adults with non-communicable diseases**. *Cochrane Database Syst. Rev.* (2020.0) **8** 013461. DOI: 10.1002/14651858.CD013461.pub2 34. García-Lara RA. **Prevalence of depression and related factors among patients with chronic disease during the COVID-19 pandemic: A systematic review and meta-analysis**. *Diagnostics* (2022.0) **12** 123094. DOI: 10.3390/diagnostics12123094 35. Sridhar GR. **Can the management of depression in type 2 diabetes be democratized?**. *World J. Diabetes* (2022.0) **13** 203-212. DOI: 10.4239/wjd.v13.i3.203 36. Lovatt S. **Smoking cessation after acute coronary syndrome: A systematic review and meta-analysis**. *Int. J. Clin. Pract.* (2021.0) **75** e14894. DOI: 10.1111/ijcp.14894 37. Yun JY, Kim YK. **Phenotype network and brain structural covariance network of major depression**. *Adv. Exp. Med. Biol.* (2021.0) **1305** 3-18. DOI: 10.1007/978-981-33-6044-0_1 38. Suh WY, Lee J, Yun JY, Sim JA, Yun YH. **A network analysis of suicidal ideation, depressive symptoms, and subjective well-being in a community population**. *J. Psychiatr. Res.* (2021.0) **142** 263-271. DOI: 10.1016/j.jpsychires.2021.08.008 39. Yun JY. **Impact of COVID-19 on lifestyle, personal attitudes, and mental health among Korean medical students: Network analysis of associated patterns**. *Front. Psychiatry* (2021.0) **12** 702092. DOI: 10.3389/fpsyt.2021.702092 40. Cohen SR, Mount BM, Tomas JJ, Mount LF. **Existential well-being is an important determinant of quality of life. Evidence from the McGill quality of life questionnaire**. *Cancer* (1996.0) **77** 576-586. DOI: 10.1002/(SICI)1097-0142(19960201)77:3<576::AID-CNCR22>3.0.CO;2-0 41. Albers G. **Evaluation of quality-of-life measures for use in palliative care: A systematic review**. *Palliat. Med.* (2010.0) **24** 17-37. DOI: 10.1177/0269216309346593 42. Cohen SR. **Validity of the McGill quality of life questionnaire in the palliative care setting: A multi-centre Canadian study demonstrating the importance of the existential domain**. *Palliat. Med.* (1997.0) **11** 3-20. DOI: 10.1177/026921639701100102 43. Hyun Kim S. **Validation study of the Korean version of the McGill quality of life questionnaire**. *Palliat. Med.* (2007.0) **21** 441-447. DOI: 10.1177/0269216307079816 44. Cohen SR. **Measuring the quality of life of people at the end of life: The McGill quality of life questionnaire-revised**. *Palliat. Med.* (2017.0) **31** 120-129. DOI: 10.1177/0269216316659603 45. Kim ES, Shiba K, Boehm JK, Kubzansky LD. **Sense of purpose in life and five health behaviors in older adults**. *Prev. Med.* (2020.0) **139** 106172. DOI: 10.1016/j.ypmed.2020.106172 46. Nakamura JS, Chen Y, VanderWeele TJ, Kim ES. **What makes life purposeful? Identifying the antecedents of a sense of purpose in life using a lagged exposure-wide approach**. *SSM Popul. Health* (2022.0) **19** 101235. DOI: 10.1016/j.ssmph.2022.101235 47. Cheawchanwattana A. **Spiritual well-being and its relationship with patient characteristics and other patient-reported outcomes in peritoneal dialysis patients: Findings from the PDOPPS**. *Nephrology (Carlton)* (2022.0) **27** 621-631. DOI: 10.1111/nep.14034 48. Hong JH. **The positive influence of sense of control on physical, behavioral, and psychosocial health in older adults: An outcome-wide approach**. *Prev. Med.* (2021.0) **149** 106612. DOI: 10.1016/j.ypmed.2021.106612 49. Ware J, Kosinski M, Keller SD. **A 12-item short-form health survey: Construction of scales and preliminary tests of reliability and validity**. *Med. Care* (1996.0) **34** 220-233. DOI: 10.1097/00005650-199603000-00003 50. Gandek B. **Cross-validation of item selection and scoring for the SF-12 health survey in nine countries: Results from the IQOLA project. International quality of life assessment**. *J. Clin. Epidemiol.* (1998.0) **51** 1171-1178. DOI: 10.1016/s0895-4356(98)00109-7 51. Auld JP, Mudd JO, Gelow JM, Hiatt SO, Lee CS. **Self-care moderates the relationship between symptoms and health-related quality of life in heart failure**. *J. Cardiovasc. Nurs.* (2018.0) **33** 217-224. DOI: 10.1097/jcn.0000000000000447 52. Cramm JM, Nieboer AP. **Self-management abilities, physical health and depressive symptoms among patients with cardiovascular diseases, chronic obstructive pulmonary disease, and diabetes**. *Patient Educ. Couns.* (2012.0) **87** 411-415. DOI: 10.1016/j.pec.2011.12.006 53. Guo J. **Diabetes self-management, depressive symptoms, quality of life and metabolic control in youth with type 1 diabetes in China**. *J. Clin. Nurs.* (2013.0) **22** 69-79. DOI: 10.1111/j.1365-2702.2012.04299.x 54. Yun JY, Sim JA, Lee S, Yun YH. **Stronger association of perceived health with socio-economic inequality during COVID-19 pandemic than pre-pandemic era**. *BMC Public Health* (2022.0) **22** 1757. DOI: 10.1186/s12889-022-14176-8 55. Jerant A, Kravitz R, Moore-Hill M, Franks P. **Depressive symptoms moderated the effect of chronic illness self-management training on self-efficacy**. *Med. Care* (2008.0) **46** 523-531. DOI: 10.1097/MLR.0b013e31815f53a4 56. Lightfoot CJ. **The codevelopment of "my kidneys & me": A digital self-management program for people with chronic kidney disease**. *J. Med. Internet Res.* (2022.0) **24** e39657. DOI: 10.2196/39657 57. Wei X. **Effects of different interventions on smoking cessation in chronic obstructive pulmonary disease patients: A systematic review and network meta-analysis**. *Int. J. Nurs. Stud.* (2022.0) **136** 104362. DOI: 10.1016/j.ijnurstu.2022.104362 58. Salmoirago-Blotcher E. **Mind your heart: Exploring feasibility, acceptability, and preliminary effects of phone-delivered mindfulness training on medication adherence in outpatients with chronic heart failure**. *J. Cardiovasc. Nurs.* (2022.0) **37** 595-602. DOI: 10.1097/jcn.0000000000000891 59. Fried EI. **Using network analysis to examine links between individual depressive symptoms, inflammatory markers, and covariates**. *Psychol. Med.* (2020.0) **50** 2682-2690. DOI: 10.1017/s0033291719002770 60. Elovainio M. **Is moderate depression associated with sleep stage architecture in adolescence? Testing the stage type associations using network and transition probability approaches**. *Psychol. Med.* (2021.0) **51** 426-434. DOI: 10.1017/s0033291719003453 61. Fischer R. **Association of burnout with depression and anxiety in critical care clinicians in Brazil**. *JAMA Netw. Open* (2020.0) **3** e2030898. DOI: 10.1001/jamanetworkopen.2020.30898 62. Zavlis O. **How does the COVID-19 pandemic impact on population mental health? A network analysis of COVID influences on depression, anxiety and traumatic stress in the UK population**. *Psychol. Med.* (2021.0). DOI: 10.1017/s0033291721000635 63. Briganti G, Scutari M, McNally RJ. **A tutorial on Bayesian networks for psychopathology researchers**. *Psychol. Methods* (2022.0). DOI: 10.1037/met0000479 64. Yun JY, Shim G, Jeong B. **Verbal abuse related to self-esteem damage and unjust blame harms mental health and social interaction in college population**. *Sci. Rep.* (2019.0) **9** 5655. DOI: 10.1038/s41598-019-42199-6 65. Haslbeck JMB, Waldorp LJ. **mgm: Estimating time-varying mixed graphical models in high-dimensional data**. *J. Stat. Softw.* (2020.0) **93** 1-46. DOI: 10.18637/jss.v093.i08 66. Otieno P. **Multimorbidity of cardiometabolic diseases: A cross-sectional study of patterns, clusters and associated risk factors in sub-Saharan Africa**. *BMJ Open* (2023.0) **13** e064275. DOI: 10.1136/bmjopen-2022-064275 67. Schiltz NK. **Prevalence of multimorbidity combinations and their association with medical costs and poor health: A population-based study of US adults**. *Front. Public Health* (2022.0) **10** 953886. DOI: 10.3389/fpubh.2022.953886 68. Alter P. **Disease progression and age as factors underlying multimorbidity in patients with COPD: Results from COSYCONET**. *Int. J. Chron. Obstruct. Pulmon. Dis.* (2022.0) **17** 1703-1713. DOI: 10.2147/copd.S364812 69. Sullivan MK. **The presence and impact of multimorbidity clusters on adverse outcomes across the spectrum of kidney function**. *BMC Med.* (2022.0) **20** 420. DOI: 10.1186/s12916-022-02628-2 70. Kuan V. **Identifying and visualising multimorbidity and comorbidity patterns in patients in the English National Health Service: A population-based study**. *Lancet Digit. Health* (2023.0) **5** e16-e27. DOI: 10.1016/s2589-7500(22)00187-x 71. Tran TN, Lee S, Oh CM, Cho H. **Multimorbidity patterns by health-related quality of life status in older adults: An association rules and network analysis utilizing Korean National Health and Nutrition Examination Survey**. *Epidemiol. Health.* (2022.0). DOI: 10.4178/epih.e2022113 72. Kim S-H. **Assessment of psychometric properties of the Korean SF-12 v2 in the general population**. *BMC Public Health* (2014.0) **14** 1086-1086. DOI: 10.1186/1471-2458-14-1086 73. Kroenke K, Spitzer RL, Williams JB. **The PHQ-9: Validity of a brief depression severity measure**. *J. Gen. Intern. Med.* (2001.0) **16** 606-613. DOI: 10.1046/j.1525-1497.2001.016009606.x 74. Park SJ, Choi HR, Choi JH, Kim K, Hong JP. **Reliability and validity of the Korean version of the patient health questionnaire-9 (PHQ-9)**. *Anxiety Mood* (2010.0) **6** 119-124. DOI: 10.35144/ghn.2017.7.2.71 75. Kang E, Kim S, Rhee YE, Yun YH. **Development and validation of the highly effective health behavior pattern inventory—Short form**. *Chronic Illn.* (2021.0) **17** 81-94. DOI: 10.1177/1742395319843166 76. Park SJ, Choi HR, Choi JH, Kim K, Hong JP. **Reliability and validity of the Korean version of the patient health questionnaire-9 [PHQ-9]**. *Anxiety Mood* (2010.0) **6** 119-124 77. Epskamp S, Fried EI. **A tutorial on regularized partial correlation networks**. *Psychol. Methods* (2018.0) **23** 617-634. DOI: 10.1037/met0000167 78. Tu YK. **Commentary: Is structural equation modelling a step forward for epidemiologists?**. *Int. J. Epidemiol.* (2009.0) **38** 549-551. DOI: 10.1093/ije/dyn346 79. Kaplan D, Lee C. **Bayesian model averaging over directed acyclic graphs with implications for the predictive performance of structural equation models**. *Struct. Equ. Model.* (2016.0) **23** 343-353. DOI: 10.1080/10705511.2015.1092088 80. Haslbeck JMB, Waldorp LJ. **How well do network models predict observations? On the importance of predictability in network models**. *Behav. Res. Methods* (2018.0) **50** 853-861. DOI: 10.3758/s13428-017-0910-x 81. Wang E, Reardon B, Cherian B, George WT, Xavier RM. **Disorder agnostic network structure of psychopathology symptoms in youth**. *J. Psychiatr. Res.* (2021.0) **143** 246-253. DOI: 10.1016/j.jpsychires.2021.09.026 82. Tibshirani R. **Regression shrinkage and selection via the lasso**. *J. R. Stat. Soc. Ser. B (Methodol.)* (1996.0) **58** 267-288 83. Epskamp S, Borsboom D, Fried EI. **Estimating psychological networks and their accuracy: A tutorial paper**. *Behav. Res. Methods* (2018.0) **50** 195-212. DOI: 10.3758/s13428-017-0862-1 84. Epskamp S, Cramer AOJ, Waldorp LJ, Schmittmann VD, Borsboom D. **qgraph: Network visualizations of relationships in psychometric data**. *J. Stat. Softw.* (2012.0) **48** 1-18. DOI: 10.18637/jss.v048.i04 85. Rubinov M, Sporns O. **Complex network measures of brain connectivity: Uses and interpretations**. *Neuroimage* (2010.0) **52** 1059-1069. DOI: 10.1016/j.neuroimage.2009.10.003 86. Haslbeck JMB, Fried EI. **How predictable are symptoms in psychopathological networks? A reanalysis of 18 published datasets**. *Psychol. Med.* (2017.0) **47** 2767-2776. DOI: 10.1017/s0033291717001258 87. Scutari M. **Learning Bayesian networks with the bnlearn R package**. *J. Stat. Softw.* (2010.0) **35** 2010. DOI: 10.18637/jss.v035.i03 88. Lazarov A. **Symptom structure of PTSD and co-morbid depressive symptoms—A network analysis of combat veteran patients**. *Psychol. Med.* (2020.0) **50** 2154-2170. DOI: 10.1017/s0033291719002034 89. McNally RJ, Heeren A, Robinaugh DJ. **A Bayesian network analysis of posttraumatic stress disorder symptoms in adults reporting childhood sexual abuse**. *Eur. J. Psychotraumatol.* (2017.0) **8** 1341276. DOI: 10.1080/20008198.2017.1341276 90. Scutari M, Nagarajan R. **Identifying significant edges in graphical models of molecular networks**. *Artif. Intell. Med.* (2013.0) **57** 207-217. DOI: 10.1016/j.artmed.2012.12.006 91. Moradi S. **Network modeling of major depressive disorder symptoms in adult women**. *Psychol. Med.* (2022.0). DOI: 10.1017/s0033291722002604 92. Scutari M. **Learning Bayesian networks with the bnlearn R package**. *J. Stat. Softw.* (2010.0) **35** 1-22. DOI: 10.18637/jss.v035.i03 93. Khidri FF. **Physical activity, dietary habits and factors associated with depression among medical students of Sindh, Pakistan, during the COVID-19 pandemic**. *Psychol. Res. Behav. Manag.* (2022.0) **15** 1311-1323. DOI: 10.2147/prbm.S364540 94. Grabich S, Farrelly E, Ortmann R, Pollack M, Wu SS. **Real-world burden of systemic lupus erythematosus in the USA: A comparative cohort study from the Medical Expenditure Panel Survey (MEPS) 2016–2018**. *Lupus Sci. Med.* (2022.0) **9** 640. DOI: 10.1136/lupus-2021-000640 95. Kochovska S. **Persisting breathlessness and activities reduced or ceased: A population study in older men**. *BMJ Open Respir. Res.* (2022.0) **9** 1168. DOI: 10.1136/bmjresp-2021-001168 96. Yohannes AM. **Depressive and anxiety symptoms in patients with COPD: A network analysis**. *Respir. Med.* (2022.0) **198** 106865. DOI: 10.1016/j.rmed.2022.106865 97. Badura-Brzoza K. **Comparative assessment of the relationship between emotional factors and quality of life in a group of patients with type 1 and type 2 diabetes—Preliminary report**. *Psychiatr. Pol.* (2022.0) **56** 123-135. DOI: 10.12740/PP/OnlineFirst/122461 98. Santos D. **Association between frailty and depression among hemodialysis patients: A cross-sectional study**. *Sao Paulo Med. J.* (2022.0) **140** 406-411. DOI: 10.1590/1516-3180.2021.0556.R1.14092021 99. Tanaka T, Okita M, Jenkins S, Kozu R. **Clinical and psychological impact of chronic pain in people with chronic obstructive pulmonary disease**. *Int. J. Chron. Obstruct. Pulmon. Dis.* (2022.0) **17** 893-903. DOI: 10.2147/copd.S359223 100. Gulley LD. **Examining cognitive-behavioral therapy change mechanisms for decreasing depression, weight, and insulin resistance in adolescent girls at risk for type 2 diabetes**. *J. Psychosom. Res.* (2022.0) **157** 110781. DOI: 10.1016/j.jpsychores.2022.110781 101. Vlachou E. **Effectiveness of cognitive behavioral therapy-based interventions on psychological symptoms in adults with type 2 diabetes mellitus: An update review of randomized controlled trials**. *J. Diabetes Complic.* (2022.0) **36** 108185. DOI: 10.1016/j.jdiacomp.2022.108185 102. Ng L, Amatya B, Khan F. **Outcomes of a peer support program in multiple sclerosis in an Australian community cohort: A prospective study**. *J. Neurodegener. Dis.* (2013.0) **2013** 429171. DOI: 10.1155/2013/429171 103. McPherson P, Sall S, Santos A, Thompson W, Dwyer DS. **Catalytic reaction model of suicide**. *Front. Psychiatry* (2022.0) **13** 817224. DOI: 10.3389/fpsyt.2022.817224 104. Bobevski I. **Demoralisation and its link with depression, psychological adjustment and suicidality among cancer patients: A network psychometrics approach**. *Cancer Med.* (2022.0) **11** 815-825. DOI: 10.1002/cam4.4406 105. McNally RJ, Robinaugh DJ, Deckersbach T, Sylvia LG, Nierenberg AA. **Estimating the symptom structure of bipolar disorder via network analysis: Energy dysregulation as a central symptom**. *J. Psychopathol. Clin. Sci.* (2022.0) **131** 86-97. DOI: 10.1037/abn0000715 106. Wu VX. **Development of a community-based e-health program for older adults with chronic diseases: Pilot pre-post study**. *JMIR Aging* (2022.0) **5** e33118. DOI: 10.2196/33118 107. Bilgin A, Muz G, Yuce GE. **The effect of motivational interviewing on metabolic control and psychosocial variables in individuals diagnosed with diabetes: Systematic review and meta-analysis**. *Patient Educ. Couns.* (2022.0). DOI: 10.1016/j.pec.2022.04.008 108. Chirumbolo A, Callea A, Urbini F. **The effect of job insecurity and life uncertainty on everyday consumptions and broader life projects during COVID-19 pandemic**. *Int. J. Environ. Res. Public Health* (2021.0) **18** 105363. DOI: 10.3390/ijerph18105363 109. Lynch KA. **Navigating a newly diagnosed cancer through clinician-facilitated discussions of health-related patient values: A qualitative analysis**. *BMC Palliat. Care* (2022.0) **21** 29. DOI: 10.1186/s12904-022-00914-7 110. Duarte-Díaz A. **Increased patient empowerment is associated with improvement in anxiety and depression symptoms in type 2 diabetes mellitus: Findings from the INDICA study**. *Int. J. Environ. Res. Public Health* (2022.0) **19** 084818. DOI: 10.3390/ijerph19084818 111. Leão D. **The importance of spirituality for women facing breast cancer diagnosis: A qualitative study**. *Int. J. Environ. Res. Public Health* (2021.0) **18** 126415. DOI: 10.3390/ijerph18126415 112. Nakata C. **Narrative arcs and shaping influences in long-term medication adherence**. *Soc. Sci. Med.* (2021.0) **285** 114264. DOI: 10.1016/j.socscimed.2021.114264 113. Costanza A. **Meaning in life among patients with chronic pain and suicidal ideation: Mixed methods study**. *JMIR Form. Res.* (2021.0) **5** e29365. DOI: 10.2196/29365 114. Donatti L, Malvezzi H, Azevedo BC, Baracat EC, Podgaec S. **Cognitive behavioral therapy in endometriosis, psychological based intervention: A systematic review**. *Rev. Bras. Ginecol. Obstet.* (2022.0) **44** 295-303. DOI: 10.1055/s-0042-1742406 115. Cuenca-Martínez F. **Behavior modification techniques on patients with chronic pain in the context of COVID-19 telerehabilitation: An umbrella review**. *Int. J. Environ. Res. Public Health* (2022.0) **19** 095260. DOI: 10.3390/ijerph19095260 116. Gagnon P. **A cognitive-existential intervention to improve existential and global quality of life in cancer patients: A pilot study**. *Palliat. Support Care* (2015.0) **13** 981-990. DOI: 10.1017/s147895151400073x
--- title: 40 years of actigraphy in sleep medicine and current state of the art algorithms authors: - Matthew R. Patterson - Adonay A. S. Nunes - Dawid Gerstel - Rakesh Pilkar - Tyler Guthrie - Ali Neishabouri - Christine C. Guo journal: NPJ Digital Medicine year: 2023 pmcid: PMC10039037 doi: 10.1038/s41746-023-00802-1 license: CC BY 4.0 --- # 40 years of actigraphy in sleep medicine and current state of the art algorithms ## Abstract For the last 40 years, actigraphy or wearable accelerometry has provided an objective, low-burden and ecologically valid approach to assess real-world sleep and circadian patterns, contributing valuable data to epidemiological and clinical insights on sleep and sleep disorders. The proper use of wearable technology in sleep research requires validated algorithms that can derive sleep outcomes from the sensor data. Since the publication of the first automated scoring algorithm by Webster in 1982, a variety of sleep algorithms have been developed and contributed to sleep research, including many recent ones that leverage machine learning and / or deep learning approaches. However, it remains unclear how these algorithms compare to each other on the same data set and if these modern data science approaches improve the analytical validity of sleep outcomes based on wrist-worn acceleration data. This work provides a systematic evaluation across 8 state-of-the-art sleep algorithms on a common sleep data set with polysomnography (PSG) as ground truth. Despite the inclusion of recently published complex algorithms, simple regression-based and heuristic algorithms demonstrated slightly superior performance in sleep-wake classification and sleep outcome estimation. The performance of complex machine learning and deep learning models seem to suffer from poor generalization. This independent and systematic analytical validation of sleep algorithms provides key evidence on the use of wearable digital health technologies for sleep research and care. ## Introduction Sleep has intrigued researchers in both basic and clinical sciences for more than a century. With the advancement of sensor technology in the 1970s, sleep researchers started to experiment with the use of accelerometers worn on the wrist to objectively assess sleep and circadian patterns. This research grew rapidly in the 90s with improved sensor technology, increased storage size and improved usability of the devices. The rich body of sleep research using wearable devices has contributed substantially to our understanding on the importance of sleep to overall health1 with poor sleep linked to the progression of many diseases, including depression2, hypertension3, obesity4 and neurodegenerative diseases5. Since the 2010s, wearable devices have been increasingly used as digital heath technologies (DHT) to provide patient-centric clinical outcomes for drug development, where sleep outcomes represent one of the major use cases. Sleep outcomes in clinical trials have traditionally been provided by polysomnography (PSG) and / or questionnaires. While PSG is the gold standard for assessing sleep physiology, its cost and burden can be prohibitive for large-scale deployment. The ecological validity of PSG-based sleep outcomes is also poor due to the need to attach a variety of sensors to the participant and the need for the participant to sleep in a supervised laboratory environment. Due to these limitations, some sleep studies use sleep diaries and questionnaires, but such sleep ratings are known to suffer from subjective and recall bias6. Wrist-worn accelerometer devices therefore offer an attractive solution for objective sleep assessments over multiple nights in the natural home environment. Such devices have low patient burden and can collect high-resolution data continuously for multiple weeks without recharging, thus minimizing the burden on the participants. To derive sleep outcomes, automated scoring algorithms are used to classify sleep and wake based on wrist acceleration. The first such algorithm was developed using simple linear regression and validated against PSG in 19827, with the coefficients of the linear regression equation being updated in 19928. The latter, referred to as Cole-Kripke, has become one of the most used sleep algorithms to date. Due to technical limitations, early actigraphy devices would perform data reduction on the device to convert the raw acceleration data into activity counts and only save the latter9. Many legacy sleep algorithms such as Cole-*Kripke thus* use counts as input features for sleep-wake classification8. These legacy algorithms follow the same format of quantifying activity count-based features around the epoch of interest and applying them in a linear or logistic regression equation to make the binary prediction of sleep or wake10,11. More recently, with the availability of large sleep datasets, machine learning and deep learning methods were applied to sleep-wake classification. Using the Multi-Ethnic Study of Atherosclerosis (MESA) sleep dataset of 1817 participants12, deep learning models were shown to have higher sensitivity and specificity compared to legacy regression-based algorithms13. With technological advancement and needs from the research community, manufacturers started to provide raw acceleration data from actigraphy devices, making it possible to develop sleep algorithms based directly on raw acceleration, not the aggregated activity counts. van Hees et al. [ 2015] used raw acceleration data to derive the angle of the forearm and developed a sleep-wake classification method based on the range of the angle over time14. This group later developed a random forest model on raw acceleration data and showed it to be more accurate than the original model and two legacy algorithms (Cole-Kripke and Sadeh)15. To fully leverage the potential of wrist-worn wearables, the sleep research community can benefit from a systematic evaluation on the analytical validity of the various sleep algorithms developed and used for the last 40 years. Such evidence can support fit-for-purpose use of the proper algorithms in sleep research and therapeutics development. The current manuscript provides the first systematic comparison across simple regression to complex deep-learning models applied to either count or raw data. We assessed the performance of these algorithms in terms of classification accuracy and validity in estimating sleep outcomes such as WASO for direct relevance to clinical applications. ## Performance on Sleep-Wake Classification Sleep-wake classification is an unbalanced classification problem, in that the datasets typically contain more sleep than wake. For such problems, the F1 score is typically used to rank algorithm performance, as it balances precision and sensitivity. However, the algorithms with the highest F1 scores had low specificity (CNN-50, Oakley, all Sleep) (Table 1). Since specificity, which represents classifying wake correctly, is essential to quantifying sleep disturbances, it should be given strong consideration in the evaluation. We thus selected only algorithms that passed minimal levels of sensitivity ($75\%$) and specificity ($45\%$) (Fig. 1; bold, Table 1). Among these algorithms, Oakley-rescore has the highest specificity at $62.8\%$, and van Hees has the best sensitivity of 83.6 and the highest F1 score of 79.1.Table 1Confusion matrix results for the binary classification problem of identifying sleep-wake from various algorithms compared to gold standard PSG annotations. AccuracySensitivitySpecificityPrecisionF1Reference All Sleep69.2 (0.2)100 (0.0)0 (0.0)69.2 (0.2)79.6 (0.2) All Wake30.8 (0.2)0 (0.0)100 (0.0)0 (0.0)0 (0.0)Raw Acceleration van Hees76.2 (0.1)83.6 (0.2)47.5 (0.2)76.2 (0.2)79.1 (0.2) Random Forest73.3 (0.1)77.5 (0.2)55.5 (0.2)77.7 (0.2)76.4 (0.2)Deep Learning Count CNN-5077.3 (0.2)85.6 (0.2)42 (0.2)76.1 (0.2)79.6 (0.2) CNN-2076.2 (0.2)86.2 (0.2)38.2 (0.2)74.9 (0.2)79.4 (0.2) CNN-10077 (0.2)79.8 (0.2)54.2 (0.2)78.4 (0.2)77.8 (0.2) LSTM-5076.3 (0.1)77.8 (0.2)59.0 (0.2)78.8 (0.2)77.1 (0.2) LSTM-10077.2 (0.1)78.6 (0.2)58.9 (0.2)79.7 (0.2)77.8 (0.2) LSTM-2073.5 (0.1)74.8 (0.2)58.5 (0.2)77.9 (0.2)75.4 (0.2)Legacy Count Oakley75.1 (0.1)86.2 (0.1)42 (0.2)75.1 (0.2)79.2 (0.2) Oakley rsc76.4 (0.1)76.9 (0.2)62.8 (0.2)80.6 (0.2)77.8 (0.2) Sadeh75.3 (0.1)82.6 (0.2)49.7 (0.2)76.4 (0.2)78.5 (0.2) Sadeh rsc69.4 (0.1)61.8 (0.2)75.1 (0.2)83 (0.2)68.9 (0.2) Cole-Kripke74.5 (0.1)81.4 (0.2)50.3 (0.2)76.3 (0.2)78 (0.2) Cole-Kripke rsc71.8 (0.1)66.1 (0.2)73.4 (0.2)82.5 (0.2)72.2 (0.2) Sazonov71.3 (0.1)73.3 (0.2)60.3 (0.2)78.2 (0.2)74.8 (0.2) Sazonov rsc63.5 (0.1)51.3 (0.2)83.1 (0.2)84.1 (0.2)61.8 (0.2)The mean values of all cases are presented with standard deviation in brackets. The top performing algorithms are shown in bold. Rsc rescore. Fig. 1Sensitivity and specificity for each algorithm predicting sleep or wake compared to PSG.Sensitivity values are shown in blue and specific values are shown in orange. Algorithms that met the sensitivity and specificity thresholds are indicated with an asterisk. ## Performance on sleep outcome estimation For sleep studies, the goal of sleep-wake classification is to derive the sleep outcomes of interest. We thus examined the performance of the algorithms in estimating sleep outcomes using the Bland-Altman validation approach. Wake after sleep onset (WASO) is a commonly used sleep outcome. Among the selected algorithms, Oakley-rescore showed the lowest root mean squared error (RMSE) and narrowest confidence interval in estimating WASO as compared to the PSG-derived WASO (Table 2, Fig. 2). Cole-Kripke and van Hees also presented low RMSE scores and narrow confidence intervals. These three algorithms also showed much higher correlation with PSG-derived WASO than other algorithms. As for the more complex, machine learning models, the best performers amongst them (Random Forest, LSTM-50 and LSTM-100) showed low mean errors, but high RMSE scores, wide confidence intervals and low correlation with PSG-derived WASO. The performance of all the algorithms on the estimations of WASO, total sleep time (TST) and sleep efficiency (SE) showed similar patterns (Supplementary Tables 1 to 3).Table 2Bland-Altman statistics for wake after sleep onset (WASO) comparison between PSG and the sleep algorithms applied to wrist-based acceleration. MERMSECorrelationCI-$95\%$+CI-$95\%$-CI-widthRaw Acceleration van Hees39.391.10.786201.9−123.4325.3 Random Forest1.093.00.717184.9−183.0367.9Deep Learning Count CNN-10011.999.80.712207.9−184.1392.0 LSTM-500.791.60.743181.9−180.5362.4 LSTM-1002.693.90.736188.3−183.1371.4Legacy Count Oakley rsc−10.481.70.795149.9−170.7320.6 Sadeh33.693.60.756206.3−139.0345.3 Cole-Kripke28.490.40.770198.1−141.3339.4ME mean error, RMSE root mean squared error, CI confidence interval, rsc rescore. Fig. 2Bland-Altman validation plots for estimated WASO from the top performing algorithms in each category compared to PSG.The upper plots show the comparison of each algorithm estimated WASO to PSG WASO in circles, with a dashed line of slope equal to one representing where points would fall if the algorithm and PSG matched perfectly. The lower plots show the PSG WASO plotted against the difference between the algorithm and PSG for each data point. The solid horizontal line represents the mean error and the dashed horizontal lines represent the $95\%$ confidence intervals of the differences. ## Discussion Wrist-worn accelerometers are the most used wearable DHTs in clinical trials and research today. Accelerometer-based DHTs continue to play a central role in clinical research, thanks to the low cost, lightweight, long battery life and decades of research findings and datasets. The algorithms used to derive sleep outcomes from wrist acceleration data have been the focus of technical improvement and continue to evolve with the advancement of data science. Our systematic evaluation of the most common sleep algorithms developed over the past 40 years provides researchers with an evidence-based approach to use them effectively in sleep research. Our findings suggest that the current application of machine learning and deep learning techniques to predict sleep-wake classification are not as robust in estimating sleep outcomes as simple heuristic (van Hees) and legacy regression models (Oakley-rescore and Cole-Kripke). This is surprising, as the deep learning and random forest models were trained on large datasets and thus would be expected to better model the complex relationship between wrist movements and sleep than simple models developed on smaller datasets13,15. The fact that these machine and deep learning models do not perform better may be due to the difference in activity counts used across sleep data sets, suboptimal model architectures, the intrinsic challenge with using motion data only to estimate sleep physiology and different PSG annotation styles between different data sets. Computation of activity counts from raw accelerometer data is a common data reduction step from the early days of actigraphy. The conversion from raw acceleration to activity counts is not always well documented nor understood. Early studies presenting the legacy count algorithms did not provide any information about how the counts were obtained7,8,10,16. In addition, manufacturers might not disclose the way they derive counts from raw accelerometer data. The activity count calculation is a crucial step in the count-based sleep algorithms and differences in the counts would lead to differences in sleep-wake classifications. Despite this, the current research confirmed that legacy algorithms estimate sleep outcomes with high validity on counts computed using the open-source agcounts Python package9. The deep learning count-based algorithms were trained on proprietary counts from the MESA dataset and may not generalize to different types of counts as readily as the simple legacy count algorithms due to the model’s overfitting the PSG annotation style in the MESA dataset13. Due to the lack of available raw acceleration data in large data sets, no raw acceleration-based deep learning models have been presented in the literature. The top-performing deep learning algorithms performed slightly worse than the much simpler heuristic and legacy algorithms. It is possible that the model architectures used in the current deep-learning algorithms may not be optimal15. In particular, the models are very simple, having only one layer of convolution filters or LSTM cells connected to a dense layer. Most deep learning models employ several layers (hence the term deep) which helps them capture more details from the training set. Their training also did not involve regularization techniques designed to avoid overfitting (and thus improve generalization) such as dropout or early stopping. The fact that the models were trained for only 30 epochs may limit their performance compared to heuristic algorithms15. In short, while deep learning algorithms hold promise, there is still work to be done. Polysomnography (PSG) is considered the gold standard for sleep assessment and provides clinical diagnosis of sleep disorders such as apneas, hypopneas and rapid eye movement (REM) disorders17. Using PSG scoring as the ground truth for actigraphy sleep algorithms, however, has some intrinsic challenges. PSG measures physiological changes during sleep, while wrist actigraphy measures the movement of the distal forearm. Physiology and movement present highly structured and correlated patterns during sleep and wake cycles, which is the fundamental principle behind actigraphy use in sleep research. But the intrinsic difference between the two types of source signals means that there is a limit to how close one can be used to estimate the other. This does not mean wrist-based accelerometer assessment of sleep is inferior to PSG, as this method is superior in longitudinal and reliable assessment of sleep patterns in free-living environments. To facilitate the proper use of wrist accelerometer-based sleep outcomes, it may be necessary to interpret actigraphy-quantified sleep endpoints in their own right, and not expect them to perfectly match PSG. Due to the subjective nature of PSG scoring, it may be difficult for a model to generalize between different data sets with different scorers. PSG needs to be scored by trained technicians to derive sleep outcomes. The scoring process takes 2–4 h to score one night of sleep and is also known to have high inter-rater variability, especially in pathological sleep populations18. To improve objectivity and reduce variability, the American Academy of Sleep Medicine (AASM) guidelines provide a series of rules that the PSG technician applies while they score raw PSG data17. For example, the scoring rule for wake is when more than $50\%$ of the epoch contains an alpha rhythm (8–13hz) over the occipital region or eye blinks at 0.5 to 2hz or rapid eye movements associated with normal / high chin muscle tone or reading eye movements17. Such scoring criteria is inherently subjective and leaves room for different raters to score the same segment differently. Automated software packages have been developed to score PSG data; however, these are not considered gold-standard18. With the presence of high inter-rater variability, it may be difficult for a model trained on the relationship between PSG scoring and movement patterns to generalize robustly to different data sets scored by different raters. Deep learning and machine learning models run the risk of overfitting to data set-specific PSG annotator style if they do not include proper model architecture to enhance generalization. While this work is the first systematic comparison across simple regression to complex acceleration-based machine learning sleep algorithms, a subset of these models has been evaluated in previous literature. Sundararajan et al. 2021 reported slightly worse results for Sadeh (F1 68.1 to $78.5\%$), Cole-Kripke (F1 67.5 to $78.0\%$), van Hees (F1 70.1 to $79.1\%$) and Random Forest (F1 73.9 to $76.4\%$) than the present study15. There are several potential reasons for this difference. One, no data was dropped in the current study. Sundararajan et al. 2021 had 24 participants in their test set, while the current work used all 28 participants from the Newcastle PSG dataset. Second, accuracy, sensitivity and other statistics were calculated for each subject and then averaged in the current work, whereas Sundararajan et al. 2021 combined all epochs from all subjects together and calculated the evaluation metrics. The advantage of averaging evaluation metrics from each participant is that a Bland-Altman style validation analysis can be performed on the sleep outcomes. Rescoring is a series of heuristic rules that was developed in conjunction with the original legacy sleep algorithm to rescore periods as wake or sleep based on the length of a period and the length of the surrounding periods7. Previous research showed that rescoring improved performance for all legacy algorithms on the MESA data set13. However, on the Newcastle PSG data set, rescoring resulted in poorer performance for Cole-Kripke (RMSE 13.0 to 18.5), Sadeh (RMSE 13.6 to 23.4) and Sazonov (RMSE 14.5 to 30.5) algorithms, while it improved performance for the Oakley algorithm (RMSE 15.9 to 12.7). In both studies rescoring decreased sensitivity and increased specificity, so algorithms with high sensitivity and low specificity to begin with were improved with rescoring and algorithms with reasonable sensitivity and specificity to begin with were made worse with rescoring. Supplementary Table 1 summarizes sensitivity, specificity and accuracy from previous studies presenting algorithms to predict sleep-wake from wrist accelerometry. The Sadeh model on the *Sadeh data* set had the highest specificity and accuracy of all other model / data set combinations, however the *Sadeh data* set consisted of only healthy sleepers. On the MESA data set, rescoring improved performance for the Sadeh, Cole-Kripke and Oakley algorithms. The van Hees and Sundararajan (random forest) algorithms could not be run on MESA because they require raw acceleration and the MESA data set contains only activity counts. The current work has several limitations. Since the PSG scoring process was not detailed in the open-source Newcastle PSG dataset, we do not know how it was performed. The test data set in the current work was from one PSG study, future work should consider using multiple PSG studies as test data sets, to ensure generalizability to different PSG annotation styles. The challenge with this currently is that many open-source PSG data sets (MESA, STAGES) only include activity count data and not raw acceleration, making it impossible to test the raw acceleration-based algorithms. The current work considered algorithms that used acceleration only. Heart rate and other physiological signals have the potential to improve sleep classification and staging, however, the trade off with more sensors is a decrease in battery life19. This is an important area for future research. ## Sleep dataset PSG data was obtained from the open-source Newcastle PSG dataset14. The study design included ethics approval from the NRES North East Sunderland ethics committee (12/NE/0406) and participants provided written informed consent. PSG data was collected from 28 adult patients (11 female). Mean age of the participants was 44.9 years (14.9 years standard deviation). Concurrent wrist accelerometry data (GENEActiv, Kimboloton, UK) was collected at 85.7hz. All 28 patients had complete data from a left wrist accelerometer and 27 patients had complete data from a right wrist accelerometer. A single night PSG (Embletta, Denver) was performed using a standard procedure that included electroencephalogram (leads C4-A1 and C3-A2), video recording, bilateral eye movements, oxygen saturation, bilateral anterior tibialis EMG, abdominal and chest inductance bands, and submental EMG. All sleep stages were scored according to standard AASM criteria20. Twenty of the participants had at least one sleep disorder. Sleep disorders included idiopathic hypersomnia, restless leg syndrome, sleep apnea, narcolepsy, sleep paralysis, nocturia, obstructive sleep apnea, REM sleep disorder, parasomnia and insomnia. ## Accelerometer data processing Raw tri-axial acceleration data was calibrated using GGIR, an R-package to process multi-day raw accelerometer data21. Analysis was performed by creating scripts in Python (v 3.8.13) that called each algorithm. For each algorithm, sleep-wake classifications were found in 30-second, non-overlapping windows and compared to the PSG-annotated sleep stages. All algorithms are summarized in Table 1. In addition, the random forest model presented in Sundararajan et al. 2021 was included. For the van Hees et al. 2015 algorithm, a python implementation of the algorithm described in the paper was used14. Table 3.Table 3Summary of the compared algorithms. PaperModel TypeInputModel ComplexityDescriptionVan HeesHeuristicAccelerationLowConsiders range of angle-z over 5 min window to classify it as sleep or wakeSundararajanRandom ForestAccelerationHighRandom forest machine learning model trained on 136 sleep patients from two different data setsPalottiCNNCountsHighDeep neural network trained on MESA data set (training $$n = 1454$$) using a convolutional neural network layerPalottiLSTMCountsHighDeep neural network trained on MESA data set (training $$n = 1454$$) using a long-short term memory layerCole-KripkeRegressionCountsMediumCommonly used linear regression modelOakleyRegressionCountsMediumLinear regression model with a trained thresholdSadehRegressionCountsMediumCommonly used linear regression modelSazonovRegressionCountsMediumLogistic regression model trained on infants from sensor on diaper location For the count-based algorithms, activity counts were calculated by first downsampling the raw acceleration data to 40hz by taking the mean in 25 milli-second windows and then using the open-source Python agcounts package to obtain activity counts in 30 s epochs9. The legacy count-based algorithms implemented were Cole-Kripke8, Oakley10, Sadeh16 and Sazonov11. In addition, Long Short-Term Memory (LSTM) and convolutional neural network (CNN) with sequences of 20, 50, and 100 samples were implemented with the same weights provided in Palotti et al.13. For the deep learning algorithms, the activity counts were combined by taking the 6th root of the sum of the squared 3 axes values, then the values were normalized. Directly using the scaler from the MESA training set resulted in poor performance, likely due to differences in activity counts. A rescoring was implemented on the legacy count-based algorithms that was originally proposed in Webster et al. 1982 and was implemented more recently in Palotti et al.7,13. The rescoring was only applied to the legacy algorithms as it was developed for the legacy algorithms and Palotti et al. 2019 showed that rescoring did not improve deep learning models performance. Original performance statistics from all algorithms are reported in Supplementary Table 4. ## Statistics A confusion matrix was obtained for predictions from each algorithm for each subject compared to the gold standard PSG sleep stages at 30-second epochs. An epoch was true positive (TP) if both the PSG and the algorithm prediction labelled it sleep, an epoch was true negative (TN) if both the PSG and algorithm prediction were wake, an epoch was false positive (FP) if the PSG was wake and the algorithm prediction was sleep, finally, an epoch was false negative (FN) if the PSG was sleep and the algorithm prediction was wake. These equations were used to calculate the following statistics:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Sensitivity = \frac{{True\,Positives}}{{True\,Positives + False\,Negatives}}$$\end{document}Sensitivity=TruePositivesTruePositives+FalseNegatives2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Specificity = \frac{{True\,Negatives}}{{True\,Negatives + False\,Positives}}$$\end{document}Specificity=TrueNegativesTrueNegatives+FalsePositives3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Precision = \frac{{True\,Positives}}{{True\,Positives + False\,Positives}}$$\end{document}Precision=TruePositivesTruePositives+FalsePositives4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F1\,Score = 2*\frac{{Precision\,*\,Sensitivity}}{{Precision + Sensitivity}}$$\end{document}F1Score=2*Precision*SensitivityPrecision+Sensitivitysensitivity, Eq. [ 1], was calculated as the percentage of PSG scored sleep that the algorithm scored correctly. Specificity, Eq. [ 2], was calculated as the percentage of PSG scored wake, that the algorithm scored correctly. Precision, Eq. [ 3], was calculated as the percentage of algorithm detected sleep that was correct according to PSG. F1 score, Eq. [ 4], was calculated as the harmonic mean of sensitivity and precision. Wake after sleep onset (WASO) was calculated by summing the amount of wake within the sleep period. Sleep efficiency (SE) was calculated as the percentage of sleep within the period in which there were PSG annotations. Total sleep time (TST) was calculated as the sum of sleep epoch classifications per night. A Bland-Altman style validation statistical approach was applied to the sleep endpoints WASO, TST and SE because this provides validation statistics that are specific to the sleep endpoints that are commonly used in clinical research22. Mean error (ME) was calculated as the average of the differences between the predicted measure and the PSG derived measure for all patients. Mean error indicates if there is a bias in a model output. Root mean squared error (RMSE) was calculated as the square root of the squared mean error and is useful because positive and negative values cannot cancel each other out. Pearson’s correlation coefficient is calculated as well as the $95\%$ confidence intervals, which is the range for which $95\%$ of the differences between the predicted endpoint and the PSG scored endpoint exist22. ## Reporting summary Further information on research design is available in the Nature Research Reporting Summary linked to this article. ## Supplementary information Supplementary material REPORTING SUMMARY The online version contains supplementary material available at 10.1038/s41746-023-00802-1. ## References 1. Luyster FS, Strollo Jr PJ, Zee PC, Walsh JK. **Sleep: a health imperative**. *Sleep* (2012.0) **35** 727-734. DOI: 10.5665/sleep.1846 2. Tsuno N, Besset A, Ritchie K. **Sleep and depression**. *J. clin. psych.* (2005.0) **66** 1254-1269. DOI: 10.4088/JCP.v66n1008 3. Gottlieb DJ. **Association of usual sleep duration with hypertension: The Sleep Heart Health Study**. *Sleep* (2006.0) **29** 1009-1014. DOI: 10.1093/sleep/29.8.1009 4. Cooper CB, Neufeld EV, Dolezal BA, Martin JL. **Sleep deprivation and obesity in adults: a brief narrative review**. *BMJ Open Sport Exerc Med.* (2018.0) **4** e000392. DOI: 10.1136/bmjsem-2018-000392 5. 5.Iranzo, A. & Santamaria, J. Sleep in neurodegenerative diseases. Sleep Medicine: A Comprehensive Guide to Its Development, Clinical Milestones, and Advances in Treatment. 271–283 (2015). 6. Moore CM, Schmiege SJ, Matthews EE. **Actigraphy and Sleep Diary Measurements in Breast Cancer Survivors: Discrepancy in Selected Sleep Parameters**. *Behav. Sleep Med.* (2015.0) **13** 472-490. DOI: 10.1080/15402002.2014.940108 7. Webster JB, Kripke DF, Messin S, Mullaney DJ, Wyborney G. **An activity-based sleep monitor system for ambulatory use**. *Sleep* (1982.0) **5** 389-399. DOI: 10.1093/sleep/5.4.389 8. Cole RJ, Kripke DF, Gruen W, Mullaney DJ, Gillin JC. **Automatic sleep/wake identification from wrist activity**. *Sleep* (1992.0) **15** 461-469. DOI: 10.1093/sleep/15.5.461 9. Neishabouri A. **Quantification of acceleration as activity counts in ActiGraph wearable**. *Sci. Rep.* (2022.0) **12** 11958. DOI: 10.1038/s41598-022-16003-x 10. 10.Oakley, N. Validation with Polysomnography of the Sleepwatch Sleep/Wake Scoring Algorithm used by the Actiwatch Activity Monitoring System. Technical Report to Mini-Mitter Co., Inc. (1997). 11. Sazonov E, Sazonova N, Schuckers S, Neuman M. **Activity-based sleep-wake identification in infants**. *Physiol. Meas.* (2004.0) **25** 1291. DOI: 10.1088/0967-3334/25/5/018 12. Lutsey PL. **Objectively measured sleep characteristics and prevalence of coronary artery calcification: The Multi-Ethnic Study of Atherosclerosis Sleep study**. *Thorax* (2015.0) **70** 880-887. DOI: 10.1136/thoraxjnl-2015-206871 13. Palotti J. **Benchmark on a large cohort for sleep-wake classification with machine learning techniques**. *NPJ Digit. Med.* (2019.0) **2** 50. DOI: 10.1038/s41746-019-0126-9 14. van Hees VT. **A novel, open access method to assess sleep duration using a wrist-worn accelerometer**. *PLoS One* (2015.0) **10** e0142533. DOI: 10.1371/journal.pone.0142533 15. Sundararajan K. **Sleep classification from wrist-worn accelerometer data using random forests**. *Sci. Rep* (2021.0) **11** 24. DOI: 10.1038/s41598-020-79217-x 16. Sadeh A, Sharkey KM, Carskadon MA. **Activity-based sleep-wake identification: An empirical test of methodological issues**. *Sleep* (1994.0) **17** 201-207. DOI: 10.1093/sleep/17.3.201 17. 17.Berry, R. B., Brooks, R., Gamaldo, C. E., Harding, S. M., Marcus, C. & Vaughn, B. V. The AASM manual for the scoring of sleep and associated events. Rules, Terminology and Technical Specifications. (American Academy of Sleep Medicine, 2012). 18. Lee YJ, Lee JY, Cho JH, Choi JH. **Interrater reliability of sleep stage scoring: a meta-analysis**. *J. Clin. Sleep Med.* (2022.0) **18** 193-202. DOI: 10.5664/jcsm.9538 19. Walch O, Huang Y, Forger D, Goldstein C. **Sleep stage prediction with raw acceleration and photoplethysmography heart rate data derived from a consumer wearable device**. *Sleep* (2019.0) **42** zsz180. DOI: 10.1093/sleep/zsz180 20. 20.Iber, C., Ancoli-Israel, S., Chesson, A. & Quan, S. The AASM manual for the scoring of sleep and associated events: rules, terminology and technical specifications. 1st ed. (American Academy of Sleep Medicine, 2007). 21. Migueles JH, Rowlands AV, Huber F, Sabia S, van Hees VT. **GGIR: A Research Community–Driven Open Source R Package for Generating Physical Activity and Sleep Outcomes From Multi-Day Raw Accelerometer Data**. *J. Meas. Phys. Behav.* (2019.0) **2** 188-196. DOI: 10.1123/jmpb.2018-0063 22. Martin Bland J, Altman DG. **Statistical methods for assessing agreement between two methods of clinical measurement**. *Lancet* (1986.0) **327** 307-310. DOI: 10.1016/S0140-6736(86)90837-8
--- title: Nasal microbiota profiles in shelter dogs with dermatological conditions carrying methicillin-resistant and methicillin-sensitive Staphylococcus species authors: - Sara Horsman - Erika Meler - Deirdre Mikkelsen - John Mallyon - Hong Yao - Ricardo J. Soares Magalhães - Justine S. Gibson journal: Scientific Reports year: 2023 pmcid: PMC10039040 doi: 10.1038/s41598-023-31385-2 license: CC BY 4.0 --- # Nasal microbiota profiles in shelter dogs with dermatological conditions carrying methicillin-resistant and methicillin-sensitive Staphylococcus species ## Abstract Dermatological conditions may be complicated by Staphylococcus spp. infections influencing skin and nasal microbiota. We investigated the associations between the resident nasal microbiota of shelter dogs with and without dermatological conditions carrying methicillin-resistant and -sensitive Staphylococcus spp. Nasal sampling of 16 dogs with and 52 without dermatological conditions were performed upon shelter admission (baseline), and then bi-weekly until discharge (follow-up). All samples were cultured for Staphylococcus spp., while 52 samples underwent microbiota analysis. Two elastic net logistic regression (ENR) models (Model 1—baseline samples; Model 2—follow-up samples) were developed to identify predictive associations between dermatological conditions and the variables: signalment, antimicrobial treatment, and nasal microbial genera. Follow-up nasal samples of dogs with dermatological conditions had decreased microbiota diversity and abundance compared to dogs without dermatological conditions. Our ENR models identified predictive differences in signalment and nasal microbial genera between baseline and follow-up samples. Co-occurrence networks showed nasal microbial genera were more dissimilar when comparing dogs with and without dermatological conditions at follow-up. Overall, this study is the first to investigate Staphylococcus spp. carriage effects on nasal microbial genera in a canine animal shelter population, and ultimately reveals the importance of investigating decolonisation and probiotic therapies for restoring nasal microbiota. ## Introduction Dogs admitted to animal shelters often arrive with dermatological conditions caused by infectious agents including mites and dermatophytes, and allergens associated with fleas, food, and the environment1. Dogs with these conditions often have secondary bacterial infections predominately caused by Staphylococcus spp., particularly Staphylococcus pseudintermedius2,3. Staphylococcus pseudintermedius forms part of the skin and nasal resident microbiota of healthy dogs2,3. As such, it is important to investigate the carriage of methicillin-resistant Staphylococcus (MRS) spp., as infection with this bacterium may further complicate treatment outcomes4. The repeated prescription of antimicrobials and longer treatment duration to treat secondary bacterial infections increases the risk of developing antimicrobial resistance, particularly methicillin resistance5. Previous studies have reported the carriage of methicillin-resistant and -sensitive S. pseudintermedius (MRSP and MSSP, respectively), and methicillin-resistant and -sensitive S. aureus (MRSA and MSSA, respectively), from the nares, mouth, and/or perineum in healthy dogs attending veterinary clinics and in shelter dogs6–10. Studies have also isolated S. pseudintermedius from up to $92\%$ of canine pyoderma cases11–14, with $12.7\%$ to $43.1\%$ being methicillin-resistant12–15. S. pseudintermedius was isolated from 20 to $95\%$ of canine otitis externa cases16–20 and of these, $8.7\%$ to $50\%$ were methicillin-resistant17,19,20. Prior bacterial infections have been identified as a risk factor for MRSP nasal carriage in dogs21. Additionally, nasal MRSA-colonising isolates from children with atopic dermatitis (AD) have been linked to MRSA-infecting isolates in children with concurrent skin and soft tissue infections and AD22. Hence, understanding the role of MRS and methicillin-sensitive Staphylococcus (MSS) spp. nasal carriage in dogs with dermatological conditions is vital in deciding whether decolonisation therapies are required to decrease secondary infection risk23. Few publications to date have investigated the microbiota of dogs with skin allergies24–27. Using 16S rRNA gene amplicon sequencing, it has been reported that the nares of allergic dogs had lower species richness compared to healthy dogs24. The predominant genera in the nares of allergic dogs in the study by Rodrigues Hoffmann, et al. 24 were Streptococcus, Diaphrobacter, and Sphingomonas, whereas, Ralstonia was the most abundant genera in the nares of healthy dogs. Other studies have identified Moraxella, Cardiobacteriaceae, Phyllobacterium, Porphyromonas, Staphylococcus, and *Streptococcus as* the predominant genera in the nares of healthy dogs28,29. Furthermore, studies have reported a decrease in bacterial diversity and an increase in the relative abundance of the *Staphylococcus genus* on the skin of dogs with dermatitis compared to healthy dogs24,25,27. Despite this, *Staphylococcus is* also considered to be the predominant genus on healthy dog’s skin but with varying relative abundance, in addition to Corynebacterium, Kocuria, Macrococcus, Porphyromonas, Propionibacterium, Pseudomonas, and Streptococcus24–27. Interestingly, a previous study has investigated the effects of MRSA and MSSA colonisation on the lesional skin microbiota of humans with AD, and determined that those colonised with MRSA had reduced microbial diversity compared to MSSA colonisers30. This study also identified a decrease in the relative abundance of Streptococcus, Propionibacterium, and Corynebacterium on MRSA-colonised lesional skin30. Since various studies have identified a decrease in the diversity and abundance of the skin and nasal microbiota in dogs and humans with MRS and MSS spp. carriage24,25,30, it is likely that oral or topical probiotics may be useful in restoring the skin and nasal microbiota of dogs with dermatological conditions31. To date, to the best of our knowledge, no studies have investigated whether nasal carriage of MRS and MSS spp. influences the nasal microbiota of dogs with dermatological conditions. Hence, the aims of this study were to: (a) investigate the nasal carriage of MRS and MSS spp.; ( b) determine the risk factors associated with dermatological conditions; (c) explore associations between the resident nasal microbiota and MRS and MSS spp. nasal carriage in dogs with and without dermatological conditions; and (d) identify predictive signalment data and canine nasal microbial genera in dogs with dermatological conditions, in an animal shelter in Brisbane, Queensland, Australia. ## Study population Nares of 70 shelter dogs were sampled ($$n = 186$$ samples). Medical histories were not available for two dogs, and signalment data was not available for one dog. Twenty-four percent ($\frac{16}{68}$) of dogs had dermatological conditions over the sampling period. Fifteen of these dogs were diagnosed with dermatological conditions on admission, and one developed a condition during their stay. Eleven of these 16 dogs had skin conditions only, two had ear conditions only, and three had both skin and ear conditions. Seventy-six percent ($\frac{52}{68}$) of dogs had no dermatological conditions, yet some had other health issues including lameness, vomiting, diarrhoea, renal disease, or kennel cough. A total of 183 nasal samples from the dogs with available medical histories were taken. Of the baseline nasal samples ($$n = 68$$), most dogs ($69.1\%$; $\frac{47}{68}$) were sampled within the first 24 h of arrival. A further $8.8\%$ ($\frac{6}{68}$) were sampled on day two, $14.7\%$ ($\frac{10}{68}$) were sampled between days three to seven from arrival, while $4.4\%$ ($\frac{3}{68}$) were sampled between days eight to 14, and $3\%$ ($\frac{2}{68}$) were sampled at days 18 and 21 (baseline samples after being in the shelter for > 24 h). Although, 23 of the 68 baseline samples corresponded to a unique dog that only stayed in the shelter for 1 day (one baseline sample taken per dog). There were 115 follow-up samples from 45 dogs, where 28 dogs stayed four or more days totalling one baseline and two or more follow-up samples taken per dog, and 17 dogs stayed for two days totalling one baseline and one follow-up sample only per dog. A higher proportion of nasal samples were taken from dogs with no dermatological conditions ($78.7\%$; $\frac{144}{183}$ samples). For dogs with dermatological conditions, $51.3\%$ ($\frac{20}{39}$) of nasal samples were from dogs with skin conditions only, $30.8\%$ ($\frac{12}{39}$) with ear conditions only, and $17.9\%$ ($\frac{7}{39}$) with both skin and ear conditions. Forty-four percent ($\frac{7}{16}$) of dogs with dermatological conditions were treated with topical antimicrobials only ($$n = 23$$ nasal samples). Two percent ($\frac{1}{52}$) of dogs without dermatological conditions were treated using topical antimicrobials (n = one sample), $13.5\%$ ($\frac{5}{52}$) were treated using oral antimicrobials ($$n = 12$$ samples), and $1\%$ ($\frac{1}{52}$) were treated using parenteral antimicrobials as a subcutaneous injection (n = one sample). Refer to Tables S10 and S11 of the Supplementary Results for the list of topical and systemic antimicrobials used. ## Methicillin-resistant and methicillin-sensitive Staphylococcus spp. nasal carriage At admission (baseline), $73\%$ ($\frac{11}{15}$) of dogs with dermatological conditions carried Staphylococcus spp. and upon discharge, $100\%$ ($\frac{16}{16}$) carried at least one Staphylococcus spp. Sixty percent ($\frac{6}{10}$) of these dogs that had two or more samples carried multiple Staphylococcus spp. in their nares. Seventy-three percent ($\frac{38}{52}$) of dogs without dermatological conditions carried Staphylococcus spp. and upon discharge, $86.5\%$ ($\frac{45}{52}$) carried at least one Staphylococcus spp. Forty-three percent ($\frac{15}{35}$) of these dogs that had two or more samples carried multiple Staphylococcus spp. in their nares. Overall, irrespective of dermatological conditions, the shelter dogs that stayed two or more days had a higher rate of staphylococci nasal carriage ($75.5\%$; $\frac{34}{45}$) upon discharge. Methicillin-resistant Staphylococcus spp. was present in the nares of $26.7\%$ ($\frac{4}{15}$) of the baseline samples from dogs with dermatological conditions at admission, and throughout the sampling period, $20.5\%$ ($\frac{8}{39}$) of nasal samples were MRS spp. positive. Of the dogs without dermatological conditions, $11.5\%$ ($\frac{6}{52}$) had MRS spp. in their nares at admission, with $13.2\%$ ($\frac{19}{144}$) of nasal samples being positive for MRS spp. over the sampling period. Detailed bacterial culture and antimicrobial susceptibility testing results for all 186 nasal samples are presented in Tables S1, S2, and S3 of the Supplementary Results. ## Risk factors associated with dogs having dermatological conditions Our univariable models indicated that the variables that should be considered (i.e. p ≤ 0.20) in the full multivariable model included sex, dog population, the original location of the dogs before entering any shelter, the dog’s previous shelter location (if any), length of stay at the sampling shelter, the number of days the dogs were in the shelter prior to the baseline swab being taken, antimicrobial usage, nasal carriage and sampling location within the shelter (Table S4 of the Supplementary Results). Our final multivariable model showed that dogs with dermatological conditions had higher odds of being female [Odds Ratio (OR): 3.73 ($95\%$ CI: 2.16–6.44); p ≤ 0.001], being present in the shelter two to seven days prior to the baseline swab being taken [OR: 5.02 ($95\%$ CI: 1.71–14.75); $$p \leq 0.003$$] (Table 1), being positive for MRSA carriage [OR: 11.54 ($95\%$ CI: 2.51–53.07); $$p \leq 0.002$$], and being treated with antimicrobials [OR: 8.92 ($95\%$ CI: 2.18–36.55); $$p \leq 0.002$$] compared to dogs without these conditions. Our multivariable results indicated that dogs with dermatological conditions had lower odds of being from the owner surrendered dog population compared to dogs without dermatological conditions [OR: 0.19 ($95\%$ CI: 0.05–0.74); $$p \leq 0.016$$].Table 1Multivariable analysis of the risk factors associated with shelter dogs with dermatological conditions. VariableMultivariable analysisVariableMultivariable analysisnOdds ratio ($95\%$ CI)p-valuenOdds ratio ($95\%$ CI)p-valueSexLength of stay (days) Male69Reference0–767Reference Female1143.73 (2.16–6.44) ≤ 0.0018–14230.55 (0.22–1.36)0.198Dog population15–21450.23 (0.04–1.32)0.101 Stray dogs90Reference ≥ 22481.25 (0.21–7.33)0.804 Owner surrendered360.19 (0.05–0.74)0.016Days in shelter prior to baseline swab being taken Humane officer seized572.82 (0.67–11.88)0.159 ≤ 1119ReferenceOriginal location of dogs 2–7525.02 (1.71–14.75)0.003 Brisbane33Reference > 7122.69 (0.39–18.35)0.311 North of Brisbane410.76 (0.07–8.31)0.825Previous shelter location South of Brisbane251.22 (0.24–6.17)0.239 Shelter one (sampling shelter)120Reference West of Brisbane840.46 (0.04–4.83)0.519 Shelter two280.32 (0.01–7.49)0.482 Shelter three350.84 (0.23–3.03)0.793Antimicrobial usageNasal carriage No146Reference Culture negative for all bacteria16Reference Yes378.92 (2.18–36.55)0.002 Staphylococci culture negative261.00 (0.07–12.99)0.997Sampling location within the sheltera MRSP102.70 (0.17–42.72)0.480 Veterinary clinic89Reference MSSP741.95 (0.70–5.45)0.201 Veterinary clinic dog holding191.09 (0.35–3.43)0.877 MRSA511.54 (2.51–53.07)0.002 Shelters’ dog holdings630.74 (0.44–1.26)0.272 MSSA160.19 (0.02–1.90)0.158 Shelter’s adoption centreb12NA MR-CoNS121.77 (0.11–12.98)0.894 MS-CoNS240.81 (0.11–5.86)0.835For all variables in the multivariable model, there were 183 nasal samples included.n the number of individual samples per variable. MRSP methicillin-resistant S. pseudintermedius, MSSP methicillin-sensitive S. pseudintermedius, MRSA methicillin-resistant S. aureus, MSSA methicillin-sensitive S. aureus, MR-CoNS methicillin-resistant coagulase-negative staphylococci, MS-CoNS methicillin-sensitive coagulase-negative staphylococci.aThe shelter includes a veterinary clinic, a veterinary clinic dog holding for dogs still requiring veterinary care, dog holdings, and an adoption centre.bThe shelters’ adoption centre was omitted from the multivariable model as it predicts failure perfectly (no dogs with dermatological conditions located at the adoption centre). N the number of total samples per variable, n the number of individual samples per variable, NA not applicable. ## Nasal microbiota study population Of the 52 samples, 28 dogs had one baseline nasal sample only, seven dogs had one baseline and one follow-up sample each, one dog had one baseline and two follow-up samples, one dog had one follow-up sample, and three dogs had only two follow-up samples each (no baseline samples). Thirteen of these 52 samples were from ten dogs with dermatological conditions (n = six baseline and seven follow-up nasal samples). Thirty-nine of these 52 samples were from 30 dogs without those conditions ($$n = 28$$ baseline and 11 follow-up samples). Four dogs with dermatological conditions (n = one baseline and five follow-up samples) corresponded with a topical antimicrobial treatment. Two dogs without dermatological conditions were treated with oral antimicrobials (n = two baseline and one follow-up sample). A total of 3,613,386 sequences were amplified from all 52 nasal samples, with $86.7\%$ (3,132,$\frac{642}{3}$,613,386 sequences) passing quality control. The minimum reads per sample were 9,461, while the maximum reads per sample were 206,747. These sequence reads were then taxonomically classified across 895 amplicon sequence variants (ASVs), with 562 being identified at the genus level. At the genus level there were 82 uncultured genera and 18 unknown/unassigned genera, respectively. The nasal carriage status of dogs with and without dermatological conditions are presented in Table S5 of the Supplementary Results. ## Relative abundance, alpha diversity, and core microbiota In all nasal microbiota samples ($$n = 52$$), Psychrobacter, Moraxella, Massilia, Elizabethkingia, and *Streptococcus were* the top five genera, followed by the genus Staphylococcus (Fig. 1a). In the follow-up samples of dogs without dermatological conditions, Psychrobacter, Moraxella, Massilia, and *Staphylococcus were* the dominant genera (Fig. 1a). Staphylococcus relative abundance was low for dogs with dermatological conditions for both sample timings, while *Streptococcus relative* abundance was higher (Fig. 1a).Figure 1Microbiota analysis of nasal samples from dogs with (yes) and without (no) dermatological conditions displaying (a) relative abundance (%) of the top 20 genera for the baseline and follow-up samples; (b) alpha diversity plots of observed richness, Chao1, Shannon Index, and Simpson’s Index for the baseline and follow-up samples using Wilcoxon rank-sum test significance [$p \leq 0.05$ = NS (not significant), *$p \leq 0.05$, **$p \leq 0.01$]; (c) core microbiota at genus level from nasal samples of dogs with and without dermatological conditions; (d) core microbiota at genus level of nasal samples from dogs with dermatological conditions that were prescribed antimicrobials (yes—antimicrobial usage) and not prescribed antimicrobials (yes—no antimicrobial usage) and from dogs without those conditions prescribed antimicrobials (no—antimicrobial usage) and not prescribed antimicrobials (no—no antimicrobial usage). There were no significant differences ($p \leq 0.05$) in the alpha diversity for dogs with dermatological conditions between their baseline and follow-up samples (Fig. 1b). For dogs without dermatological conditions, the observed richness of the follow-up nasal samples was statistically higher ($p \leq 0.05$), compared to the baseline samples (Fig. 1b). Also, between the baseline and follow-up samples, Shannon’s and Simpson’s diversity indices differed significantly ($p \leq 0.01$ and $p \leq 0.05$, respectively) (Fig. 1b). Additionally, when comparing the follow-up samples of dogs with and without dermatological conditions, the differences in both Shannon’s and Simpson’s diversity indices were statistically significant ($p \leq 0.05$) (Fig. 1b). There was also no significant difference observed in alpha diversities between dog groups for nasal carriage (MRS spp., MSS spp., and culture negative for MRS and MSS spp.), and antimicrobial usage (Figs. S1 and S2 of the Supplementary Results). When focusing on sequence reads with ≥ $1\%$ relative abundance, only 44 genera were identified and used for core microbiota analyses to determine shared and unique genera, irrespective of the sample timing. When comparing the canine nasal microbial genera in dogs with and without dermatological conditions, seven genera including Acidibacter, Bacteroides, Chloroplast, Cladosporium, Faecalibacterium, Serratia, and an unclassified Solirubrobacterales referred to as 67–14 were unique to the nares of dogs with dermatological conditions (Fig. 1c). *Sixteen* genera were unique to dogs without dermatological conditions, while 20 genera were shared (core genera) between the dog groups (Fig. 1c). Chloroplast and Cladosporium were uniquely identified in the nares of dogs with dermatological conditions treated with antimicrobials (Fig. 1d), whilst Acidibacter, Bacteroides, Faecalibacterium, Serratia, and an unclassified Solirubrobacterales were unique to dogs with dermatological conditions not treated with antimicrobials (Fig. 1d). Additionally, an UpSet plot was used to determine the core microbiota present for nasal carriage of dogs with and without dermatological conditions (Fig. S3 of the Supplementary Results). The full list of shared and unique genera for dogs with and without dermatological conditions for the core microbiota analyses are presented in Tables S6, S7, and S8 of the Supplementary Results. ## Beta diversity Principal coordinate analysis (PCoA) based on Bray–Curtis dissimilarity metrics depicted $52.3\%$ of the microbial communities between the nasal samples from dogs with and without dermatological conditions, and the signalment variables (Fig. 2). No clustering was observed based on condition alone, in addition to no clustering for sex, age, breed size, neuter status and nasal carriage (Fig. 2a–f), and antimicrobial usage (Fig. S2b of the Supplementary Results). However, without accounting for the condition, the microbiota of dogs which cultured negative for all bacteria in the nares on the selective agar plates were less dissimilar, clustering together in the lowest horizontal quadrant of the PCoA plot (Fig. 2f).Figure 2PCoA plots with Bray–Curtis displaying the amplicon sequence variants (ASVs) from the nasal samples ($$n = 52$$) from (a) dogs with (yes) and without (no) dermatological conditions and comparing that to (b) sex; (c) age; (d) breed size; (e) neuter status; and (f) nasal carriage represented by the corresponding symbols per plot. ## Associations between canine dermatological conditions, signalment, and nasal microbiota diversity From a total of 580 variables, 30 of these were selected using the information value results. For Model 1, the 34 baseline nasal microbiota samples included $17.6\%$ ($\frac{6}{34}$) of samples from dogs with dermatological conditions and $82.4\%$ ($\frac{28}{34}$) of samples from dogs without those conditions. For Model 2, the 18 follow-up nasal microbiota samples included $38.9\%$ ($\frac{7}{18}$) of samples from dogs with dermatological conditions and $61.1\%$ ($\frac{11}{18}$) of samples from dogs without those conditions. Both models were highly accurate in being able to classify variables as predictors of dermatological conditions in shelter dogs (Model 1 and Model 2: AUC = 1; sensitivity and specificity = 1). The coefficient values for both models are presented in Table S9, in Supplementary Results. In Model 1, Faecalibaculum, Gracilibacteria, and Defluviitaleaceaea_UCD-011 were identified as the most predictive genera in the nasal microbiota (Fig. 3a). Sex (females) and MRSP nasal carriage were the only signalment variables included for Model 1 (Fig. 3a). For Model 2, the variables considered to be most predictive of dermatological conditions in shelter dogs’ nares were Acinetobacter and *Arachnida* genera, followed by antimicrobial usage (Fig. 3b). Topical antimicrobial usage corresponded to $16.6\%$ ($\frac{1}{6}$) of nasal microbiota samples from dogs with dermatological conditions in Model 1. For Model 2, $71.4\%$ ($\frac{5}{7}$) of nasal microbiota samples from dogs with dermatological conditions corresponded to topical antimicrobial usage (Refer to Tables S10 and S11 of the Supplementary Results for the list of antimicrobials used). Repeating the selected variables from the final models (represented as error bars in Fig. 3a,b), revealed that the variable importance scores differed highly depending on each individual repeated model. Figure 3Barplots representing the top 20 most predictive variables identified by the elastic net logistic regression model for dogs with dermatological conditions. ( a) Model 1 used only the baseline nasal samples and signalment data; and (b) Model 2 used only the follow-up nasal samples and signalment data. The whiskers (error bars) represent the standard deviations from training the selected features (variables) from the highest AUC models which were repeated 10 times (AUC ≥ 0.7). The coloured bars represent the different phyla. Allorhizobuim–Neorhizobium–Pararhizobium–Rhizobium was abbreviated to Allo-Neo-Para-Rhizobium in this graph. MRSP methicillin-resistant S. pseudintermedius. ## Co-occurrence networks based on elastic net logistic regression model outcomes Using nasal microbiota data collected at baseline (Model 1) and follow-up (Model 2), we were able to identify a total of 25 and 20 genera, respectively. Using those genera, our results indicate that the nasal microbiota at baseline (Model 1) were similar for dogs with and without dermatological conditions (Fig. 4a). At follow-up (Model 2), the nasal microbiota were dissimilar for dogs with and without dermatological conditions (Fig. 5a).Figure 4Co-occurrence network analysis using the genera with coefficient values from the elastic net logistic regression Model 1 (baseline nasal microbiota samples) using Bray–Curtis dissimilarly and maximum distance of 0.8 displaying the connections between (a) baseline nasal microbiota samples of dogs with (yes) and without (no) dermatological conditions and (b) the genera. Figure 5Co-occurrence network analysis using the genera with coefficient values from the elastic net logistic regression Model 2 (follow-up nasal microbiota samples) using Bray–Curtis dissimilarly and maximum distance of 0.8 displaying the connections between (a) follow-up nasal microbiota samples of dogs with (yes) and without (no) dermatological conditions and (b) the genera. Allorhizobuim–Neorhizobium–Pararhizobium–Rhizobium was abbreviated to Allo-Neo-Para-Rhizobium in this graph. When investigating the co-occurrence of the nasal microbiota at the genera level (not dog group), of the 25 genera identified in Model 1, only 17 remained in the network, with eight of those being positively correlated to dogs with dermatological conditions (Fig. 4b). There were connections between both the positively and negatively associated genera, based on the coefficient output from Model 1 (Fig. 4b). For instance, Bacillus (negative coefficient) co-occurred with Nocardioides, and Pseudonocardia (both positive coefficients), within its own cluster (Fig. 4b). A total of 12 out of 20 genera from Model 2 were retained, with eight correlating to dogs with dermatological conditions (Fig. 5b). For the follow-up samples, all of the positively associated genera from Model 2 formed one cluster, with the four negatively associated genera forming two separate clusters (Fig. 5b). ## Discussion In this study, we compared the nasal microbiota of shelter dogs carrying MRS and MSS spp., with and without dermatological conditions, using conventional microbiota bioinformatics analyses and machine learning techniques. Significant differences in nasal microbiota diversity and abundance were only observed based on dog group and sample timing using standard microbiota analyses. Utilising elastic net logistic regression, we identified nasal microbial genera of shelter dogs that were predictive of dermatological conditions at baseline and follow-up, whilst accounting for signalment, nasal carriage data, and antimicrobial usage. Our study shows that upon admission and throughout the study period when accounting for all nasal samples, dogs with dermatological conditions had a higher isolation of MRS spp. in the nares compared to dogs without those conditions. The high isolation of MRS spp. in dogs with dermatological conditions is concerning due to the potential for the carriage of nasal MRS spp. to cause secondary bacterial skin infections. This has been identified in children with AD, whereby, the nasal MRSA-colonising isolates were clonally related to the MRSA-infecting isolates22. Additionally, as dogs with dermatological conditions often have secondary bacterial infections2, decolonisation of the MRS spp. may be beneficial in reducing the risk of these bacteria, particularly MRSP, in causing skin and ear infections in shelter dogs. When investigating the staphylococci nasal carriage as risk factors for dogs with dermatological conditions using all samples, only MRSA was significant. This was surprising as S. pseudintermedius is the most frequently isolated bacterium from dogs with skin diseases18,32. However, as there was a large $95\%$ confidence interval, and the result was only represented by two dogs, this finding should be interpreted cautiously. Dogs with dermatological conditions in this study also had higher odds of being female compared to dogs without those conditions. This is contradictory to other studies investigating the risk factors of dogs with AD and otitis externa, which reported either no differences between sexes or that males had a higher risk33–37. The difference in target populations between our study (shelter animals) and other studies (veterinary clinics or insurance databases)33–37, may partially explain the sex differences. The findings that antimicrobials were identified as a risk factor for dogs with dermatological conditions were expected, in that topical antimicrobials such as shampoos and ointments are often prescribed for dogs with pyoderma and otitis externa38,39. The significantly lower odds of dogs with dermatological conditions being from an “owner surrendered dog population” may be due to the care and treatment of dogs before entering the shelter. The higher relative abundance of Staphylococcus spp. in the nares of dogs without dermatological conditions was contradictory to other studies investigating the nasal and skin microbiota of dogs with allergic or AD24,25,27. Our results echo those reported by Rodrigues Hoffmann, et al. 24, who showed that the nares of dogs with allergic dermatitis were predominantly colonised by Streptococcus spp. Additionally, the significant differences in alpha diversities between baseline and follow-up nasal samples from dogs without dermatological conditions suggests that length of stay is likely to influence the diversity and abundance of dog’s nasal microbiota. This was also evident when comparing follow-up samples of dogs with and without dermatological conditions. Regarding the microbiota analyses of this study, we are the first to our knowledge to report core microbiota analyses while investigating Staphylococcus spp. nasal carriage and antimicrobial therapy effect on the nasal microbiota. To date, only two studies have examined the skin microbiota in dogs with AD before and after antimicrobial therapy25,27, where core microbiota analyses were not reported. However, our beta diversity plots which identified no clustering between the nasal microbiota samples from dogs with and without dermatological conditions (Fig. 2), were similar to a previously reported finding by Rodrigues Hoffmann, et al. 24, who investigated the microbiota of allergic and healthy dogs. As our alpha diversity measures demonstrated a significant difference between dogs with and without dermatological conditions for the follow-up samples, we adopted an ENR machine learning approach to further explore this association40–43. This allowed for the identification of predictive nasal microbiota genera of dermatological conditions accounting for important multicollinear variables40–43. The ENR approach was also selected for this investigation due to its ability to effectively analyse data with a small sample size and a large number of variables commonly represented in microbial data44. The ENR models identified predictive genera of disease, and along with the co-occurrence networks, demonstrated that there was a relationship between multiple canine nasal microbial genera in dogs with dermatological conditions. Furthermore, nasal carriage at the species level was unable to be investigated for the alpha diversities using conventional microbiota analyses due to the low sample size. However, when using ENR, MRSP was identified as a predictive signalment variable, along with female dogs. This is an important finding as S. pseudintermedius causes both pyoderma and otitis externa, and MRSP is becoming more common in pyoderma cases45–47. Additionally, this machine learning analyses shows that there were changes in predictive signalment data and microbial genera in the nasal microbiota between baseline (Model 1) and follow-up (Model 2) samples, in that we identified differences in signalment and predictive genera and genera ranking in the two ENR models. The co-occurrence networks indicated that the nasal microbiota of the dog groups was more similar for the baseline samples. This is despite the dogs originally being from different geographical locations and having different medical histories upon arrival. The changes in microbial genera identified in the ENR models were also indicated in the co-occurrence network for the follow-up samples suggesting that factors within the shelter altered the microbiota between the dog groups. The co-occurrence networks further demonstrated the dissimilarity of the nasal microbial genera, as the connections between the genera were based on whether the genera positively or negatively predicted dermatological conditions for the follow-up samples. Due to this dissimilarity and connections between the genera, our findings for the follow-up samples suggest antimicrobial usage is the only non-microbial predictor in ENR Model 2, highlighting the significance of antimicrobial usage in dogs with dermatological conditions as a factor influencing the nasal microbiota in shelter dogs. Overall, combining signalment, MRS and MSS spp. nasal carriage, and antimicrobial treatment with nasal microbial data in machine learning aids in understanding changes in the importance of those variables in dogs with dermatological conditions. It should be noted that our findings may be influenced by the fact that only 28 dogs were sampled for ≥ 4 days, a larger longitudinal study was not possible due to dogs being moved through the shelter or being transferred to another shelter faster than expected. Additionally, some dogs were also unable to be resampled after the baseline sample was taken due to behavioural and/or medical concerns, thus limiting the opportunities for more follow-up samples to be collected. Furthermore, due to one swab being taken per dog from both nostrils, only 52 nasal samples were submitted for 16S rRNA gene amplicon sequencing, due to DNA quantities. Only one sample had less than the minimum sequence reads per sample threshold of 10,000 (9,461 reads), and thus it is unlikely that the bioinformatics analyses would have been biased. Future studies should aim to collect two nasal swabs per dog each sampling time which could be pooled to increase DNA yield48, however, this was beyond the scope of our ethics agreement. Also, as this study was conducted at an animal shelter undertaking their normal routines, for ethical reasons it was not feasible to control for antimicrobial usage in the sampled dogs. Yet after accounting for antimicrobial usage, ENR Model 2 demonstrated the importance of this variable for the nasal microbiota of dogs with dermatological conditions and changes in predictive genera. Thus, when interpreting these results, it is important to identify the potential effect of antimicrobials on the nasal microbiota, in addition to canine dermatological conditions. Lastly, $69.1\%$ ($\frac{47}{68}$) of dogs were sampled within the first 24 h of arrival at the shelter, while this was not the case for $30.9\%$ ($\frac{21}{68}$) of dogs. This was due to the busyness of the shelter. Future studies could further investigate the relationship between MRS and MSS spp. nasal and skin carriage in dogs with allergic dermatitis or AD whilst accounting for disease severity in both animal shelters and veterinary clinics and shelter/practice-based data. Indeed, two human-based studies have observed that persistent nasal S. aureus carriers experienced more severe AD49,50. Additionally, future studies could use machine learning techniques like ENR models to investigate predictive genera of dogs with skin and ear conditions from their corresponding microbiotas, to determine if there are any differences between body sites or condition types. This analysis also has the potential to incorporate results from gene marker sequencing or whole-genome sequencing of cultured MRS and MSS spp. isolates to better understand the relationship between the microbiota and the organisms’ genetics. By using ENR, significant associations between MRSP nasal carriage and dogs with dermatological conditions were revealed, whilst accounting for genera in the nares at baseline. This was despite our overall results showing no association between MRS and MSS spp. nasal carriage and microbiota abundance and diversity, using standard microbiota bioinformatics analyses. Additionally, due to the continual isolation of MRS spp. throughout the dog’s time at the shelter, our study highlights the importance of determining if decolonisation therapies are necessary to reduce the infection risk of dogs with dermatological conditions. The ENR models not only identified similar signalment risk factors indicating their importance for dermatological conditions in shelter dogs such as antimicrobial usage but also highlighted changes in predictive genera between the baseline and follow-up nasal microbiota samples. Lastly, as our study showed that the follow-up nasal microbiota samples were statistically different between dog groups, indicating lowered diversity and abundance for dogs with dermatological conditions, these dogs may benefit from the use of probiotic treatments to restore the nasal microbiota. The clinical relevance of such an approach deserves further investigation. ## Ethics statement This animal study was approved by the Production and Companion Animals ethics committee, School of Veterinary Science, The University of Queensland (The University of Queensland Animal Ethics SVS/$\frac{487}{15}$/KIBBLE) and was performed in accordance with all relevant guidelines and regulations. The methods described in the current study and reported results were compliant with ARRIVE guidelines. Consent was given by the animal shelter to sample the animals housed within the shelter. ## Animal sample collection Nasal swabs were taken from August 13th to November 15th, 2019, using convenience sampling of the shelter dogs for the purpose of identifying MRS and MSS spp. carriage and characterising the resident microbiota. A baseline sample was taken ideally within 24 h of arrival at the animal shelter at either a veterinary check-up or desexing. However, if this did not happen, the length of time since admission was recorded and the animal was still included in the study. Follow-up samples were taken twice a week until discharge (e.g., adoption, foster, moved to another animal shelter, or humane euthanasia). Both nares of the dogs were swabbed using one ‘Eswab’ swab (481CE; Copan Diagnostics Inc., California, USA), per sampling timepoint by inserting the swab into the nostril and rotating carefully. Gloves were changed between dogs to prevent cross-contamination during sampling. Thereafter, swabs were aseptically snapped into liquid *Amies medium* (1 mL) (Copan Diagnostics Inc., California, USA). The baseline samples were stored (4 °C; for ≤ 48 h) at the shelter and collected on the same day as all follow-up samples. Samples were transported (4 °C) to the laboratory for processing. Bacterial isolation was conducted within the 48 h of collection, whereas, identification of nasal MRS and MSS spp., antimicrobial susceptibility testing, and identification of the mecA gene which confers methicillin resistance in Staphylococcus spp. were conducted in batches (refer to Supplementary Methods). ## Data collection Medical history and signalment data were retrospectively extracted from the animal shelter’s database for the sampling period. Dogs were grouped into whether they had or were free of a dermatological condition. For dogs with veterinary medical records that included terms such as otitis externa (bacterial and yeast), pruritus, interdigital dermatitis, chronic dermatitis, flea allergic dermatitis, erythema, lichenification, hyperpigmentation, excoriation, alopecia, papules, and sarcoptic mange mite, were identified as having a dermatological condition. Information regarding antimicrobial usage was collected and identified by manually reviewing all veterinary consultation notes. Veterinary notes were correlated to the date that the swabs were taken. Refer to Supplementary Methods for the full list of search terms. For each dog, the signalment data included: estimations of birth date or age unless available using microchip data, sex, breed, neuter status, the date the dogs entered and left the animal shelter, and the dog population (stray, owner surrendered, and humane officer seized/surrendered). The location of where the dog originated from and whether the dog was located at a different animal shelter prior to being sampled were also included. ## DNA extraction for microbiota analysis Nasal samples were processed for microbiota analyses by thawing at 4 °C and centrifuging (13,500×g; 4 °C; 5 min) to pellet the cells. Pellets were washed with 1 × phosphate buffered saline (PBS; 1 mL) and stored (−20 °C) until the DNA was extracted as described by Yao, et al. 51 using 600 μL lysis buffer (50 mM Tris–HCl at pH 8.0, $4\%$ sodium dodecyl sulphate, 500 mM NaCl, 50 mM EDTA). Total genomic DNA (gDNA) was extracted using the Maxwell®16 Instrument (Promega, Wisconsin, USA) and the Maxwell® 16 SEV Cell DNA Purification Kit (AS1020, Promega, Wisconsin, USA), as per the manufacturer’s instructions. The DNA quality and concentration was determined using the NanoDrop™ 8000 Spectrophotometer (Thermo Fisher Scientific, Massachusetts, USA), and samples with low DNA concentrations underwent ethanol precipitation. Only samples with a DNA concentration of ≥ 1.88 µg/µL were submitted for 16S rRNA gene amplicon sequencing. ## Microbiota sequencing and bioinformatics analysis 16S rRNA gene amplicon sequencing of 52 of the 183 nasal samples were selected based on DNA quantity and was carried out by the Australian Centre for Ecogenomics (ACE; the University of Queensland, Queensland, Australia), using the Illumina MiSeq Platform (Illumina, California, USA), where the V6 to V8 variable region was targeted using the primers 926F and 1392wR52. *The* generated sequence reads were imported into QIIME 253 and denoised with DADA254. The SILVA rRNA database55 was used for the taxonomic classification of representative sequences. In R statistical software v4.1.2, the data was rarefied to an analysable 9,445 reads per sample and alpha diversity parameters (observed richness, Shannon’s Index, Chao1, and Simpson’s Index) were performed with respect to the dog groups’ nasal samples and sample timing, nasal carriage, and antimicrobial usage. Wilcoxon rank-sum test was carried out for each alpha diversity parameter to determine whether the results were significant. Principal coordinates analysis (PCoA; beta diversity) was conducted to determine the microbial diversity between the dog groups’ nasal samples regardless of sample timing. The relative abundances of the top 20 genera were identified and then visualised as a bar chart using Excel56 with respect to the dog groups’ nasal samples and sample timing. Using Venn diagrams57 and an UpSet plot58, the core microbiota at the genus level were displayed for the dog groups, including only genera with a relative abundance of ≥ $1\%$. For the full list of R packages, refer to Supplementary Methods. ## Risk factor analysis Using Stata v17.0 (Stata Corporation, Texas, USA), a Bernoulli logistic regression model was used to identify risk factors associated with dogs with dermatological conditions ($$n = 183$$; the outcome of interest) and the variables of interest at the shelter whilst adjusting for the resampling of the dogs. For the univariable model, a cut-off overall p-value of ≤ 0.20 per variable was considered significant and were retained in the multivariable model. To identify confounders, a manual backward stepwise variable selection procedure was conducted. If a removed variable had a ≥ $25\%$ change on any other variables’ coefficient, then that variable was retained in the model as a confounder. A p-value of 0.05 in the multivariable analysis was considered significant. The smallest estimate of the Akaike information criterion (AIC) was used to determine the final multivariable model. ## Predictive modelling Elastic net logistic regression (ENR) models were conducted to investigate the associations between signalment, staphylococci nasal carriage, antimicrobial treatment, and the relative abundances of the nasal microbiota identified at the genus level ($$n = 580$$ predictor variables) in shelters dogs with dermatological conditions ($$n = 52$$ nasal samples; the outcome of interest). Two models were run using only the baseline nasal samples ($$n = 34$$ nasal samples; Model 1) and only the follow-up nasal samples ($$n = 18$$ nasal samples; Model 2). For both ENR models, the variables included sex, neuter status, breed size, age, nasal carriage, and antimicrobial usage as dummy variables (if required), along with the relative abundances of 562 genera, totalling 580 variables. A $\frac{70}{30}$ training/testing dataset split was used for both models. The training dataset’s predictor variables were evaluated using information value59 to select variables to include in a reduced training and testing model. The selected variables were then trained with a repeated ten-fold cross validation (CV) with five repeats and underwent pre-processing to further normalise the data using “nzv”, “centre” and “scale”60. The accuracy of the model was determined using the reduced testing dataset using the area under the (receiver operation characteristics) curve (AUC) for Model 1 and Model 2, separately. Models with an AUC of ≥ 0.7 were classified as acceptable models61. To identify the top 20 predictors of canine dermatological conditions, the scaled variable importance scores which ranked the absolute values of the coefficients of the selected variables from the training model were visualised using barplots colour-coded at phylum level. The final Model 1 and Model 2 selected variables were repeated 10 times by shuffling the dataset each time to calculate the standard deviation from the variable importance scores, represented as error bars in the barplots. Only the repeated models with an AUC ≥ 0.7 were included. The analysis was performed using the caret62, information63, information value59, metrics64, and glmnet65 packages in R statistical software v4.1.2. ## Co-occurrence network analysis based on the predictive modelling outcomes To better understand the associations of genera with coefficient values identified in the final ENR Model 1 and Model 2 from the nasal samples of dogs with and without dermatological conditions, co-occurrence networks were created. Two networks were created per model using Bray–Curtis dissimilarity and a maximum distance of 0.8. The first network displayed the nasal microbiota samples associated with dogs with or without dermatological conditions, connected based on the genera from Model 1 and Model 2, separately. The second network displayed the connections between the genera using samples from dogs with and without dermatological conditions from Model 1 and Model 2. All networks were created in R statistical software v4.1.2 using igraph version 1.3.266 and phyloseq67. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31385-2. ## References 1. Newbury S, Moriello KA. **Skin diseases of animals in shelters: Triage strategy and treatment recommendations for common diseases**. *Vet. Clin. N. Am. Small Anim.* (2006.0) **36** 59-88. DOI: 10.1016/j.cvsm.2005.09.007 2. DeBoer DJ, Marsella R. **The ACVD task force on canine atopic dermatitis (XII): The relationship of cutaneous infections to the pathogenesis and clinical course of canine atopic dermatitis**. *Vet. Immunol. Immunopathol.* (2001.0) **81** 239-249. DOI: 10.1016/S0165-2427(01)00345-2 3. Hillier A. **Guidelines for the diagnosis and antimicrobial therapy of canine superficial bacterial folliculitis (Antimicrobial Guidelines Working Group of the International Society for Companion Animal Infectious Diseases)**. *Vet. Dermatol.* (2014.0) **25** 163-e143. DOI: 10.1111/vde.12118 4. János D. **Carriage of multidrug resistance staphylococci in shelter dogs in Timisoara, Romania**. *Antibiotics* (2021.0) **10** 801. DOI: 10.3390/antibiotics10070801 5. Zur G, Gurevich B, Elad D. **Prior antimicrobial use as a risk factor for resistance in selected**. *Vet. Dermatol.* (2016.0) **27** 468-e125. DOI: 10.1111/vde.12382 6. Kjellman EE, Slettemeås JS, Small H, Sunde M. **Methicillin-resistant**. *MicrobiologyOpen* (2015.0) **4** 857-866. DOI: 10.1002/mbo3.258 7. Boost MV, O'Donoghue MM, James A. **Prevalence of**. *Epidemiol. Infect.* (2008.0) **136** 953-964. DOI: 10.1017/s0950268807009326 8. Gingrich EN, Kurt T, Hyatt DR, Lappin MR, Ruch-Gallie R. **Prevalence of methicillin-resistant staphylococci in northern Colorado shelter animals**. *J. Vet. Diagn. Invest.* (2011.0) **23** 947-950. DOI: 10.1177/1040638711407301 9. Machado AB, Machado MFR, Picoli SU. **An investigation of methicillin-resistant**. *Rev. Bras. Saude Prod. Anim.* (2017.0) **18** 542-548. DOI: 10.1590/S1519-99402017000400005 10. Huang TM, Chou CC. **Methicillin-sensitive and methicillin-resistant**. *J. Appl. Microbiol.* (2019.0) **126** 1899-1909. DOI: 10.1111/jam.14266 11. Huerta B. **Risk factors associated with the antimicrobial resistance of staphylococci in canine pyoderma**. *Vet. Microbiol.* (2011.0) **150** 302-308. DOI: 10.1016/j.vetmic.2011.02.002 12. Yoo JH, Yoon JW, Lee SY, Park HM. **High prevalence of fluoroquinolone- and methicillin-resistant**. *J. Microbiol. Biotechnol.* (2010.0) **20** 798-802. PMID: 20467256 13. Bryan J. **Treatment outcome of dogs with meticillin-resistant and meticillin-susceptible**. *Vet. Dermatol.* (2012.0) **23** 361-e365. DOI: 10.1111/j.1365-3164.2012.01034.x 14. Wang Y. **Methicillin-resistant**. *J. Appl. Microbiol.* (2012.0) **112** 623-630. DOI: 10.1111/j.1365-2672.2012.05233.x 15. Beck KM, Waisglass SE, Dick HLN, Weese JS. **Prevalence of meticillin-resistant**. *Vet. Dermatol.* (2012.0) **23** 369-e367. DOI: 10.1111/j.1365-3164.2012.01035.x 16. Bugden D. **Identification and antibiotic susceptibility of bacterial isolates from dogs with otitis externa in Australia**. *Aust. Vet. J.* (2012.0) **91** 43-46. DOI: 10.1111/avj.12007 17. Dziva F. **First identification of methicillin-resistant**. *Infect. Ecol. Epidemiol.* (2015.0) **5** 29170. DOI: 10.3402/iee.v5.29170 18. Zur G, Lifshitz B, Bdolah-Abram T. **The association between the signalment, common causes of canine otitis externa and pathogens**. *J. Small. Anim. Pract.* (2011.0) **52** 254-258. DOI: 10.1111/j.1748-5827.2011.01058.x 19. Sim JXF, Khazandi M, Chan WY, Trott DJ, Deo P. **Antimicrobial activity of thyme oil, oregano oil, thymol and carvacrol against sensitive and resistant microbial isolates from dogs with otitis externa**. *Vet. Dermatol.* (2019.0) **30** 524-e159. DOI: 10.1111/vde.12794 20. Scherer C. **Frequency and antimicrobial susceptibility of**. *Cienc. Rural.* (2018.0). DOI: 10.1590/0103-8478cr20170738 21. Rynhoud H. **Epidemiology of methicillin resistant**. *Res. Vet. Sci.* (2021.0) **136** 138-142. DOI: 10.1016/j.rvsc.2021.02.012 22. Lo W-T. **Comparative molecular analysis of meticillin-resistant**. *Br. J. Dermatol.* (2010.0) **162** 1110-1116. DOI: 10.1111/j.1365-2133.2010.09679.x 23. Sai N, Laurent C, Strale H, Denis O, Byl B. **Efficacy of the decolonization of methicillin-resistant**. *Antimicrob. Resist. Infect. Control* (2015.0) **4** 56. DOI: 10.1186/s13756-015-0096-x 24. Rodrigues Hoffmann A. **The skin microbiome in healthy and allergic dogs**. *PLoS ONE* (2014.0) **9** e83197. DOI: 10.1371/journal.pone.0083197 25. Bradley CW. **Longitudinal evaluation of the skin microbiome and association with microenvironment and treatment in canine atopic dermatitis**. *J. Invest. Dermatol.* (2016.0) **136** 1182-1190. DOI: 10.1016/j.jid.2016.01.023 26. Apostolopoulos N. **Description and comparison of the skin and ear canal microbiota of non-allergic and allergic German shepherd dogs using next generation sequencing**. *PLoS ONE* (2021.0) **16** e0250695. DOI: 10.1371/journal.pone.0250695 27. Chermprapai S. **The bacterial and fungal microbiome of the skin of healthy dogs and dogs with atopic dermatitis and the impact of topical antimicrobial therapy, an exploratory study**. *Vet. Microbiol.* (2019.0) **229** 90-99. DOI: 10.1016/j.vetmic.2018.12.022 28. Tress B. **Bacterial microbiome of the nose of healthy dogs and dogs with nasal disease**. *PLoS ONE* (2017.0) **12** e0176736. DOI: 10.1371/journal.pone.0176736 29. Cuscó A, Sánchez A, Altet L, Ferrer L, Francino O. **Individual signatures define canine skin microbiota composition and variability**. *Front. Vet. Sci.* (2017.0). DOI: 10.3389/fvets.2017.00006 30. Shi B, Leung DYM, Taylor PA, Li H. **Methicillin-resistant**. *J. Invest. Dermatol.* (2018.0) **138** 1668-1671. DOI: 10.1016/j.jid.2018.01.022 31. Hyejin K. **A double-blind, placebo controlled-trial of a probiotic strain**. *J. Microbiol. Biotechnol.* (2015.0) **25** 1966-1969. DOI: 10.4014/jmb.1506.06065 32. Hensel P, Santoro D, Favrot C, Hill P, Griffin C. **Canine atopic dermatitis: Detailed guidelines for diagnosis and allergen identification**. *BMC Vet. Res.* (2015.0) **11** 196. DOI: 10.1186/s12917-015-0515-5 33. Kumar S, Hussain K, Sharma R, Chhibber S, Sharma N. **Prevalence of canine otitis externa in Jammu**. *J. Anim. Res.* (2014.0) **4** 121-129. DOI: 10.5958/2277-940X.2014.00083.7 34. Harvey ND, Shaw SC, Craigon PJ, Blott SC, England GCW. **Environmental risk factors for canine atopic dermatitis: A retrospective large-scale study in Labrador and golden retrievers**. *Vet. Dermatol.* (2019.0) **30** 396-e119. DOI: 10.1111/vde.12782 35. Nødtvedt A, Egenvall A, Bergval K, Hedhammar Å. **Incidence of and risk factors for atopic dermatitis in a Swedish population of insured dogs**. *Vet. Rec.* (2006.0) **159** 241-246. DOI: 10.1136/vr.159.8.241 36. O’Neill DG. **Frequency and predisposing factors for canine otitis externa in the UK—A primary veterinary care epidemiological view**. *Canine Med. Genet.* (2021.0) **8** 7. DOI: 10.1186/s40575-021-00106-1 37. O'Neill DG, Coulson NR, Church DB, Brodbelt DC. **Demography and disorders of German shepherd dogs under primary veterinary care in the UK**. *Canine Genet. Epidemiol.* (2017.0) **4** 7-7. DOI: 10.1186/s40575-017-0046-4 38. Olivry T. **Treatment of canine atopic dermatitis: 2010 clinical practice guidelines from the International Task Force on Canine Atopic Dermatitis**. *Vet. Dermatol.* (2010.0) **21** 233-248. DOI: 10.1111/j.1365-3164.2010.00889.x 39. Bajwa J. **Canine otitis externa—Treatment and complications**. *Can. Vet. J.* (2019.0) **60** 97-99. PMID: 30651659 40. Flemer B. **The oral microbiota in colorectal cancer is distinctive and predictive**. *Gut* (2018.0) **67** 1454. DOI: 10.1136/gutjnl-2017-314814 41. Walters WA, Xu Z, Knight R. **Meta-analyses of human gut microbes associated with obesity and IBD**. *FEBS Lett.* (2014.0) **588** 4223-4233. DOI: 10.1016/j.febslet.2014.09.039 42. Zeller G. **Potential of fecal microbiota for early-stage detection of colorectal cancer**. *Mol. Syst. Biol.* (2014.0) **10** 766-766. DOI: 10.15252/msb.20145645 43. Qin N. **Alterations of the human gut microbiome in liver cirrhosis**. *Nature* (2014.0) **513** 59-64. DOI: 10.1038/nature13568 44. Zou H, Hastie T. **Regularization and variable selection via the elastic net**. *J. R. Stat. Soc. Ser. B (Stat. Methodol.)* (2005.0) **67** 301-320. DOI: 10.1111/j.1467-9868.2005.00503.x 45. Kawakami T. **Antimicrobial susceptibility and methicillin resistance in**. *J. Vet. Med. Sci.* (2010.0) **72** 1615-1619. DOI: 10.1292/jvms.10-0172 46. Siak M. **Characterization of meticillin-resistant and meticillin-susceptible isolates of**. *J. Med. Microbiol.* (2014.0) **63** 1228-1233. DOI: 10.1099/jmm.0.076117-0 47. Kania SA. **Methicillin resistance of staphylococci isolated from the skin of dogs with pyoderma**. *Am. J. Vet. Res.* (2004.0) **65** 1265-1268. DOI: 10.2460/ajvr.2004.65.1265 48. Isaiah A. **Characterization of the nasal and oral microbiota of detection dogs**. *PLoS ONE* (2017.0) **12** e0184899. DOI: 10.1371/journal.pone.0184899 49. Masiuk H, Wcisłek A, Jursa-Kulesza J. **Determination of nasal carriage and skin colonization, antimicrobial susceptibility and genetic relatedness of**. *BMC Infect. Dis.* (2021.0) **21** 701. DOI: 10.1186/s12879-021-06382-3 50. Alsterholm M. **Variation in**. *Acta Derm. Venereol.* (2017.0) **97** 802-807. DOI: 10.2340/00015555-2667 51. Yao H. **Absolute abundance values reveal microbial shifts and co-occurrence patterns during gut microbiota fermentation of dietary fibres in vitro**. *Food Hydrocoll.* (2022.0) **127** 107422. DOI: 10.1016/j.foodhyd.2021.107422 52. Engelbrektson A. **Experimental factors affecting PCR-based estimates of microbial species richness and evenness**. *ISME J.* (2010.0) **4** 642-647. DOI: 10.1038/ismej.2009.153 53. Bolyen E. **Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2**. *Nat. Biotechnol.* (2019.0) **37** 852-857. DOI: 10.1038/s41587-019-0209-9 54. Callahan BJ. **DADA2: High-resolution sample inference from Illumina amplicon data**. *Nat. Methods.* (2016.0) **13** 581-583. DOI: 10.1038/nmeth.3869 55. Quast C. **The SILVA ribosomal RNA gene database project: Improved data processing and web-based tools**. *Nucleic Acids Res.* (2013.0) **41** D590-D596. DOI: 10.1093/nar/gks1219 56. 56.Microsoft 365. Microsoft Excel. https://www.microsoft.com/en-au/microsoft-365/excel?legRedir=true&CorrelationId=8ebc73df-5955-4130-b367-499fea70f562&rtc=1 (2022). 57. 57.Yan, L. ggvenn: Draw Venn Diagram by 'ggplot2'. https://CRAN.R-project.org/package=ggvenn (2021). 58. 58.Gehlenborg, N. UpSetR: A More Scalable Alternative to Venn and Euler Diagrams for Visualizing Intersecting Sets. https://CRAN.R-project.org/package=UpSetR (2019). 59. 59.Prabhakaran, S. InformationValue: Performance Analysis and Companion Functions for Binary Classification Models. https://CRAN.R-project.org/package=InformationValue (2016). 60. 60.Kuhn, M. et al. Classification and Regression Training: Package 'Caret'. https://cran.r-project.org/web/packages/caret/caret.pdf (2022). 61. 61.Hosmer, D. W., Lemeshow, S. & Sturdivant, R. X. Applied Logistic Regression (eds. Hosmer, D.W., Lemeshow, S., & Sturdivant, R.X.). 153–225 (Wiley, 2013). 62. 62.Kuhn, M. caret: Classification and Regression Training. https://CRAN.R-project.org/package=caret (2021). 63. 63.Larsen, K. Information: Data Exploration with Information Theory (Weight-of-Evidence and Information Value). https://CRAN.R-project.org/package=Information (2016). 64. 64.Hamner, B. & Frasco, M. Metrics: Evaluation Metrics for Machine Learning. https://CRAN.R-project.org/package=Metrics (2018). 65. Friedman JH, Hastie T, Tibshirani R. **Regularization paths for generalized linear models via coordinate descent**. *J. Stat. Softw.* (2010.0) **33** 1-22. DOI: 10.18637/jss.v033.i01 66. Csardi G, Nepusz T. **The igraph software package for complex network research**. *Int. J. Complex Syst.* (2006.0) **1695** 1-9 67. McMurdie PJ, Holmes S. **phyloseq: An R package for reproducible interactive analysis and graphics of microbiome census data**. *PLoS ONE* (2013.0) **8** e61217. DOI: 10.1371/journal.pone.0061217
--- title: Meta-analysis of the normal diffusion tensor imaging values of the peripheral nerves in the upper limb authors: - Ryckie G. Wade - Fangqing Lu - Yohan Poruslrani - Chiraag Karia - Richard G. Feltbower - Sven Plein - Grainne Bourke - Irvin Teh journal: Scientific Reports year: 2023 pmcid: PMC10039047 doi: 10.1038/s41598-023-31307-2 license: CC BY 4.0 --- # Meta-analysis of the normal diffusion tensor imaging values of the peripheral nerves in the upper limb ## Abstract Peripheral neuropathy affects 1 in 10 adults over the age of 40 years. Given the absence of a reliable diagnostic test for peripheral neuropathy, there has been a surge of research into diffusion tensor imaging (DTI) because it characterises nerve microstructure and provides reproducible proxy measures of myelination, axon diameter, fibre density and organisation. Before researchers and clinicians can reliably use diffusion tensor imaging to assess the ‘health’ of the major nerves of the upper limb, we must understand the “normal” range of values and how they vary with experimental conditions. We searched PubMed, Embase, medRxiv and bioRxiv for studies which reported the findings of DTI of the upper limb in healthy adults. Four review authors independently triple extracted data. Using the meta suite of Stata 17, we estimated the normal fractional anisotropy (FA) and diffusivity (mean, MD; radial, RD; axial AD) values of the median, radial and ulnar nerve in the arm, elbow and forearm. Using meta-regression, we explored how DTI metrics varied with age and experimental conditions. We included 20 studies reporting data from 391 limbs, belonging to 346 adults (189 males and 154 females, ~ 1.2 M:1F) of mean age 34 years (median 31, range 20–80). In the arm, there was no difference in the FA (pooled mean 0.59 mm2/s [$95\%$ CI 0.57, 0.62]; I2 $98\%$) or MD (pooled mean 1.13 × 10–3 mm2/s [$95\%$ CI 1.08, 1.18]; I2 $99\%$) of the median, radial and ulnar nerves. Around the elbow, the ulnar nerve had a $12\%$ lower FA than the median and radial nerves ($95\%$ CI − 0.25, 0.00) and significantly higher MD, RD and AD. In the forearm, the FA (pooled mean 0.55 [$95\%$ CI 0.59, 0.64]; I2 $96\%$) and MD (pooled mean 1.03 × 10–3 mm2/s [$95\%$ CI 0.94, 1.12]; I2 $99\%$) of the three nerves were similar. Multivariable meta regression showed that the b-value, TE, TR, spatial resolution and age of the subject were clinically important moderators of DTI parameters in peripheral nerves. We show that subject age, as well as the b-value, TE, TR and spatial resolution are important moderators of DTI metrics from healthy nerves in the adult upper limb. The normal ranges shown here may inform future clinical and research studies. ## Introduction Peripheral neuropathy and nerve injury are common, affecting approximately 1 in 10 adults over the age of 40 years1. Given the absence of a reliable diagnostic test, there has been a surge of research related to diffusion-weighted magnetic resonance imaging (dMRI) for the evaluation of peripheral nerves. dMRI characterises tissue microstructure and provides reproducible2–5 proxy measures of nerve health which are sensitive to axon type, diameter, density, myelination and organisation6–9. The most prevalent form of dMRI in peripheral nerves is diffusion tensor imaging (DTI). This typically generates the following voxel-wise parameters; fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD) and radial diffusivity (RD). FA is a scalar value between zero and one; an FA of zero implies isotropic diffusion within a voxel, whilst (in the context of peripheral nerves) a FA nearing one implies diffusion predominantly along a single axis (i.e., axoplasmic along nerves). MD describes the average molecular diffusion rate of the tensor; AD describes the diffusion rate in the long axis and RD represents diffusion perpendicular to the long axis. The diffusion of water and therefore dMRI signal is affected by factors other than tissue microstructure and physiology. Studies in the brain have shown that dMRI outputs are dependent on scanner hardware, acquisition settings, preprocessing techniques, reconstruction algorithms and extraction methods2,10–15. Recent work has shown similar dependence in the brachial plexus16 and median nerve in the hand17. Therefore, before researchers and clinicians can reliably use dMRI to assess the ‘health’ of the major nerves of the upper limb, there is a need to define the range of “normal” values and how they vary with experimental conditions. These knowledge gaps form the rationale for this review. ## Methods This review was registered on the PROSPERO database (CRD42021275343). It was designed and conducted in accordance with the Cochrane Handbook of Systematic Reviews18 and reported in accordance with the PRISMA checklist19. ## Types of studies We included studies which reported the findings of DTI of the arm, elbow or forearm in healthy adults. There were no language restrictions. Case reports were excluded. ## Participants This review considers adults (aged ≥ 16 years) with no known pathology (past or present) affecting any peripheral nerve(s) of the upper limb. ## Image acquisition The included studies must have reported the DTI parameters from any of the median, ulnar or radial nerve in the forearm. The forearm was defined as the anatomical region distal to the elbow joint and proximal to the radiocarpal joint. ## Search strategy In accordance with our search strategy (Appendix 1), PubMed and Embase were interrogated using the NICE Healthcare Databases (hdas.nice.org.uk), and medRxiv and bioRxiv were searched using the R package medrxivr20 from inception to the 5th January 2022. This yielded 93 hits in PubMed, 509 in Medline but none in the preprint archives. Later, CitationChaser21 was used for forward and backward chasing of 452 citations, which yielded a further 371 records on 28th January 2022. ## Study selection Three review authors (FL, YP and RGW) independently screened titles and abstracts for relevance, in accordance with the eligibility criteria. The full texts of potentially eligible articles were obtained and again independently assessed by the same authors. Disagreements were resolved by discussion. ## Data extraction Three review authors (FL, YP and RGW) independently extracted data in duplicate, after which a 4th review author (CK) independently performed complete data validation (i.e. triplicate extraction). Where bilateral or repeated (e.g., test–retest) measurements were reported, the values were averaged given that the variability of DTI metrics from peripheral nerves on the right and left side22–24, intrasessional and intersessional variability is less than $5\%$25. When data were missing or unclear, the corresponding author was contacted by email and if no reply was received, values were back-calculated26 or extracted from graphs using metaDigitise27. Some studies used the term apparent diffusion coefficient (ADC) rather than MD. By convention, MD is used to describe the average of the diffusion tensor eigenvalues in DTI. As all the included studies fitted DTI models to their data, we refer to their reported results as MD rather than ADC. Two authors provided additional information upon request28,29. ## Outcomes We planned to estimate the normal FA and diffusivity values (MD, RD and AD) of the major upper limb nerves in healthy adults. Thereafter, we planned to explore how DTI metrics varied with age, anatomical location, and experimental conditions, such as the b-value(s), echo time(s) (TE), repetition times (TR), resolution (in cubic millimetres, mm3) and the number of diffusion encoding gradient directions (ND) sampled per shell. When the anatomical location was described we categorised data into 3 distinct regions: [1] the arm, which included data distal to the shoulder joint and ~ 5 cm proximal to the elbow joint; [2] the elbow, which included data ~ 5 cm either side of the elbow joint; and (c) the forearm, defined as 5 cm distal to the elbow joint and proximal to the radiocarpal joint. ## Methodological quality assessment There is no consensus on the appropriate tool to assess the risk of methodological bias in observational studies of healthy volunteers, so no risk of bias assessment was performed. ## Statistical analysis The datasets generated and/or analysed during the current study (including the outputs of metaDigitise and additional data shared by authors) are available in the Open Science Framework repository, https://osf.io/8yzst/. The PRISMA2020 tool30 was used to create the flow diagram. Two studies were excluded from the meta-analysis given that they were performed at 1.5T45 and 7T31. Data was analysed in Stata/MP v17 (StataCop LLC, Texas) and graphs customised using grstyle32,33. Using the meta suite, the aggregate mean FA, MD, RD and AD from studies were pooled to estimate the normative values, subgrouped by anatomical location and nerve. Restricted maximum likelihood was used to estimate the between-study variance (tau2), with the Knapp and Hartung modification. Heterogeneity was quantified by I234. Sensitivity leave-one-out meta-analyses were performed to identify potential outlier studies. Mixed-effects meta-regression was then used to explore heterogeneity with FA as the dependent variable. We selected the moderator variables in the protocol phase through the production of a directed acyclic graph (http://dagitty.net/dags.html?id=cgJvh9#)35. The continuous fixed effects were age, resolution (mm3), echo time (TE in ms), b-value (s/mm2) and number of diffusion encoding gradient directions (ND) whilst the categorical fixed effect was the location (arm, elbow and forearm). For each moderator (fixed-effect), we used the minimum adjustment dataset as prescribed by DAGitty (eFigs. S1–S7). Thereafter, variance inflation factors (used to quantify potential multicollinearity) were calculated36,37. Confidence intervals (CI) were generated to the $95\%$ level. ## Results Overall, 20 studies38 were included (eFig. S8). ## Study characteristics Study characteristics are detailed in eTable S1. We included data from 391 limbs belonging to 346 adults (189 males and 154 females [3 were of unknown sex], translating to ~ 1.2 M:1F) of mean age 34 years (median 31, range 20–80). The median number of authors per paper was 7 (range 5–10). Studies were performed most commonly on Siemens (9 studies31,38–43, $43\%$) or Philips (9 studies29,44,44–51, $43\%$) scanners with the remainder using GE (3 studies28,52,53, $14\%$). Nineteen studies were performed at 3T28,29,38–44,46–54, with one at 1.5T45 and another at 7T31. The majority of studies used single-shot echo planar imaging (ssEPI; 17 studies, $85\%$), one compared ssEPI to readout-segmented echo planar imaging (rsEPI)39 and 3 studies28,31,40 did not specify the type of sequence. Subjects were most commonly in the “superman” position (i.e., prone with the shoulder and elbow extended, with the elbow positioned in the isocentre of the magnet; 17 studies28,29,31,38,38,39,41–43,45,46,48–52,54, $85\%$) whilst the others positioned individuals supine with their arm extended overhead40, supine in the anatomical position44,47 or did not report the position53. The receiver coils used were most commonly flexible extremity coils (11 studies29,31,39,41–44,47,51,54, $55\%$) with the remainder using head48, knee38,46,50,52, wrist28,45 or unspecified53 coils. The receiver coils had a median 8 channels (IQR 8–12, range 2–32). The mean TE and TR were 82 ms (range 65, 105) and 5459 ms (range 2800, 10,000), respectively. The mean in-plane resolution was 1.17mm2 (range 0.12–1.8). The mean slice thickness was 3.5 mm (range 2–4). *This* generated a mean voxel volume of 5.11 mm3 (range 0.06–9.72). Nine studies used parallel imaging techniques (GRAPPA39–42,42, SENSE29,29,49–51, ASSET52, and one unspecified method31) whilst 9 studies28,38,44–48,53,54 did not specify this parameter. Four studies used partial Fourier acquisition39–41,51, one reported full k-space acquisition52 but the majority of studies28,29,31,38,42–50,53,54 did not report this information. Nineteen studies captured a single (maximum) b-value of mean 1045 s/mm2 (range 700–1300). One study52 captured several b-values (300, 450, 600, 750 and 900 s/mm2) to calculate track-weighted DTI metrics via multi-shell multi-tissue constrained spherical deconvolution. All other studies reconstructed their data using 2nd order tensors. The median ND was 20 (range 6–64). Most studies did not specify the diffusion encoding waveform28,29,31,42–51,53,54 whilst 6 studies38–41,52 used “monopolar” with no further explanation. Four studies reported preprocessing their data29,31,41,52,52. This included MC-PCA denoising52; Gibbs ringing correction52; correction of artefacts related to susceptibility, motion and eddy currents using MRtrix352, FSL31,41 or ExploreDTI55; bias correction using Advanced Normalisation Tools52; and interpolation of slice thickness from 3 to 1 mm52. Two studies42,51 had a single image reader/reporter whilst the remainder had two reporting clinicians/scientists28,29,31,38,38–41,43–50,52–54. ## Evidence synthesis: the arm In the arm, the normal FA of the median40,42,43,52, radial40,42,43,47 and ulnar40–42,42,46,52,54 nerves are shown in Fig. 1. There was no significant difference in FA between the three nerves ($$p \leq 0.554$$, I2 $96\%$). The normal MD of the median40,42,43,52, radial40,42,43,47,47 and ulnar40–43,52,54 nerves was 1.13 × 10–3 mm2/s (CI 1.08, 1.18; Fig. 1) with no significant difference between nerves ($$p \leq 0.95$$, I2 $99\%$). The normal RD of the median40,42,52, radial40,42 and ulnar40–42,52 nerves was 0.70 × 10–3 mm2/s (CI 0.73, 0.76; Fig. 1) with no significant difference between nerves ($$p \leq 0.80$$, I2 $95\%$). The normal AD of the median40,42,52, radial40,42 and ulnar40–42,52 nerves was 1.99 × 10–3 mm2/s (CI 1.91, 2.08; eFigure S10) with no significant difference between nerves ($$p \leq 0.74$$, I2 $99\%$). Leave-one-out meta-analysis did not detect any outlier studies. Figure 1Forest plot of the normal FA of the median, ulnar and radial nerves in the arm, sorted by the echo time and b-value. ## Evidence synthesis: the elbow Around the elbow, the normal FA of the median51,52, radial51 and ulnar38,39,41,44,46,49,51,52,54 nerves are shown in Fig. 2.Figure 2Forest plot of the normal FA of the median, ulnar and radial nerves around the elbow, sorted by the echo time and resolution. In the region of the elbow, the ulnar nerve appeared to have a $12\%$ lower FA than the median and radial nerves (CI − 0.25, 0.00; $p \leq 0.001$, I2 $96\%$). The normal MD of the median51,52, radial51 and ulnar39,41,44,46,49,51,52,54,54 nerves in the elbow region was 1.01 × 10–3 mm2/s (CI 0.68, 1.34), 0.71 × 10–3 mm2/s (CI 0.64, 0.78) and 1.11 × 10–3 mm2/s (CI 1.01, 1.20), respectively (eFigure S12). The radial nerve had lower MD than the ulnar nerve (− 0.40 × 10–3 mm2/s CI − 0.76, − 0.04]; $p \leq 0.001$, I2 $93\%$) but was similar to the median nerve. The normal RD of the median51,52, radial51 and ulnar41,51,52 nerves around the elbow was 0.57 × 10–3 mm2/s (CI 0.41, 0.71; eFig. S13) with no significant difference between the nerves ($$p \leq 0.409$$). The normal AD of the median51,52, radial51 and ulnar41,51,52 nerves at the level of the elbow was 1.94 × 10–3 mm2/s (CI 1.71, 2.17), 1.35 × 10–3 mm2/s (CI 1.23, 1.47) and 1.88 × 10–3 mm2/s (CI 1.74, 2.03), respectively (eFig. S14). The radial nerve had a lower AD than both the ulnar nerve (mean difference 0.53 × 10–3 mm2/s [CI 0.25, 0.82]) and median nerve (mean difference 0.59 [CI 0.28, 0.91]; $p \leq 0.001$, I2 $90\%$). Leave-one-out meta-analysis did not detect any outlier studies. ## Evidence synthesis: the forearm In the forearm, the normal FA of the median51,52, radial51 and ulnar38,39,39,41,44,46,49,51,52,54 nerves are shown in Fig. 3 and there was no significant difference between the nerves ($$p \leq 0.690$$, I2 $96\%$). The normal MD of the median29,48,50,51,53, radial51 and ulnar29,41,46,50,51 nerves around the elbow was 1.03 × 10–3 mm2/s (CI 0.94, 1.12; eFig. S15) with no significant difference between the nerves ($$p \leq 0.409$$, I2 $97\%$). The normal RD of the median, radial and ulnar nerves in the forearm was 0.64 × 10–3 mm2/s (CI 0.51, 0.77; eFig. S16) with no significant difference between the nerves ($$p \leq 0.752$$, I2 $98\%$). The normal AD of the median, radial and ulnar nerves in the forearm was 1.91 × 10–3 mm2/s (CI 1.77, 2.04; eFig. S17) with no significant difference between the nerves ($$p \leq 0.562$$, I2 $95\%$). Leave-one-out meta-analysis did not detect any outlier studies. Figure 3Forest plot of the normal FA of the median, ulnar and radial nerves within the forearm, sorted by the echo time and b-value. ## Meta-regression Multivariable meta regression showed that the TE, TR, b-value, spatial resolution, anatomical location and age of the subject moderated DTI metrics within peripheral nerves (Table 1 and Fig. 4).Table 1Multivariable meta-regression of factors associated with FA and MD.ModeratorsFAMDAdjusted β ($95\%$ CI)Adjusted β ($95\%$ CI)Echo time (ms)− 5.6 × 10–3 (− 7.9 × 10–3, − 3.3 × 10–3)5.5 × 10–3 (− 5.6 × 10–4, 0.01)Age in years− 4.8 × 10–3 (− 7.1 × 10–3, − 2.4 × 10–3)7.8 × 10–6 (3.3 × 10–6, 1.2 × 10–5)b-value (s/mm2)8.9 × 10–5 (− 5.9 × 10–5, 2.4 × 10–4)− 3.8 × 10–4 (− 6.4 × 10–4, − 1.2 × 10–4)Location ArmReferentReferent Elbow1.5 × 10–3 (− 0.06, 0.06)− 0.06 (− 0.17, 0.05) Forearm0.06 (2.5 × 10–4, 0.13)− 0.10 (− 0.17, 0.05)Repetition time (s)7.9 × 10–6 (− 1.4 × 10–5, 3.0 × 10–5)− 5.1 × 10–5 (− 9.3 × 10–5, − 9.6 × 10–6)Resolution (mm3)− 0.02 (− 0.03, − 0.004)0.02 (9.1 × 10–4, 0.04)ND1.8 × 10–3 (− 3.9 × 10–4, 4.0 × 10–3)3.6 × 10–3 (− 7.8 × 10–3, 4.8 × 10–4)ND = the number of diffusion encoding gradient directions. Note that each variable has been differently adjusted, as defined by our DAGs (http://dagitty.net/mcgJvh9, eFigs. S11–S17).Figure 4Bubble plots showing the linear dependence of FA (left column) and MD (right column) on the maximum b-value, echo time and aggregate mean age of participants. The size of the points corresponds to the precision (inverse variance) of the study. Both anisotropy and diffusivity were dependent on age whereby each decade of life reduced the FA by 0.05 (CI 0.007, 0.02) and increased MD by 7.8 × 10–5 mm2/s (CI 3.3 × 10–5, 1.2 × 10–4). Increments in the b-value of 100 s/mm2 reduced the observed MD by approximately 0.038 mm2/s (CI 0.064–0.012) without affecting the FA. increments of 10 ms in the TE reduced the FA within peripheral nerves by approximately 0.056 (CI 0.079, 0.033) without affecting the MD. Increasing the spatial resolution by 1mm3 downwardly biased the FA by $2\%$ (CI 3–4) and upwardly biased the MD by 0.02 (CI 9.1 × 10–4, 0.04). The nerves within the forearm had a $6\%$ higher FA than nerves within the arm. FA and MD appeared to be robust to ND. ## Discussion This work shows that dMRI metrics from healthy nerves in the upper limb are dependent on experimental conditions and age, and differ throughout the length of the limb. Importantly, we show that seemingly small alterations to acquisition parameters (e.g., changing the b-value by 100 mm2/s or TE by 10 ms) is associated with meaningful changes to DTI measurements. Equally, we show that DTI metrics from the median, ulnar and radial nerves are age-dependent, which has important ramifications. Our work corroborates prior dMRI studies in peripheral nerves16,17 (and the brain56) which demonstrate that nerves exhibit more isotropic diffusion with advancing age. This is expected because aging axons lose their integrity, axoplasmic transport is slowed and the myelin sheath deteriorates which gives way to segmental demyelination and axonal loss without remyeliation57. These morphological changes lead to an increase in extra-cellular fluid and decline in both the density and integrity of microstructures which hinder/restrict diffusion. On a practical level, the observation that aging nerves exhibit more isotropic diffusion is important because it shows that when comparing group differences or longitudinal changes in dMRI metrics, adjustment for age is likely to be required. In keeping with the prior literature, this work solidifies a clinically important and unique features of the pattern of diffusion within the ulnar nerve. At the level of the elbow and more distally within the forearm, diffusion within the ulnar nerve appears to become more isotropic compared to its proximal course. This may be a manifestation of microstructural changes due to hardships endured around the elbow, namely repetitive mechanical deformation as the elbow moves, cumulative external trauma from knocks (aka the “funny bone”) and the resistance created due to the passage through a relatively tight fibrosseous (cubital) tunnel. These factors together may contribute to systematic differences in the microstructure of the ulnar nerve which render its diffusion more isotropic within and beyond the cubital tunnel. In our multivariable meta-regression model, prolongation of the TE was associated with lower estimates of anisotropy. Whilst we show the same lack of association between TE and MD as observed in the brain58, the observation that TE downwardly biases FA in the limb does not agree with the literature in the human brain (at both 1.5 and 3 T58) whereby a positive linear correlation was observed between TE and the FA within white matter. However, the opposite was observed in this study, which is difficult to explain, so we offer some hypotheses. We observed a positive linear relationship between TE and aggregate age (β 0.40 [CI 0.21, 0.58]) which might explain why studies with longer TE (older participants) had lower FA. Secondly, there was a linear correlation between TE and b-value ($r = 0.42$) but the variance inflation factor for TE and b-value was 4.9 and 1.2, respectively in the model with TE as the exposure. Consensus amongst the statistical community is that a correlation coefficient > 0.7 between predictor variables or variance inflation factors > 10 is evidence of multicollinearity and should lead to the exclusion of colinear variables. Our models appear to have some collinearity (not enough to warrant variable removal) and we feel that this might contribute but does not completely explain the relationship between TE and FA. Finally, FA decreasing with TE might represent differences in the T2 of intracellular water and extracellular water in peripheral nerves. Compounding this is the problem of myelin’s magnetic susceptibility which alters the off-resonance field59 for intra-axonal water, meaning that at longer echo times (after diffusion-weighting) there may be phase offsets which amplify differences between intra- and extra-axonal water. Also, at longer echo times there may be more sensitivity to non-gaussian diffusion in peripheral nerves, which has been observed to start from lower b-values than in the brain (~ 700 s/mm2)60. Finally, the included studies did not report the diffusion time (and many other important methods), so it is plausible that the relationship between TE and FA was confounded by something else. Future studies should seek to: (a) fully report the parameters of their sequences, (b) report methods of pre- and postprocessing, and (c) make their anonymised data available open source to enable individual patient-data meta-analysis. There are some important limitations to our study. Non-gaussian diffusion has been observed at b-values above ~700 s/mm2 and consequently, monoexponential fitting (i.e., a 2nd order tensor) may be influenced by restricted diffusion at higher b-values60. It is widely accepted that preprocessing of dMRI data improves the accuracy of metrics and tractography61, and differences in preprocessing practices and pipelines generate important differences in results2 which negatively impacts reproducibility15. In this review, most studies failed to report if or what preprocessing was performed, and this may be a source of variability. Finally, few authors described their postprocessing methods (e.g. the size and position of regions-of-interest used to extract DTI metrics, how they were drawn, etc.) which is important because recent work has shown that subtle variability in the size and position of regions of interest have downstream effects on DTI metrics10. Some readers may decry our decision to meta-analyse statistically heterogeneous data, but this was done purposively because forest plots provide an important graphical representation of measurement variation in relation to experimental conditions (e.g., b-values and ND) and they summarise a large amount of information in an easy-to-interpret format. Furthermore, by making this choice we could deploy meta-regression to explore potential moderators. Ultimately, our choice to meta-analyse heterogenous data has provided important insight into factors which appear to moderate FA and diffusivity within the nerves of the upper limb. Non-biological variability in dMRI metrics undermines the reliability of multi-site and/or longitudinal studies. Therefore, there remains a need for robust harmonisation techniques62,63. Harmonisation is a mathematical approach (regression, interpolation or machine learning) which seeks to reduce the unwanted (non-biological) variability in dMRI datasets whilst retaining information which pertains to the underlying microstructure and physiology64. A recent review of Harmonisation of dMRI64 showed the benefits of such an approach. By summarising the effects of non-biological variability in dMRI of the arm, elbow and forearm, we provide information which may inform harmonisation efforts in the limb by quantifying the direction and magnitude of dMRI metric variation in relation to non-biological factors. In conclusion, we show that dMRI metrics from healthy nerves in the upper limb are age-dependent, and that the b-value, echo time, repetition time and resolution are clinically important sources of variability. We provide summary estimates of the normal values of the median, ulnar and radial nerves in different experimental settings which may be of value to researchers and clinicians alike. ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-31307-2. ## References 1. Hicks CW, Wang D, Windham BG. **Prevalence of peripheral neuropathy defined by monofilament insensitivity in middle-aged and older adults in two US cohorts**. *Sci. Rep.* (2021.0) **11** 19159. DOI: 10.1038/s41598-021-98565-w 2. Nath V, Schilling KG, Parvathaneni P. **Tractography reproducibility challenge with empirical data (TraCED): The 2017 ISMRM diffusion study group challenge**. *J. Magn. Reson. Imaging.* (2020.0) **51** 234-249. DOI: 10.1002/jmri.26794 3. Vavasour IM, Meyers SM, Mädler B. **Multicenter measurements of T 1 relaxation and diffusion tensor imaging: intra and intersite reproducibility**. *J. Neuroimaging.* (2019.0) **29** 42-51. DOI: 10.1111/jon.12559 4. Prohl AK, Scherrer B, Tomas-Fernandez X. **Reproducibility of structural and diffusion tensor imaging in the TACERN multi-center study**. *Front. Integr. Neurosci.* (2019.0) **13** 1-15. PMID: 30705624 5. Kimura M, Yabuuchi H, Matsumoto R. **The reproducibility of measurements using a standardization phantom for the evaluation of fractional anisotropy (FA) derived from diffusion tensor imaging (DTI)**. *Magn. Reson. Mater. Phys. Biol. Med.* (2019.0) **1** 15-19 6. Heckel A, Weiler M, Xia A. **Peripheral nerve diffusion tensor imaging: Assessment of axon and myelin sheath integrity**. *PLoS ONE* (2015.0) **10** e0130833. DOI: 10.1371/journal.pone.0130833 7. Andersson G, Orädd G, Sultan F. **In vivo diffusion tensor imaging, diffusion kurtosis imaging, and tractography of a sciatic nerve injury model in rat at 9.4T**. *Sci. Rep.* (2018.0) **8** 12911. DOI: 10.1038/s41598-018-30961-1 8. Friedrich P, Fraenz C, Schlüter C. **The relationship between axon density, myelination, and fractional anisotropy in the human corpus callosum**. *Cereb. Cortex.* (2020.0) **30** 2042-2056. DOI: 10.1093/cercor/bhz221 9. Pušnik L, Serša I, Umek N, Cvetko E, Snoj Z. **Correlation between diffusion tensor indices and fascicular morphometric parameters of peripheral nerve**. *Front. Physiol.* (2023.0) **14** 1070227. DOI: 10.3389/fphys.2023.1070227 10. 10.Rheault, F., Schilling, K., & Obaid, S., et al. The influence of regions of interest on tractography virtual dissection protocols: General principles to learn and to follow. Res. Sq. (2022). 11. Alexander AL, Lee JE, Wu Y-C. **Comparison of diffusion tensor imaging measurements at 3.0 T versus 1.5 T with and without Parallel imaging**. *Neuroimaging Clin. N. Am.* (2006.0) **16** 299-309. DOI: 10.1016/j.nic.2006.02.006 12. Ni H, Kavcic V, Zhu T. **Effects of number of diffusion gradient directions on derived diffusion tensor imaging indices in human brain**. *AJNR Am. J. Neuroradiol.* (2006.0) **27** 1776-1781. PMID: 16971635 13. Vollmar C, O’Muircheartaigh J, Barker GJ. **Identical, but not the same: Intra-site and inter-site reproducibility of fractional anisotropy measures on two 3.0T scanners**. *Neuroimage* (2010.0) **51** 1384-1394. DOI: 10.1016/j.neuroimage.2010.03.046 14. Grech-Sollars M, Hales PW, Miyazaki K. **Multi-centre reproducibility of diffusion MRI parameters for clinical sequences in the brain**. *NMR Biomed.* (2015.0) **28** 468-485. DOI: 10.1002/nbm.3269 15. Esteban O, Daducci A, Caruyer E. **IEEE 11th International Symposium on Biomedical Imaging (ISBI). Vol 582**. *IEEE* (2014.0) **2014** 738-741 16. Wade RG, Whittam A, Teh I. **Diffusion tensor imaging of the roots of the brachial plexus: A systematic review and meta-analysis of normative values**. *Clin. Transl. Imaging.* (2020.0) **8** 419-431. DOI: 10.1007/s40336-020-00393-x 17. Rojoa D, Raheman F, Rassam J. **Diffusion tensor imaging of the median nerve: A systematic review and meta-analysis of normal values in asymptomatic adults and how they change in carpal tunnel syndrome**. *SSRN Electron. J.* (2021.0) **1** 1-11 18. 18.Higgins, J.P.T., & Green, S. (editors). Cochrane handbook for systematic reviews of interventions version 5.1.0 [updated March 2011]. Cochrane Collab. (2011). 19. Moher D, Liberati A, Tetzlaff J. **Systematic reviews and meta-analyses: The PRISMA statement**. *Annu. Intern. Med.* (2009.0) **151** 264-269. DOI: 10.7326/0003-4819-151-4-200908180-00135 20. McGuinness L, Schmidt L. **medrxivr: Accessing and searching medRxiv and bioRxiv preprint data in R**. *J. Open Source Softw.* (2020.0) **5** 2651. DOI: 10.21105/joss.02651 21. 21.Haddaway, N.R. citationchaser: An R package for forward and backward citations chasing in academic searching. 2021. Available at: https://zenodo.org/record/4533747. Accessed 15 Feb 2021. 22. Wade RG, Teh I, Andersson G. **Fractional anisotropy thresholding for deterministic tractography of the roots of the brachial plexus**. *Sci. Rep.* (2021.0) **11** 80. DOI: 10.1038/s41598-020-79840-8 23. Andreisek G, White LM, Kassner A. **Evaluation of diffusion tensor imaging and fiber tractography of the median nerve: Preliminary results on intrasubject variability and precision of measurements**. *AJR Am. J. Roentgenol.* (2010.0) **194** W65. DOI: 10.2214/AJR.09.2517 24. Holmes SA, Staffa SJ, Karapanagou A. **Biological laterality and peripheral nerve DTI metrics**. *PLoS ONE* (2020.0) **10** 1 25. Cai LY, Yang Q, Kanakaraj P. **MASiVar: Multisite, multiscanner, and multisubject acquisitions for studying variability in diffusion weighted MRI**. *Magn. Reson. Med.* (2021.0) **1** 28926 26. Wan X, Wang W, Liu J. **Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range**. *BMC Med. Res. Methodol.* (2014.0) **14** 135. DOI: 10.1186/1471-2288-14-135 27. 27.Pick, J.L., Nakagawa, S., & Noble, D.W.A. Reproducible, flexible and high-throughput data extraction from primary literature: The metaDigitise R package. Evolutionary Biology; 2018. 10.1101/247775. Accessed 10 Jan 2022. 28. 28.Ohana, M.M.T., Meyer, N., Zorn, P.E., Liverneaux, P., Dietemann, J.L. 3T tractography of the median nerve: Optimisation of acquisition parameters and normative diffusion values. Diagn. Interv. Imaging. 93(10), 775–784 (2012). 29. 29.Haakma, W.J.B.A., Froeling, M., Goedee, H.S., Bos, C., Leemans, A., van den Berg, L.H., & Hendrikse, J., van der Pol, W.L. MRI shows thickening and altered diffusion in the median and ulnar nerves in multifocal motor neuropathy. Eur. Radiol.27(5), 2216–2224 (2016). 30. 30.Haddaway, N.R., McGuinness, L.A., & Pritchard, C.C. PRISMA2020: R package and ShinyApp for producing PRISMA 2020 compliant flow diagrams. 2021. Available at: https://zenodo.org/record/4287834. Accessed March 24, 2022. 31. Raval SB, Cynthia A, Zhao T, Krishnamurthy N, Santini T, Gorantla VS, Ibrahim TS. **Ultra-high field upper extremity peripheral nerve and non-contrast enhanced vascular imaging**. *PLoS ONE* (2017.0) **12** e0175629. DOI: 10.1371/journal.pone.0175629 32. 32.Jann, B. Customizing stata graphs made easy (Part 1). Stata J.18(3), 491–502. 33. 33.Jann, B. Customizing stata graphs made easy (Part 2). Stata J.18(4), 786–802. 34. Higgins JPT, Thompson SG. **Quantifying heterogeneity in a meta-analysis**. *Stat. Med.* (2002.0) **21** 1539-1558. DOI: 10.1002/sim.1186 35. 35.Textor, J., van der Zander, B., & Gilthorpe, M.S., et al. Robust causal inference using directed acyclic graphs: the R package ‘dagitty.’ Int. J. Epidemiol. dyw341 (2017). 36. 36.Viechtbauer, W. Conducting meta-analyses in R with the metafor package. J. Stat. Softw.36(3) (2010). Available at: http://www.jstatsoft.org/v36/i03/. Accessed February 26, 2021. 37. 37.Wickham, H., & Miller, E. Haven: Import and export “SPSS”, “Stata” and “SAS” Files. R package version 2.5. Available at: https://CRAN.R-project.org/package=haven. 38. Bäumer PPM, Ruetters M, Heiland S, Heckel A, Radbruch A, Bendszus M, Weiler M. **Peripheral neuropathy: Detection with diffusion-tensor imaging**. *Radiology* (2014.0) **273** 185-193. DOI: 10.1148/radiol.14132837 39. Ho M, Becker A, Ulbrich E. **Comparison of simultaneous multi-slice readout-segmented EPI and conventional single-shot EPI for diffusion tensor imaging of the ulnar nerve**. *Heliyon.* (2018.0) **4** e00853. DOI: 10.1016/j.heliyon.2018.e00853 40. Breckwoldt MO, Christian S, Xia A, Heckel A, Bendszus M, Pham M, Heiland S, Bäumer P. **Diffusion tensor imaging adds diagnostic accuracy in magnetic resonance neurography**. *Invest. Radiol.* (2015.0) **50** 498-504. DOI: 10.1097/RLI.0000000000000156 41. Griffiths TTFR, Teh I, Haroon HA, Shelley D, Plein S, Bourke G, Wade RG. **Diffusion tensor imaging in cubital tunnel syndrome**. *Sci. Rep.* (2021.0) **11** 14982. DOI: 10.1038/s41598-021-94211-7 42. Kronlage MPK, Schwarz D, Godel T, Heiland S, Yoon MS, Bendszus M, Bäumer P. **Diffusion tensor imaging in chronic inflammatory demyelinating polyneuropathy: Diagnostic accuracy and correlation with electrophysiology**. *Invest. Radiol.* (2017.0) **52** 701-707. DOI: 10.1097/RLI.0000000000000394 43. Kronlage M, Schwehr V, Schwarz D. **Peripheral nerve diffusion tensor imaging (DTI): Normal values and demographic determinants in a cohort of 60 healthy individuals**. *Eur. Radiol.* (2018.0) **28** 1801-1808. DOI: 10.1007/s00330-017-5134-z 44. Breitenseher JKG, Hold A, Berzaczy D, Nemec SF, Sycha T, Weber M, Prayer D, Kasprian G. **MR neurography of ulnar nerve entrapment at the cubital tunnel: a diffusion tensor imaging study**. *Eur. Radiol.* (2015.0) **25** 1911-1918. DOI: 10.1007/s00330-015-3613-7 45. Edward R, Abdelalim AM, Ashour AS. **A study of diffusion tensor imaging of median nerve in diabetic peripheral neuropathy**. *Egypt. J. Neurol. Psychiatry Neurosurg.* (2020.0) **56** 42. DOI: 10.1186/s41983-020-00172-5 46. Jambawalikar SBJ, Button TM, Li H, Geronimo V, Gould ES. **Diffusion tensor imaging of peripheral nerves**. *Skeletal Radiol.* (2010.0) **39** 1073-1079. DOI: 10.1007/s00256-010-0974-5 47. Jengojan SK, Florian M, Breitenseher J, Weber M, Prayer D, Kasprian G. **Acute radial nerve entrapment at the spiral groove: Detection by DTI-based neurography**. *Eur. Radiol.* (2015.0) **25** 1678-1683. DOI: 10.1007/s00330-014-3562-6 48. Kabakci N, Bengi G, Firat Z, Bayram A, Uluğ AM, Kovanlikaya A, Kovanlikaya I. **Diffusion tensor imaging and tractography of median nerve: normative diffusion values**. *AJR Am. J. Roentgenol.* (2007.0) **189** 923-927. DOI: 10.2214/AJR.07.2423 49. Park S-YK, Hye S, Lee I, Lee K, Lee Y. **Determination of optimum pixel size and slice thickness for tractography and ulnar nerve diffusion tensor imaging at the cubital tunnel using 3T MRI**. *Acta Radiol. Stockh. Swed.* (2020.0) **62** 1063-1071 50. Zhou YKM, Patel VS, Sheikh KA, Narayana PA. **Diffusion tensor imaging of forearm nerves in humans**. *J. Magn. Reson. Imaging JMRI.* (2012.0) **36** 920-927. DOI: 10.1002/jmri.23709 51. Zhou YNPA, Kumaravel M, Athar P, Patel VS, Sheikh KA. **High resolution diffusion tensor imaging of human nerves in forearm**. *J. Magn. Reson. Imaging JMRI.* (2014.0) **39** 1374-1383. DOI: 10.1002/jmri.24300 52. 52.D’Souza, A.W.C., Tu, S., Soligo, D.J., Kiernan, M.C., Barnett, M., & Calamante, F. A robust framework for characterising diffusion metrics of the median and ulnar nerves: Exploiting state-of-the-art tracking methods. J. Peripher. Nerv. Syst. JPNS. (2021). 53. Arnon SDN, Pasternak O, Graif M, Patish H, Schwimmer E, Ziv E, Assaf Y. **Diffusion tensor imaging of the median nerve in healthy and carpal tunnel syndrome subjects**. *J. Magn. Reson. Imaging JMRI.* (2009.0) **29** 657-662. DOI: 10.1002/jmri.21553 54. 54.Ho, M.J., Ulrike, H., Steigmiller, K., Manoliu, A., Schiller, A., Hinzpeter, R., Lanz, C., Martinoli, C., Jung, H.H., & Petersen, J.A. Comparison of electrodiagnosis, neurosonography and MR neurography in localization of ulnar neuropathy at the elbow. J. Neuroradiol. (2021). 55. 55.Leemans, A, Jeurissen, B., Sijbers, J., et al. ExploreDTI: A graphical toolbox for processing, analyzing, and visualizing diffusion MR data. 2009. Available at: https://www.semanticscholar.org/paper/ExploreDTI%3A-a-graphical-toolbox-for-processing%2C-and-Leemans-Jeurissen/f92f4418d4a4523a5760414144ecce8ab6472fde. Accessed March 24, 2022. 56. Watanabe M, Sakai O, Ozonoff A. **Age-related apparent diffusion coefficient changes in the normal brain**. *Radiology* (2013.0) **266** 575-582. DOI: 10.1148/radiol.12112420 57. Verdú E, Ceballos D, Vilches JJ. **Influence of aging on peripheral nerve function and regeneration**. *J. Peripher. Nerv. Syst.* (2000.0) **18** 1 58. Qin W, Shui YuC, Zhang F. **Effects of echo time on diffusion quantification of brain white matter at 1.5T and 3.0T**. *Magn. Reson. Med.* (2009.0) **61** 755-760. DOI: 10.1002/mrm.21920 59. Wharton S, Bowtell R. **Effects of white matter microstructure on phase and susceptibility maps: Effects of microstructure on phase and QSM**. *Magn. Reson. Med.* (2015.0) **73** 1258-1269. DOI: 10.1002/mrm.25189 60. Foesleitner O, Sulaj A, Sturm V. **Diffusion MRI in peripheral nerves: Optimized**. *Radiology* (2022.0) **302** 153-161. DOI: 10.1148/radiol.2021204740 61. 61.Maffei, C. New insights from the IronTract challenge: Simple post-processing enhances the accuracy of diffusion tractography. Abstr. 2021 ISMRM SMRT Annu. Meet. Exhib. 62. 62.Jenkins, J., Chang, L.-C., Hutchinson, E., et al. Harmonization of methods to facilitate reproducibility in medical data processing: Applications to diffusion tensor magnetic resonance imaging. In: 2016 IEEE International Conference on Big Data (Big Data). Washington DC, USA: IEEE; 2016:3992–3994. Available at: http://ieeexplore.ieee.org/document/7841086/. Accessed April 8, 2022. 63. Jovicich J, Barkhof F, Babiloni C. **Harmonization of neuroimaging biomarkers for neurodegenerative diseases: A survey in the imaging community of perceived barriers and suggested actions**. *Alzheimers Dement. Diagn. Assess. Dis. Monit.* (2019.0) **11** 69-73 64. Pinto MS, Paolella R, Billiet T. **Harmonization of brain diffusion MRI: Concepts and methods**. *Front. Neurosci.* (2020.0) **14** 396. DOI: 10.3389/fnins.2020.00396
--- title: Phase separation of α-crystallin-GFP protein and its implication in cataract disease authors: - Jie Shi - Ya-Xi Zhu - Rui-Yan Huang - Shao-Mei Bai - Yu-Xing Zheng - Jian Zheng - Zhao-Xia Xia - Yun-Long Wang journal: Scientific Reports year: 2023 pmcid: PMC10039050 doi: 10.1038/s41598-023-31845-9 license: CC BY 4.0 --- # Phase separation of α-crystallin-GFP protein and its implication in cataract disease ## Abstract Cataract, the leading cause of blindness worldwide, is caused by crystallin protein aggregation within the protected lens environment. Phase separation has been implicated as an important mechanism of protein aggregation diseases, such as neurodegeneration. Similarly, cataract has been proposed to be a protein condensation disease in the last century. However, whether crystallin proteins aggregate via a phase separation mechanism and which crystallin protein initiates the aggregation remain unclear. Here, we showed that all types of crystallin-GFP proteins remain soluble under physiological conditions, including protein concentrations, ion strength, and crowding environments. However, in age or disease-induced aberrant conditions, α-crystallin-GFP, including αA- and αB-crystallin-GFP, but not other crystallin-GFP proteins, undergo phase separation in vivo and in vitro. We found that aging-related changes, including higher crystallin concentrations, increased Na+, and decreased K+ concentrations, induced the aggregation of α-crystallin-GFP. Furthermore, H2O2, glucose, and sorbitol, the well-known risk factors for cataract, significantly enhanced the aggregation of αB-crystallin-GFP. Taken together, our results revealed that α-crystallin-GFP forms aggregates via a phase transition process, which may play roles in cataract disease. Opposite to the previously reported function of enhancing the solubility of other crystallin, α-crystallin may be the major aggregated crystallin in the lens of cataract patients. ## Introduction Cataract is the leading cause of blindness, which accounts for one third of visual impairments and about half of blindness cases worldwide1. Cataract is caused by the aggregation of crystallin proteins in the eye lens. The lens is a nearly transparent biconvex structure that is suspended behind the iris of each eye and it is partially responsible for the focusing of light onto the retina. Therefore, the insoluble protein aggregates in the lens block the light and impair visibility. Crystallin proteins consists of the members of the α-, β-, and γ-crystallin families, and they account for $90\%$ of the proteins in the mature lens2. α-crystallin is a member of the small heat shock protein family and it serves as an ATP-independent chaperone that efficiently binds to damaged or partially unfolded proteins to prevent widespread protein aggregation. β- and γ-crystallin have a common two-domain structure comprising four repeated “Greek key” motifs, which are essential for their high stability3. All of these types of crystallin proteins have been found in the aggregates of cataract patients. Although several risk factors for cataract have been identified, including oxidation environments, diabetes, and radiation exposure4, the detailed mechanism underlying crystallin protein aggregation remains poorly understood. In the last decade, phase separation has been implicated as an important mechanism of protein aggregation in neurodegenerative disease, such as Parkinson’s and Alzheimer’s diseases5. For example, FUS, a protein with a prion-like domain at its N-terminal, undergoes liquid–liquid phase separation to form highly concentrated droplets or condensates in cells. This condensate has liquid properties and can be resolved again. However, when age increases, the liquid-like FUS condensates will develop a solid-like phase or aggregates, which are hard to be resolved6. Similar to neurodegenerative diseases, cataract has been proposed to be a protein condensation disease in the last century, due to observations that crystallin proteins could separate into a protein-rich region and protein-poor region7. However, whether crystallin proteins aggregate via a phase separation mechanism remains unclear. Here, our data indicate that α-crystallin-GFP proteins form aggregates via a phase separation mechanism in vitro and in vivo. Specifically, both αA- and αB-crystallin-GFP proteins form aggregates under the condition of aging-related cataracts. Whereas, under the conditions of oxidation- or diabetes-induced cataracts, αB-crystallin-GFP is the major aggregated crystallin. ## α-Crystallin-GFP forms puncta in SRA01/04 and HLE-B3 cells To examine whether crystallin proteins underwent phase separation, members of the crystallin-GFP proteins were ectopically expressed in SRA$\frac{01}{04}$ cells, an immortalized human lens epithelial cell line. Interestingly, only αA- and αB-crystallin-GFP formed high concentrated puncta, whereas other crystallin-GFP remained diffuse in cells (Fig. 1a). The formation of αA- and αB-crystallin-GFP puncta was further verified in another human lens epithelial cell line HLE-B3 (Fig. 1b). A fluorescence recovery after photobleaching (FRAP) assay was performed to examine the dynamic exchange between puncta and diffused phase. αA- and αB-crystallin-GFP puncta were photobleached and continuously observed for 120 s. Surprisingly, almost no fluorescence recovery of αA- or αB-crystallin-GFP was found after photobleaching (Fig. 1c), suggesting that those condensates were solid-like phases or aggregates. We next asked whether α-crystallin-GFP condensates exhibited features of aggresomes, the pericentriolar accumulations of aggregated protein8. As shown in Fig. 1d, α-crystallin puncta did not co-localize with vimentin and dynein, two known components of aggresomes, suggesting that α-crystallin-GFP condensates were not aggresomes. Surprisingly, immunofluorescence assays showed that endogenous αA- and αB-crystallin did not form puncta in SRA$\frac{01}{04}$ and HLE-B3 cells (Fig. 1e). We thought the different phenotypes between endogenously and exogenously expressed αA- and αB-crystallin may be ascribed to the lower level of endogenous αA- and αB-crystallin. Consistently, reducing the amount of plasmid significantly decreased the puncta formation of exogenously expressed αA- and αB-crystallin-GFP (Fig. 1f).Figure 1α-Crystallin-GFP form puncta in lens epithelial cells. ( a,b) Ectopic expression of crystallin-GFP recombinant proteins in SRA$\frac{01}{04}$ and HLE-B3 cells. ( c) FRAP assay showed that αA- and αB-crystallin-GFP condensates contained no mobile fraction in SRA$\frac{01}{04}$ cells. The bleached punctum was labeled by red circle and unbleached control punctum were labeled by green circle. Data are mean ± SD from 3 independent puncta. Three bleached puncta were included. ( d) αA- and αB-crystallin-GFP puncta were not co-localized with vimentin and dynein in SRA$\frac{01}{04}$ cells. For (a–d), cells were transfected with 1 μg plasmids. ( e) Immunofluorescence of endogenous αA- and αB-crystallin in SRA$\frac{01}{04}$ and HLE-B3 cells. ( f) Overexpression of αA- and αB-crystallin-GFP recombinant proteins in different amounts of plasmid (0.5 μg, 1 μg and 2 μg) in SRA$\frac{01}{04}$ and HLE-B3 cells. Data are the mean ± SD. $$n = 10$$ images for each group. ** $P \leq 0.01$; ***$P \leq 0.001.$ ## Crystallin-GFP remains soluble in physiological conditions in vitro To verify the aggregation of crystallin proteins in vitro, we expressed and purified most types of crystallin-GFP recombinant proteins from E. coli, including αA-, αB-, βA1/A3-, βA2-, βA4-, βB1-, βB2-, βB3-, γA-, γB-, γC-, γD-, and γS-crystallin-GFP proteins (Fig. 2a). It has been reported that in normal lenses, crystallin proteins were dissolved in a buffer containing 20 mM Na+, 10 μM Ca2+, and 120 mM K+ (physiological buffer)9,10. The purified recombinant proteins were diluted in a physiological buffer to a final concentration of 2.7 μM (αA/B), 4.0–4.5 μM (βA/B), and 0.13 μM (γA/B/C/D/S). The concentration of these purified crystallin-GFP proteins were calculated according to the mass percentage of crystallin, which accounted for ∼$90\%$ of the ocular proteins in the lens2. In agreement with the high solubility of crystallin, all crystallin-GFP proteins remained soluble and no aggregates formed (Fig. 2b,c). To verify the quality of purified proteins, the chaperone activity of purified αA- and αB-crystallin-GFP was examined. As shown in Fig. 2d, both αA- and αB-crystallin-GFP prevented the DTT-induced aggregation of insulin, indicating good quality purified proteins. Figure 2In vitro aggregation of recombinant crystallin-GFP proteins in physiological buffer. ( a) The purified crystallin-GFP recombinant proteins were verified with Coomassie Staining, including αA-, αB-, βA1/A3-, βA2-, βA4-, βB1-, βB2-, βB3-, γA-, γB-, γC-, γD-, and γS-crystallin-GFP proteins. ( b) The liquid-to-solid phase transition of αA- and αB-crystallin-GFP proteins (2.7 μM, 5 μM, or 8.2 μM) were explored in physiological buffer (20 mM NaCl, 10 μM CaCl2, 20 mM KCl, 20 mM Tris–HCl pH7.4, 0.6 mM glucose, 12 mM glutathione, 1 mM vitamin C, and 5.9 mM inositol), and with or without $5\%$ PEG8000. The fluorescence intensity is presented as the area × mean intensity (A. × M.). Data are the mean ± SD. $$n = 3$$ images for each group. ** $P \leq 0.01.$ ( c) The aggregation of βA1/A3 (4.2 μM)-, βA2 (4.5 μM)-, βA4 (4.5 μM)-, βB1 (4.0 μM)-, βB2 (4.4 μM)-, βB3 (4.3 μM)-, γA (0.13 μM)-, γB (0.13 μM)-, γC (0.13 μM)-, γD (0.13 μM)-, and γS (0.13 μM)-crystallin-GFP recombinant proteins in vitro was examined in physiological buffer, with or without $5\%$ PEG8000. For (b,c), the incubation time was 10 min at 4 °C. ( d) An insulin aggregation assay of chaperone activity of purified recombinant αA- or αB-crystallin-GFP proteins at 25 °C. Assay mixture (200 μl) contained insulin (0.35 mg/mL), αA-crystallin-GFP (1.05 mg/mL), or αB-crystallin-GFP (0.35 mg/mL) or both of αA- and αB-crystallin-GFP, and 20 mM DTT in 50 mM PBS (pH 7.2) were monitored after the indicated incubation time (0, 15, 30, 45, or 60 min). ## α-Crystallin-GFP is the major aggregated crystallin of aging-related cataracts The crowding condition in the lens increases as age increases and this is a risk factor of crystallin protein aggregation11. Consistently, when PEG8000 was added to the solution to mimic the crowding environment in lens12,13, αA-, αB-, and βA2-crystallin-GFP proteins became aggregated, whereas other crystallin-GFP proteins remained soluble (Fig. 2b,c), suggesting that aggregation of αA-, αB-, and βA2-crystallin-GFP may be an early event of cataract. The concentration of crystallin proteins in human lenses is extremely high (about 450 mg/ml)2. When we increased the concentration of recombinant αA- or αB-crystallin-GFP proteins in the solution, the aggregates and the opacity were significantly increased (Figs. 2b, 3a, S1b,c). Figure 3In vitro aggregation of recombinant crystallin-GFP proteins in pathological buffer. ( a) The liquid-to-solid phase transition of αA- and αB-crystallin-GFP proteins (2.7 μM, 5 μM, or 8.2 μM) were detected in pathological buffer (150 mM NaCl, 30 mM CaCl2, 5 mM KCl, 20 mM Tris–HCl pH7.4, 0.6 mM glucose, 12 mM glutathione, 1 mM vitamin C, and 5.9 mM inositol), and with or without $5\%$ PEG8000. The fluorescence intensity is presented as the area × mean intensity (A. × M.). Data are the mean ± SD. $$n = 3$$ images for each group. ( b) The aggregation of βA1/A3 (4.2 μM)-, βA2 (4.5 μM)-, βA4 (4.5 μM)-, βB1 (4.0 μM)-, βB2 (4.4 μM)-, βB3 (4.3 μM)-, γA (0.13 μM)-, γB (0.13 μM)-, γC (0.13 μM)-, γD (0.13 μM)-, and γS (0.13 μM)-crystallin-GFP recombinant proteins in vitro was examined in pathological buffer, with or without $5\%$ PEG8000. For (a,b), the incubation time was 10 min at 4 °C. * $P \leq 0.05$; **$P \leq 0.01.$ Disruption of lens epithelium due to increasing age or radiation-induced injury, the concentration of Na+ and Ca2+ will increase and K+ will decrease in the lens nucleus, which is one of the risk factors for cataract14. Consistently, the aggregation of recombinant αA-, αB-, and βA2-crystallin-GFP proteins largely increased in the solution containing high Na+, Ca2+, and low K+ (pathological buffer; Fig. 3a). Interestingly, several crystallin-GFP proteins that are soluble in physiological buffer formed aggregates in pathological buffer (Fig. 3b), indicating ion conditions are a strong inducer of crystallin aggregations. On the other hand, when we treated the crystallin-GFP-overexpressed lens epithelial cells with pathological buffer, only αA- and αB-crystallin-GFP proteins showed enhanced aggregation, while endogenous αA- and αB-crystallin remained soluble under a pathological environment (Figs. 4a–e, S1a). Most importantly, aggregated αA- and αB-crystallin proteins were also observed in aging-related cataractous lenses of patients (Fig. 4f). Figure 4Live cell image of crystallin-GFP proteins in SRA$\frac{01}{04}$ cells in physiological or pathological buffer. ( a,b) The crystallin-GFP protein-expressed SRA$\frac{01}{04}$ cells were incubated in physiological buffer. ( c,d) The crystallin-GFP protein-expressed SRA$\frac{01}{04}$ cells were incubated in pathological buffer. For (a–d), cells were transfected with the indicated plasmid (1 μg) for 24 h, following by incubation with physiological buffer or pathological buffer for 30 min before imaging. ( e) The fluorescence intensity of the α-crystallin-GFP puncta is presented as the area of the puncta. Data are the mean ± SD. $$n = 10$$ images for each group. Phy physiological buffer, Path pathological buffer. ( f) Immunofluorescence of αA- and αB-crystallin in aging-related cataractous lens capsular epithelia. White arrows indicate αA- and αB-crystallin aggregates. ( g) Treatment of mouse lenses with 10 mU/ml GO induced in vitro cataract. ( h) Immunofluorescence of αA- and αB-crystallin in GO-treated mouse lenses. *** $P \leq 0.001.$ ## αB-crystallin-GFP is the major aggregated crystallin in oxidation- or diabetes-induced cataracts Oxidation is a major cause of age-related crystallin aggregation15. However, adding H2O2 to crystallin protein solutions only slightly promoted aggregation of αB-crystallin-GFP (Fig. 5a,d). Diabetes is a risk factor for cataract, which is mainly ascribed to hyperglycemia and the production of sorbitol16,17. Adding high glucose and sorbitol to crystallin protein solutions slightly promoted aggregation of αB-crystallin-GFP, but had no impact on αA-crystallin-GFP (Fig. 5b,c,e,f). The liquid-to-solid phase separation will increase along with time accumulation. Interestingly, as time increased, the aggregation of αB-crystallin-GFP was significantly enhanced, whereas αA-crystallin-GFP remained soluble (Fig. 5a–f), in the presence of H2O2, high glucose, or sorbitol. Consistently, adding H2O2, high glucose, or sorbitol to culture medium significantly enhanced the formation of αB-crystallin-GFP puncta in SRA$\frac{01}{04}$ cells, while endogenous αA- and αB-crystallin remained diffuse (Figs. 5g, S1a). Additionally, the aggregation of α-crystallin (mostly αB-crystallin) was also observed in glucose oxidase (GO)-induced cataracts of mouse lenses (Fig. 4g,h), indicating that the aggregation of α-crystallin was an early event of cataract. Figure 5αB-crystallin-GFP is the major aggregated crystallin in oxidation- or diabetes-induced cataracts. ( a–c) H2O2 (100 μM or 200 μM), high glucose (20 mM or 40 mM), and sorbitol (30 mM or 60 mM) slightly enhanced αB-crystallin-GFP (2.7 μM) but not αA-crystallin-GFP (2.7 μM) aggregation after incubation for 10 min at 4 °C in vitro. ( a–c) αB-crystallin-GFP (2.7 μM) aggregation accumulated along with increases in the incubation times (1 h, 12 h, and 24 h), whereas αA-crystallin-GFP (2.7 μM) remained soluble. For (a–c), an in vitro assay was performed in physiological buffer with $5\%$ PEG8000. ( d-f) The fluorescence intensity is presented as the area × mean intensity (A. × M.). Data are the mean ± SD. $$n = 3$$ images for each group. ( g) After transfection with αA- or αB-crystallin-GFP plasmids (1 μg) for 6 h, SRA$\frac{01}{04}$ cells were treated with H2O2 (200 μM), high glucose (40 mM), and sorbitol (60 mM) for 24 h before imaging. The fluorescence intensity of α-crystallin-GFP puncta is presented as the area of the puncta. Data are the mean ± SD. $$n = 10$$ images for each group. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$, ns no significance. ## Discussion Cataract, one of the leading causes of blindness worldwide, is the result of crystallin protein aggregation7. Early research showed that under pathological conditions, such as aging or radiation exposure, crystallin proteins could spatially separate into protein-rich regions and protein-poor regions, causing opacification and thus visual impairment7,18. In this study, we demonstrated that α-crystallin-GFP proteins formed aggregates via a phase separation mechanism. Previous studies reported that α-crystallin, a member of the molecular chaperones19, prevents aberrant aggregation of damaged β- and γ-crystallin by interacting with the client protein using a variety of binding modes20. α-crystallin chaperone activity can be compromised by mutation or posttranslational modifications, leading to large-scale crystallin aggregation and cataract formation21. Surprisingly, in this study, we found that α-crystallin-GFP, without changes such as mutation or modification, could form condensates upon several risk factor stimulations. These observations suggested that α-crystallin may be the major aggregated crystallin in the early stage of cataract disease. In the last century, cataract has been designated as a molecular condensation disease. Our results showed that although crystallin proteins remain soluble under normal conditions, aberrant crystallin condensates were largely induced under pathological conditions, such as aging and diabetes. Such aberrant condensates are also involved in neurodegenerative diseases22–24. For example, liquid droplets of FUS protein convert with time from a liquid to an aggregated state22. These findings indicated that aberrant phase transitions within liquid-like compartments are central for age-related cataracts. Previous studies have determined the phase separation of a protein-water mixture in cold cataract and selenite-induced cataract, which was associated with abnormal variation in temperature18,25. Annunziata et al. reported that γS-crystallin underwent liquid–liquid phase separation (LLPS), but this process needed an extremely low temperature (as low as − 8 °C)26. We did not observe the LLPS of γS-crystallin at room temperature, which was more like the situation peoples met. We thought that the low temperature might increase the multivalent interaction among γS-crystallin proteins, thereby accelerating the LLPS of γS-crystallin. Previous reports suggested that protein oxidation can lead to formation of insoluble, light-scattering protein aggregates27. Another main risk factor of age-related cataract is diabetes28. For diabetes-related cataract, increased glucose and sorbitol concentrations in the lens are major initiators for crystallin aggregation17. Here, we found that although H2O2, glucose, or sorbitol only slightly promoted aggregation of αB-crystallin-GFP within a short time in vitro, as time increased, H2O2, glucose, or sorbitol significantly enhanced aggregation of αB-crystallin-GFP, whereas αA-crystallin-GFP remained soluble regardless the incubation time. These results, coupled with previous reports, illustrate that early oxidative and diabetic damage in crystallin proteins may be spontaneously reversed if oxidant and sorbitol are removed in time. There are some shortages in this study. Firstly, all in vitro experiments were performed with crystallin-GFP protein. The big GFP tag may influence the properties of crystallin protein. Additionally, it is also reported that α-crystallin binding to lens membrane contributes to cataract formation29, and whether the aggregated α-crystallin bind to the lens membrane remains unclear. Finally, it needs to be further elucidated whether α-crystallin aggregation associates with other well-known causes of cataract, such as genetics, high myopia, smoking, medications, significant alcohol consumption, obesity, and hypertension30. ## Cell culture Immortalized human lens epithelial cell line SRA$\frac{01}{04}$ was purchased from Shanghai Baifeng Biotech Co., Ltd. Another immortalized human lens epithelial cell line HLE-B3 was a gift form Prof. Ming-Xing Wu (State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-Sen University). Both cell lines were mycoplasma-free and were authenticated using STR profiling by Guangzhou Cellcook Biotech Co., Ltd or American Type Culture Collection (ATCC). SRA$\frac{01}{04}$ and HLE-B3 were maintained at 37℃ in a $5\%$ CO2 atmosphere and cultured in Dulbecco’s modified Eagle’s medium (DMEM, low glucose, Gibco, ThermoFisher Scientific, Waltham, MA, USA) supplemented with $10\%$ fetal bovine serum (FBS, Gibco) and 100 units/mL penicillin–streptomycin (15,140,122, HyClone, South Logan, UT, USA). Cells were grown to 50–$60\%$ confluence before transfection with plasmids using Lipofectamine 2000 transfection reagents (ThermoFisher Scientific) according to the manufacturer’s instructions. For crystallin aggregation induction, cells were incubated with complete medium containing H2O2 (50 μM or 200 μM), glucose (10 mM or 40 mM), and sorbitol (15 mM or 60 mM) for 24 h, or treated with physiological buffer (20 mM NaCl, 10 μM CaCl2, 120 mM KCl, 20 mM Tris–HCl pH7.4, 0.6 mM glucose, 12 mM glutathione, 1 mM vitamin C, 5.9 mM inositol) or pathological buffer (150 mM NaCl, 30 mM CaCl2, 5 mM KCl, 20 mM Tris–HCl pH7.4, 0.6 mM glucose, 12 mM glutathione, 1 mM vitamin C, 5.9 mM inositol)9,10,31 for 30 min at 37 °C before subsequent analysis. ## Plasmid construction The cDNA encoding crystallin was cloned into pGEX-6P-1 or pcDNA3.0 vectors. The cDNA fragments encoding our proteins of interest were generated with PCR using a NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs, Beijing, China) and inserted in-frame before EGFP using the restriction enzyme sites including BamH I and EcoR I. Plasmid inserts were confirmed by Sanger sequencing (Tsingke, Guangzhou, China) and reading the full length of the insert. The human crystallin genes used in this study were as follows: α-crystallin (αA, αB), β-crystallin (βA1/A3, βA2, βA4, βB1, βB2, and βB3), and γ-crystallin (γA, γB, γC, γD, γN, and γS). ## Live cell imaging and fluorescence recovery after photobleaching (FRAP) SRA$\frac{01}{04}$ or HLE-B3 cells were: [1] seeded on glass plates and transfected with crystallin-GFP plasmids for 24 h; [2] incubated with H2O2 (200 μM), glucose (40 mM), and sorbitol (60 mM) for 24 h, or [3] treated with the physiological buffer for 30 min before imaging; then, the physiological buffer was discarded and cells were further incubated with the pathological buffer for another 30 min before imaging. Confocal images were taken with a Zeiss LSM880 confocal microscope with a 488 nm laser using a 60X oil immersion lens. Images were processed with ZEN software (Blue edition, 3.1). Fluorescence intensity was measured with Image J. For FRAP experiments, the green puncta were bleached with $100\%$ laser power (488 nm), and time-lapse images were captured every 1 s. Images were further processed using ZEN3.1, and the fluorescence intensity was normalized to the prebleaching time points. GraphPad Prism was used to plot and analyze the FRAP results. ## Protein purification Crystallin-GFP recombinant proteins were expressed in *Escherichia coli* BL21(DE3). E. coli cells were grown to OD600 of 0.6 at 37 °C and induced with 0.5 mM IPTG (R0393, Invitrogen, Waltham, MA, USA) at 16 °C for 16 h. Cells were harvested by centrifugation at 4000g for 10 min at 4 °C, resuspended in 1 × PBS (supplemented with 1 mM PMSF), and then lysed by sonication. Lysates were centrifuged twice at 10,000g for 20 min at 4 °C. The supernatant was subjected to the purification of crystallin-GFP proteins using a GST-tag protein purification kit (P2262, Beyotime, Shanghai, China) according to the manufacturer’s protocol. Consequently, the eluted proteins were confirmed by SDS-PAGE and stored at − 80 °C. ## In vitro aggregation For in vitro aggregation experiments, purified crystallin-GFP proteins were diluted to the indicated concentrations in physiological buffer or pathological buffer with or without $5\%$ PEG8000, and then incubated at 4 °C for 10 min. In vitro aggregation experiments were also performed to investigate crystallin protein aggregation in response to H2O2 (100 μM, 200 μM), glucose (20 mM, 40 mM), and sorbitol (30 mM, 60 mM). Finally, 10 μL of each mixture were placed on a glass slide or a 384 well glass bottom plate for imaging with a Zeiss LSM880 confocal microscope. All images were processed with ZEN3.1. Fluorescence intensity was measured with Image J. For opacity analysis, we performed 200 μL reaction mixtures and measured the apparent absorbance at 400 nm to detect the opacity of α-crystallin aggregates in vitro. ## Assay of chaperone activity Chaperone activity of purified recombinant αA- and αB-crystallin-GFP proteins was measured at 25 °C using an insulin B-chain aggregation assay as described previously32. Briefly, insulin (0.35 mg/mL, in 50 mM PBS pH 7.2) was reduced with 20 mM DTT. Aggregation was monitored in the presence of αA-crystallin (1.05 mg/mL), αB-crystallin (0.35 mg/mL), or both αA- and αB-crystallin in a 96-well plate by measuring the apparent absorbance at 400 nm after the indicated incubating time (0, 15, 30, 45, or 60 min). ## Collection of human lens capsular epithelial samples Collection of human capsular epithelia from cataract lenses was approved by the Institutional Research Ethics Committee of the Sixth Affiliated Hospital of Sun Yat-sen University. Informed consent was obtained from each of the cataract patients. All procedures followed the ethical principles of the World Medial Association (WMA) Declaration of Helsinki. The lens capsules from 10 cataract patients were collected at surgery by the physicians in Guangdong Provincial People’s Hospital. The clinical classifications of cataract patients are summarized in Table 1. Cataract grade was evaluated according to the Lens Opacities Classification System III33.Table 1Cataract patients’ information. NumberGenderAgeClassification1Male55C2N22Female58C2N23Female76C2N24Male61C2N2P25Male68C2N36Male71C3N27Male67C2N3P18Male55C2N3P39Male44C3N4P210Female63C3N5P2C cortex, N nuclei, P posterior capsule. ## Animals Animal experiments were approved by the Institutional Animal Care and Use Committee of the Sixth Affiliated Hospital of Sun Yat-sen University. The experimental procedures with animals complied with ARRIVE guidelines and were performed in accordance with the U.K. Animals (Scientific Procedures) Act, 1986. Four-week-old male C57BL/6 J mice were purchased from Gempharmatech-GD (Guangdong, China). The eyeballs of the mice were removed and the lenses were carefully dissected after sacrifice with CO2 inhalation. Dissected lenses were placed in a 10-cm dish containing 20 ml Medium 199 (M4530, Sigma-Aldrich), and incubated at 37 °C in a $5\%$ CO2 atmosphere for 12 h. Then, three transparent lenses were transferred into a 6-cm dish and incubated with 8 ml Medium 199 containing 10 mU/ml glucose oxidase (GO, G7141, Sigma-Aldrich)34, which continuously generated oxidative stress and induced crystallin aggregation. The morphological and crystallin protein changes of the lenses were analyzed at 0 and 24 h after GO treatment. ## Immunofluorescence analysis SRA$\frac{01}{04}$ or HLE-B3 cells seeded on glass coverslips were fixed with $4\%$ paraformaldehyde for 15 min. Frozen human lens capsular epithelia from cataract lenses and the entire GO-treated mouse lenses were fixed with cold acetonum for 10 min. After fixation, cells or samples were incubated with blocking buffer ($5\%$ goat serum, $0.3\%$ Triton X-100 in 1 × PBS) for 1 h and primary antibodies containing blocking buffer for 2 h at room temperature. After three washes in 1 × PBS, cells or samples were incubated with secondary antibodies tagged with Alexa Fluor 488, 555, or 647 (4408S, 4413S, or 4414S, Cell Signalling Technology) for 1 h at room temperature in the dark, following by DAPI staining (D9542, Sigma-Aldrich) for 5 min. Images were acquired using a Zeiss LSM880 confocal microscope and processed with ZEN3.1. The primary antibodies used in immunofluorescence analysis included: CRYAA (A5725, ABclonal), CRYAB (A9633, ABclonal), Vimentin (A19607, ABclonal), and DYNC1H1 (12345-1-AP, Proteintech). ## Statistical analysis All data were expressed as mean ± standard deviation (SD) of independent experiments performed in triplicate. Statistical analyses were performed with SPSS 20.0 software (SPSS, Inc., Chicago, IL). Unpaired t test was used to assess the difference between two groups and one-way analysis of variance were used when more than two groups were compared. The p-value < 0.05 was considered statistically significant. *, **, and *** represent $p \leq 0.05$, $p \leq 0.01$, and $p \leq 0.001$, respectively. ## Supplementary Information Supplementary Figures. The online version contains supplementary material available at 10.1038/s41598-023-31845-9. ## References 1. Song P, Wang H, Theodoratou E, Chan KY, Rudan I. **The national and subnational prevalence of cataract and cataract blindness in China: A systematic review and meta-analysis**. *J. Glob. Health* (2018.0) **8** 10804. DOI: 10.7189/jogh.08.010804 2. He Y, Kang J, Song J. **ATP differentially antagonizes the crowding-induced destabilization of human γS-crystallin and its four cataract-causing mutants**. *Biochem. Biophys. Res. Commun.* (2020.0) **533** 913-918. DOI: 10.1016/j.bbrc.2020.09.090 3. Gupta R, Srivastava K, Srivastava OP. **Truncation of motifs III and IV in human lens betaA3-crystallin destabilizes the structure**. *Biochemistry* (2006.0) **45** 9964-9978. DOI: 10.1021/bi060499v 4. Finger PT, Chin KJ, Yu GP, Patel NS. **Risk factors for cataract after palladium-103 ophthalmic plaque radiation therapy**. *Int. J. Radiat. Oncol. Biol. Phys.* (2011.0) **80** 800-806. DOI: 10.1016/j.ijrobp.2010.02.051 5. Kanaan NM, Hamel C, Grabinski T, Combs B. **Liquid-liquid phase separation induces pathogenic tau conformations in vitro**. *Nat. Commun.* (2020.0) **11** 2809. DOI: 10.1038/s41467-020-16580-3 6. Portz B, Lee BL, Shorter J. **FUS and TDP-43 phases in health and disease**. *Trends Biochem. Sci.* (2021.0) **46** 550-563. DOI: 10.1016/j.tibs.2020.12.005 7. Benedek GB. **Cataract as a protein condensation disease: The proctor lecture**. *Investig. Ophth. Vis. Sci.* (1997.0) **38** 1911-1921 8. Kawaguchi Y. **The deacetylase HDAC6 regulates aggresome formation and cell viability in response to misfolded protein stress**. *Cell* (2003.0) **115** 727-738. DOI: 10.1016/S0092-8674(03)00939-5 9. Garner WH, Hilal SK, Lee SW, Spector A. **Sodium-23 magnetic resonance imaging of the eye and lens**. *Proc. Natl. Acad. Sci. U. S. A.* (1986.0) **83** 1901-1905. DOI: 10.1073/pnas.83.6.1901 10. Dilsiz N, Olcucu A, Atas M. **Determination of calcium, sodium, potassium and magnesium concentrations in human senile cataractous lenses**. *Cell Biochem. Funct.* (2000.0) **18** 259-262. DOI: 10.1002/1099-0844(200012)18:4<259::AID-CBF881>3.0.CO;2-O 11. Yan H, Wang NL, Yao K. *Ophthalmology* (2016.0) 12. Sabari BR. **Coactivator condensation at super-enhancers links phase separation and gene control**. *Science* (2018.0) **361** eaar3958. DOI: 10.1126/science.aar3958 13. Grosas AB, Rekas A, Mata JP, Thorn DC, Carver JA. **The aggregation of αB-crystallin under crowding conditions is prevented by αA-crystallin: Implications for α-crystallin stability and lens transparency**. *J. Mol. Biol.* (2020.0) **432** 5593-5613. DOI: 10.1016/j.jmb.2020.08.011 14. Duncan G, Bushell AR. **Ion analyses of human cataractous lenses**. *Exp. Eye Res.* (1975.0) **20** 223-230. DOI: 10.1016/0014-4835(75)90136-0 15. Truscott RJ. **Age-related nuclear cataract-oxidation is the key**. *Exp. Eye Res.* (2005.0) **80** 709-725. DOI: 10.1016/j.exer.2004.12.007 16. Chung SS, Ho EC, Lam KS, Chung SK. **Contribution of polyol pathway to diabetes-induced oxidative stress**. *J. Am. Soc. Nephrol.* (2003.0) **14** S233-S236. DOI: 10.1097/01.ASN.0000077408.15865.06 17. Brownlee M, Cerami A. **The biochemistry of the complications of diabetes mellitus**. *Annu. Rev. Biochem.* (1981.0) **50** 385-432. DOI: 10.1146/annurev.bi.50.070181.002125 18. Tanaka T, Ishimoto C, Chylack LJ. **Phase separation of a protein-water mixture in cold cataract in the young rat lens**. *Science* (1977.0) **197** 1010-1012. DOI: 10.1126/science.887936 19. Horwitz J. **Alpha-crystallin can function as a molecular chaperone**. *Proc. Natl. Acad. Sci. U. S. A.* (1992.0) **89** 10449-10453. DOI: 10.1073/pnas.89.21.10449 20. Sprague-Piercy MA, Rocha MA, Kwok AO, Martin RW. **Α-crystallins in the vertebrate eye lens: Complex oligomers and molecular chaperones**. *Annu. Rev. Phys. Chem.* (2021.0) **72** 143-163. DOI: 10.1146/annurev-physchem-090419-121428 21. Carver JA, Ecroyd H, Truscott R, Thorn DC, Holt C. **Proteostasis and the regulation of intra- and extracellular protein aggregation by ATP-independent molecular chaperones: Lens α-crystallins and milk caseins**. *Acc. Chem. Res.* (2018.0) **51** 745-752. DOI: 10.1021/acs.accounts.7b00250 22. Patel A. **A liquid-to-solid phase transition of the ALS protein FUS accelerated by disease mutation**. *Cell* (2015.0) **162** 1066-1077. DOI: 10.1016/j.cell.2015.07.047 23. Wegmann S. **Tau protein liquid-liquid phase separation can initiate tau aggregation**. *Embo J.* (2018.0) **37** e98049. DOI: 10.15252/embj.201798049 24. Peskett TR. **A liquid to solid phase transition underlying pathological Huntingtin Exon1 aggregation**. *Mol. Cell.* (2018.0) **70** 588-601. DOI: 10.1016/j.molcel.2018.04.007 25. Clark JI, Steele JE. **Phase-separation inhibitors and prevention of selenite cataract**. *Proc. Natl. Acad. Sci. U. S. A.* (1992.0) **89** 1720-1724. DOI: 10.1073/pnas.89.5.1720 26. Annunziata O, Ogun O, Benedek GB. **Observation of liquid-liquid phase separation for eye lens gammaS-crystallin**. *Proc. Natl. Acad. Sci. U. S. A.* (2003.0) **100** 970-974. DOI: 10.1073/pnas.242746499 27. Moreau KL, King JA. **Protein misfolding and aggregation in cataract disease and prospects for prevention**. *Trends Mol. Med.* (2012.0) **18** 273-282. DOI: 10.1016/j.molmed.2012.03.005 28. Asbell PA. **Age-related cataract**. *Lancet* (2005.0) **365** 599-609. DOI: 10.1016/S0140-6736(05)70803-5 29. Grami V. **Alpha-crystallin binding in vitro to lipids from clear human lenses**. *Exp. Eye Res.* (2005.0) **81** 138-146. DOI: 10.1016/j.exer.2004.12.014 30. Ang MJ, Afshari NA. **Cataract and systemic disease: A review**. *Clin. Exp. Ophthalmol.* (2021.0) **49** 118-127. DOI: 10.1111/ceo.13892 31. 31.Bobrow, J. C. 2014-2015 Basic and Clinical Science Course(BCSC): Section 11: Lens and Cataract. American Academy of Ophthalmology (2014). 32. Das KP, Surewicz WK. **Temperature-induced exposure of hydrophobic surfaces and its effect on the chaperone activity of alpha-crystallin**. *FEBS Lett.* (1995.0) **369** 321-325. DOI: 10.1016/0014-5793(95)00775-5 33. Chylack LJ. **The lens opacities classification system III. The longitudinal study of cataract study group**. *Arch. Ophthalmol.* (1993.0) **111** 831-836. DOI: 10.1001/archopht.1993.01090060119035 34. Gong L. **Heterochromatin protects retinal pigment epithelium cells from oxidative damage by silencing P53 target genes**. *Proc. Natl. Acad. Sci. U. S. A.* (2018.0) **115** E3987-E3995. DOI: 10.1073/pnas.1715237115
--- title: Automated deep learning for classification of dental implant radiographs using a large multi-center dataset authors: - Wonse Park - Jong-Ki Huh - Jae-Hong Lee journal: Scientific Reports year: 2023 pmcid: PMC10039053 doi: 10.1038/s41598-023-32118-1 license: CC BY 4.0 --- # Automated deep learning for classification of dental implant radiographs using a large multi-center dataset ## Body Dental implants are among the most widely used and commonly accepted treatment modalities for oral rehabilitation of partially and completely edentulous patients1,2. The occurrence of various major or critical mechanical (such as fractures of screws or fixtures) and biological (such as peri-implantitis) problems is steadily and inevitably increasing, affecting long-term survival and reintervention outcomes3,4. Therefore, dental implant-related complications are a growing concern in the dental community worldwide and are a public health problem associated with a high socio-economic burden5,6. In particular, early detection and appropriate treatment of simple mechanical complications such as screw loosening can prevent more severe complications, such as fixture fracture or severe peri-implantitis, at an early stage7,8. For early and fast intervention, dental implant systems (DIS) placed in the oral cavity must be unambiguously identified and classified. However, in actual clinical practice, it is not easy to properly identify or classify the various different types of DIS after implant surgery because of various clinical and environmental factors, including the closure of a dental hospital or the loss of dental records. Although two-dimensional dental radiography, including panoramic and periapical radiographs, is the most useful tool for identifying and classifying DIS post-implant surgery, there is a fundamental and practical limit for classifying thousands of different types of DIS with similar shapes and physical properties9,10. In addition, three-dimensional cone-beam computed tomography (CBCT) has been actively used for dental implant surgery; however, whether CBCT can better classify DIS is debatable because the sharpness and resolution of CBCT is still significantly lower than that of peripheral radiographs11. Deep learning (DL), a subfield of artificial intelligence (AI), has a wide range of applications in medicine; this unique technology is associated with high accuracy in medical image analysis for edge detection, classification, or segmentation based on a cascade of multiple computational and hidden layers in a deep neural network12. When limited to dentistry, deep and convolutional neural networks have rapidly become the methodology of choice for two- and three-dimensional dental image analyses13–16. Several studies have demonstrated DL algorithms as an emerging state-of-the-art approach in terms of accuracy performance for identifying and classifying various types of DIS and often show outperforming results compared to dental professionals specialized in implantology17–25. However, since most previous studies were based on fewer than thousands of DIS images or fewer than 10 different types of DIS, available evidence is insufficient to be implemented in actual clinical practice19–24. This study aimed to evaluate the accuracy of the automated DL algorithm for the identification and classification of DIS using a large-scale and comprehensive multicenter dataset. ## Abstract This study aimed to evaluate the accuracy of automated deep learning (DL) algorithm for identifying and classifying various types of dental implant systems (DIS) using a large-scale multicenter dataset. Dental implant radiographs of pos-implant surgery were collected from five college dental hospitals and 10 private dental clinics, and validated by the National Information Society Agency and the Korean Academy of Oral and Maxillofacial Implantology. The dataset contained a total of 156,965 panoramic and periapical radiographic images and comprised 10 manufacturers and 27 different types of DIS. The accuracy, precision, recall, F1 score, and confusion matrix were calculated to evaluate the classification performance of the automated DL algorithm. The performance metrics of the automated DL based on accuracy, precision, recall, and F1 score for 116,756 panoramic and 40,209 periapical radiographic images were $88.53\%$, $85.70\%$, $82.30\%$, and $84.00\%$, respectively. Using only panoramic images, the DL algorithm achieved $87.89\%$ accuracy, $85.20\%$ precision, $81.10\%$ recall, and $83.10\%$ F1 score, whereas the corresponding values using only periapical images achieved $86.87\%$ accuracy, $84.40\%$ precision, $81.70\%$ recall, and $83.00\%$ F1 score, respectively. Within the study limitations, automated DL shows a reliable classification accuracy based on large-scale and comprehensive datasets. Moreover, we observed no statistically significant difference in accuracy performance between the panoramic and periapical images. The clinical feasibility of the automated DL algorithm requires further confirmation using additional clinical datasets. ## Results The performance metrics of the automated DL algorithm based on the accuracy, precision, recall, and F1 score for total of 156,965 panoramic and periapical radiographic images were $88.53\%$, $85.70\%$, $82.30\%$, and $84.00\%$, respectively. Using only panoramic images ($$n = 116$$,756), the DL algorithm achieved $87.89\%$ accuracy, $85.20\%$ precision, $81.10\%$ recall, and $83.10\%$ F1 score, whereas the corresponding values using only periapical images ($$n = 40$$,209) achieved $86.87\%$ accuracy, $84.40\%$ precision, $81.70\%$ recall, and $83.00\%$ F1 score, respectively. No statistically significant difference in the classification accuracy was observed between the three groups, and the detailed accuracy performances of DL in the classification of DIS are listed in Table 1.Table 1Accuracy performance of automated deep learning algorithm. ManufacturesSystemPanoramic images(Accuracy = $87.89\%$)Periapical images(Accuracy = $86.87\%$)Panoramic and Periapical images(Accuracy = $88.53\%$)Precision (%)Recall(%)F1 score(%)Precision (%)Recall(%)F1 score(%)Precision (%)Recall(%)F1 score(%)All dental implant systems85.2081.1083.1084.4081.7083.0085.7082.3084.00NeobiotechIS I88.3093.0090.6087.6093.0090.2091.3092.1091.70IS II66.7020.0030.8060.0030.0040.0066.7020.0030.80IS III87.8081.1084.3074.1075.5074.8079.4094.3086.20EB94.0090.4092.2098.0092.3095.0097.9088.5092.90Nobel biocareBranemark92.9076.5083.9096.0070.6081.40100.076.5086.70DentsplyAstra93.40100.096.6093.40100.096.6093.40100.096.60Xive98.50100.099.20100.0100.0100.098.50100.099.20DentiumImplantium95.2096.2095.7095.9095.2095.5095.4095.7095.60Superline95.0095.4095.2094.2093.7093.9096.0095.6095.80DioimplantUF92.9096.3094.5096.2092.6094.30100.096.3098.10UF II86.2086.2086.2086.2086.2086.2086.2086.2086.20MegagenAny ridge92.3092.3092.3085.7092.3088.9084.6084.6084.60Anyone internal73.8082.9078.1072.071.9072.4079.4064.1070.90Anyone external66.4059.5062.8058.0065.90961.7056.5076.2064.90Exfeel external100.075.085.70100.083.3090.90100.083.3090.90StraumannTS standard84.2088.9086.5088.9088.9088.9084.2088.9086.50TS standard plus90.3093.3091.8093.80100.096.8093.5096.7095.10Bone level99.2096.0097.6098.4096.0097.2099.2094.4096.70ShinhungLuna97.5088.6092.9083.0088.6085.7092.9088.6090.70OsstemGS II86.1093.9089.9093.8090.9092.3081.6093.9087.30SS II77.8058.3066.7090.0075.0081.8080.0066.7072.70TS III97.4096.8097.1097.3096.2096.8097.6097.6097.60US II91.5095.3093.4090.5097.0093.7091.0098.7094.70US III100.0078.9088.2093.3073.7082.4093.3073.7082.40WarantecHexplant80.2064.6071.6079.6062.1069.7078.9072.8075.80Internal54.5071.5061.8052.5071.9061.3060.1067.2063.50IT19.0019.0019.0019.2023.8021.3035.3028.6031.60 Figure 1 shows the normalized confusion matrices, containing a summary of the classification of the 27 different types of DIS based on the automated DL algorithm (full details are provided in Appendix 3). Using panoramic and periapical images, the classification accuracy of DL was the highest for Nobel Biocare Branemark ($100.0\%$) and Megagen Exfeel external ($100.0\%$), and the lowest for Warantec IT ($35.3\%$). Using only panoramic images, the classification accuracy was the highest for Osstem US III ($100.0\%$) and Megagen Exfeel external ($100.0\%$), and the lowest for Warantec IT ($19.0\%$). When using only periapical images, the classification accuracy was the highest for Megagen Exfeel external ($100.0\%$) and Dentsply Xive ($100.0\%$), and the lowest for Warantec IT ($19.2\%$).Figure 1Schematic illustration of dataset collection and verification. All study protocols and related procedures were supervised by the National Information Society Agency (NIA) and the Korean Academy of Oral & Maxillofacial Implantology (KAOMI). ## Discussion AI-based large-scale machine learning and DL in the late 2010s, which facilitated the accurate diagnosis of medical radiographic images, garnered attention in biomedical engineering and provided novel insights into precision medicine26–28. More recently, deep convolutional neural network algorithms have gained popularity in dentistry, and have also achieved considerable success in analyzing dental radiographic images29. The potential clinical applications of DL technology are closely related to [1] deeper and more sophisticated neural network structures and [2] large annotated and high-quality datasets. Particularly, a gold-standard dataset annotated and verified by medical and dental professionals is essential to create a reliable radiographic image-based DL model in the medical and dental fields26,27. To evaluate the performance of DL-based identification and classification of various types of DIS in actual clinical practice, a large, highly accurate, and reliable dataset is necessary. Recently, a large-scale and comprehensive multicenter dataset that could be used in the clinical field for DL-based identification and classification of DIS was collected and released openly by the national initiative. To our knowledge, the dataset used in the present study contained the larger number of radiographic images and types of DIS than any previously reported implant-related dataset. Because we used this dataset in the current study, it is expected to show higher feasibility than that of any previous implant-related DL research. Most previous studies evaluated the accuracy performance of the conventional or minimally modified DL architectures (e.g., YOLO, SqueezeNet, ResNet, GoogLeNet, and VGG-$\frac{16}{19}$) using less than a few thousand dental radiographic images, and usually fewer than 10 different types of DIS in their datasets, identifying a classification accuracy ranging from 70 to $100\%$17–25. One study that utilized a ResNet architecture based on 12 types of 9767 panoramic images reported a high accuracy of $98\%$ or more23. Our previous pilot study that utilized automated DL based on six different types of 11,980 DIS images also showed reliable outcomes and achieved a very high accuracy of $95.4\%$ (sensitivity:$95.5\%$ and specificity:$85.3\%$)18. Conversely, another study based on Yolov3 using 1282 panoramic images showed a relatively low accuracy in the $70\%$ range on average22. The automated DL algorithm used in this study, based on the combination of periapical and panoramic radiographs, achieved an AUC of 0.885. When only panoramic radiographs were used, the AUC was 0.878, and when only periapical radiographs were used, the AUC was 0.868. Specifically, periapical and panoramic images had the highest classification accuracy, and periapical images alone had the lowest accuracy, but there was no statistically significant difference between the three groups. These outcomes are consistent with the previously reported absence of a significant difference in classification accuracy between panoramic and periapical images and are also likely due to the fact that almost three times more panoramic images ($$n = 105$$,080) than periapical images ($$n = 36$$,188) were used for training and validation17,18. Specifically, the Nobel Biocare Branemark, Megagen Exfeel external, Osstem US III, and Dentsply Xive showed a high classification accuracy of $100.0\%$, whereas Warantec IT showed a low accuracy performance (accuracy: 19.0–$35.3\%$) due to the relatively small number of radiographic images, including only 238 panoramic and 208 periapical images, despite having a conventional fixture morphology with an internally tapered shape. From this perspective, DL has great advantages in identifying and classifying similar types of DIS; however, the accuracy performance varies significantly depending on the amount of datasets required for training, which is considered a fundamental limitation of the existing DL algorithms. Further research should be conducted to confirm whether the number of datasets required for training can be reduced by adopting an algorithm that is more specialized than the algorithm in this study for DIS classification. In the radiographs used in this study, the main ROI was the implant fixture, but a number of other confounding conditions (such as surrounding alveolar bone, cover screw, healing abutment, provisional or definitive prosthesis) were included. To be used in actual clinical practice, implant fixtures with different confounding conditions and angles should be used as datasets, rather than implant fixtures with perfect/intact shapes and standard angles. Several previous studies, including this one, have confirmed that implant datasets with different angles and confounding conditions have a high accuracy performance of over $80\%$17,18,25. Furthermore, using the Gradient-Weighted Class Activation Mapping technique, it was found that the types of DIS were classified by focusing on the implant fixture itself rather than the various confounding components of the DIS. Therefore, various confounding factors and angles do not appear to have a significant impact on the accuracy performance of DL-based implant system classification. In a recent study wherein healthcare professionals with no coding experience evaluated the feasibility of automated DL models using five publicly available and open-source medical image datasets, most classification models showed accuracy performance and diagnostic properties comparable to those of state-of-the-art DL algorithms30. Developing customized DL models according to the types and characteristics of datasets requires highly specialized skills and expertise. This study confirmed that the DL algorithm itself, not computer scientists and engineers, built an automated DL model without coding and showed excellent classification accuracy of over $86\%$ in 27 similar design but different types of multiple classifications. Identifying and classifying DIS with varying features and characteristics and limited clinical and radiographic information is a challenge not only for inexperienced dental professionals, but also for dentists with sufficient experience in implant surgery and prosthetics. In the past, several studies have identified DIS from a forensic perspective based on radiographs, and until recently, efforts have been made to classify DIS, but most of these are based on empirical evidence, making it difficult to achieve high reliability9,10,31. More recently, computer-based implant recognition software and web-based DIS classification platforms have been developed and used; however, most require manual classification of DIS features (such as coronal interface, flange, thread type, taper and apex shape) or contain only a small number of DIS datasets, limiting their active use in clinical practice32. The first end goal based on this research was to obtain a database of almost all types of DIS used worldwide and train it with sophisticated and refined DL algorithms optimized for DIS classification to achieve a high level of reliability that can be used in actual clinical practice. The second goal was to create a web or cloud-based environment where datasets can be freely stored, trained, and validated in real time. Achieving these goals requires the proactive development of standard protocols to facilitate data sharing and integration, secure transmission and storage of large datasets, and enable federated learning33,34. This study had several limitations. Collecting a dataset using supervised learning requires considerable tangible and intangible resources including finances, time, trained personnel, hardware, and software. Therefore, unsupervised learning, a technique for overcoming small-scale and imbalanced datasets, has been introduced and tested with caution in dentistry; however, it remains a challenging approach35. Large-scale and multicenter datasets may be useful for future DL-based research and actual clinical trials to identify and classify various types of DIS. Nevertheless, the dataset used in this study had inherent limitations regarding the interpretability of the results. Although the raw NIA dataset consisted of 165,700 radiographs and 42 different types of DIS, the number of panoramic and periapical images for each type of DIS was highly heterogeneous. In addition, DIS manufactured by foreign companies or using non-titanium materials (such as non-metallic ceramic zirconia), which are rarely used in South Korea, were few or not included in the raw dataset. To overcome the potential problem of overfitting and selective bias, we selected only DIS that contained more than 100 images of panoramic and periapical radiographs. ## Conclusion We verified that automated DL shows a high classification accuracy based on large-scale and multicenter datasets. Furthermore, no significant difference in accuracy was observed between panoramic and periapical radiographic images. The clinical feasibility of the automated DL algorithm will have to be confirmed using additional datasets and clinical research in the future. ## Ethics This study was approved and conducted in accordance with the following Institutional Review Board (IRB): Seoul National University Dental Hospital (ERI21024), Yonsei University Dental Hospital [2-2021-0049], Gangnam Severance Dental Hospital [3-2021-0175], Wonkwang University Daejeon Dental Hospital (W$\frac{2104}{003}$-002), Dankook University Dental Hospital [2021-8-004], and national public IRB (P01-202109-21-020). IRBs of Seoul National University Dental Hospital, Yonsei University Dental Hospital, Gangnam Severance Dental Hospital, Wonkwang University Daejeon Dental Hospital, Dankook University Dental Hospital, and national public approved a waiver of informed consent for retrospective large-scale and multicenter data analysis. All methods in this study were performed in accordance to relative guidelines and regulations. ## Dataset collection and verification All included dental radiographic images were managed and supervised by the National Information Society Agency (NIA) under the Ministry of Science and the Korean Academy of Oral and Maxillofacial Implantology (KAOMI). The dataset was collected from five college dental hospitals (including Seoul National University Dental Hospital, Yonsei University Dental Hospital, Yonsei University Gangnam Severance Dental Hospital, Wonkwang University Daejeon Dental Hospital, and Dankook University Dental Hospital) and 10 private dental clinics. Appendix 1 summarizes the detailed consortium of the dataset collection. Digital imaging and communication in medicine-format panoramic and periapical images were converted into either de-identified 512 × 512 pixels or smaller JPEG- or PNG-format images, and one implant fixture per image was cropped as a region of interest (ROI). The collected ROI images were reviewed to ensure that cropping, resolution, and sharpness were properly performed by 10 dental professionals employed by the KAOMI. Subsequently, based on the medical and dental records provided by college dental hospitals and private clinics, each implant fixture was labeled with the manufacturer, brand and system, diameter and length, placement position, surgery date, age, and sex using customized labeling and annotation tools. All included radiographic images were validated by a board-certified oral and maxillofacial radiologist who was not involved in dataset management. The final dataset consisted of 165,700 panoramic and periapical radiographic images and 42 types of DIS. Appendix 2 provides a detailed list of the raw NIA dataset (Fig. 2).Figure 2Dataset containing 116,756 panoramic and 40,209 periapical radiographic images and comprising 10 manufacturers and 27 types of dental implant systems. We included only DIS that contained at least 100 periapical and 100 panoramic images from the raw NIA dataset. Finally, the dataset used in this study contained 116,756 panoramic and 40,209 periapical images, comprised 10 manufacturers and 27 types of DIS. Specifically, the dataset included Neobiotech ($$n = 21$$,260, $13.54\%$), Nobel biocare ($$n = 3644$$, $2.32\%$), Dentsply ($$n = 15$$,296, $9.74\%$), Dentium ($$n = 41$$,096, $26.18\%$), Dioimplant ($$n = 1530$$, $0.97\%$), Megagen ($$n = 7801$$, $4.97\%$), Straumann ($$n = 4977$$, $3.17\%$), Shinhung ($$n = 3376$$, $2.15\%$), Osstem ($$n = 42$$,920, $27.34\%$), and Warantec ($$n = 15$$,065, $9.60\%$). The detailed types of DIS are listed in Table 2 and illustrated in Fig. 3.Table 2Number of panoramic and periapical radiographic images for each dental implant system. ManufacturesSystemPanoramic images($$n = 116$$,756)Periapical images($$n = 40$$,209)Total images($$n = 156$$,965)NeobiotechIS I$67085.75\%$$11392.83\%$$78475.00\%$IS II$27742.38\%$$1040.26\%$$28781.83\%$IS III$75946.50\%$$5331.33\%$$81275.18\%$EB$18901.62\%$$5181.29\%$$24081.53\%$Nobel biocareBranemark$33022.83\%$$3420.85\%$$36442.32\%$DentsplyAstra13,$40411.48\%$$5711.42\%$13,$9758.90\%$Xive$6670.57\%$$6541.63\%$$13210.84\%$DentiumImplantium14,$99312.84\%$$416210.35\%$19,$15512.20\%$Superline16,$73414.33\%$$520712.95\%$21,$94113.98\%$DioimplantUF$5250.45\%$$2730.68\%$$7980.51\%$UF II$4470.38\%$$2850.71\%$$7320.47\%$MegagenAny ridge$2170.19\%$$1350.34\%$$3520.22\%$Anyone internal$16401.40\%$$21675.39\%$$38072.43\%$Anyone external$12901.10\%$$12633.14\%$$25531.63\%$Exfeel external$9740.83\%$$1150.29\%$$10890.69\%$StraumannTS standard$11510.99\%$$1760.44\%$$13270.85\%$TS standard plus$7410.63\%$$3010.75\%$$10420.66\%$Bone level$13471.15\%$$12613.14\%$$26081.66\%$ShinhungLuna$29352.51\%$$4411.10\%$$33762.15\%$OsstemGS II$14011.20\%$$3270.81\%$$17281.10\%$SS II$7170.61\%$$1160.29\%$$8330.53\%$TS III19,$22216.46\%$10,$53626.20\%$29,$75818.96\%$US II$72956.25\%$$23635.88\%$$96586.15\%$US III$7580.65\%$$1850.46\%$$9430.60\%$WarantecHexplant$45683.91\%$$427110.62\%$$88395.63\%$Internal$32242.76\%$$25566.36\%$$57803.68\%$IT$2380.20\%$$2080.52\%$$4460.28\%$Figure 3Multi-label classification confusion matrix with normalization using panoramic and periapical radiographic images. ## Implementation of automated DL algorithm For the identification and classification of 156,965 radiographic images, a customized automatic DL engine (Neuro-T version 3.0.1, Neurocle Inc., Seoul, Korea) was adopted in this study. Within the available computing resources and training time, an automated DL algorithm is a self-training architecture that selects appropriate DL models and optimizes the hyperparameters (including the resize method, number of network convolutional layers, decay method, learning rate, dropout rate, batch size, number of epochs and patience, and optimizer) to fit the model in the customized dataset36. The dataset comprised three groups: panoramic ($$n = 116$$,756), periapical ($$n = 40$$,209), and panoramic and periapical ($$n = 156$$,965) images. Each dataset group was randomly and evenly divided into three subgroups: training ($80\%$), validation ($10\%$), and testing ($10\%$). After the dataset division, the training dataset was augmented by ten times, with random rotations (90°), hue (range of − 0.1 to 0.1), brightness (range of − 0.12 to 0.12), saturation (range of 0.6–1.5), contrast (range of 0.6–1.5), noise (0.05), and horizontal and vertical flips. We trained our approach on two NVIDIA A6000 graphic processing units (48 GB memory, NVIDIA, Mountain View, CA, USA). The models were trained for a maximum of 500 epochs and stopped if the validation set loss did not improve for more than 20 epochs. ## Statistical analysis Categorical and continuous variables were expressed as frequencies (n) and ratios (%). The performance metrics were evaluated as accuracy, precision, recall, and F1 score (Eqs. [ 1]–[4], TP: true positive, FP: false positive, FN: false negative, and TN: true negative):1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{Accuracy}} = \frac{{\left({{\text{TP}} + {\text{TN}}} \right)}}{{\left({{\text{TP}} + {\text{FP}} + {\text{TN}} + {\text{FN}}} \right)}} $$\end{document}Accuracy=TP+TNTP+FP+TN+FN2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{Precison }} = \frac{{{\text{TP}}}}{{\left({{\text{TP}} + {\text{FP}}} \right)}} $$\end{document}Precison=TPTP+FP3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{Recall}} = \frac{{{\text{TP}}}}{{\left({{\text{TP}} + {\text{FN}}} \right)}} $$\end{document}Recall=TPTP+FN4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{F}}1{\text{ score}} = 2{ } \times { }\frac{{\left({{\text{Precision }} \times {\text{ Recall}}} \right)}}{{\left({{\text{Precision }} + {\text{Recall}}} \right)}} $$\end{document}F1score=2×Precision×RecallPrecision+Recall Additionally, a normalized confusion matrix for each DIS was calculated based on the test dataset. All data processing and statistical analyses were conducted using a commercial statistical package (Neuro-T version 3.0.1, Neurocle Inc., Seoul, Korea) and non-commercial statistical package (R version 4.2.0, R Foundation for Statistical Computing, Vienna, Austria). ## Supplementary Information Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-32118-1. ## References 1. Howe MS, Keys W, Richards D. **Long-term (10-year) dental implant survival: A systematic review and sensitivity meta-analysis**. *J. Dent.* (2019.0) **84** 9-21. DOI: 10.1016/j.jdent.2019.03.008 2. Elani HW, Starr JR, Da Silva JD, Gallucci GO. **Trends in dental implant use in the U.S., 1999–2016, and projections to 2026**. *J. Dent. Res.* (2018.0) **97** 1424-1430. DOI: 10.1177/0022034518792567 3. Lee JH, Kim YT, Jeong SN, Kim NH, Lee DW. **Incidence and pattern of implant fractures: A long-term follow-up multicenter study**. *Clin. Implant Dent. Relat. Res.* (2018.0) **20** 463-469. DOI: 10.1111/cid.12621 4. Lee JH, Lee JB, Park JI, Choi SH, Kim YT. **Mechanical complication rates and optimal horizontal distance of the most distally positioned implant-supported single crowns in the posterior region: A study with a mean follow-up of 3 years**. *J. Prosthodont.* (2015.0) **24** 517-524. DOI: 10.1111/jopr.12306 5. Albrektsson T, Donos N, Working G. **Implant survival and complications. The Third EAO consensus conference 2012**. *Clin. Oral Implants Res.* (2012.0) **23** 63-65. DOI: 10.1111/j.1600-0501.2012.02557.x 6. Dreyer H. **Epidemiology and risk factors of peri-implantitis: A systematic review**. *J. Periodontal Res.* (2018.0) **53** 657-681. DOI: 10.1111/jre.12562 7. Lee DW. **Implant fracture failure rate and potential associated risk indicators: An up to 12-year retrospective study of implants in 5,124 patients**. *Clin. Oral Implants Res.* (2019.0) **30** 206-217. DOI: 10.1111/clr.13407 8. Lee DW, Kim SY, Jeong SN, Lee JH. **Artificial intelligence in fractured dental implant detection and classification: Evaluation using dataset from two dental hospitals**. *Diagnostics (Basel)* (2021.0) **11** 233. DOI: 10.3390/diagnostics11020233 9. Nuzzolese E, Lusito S, Solarino B, Di Vella G. **Radiographic dental implants recognition for geographic evaluation in human identification**. *J. Forensic Odontostomatol.* (2008.0) **26** 8-11. PMID: 22689351 10. Berketa JW, Hirsch RS, Higgins D, James H. **Radiographic recognition of dental implants as an aid to identifying the deceased**. *J. Forensic Sci.* (2010.0) **55** 66-70. DOI: 10.1111/j.1556-4029.2009.01226.x 11. Correa LR. **Planning of dental implant size with digital panoramic radiographs, CBCT-generated panoramic images, and CBCT cross-sectional images**. *Clin. Oral Implants Res.* (2014.0) **25** 690-695. DOI: 10.1111/clr.12126 12. LeCun Y, Bengio Y, Hinton G. **Deep learning**. *Nature* (2015.0) **521** 436-444. DOI: 10.1038/nature14539 13. Lee JH, Kim DH, Jeong SN, Choi SH. **Diagnosis and prediction of periodontally compromised teeth using a deep learning-based convolutional neural network algorithm**. *J. Periodontal Implant Sci.* (2018.0) **48** 114-123. DOI: 10.5051/jpis.2018.48.2.114 14. Lee JH, Kim DH, Jeong SN, Choi SH. **Detection and diagnosis of dental caries using a deep learning-based convolutional neural network algorithm**. *J. Dent.* (2018.0) **77** 106-111. DOI: 10.1016/j.jdent.2018.07.015 15. Lee JH, Kim DH, Jeong SN. **Diagnosis of cystic lesions using panoramic and cone beam computed tomographic images based on deep learning neural network**. *Oral Dis.* (2020.0) **26** 152-158. DOI: 10.1111/odi.13223 16. Schwendicke F. **Deep learning for cephalometric landmark detection: Systematic review and meta-analysis**. *Clin. Oral Investig.* (2021.0) **25** 4299-4309. DOI: 10.1007/s00784-021-03990-w 17. Lee JH, Jeong SN. **Efficacy of deep convolutional neural network algorithm for the identification and classification of dental implant systems, using panoramic and periapical radiographs: A pilot study**. *Medicine (Baltimore)* (2020.0) **99** e20787. DOI: 10.1097/MD.0000000000020787 18. Lee JH, Kim YT, Lee JB, Jeong SN. **A performance comparison between automated deep learning and dental professionals in classification of dental implant systems from dental imaging: A multi-center study**. *Diagnostics (Basel)* (2020.0) **10** 910. DOI: 10.3390/diagnostics10110910 19. Sukegawa S. **Deep neural networks for dental implant system classification**. *Biomolecules* (2020.0) **10** 984. DOI: 10.3390/biom10070984 20. Hadj Said M, Le Roux MK, Catherine JH, Lan R. **Development of an artificial intelligence model to identify a dental implant from a radiograph**. *Int. J. Oral. Maxillofac. Implants* (2020.0) **36** 1077-1082. DOI: 10.11607/jomi.8060 21. Kim JE. **Transfer learning via deep neural networks for implant fixture system classification using periapical radiographs**. *J. Clin. Med.* (2020.0) **9** 1117. DOI: 10.3390/jcm9041117 22. Takahashi T. **Identification of dental implants using deep learning-pilot study**. *Int. J. Implant Dent.* (2020.0) **6** 53. DOI: 10.1186/s40729-020-00250-6 23. Sukegawa S. **Multi-task deep learning model for classification of dental implant brand and treatment stage using dental panoramic radiograph images**. *Biomolecules* (2021.0) **11** 815. DOI: 10.3390/biom11060815 24. da Mata Santos RP. **Automated identification of dental implants using artificial intelligence**. *Int. J. Oral Maxillofac. Implants* (2021.0) **36** 918-923. DOI: 10.11607/jomi.8684 25. Lee JH, Kim YT, Lee JB, Jeong SN. **Deep learning improves implant classification by dental professionals: A multi-center evaluation of accuracy and efficiency**. *J. Periodontal Implant Sci.* (2022.0) **52** 220-229. DOI: 10.5051/jpis.2104080204 26. Gulshan V. **Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs**. *JAMA* (2016.0) **316** 2402-2410. DOI: 10.1001/jama.2016.17216 27. Esteva A. **Dermatologist-level classification of skin cancer with deep neural networks**. *Nature* (2017.0) **542** 115-118. DOI: 10.1038/nature21056 28. Filipp FV. **Opportunities for artificial intelligence in advancing precision medicine**. *Curr. Genet. Med. Rep.* (2019.0) **7** 208-213. DOI: 10.1007/s40142-019-00177-4 29. Mohammad-Rahimi H. **Deep learning in periodontology and oral implantology: A scoping review**. *J. Periodontal Res.* (2022.0) **57** 942-951. DOI: 10.1111/jre.13037 30. Faes L. **Automated deep learning design for medical image classification by health-care professionals with no coding experience: A feasibility study**. *Lancet Digit. Health* (2019.0) **1** e232-e242. DOI: 10.1016/S2589-7500(19)30108-6 31. Sahiwal IG, Woody RD, Benson BW, Guillen GE. **Radiographic identification of nonthreaded endosseous dental implants**. *J. Prosthet. Dent.* (2002.0) **87** 552-562. DOI: 10.1067/mpr.2002.124431 32. Michelinakis G, Sharrock A, Barclay CW. **Identification of dental implants through the use of Implant Recognition Software (IRS)**. *Int. Dent. J.* (2006.0) **56** 203-208. DOI: 10.1111/j.1875-595X.2006.tb00095.x 33. Rischke R. **Federated learning in dentistry: Chances and challenges**. *J. Dent. Res.* (2022.0) **101** 1269-1273. DOI: 10.1177/00220345221108953 34. Dayan I. **Federated learning for predicting clinical outcomes in patients with COVID-19**. *Nat. Med.* (2021.0) **27** 1735-1743. DOI: 10.1038/s41591-021-01506-3 35. Prados-Privado M, Garcia Villalon J, Martinez-Martinez CH, Ivorra C, Prados-Frutos JC. **Dental caries diagnosis and detection using neural networks: A systematic review**. *J. Clin. Med.* (2020.0) **9** 3579. DOI: 10.3390/jcm9113579 36. 36.Jin, H., Song, Q. & Hu, X. Auto-Keras: An efficient neural architecture search system. arXiv e-print, arXiv:1806.10282 (2019).
--- title: Direct reprogramming of human fibroblasts into insulin-producing cells using transcription factors authors: - Marta Fontcuberta-PiSunyer - Ainhoa García-Alamán - Èlia Prades - Noèlia Téllez - Hugo Alves-Figueiredo - Mireia Ramos-Rodríguez - Carlos Enrich - Rebeca Fernandez-Ruiz - Sara Cervantes - Laura Clua - Javier Ramón-Azcón - Christophe Broca - Anne Wojtusciszyn - Nuria Montserrat - Lorenzo Pasquali - Anna Novials - Joan-Marc Servitja - Josep Vidal - Ramon Gomis - Rosa Gasa journal: Communications Biology year: 2023 pmcid: PMC10039074 doi: 10.1038/s42003-023-04627-2 license: CC BY 4.0 --- # Direct reprogramming of human fibroblasts into insulin-producing cells using transcription factors ## Abstract Direct lineage reprogramming of one somatic cell into another without transitioning through a progenitor stage has emerged as a strategy to generate clinically relevant cell types. One cell type of interest is the pancreatic insulin-producing β cell whose loss and/or dysfunction leads to diabetes. To date it has been possible to create β-like cells from related endodermal cell types by forcing the expression of developmental transcription factors, but not from more distant cell lineages like fibroblasts. In light of the therapeutic benefits of choosing an accessible cell type as the cell of origin, in this study we set out to analyze the feasibility of transforming human skin fibroblasts into β-like cells. We describe how the timed-introduction of five developmental transcription factors (Neurog3, Pdx1, MafA, Pax4, and Nkx2-2) promotes conversion of fibroblasts toward a β-cell fate. Reprogrammed cells exhibit β-cell features including β-cell gene expression and glucose-responsive intracellular calcium mobilization. Moreover, reprogrammed cells display glucose-induced insulin secretion in vitro and in vivo. This work provides proof-of-concept of the capacity to make insulin-producing cells from human fibroblasts via transcription factor-mediated direct reprogramming. Human foreskin fibroblasts are directly reprogrammed into insulin-producing cells using 5 pancreatic transcription factors, without pluripotency induction. ## Introduction Direct lineage reprogramming entails the direct conversion of one differentiated cell type into another bypassing an intermediate pluripotent stage. This strategy is often based on the forced expression of cocktails of transcription factors that function as potent fate determinants during development of the cell type of interest1–3. As the number of cell types produced through direct conversion has rapidly increased in recent years, this strategy has emerged as a possible method for creating cell types with potential for use in therapeutic settings. Pancreatic beta (β) cells produce insulin, which controls whole body glucose homeostasis. Diabetes is characterized by a relative or total lack of functional β cells, and cell replacement therapy has consequently emerged as a promising therapeutic option to treat and ultimately cure this disease. One of the strategies pursued to produce replacement β cells has been direct lineage reprogramming. A major breakthrough in this area was the discovery that three developmental transcription factors, namely Pdx1, Neurog3, and MafA, promoted the in situ conversion of acinar cells into insulin-producing cells in the mouse pancreas4,5. Since then, studies have shown that various combinations of these and other transcription factors can promote conversion toward a β-like fate in other pancreatic cell lineages, including ductal and glucagon-expressing α-cells, and in extra-pancreatic related endodermal cell lineages, such as liver, gallbladder, and gastrointestinal tract cells6–11. One of the most important aspects of direct reprogramming strategies is the choice of the cell source, especially when taking into account their clinical application. The initial material should ideally be available, simple to handle and to grow in the laboratory. In this regard, skin fibroblasts have been the preferred cell source for many reprogramming protocols and they have so far been successfully transformed into a variety of somatic cell types including cardiomyocytes12, chondrocytes13, neurons14, oligodendrocyte progenitors15, hepatocytes16 or endothelial cells17. However, research to date suggests that fibroblasts are resistant to being transformed into β-like cell using lineage-specific transcription factors4,9,18. In light of the expanding number of cell types produced by direct lineage conversion procedures and the therapeutic interest of insulin-producing cells, we chose to thoroughly consider the viability of using fibroblasts as cells of origin in direct reprogramming protocols to generate β-like cells. Here we present a protocol based on a cocktail of five endocrine transcription factors that induces human fibroblasts to activate the β-cell transcriptional program while downregulating their native fibroblastic transcriptional program, resulting in the generation of cells that produce and secrete insulin in vitro and in vivo. We believe these findings demonstrate the feasibility of this approach and set the basis to further explore this alternative path for generation of β-like cells for disease modeling and cellular therapy. ## Exogenous expression of the transcription factors Pdx1, Neurog3, and MafA in human fibroblasts We first sought to examine whether the transcription factors Neurog3, Pdx1, and MafA could induce expression of the INSULIN (INS) gene in human fibroblasts as readout of the capacity of these cells to be transformed toward a β-cell fate. To deliver these factors we employed a polycistronic adenoviral vector carrying the three transgenes (Ad-NPM hereafter), which had been previously used to promote β-cell reprogramming from pancreatic acinar cells19. After optimization of adenoviral transduction in fibroblasts (see Methods), abundant (>$80\%$) cells positive for Cherry, which is also encoded by Ad-NPM, were easily observable three and seven days after viral infection (Fig. 1a). Likewise, high levels of transcripts encoding the NPM factors were expressed at both time points (Fig. 1b). Three days after addition of Ad-NPM we detected marginal levels of INS mRNA that were increased >10-fold by day 7 (Fig. 1c). As cell culture formulations can have a major impact on gene expression events and cellular reprogramming, we tested different conditions after Ad-NPM infection. We observed that moving to RPMI-1640 and, to a lesser extent, CMRL-1066 medium and lowering the fetal calf serum concentration to $6\%$ dramatically boosted INS gene activation, reaching values that were $0.12\%$ those of human islets (Fig. 1d and Supplementary Figure 1). Under the same culture conditions, only a very marginal induction of the INS gene occurred when the N + P + M factors were delivered simultaneously via distinct adenoviruses to human fibroblasts (Supplementary Fig. 2). In addition to INS, we discovered that the NPM factors also activated the hormone genes GLUCAGON (GCG) and SOMATOSTATIN (SST), albeit at lower levels than INS as indicated by decreased relative expression values (compared to the housekeeping gene TBP) (Fig. 1e). The NPM factors also induced expression of genes encoding islet differentiation transcription factors including NEUROD1, INSM1, PAX4, NKX2-2, and ARX (Fig. 1f).Fig. 1Introduction of the transcription factors Neurog3, Pdx1, and MafA activates pancreatic endocrine gene expression in human fibroblasts. Human fibroblasts (HFF1) were infected with a polycistronic recombinant adenovirus encoding the transcription factors Neurog3, Pdx1, MafA, and the reporter protein Cherry (Ad-NPM). Untreated parental fibroblasts were used as controls (indicated as C in graphs). a Bright field images and Cherry immunofluorescence of control fibroblasts and fibroblasts infected with Ad-NPM at day 3 and 7 post-infection. Scale bar, 100 μm. b qPCR of transgenes at day 3 ($$n = 11$$) and 7 ($$n = 6$$) after infection with Ad-NPM. c qPCR of human INS at day 3 ($$n = 7$$) and 7 ($$n = 13$$) after infection with Ad-NPM. d qPCR of human INS in fibroblasts maintained in the indicated culture media (DM = DMEM; CM = CMRL; RP = RPMI) during 7 days after infection with Ad-NPM or with an adenovirus expressing B-galactosidase (B-gal) ($$n = 3$$–10). In yellow, INS mRNA levels in isolated human islets ($$n = 10$$). e, f qPCR of islet hormone genes (GCG, SST) and islet differentiation transcription factors (NEUROD1, INSM1, PAX4, NKX2-2, ARX) at day 7 post-NPM ($$n = 8$$–15). g qPCR of the indicated fibroblast markers at day 3 ($$n = 5$$–11) and day 7 post-NPM ($$n = 5$$–6). In b–f, expression levels are expressed relative to TBP. In g, expression is expressed relative to control fibroblasts, given the value of 1 (dotted line). Data are presented as the mean ± SEM for the number of samples indicated in parentheses. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001$, between indicated conditions using unpaired t-test (b–f), or relative to control fibroblasts using one sample t-test (g). To further establish if the NPM factors promoted cell fate conversion and not simply activated their target genes in fibroblasts, we surveyed expression of genes associated with the fibroblastic signature, including several factors involved in maintenance of the fibroblastic transcriptional network such as TWIST2, PRRX1, and LHX920. We found that these genes were downregulated as early as three days after NPM introduction. Other fibroblast markers exhibited a more delayed response but, by day 7 post-NPM, all tested genes exhibited significant down-regulation (Fig. 1g). Together, these experiments validate that islet cell fate can be induced in human fibroblasts using a defined set of transcription factors. ## Addition of exogenous Pax4 and Nkx2-2 after the NPM reprogramming cocktail in human fibroblasts The observed induction of the islet hormone genes GCG and SST implied that the NPM factors might not specifically endorse β-cell fate in fibroblasts. Furthermore, we found that these factors did not induce NKX6-1, which encodes a β-cell specific factor required for the formation of pancreatic β cells during development21 and key for optimal maturation of stem cell β cells in vivo22,23 (Fig. 2a). These findings indicated that the NPM factors sub-optimally promoted a β-cell state in human fibroblasts. In order to enhance β-cell fate over other islet cell identities, we opted to add new transcription factors to the reprogramming cocktail. Fig. 2Sequential addition of the transcription factors Pax4 and Nkx2-2 enhances β-cell fate in human fibroblasts expressing Neurog3, Pdx1, and MafA.Human fibroblasts (HFF1) were infected with Ad-NPM alone or sequentially with Ad-NPM and adenoviruses encoding the transcription factors Pax4 and Nkx2-2. Ad-Pax4 and Ad-Nkx2-2 were added three days after NPM in the two-virus conditions. In the three-virus condition, Pax4 was added three days and Nkx2-2 and six days after NPM (condition called 5TF). All cells were collected ten days after infection with Ad-NPM. a qPCR of the indicated transgenes and endogenous genes. Expression levels are calculated relative to TBP. Values represent the mean ± SEM ($$n = 4$$–12). b Representative immunofluorescence images showing insulin staining (in red) using two different antibodies, one against C-PEP, in untreated fibroblasts and in fibroblasts infected with Ad-NPM alone or with 5TF. Nuclei were stained with Hoechst (in blue). Scale bar, 25 μm. * $P \leq 0.05$; ****$P \leq 0.0001$ relative to NPM in (b) using one-way ANOVA and Tukey’s multiple comparison test. Pax4 is activated downstream of Neurog3 during development24 and has been shown to favor β- over α-cell specification25,26, and to contribute to maintenance of the expression of Nkx6.1 in differentiating β cells27. Despite that the NPM factors induced endogenous PAX4 mRNA, the expression levels attained might not be sufficient to endorse β- over α-cell fate. Hence, we treated fibroblasts with an adenovirus encoding Pax4 three days after NPM (Fig. 2a). This resulted in the significant enhancement of INS expression as compared to NPM alone but, unexpectedly, GCG expression was also increased (Fig. 2a), indicating that ectopic Pax4 improved islet hormone gene expression without apparent impact on β- versus α-cell fate conversion in human fibroblasts. As the NKX6-1 gene remained silent in response to NPM + Pax4 (Fig. 2a), we tried directly adding Nkx6-1 to the NPM reprogramming cocktail. However, exogenous Nkx6-1 resulted in considerable cell death irrespective of level of expression or timing of introduction. As an alternate approach, we added exogenous Nkx2-2, which also regulates early β-cell differentiation and is an upstream activator of Nkx6-1 during mouse islet development21. Treatment with an adenovirus encoding Nkx2-2 three days after NPM led to endogenous activation of NKX6-1 expression with no compromise of fibroblast viability (Fig. 2a). Nkx2-2 also induced PAX6, a pan-endocrine gene required to achieve high levels of islet hormone gene expression during mouse pancreas development28,29. Remarkably, ectopic Nkx2-2 reduced NPM-induced GCG gene activation without affecting INS gene expression (Fig. 2a). During development, Pax4 and Nkx2-2 are found in β-cell precursors at around the same time, and their parallel activities are thought to enable the β-cell differentiation program27. Hence, we tested the effects of including both transcription factors in the reprogramming cocktail. To ensure optimal expression of each transcription factor, we treated cells with Ad-Pax4 and Ad-Nkx2.2 sequentially, at day 3 and day 6 post-NPM, respectively. Following this protocol, the blockade of GCG gene activation and the induction of the NKX6.1 and PAX6 genes seen with NPM + Nkx2.2, and the higher INS expression elicited by NPM + Pax4 relative to NPM alone were all maintained (Fig. 2a). Neither Pax4 nor Nkx2-2, added alone or together, had any impact on the minimal INS gene induction shown when the N + P + M factors were delivered via separate adenoviruses to human fibroblasts (Supplementary Fig. 2). Consistent with the gene expression data, staining for insulin protein was more robust in cells reprogrammed with NPM + Pax4 + Nkx2.2 than in cells reprogrammed with NPM as assessed using two different antibodies, one against human insulin and another against human C-PEP to exclude possible insulin uptake from the media (Fig. 2b). We quantified the immunofluorescence images and found that 67.9 ± $6.2\%$ of cells in the culture were INS + at day 10. ## Characterization of cells generated from fibroblasts using the 5TF- reprogramming cocktail From here on, we used the sequential introduction of the five transcription factors (5TF protocol, Fig. 3a) to generate insulin-producing cells from human fibroblasts (reprogrammed cells will be referred as 5TF cells). At day 10, 5TF cells displayed an epithelial morphology (Fig. 3b) and hadn’t grown as much as untreated fibroblasts (day 10; 5TF: 44 × 103 ± 3 × 103 cells/well; control: 238 × 103 ± 18 × 103 cells/well, $$n = 18$$). This decreased cell number was likely due to diminished proliferation, which was evident as soon as one day following Ad-NPM infection (Fig. 3c). The capacity of cells to reduce the MTT compound, in contrast, was comparable to that of fibroblasts, indicating that viability was not compromised (Fig. 3d).Fig. 3The 5TF protocol results in cell growth arrest and activation of endogenous β-cell differentiation transcription factors and β-cell marker genes in human fibroblasts.a Scheme of the reprogramming protocol 5TF (NPM + Pax4 + Nkx2.2) showing the sequence of addition of adenoviruses encoding the indicated transcription factor/s. Duration of incubation with each adenovirus is represented with a line. Cells were studied at days 10-11 after initial addition of Ad-NPM. b Representative bright field image of parental fibroblasts and 5TF reprogrammed fibroblasts at day 10. Scale bar, 75 μm. c Cell proliferation measured by BrdU incorporation and d cell viability measured by MTT assay for $$n = 3$$ independent reprogramming experiments. Bars represent values relative to control fibroblasts (given the value of 1, represented by a dotted line). Note that day 4 values are before Pax4 introduction. e, f qPCR of islet/β-cell transcription factor and β-cell function genes in untreated control fibroblasts (C, $$n = 5$$–22), in fibroblasts infected with Ad-NPM alone ($$n = 3$$–22) or with 5TF ($$n = 5$$–22). Expression levels were calculated relative to TBP. qPCR of the indicated endogenous genes (g) and transgenes (h) at day 21 after initiation of reprogramming ($$n = 9$$–13, from 7 reprogramming experiments). Transcript levels are expressed relative to levels in cells at day 10 of the reprogramming protocol (given the value of 1, shown with a dotted line). Data are mean ± SEM for the number of n indicated in parentheses. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ compared to control fibroblasts (c, d), or between indicated bars using unpaired t-test (e, f), or compared to day 10 5TF cells (g, h) using one-sample t-test. Next we studied expression of selected differentiation transcription factor genes at days 10-11 of the protocol. *All* genes tested, except PAX4, were more expressed in 5TF relative to NPM (NEUROD1, INSM1, HNF1B, MAFB, PDX1, NEUROG3, NKX2.2) (Fig. 3e). Likewise, several genes (PCSK1, KCNJ11, GLP1R, NCAM1) that are linked to β-cell function were increased in 5TF cells as compared to NPM cells (Fig. 3f). Remarkably, some genes were induced de novo by 5TF (ABCC8, GIPR) (Fig. 3f). In line with a loss of GCG activation, the pro-convertase gene PCSK2, which is expressed at higher levels in α than in β cells30, was reduced by 5TF as compared to NPM (Fig. 3f). These results support that sequential introduction of Pax4 and Nkx2-2 after NPM endorses the β-cell differentiation program in human fibroblasts. β-cell gene activation was sustained for at least twenty-one days after initiation of the protocol despite reduced expression of the reprogramming factor transgenes (Fig. 3g, h). Furthermore, expression of several of the tested genes increased with time in culture including NKX6-1, PCSK1, KCNJ11, ABCC8 and CHGB among others (Fig. 3g), suggestive of permanent cell lineage conversion. Glucose-induced insulin secretion by β cells is mediated by cellular glucose metabolism, closure of ATP-dependent potassium channels, membrane depolarization and opening of voltage-dependent calcium channels, resulting in an increase in cytosolic Ca2+ that triggers insulin exocytosis. We investigated whether 5TF cells increased intracellular Ca2+ in response to glucose and membrane depolarization elicited by high potassium. We found that $65\%$ of the cells exhibited a response to glucose, high potassium, or both, whilst $35\%$ of cells were unresponsive to either stimulus (Fig. 4a and Supplementary Video 1). Parental fibroblasts not engineered for 5TF expression were unresponsive to these stimuli (Fig. 4b and Supplementary Video 2). Among responsive cells, approximately half responded to both glucose and high potassium and half responded only to potassium (Fig. 4a). We observed heterogeneity in the amplitude and kinetics of responses among individual cells (Fig. 4c). Next, we performed static incubation assays to study GSIS and found that 5TF cells released similar amounts of human insulin at low (2 mM) and high (20 mM) glucose concentrations (Fig. 4d). Thus, even though 5TF cells increased their intracellular calcium in response to glucose and membrane depolarization, they secreted insulin in a constitutive manner. Fig. 45TF cells increase intracellular calcium in response to glucose and KCl.5TF cells were loaded with the calcium indicator Fluo-4-AM at day 10 of the reprogramming protocol. Single-cell imaging to detect cytosolic calcium was performed in the following sequence: low glucose (2 mM, G2), high glucose (20 mM, G20) and membrane depolarization with KCl (30 mM). a Quantification of the frequency of cells ($$n = 200$$, from six independent reprogramming experiments) that responded to glucose, membrane depolarization elicited by high potassium or both. Representative measurements of dynamic Fluo-4 fluorescence for (b) six fibroblasts and (c) four 5TF cells. d In vitro insulin secretion by 5TF cells. ELISA determination of secreted human insulin by control fibroblasts ($$n = 4$$–13) and 5TF cells ($$n = 16$$) under non-stimulatory conditions (glucose 2 mM) and under stimulatory conditions (glucose 20 mM). Data are mean ± SEM and correspond to six independent reprogramming experiments, 2–4 biological replicates per experiment. ## Generation of 5TF cell spheroids and transcriptome-wide analysis The differentiation and functionality of many cell types vary dramatically between three-dimensional (3D) and two-dimensional (2D) monolayer cultures, the former being closer to the natural 3D microenvironment of cells in a living organism. Thus, we generated spheroids of 5TF cells (1200-1800 cells/spheroid; average diameter of 128 ± 27 µm) one day after the introduction of Nkx2-2 and maintained them in culture for three additional days (Fig. 5a). At the time of collection, insulin-positive staining was easily identified but glucagon and somatostatin staining was undetectable (Fig. 5b and Supplementary Fig. 3). While INS transcript levels were nearly 2-fold higher in 5TF cell spheroids compared to 5TF cells kept in monolayer, other β-cell marker genes, such as the prohormone convertase PCSK1 and the ATP-sensitive potassium channel subunits KCNJ11 and ABCC8, showed a higher response (4 to 5-fold) to 3D culture (Fig. 5c). Thus, cell aggregation during the last stage of reprogramming (note that total length of the protocol was not changed) conferred improved activation of genes associated to β-cell function. Despite increased gene activation, β-cell gene expression in 5TF cell spheroids remained lower than in human islets, with differences ranging widely among examined genes (Fig. 5c).Fig. 5Generation and transcriptomic characterization of 5TF cell spheroids.a Schematic representation of the modified 5TF protocol (5TF-3D): cells were moved from 2D to 3D culture during the last three days (days 7–10) of the protocol. Representative bright field image of 5TF cell spheroids. Scale bar, 100 μm. b Representative immunofluorescence image showing insulin staining in red and nuclei in blue (marked with Hoechst) of a 5TF cell spheroid at the end of the reprogramming protocol. Scale bar is 50 μm. c qPCR of the indicated genes in 5TF cell spheroids. Transcript levels are expressed as fold relative to levels in 5TF cells maintained in 2D culture throughout the 10-day protocol (given the value of 1, dotted line). Data are mean ± SEM for $$n = 4$$–12. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$ relative to 2D culture using one-sample t-test. Fold-change differences in expression levels between human islets and 3D-5TF reprogrammed cells are shown in the upper yellow box. d Heat map of differentially expressed genes between parental fibroblasts (C) and 5TF cell spheroids ($$n = 3$$ reprogramming experiments). e GSEA plots on indicated gene sets and pathways. f Dot plots showing the enrichment analysis on Gene Ontology (GO) and KEGG categories of differentially expressed genes (gained in red, lost in blue) between fibroblasts (C) and 5TF cells. The X-axis represents the adjusted p value, the size of the dot represents the number of enriched genes (count) and the color intensity of the dots represents the percentage of hits in each category. g GSEA plot on β-cell disallowed genes. h Relative expression levels of β-cell disallowed genes repressed in 5TF cell spheroids as compared to fibroblasts (given the value of 1) based on RNA-seq data normalized expression values. Data are mean ± SEM ($$n = 3$$). Insets show mRNA expression of the indicated genes in untreated control fibroblasts ($$n = 5$$), 5TF cell spheroids ($$n = 6$$) and human islets ($$n = 5$$) as assessed by qPCR. Expression levels were calculated relative to TBP. Data are mean ± SEM. * $P \leq 0.05$, **$P \leq 0.01$ relative to control fibroblasts using unpaired t-test. To obtain a more comprehensive understanding of the cell identity switch induced by the 5TF-3D reprogramming protocol, we performed RNA-sequencing of 5TF cell spheroids and parental fibroblasts. A total of 2806 genes (1186 upregulated, 1620 downregulated) were differentially expressed between both cell populations (adjusted p-value <0.05 and fold-change (FC) > 2) (Fig. 5d and Supplementary Data 1). Gene set enrichment analysis (GSEA) showed that pancreas/β-cell and peptide hormone metabolism gene sets were enriched in 5TF cells (Fig. 5e). Biological functions associated with gained genes included epithelium development, synaptic signaling, ion transport, calcium sensing and secretion (Fig. 5f). Among the upregulated genes related to stimulus-secretion coupling, there were synaptotagmins (SYT1,2,3,6,13,17), syntaxins (SYN2 SYN3), calcium sensors (SCG2) and SNARE protein complexes (VAMP1). Correlating with our previous results, cell cycle and mitotic function genes were enriched among repressed genes (Fig. 5e, f). Additionally, GSEA demonstrated that 5TF cells had a lower expression of the gene set associated with the epithelial-mesenchymal transition (Fig. 5e). In agreement, functions including cytoskeleton organization and cellular migration were overrepresented among lost genes (Fig. 5f). Interestingly, GSEA also revealed that the β-cell disallowed gene set, which includes genes that are selectively suppressed in β cells and believed to be detrimental for β cell function31–33, was reduced in 5TF cells (Fig. 5g). A total of 23 previously recognized β-cell disallowed were significantly downregulated in 5TF cells (Fig. 5h). By using qPCR, we confirmed the repression of three of these genes -OAT, LDHA, and SMAD3- which are regarded as part of the core disallowed unit33. Of note, the levels of these genes in 5TF cells matched those of human islets (Fig. 5h). Collectively, these results show that 5TF-3D reprogramming promotes a change in the fibroblast transcriptome, including selective gene activation along with specific gene repression events, enabling a change in cell identity from fibroblast towards a β-cell fate. ## Ultrastructure and insulin secretory features of 5TF cell spheroids Consistent with gene activation events identified in prior gene expression analyses, immunofluorescence staining showed the presence of the mature β-cell markers PCSK1, NCAM1, and KCNJ11 (Kir6.2) in many insulin-positive 5TF cells. PTPRN (IA2) was also expressed albeit more sporadically in insulin-positive 5TF cells (Fig. 6a). Using conventional electron microscopy, we looked for the existence of secretory granules and discovered that most cells contained multiple spherical electron-dense prototypical secretory vesicles (Fig. 6b). These vesicles showed a high degree of morphological heterogeneity, presumably as consequence of their degree of maturation and/or loading. Although they did not have the appearance of typical insulin-containing granules from primary β cells, which are characterized by a clear halo surrounding a dark polygonal dense core34, some of the vesicles exhibited a gray or less electron dense halo and looked like the granules described in immature insulin-positive cells generated in early stem cell differentiation protocols35,36.Fig. 6Insulin secretion by 5TF cell spheroids.a Representative confocal images of 5TF cell spheroids immunostained with the indicated antibodies. Scale bar, 10 μm. b Conventional transmission electron microscopy showing a representative image of a 5TF cell spheroid. Prototypical electron dense secretory vesicles (asterisks) are observed dispersed in the cytoplasm. Well-preserved mitochondria (mit), endoplasmic reticulum (ER), Golgi membranes (G) and lipid droplets (LD) are also observed. Inset shows a detail of a secretory vesicle with an average diameter of 450 nm. N, nucleus. Scale bars are 200 nm (inset) and 500 nm. c In vitro glucose-induced insulin secretion by 5TF cell spheroids ($$n = 14$$, from 8 reprogramming experiments). Secretion by control spheroids composed of parental fibroblasts ($$n = 5$$) is also shown. d Glucose stimulation Index (ratio between insulin secreted at 20 mM glucose vs. 2 mM glucose) of 5TF cells maintained in 2D or in 3D (spheroid) cultures ($$n = 16$$–18, from 8 to 10 reprogramming experiments). e Glucose dose curve of insulin secretion by 5TF cell spheroids ($$n = 4$$–12, 5 reprogramming experiments). Data are presented as the mean ± SEM for the number of n indicated in parentheses. * $P \leq 0.05$; ***$P \leq 0.001$ between the indicated conditions using unpaired t-test (c), one sample t-test (d) or one-way ANOVA (e). We next performed static incubation GSIS assays. 5TF cell spheroids exhibited significant insulin secretory response to glucose (fold 20 mM/2 mM: 2.02 ± 0.18) as compared to 2D cultures (fold 20 mM/2 mM: 1.08 ± 0.15) (Fig. 6c, d). To establish the glucose threshold for stimulation of insulin secretion, 5TF cell spheroids were subjected to either 2, 5, 11 or 20 mM glucose. Between 2 mM and 11 mM/20 mM glucose, 5TF spheroids showed a 2.3-fold increase in insulin production on average (Fig. 6e). In contrast, although there was some variability, they did not show a statistically significant increase in insulin secretion between 2 mM and 5 mM glucose (Fig. 6e). These observations indicate that 5TF cell spheroids are stimulated at higher glucose threshold; it is interesting to note that human islets have a glucose threshold at 3 mM and a maximal response at 15 mM37. The 5TF-3D protocol was repeated on an additional HFF line and produced results that were comparable (Supplementary Fig. 4) proving the reproducibility of the reprogramming protocol. ## Transplantation of 5TF cell spheroids Finally, we studied the stability of reprogramming in vivo. With this aim, we transplanted 300 5TF cell spheroids (1000–1200 cells/spheroid) into the anterior chamber of the eye (ACE) of non-diabetic immune-deficient NOD scid gamma (NSG) mice (Fig. 7a). The ACE allows fast engraftment38 and in vivo imaging39. Ten days following transplantation, we used two-photon microscopy to evaluate in vivo graft re-vascularization and confirmed the presence of functioning vessels in the grafts (Fig. 7b). Additionally, by observing the long-term tracer CFDA’s fluorescence, we confirmed that the transplanted cells were alive (Fig. 7b). To assess the maintenance of insulin expression in vivo, we harvested the eye grafts at day 10 for RNA extraction and immunostaining. Human INS mRNA was readily detectable and levels, calculated relative to human TBP, were comparable to those in 5TF cell clusters prior to transplantation (Fig. 7c). In agreement, abundant HLA + (human cell marker) cells that stained for insulin were detected in the eye grafts by immunofluorescence staining (43.5 ± $2.8\%$ INS+HLA+/total HLA+, $$n = 5$$) (Fig. 7d, e and Supplementary Fig. 5). We observed positive staining for the reprogramming transcription factors in 20–$30\%$ of the INS + cells (Supplementary Figure 6). Although we were unable to discriminate between the two, high transgene expression found by qPCR analysis in eye grafts (Supplementary Fig. 6) indicated that the staining represented virally encoded exogenous protein rather than endogenous protein. Since adenoviral vectors do not normally integrate into the host DNA, we speculate that the cessation of cell division induced by reprogramming may explain persistent transgene expression in 5TF cells. In fact, similar findings were reported in reprogrammed human duct-derived insulin-producing cells9. We were able to identify INS + cells in 4 (of 5) grafts harvested one month after transplantation even though their number was reduced relative to day 10 grafts (Supplementary Fig. 7). The proportion of INS+HLA+ cells in 30-day grafts was more heterogeneous than in 10-day grafts, and in 3 (of 5) grafts, it was comparable or even higher than that of 10-day grafts, demonstrating the maintenance of reprogramming (Supplementary Fig. 7).Fig. 7In vivo characterization of 5TF cell spheroids.a Schematic illustration and image showing 5TF cell spheroids transplanted into the anterior chamber of the eye (ACE) of a normoglycemic NSG mouse. b Vascularization of 5TF cell grafts ten days following transplantation into the ACE. Representative in vivo image depicting functional vessels (RITC-dextran, red) and viable 5TF cells (CFDA, green). Scale bar, 100 μm. c qPCR of INS and TBP transcripts in eyes of non-transplanted mice (nt, $$n = 3$$) and mice transplanted with either control fibroblast spheroids (C, $$n = 3$$) or 5TF cell spheroids ($$n = 5$$) collected ten days post-transplantation. INS gene expression in 5TF cell spheroids prior to transplantation is depicted in the blue bar ($$n = 6$$). INS gene expression is calculated relative to TBP. Expression of TBP relative to mouse *Tbp is* shown to prove the presence of human cells in eyes receiving control and 5TF spheroids. Data are presented as mean ± SEM. d Representative immunofluorescence images showing HLA staining in red and insulin staining in green in 5TF cell grafts ten days post-transplantation. Scale bar, 25 μm. e Percentage of cells doubly positive for insulin and HLA (relative to total HLA + cells) in 5TF cell grafts at day 10 following transplantation. Each dot corresponds to one eye graft ($$n = 5$$). f ELISA determination of human insulin in the aqueous humor in un-transplanted mice ($$n = 7$$), in mice transplanted with either 300 fibroblast spheroids ($$n = 14$$) or 300 5TF cell spheroids ($$n = 17$$) at day 10 post-transplantation and in mice transplanted with 150–200 human islets ($$n = 4$$) at day 12–15 post-transplantation. Data are presented as mean ± SEM for the number of n indicated in parentheses. *** $P \leq 0.001$; ***$P \leq 0.0001$ between indicated samples using unpaired t-test. To study if 5TF cells secreted insulin in vivo, we first measured the presence of human insulin by ELISA in the aqueous humor of the transplanted eyes. Human insulin was readily detectable in eyes carrying 5TF cell grafts (17 of 17, ranging from 76 to 1103 pmol/L) whilst no insulin was detected in eyes transplanted with parental fibroblast clusters or in non-transplanted mice (Fig. 7f). For comparison, eyes containing 300 5TF spheroids showed on average approximately 20-fold lower levels of human insulin than eyes containing 150–200 human islets (Fig. 7f). Due to space limitations in the ACE, we transplanted a larger number of spheroids (3500–5000) into the omentum of normoglycemic NSG mice in order to detect circulating human insulin in host animals. We measured low amounts of human insulin in the plasma of most transplanted mice, and these levels increased in 6 (of 10) mice after receiving an intraperitoneal glucose injection on day 30 post-transplantation (3.6 ± 0.9 vs 13.9 ± 3.7pmol/L, $$p \leq 0.014$$) (Supplementary Fig. 8). Transplants were repeated in other locations yielding similar results (Supplementary Table 2). As observed in the ACE grafts, a low number of INS + cells were identified in omentum grafts harvested at 30 days post-transplantation (Supplementary Fig. 8). These findings show that, despite restricted survival, reprogramming is maintained and 5TF cells maintain the capacity to release insulin in an in vivo setting. ## Discussion This study describes a direct reprogramming protocol based on the sequential introduction of five lineage-determining transcription factors that induces β-cell fate in human fibroblasts. Reprogrammed fibroblasts exhibit the concomitant activation of β-cell genes and the repression of fibroblastic and β-cell disallowed genes. Significantly, reprogrammed cells display functional features of β cells, including the ability to mobilize calcium and secrete insulin upon glucose stimulation. To the best of our knowledge, this is the first instance where it has been shown that skin fibroblasts can serve as cells of origin for β-cell derivation using transcription factor-based direct conversion methodologies. The N + P + M cocktail was initially described for the conversion of pancreatic exocrine cells into β cells in the mouse pancreas in situ4. Following studies demonstrated that these factors were also able to reprogram other endoderm-derived gastrointestinal tract cell lineages towards β-like cells6–9. Contrarily, information in the literature suggested that these transcription factors were ineffective in promoting β-cell fate from mesoderm-derived mouse and human fibroblasts4,9,18. However, in our view, the available studies did not fully analyze this option. In the present work, we examined a range of simultaneous and sequential transcription factor combinations, and were able to show that the N + P + M cocktail also works in human fibroblasts. Successful reprogramming is known to depend on a number of variables in addition to the selection of the proper reprogramming cocktail. We found that reprogramming was only possible when the three transcription factors where supplied by a single polycistronic adenovirus rather than multiple adenoviruses that expressed them separately, which is line with earlier research employing this same vector in acinar cells19. This may be owing to the random nature of viral co-infection, which makes it impossible to ensure that every infected cell receives every transgene when employing distinct viruses. Two other crucial parameters in reprogramming are the stoichiometry and the expression level of the reprogramming factors2,3,40,41. Regarding the former, a polycistronic expression system instead of using several vectors offers a more homogenous TF stoichiometry across the infected cells. We speculate that successful reprogramming in our study was made possible by the use of this polycistronic construct and the high levels of transgene expression achieved in fibroblasts through a minor modification of the infection protocol. It is interesting that the NPM factors, identified as a β-cell promoting reprogramming cocktail in acinar cells, induced expression of the GCG and SST genes in human fibroblasts. Despite the fact that INS was the most activated by difference, we used other factors (Pax4, Nkx2-2) to confer enhanced β-cell specificity in fibroblasts. These observations highlight the need to customize conversion transcription factor cocktails to the selected cell source. The 5TF-3D protocol used in our study promoted transcriptome alterations consistent with a change in cell identity from fibroblast toward a β-like cell. It is noteworthy that nearly $60\%$ of the detected transcriptional changes are gene repression events. In addition to the downregulation of fibroblast-specific genes, we found that 5TF cells had selectively suppressed the group of β-cell disallowed genes. This finding suggests that, in a cellular environment of a different lineage, lineage-specific transcription factors can promote the precise suppression of potentially damaging genes for their lineage. Intriguingly, SLC16A1 (MCT1), one of the founding members of the β-cell disallowed gene set, was not repressed in 5TF cells. Developmentally, β-cell disallowed genes are marked for repression by Polycomb Group (PcG) proteins in pancreatic progenitors during pancreas organogenesis42,43. However, the mechanisms that support their silencing in adult cells are still poorly understood and might vary amongst genes. These results spur future experiments into the way developmental transcription factors repress undesirable genes during the reprogramming process. Experimental data showing that intracellular calcium concentration and insulin release were stimulated by glucose suggest that fibroblast-derived 5TF cells are progressing toward a β-cell at the functional level. The levels of INS gene activation reached in 5TF cells were higher than those of reprogrammed human hepatocytes44 and comparable to those reported for reprogrammed human pancreatic duct cells9. They were, nonetheless, less than those of human islets. 5TF cells share traits with first generation pluripotent-derived β-like cells, such as decreased insulin production, low β-cell gene expression, and the absence of mature insulin granules36,45. Hence, we acknowledge that there is room for advancement and that optimization of our current reprogramming procedure is required in order to generate cells that are more similar to a primary β cell. A potential strategy is to modify the cell culture settings. Here, for instance, we demonstrated how switching to a different culture media after the addition of the NPM factors noticeably improved INS gene activation. We also showed how switching 5TF cells from a 2D to a 3D culture system increased expression of key β-cell genes and enabled converted cells to develop the ability to release insulin in response to glucose. Thus, additional adjustments to culture conditions, like extended culture times or agitation, may be helpful to improve quality of the generated cells46. In this same line, inclusion of soluble signaling molecules during or after introduction of the conversion transcription factors could be employed47–49, as it has been shown in direct reprogramming examples towards neurons50. In recent years, stem cell research has developed a significant body of knowledge on how to improve the maturity of pluripotent cell-derived insulin-secreting cells created in vitro22,23,35,36,51–56. We anticipate that this information can be very helpful in the effort to optimize direct reprogramming approaches from somatic cell types toward the β-cell lineage. Poor long-term survival of 5TF cells after transplantation precluded physiological studies in mouse models. Although the causes of cell loss remain to be elucidated, limited graft survival is a prevalent concern in transplantation of cadaveric donor islets and in vitro created islet tissue57. In our case, it is possible that additional endocrine and/or non-endocrine cell types normally found in islets such as endothelial cells are required for better engraftment and prolonged survival58,59. Additionally, our observation of sustained expression of exogenous reprogramming factors in 5TF cells raises the question of what effects this might have, especially the continued presence of Neurog3 and Pax4 given that these factors are not present in mature β cells27,60. Future effort will be required to develop reprogramming strategies to guarantee that the conversion factors are turned off if necessary. *The* generation of substitute β cells from a cell source that can be replenished has been a long-standing major goal in diabetes research. The most advanced approach to date involves the guided differentiation of pluripotent stem cells to islet cells. The first clinical studies involving actual patients have been made possible by the enormous advancements made in this field over the past fifteen years36,45,46,54,56,61,62. Recent publication of the first mid-term results of one of these trials shows positive outcomes as well as the need to continue improving current differentiation protocols and transplantation techniques63,64. On the other hand, the idea of direct cellular reprogramming to produce therapeutically relevant cell types, as an alternative to their derivation from stem cells, regained momentum with the discovery of iPS cells65,66. Since then, there has been a notable increase in the amount of somatic cells produced from readily accessible cell types, such as fibroblasts67–69. In the β-cell field, however, available evidence indicated that human fibroblasts were not keen to change identity toward a β cell via direct reprogramming. Our results challenge this view and show that this is possible if the appropriate combination of conversion factors and conditions are found. The value of our approach is related to the major benefit of using a cell source that is readily available, such as the skin fibroblast, in terms of translational potential. Two further assets that should be taken into account from a clinical standpoint are the possibility of auto-transplantation and the avoidance of tumor-related concerns linked to pluripotent cell states. Finally, the relative simplicity of direct reprogramming methodology compared to pluripotent stem cell derivation supports the ongoing interest in developing this kind of approaches. In conclusion, here we demonstrate that human fibroblasts can be directly converted toward a β-cell fate using a defined set of developmental transcription factors. Further research should refine this strategy so that generated insulin-producing cells more closely resemble primary β cells. These findings provide a promising starting point for future investigation into an alternative pathway to produce β-like cells for therapeutic and modeling purposes. ## Fibroblasts Human fibroblasts were obtained from a child foreskin biopsy after signed informed content and approval of the institutional Review Board of the Center of Regenerative Medicine in Barcelona. In brief, skin biopsy was collected in sterile saline solution, divided into small pieces, and allowed to attach to cell culture dishes before adding Iscove’s modified Dulbecco’s medium (Invitrogen, Carlsbad, CA, USA) supplemented with $10\%$ human serum (Sigma, St. Louis, MO, USA) and penicillin/streptomycin (0.5X) (Invitrogen). After 10 days of culture at 37 °C, $5\%$ CO2, fibroblast outgrowths were dissociated and split 1:4 using a recombinant trypsin‐like enzyme (TrypLE Select, Invitrogen). Preparation was negative for hematopoietic markers, including CD34. This fibroblast line (HFF1) was used to design the reprogramming protocol. Once established, the protocol was validated in another HFF preparation (HFF2) that was purchased from a commercial source (SCRC1041TM, ATCC, Manassas, VA, USA). ## Human islets Human islets were prepared by collagenase digestion followed by density gradient purification at the Laboratory of Cell Therapy for Diabetes (Hospital Saint-Eloi, Montpellier, France), as previously described70. After reception in Barcelona, human islets were maintained in culture at 37 °C, $5\%$ CO2 for 1–3 days in RPMI-1640 with 5.5 mM glucose, $10\%$ fetal bovine serum (FBS) and antibiotics, before performing the experiments. Experiments were performed in agreement with the local ethic committee (CHU, Montpellier) and the institutional ethical committee of the French Agence de la Biomédecine (DC Nos. 2014-2473 and 2016-2716). Informed consent was obtained for all donors. ## Recombinant adenoviruses The adenoviral expression vector pAd/CMV/V5-DEST carrying mouse Neurog3, Pdx1, MafA and 2A-Cherry under the CMV promoter and separated by self-cleaving 2A peptides was kindly provided by Dr. Q. Zhou, Cornell University19. The recombinant adenovirus (hereafter termed Ad-NPM) was generated after Pac1 digestion and transfection into HEK293 cells. The recombinant adenovirus encoding Pdx1 was kindly provided by the Beta Cell Biology Consortium. The recombinant adenovirus encoding MafA was purchased from Vector Biolabs (Chicago, IL, USA). All other recombinant adenoviruses encoding single transcription factors (Neurog3, NKX2-2, Nkx6.1, and Pax4) were described previously71,72. Crude virus lysates were used for infection of fibroblasts. ## Reprogramming protocol Fibroblasts were grown in DMEM-F12 media supplemented with $10\%$ (v/v) fetal bovine serum (FBS), 100 U/ml penicillin, 100 µg/ml streptomycin, and $1\%$ Glutamax. They were plated onto 96-well plates (9500 cells per well) for MTT and BrdU assays, onto 12-well plates (1.25 × 105 cells/well) for gene expression, insulin secretion, immunofluorescence and caspase assays and onto 10 cm plates (1.5 × 106 cells) or T-75 flasks (3.0 × 106 cells) for transplantation experiments. Reprogramming was initiated when fibroblasts reached $80\%$ confluence, normally 1–2 days post seeding. Cells were sequentially incubated with 15 moi (multiplicity of infection) of Ad-NPM (day 1), 50 moi of Ad-Pax4 (day 4), and 50 moi of Ad-Nkx2-2 (day 7). As fibroblasts show limited infection by adenoviral vectors73,74, we added a DNA transfection reagent to the virus incubation conditions to improve transgene expression. This small change significantly increased viral transduction efficiency and allowed human fibroblasts to express significant amounts of the reprogramming factors (Supplementary Fig. 9). This reagent was Superfect (Qiagen, Venlo, Netherlands) or jet-PEI (Polyplus, IIlkirch, France). In brief, for one well of a 12-well plate, the appropriate amount of virus was pre-incubated with 1.5–2.0 μl of Superfect or JetPEI in 0.1 ml un-supplemented media for 10–15 min at room temperature. Then, the virus/transfection reagent mix was gently mixed with 0.9 ml supplemented media and added to the cells. Cells were incubated with the virus for 16–18 h (NPM) or 6–8 h (Pax4/Nkx2-2). The volume of culture media, adenovirus crude lysate, and transfection reagent used was scaled down or up according to the well size. After NPM virus removal, media was changed to RPMI-1640 medium containing $6\%$ FBS and antibiotics, and this media formulation was maintained throughout the remaining of the reprogramming protocol. Adenovirus doses and/or time of addition were picked in pilot experiments based on a compromise between high INS gene expression and low cytotoxicity. Spheroidal cell aggregates (spheroids) were prepared one day after Ad-Nkx2.2 infection. Cells were trypsinized and transferred to 96-well Nunclon Sphera plates (Thermo Scientific) to generate spheroids of 1200–1800 cells each, which were used for gene expression, immunofluorescence and insulin secretion assays. For transplantation purposes, spheroids (1000–1200 cells/spheroid) were generated in AggreWell-400 plates (StemCell Technologies, Saint Égrève, France). Cell aggregation was performed in RPMI-1640 media supplemented as detailed above. ## Gene expression assays Total RNA from cultured cells was isolated using NucleoSpin®RNA (Macherey-Nagel Düren, Germany) following the manufacturer’s manual. Total RNA from eyes was extracted using Trizol reagent (Sigma) and then cleaned and DNAse-treated using RNeasy mini columns (Qiagen) prior to cDNA synthesis. First-strand cDNA was prepared using Superscript III Reverse Transcriptase (Invitrogen) and random hexamers in a total volume of 20 ul and $\frac{1}{40}$ to $\frac{1}{200}$ of the resulting cDNA was used as a template for real time PCR reactions. Real time PCR was performed on an ABI Prism 7900 detection system using Gotaq master mix (Promega, Madison, WI, USA). Expression relative to the housekeeping gene TBP was calculated using the delta(d)Ct method and expressed as 2^(-dCT) unless otherwise indicated. Primer sequences are provided in Supplementary Table 1. ## Proliferation and viability assays For quantification of cell proliferation, cells in 96-well plates were cultured overnight with medium containing 5-bromo-2’-deoxuridine (BrdU). BrdU incorporation was determined colorimetrically with the Cell proliferation ELISA kit (Roche, Basilea, Switzerland) following the manufacturer’s instructions. For assessing cell viability, cells grown in 96-well plates were incubated with medium containing 0.75 mg/ml of 3-(4,5-dimethythiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) for 3 h at 37 °C. The resulting formazan crystals were solubilized in Isopropanol/0.04 N HCl solution and optical density was read at 575 and 650 nm using a Synergy HT reader (BIO-TEK Instruments, Winooski, VT, USA). The OD [575-650] was expressed relative to control fibroblasts, which were given the value of $100\%$. ## Calcium imaging To study glucose-dependent calcium influx, cells were washed with Hanks’ Balanced Salt Solution (HBSS, Sigma) and incubated with Fluo4-AM (Life Technologies) in fresh Hepes-buffered Krebs-Ringer buffer (Krb) containing 2 mM glucose for 1 h at 37 °C in the dark. Intracellular calcium fluorescence was recorded from fluo-4-loaded cells using a Leica TCS SPE confocal microscope with an incubation chamber set at 37 °C, and a 40× oil immersion objective. Fluorescent images and average fluorescence intensity were acquired at 600 Hz every 1.8 s, using a 488 nm excitation laser, an emission set at 520 with a bandwidth of 10 nm. Image registry consisted of: 5 min in 2 mM glucose-Krb buffer, 10 min in 22 mM glucose-Krb buffer and 5 min in 30 mM KCl-Krb buffer. Average fluorescence intensity images of each individual cell were analyzed with LAS AF Lite and Fuji programs. ## Insulin secretion assays Cells in 2D or 3D aggregates were washed with phosphate-buffered saline (PBS) and then incubated in Krb for 45 min at 37 °C with low glucose (2 mM) to remove residual insulin. Cells were incubated in Krb with low glucose for 90 min and supernatant collected. Then clusters were incubated in Krb with high glucose (20 mM or as indicated in figure) for 90 min and supernatant collected. Human insulin in the collected supernatants was measured using a human Insulin ELISA kit (Crystal Chem, Zaandam, Netherlands). ## RNA sequencing Total RNA was extracted using NucleoSpin®RNA (Macherey-Nagel). Quantity and quality of the RNA was assessed using Qubit fluorimeter and TapeStation Instrument, respectively. All samples used for sequencing had RIN > 9. mRNA strand-specific RNA libraries were generated using 200 ng of total RNA with the Illumina® Stranded mRNA Prep Ligation kit and IDT for Illumina RNA UD Indexes following the manufacturer’s instructions. Each library was sequenced on an Illumina NextSeq2000 (Illumina, Inc.) in paired-end mode with a read length of 2 × 50 base pair. More than 60 million paired-end reads were generated for each sample/condition. Quality of sequenced reads was assessed using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). To better discern between reads coming from mouse transgenes and from human endogenous genes, we generated a custom transcriptome by adding the Gencode M10 mouse transcripts from the transgenes Pdx1, Pax4, Neurog3 and Mafa to the human GENCODE release 39 human transcripts. We assigned reads to this custom transcriptome using Salmon v.1.3.075 with parameters “-l A --validateMappings”. We obtained > 40 million aligned reads per sample. Next, we summarised the transcript counts to gene counts and used this as input for downstream analysis. Normalization and differential analysis were performed using the DESeq2 R package v.1.36.076. Threshold for significance was set at an FDR-adjusted p-value < 0.05 and an absolute log2 fold change (FC) > 1. *All* genes that did not reach significance or did not pass the log2 FC cutoff were classified as stable. As input for the heatmap in Fig. 5d, we produced a matrix of regularised log-transformed gene counts applying the “rlog” function from DESeq2. Gene Ontology enrichment analysis was performed using the goseq R package v.1.48.077 and the resulting p-values were adjusted for multiple testing using the FDR method. KEGG enrichment analysis was produced using the ClusterProfiler R package v.4.4.178. Gene Set Enrichment Analysis (GSEA) was conducted using the fgsea function from the fgsea R package v.1.22.0 with default parameters, except in the case of “Beta cell disallowed genes” in which we specified the expected direction of enrichment with the argument “scoreType = ‘neg’“. In all the above analyses, terms were considered statistically significant when adjusted p-values < 0.05. ## Electron microscopy Cell spheroids were collected, washed with PBS and fixed with $4\%$ paraformaldehyde/ $0.5\%$ glutaraldehyde (Sigma) mixture in 0.1 M phosphate buffer (PB) pH7.4 for 30 min at 4 °C and gentle agitation. Cells were then transferred to fresh fixation solution and maintained at 4 °C until secondarily fixed with $1\%$ uranyl acetate and $1\%$ osmium tetroxide. Cells were then dehydrated, embedded in Spurr’s Resin and sectioned using Leica ultramicrotome (Leica Microsystems). Conventional transmission electron microscopy (TEM) images were acquired from thin sections using a JEOL-1010 electron microscopy equipped with an SC1000 ORIUS-CCD digital camara (Gatan). ## Mouse studies Adult (8–20 week old) normoglycemic male NOD scid gamma (NSGTM) mice (catalog no 005557, Jackson Laboratories) were used as transplantation recipients. Approximately 300 cell spheroids (reprogrammed cells or parental fibroblasts) or 150–200 human islets were transplanted into the anterior chamber of the eye (ACE)58. In brief, an incision was made in the cornea near the corneoscleral junction, and a cannula (0.4 mm internal diameter) loaded with spheroids/islets, connected to a 500 µL syringe, was introduced into the incision. Cells were carefully injected without damaging the iris. The omentum, subcutatenous space and kidney were used for transplantation of higher number of cell spheroids (between 3500 and 5000). For transplantation in the omentum and subcutaneous space, cell spheroids were preloaded in collagen/Matrigel scaffolds. Briefly, the cell-laden collagen/Matrigel hydrogel was prepared by first mixing 110 μl of a 4 mg/ml rat tail type I collagen (Corning, NY, USA) solution with 40 μl of Matrigel (Corning) and then adding the mix to 5TF cell spheroids (pelleted by gentle centrifugation) and poured onto a cylindrical 8 mm diameter × 1 mm thick PDMS mold (Dow Corning Sylgard 184 Silicone Elastomer). The hydrogel was polymerized for 20 min at 37 °C, detached from the mold and maintained in tissue culture dishes with warm RPMI-1640 medium until transplant (usually 2–3 h). Constructs were placed on the omentum close to the duodenal-stomach junction. Alternatively, constructs were introduced in the abdominal subcutaneous space through a small (5 mm) incision. Transplantation in the kidney was performed following standard procedures79. NSG mice transplanted with fibroblast spheroids an/or non-transplanted NSG mice were used as controls in transplantation experiments. To assess vascularization and cell viability in ACE implants, 5TF cell spheroids were labeled with the long-term tracer for viable cells Vybrant CFDA SE (Invitrogen) before transplantation. At day 10 post-transplantation, mice received an intravenous injection of RITC-dextran and in vivo imaging was used to assess functional vascularization and cell viability58. For the determination of glucose-induced insulin secretion, mice were fasted for 5–6 h and then injected intraperitoneally with glucose (3 g/Kg). Tail blood was collected before and after (20 min) the glucose challenge. Aqueous humor from mice with ACE implants was obtained at time of sacrifice and kept frozen until human insulin determination. Human insulin in plasma and in aqueous humor was determined using an ultrasensitive Human Insulin ELISA (Chrystal Chem). The Animal Research Committee of the University of Barcelona approved all animal procedures. European and local guidelines (Generalitat de Catalunya) on accommodation and care of laboratory animal were followed. ## Immunofluorescence and morphometric measurements Cells grown in 2D were fixed with $4\%$ (v/v) paraformaldehyde (PFA) during 15 min and incubated with blocking solution ($0.25\%$ (v/v) Triton, $6\%$ (v/v) donkey serum, $5\%$ (w/v) BSA in PBS for 1 h at room temperature. Slides were then incubated with primary antibodies diluted in PBS-triton $0.1\%$ (v/v) containing $1\%$ donkey serum overnight at 4 °C. 5TFcells grown in spheroids were fixed with $4\%$ (v/v) paraformaldehyde (PFA) for 15 min at 4 °C, permeabilized with $0.5\%$ (v/v) Triton in PBS for 20 min and blocked with $0.5\%$ (v/v) Triton/ FBS $10\%$ (v/v) in PBS during 1 h at room temperature. Slides were then incubated with primary antibodies diluted in blocking solution overnight at 4 °C. Eyes and omentum implants were fixed overnight in 2 and $4\%$ (v/v) PFA, respectively, dehydrated with ethanol gradient, cleared with xylene and paraffin-embedded. 3 µm thick eye sections were used for standard immunofluorescence staining protocol. Primary antibodies used were: Insulin (DAKO, 1:400, Fig. 2b); Insulin/C-PEP (Hybridoma Bank; 1:40, used in all figures); HLA (Abcam, 1:100), PCSK1 (Gene Tex; 1:100); KCNJ11, PTPRN, and NCAM1 (Santa Cruz, 1:50). The antigen-primary antibody immune complex was visualized with secondary antibodies conjugated to Alexa Fluor 488 (Jackson Immunoresearch, 1:250), Alexa fluor 555 (Molecular Probes, 1:400) or Alexa fluor 647 (Jackson Immunoreserach, 1:250). Cell nuclei were counterstained with Hoechst 33258 (SIGMA, 1:500). Fluorescent images were captured using a Leica TCS SPE confocal microscope. For morphometric analysis, total ACE grafts were sectioned at 3 μm and distributed as serial sections onto two sets of 10 slides each. At least 10 sections per graft, 90μm apart, were used to quantify the number of HLA+/INS+ and transcription factor+/INS+ cells using ImageJ/Fiji (National Institutes of Health, Bethesda, MD, USA; http://rsb.info.nih.gov/ij/) software. In 10-day grafts, to determine the percentage of INS+/HLA+ cells, 600–2000 HLA+ cells per graft were counted; and to determine the percentage of transcription factor+/INS+ cells, 200–1000 INS+ cells per graft were counted. ## Statistics and reproducibiltiy Data are presented as mean ± standard error of the mean (SEM) from at least three independent reprogramming experiments, with one to four biological replicates per experiment. Significant differences between the means were analyzed by the two-tailed unpaired Student’s t-test, one sample t-test or one-way ANOVA followed by Tukey’s or Dunnett’s multiple comparison tests as indicated in the figure legends. Statistical analysis was performed with GraphPad Prism 8.00 and Microsoft Office Excel 2007 and differences were considered significant at $P \leq 0.05.$ No methods were used to determine whether the data met assumptions of the statistical approach (e.g., test for normal distribution). ## Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. ## Supplementary information Peer Review File Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Video 1 Supplementary Video 2 Reporting Summary The online version contains supplementary material available at 10.1038/s42003-023-04627-2. ## Peer review information Communications Biology thanks Nidheesh Dadheech and Qiao Zhou for their contribution to the peer review of this work. Primary Handling Editor: Eve Rogers. Peer reviewer reports are available. ## References 1. Graf T. **Historical origins of transdifferentiation and reprogramming**. *Cell Stem Cell* (2011) **9** 504-516. DOI: 10.1016/j.stem.2011.11.012 2. Vierbuchen T, Wernig M. **Molecular roadblocks for cellular reprogramming**. *Mol. Cell* (2012) **47** 827-838. DOI: 10.1016/j.molcel.2012.09.008 3. Xu J, Du Y, Deng H. **Direct lineage reprogramming: strategies, mechanisms, and applications**. *Cell Stem Cell* (2015) **16** 119-134. DOI: 10.1016/j.stem.2015.01.013 4. Zhou Q, Brown J, Kanarek A, Rajagopal J, Melton DA. **In vivo reprogramming of adult pancreatic exocrine cells to beta-cells**. *Nature* (2008) **455** 627-632. DOI: 10.1038/nature07314 5. Li W. **long-term persistence and development of induced pancreatic beta cells generated by lineage conversion of acinar cells**. *Nat. Biotechnol.* (2014) **32** 1223-1232. DOI: 10.1038/nbt.3082 6. Ariyachet C. **Reprogrammed stomach tissue as a renewable source of functional beta cells for blood glucose regulation**. *Cell Stem Cell* (2016) **18** 410-421. DOI: 10.1016/j.stem.2016.01.003 7. Chen YJ. **De novo formation of insulin-producing “neo-beta cell islets” from intestinal crypts**. *Cell Rep.* (2014) **6** 1046-1058. DOI: 10.1016/j.celrep.2014.02.013 8. Furuyama K. **Diabetes relief in mice by glucose-sensing insulin-secreting human alpha-cells**. *Nature* (2019) **567** 43-48. DOI: 10.1038/s41586-019-0942-8 9. Lee J. **Expansion and conversion of human pancreatic ductal cells into insulin-secreting endocrine cells**. *Elife* (2013) **2** e00940. DOI: 10.7554/eLife.00940 10. Galivo F. **Reprogramming human gallbladder cells into insulin-producing β-like cells**. *PLoS One* (2017) **12** e0181812. DOI: 10.1371/journal.pone.0181812 11. Yechoor V. **Neurogenin3 is sufficient for transdetermination of hepatic progenitor cells into neo-islets in vivo but not transdifferentiation of hepatocytes**. *Dev. Cell* (2009) **16** 358-373. DOI: 10.1016/j.devcel.2009.01.012 12. Ieda M. **Direct reprogramming of fibroblasts into functional cardiomyocytes by defined factors**. *Cell* (2010) **142** 375-386. DOI: 10.1016/j.cell.2010.07.002 13. Hiramatsu K. **Generation of hyaline cartilaginous tissue from mouse adult dermal fibroblast culture by defined factors**. *J. Clin. Invest.* (2011) **121** 640-657. DOI: 10.1172/JCI44605 14. Vierbuchen T. **Direct conversion of fibroblasts to functional neurons by defined factors**. *Nature* (2010) **463** 1035-1041. DOI: 10.1038/nature08797 15. Najm FJ. **Transcription factor-mediated reprogramming of fibroblasts to expandable, myelinogenic oligodendrocyte progenitor cells**. *Nat. Biotechnol.* (2013) **31** 426-433. DOI: 10.1038/nbt.2561 16. Huang P. **Direct reprogramming of human fibroblasts to functional, expandable hepatocytes**. *Cell Stem Cell* (2014) **14** 370-384. DOI: 10.1016/j.stem.2014.01.003 17. Han J-K. **Direct conversion of adult skin fibroblasts to endothelial cells by defined factors**. *Circulation* (2014) **130** 1168-1178. DOI: 10.1161/CIRCULATIONAHA.113.007727 18. Akinci E. **Reprogramming of various cell types to a beta-like state by Pdx1, Ngn3, and MafA**. *PLoS One* (2013) **8** e82424. DOI: 10.1371/journal.pone.0082424 19. Li W. **In vivo reprogramming of pancreatic acinar cells to three islet endocrine subtypes**. *Elife* (2014) **3** e01846. DOI: 10.7554/eLife.01846 20. Tomaru Y. **A transient disruption of fibroblastic transcriptional regulatory network facilitates trans-differentiation**. *Nucleic Acids Res.* (2014) **42** 8905-8913. DOI: 10.1093/nar/gku567 21. Sander M. **Homeobox gene Nkx6.1 lies downstream of Nkx2.2 in the major pathway of beta-cell formation in the pancreas**. *Development* (2000) **127** 5533-5540. DOI: 10.1242/dev.127.24.5533 22. Rezania A. **Enrichment of human embryonic stem cell-derived NKX6.1-expressing pancreatic progenitor cells accelerates the maturation of insulin-secreting cells in vivo**. *Stem Cells* (2013) **31** 2432-2442. DOI: 10.1002/stem.1489 23. Rezania A. **Maturation of human embryonic stem cell-derived pancreatic progenitors into functional islets capable of treating pre-existing diabetes in mice**. *Diabetes* (2012) **61** 2016-2029. DOI: 10.2337/db11-1711 24. Smith SB. **Neurogenin3 and hepatic nuclear factor 1 cooperate in activating pancreatic expression of Pax4**. *J. Biol. Chem.* (2003) **278** 38254-38259. DOI: 10.1074/jbc.M302229200 25. Collombat P. **The simultaneous loss of Arx and Pax4 genes promotes a somatostatin-producing cell fate specification at the expense of the alpha- and beta-cell lineages in the mouse endocrine pancreas**. *Development* (2005) **132** 2969-2980. DOI: 10.1242/dev.01870 26. Sosa-Pineda B, Chowdhury K, Torres M, Oliver G, Gruss P. **The Pax4 gene is essential for differentiation of insulin-producing beta cells in the mammalian pancreas**. *Nature* (1997) **386** 399-402. DOI: 10.1038/386399a0 27. Wang J. **The concerted activities of Pax4 and Nkx2.2 are essential to initiate pancreatic b-cell differentiation**. *Dev. Biol.* (2004) **266** 178-189. DOI: 10.1016/j.ydbio.2003.10.018 28. Sander M. **Genetic analysis reveals that PAX6 is required for normal transcription of pancreatic hormone genes and islet development**. *Genes Dev.* (1997) **11** 1662-1673. DOI: 10.1101/gad.11.13.1662 29. St-Onge L, Sosa-Pineda B, Chowdhury K, Mansouri A, Gruss P. **Pax6 is required for differentiation of glucagon-producing alpha-cells in mouse pancreas**. *Nature* (1997) **387** 406-409. DOI: 10.1038/387406a0 30. Ramzy AK, T J. **Revisiting proinsulin processing: evidence that human β-cells process proinsulin with prohormone convertase (PC) 1/3 but not PC2**. *Diabetes* (2020) **69** 1451-1462. DOI: 10.2337/db19-0276 31. Schuit F. **b-Cell-specific gene repression: a mechanism to protect against inappropriate or maladjusted insulin secretion?**. *Diabetes* (2012) **61** 969-975. DOI: 10.2337/db11-1564 32. Pullen TJ. **Identification of genes selectively disallowed in the pancreatic islet**. *Islets* (2010) **2** 89-95. DOI: 10.4161/isl.2.2.11025 33. Pullen TJ, Huising MO, Rutter GA. **Analysis of purified pancreatic islet beta and alpha cell transcriptomes reveals 11β-hydroxysteroid dehydrogenase (Hsd11b1) as a novel disallowed gene**. *Front. Genet.* (2017) **8** 41. DOI: 10.3389/fgene.2017.00041 34. Brereton MF, Vergari E, Zhang Q, Clark A. **Alpha-, delta- and PP-cells: are they the architectural cornerstones of islet structure and co-ordination?**. *J. Histochem Cytochem* (2015) **63** 575-591. DOI: 10.1369/0022155415583535 35. D’Amour KA. **Production of pancreatic hormone-expressing endocrine cells from human embryonic stem cells**. *Nat. Biotechnol.* (2006) **24** 1392-1401. DOI: 10.1038/nbt1259 36. Pagliuca FW. **Generation of functional human pancreatic beta cells in vitro**. *Cell* (2014) **159** 428-439. DOI: 10.1016/j.cell.2014.09.040 37. Henquin J, Dufrane D, Kerr-Conte J, Nenquin M. **Dynamics of glucose-induced insulin secretion in normal human islets**. *Am. J. Physiol. Endocrinol. Metab.* (2015) **309** E640-E650. DOI: 10.1152/ajpendo.00251.2015 38. Adeghate E, Donath T. **Morphological findings in long-term pancreatic tissue transplants in the anterior eye chamber of rats**. *Pancreas* (1990) **5** 298-305. DOI: 10.1097/00006676-199005000-00009 39. Leibiger IB, Berggren PO. **Intraocular in vivo imaging of pancreatic islet cell physiology/pathology**. *Mol. Metab.* (2017) **6** 1002-1009. DOI: 10.1016/j.molmet.2017.03.014 40. Carey BW. **Reprogramming factor stoichiometry influences the epigenetic state and biological properties of induced pluripotent stem cells**. *Cell Stem Cell* (2011) **9** 588-598. DOI: 10.1016/j.stem.2011.11.003 41. Brouwer M, Zhou H, Kasri NN. **Choices for Induction of Pluripotency: Recent Developments in Human Induced Pluripotent Stem Cell Reprogramming Strategies**. *Stem Cell Rev. Rep.* (2016) **12** 54-72. DOI: 10.1007/s12015-015-9622-8 42. van Arensbergen J. **Derepression of Polycomb targets duringpancreatic organogenesis allows insulin-producing b-cellsto adopt a neural gene activity program**. *Genome Res.* (2010) **20** 722-732. DOI: 10.1101/gr.101709.109 43. van Arensbergen J. **Ring1b bookmarks genes in pancreaticembryonic progenitors for repressionin adult b cells**. *Genes Dev.* (2013) **27** 52-63. DOI: 10.1101/gad.206094.112 44. Sapir T. **Cell-replacement therapy for diabetes: Generating functional insulin-producing tissue from adult human liver cells**. *Proc. Natl Acad. Sci. USA* (2005) **102** 7964-7969. DOI: 10.1073/pnas.0405277102 45. Kroon E. **Pancreatic endoderm derived from human embryonic stem cells generates glucose-responsive insulin-secreting cells in vivo**. *Nat. Biotechnol.* (2008) **26** 443-452. DOI: 10.1038/nbt1393 46. Balboa D. **Functional, metabolic and transcriptional maturation of human pancreatic islets derived from stem cells**. *Nat. Biotechnol.* (2022) **40** 1042-1055. DOI: 10.1038/s41587-022-01219-z 47. Fontcuberta-PiSunyer M. **Modulation of the endocrine transcriptional program by targeting histone modifiers of the H3K27me3 mark**. *Biochim. Biophys. Acta* (2018) **1861** 473-480. DOI: 10.1016/j.bbagrm.2018.03.003 48. Liu ML. **Small molecules enable neurogenin 2 to efficiently convert human fibroblasts into cholinergic neurons**. *Nat. Commun.* (2013) **4** 2183. DOI: 10.1038/ncomms3183 49. Zhu S. **Reprogramming of human primary somatic cells by OCT4 and chemical compounds**. *Cell Stem Cell* (2010) **7** 651-655. DOI: 10.1016/j.stem.2010.11.015 50. Sepehrimanesh M, Akter M, Ding B. **Direct conversion of adult fibroblasts into motor neurons**. *Star. Protoc.* (2021) **2** 100917. DOI: 10.1016/j.xpro.2021.100917 51. Russ HA. **Controlled induction of human pancreatic progenitors produces functional beta-like cells in vitro**. *EMBO J.* (2015) **34** 1759-1772. DOI: 10.15252/embj.201591058 52. Liu H. **Chemical combinations potentiate human pluripotent stem cell-derived 3D pancreatic progenitor clusters toward functional β cell**. *Nat. Commun.* (2021) **12** 3330. DOI: 10.1038/s41467-021-23525-x 53. Hrvatin S. **Differentiated human stem cells resemble fetal, not adult, beta cells**. *Proc. Natl Acad. Sci. USA* (2014) **111** 3038-3043. DOI: 10.1073/pnas.1400709111 54. D’Amour KA. **Efficient differentiation of human embryonic stem cells to definitive endoderm**. *Nat. Biotechnol.* (2005) **23** 1534-1541. DOI: 10.1038/nbt1163 55. Bruin JE. **Maturation and function of human embryonic stem cell-derived pancreatic progenitors in macroencapsulation devices following transplant into mice**. *Diabetologia* (2013) **56** 1987-1998. DOI: 10.1007/s00125-013-2955-4 56. Nair GG. **Recapitulating endocrine cell clustering in culture promotes maturation of human stem-cell-derived β cells**. *Nat. Cell Biol.* (2019) **21** 263-274. DOI: 10.1038/s41556-018-0271-4 57. Gamble A, Pepper AR, Bruni A, Shapiro AM. **The journey of islet cell transplantation and future development**. *Islets* (2018) **10** 80-94. DOI: 10.1080/19382014.2018.1428511 58. Figueiredo H. **Targeting pancreatic islet PTP1B improves islet graft revascularization and transplant outcomes**. *Sci. Transl. Med.* (2019) **11** eaar6294. DOI: 10.1126/scitranslmed.aar6294 59. Aghazadeh Y. **Microvessels support engraftment and functionality of human islets and hESC-derived pancreatic progenitors in diabetes models**. *Cell Stem Cell* (2021) **28** 1936-1949. DOI: 10.1016/j.stem.2021.08.001 60. Schwitzgebel VM. **Expression of neurogenin3 reveals an islet cell precursor population in the pancreas**. *Development* (2000) **127** 3533-3542. DOI: 10.1242/dev.127.16.3533 61. Maehr R. **Generation of pluripotent stem cells from patients with type 1 diabetes**. *Proc. Natl Acad. Sci. USA* (2009) **106** 15768-15773. DOI: 10.1073/pnas.0906894106 62. Rezania A. **Reversal of diabetes with insulin-producing cells derived in vitro from human pluripotent stem cells**. *Nat. Biotechnol.* (2014) **32** 1121-1133. DOI: 10.1038/nbt.3033 63. Ramzy A. **Implanted pluripotent stem-cell-derived pancreaticendoderm cells secrete glucose-responsiveC-peptide in patients with type 1 diabetes**. *Cell Stem Cell* (2021) **28** 2047-2061. DOI: 10.1016/j.stem.2021.10.003 64. Shapiro AM. **Insulin expression and C-peptide in type 1 diabetessubjects implanted with stem cell-derived pancreaticendoderm cells in an encapsulation device**. *Cell Rep. Med.* (2021) **2** 100466. DOI: 10.1016/j.xcrm.2021.100466 65. Takahashi K. **Induction of pluripotent stem cells from adult human fibroblasts by defined factors**. *Cell* (2007) **131** 861-872. DOI: 10.1016/j.cell.2007.11.019 66. Takahashi K, Yamanaka S. **Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors**. *Cell* (2006) **126** 663-676. DOI: 10.1016/j.cell.2006.07.024 67. Wang J, Yang YH, Liu JS, L Q. **Direct cell reprogramming: approaches, mechanisms, and progress**. *Nat. Rev. Mol. Cell Biol.* (2021) **22** 410-424. DOI: 10.1038/s41580-021-00335-z 68. Bocchi R, Masserdotti G, Götz M. **Direct neuronal reprogramming: fast forward from new concepts toward therapeutic approaches**. *Neuron* (2022) **110** 366-393. DOI: 10.1016/j.neuron.2021.11.023 69. Srivastava D, DeWitt N. **In vivo cellular reprogramming: the next generation**. *Cell* (2016) **166** 1386-1396. DOI: 10.1016/j.cell.2016.08.055 70. Broca C. **Proteasome dysfunction mediates high glucose-induced apoptosis in rodent beta cells and human islets**. *PLoS One* (2014) **9** e92066. DOI: 10.1371/journal.pone.0092066 71. Gasa R. **Proendocrine genes coordinate the pancreatic islet differentiation program in vitro**. *Proc. Natl Acad. Sci. USA* (2004) **101** 13245-13250. DOI: 10.1073/pnas.0405301101 72. Lynn FC, Sanchez L, Gomis R, German MS, Gasa R. **Identification of the bHLH factor Math6 as a novel component of the embryonic pancreas transcriptional network**. *PLoS One* (2008) **3** e2430. DOI: 10.1371/journal.pone.0002430 73. Granio O. **Improved adenovirus type 5 vector-mediated transduction of resistant cells by piggybacking on coxsackie B-adenovirus receptor-pseudotyped baculovirus**. *J. Virol.* (2009) **83** 6048-6066. DOI: 10.1128/JVI.00012-09 74. Toh ML. **Enhancement of adenovirus-mediated gene delivery to rheumatoid arthritis synoviocytes and synovium by fiber modifications: role of arginine-glycine-aspartic acid (RGD)- and non-RGD-binding integrins**. *J. Immunol.* (2005) **175** 7687-7698. DOI: 10.4049/jimmunol.175.11.7687 75. Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. **Salmon provides fast and bias-aware quantification of transcript expression**. *Nat. Methods* (2017) **14** 417-419. DOI: 10.1038/nmeth.4197 76. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol.* (2014) **15** 550-571. DOI: 10.1186/s13059-014-0550-8 77. Young MD, Wakefield MJ, Smyth GK, Oshlack A. **Gene ontology analysis for RNA-seq: accounting for selection bias**. *Genome Biol.* (2010) **11** R14. DOI: 10.1186/gb-2010-11-2-r14 78. Wu T. **clusterProfiler 4.0: a universal enrichment tool for interpreting omics data**. *Innovation* (2021) **2** 100141. PMID: 34557778 79. Marin-Cañas S. **Pancreatic ductal cells may have a negative effect on human islet transplantation**. *PLoS One* (2019) **14** e0220064. DOI: 10.1371/journal.pone.0220064
--- title: Querying a Clinical Data Warehouse for Combinations of Clinical and Imaging Data authors: - Mathias Kaspar - Leon Liman - Caroline Morbach - Georg Dietrich - Lea Katharina Seidlmayer - Frank Puppe - Stefan Störk journal: Journal of Digital Imaging year: 2022 pmcid: PMC10039164 doi: 10.1007/s10278-022-00727-3 license: CC BY 4.0 --- # Querying a Clinical Data Warehouse for Combinations of Clinical and Imaging Data ## Abstract This study aims to show the feasibility and benefit of single queries in a research data warehouse combining data from a hospital’s clinical and imaging systems. We used a comprehensive integration of a production picture archiving and communication system (PACS) with a clinical data warehouse (CDW) for research to create a system that allows data from both domains to be queried jointly with a single query. To achieve this, we mapped the DICOM information model to the extended entity–attribute–value (EAV) data model of a CDW, which allows data linkage and query constraints on multiple levels: the patient, the encounter, a document, and a group level. Accordingly, we have integrated DICOM metadata directly into CDW and linked it to existing clinical data. We included data collected in 2016 and 2017 from the Department of Internal Medicine in this analysis for two query inquiries from researchers targeting research about a disease and in radiology. We obtained quantitative information about the current availability of combinations of clinical and imaging data using a single multilevel query compiled for each query inquiry. We compared these multilevel query results to results that linked data at a single level, resulting in a quantitative representation of results that was up to $112\%$ and $573\%$ higher. An EAV data model can be extended to store data from clinical systems and PACS on multiple levels to enable combined querying with a single query to quickly display actual frequency data. ## Introduction Clinical data warehouses (CDW) enable quick queries on homogenized data of a large number of patients and data of a multitude of clinical subsystems as shown by many examples [1–5]. CDWs can be used for a variety of reasons, including rapid feasibility testing or long-term data processing support for individual studies [6, 7]. Medical imaging data, unlike the types of data commonly documented in CDWs (i.e., numeric, categorical, and textual data from various clinical subsystems), are less frequently integrated into CDWs and tend to be more segregated from them [8, 9]. Medical imaging data has distinct features that render its use in CDWs more challenging. In particular, its pixel-based information results in much larger data sizes, which impedes its simple pseudonymized duplication into a CDW and hence its immanent usability. The increasingly better-defined analytic strategies that are supported by deep learning and artificial intelligence render the combination of clinical and imaging data a highly attractive research and development area [9–11]. But not only can the search for information in the pixel data itself be useful for this, but also the enhanced search in DICOM metadata [12, 13]. We previously showed the overall feasibility of a comprehensive integration architecture of a production PACS (including identified data) to a research CDW (including pseudonymized data) using ad hoc pseudonymization [14]. The most comprehensive PACS-CDW integration of the related work so far has been shown for i2b2, which primarily requires selecting a patient population in the CDW using clinical data that then becomes the basis for a PACS query in another downstream module [15]. Other CDW systems often only connect to a dedicated research PACS including image data of patient subgroups [16–22], a CDW dedicated to a specific disease [23] or are rather specialized for imaging analysis and less on clinical data [12, 24–28]. CDWs are often based on an entity–attribute–value (EAV) data schema (e.g., i2b2 [4]), a single data model into which data from various source systems (e.g., clinical subsystems) and their data models (e.g., the data models of specific structured forms) must be integrated. A common method used to provide data in EAV-based CDWs is an early aggregation of single or multiple values of different data models into individually usable variables. Such variables can typically be searched on the level of a patient or encounter, e.g., search for encounters with a laboratory NT-proBNP value > 1000 pg/ml and an ICD1 code = “I50” to detect patient encounters with heart failure. However, some queries require a more detailed integration of subsets of the original data model to allow comparing parameters at the level of a document or more detailed groupings, e.g., search for patient encounters with elevated troponin T and elevated creatine kinase-MB (CK-MB) levels in the same laboratory report to detect patients with an ST elevation myocardial infarction (STEMI). Such a search requires integration of individual data elements linked not only via the patient and encounter identifier, but also via a document identifier (i.e., linkage of laboratory values to the laboratory report). The integration of a structured form containing further groupings (e.g., a table) would require an even more detailed linkage. A detailed example is shown in the online supplement. Radiology datasets often consist of narrative radiology reports and associated imaging data that is structured in the multilevel DICOM data model. A detailed search for combinations of clinical and imaging data could also benefit from data integration at multiple grouping levels, i.e., to search for specific DICOM series instead of just discovering whether or not a patient or case has assigned images. However, we have not found another CDW that offers the selection and search on the document and further detailed levels for imaging metadata. Thus, querying and extracting related data from PACS and other clinical subsystems for research on large numbers of patients is impeded by limited integration into CDWs and manual downstream processing steps. ## Objectives The objective of this study is to show the feasibility and benefit of single queries in a research data warehouse combining data from a hospital’s clinical and imaging systems. ## Evaluation Methodology We first contrast the data models commonly used in the clinical and imaging domain. Then, we describe the CDW used, its abstract data model, and the PACS middleware we used for the DICOM data integration. This basis is used to describe the method we used to integrate the DICOM data model into the CDW’s abstract data model, by preserving the linkages of the DICOM data model as far as possible. Since the existing generic CDW query interface was only able to query variables on either of the patient, encounter, or document level, we extended the PACS-CDW-middleware with a simple graphical query interface optimized for the combined query of imaging and clinical data. Finally, we demonstrate the viability of a combined query using data entries from clinical and imaging systems using multiple grouping levels. Therefore, we selected the following two real inquiries from researchers, both of which mandate combinations of clinical and imaging data but different query approaches:Radiology-oriented: Retrieve combinations of a radiology report and associated DICOM series of cardiac magnetic resonance imaging (MRI).Disease-oriented: Detect patients with heart failure and retrieve combinations of the radiology report and associated MRI DICOM series Whereas, the first inquiry has its focus rather on identifying DICOM series with similar image acquisition characteristics in combination with a radiology report, the second has its focus on retrieving DICOM series and clinical data for a specific disease. Both inquiries had a similar goal of providing potentially large amounts of data to researchers who planned a data analysis using deep learning approaches. However, in this work, we only focus on the optimized data retrieval in contrast to an overall retrieval and analysis pipeline. In addition, we chose very simple queries to show the benefit of a multilevel query, as opposed to more complex queries, as e.g., including DICOM tags that are documented only by special request from the manufacturer of a research MRI. A single query was defined for each inquiry and executed using the new query interface. Resulting SQL queries were analyzed using the CDW’s relational database. Extracted query results were characterized using descriptive statistics. To show the benefit of a multilevel query, the query results were compared with the results of queries performed only on individual grouping levels (e.g., document or encounter level) in order to simulate systems that do not support multiple grouping levels. We then discuss these results with existing and potentially future integration into a very similar EAV-based CDW and a refined relational data model. ## Medical Data From a data homogenization perspective when using a CDW, the data models of most clinical subsystems are fairly similar, but differ regarding the data model employed in PACS. ## Clinical Data Model Hospital information systems (HIS) may consist of many subsystems, each with own specific data models. They may use open standards (e.g., OpenEHR,2 FHIR,3 IHE4) for storage and exchange, or proprietary data models. Each item is connected to a patient with a unique patient identifier and an acquisition time. Commonly, data is also mapped to distinct patient hospitalizations using encounter identifier. Multiple encounters might be further linked using an episode of care (e.g., to link data for the treatment of a specific disease). The data is often stored as document (e.g., a structured form) or in various sub-data models (e.g., OpenEHR archetypes or FHIR resources), which may be linked to a document identifier. A document may further contain data values that are linked on multiple levels, e.g., all values of a table or a table row. Overall, this data compilation is very heterogenous. A simplified data hierarchy is illustrated in Fig. 1A.Fig. 1Example of a clinical and DICOM (PACS) data model. Only the links to the next level in the hierarchy are shown. Elements with a gray background structure the data at different levels. ( MRI, magnetic resonance imaging) ## DICOM Data Model Images are often stored using DICOM in a PACS. The DICOM data model is hierarchically structured and based on a real radiological examination: a single image (an instance), a series of images (e.g., images of a single MRI sequence), and a study containing all series required for the examination of a specific diagnosis. Often, a DICOM study is linked to a radiology report in a radiology information system (RIS) via the accession number (e.g., an order identifier). Furthermore, a patient identifier can be stored. An illustration of the DICOM data model and its linkage to clinical data is shown in Fig. 1B. ## Clinical Data Warehouse The CDW contains homogenized and pseudonymized data of a large part of the hospital’s data [29]. Its query system enables constraints on structured data and free text searches in narrative texts (e.g., discharge letters and radiology reports) using regular expressions [30]. Figure 2 illustrates the CDW’s abstract data model in the leftmost column. It is based on an extended EAV model within a relational data base and is similar yet not identical to the star schema of i2b2 [4, 31]. It is derived from the clinical data model described in Fig. 1A and provides the four grouping levels “patient,” “encounter,” “document,” and “group”. Depending on the source data model, values in the CDW are always connected at patient level, almost always at encounter and document level, and sometimes also at group level. Each value is further stored with a concept identifier which links to the value’s metadata. Fig. 2Illustration of the abstract CDW data model (left) and the method we used to map and store parameters of the radiology report (center) and DICOM instances (right). Only connections to the next higher level are shown in the data model. A value can be connected to any level or only to one patient The main component of the DICOM integration used in this work is a separate middleware (PACS-to-CDW; P2D) that allows pseudonymous querying of a production PACS whose overall architecture was described previously [6]. The middleware essentially accepts the pseudonymous identifiers used in the CDW as input via a REST-style interface, which it maps to the identifiers used in the PACS via an identity management system to perform the most basic DICOM queries C-FIND and C-MOVE (which are most likely available in any production PACS). ## DICOM-CDW Integration Within the CDW, we linked the DICOM imaging data with the clinical data via the radiology report, which are also linked in the clinical domain via the accession number. The report’s data elements were stored with a patient, encounter, and document identifier. Since we need a direct linkage between a report and the DICOM data within the CDW, we adopted these three identifiers to the DICOM data via the accession number. DICOM data originally has a patient identifier, but no encounter and document identifier. Thus, we only have a single grouping level left in the CDW’s data model, which we used to encode the DICOM series via the Series Instance UID. In our experience, this grouping level is of greater interest for research than the single image or the study (potentially containing several modalities and specific examinations). Some DICOM images may not be associated with an accession number nor a radiology report in the local hospital, e.g., if transmitted from external providers. Such DICOM data was imported with a patient identifier, an empty encounter identifier, and a document identifier that was based on the Study Instance Unique Identifiers (UID). Each value of the DICOM header was stored as separate row within the EAV schema. ## Multilevel Query We developed a prototypic query system that provides functionality to transform a query defined in a self-developed query language into SQL/Solr queries. The query language allows to constrain the patient selection by filtering the admission date. Most importantly, it allows conditioning of CDW concepts using the concept identifier and various operators (e.g., text contains x, number greater than x) using the operators ‘AND’ and ‘OR’, specifically for each of the available grouping levels. Running a query results in a multilevel collection of data, starting with the filter and an SQL query on the most comprehensive level (i.e., patient) and ending with the most restrictive level (i.e., group). The data selection of an upper group is used as input for the next level. Levels that are not defined are simply skipped, e.g., when querying for a radiology report without a patient/encounter condition, all existing documents of all patients and encounters are selected. Depending on the need, the query results of each level (which may be particularly high at the patient level) or only those of the most restrictive level can be extracted. After entering a query, the system extracts descriptive statistics and a data export of all related data from the CDW by pressing a single button. With a second button press, it extracts all DICOM images of the DICOM series indicated in the query results from the PACS. This process is illustrated in Fig. 3.Fig. 3Illustration of the query process: A clinical researcher and data scientist jointly define a query. After execution (2.) the query results are retrieved from the CDW data base. If the query is defined final, the images can be downloaded on another button press (3.) ## Radiology-Oriented Query For inquiry #1, we required to search for radiology reports without conditions on patients or encounters. Thus, resulting data sets may contain data from any patient with any disease. We started the query definition on the document level, by selecting a combination of multiple radiology order identifiers defined by a regular expression (each defines a specific radiological examination). A free text search in the DICOM study description may extend the number of resulting documents. On the group level, this data selection is refined to selectively include images acquired by the modality MRI. Figure 4 illustrates the resulting radiology-oriented query. Since multiple reports may exist per encounter, the report and DICOM parameters needed to be joined on the document level (Fig. 4: 1–3). Of this subgroup, only DICOM series acquired with MRI needed to be queried. Thus, the document level query (Fig. 4: 1–3) needed to be refined to the Series Instance UIDs, joined over the group (Fig. 4: 3–4). Table 1 presents the descriptive statistics provided by the query interface after performing the query illustrated in Fig. 4. Each row shows metrics of the data subset on the specific grouping level. The query results in 28,651 DICOM series with a mean [SD] 17.6 [6.6] per study. The query includes images from different MRI sequences (that could have been further constrained with additional parameters, e.g., series description, which was not requested by the researchers). Each study was connected to a radiology report. Table 1Descriptive statistics produced by the query interface described in Fig. 4. The rightmost column presents the number of attributes or unique texts of the query attributes associated with the levelLevel#Patients#Encounter#Documents#Groups#Values#Values per attributeDocument20392185223349,80758,738Unique:Study description = 13Radiology order = 15Radiology short finding = 2175Radiology finding = 1833Radiology assessment = 2109Group15041613163128,65157,302Unique:Modality = 1Study instance UID = 1632Series instance UID = 28,651 This query was joined on two levels in order to arrive at the final result. Assuming we were to join all the data elements of the query at a single level, the query would result in 41,421 (+ $45\%$) DICOM series when using the document level and 60,701 (+ $112\%$) DICOM series when using the encounter level. Joining at the document level would lead to the additional selection of the modality PR (presentation state, > $99\%$); in case of the encounter level, there were 16 additional modalities, mainly PR ($43\%$), XA (X-ray angiography, $25\%$), CT (computed tomography, $10\%$), MR (magnetic resonance, $10\%$), and US (ultrasound, $4\%$). ## Disease-Oriented Query For inquiry #2, we required to primarily search for patients with the specific condition of having a heart failure. We started the query definition on the encounter level by selecting parameters indicative of the disease, following an evaluated heart failure detection algorithm [32]. This algorithm required the search of certain occurrences of free text in discharge letters, conditions on numerical values from echocardiography reports, and the selection of ICD codes. On the document level, we only restrict the data selection to patients with an existing radiology report. Conditions on the group level are similar to the radiology-oriented query. The added complexity is that not only data elements of single documents or a more detailed structure have to be compared to each other, but also data elements of multiple documents of an encounter or a patient. Figure 5 illustrates a disease-oriented query with the conditions from a specific research project. Most query parameters (clinical or DICOM) originated from different processes, therefore had separate document identifiers, and could only be linked at the patient or encounter level. The clinical values had to be joined on the encounter level, since we are only looking for patients with heart failure and associated images (Fig. 5: 1–4). In contrast, a linkage at the patient level could also reveal images that predate the development of the disease. Fig. 4Illustration of the grouping levels we needed to extract data for the first inquiryFig. 5Illustration of the grouping levels we needed to extract data for the second inquiry Table 2 presents the descriptive statistics produced by the query interface illustrated in Fig. 5. The query resulted in 12,834 DICOM series with a mean [SD] 1,4 [0,5] per study. Table 2Descriptive statistics provided by the query interface after performing the query described in Fig. 5. The rightmost column presents the number of attributes or unique texts of the query attributes associated with the level. ( LVEF = left ventricular ejection fraction)Level#Patients#Encounter#Documents#Groups#Values#Values per attributeEncounter818513,23414,2236838,512Unique:Physician letter = 5125Count:LVEF = 7759I11 = 4760I13 = 71I50 = 11,638Document5181665014,670158,692Unique:Radiology order = 407Radiology short finding = 14,245Radiology finding = 10,443Radiology assessment = 13,805Group41105152949912,83525,670Unique:Modality = 2Study instance UID = 9529Series instance UID = 12,834 This query was queried on three grouping levels in order to arrive at correct results, because joining on solely one level would not result in meaningful data since multiple types of documents are required on the encounter level. By contrast, joining data on the most detailed level available at every data entry, i.e., the encounter, such query would result in the same number of encounters and DICOM studies, but would contain 86,346 DICOM series (+ $573\%$). A query on the encounter level would result in 18 additional modalities, mainly in XA ($38\%$), CT ($30\%$), MR ($12\%$), US ($7\%$), and PR ($4\%$). ## Results For the two queries, we selected all patient encounters stored in the CDW from the Department of Internal Medicine of the years 2016 and 2017 (“baseline data set”). This data set included 32,153 patients, 64,018 encounters (mean [standard deviation (SD)] of 2546 [629] per month), 31,205 radiology reports (1223 [305] per month), 52,882 DICOM studies (2090 [476] per month), and about 292,369 DICOM series (11,574 [2765] per month). A DICOM study had a mean 5.0 [7.6] series, most often with a single modality ($83\%$), followed by 2 ($16\%$) and 3 ($2\%$) modalities. ## Discussion We showed the feasibility and benefit of querying combinations of data from medical imaging and clinical subsystems in a CDW based on an EAV data schema. The proposed method integrates DICOM metadata directly into the CDW and allows to reliably obtain quantitative information on the current availability of data with reduced manual steps using a single query (e.g., for feasibility studies). Querying the details of DICOM series in addition to clinical data may be particularly important when imaging data from a large number of patients is required, e.g., for deep learning approaches including both imaging and clinical data. To our knowledge, such a data integration and query capability with constraints at multiple grouping levels in a single query has not been demonstrated before. In order to provide such a query capability, we advanced an approach that has been started by existing CDWs. CDWs mainly provide the two grouping levels patient and encounter, to which we have added the levels document and group. This allows to integrate data values that may be linked on four levels, which can be used to include small subsets of the original data models, e.g., the DICOM information model. Since the data models of the different data domains differ, the meaning of the grouping layers also differs. We demonstrated the feasibility of a combined query that we constrained on all four levels for a single sub-model, i.e., DICOM. The execution of the query leads to the provision of descriptive statistics, partitioned into the grouping levels used in the query, e.g., count of patient encounter, radiology reports (document), and DICOM series (group). Another usage is conceivable that requires such constraining on all levels for multiple data domains, e.g., to identify patients with a myocardial infarction based on two variables of a single blood sample (sub-model 1) and specific DICOM series (sub-model 2). We currently enable the mapping of source data models with up to four levels to the EAV data model. Especially the document and group level could be mapped to any association of data elements that is of interest. However, we do not yet provide a generic solution to allow for more than four levels. We neither provide a fully standardized method to annotate the meaning of the group levels in the CDW with its usage in the data model (e.g., in data domain “imaging,” a group refers to the linkage using the Series Instance UID). Furthermore, we developed an imaging-optimized prototype that can be applied for similarly framed used cases, but is not as easy to use as our standard CDW query interface. We illustrated, how we were able to improve the accuracy of the presented data using multiple grouping levels compared to queries that only combine data elements on a single level. Accordingly, merging all data elements of the radiology-oriented query at the document level would result in a near doubling of the number of identified DICOM series. Since these were almost exclusively of type PR (presentation state), this would not have significantly affected the image download from the PACS, but might have distorted the display of the correct number of series. In any case, the impact of using multiple grouping levels on the numbers presented at the encounter level would be large. We integrated data into the CDW as provided by the DICOM data model and then constrained it while querying. An alternative to the integration of granular data to the CDW is to them as pre-aggregated data. However, such strategy would impede a quick usage for new usage scenarios. Furthermore, since duplication of a complete PACS is not practical, we would have to access the production PACS each time a new variable of interest is being defined. We showed queries on DICOM parameters that were retrieved via DICOM C-FIND. While an extraction of all DICOM parameters using C-MOVE would be possible, it would require the transfer of at least one image from each DICOM series. Since this might result in a workload that impacts on the performance of the clinical systems or demands much more time, we only did this when required for subsets of patients yet. A quicker access to all DICOM metadata would be the usage of a PACS that provides web-based parts of the DICOM standard, e.g., QIDO and WADO.5 However, not every PACS supports these nor might a hospital department easily acquire such an extension for research purpose. Even then, metadata extraction from a pre-filled CDW would still be faster for large queries. The direct access to the internal data structure of a PACS would provide the fastest solution to access DICOM files of a PACS [28], but could provide a major vulnerability for a production PACS. The solution presented here is most interesting for CDWs that are based on an EAV data schema. The CDW that is probably most often used is i2b2 [15], which provides an EAV schema that allows storing values with a patient, an encounter, and a modifier. The latter could be used similarly to one of the levels “document” or “group,” i.e., to providing not only a single value per concept, but also its complementary attributes. In practice, however, the i2b2 query user interface usually queries on the patient or encounter level.6 An alternative would be the usage of common data models (CDM) that separate data domains into specific database tables. The OMOP-CDM7 for example does not provide a single table for all values, but provides a multitude of tables each with a special purpose, and consists of combinations of relational and EAV data schemas. The fastest way to include DICOM data might be the usage of an existing table (e.g., “measurement” or “observation,” both with an EAV schema). However, the OMOP-CDM only links a value to a patient or an encounter. Importantly, a document or group in terms of a collection of parameters measured from a single specimen at a specific time is not provided.8 Thus, it might be more straightforward to extend the CDM with a specific table for DICOM data. If we were to follow this reasoning from an a single EAV table to domain-specific tables further, we could also use the data models of HL7 FHIR or OpenEHR itself, as suggested by Paff et al. [ 33]. This would allow to document even more granular data including links between data values. Multiple values of a laboratory panel for example would be documented within the FHIR resource ‘DiagnosticReport’. However, this approach has the disadvantage of many different sub-data models thus requiring multiple small queries. With “Dr. Warehouse,” Garcelon et al. [ 34] developed a CDW that centers on documents in their database schema. This allows to store documents as text, in combination to extracted parameters. The CDW then can be searched in relation to the document, the encounter or patient level also providing a free text search option to query all documents. Despite this extensive search functionality in the graphical user interface, however, it does not seem to provide possibilities to constrain multiple variables of a single document yet. Given a variety of base data models being integrated into CDWs and research data models, and a concomitant increase in data complexity, an ideal solution might be to provide a generic query interface for a general unrestricted CDW query and a domain-specific query interface for specific used cases that guides through a query as described by Horvath et al. [ 35]. We are not aware of a CDW system that integrates both clinical data and data from imaging systems themselves, but rather only data derived from image data, such as radiology reports. Instead, many CDWs offer an image store for manual image upload or semi-manual PACS download and are not optimized for access to all PACS data, as described in the introduction. The CDW system closest to the one we propose is the i2b2 PACS integration, which lacks metadata integration and offers multistage querying of a production PACS [15]. They provided a special module to i2b2 that allows for various query constraints based on a list of patients, e.g., obtained from the standard i2b2 query interface. The query is directly sent to the DICOM interface of the connected PACS. Consequently, their approach mandates multiple query steps, which heavily slows speed of result generation when querying a larger group of patients. As opposed to a single CDW system containing clinical and imaging data, a viable path could also be to virtually integrate separate imaging-oriented warehouse systems (e.g., Dicoogle) and clinical systems (i.e., i2b2), e.g., using a federated query system. Dicoogle is a comprehensive PACS developed for research purposes that enables distributed queries across multiple PACS sites. It creates indices of metadata for a fast query capability and allows for many plugins and applications on image-oriented studies [28, 36]. It is not directly used with a CDW and its query is image-centric [28]. We added metadata from DICOM files to the CDW. This is relatively simple, but can support many used cases [14]. A major goal remaining is to also integrate variables of features extracted from the imaging raw data, i.e., the pixel data. Several content-based image retrieval solutions in the context of dedicated image-oriented CDW and analysis systems have been developed [37–42]. However, such systems have not been comprehensively integrated into a productive hospital CDW, which remains an open task. ## Conclusion An EAV data model can be extended to store data from clinical systems and PACS on multiple levels to enable combined querying with a single query to quickly display actual frequency data. ## References 1. Dorda W, Wrba T, Duftschmid G, Sachs P, Gall W, Rehnelt C, Boldt G, Premauer W. **ArchiMed: a medical information and retrieval system**. *Methods Inf Med* (1999.0) **38** 16-24. DOI: 10.1055/s-0038-1634148 2. Brammen D, Katzer C, Röhrig R, Weismüller K, Maier M, Hossain H, Menges T, Hempelmann G, Chakraborty T. **An integrated data-warehouse-concept for clinical and biological information**. *Stud Health Technol Inform* (2005.0) **116** 9-14. PMID: 16160228 3. 3.Lowe HJ, Ferris TA, Hernandez PM, Weber SC. STRIDE–An integrated standards-based translational research informatics platform. AMIA Symposium 2009; 391–395. 4. Murphy SN, Weber G, Mendis M, Gainer V, Chueh HC, Churchill S, Kohane I. **Serving the enterprise and beyond with informatics for integrating biology and the bedside (i2b2)**. *J Am Med Inform Assoc* (2010.0) **17** 124-130. DOI: 10.1136/jamia.2009.000893 5. Virostko J, Hilmes M, Eitel K, Moore DJ, Powers AC. **Secondary use of clinical data: the Vanderbilt approach**. *J Biomed Inform* (2014.0) **52** 28-35. DOI: 10.1016/j.jbi.2014.02.003 6. 6.Kaspar M, Ertl M, Fette G, Dietrich G, Toepfer M, Angermann C, Störk S, Puppe F. Data Linkage from Clinical to Study Databases via an R Data Warehouse User Interface. Experiences from a Large Clinical Follow-up Study. Methods Inf Med 2016; 55:381–386. 7. 7.Prokosch HU, Ganslandt T: Perspectives for medical informatics. Reusing the electronic medical record for clinical research. Methods Inf Med 2009; 48:38–44. 8. Shi Z, Zhovannik I, Traverso A, Dankers FJ, Deist TM, Kalendralis P, Monshouwer R, Bussink J, Fijten R, Aerts HJ. **Distributed radiomics as a signature validation study using the Personal Health Train infrastructure**. *Sci. Data* (2019.0) **6** 1-8. DOI: 10.1038/s41597-019-0241-0 9. Morris MA, Saboury B, Burkett B, Gao J, Siegel EL. **Reinventing Radiology: Big Data and the Future of Medical Imaging**. *J Thorac Imaging* (2018.0) **33** 4-16. DOI: 10.1097/RTI.0000000000000311 10. Vial A, Stirling D, Field M, Ros M, Ritz C, Carolan M, Holloway L, Miller AA. **The role of deep learning and radiomic feature extraction in cancer-specific predictive modelling: a review**. *Transl Cancer Res* (2018.0) **7** 803-816. DOI: 10.21037/tcr.2018.05.02 11. Kansagra AP, Yu J-PJ, Chatterjee AR, Lenchik L, Chow DS, Prater AB, Yeh J, Doshi AM, Hawkins CM, Heilbrun ME, Smith SE, Oselkin M, Gupta P, Ali S. **Big Data and the Future of Radiology Informatics**. *Acad Radiol* (2016.0) **23** 30-42. DOI: 10.1016/j.acra.2015.10.004 12. Langer SG. **DICOM Data Warehouse: Part 2**. *J Digit Imaging* (2016.0) **29** 309-313. DOI: 10.1007/s10278-015-9830-4 13. Valente F, Costa C, Silva A. **Dicoogle, a PACS featuring profiled content based image retrieval**. *PloS one* (2013.0) **8** e61888. DOI: 10.1371/journal.pone.0061888 14. 14.Kaspar M, Liman L, Ertl M, Fette G, Seidlmayer L K, Schreiber L, Puppe F, Störk S. Unlocking the PACS DICOM Domain for its Use in Clinical Research Data Warehouses. J Digit Imaging 2020; 33:1016–1025. 15. Murphy SN, Herrick C, Wang Y, Wang TD, Sack D, Andriole KP, Wei J, Reynolds N, Plesniak W, Rosen BR, Pieper S, Gollub RL. **High throughput tools to access images from clinical archives for research**. *J Digit Imaging* (2015.0) **28** 194-204. DOI: 10.1007/s10278-014-9733-9 16. Marcus DS, Olsen TR, Ramaratnam M, Buckner RL. **The Extensible Neuroimaging Archive Toolkit: an informatics platform for managing, exploring, and sharing neuroimaging data**. *Neuroinformatics* (2007.0) **5** 11-34. DOI: 10.1385/NI:5:1:11 17. Warnock MJ, Toland C, Evans D, Wallace B, Nagy P. **Benefits of using the DCM4CHE DICOM archive**. *J Digit Imaging* (2007.0) **20** 125-129. DOI: 10.1007/s10278-007-9064-1 18. Hu H, Correll M, Kvecher L, Osmond M, Clark J, Bekhash A, Schwab G, Gao D, Gao J, Kubatin V, Shriver CD, Hooke JA, Maxwell LG, Kovatich AJ, Sheldon JG, Liebman MN, Mural RJ. **DW4TR: A Data Warehouse for Translational Research**. *J Biomed Inform* (2011.0) **44** 1004-1019. DOI: 10.1016/j.jbi.2011.08.003 19. Marés J, Shamardin L, Weiler G, Anguita A, Sfakianakis S, Neri E, Zasada SJ, Graf N, Coveney PV. **p-medicine: A Medical Informatics Platform for Integrated Large Scale Heterogeneous Patient Data**. *AMIA Symposium* (2014.0) **2014** 872-881. PMID: 25954394 20. Skripcak T, Just U, Simon M, Büttner D, Lühr A, Baumann M, Krause M. **Toward Distributed Conduction of Large-Scale Studies in Radiation Therapy and Oncology: Open-Source System Integration Approach**. *IEEE J Biomed Health Inform* (2016.0) **20** 1397-1403. DOI: 10.1109/JBHI.2015.2450833 21. Foran DJ, Chen W, Chu H, Sadimin E, Loh D, Riedlinger G, Goodell LA, Ganesan S, Hirshfield K, Rodriguez L, DiPaola RS. **Roadmap to a comprehensive clinical data warehouse for precision medicine applications in oncology**. *Cancer Inform* (2017.0) **16** 1176935117694349. DOI: 10.1177/1176935117694349 22. MohammadianForoushani H, Dhar R, Chen Y, Gurney J, Hamzehloo A, Lee J-M, Marcus DS. **The Stroke Neuro-Imaging Phenotype Repository: An Open Data Science Platform for Stroke Research**. *Front Neuroinform* (2021.0) **15** 597708. DOI: 10.3389/fninf.2021.597708 23. 23.Kortüm KU, Müller M, Kern C, Babenko A, Mayer W J, Kampik A, Kreutzer T C, Priglinger S, Hirneiss C. Using electronic health records to build an ophthalmological data warehouse and visualize patients’ data. Am J Ophthalmol 2017. 24. Wong STC, Hoo KS, Knowlton RC, Laxer KD, Cao X, Hawkins RA, Dillon WP, Arenson RL. **Design and applications of a multimodality image data warehouse framework**. *J Am Med Inform Assoc* (2002.0) **9** 239-254. DOI: 10.1197/jamia.M0988 25. Minati L, Ghielmetti F, Ciobanu V, D’Incerti L, Maccagnano C, Bizzi A, Bruzzone MG. **Bio-image warehouse system: concept and implementation of a diagnosis-based data warehouse for advanced imaging modalities in neuroradiology**. *J Digit Imaging* (2007.0) **20** 32-41. DOI: 10.1007/s10278-006-0859-2 26. Rajala T, Savio S, Penttinen J, Dastidar P, Kähönen M, Eskola H, Miettunen R, Turjanmaa V, Järvenpää R. **Development of a research dedicated archival system (TARAS) in a university hospital**. *J Digit Imaging* (2011.0) **24** 864-873. DOI: 10.1007/s10278-010-9350-1 27. 27.Toms A, Kasmai B, Williams S, Wilson P. Building an anonymized catalogued radiology museum in PACS: a feasibility study. Br J Radiol 2014. 28. Pedrosa M, Silva JM, Silva JF, Matos S, Costa C. **SCREEN-DR: Collaborative platform for diabetic retinopathy**. *Int J Med Inform* (2018.0) **120** 137-146. DOI: 10.1016/j.ijmedinf.2018.10.005 29. 29.Fette G, Ertl M, Wörner A, Kluegl P, Störk S, Puppe F. Information Extraction from Unstructured Electronic Health Records and Integration into a Data Warehouse. Lecture Notes in Informatics 2012;1238–52. 30. 30.Dietrich G, Krebs J, Fette G, Ertl M, Kaspar M, Störk S, Puppe F. Ad Hoc Information Extraction for Clinical Data Warehouses. Methods Inf Med 2018; 57:e22–e29. 31. 31.Georg F, Mathias K, Georg D, Max E, Jonathan K, Stefan S, Frank P. A customizable importer for the clinical data warehouses PaDaWaN and I2B2. Stud Health Technol Inform 2017. 32. 32.Kaspar M, Fette G, Güder G, Seidlmayer L, Ertl M, Dietrich G, Greger H, Puppe F, Störk S. Underestimated prevalence of heart failure in hospital inpatients: a comparison of ICD codes and discharge letter information. Clin Res Cardiol. 2018; 107(9):778-787. 33. 33.Pfaff ER. Cohort Identification Using Semantic Web Technologies: Ontologies and Triplestores as Engines for Complex Computable Phenotyping, Dissertation 2020. 34. Garcelon N, Neuraz A, Salomon R, Faour H, Benoit V, Delapalme A, Munnich A, Burgun A, Rance B. **A clinician friendly data warehouse oriented toward narrative reports: Dr**. *Warehouse. J Biomed Inform* (2018.0) **80** 52-63. DOI: 10.1016/j.jbi.2018.02.019 35. Horvath MM, Winfield S, Evans S, Slopek S, Shang H, Ferranti J. **The DEDUCE Guided Query tool: providing simplified access to clinical data for research and quality improvement**. *J Biomed Inform* (2011.0) **44** 266-276. DOI: 10.1016/j.jbi.2010.11.008 36. Viana-Ferreira C, Ribeiro LS, Costa C. **A framework for integration of heterogeneous medical imaging networks**. *Open Med Inform J* (2014.0) **8** 20-32. DOI: 10.2174/1874431101408010020 37. 37.Lehmann TM, Wein BB, Greenspan H. Integration of content-based image retrieval to picture archiving and communication systems, Medical Informatics Europe Conference 2003. 38. 38.Traina A, Rosa NA, Traina C. Integrating images to patient electronic medical records through content-based retrieval techniques, Computer-Based Medical Systems, 16th IEEE Symposium 2003;163–168. 39. 39.Montagnat J, Duque H, Pierson J-M, Breton V, Brunie L, Magnin I. Medical image content-based queries using the grid. HealthGrid 2003;138–147. 40. Hsu W, Antani S, Long LR, Neve L, Thoma GR. **SPIRS: a Web-based image retrieval system for large biomedical databases**. *Int J Med Inform* (2009.0) **78** S13-S24. DOI: 10.1016/j.ijmedinf.2008.09.006 41. Valente F, Silva LAB, Godinho TM, Costa C. **Anatomy of an Extensible Open Source PACS**. *J Digit Imaging* (2016.0) **29** 284-296. DOI: 10.1007/s10278-015-9834-0 42. Istephan S, Siadat M-R. **Unstructured medical image query using big data - An epilepsy case study**. *J Biomed Inform* (2016.0) **59** 218-226. DOI: 10.1016/j.jbi.2015.12.005
--- title: 'Neuroinflammatory Biomarkers for Traumatic Brain Injury Diagnosis and Prognosis: A TRACK-TBI Pilot Study' authors: - John K. Yue - Firas H. Kobeissy - Sonia Jain - Xiaoying Sun - Ryan R.L. Phelps - Frederick K. Korley - Raquel C. Gardner - Adam R. Ferguson - J. Russell Huie - Andrea L.C. Schneider - Zhihui Yang - Haiyan Xu - Cillian E. Lynch - Hansen Deng - Miri Rabinowitz - Mary J. Vassar - Sabrina R. Taylor - Pratik Mukherjee - Esther L. Yuh - Amy J. Markowitz - Ava M. Puccio - David O. Okonkwo - Ramon Diaz-Arrastia - Geoffrey T. Manley - Kevin K.W. Wang journal: Neurotrauma Reports year: 2023 pmcid: PMC10039275 doi: 10.1089/neur.2022.0060 license: CC BY 4.0 --- # Neuroinflammatory Biomarkers for Traumatic Brain Injury Diagnosis and Prognosis: A TRACK-TBI Pilot Study ## Abstract The relationship between systemic inflammation and secondary injury in traumatic brain injury (TBI) is complex. We investigated associations between inflammatory markers and clinical confirmation of TBI diagnosis and prognosis. The prospective TRACK-TBI Pilot (Transforming Research and Clinical Knowledge in Traumatic Brain Injury Pilot) study enrolled TBI patients triaged to head computed tomography (CT) and received blood draw within 24 h of injury. Healthy controls (HCs) and orthopedic controls (OCs) were included. Thirty-one inflammatory markers were analyzed from plasma. Area under the receiver operating characteristic curve (AUC) was used to evaluate discriminatory ability. AUC >0.7 was considered acceptable. Criteria included: TBI diagnosis (vs. OC/HC); moderate/severe vs. mild TBI (Glasgow Coma Scale; GCS); radiographic TBI (CT positive vs. CT negative); 3- and 6-month Glasgow Outcome Scale-Extended (GOSE) dichotomized to death/greater relative disability versus less relative disability (GOSE 1–$\frac{4}{5}$–8); and incomplete versus full recovery (GOSE <8/ = 8). One-hundred sixty TBI subjects, 28 OCs, and 18 HCs were included. Markers discriminating TBI/OC: HMGB-1 (AUC = 0.835), IL-1b (0.795), IL-16 (0.784), IL-7 (0.742), and TARC (0.731). Markers discriminating GCS 3–$\frac{12}{13}$–15: IL-6 (AUC = 0.747), CRP (0.726), IL-15 (0.720), and SAA (0.716). Markers discriminating CT positive/CT negative: SAA (AUC = 0.767), IL-6 (0.757), CRP (0.733), and IL-15 (0.724). At 3 months, IL-15 (AUC = 0.738) and IL-2 (0.705) discriminated GOSE 5–$\frac{8}{1}$–4. At 6 months, IL-15 discriminated GOSE 1–$\frac{4}{5}$–8 (AUC = 0.704) and GOSE <8/ = 8 (0.711); SAA discriminated GOSE 1–$\frac{4}{5}$–8 (0.704). We identified a profile of acute circulating inflammatory proteins with potential relevance for TBI diagnosis, severity differentiation, and prognosis. IL-15 and serum amyloid A are priority markers with acceptable discrimination across multiple diagnostic and outcome categories. Validation in larger prospective cohorts is needed. ClinicalTrials.gov Registration: NCT01565551 ## Introduction Traumatic brain injury (TBI) affects an estimated 2 to 5 million people annually in the United States and 70 million worldwide.1–3 A significant subpopulation suffers persistent deficits, leading to loss of livelihood and societal costs.4–6 Determining the extent of acute injury and long-term prognosis remains challenging because of heterogeneity in patient characteristics, pathoanatomical subtypes, and local or systemic inflammatory responses that drive secondary injury. Objective, reliable, and efficient tools for TBI diagnosis, triage, and prognosis are greatly needed. A major milestone was reached in 2018 when the U.S. Food and Drug Administration cleared two central nervous system (CNS)-specific biomarkers, glial fibrillary acidic protein (GFAP) and ubiquitin C-terminal hydrolase (UCH-L1), to aid in TBI evaluation.7 Literature on biomarker-assisted TBI evaluation, before and after the approval of GFAP and UCH-L1, has focused on brain-enriched molecules, which have good discrimination for TBI severity.8 However, because systemic inflammation can cause secondary brain injury,9 it is also important to identify promising non-CNS-specific biomarkers in TBI diagnosis and prognosis. Systemic biomarkers have potential utility in detecting not only the presence of brain injury, but also its evolution from acute to subacute and chronic phases. Primary TBI triggers reactive astrogliosis, recruitment of local and systemic immune cells to damaged neural tissue, and release of pro- and anti-inflammatory cytokines that mediate cellular repair, secondary injury, and neurodegeneration.10,11 TBI induces and modulates circulating levels of selected cytokines, chemokines, and alarmins that activate secondary injury cascades and cause blood–brain barrier (BBB) breakdown, cytotoxic and vasogenic edema, excessive immune cell infiltration, and neuronal apoptosis.12 Collectively, certain cytokines—small proteins that modulate cell-cell communication and immune reactions (e.g., interleukins [ILs], tumor necrosis factors [TNF]), chemokines—a subclass of cytokines that recruits immune cells toward lesions (e.g., macrophage-associated proteins), and alarmins—damage-associated molecular patterns that trigger and amplify inflammatory cascades (“danger signals”),13,14 constitute key signaling molecules that bridge primary and secondary TBI, with potentially dynamic roles in TBI outcome. One recent example of an alarmin with promise in TBI detection, progression, and outcome is high mobility group box 1 (HMGB-1). HMGB-1 is a ubiquitous nuclear protein released by damaged cells that initiates host defenses in acute tissue/organ damage and has been found to be prognostic of the degree of residual function in injured tissue.15 Circulating HMGB-1 activates liver-derived acute phase reactants, such as serum amyloid A (SAA) and C-reactive protein (CRP), which in turn propagate multiple cytokine and chemokine cascades to amplify systemic and neuroinflammation.16 Activation of specific secondary injury cascades may preferentially affect long-term outcome after TBI, as evidenced by the association observed between higher HMGB-1 and poorer 6-month Glasgow Outcome Scale in pediatric TBI,17 underscoring the potential value of neuroinflammatory markers as therapeutic targets in TBI recovery.18 Indeed, neuroinflammation may explain why some TBI patients develop persistent deficits whereas others progress to good recovery. Recent research has targeted the blockade of TBI-specific cytokines, using receptor antagonists and monoclonal antibodies to dampen overactive inflammatory responses and facilitate neuroprotection after CNS trauma.19,20 Determining the precise cellular interactions among candidate cytokines, chemokines, and alarmins during acute TBI will aid in discovering the inflammatory endophenotypes relevant to TBI diagnosis and outcome, similar to recent successes in traumatic microvascular and neurodegenerative studies.21,22 Identification of promising neuroinflammatory markers is the critical next step for determining therapeutic targets in cellular injury pathways after TBI. Using a multi-marker panel with robust and reliable assays from pre-clinical and clinical data,23–25 we aimed to identify acute inflammatory markers (cytokines, chemokines, and alarmins) suitable for next-phase validation in TBI detection and outcome, in a prospective cohort of acute TBI subjects and controls. ## Study overview and informed consent The prospective, multi-center TRACK-TBI Pilot (Transforming Research and Clinical Knowledge in Traumatic Brain Injury Pilot) study enrolled patients with external force trauma to the head who presented to one of three participating U.S. level 1 trauma centers and received a clinically indicated head computed tomography (CT) scan within 24 h of injury between years 2010 and 2012, as previously described (ClinicalTrials.gov Registration: NCT01565551).26 TRACK-TBI Pilot applied the American College of Emergency Physicians/Centers for Disease Control and Prevention guidelines for obtaining head CTs,27 and data were collected using the National Institutes of Health (NIH) TBI Common Data Elements (CDEs), version 1.28 *Exclusion criteria* were pregnancy, ongoing life-threatening disease (e.g., end-stage malignancy), police custody, involuntary psychiatric hold, and non-English speakers.28 A subset of TRACK-TBI Pilot subjects underwent venous blood draw within 24 h of injury and 3- and 6-month outcomes by structured interview. Eligible subjects were enrolled by convenience sampling at each participating site. Institutional review board (IRB) approval was obtained at each site, and the overall study received approval from the IRB of record at the University of California, San Francisco (UCSF; Protocol No.: 10-00111).28 Informed consent was obtained before enrollment. For subjects unable to provide consent because of the severity of their injury, consent was obtained from their legally authorized representative or surrogate next of kin. Subjects were reconsented, if cognitively able, during their clinical care and/or follow-up time points regarding continuation in study participation.28 ## Study subjects and blood sample processing The current analysis included a subset of TRACK-TBI Pilot subjects who underwent blood draw within 24 h of injury and had unused samples available for analysis. Blood collection and processing in TRACK-TBI Pilot were performed in accordance with the NIH TBI CDEs, as previously described.28,29 Four to 8 mL of whole blood was collected by peripheral venipuncture using dipotassium ethylene diamine tetraacetic acid vacutainer tubes (Becton, Dickinson and Company, Franklin Lakes, New Jersey, U.S.), which are the standard blood collection tubes used for clinical care at our institution. Fresh blood samples were placed on ice for 5 min, then processed by centrifuge at 4000 revolutions per minute for 7 min. Plasma was aliquoted into multiple 250-μL cryovials per patient and stored in −80°C freezers at the UCSF DNA Bank (San Francisco, CA). The process from blood draw to storage at −80°C was completed within 1 h. Plasma samples were stored until they were retrieved for assay analysis; the plasma samples used in the current analysis received one freeze-thaw cycle over their lifetime. In addition, orthopedic injury controls (OCs) and healthy controls (HCs) were recruited by convenience sampling and patient availability. OCs were patients who presented to a participating trauma center within 24 h of acute trauma to their limbs, pelvis, and/or thorax and had an Abbreviated Injury Scale score <4 for those regions. OCs did not have loss or alteration of consciousness, peritraumatic amnesia, or other clinical findings suggestive of TBI and did not undergo a head CT as part of their clinical care. OCs underwent the same informed consent procedure as TBI patients and received a venous blood draw within 24 h of injury. HCs without acute injuries were recruited from the community through an existing relationship with a TRACK-TBI participant or approved public advertisement within TRACK-TBI institutions and received a venous blood draw after informed consent was obtained. HCs were excluded if they had a self-reported history of TBI or polytrauma within 12 months of enrollment. Blood collection and processing for OCs and HCs were identical to TBI patients. ## Plasma biomarker analyses We assembled a multi-marker panel of 31 priority inflammatory markers for investigation. Plasma was extracted from blood samples as previously described.30 All biomarker assays were run in a blinded fashion at the University of Florida Biomarker Laboratory supervised by the senior author K.K.W.W. (Gainesville, FL). Thirty inflammatory markers were analyzed using pre-made Meso Scale Discovery (MSD) V-Plex Panels: Proinflammatory Panel 1 (Catalog #K15049D-1), Cytokine Panel 1 (#K15050D-1), Chemokine Panel 1 (#K15047D-1), and Vascular Injury Panel 2 (#K15198D-1) without using its vascular cell adhesion molecule (VCAM) assay (Meso Scale Diagnostics, LLC, Rockville, MD).23 Though the MSD Vascular Injury Panel 2 and other V-Plex Panels often included vascular and angiogenesis markers, such as VCAM, types of vascular endothelial growth factors, fibroblast growth factor, and others, these were not included in the current analysis because of being out of scope. We report data on the following markers: CRP, eotaxin, eotaxin-3, interferon gamma-induced protein 10 (IP-10), interferon-γ (IFN-γ), intercellular adhesion molecule 1 (ICAM-1), IL-1a, IL-1b, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12/IL-23 p40 protein (IL-12/IL-23p40), IL-12 p70 protein (IL-12p70), IL-13, IL-15, IL-16, IL-17a, macrophage-derived chemokine (MDC), macrophage inflammatory protein 1a (MIP-1a), MIP-1b, monocyte chemoattractant protein-1 (MCP-1), MCP-4, SAA, thymus- and activation-regulated chemokine (TARC), TNF-a, and TNF-b. MSD does not provide an assay for HMGB-1, and we selected the Shino-Test HMGB-1 enzyme-linked immunosorbent assay as a reliable assay because of its wide usage in clinical medicine studies (catalog no.: ST51011; Shino-Test Corporation, Japan, available through Tecan, Incorporated, Morrisville, NC).15,31,32 Biomarkers were run in duplicate according to manufacturing instructions, and the average value of the duplicates was used as the final value for each biomarker. The intra- and interassay coefficients of variation are provided in Supplementary Table S1. The lower limit of detection (LLOD) and dynamic range for each MSD biomarker are available at the MSD website33 and are reported in Supplementary Table S1. The HMGB-1 assay has a dynamic range of 0.31–160 ng/mL and an LLOD of 0.15 ng/mL. Values below LLOD were not used in the final data analysis. Biomarker concentrations are reported in pg/mL, with the exception of HMGB-1, which is reported in ng/mL. ## Statistical analysis Biomarker levels were summarized and compared by diagnostic groups. Comparisons were made using the Wilcoxon rank-sum test because of the skewness of the biomarkers' distribution and relatively small sample sizes. The pair-wise Spearman correlation was calculated and plotted between biomarkers among the TBI cases. Median and first to third quartile (Q1–Q3) were reported for descriptive variables, unless otherwise specified. Receiver operating characteristic (ROC) analyses were conducted to assess the performance of each biomarker in discriminating TBI versus OC, TBI versus HC, GCS 3–12 versus GCS 13–15, and CT positive (CT+) versus CT negative (CT–). ROC analyses were also performed to evaluate the ability of each biomarker to predict 3- and 6-month outcome assessed by the Glasgow Outcome Scale-Extended (GOSE), which consists of an ordinal score from 1 to 8 without units (1 = dead, 2 = vegetative state, 3 = lower severe disability, 4 = upper severe disability, 5 = lower moderate disability, 6 = upper moderate disability, 7 = lower good recovery, and 8 = upper good recovery) and is widely used as the standard measure for functional outcome after TBI.34,35 Outcome was dichotomized in two ways: 1) death/greater relative disability (GOSE 1–4: death or severe disability) versus less relative disability (GOSE 5–8: moderate disability or good recovery) and 2) incomplete recovery (GOSE <8) versus full recovery (GOSE = 8), as shown in earlier studies.36,37 Area under the ROC curve (AUC) was calculated with $95\%$ confidence intervals. *In* general, an AUC of 0.5 suggests no discrimination, 0.7–0.8 is considered acceptable, 0.8–0.9 is considered good, and >0.9 is considered excellent.38 We adopted an AUC threshold of >0.7 to identify candidate markers with acceptable discrimination for TBI diagnosis and prognosis. Because this was an exploratory secondary analysis of existing data, with known limitations in sample size of TBI patients, OCs, and HCs, a priori and post hoc power calculations were not performed. Statistical significance was assessed at $p \leq 0.05.$ Analyses were performed using R version 4.1.2. ## Demographic and clinical data The analytical cohort included 160 subjects with TBI, 28 OCs, and 18 HCs. Mean age was 44.2 years, and $65\%$ (104 of 160) were male. Seventy-nine percent (124 of 160) presented with GCS 13–15, and $49.4\%$ of patients (79 of 160) had intracranial injuries on initial head CT. At 3 months, $80\%$ (128 of 160) of subjects completed the GOSE; median was 7 (Q1–Q3: 5–8), $18\%$ (23 of 128) had death/greater relative disability (GOSE 1–4), and $25.8\%$ (33 of 128) had full recovery (GOSE = 8). At 6 months, $74.3\%$ (119 of 160) of patients completed the GOSE; median was 7 (Q1–Q3: 5–7), $15.1\%$ (18 of 119) had death/greater relative disability, and $24.4\%$ (29 of 119) had full recovery. Full demographic and clinical data are presented in Table 1. **Table 1.** | Variable | % (of N = 160) | | --- | --- | | Age (years) | | | Mean (SD) | 44.2 (18.0) | | Sex | | | Male | 104 (65.0%) | | Female | 56 (35.0%) | | Race | | | White/Caucasian | 131 (81.9%) | | African-American/African | 12 (7.5%) | | Other race | 17 (10.6%) | | Ethnicity | | | Hispanic | 26 (16.6%) | | Non-Hispanic | 131 (83.4%) | | Education | | | Below high school | 18 (11.8%) | | High school graduate | 93 (61.2%) | | College degree or above | 41 (27.0%) | | Employment | | | Full time | 59 (38.6%) | | Part time | 20 (13.1%) | | Unemployed | 38 (24.8%) | | Retired/student/disabled | 36 (23.5%) | | Loss of consciousness | | | No | 42 (26.6%) | | Yes | 101 (63.9%) | | Unknown | 15 (9.5%) | | Post-traumatic amnesia | | | No | 49 (31.0%) | | Yes | 86 (54.4%) | | Unknown | 23 (14.6%) | ## Clinical diagnosis, traumatic brain injury severity, and radiographic diagnosis Acute inflammatory biomarkers with acceptable discriminatory ability (AUC >0.7) for clinical diagnosis of TBI, TBI severity, and radiographic TBI are described below and in detail in Table 2. **Table 2.** | Clinical diagnosis: TBI vs. HC | Clinical diagnosis: TBI vs. HC.1 | Clinical diagnosis: TBI vs. HC.2 | Clinical diagnosis: TBI vs. HC.3 | Clinical diagnosis: TBI vs. HC.4 | | --- | --- | --- | --- | --- | | Biomarker | AUC | TBI | HC | Sig. (p) | | IL-6 | 0.924 [0.880–0.967] | 1.47 [0.55–4.07] pg/mL | 0.15 [0.10–0.22] pg/mL | <0.001 | | IL-10 | 0.863 [0.804–0.922] | 0.17 [0.10–0.39] pg/mL | 0.05 [0.04–0.08] pg/mL | <0.001 | | HMGB-1 | 0.860 [0.802–0.919] | 47.48 [24.35–146.79] ng/mL | 20.77 [14.88–20.77] ng/mL | <0.001 | | IL-4 | 0.819 [0.731–0.907] | 0.09 [0.07–0.15] pg/mL | 0.06 [0.06–0.07] pg/mL | <0.001 | | IL-7 | 0.764 [0.637–0.891] | 0.61 [0.25–1.29] pg/mL | 2.32 [0.90–3.67] pg/mL | <0.001 | | IL-8 | 0.764 [0.666–0.862] | 3.46 [1.53–12.58] pg/mL | 1.29 [0.50–1.64] pg/mL | 0.001 | | TARC | 0.749 [0.626–0.872] | 16.23 [10.49–29.74] pg/mL | 40.63 [22.08–56.31] pg/mL | <0.001 | | IL-5 | 0.748 [0.621–0.874] | 0.37 [0.26–0.49] pg/mL | 0.24 [0.16–0.35] pg/mL | <0.001 | | IL-16 | 0.727 [0.642–0.813] | 146.17 [107.02–309.52] pg/mL | 110.04 [98.74–114.16] pg/mL | 0.002 | Biomarkers with acceptable discrimination between TBI versus HC, with higher values in TBI, included: IL-6 (AUC = 0.924), IL-10 (0.863), HMGB-1 (0.860), IL-4 (0.819), IL-8 (0.764), IL-5 (0.748), and IL-16 (0.727). Biomarkers with acceptable discrimination between TBI versus HC, with lower values in TBI, included: IL-7 (0.764) and TARC (0.749). Biomarkers with acceptable discrimination between TBI versus OC, with higher values in TBI, included: HMGB-1 (AUC = 0.835), IL-1b (0.795), and IL-16 (0.784). Biomarkers with acceptable discrimination between TBI versus OC, with lower values in TBI, included: IL-7 (0.742) and TARC (0.731). Biomarkers with acceptable discrimination between moderate-to-severe versus mild TBI included: IL-6 (AUC = 0.747), CRP (0.726), IL-15 (0.720), and SAA (0.716). Of these, all markers were higher in the moderate-to-severe TBI. Biomarkers with acceptable discrimination for radiographic TBI included: SAA (AUC = 0.767), IL-6 (0.757), CRP (0.733), and IL-15 (0.724). Of these, all markers were higher in CT-positive patients. ## 3- and 6-month prognosis/outcome Inflammatory biomarkers with acceptable discriminatory ability for 3- and 6-month outcome are described below and in Table 3. **Table 3.** | 3-month death/greater relative disability vs. less relative disability (GOSE 1–4 vs. 5–8) | 3-month death/greater relative disability vs. less relative disability (GOSE 1–4 vs. 5–8).1 | 3-month death/greater relative disability vs. less relative disability (GOSE 1–4 vs. 5–8).2 | 3-month death/greater relative disability vs. less relative disability (GOSE 1–4 vs. 5–8).3 | 3-month death/greater relative disability vs. less relative disability (GOSE 1–4 vs. 5–8).4 | | --- | --- | --- | --- | --- | | Biomarker | AUC | GOSE 1–4 (N = 23) | GOSE 5–8 (N = 105) | Sig. (p) | | IL-15 | 0.738 [0.615–0.861] | 1.11 [0.69–1.43] pg/mL | 0.55 [0.39–0.87] pg/mL | <0.001 | | IL-2 | 0.705 [0.587–0.823] | 0.10 [0.08–0.17] pg/mL | 0.08 [0.07–0.10] pg/mL | 0.002 | For 3-month death/greater relative disability (GOSE 1–4) versus less relative disability (GOSE 5–8), biomarkers with acceptable discrimination included: IL-15 (AUC = 0.738) and IL-2 (0.705). Biomarker values were higher in those with death/greater relative disability. No biomarker had discriminatory ability above threshold for 3-month incomplete versus full recovery (GOSE <8 vs. GOSE = 8). For 6-month death/greater relative disability versus less relative disability, biomarkers with acceptable discrimination included: IL-15 (AUC = 0.704) and SAA (0.704). Biomarker values were higher in patients with death/greater relative disability. For 6-month incomplete versus full recovery, the only biomarker with acceptable discrimination was IL-15 (AUC = 0.711), and biomarker values were higher in those with incomplete recovery. Complete data with AUCs for all 31 biomarkers across clinical and diagnostic categories, and 3- and 6-month outcome categories, are provided in Supplementary Table S2. ## Correlations between biomarkers Spearman's correlation matrix was used to evaluate potential collinearity (redundancy) among diagnostic and prognostic markers (Fig. 1). Among markers within the same category of diagnostic or prognostic discrimination (in Tables 2 and 3), several correlations were of moderate strength (0.60–0.79), including IL-15/SAA (0.69), HMGB-1/IL-1b (0.63), HMGB-1/IL-16 (0.62), IL-15/CRP (0.62), and SAA/IL-6 (0.61). The SAA/CRP correlation (0.86) was the only one to exceed moderate strength. **FIG. 1.:** *Correlation matrix for 31 neuroinflammatory biomarkers after acute TBI. Spearman's correlation matrix is shown for the 31 biomarkers included in the current study. Correlation ranges from −1 to +1. Spearman's correlation was considered “moderate” in the 0.6–0.8 range and “strong” if >0.8. CRP, C-reactive protein; HMGB-1, biomarker high mobility group box 1; ICAM-1, intercellular adhesion molecule 1; IFN-γ, interferon-γ; IL, interleukin; IL-12/IL-23p40, IL-12/IL-23 p40 protein; IL-12 p70, IL-12 p70 protein; IP-10, interferon gamma-induced protein 10; MCP, monocyte chemoattractant protein; MDC, macrophage-derived chemokine; MIP-1a, macrophage inflammatory protein 1a; SAA, serum amyloid A; TARC, thymus- and activation-regulated chemokine; TBI, traumatic brain injury; TNF, tumor necrosis factor.* ## Discussion TBI patients show upregulated neuroinflammatory genes and increased expression of cytokines, chemokines, the alarmin HMGB-1, and acute phase reactants (SAA, CRP).39,40 We identified a distinct profile of neuroinflammatory proteins detectable in the systemic circulation within 24 h of acute TBI, with potential utility for objective TBI detection, severity differentiation, and prognosis. Identification of markers able to discriminate both clinical/radiographic TBI severity and better/worse outcome is an important step toward the determination of an inflammatory endophenotype in TBI and potential targets for therapeutic modulation. General TBI diagnostic criteria include external force trauma to the head causing an alteration of consciousness.41 TBI severity has been historically defined as “mild, moderate, or severe” based on GCS and head CT results. Whereas “severe” GCS and greater extent of intracranial injury portend a worse prognosis, their sensitivity for outcome prediction is limited. Objective, quantifiable biomarkers with the ability to determine TBI presence and severity have a wide range of applications, including early detection in pre-hospital settings or where neuroimaging is unavailable, confirmation of injury (e.g., patient with equivocal CT and persistent neurological deficit), and triage to appropriate resources ranging from observation to intensive care unit admission. Though CNS-specific biomarkers such as GFAP and UCH-L1 have been qualified for the evaluation of TBI,42,43 neuroinflammatory biomarkers have the added importance of comprising distinct biochemical and molecular pathways that contribute to secondary injury cascades that cross into subacute and chronic phases, and become a continuum with recovery and outcome. Validation and qualification of robust neuroinflammatory markers can enable the development of a high-yield TBI biomarker panel to serve as primary or adjunct tools for diagnosis. Downstream inflammatory cascades not only contribute to outcome prediction, but may also be promising targets for therapeutic modulation in clinical trials. In our study, few markers showed acceptable concurrent discriminability for both TBI diagnosis and prognosis. One marker was IL-15, which showed acceptable discriminability (AUC >0.7) across TBI severity, radiographic injury, and 3- and 6-month GOSE 1–4 versus 5–8. IL-15 is a proinflammatory cytokine expressed centrally by neuronal and glial cells, peripherally in macrophages and monocytes, and exists in both intracellular and secretory forms.44 Although it has low BBB permeability, peripheral IL-15 activates multiple CNS signaling pathways.45 IL-15 is robustly upregulated in neuroinflammation, induces reactive gliosis, and modulates gamma-amino butyric acid and serotonin transmission, affecting mood, memory, sleep, and activity. These cascades are relevant to acute inflammation and as contributors to persistent cognitive, behavioral, and functional disability. Substantial progress has been made in the IL-15 blockade in cellular and animal models of various neuroinflammatory conditions.46,47 If IL-15 is causally linked to secondary neurological injury in TBI, IL-15 may be a candidate for neuroprotective blockade in human trials. SAA is the second marker with acute and long-term implications (AUC >0.7 for clinical and radiographic TBI severity, as well as 6-month GOSE 1–4 vs. 5–8). As with IL-15, SAA may represent another link between acute injury and long-term inflammatory cascades. SAA is released into the circulation after major injury or infection, induces monocyte and neutrophil migration, and stimulates the production and release of cytokines, chemokines, and matrix metalloproteinases.48,49 These all have broad downstream effects in the activation of transcription factors and epigenetic regulation not only in proinflammatory states, but also for subsequent homeostasis during inflammation.48,49 Murine models have demonstrated that SAA levels correspond to injury severity after controlled cortical impact, with important roles in microglial recruitment and neutrophil infiltration that lead to substantial secondary injury.50 In our data, the concentration of SAA was 43-fold higher in GCS 3–12 versus GCS 13–15, and 30-fold higher in CT+ vs. CT– patients, congruent with its role as an acute phase reactant. SAA has been shown to transiently increase up to 1000-fold during acute injury, although it should return to baseline levels after the insult has resolved.51 In our study, patients with 6-month GOSE 1–4 had a 30-fold acute elevation of SAA compared with GOSE 5–8, underscoring the potential role of SAA in an inflammatory endophenotype connecting persistent inflammation with poor long-term outcome. Recent literature in patients with cerebral microvascular disease has implicated increased SAA and CRP with a cluster of proinflammatory cytokines (IL-6, IL-8, IL-10, and TNF-a) in persistent anxiety.52 SAA and CRP correlated strongly in our data set, but differed in the discriminability of TBI severity. Though the ubiquitous role of SAA in acute phase response makes it a more challenging therapeutic target, there is the potential for research into the neuroprotective blockade of molecules either up- or downstream to SAA in various pathways. In contrast to the small subset of markers predictive of outcomes, the markers associated with primary injury are more diverse. The five diagnostic markers of brain-specific trauma (TBI vs. OC: HMGB-1, IL-1b, IL-7, IL-16, and TARC) did not overlap with markers of TBI severity by GCS or CT criteria (SAA, CRP, IL-6, and IL-15), whereas markers for the latter were identical. This suggests that whereas inflammatory signals are induced at the time of injury, distinct clusters of markers may be induced by different TBI severities and/or injury patterns identifiable by CT. This phenomenon is reassuring, given that it suggests that these cytokine levels are not broadly and indiscriminately altered after TBI, but may be divisible into distinct biomarker profiles that are able to differentiate nuanced clinical correlates. On correlation analysis, analytical “pairs” of inflammatory markers emerged. IL-15 showed moderate correlations with SAA and CRP, implicating its involvement across acute-phase cellular cascades. The alarmin HMGB-1 was associated with IL-1b and IL-16; HMGB-1 increases chemotaxis and activation of leukocytes ex vivo, triggers microglial activation and neuroinflammation, and has been closely associated with detrimental effects of brain injury in traumatic and non-traumatic animal and cellular models.53 TBI-induced microglial activation and increased expression of proinflammatory mediators, such as HMGB-1 and IL-6, have been associated with cerebral edema and neurological deficits.16,54 Our results support the likelihood of HMGB-1 as a marker for brain-specific trauma in humans. The correlations identified in our study underscore the complex crosstalk among markers of neuroinflammation and secondary injury and inform the development of biomarker “panels” for validation in acute and chronic TBI. Finally, our data showed an overlap between markers for brain-specific trauma (TBI vs. OC) with TBI versus HC. Given the multitude and variability of systemic inflammatory pathways activated by trauma, the identification of neuroinflammatory markers with a discriminatory potential for diagnosis and prognosis should focus on brain-specific, rather than generalized, trauma. ## Limitations We recognize several limitations. We performed an exploratory secondary analysis of existing data in a relatively small sample of TBI patients, with fewer numbers of OCs and HCs attributable to limitations in convenience sampling and recruitment. Confirmatory studies with larger numbers of TBI patients and controls encompassing diverse demographics and injury severities are needed, with the additional goal of robustly quantifying differences in biomarker levels between TBIs with and without polytrauma. Changes in biomarker levels as part of non-TBI systemic trauma should also be quantified and accounted for in validation studies. To identify associations for near-term validation and clinical implementation, we dichotomized variables for radiographic injury and functional outcome and used a more stringent cutoff of AUC >0.7 to define “acceptable” discrimination and may have selected out markers with lower AUCs that would have increased with larger sample sizes. Because of the small number of markers above our AUC cutoff, we did not perform multi-variate analyses, which would have provided more definitive yield in larger validation data sets. We were limited by the assays used for this study, which did not include CNS-based biomarkers (e.g. GFAP, UCH-L1). Our study scope focused on acute inflammatory cytokines, chemokines, and alarmins, and we did not include other classes of markers, such as vascular injury and angiogenesis, that may be relevant to TBI injury cascades and outcome21,22 and/or interact with neuroinflammatory cascades. At the time of our study design, some neuroimmune cytokine assays were not yet available at MSD (e.g., IL-31),55 which may warrant inclusion in future studies. We recognize that systemic inflammatory markers may be elevated in non-TBI acute and chronic inflammatory conditions (e.g. the acute stress response, autoimmune disorders, infection, malignancy, and others).56–59 We were limited by the available data from the TRACK-TBI Pilot study, which did not collect comprehensive data on pre-existing inflammatory conditions; it would be important for validation studies to adjust for these important confounders when interpreting inflammatory biomarker values in the context of TBI diagnosis and prognosis. Important next steps include evaluating for more granular associations among cytokine markers, intracranial injury type and location, multi-dimensional outcomes, and changes in their diagnostic/prognostic ability when combined with CNS-specific biomarkers. Evaluation of temporal cascades of inflammatory biomarkers will clarify their relationship with secondary injury and recovery trajectories. Hypothesis-driven studies with appropriate power calculations should be prioritized. Advanced statistical modeling (e.g., dimension reduction) can identify clusters of markers with improved diagnostic or prognostic discriminability and elucidate the underlying “neuroinflammatory endophenotype” that may modulate TBI outcome. These limitations await imminent validation studies utilizing the 18-center prospective TRACK-TBI consortium (https://tracktbi.ucsf.edu/). ## Conclusion We identified a distinct profile of inflammatory proteins detectable in the systemic circulation within 24 h of acute TBI, which may be significant for TBI diagnosis, severity differentiation, and prognosis. The proinflammatory cytokine IL-15 and the acute phase reactant SAA had acceptable discriminatory ability for clinical and radiographic TBI, as well as for outcome after TBI. Future research is needed to validate these findings in a larger cohort and understand how levels of these biomarkers change over time as injury evolves from acute to subacute and chronic phases. This understanding may yield potential targets for therapeutic intervention. ## TRACK-TBI Investigators Neeraj Badjatia, MD (Department of Neurology, University of Maryland, Baltimore, MD); Brandon Foreman, MD (Department of Neurology, University of Cincinnati, Cincinnati, OH); Shankar Gopinath, MD (Department of Neurosurgery, Baylor College of Medicine, Houston, TX); Ramesh Grandhi, MD, MS (Department of Neurosurgery, University of Utah, Salt Lake City, UT); Ruchira M. Jha, MD, MSc (Department of Neurology, Barrow Neurological Institute, Phoenix, AZ); Hester F. Lingsma, PhD (Department of Public Health, Erasmus Medical Center, Rotterdam, The Netherlands); Christopher Madden, MD (Department of Neurosurgery, University of Texas Southwestern Medical Center, Dallas, TX); Debbie Y. Madhok, MD (Department of Emergency Medicine, University of California San Francisco, San Francisco, CA); Michael A. McCrea, PhD (Department of Neurosurgery, Medical College of Wisconsin, Milwaukee, WI); Randall Merchant, PhD (Department of Anatomy and Neurobiology, Virginia Commonwealth University, Richmond, VA); Lindsay D. Nelson, PhD (Department of Neurosurgery, Medical College of Wisconsin, Milwaukee, WI); Laura B. Ngwenya, MD, PhD (Department of Neurosurgery, University of Cincinnati, Cincinnati, OH); Claudia S. Robertson, MD (Department of Neurosurgery, Baylor College of Medicine, Houston, TX); Richard B. Rodgers, MD (Goodman Campbell Brain and Spine, Indianapolis, IN); Gabriela G. Satris, RN, MSN, MSc (Department of Neurosurgery, University of California San Francisco, San Francisco, CA); David M. Schnyer, PhD (Department of Psychology, University of Texas at Austin, Austin, TX); Alex B. Valadka, MD (Department of Neurosurgery, University of Texas Southwestern Medical Center, Dallas, TX); Thomas A. van Essen, MD, PhD (Department of Neurosurgery, Leiden University Medical Center, Leiden, The Netherlands); Ross Zafonte, DO (Department of Rehabilitation Medicine, Harvard Medical School, Boston, MA). ## Authors' Contributions John K. Yue: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing–original draft preparation, writing–review and editing. Firas H. Kobeissy: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing–original draft preparation, writing–review and editing. Sonia Jain: data curation, formal analysis, investigation, methodology, resources, validation, visualization, writing–original draft preparation, writing–review and editing. Xiaoying Sun: data curation, formal analysis, funding acquisition, investigation, methodology, resources, validation, visualization, writing–original draft preparation, writing–review and editing. Ryan R.L. Phelps: formal analysis, investigation, visualization, writing–original draft preparation, writing–review and editing. Frederick K. Korley: formal analysis, investigation, methodology, writing–original draft preparation, writing–review and editing. Raquel C. Gardner: investigation, methodology, writing–original draft preparation, writing–review and editing. Adam R. Ferguson: formal analysis, investigation, methodology, writing–original draft preparation, writing–review and editing. J. Russell Huie: formal analysis, investigation, methodology, writing–original draft preparation, writing–review and editing. Andrea L.C. Schneider: formal analysis, investigation, methodology, writing–original draft preparation, writing–review and editing. Zhihui Yang: investigation, methodology, writing–original draft preparation, writing–review and editing. Haiyan Xu: investigation, methodology, writing–original draft preparation, writing–review and editing. Cillian E. Lynch: investigation, writing–original draft preparation, writing–review and editing. Hansen Deng: investigation, methodology, writing–original draft preparation, writing–review and editing. Miri Rabinowitz: investigation, methodology, writing–original draft preparation, writing–review and editing. Mary J. Vassar: investigation, methodology, writing–original draft preparation, writing–review and editing. Sabrina R. Taylor: formal analysis, investigation, methodology, writing–original draft preparation, writing–review and editing. Pratik Mukherjee: data curation, formal analysis, funding acquisition, investigation, methodology, project administration; writing–original draft preparation, writing–review and editing. Esther L. Yuh: data curation, formal analysis, funding acquisition, investigation, methodology, project administration; writing–original draft preparation, writing–review and editing. Amy J. Markowitz: data curation, formal analysis, funding acquisition, investigation, methodology, project administration; writing–original draft preparation, writing–review and editing. Ava M. Puccio: formal analysis, investigation, methodology, writing–original draft preparation, writing–review and editing. David O. Okonkwo: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing–original draft preparation, writing–review and editing. Ramon Diaz-Arrastia: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing–original draft preparation, writing–review and editing. Geoffrey T. Manley: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing–original draft preparation, writing–review and editing. Kevin K.W. Wang: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, writing–original draft preparation, writing–review and editing. ## Funding Information This work was supported by the following grants: NINDS RC2NS069409-01, RC2NS069409-02S1, RC2NS069409-02, U01NS086090-01, U01NS086090-02S1, U01NS086090-02S2, U01NS086090-03S1, U01NS086090-02, U01NS086090-03, and #U01NS1365885; US DOD #W81XWH-13-1-0441, US DOD #W81XWH-14-2-0176 (to G.T. Manley); NINDS K23NS123340 (to A.L.C. Schneider); US DOD #W81XWH-18-2-0042 (to G.T. Manley and K.K.W. Wang); and a Neurosurgery Research and Education Foundation & Bagan Family Foundation Research Award (UCSF #A139203; to J.K. Yue). ## Author Disclosure Statement Kevin K.W. Wang is a shareholder of Gryphon Bio, Inc. ## Supplementary Material Supplementary Table S1 Supplementary Table S2 ## References 1. Dewan MC, Rattani A, Gupta S. **Estimating the global incidence of traumatic brain injury**. *J Neurosurg* (2018.0). DOI: 10.3171/2017.10.JNS17352 2. Coronado VG, McGuire LC, Sarmiento K. **Trends in traumatic brain injury in the U.S. and the public health response: 1995–2009**. *J Safety Res* (2012.0) **43** 299-307. DOI: 10.1016/j.jsr.2012.08.011 3. Taylor CA, Bell JM, Breiding MJ. **Traumatic Brain Injury-Related Emergency Department Visits, Hospitalizations, and Deaths—United States, 2007 and 2013**. *MMWR Surveill Summ* (2017.0) **66** 1-16. DOI: 10.15585/mmwr.ss6609a1 4. Seabury SA, Goldman DP. **Assessment of follow-up care after emergency department presentation for mild traumatic brain injury and concussion: results from the TRACK-TBI Study**. *JAMA Netw Open* (2018.0) **1** e180210. DOI: 10.1001/jamanetworkopen.2018.0210 5. McCrea MA, Giacino JT, Barber J. **Functional outcomes over the first year after moderate to severe traumatic brain injury in the prospective, longitudinal TRACK-TBI Study**. *JAMA Neurol* (2021.0) **78** 982-992. DOI: 10.1001/jamaneurol.2021.2043 6. Seabury SA, Temkin N. **Employment and economic outcomes of participants with mild traumatic brain injury in the TRACK-TBI Study**. *JAMA Netw Open* (2022.0) **5** e2219444. PMID: 35767257 7. Bazarian JJ, Biberthaler P, Welch RD. **Serum GFAP and UCH-L1 for prediction of absence of intracranial injuries on head CT (ALERT-TBI): a multicentre observational study**. *Lancet Neurol* (2018.0) **17** 782-789. PMID: 30054151 8. Su YS, Schuster JM, Smith DH. **Cost-effectiveness of biomarker screening for traumatic brain injury**. *J Neurotrauma* (2019.0) **36** 2083-2091. DOI: 10.1089/neu.2018.6020 9. Kinoshita K.. **Traumatic brain injury: pathophysiology for neurocritical care**. *J Intensive Care Med* (2016.0) **4** 29. DOI: 10.1186/s40560-016-0138-3 10. Burda JE, Bernstein AM, Sofroniew MV. **Astrocyte roles in traumatic brain injury**. *Exp Neurol* (2016.0) **275** 305-315. PMID: 25828533 11. Corps KN, Roth TL, McGavern DB. **Inflammation and neuroprotection in traumatic brain injury**. *JAMA Neurol* (2015.0) **72** 355-362. PMID: 25599342 12. Jarrahi A, Braun M, Ahluwalia M. **Revisiting traumatic brain injury: from molecular mechanisms to therapeutic interventions**. *Biomedicines* (2020.0) **8** 389. DOI: 10.3390/biomedicines8100389 13. Braun M, Vaibhav K, Saad NM. **White matter damage after traumatic brain injury: A role for damage associated molecular patterns**. *Biochim Biophys Acta Mol Basis Dis* (2017.0) **1863** 2614-2626. PMID: 28533056 14. Morganti-Kossman MC, Lenzlinger PM, Hans V. **Production of cytokines following brain injury: beneficial and deleterious for the damaged tissue**. *Mol Psychiatry* (1997.0) **2** 133-136. PMID: 9106236 15. Andrassy M, Volz HC, Riedle N. **HMGB1 as a predictor of infarct transmurality and functional recovery in patients with myocardial infarction**. *J Intern Med* (2011.0) **270** 245-253. PMID: 21362071 16. Paudel YN, Angelopoulou E, Piperi C. **HMGB1-mediated neuroinflammatory responses in brain injuries: potential mechanisms and therapeutic opportunities**. *Int J Mol Sci* (2020.0) **21** 4609. DOI: 10.3390/ijms21134609 17. Au AK, Aneja RK, Bell MJ. **Cerebrospinal fluid levels of high-mobility group box 1 and cytochrome C predict outcome after pediatric traumatic brain injury**. *J Neurotrauma* (2012.0) **29** 2013-2021. DOI: 10.1089/neu.2011.2171 18. Parker TM, Nguyen AH, Rabang JR. **The danger zone: systematic review of the role of HMGB1 danger signalling in traumatic brain injury**. *Brain Inj* (2017.0) **31** 2-8. PMID: 27819487 19. DeKosky ST, Styren SD, O'Malley ME. **Interleukin-1 receptor antagonist suppresses neurotrophin response in injured rat brain**. *Ann Neurol* (1996.0) **39** 123-127. PMID: 8572657 20. Li S, Eisenstadt R, Kumasaka K. **Does enoxaparin interfere with HMGB1 signaling after TBI? A potential mechanism for reduced cerebral edema and neurologic recovery**. *J Trauma Acute Care Surg* (2016.0) **80** 381-387. PMID: 26670109 21. Sandsmark DK, Bogoslovsky T, Qu BX. **Changes in plasma von Willebrand factor and cellular fibronectin in MRI-defined traumatic microvascular injury**. *Front Neurol* (2019.0) **10** 246. PMID: 30972003 22. McBride WR, Conlan CE, Barylski NA. **Blood biomarkers in brain injury medicine**. *Curr Phys Med Rehabil Rep* (2022.0) **10** 114-121. DOI: 10.1007/s40141-022-00343-w 23. **Human Cytokine Assays. n.d** 24. King E, O'Brien JT, Donaghy P. **Peripheral inflammation in prodromal Alzheimer's and Lewy body dementias**. *J Neurol Neurosurg Psychiatry* (2018.0) **89** 339-345. PMID: 29248892 25. Maloley PM, England BR, Sayles HR. **Performance of a commercially available multiplex platform in the assessment of circulating cytokines and chemokines in patients with rheumatoid arthritis and osteoarthritis**. *J Immunol Methods* (2021.0) **495** 113048. PMID: 33933473 26. Yue JK, Vassar MJ, Lingsma HF. **Transforming research and clinical knowledge in traumatic brain injury pilot: multicenter implementation of the common data elements for traumatic brain injury**. *J Neurotrauma* (2013.0) **30** 1831-1844. PMID: 23815563 27. Jagoda AS, Bazarian JJ, Bruns JJ. **Clinical policy: neuroimaging and decisionmaking in adult mild traumatic brain injury in the acute setting**. *J Emerg Nurs* (2009.0) **35** e5-e40. DOI: 10.1016/j.jen.2008.12.010 28. Yue JK, Vassar MJ, Lingsma HF, TRACK-TBI Investigators. **Transforming research and clinical knowledge in traumatic brain injury pilot: multicenter implementation of the common data elements for traumatic brain injury**. *J Neurotrauma* (2013.0) **30** 1831-1844. DOI: 10.1089/neu.2013.2970 29. Manley GT, Diaz-Arrastia R, Brophy M. **Common data elements for traumatic brain injury: recommendations from the biospecimens and biomarkers working group**. *Arch Phys Med Rehabil* (2010.0) **91** 1667-1672. PMID: 21044710 30. Diaz-Arrastia R, Wang KK, Papa L. **Acute biomarkers of traumatic brain injury: relationship between plasma levels of ubiquitin C-terminal hydrolase-L1 and glial fibrillary acidic protein**. *J Neurotrauma* (2014.0) **31** 19-25. DOI: 10.1089/neu.2013.3040 31. Lehner J, Wittwer C, Fersching D. **Methodological and preanalytical evaluation of an HMGB1 immunoassay**. *Anticancer Res* (2012.0) **32** 2059-2062. PMID: 22593488 32. Li ZC, Cheng GQ, Hu KZ. **Correlation of synovial fluid HMGB-1 levels with radiographic severity of knee osteoarthritis**. *Clin Invest Med* (2011.0) **34** E298. PMID: 21968272 33. **Cytokines & Chemokines. n.d** 34. Wilson JT, Pettigrew LE, Teasdale GM. **Structured interviews for the Glasgow Outcome Scale and the extended Glasgow Outcome Scale: guidelines for their use**. *J Neurotrauma* (1998.0) **15** 573-585. DOI: 10.1089/neu.1998.15.573 35. Beers SR, Wisniewski SR, Garcia-Filion P. **Validity of a pediatric version of the Glasgow Outcome Scale-Extended**. *J Neurotrauma* (2012.0) **29** 1126-1139. DOI: 10.1089/neu.2011.2272 36. Fiorentino M, Hwang F, Pentakota SR. **Palliative care in trauma: not just for the dying**. *J Trauma Acute Care Surg* (2019.0) **87** 1156-1163. PMID: 31658239 37. Vande Vyvere T, De La Rosa E, Wilms G, CENTER-TBI Participants and Investigators. **Prognostic validation of the NINDS Common Data Elements for the radiologic reporting of acute traumatic brain injuries: a CENTER-TBI Study**. *J Neurotrauma* (2020.0) **37** 1269-1282. DOI: 10.1089/neu.2019.6710 38. Hosmer DW, Lemeshow S, Sturdivant RX. **Applied Logistic Regression**. (2013.0) 39. Morganti-Kossmann MC, Rancan M, Stahel PF. **Inflammatory response in acute traumatic brain injury: a double-edged sword**. *Curr Opin Crit Care* (2002.0) **8** 101-105. PMID: 12386508 40. DeKosky ST, Blennow K, Ikonomovic MD. **Acute and chronic traumatic encephalopathies: pathogenesis and biomarkers**. *Nat Rev Neurol* (2013.0) **9** 192-200. PMID: 23558985 41. Ruff RM, Iverson GL, Barth JT. **Recommendations for diagnosing a mild traumatic brain injury: a National Academy of Neuropsychology education paper**. *Arch Clin Neuropsychol* (2009.0) **24** 3-10. PMID: 19395352 42. Shahim P, Politis A, van der Merwe A. **Time course and diagnostic utility of NfL, tau, GFAP, and UCH-L1 in subacute and chronic TBI**. *Neurology* (2020.0) **95** e623-e636. PMID: 32641529 43. Yue JK, Yuh EL, Korley FK. **Association between plasma GFAP concentrations and MRI abnormalities in patients with CT-negative traumatic brain injury in the TRACK-TBI cohort: a prospective multicentre study**. *Lancet Neurol* (2019.0) **18** 953-961. PMID: 31451409 44. Pan W, Wu X, He Y. **Brain interleukin-15 in neuroinflammation and behavior**. *Neurosci Biobehav Rev* (2013.0) **37** 184-192. PMID: 23201098 45. Budagian V, Bulanova E, Paus R. **IL-15/IL-15 receptor biology: a guided tour through an expanding universe**. *Cytokine Growth Factor Rev* (2006.0) **17** 259-280. PMID: 16815076 46. Gomez-Nicola D, Valle-Argos B, Nieto-Sampedro M.. **Blockade of IL-15 activity inhibits microglial activation through the NFkappaB, p38, and ERK1/2 pathways, reducing cytokine and chemokine release**. *Glia* (2010.0) **58** 264-276. PMID: 19610094 47. Lee GA, Lin TN, Chen CY. **Interleukin 15 blockade protects the brain from cerebral ischemia-reperfusion injury**. *Brain Behav Immun* (2018.0) **73** 562-570. PMID: 29959050 48. Ye RD, Sun L.. **Emerging functions of serum amyloid A in inflammation**. *J Leukoc Biol* (2015.0) **98** 923-929. PMID: 26130702 49. Lin A, Liu J, Gong P. **Serum amyloid A inhibits astrocyte migration via activating p38 MAPK**. *J Neuroinflammation* (2020.0) **17** 254. PMID: 32861245 50. Wicker E, Benton L, George K. **Serum amyloid A protein as a potential biomarker for severity and acute outcome in traumatic brain injury**. *Biomed Res Int* (2019.0) **2019** 5967816. PMID: 31119176 51. Urieli-Shoval S, Shubinsky G, Linke RP. **Adhesion of human platelets to serum amyloid A**. *Blood* (2002.0) **99** 1224-1229. PMID: 11830469 52. Shan LL, Wang YL, Qiao TC. **Association of serum interleukin-8 and serum amyloid A with anxiety symptoms in patients with cerebral small vessel disease**. *Front Neurol* (2022.0) **13** 938655. PMID: 35923828 53. Paudel YN, Shaikh MF, Chakraborti A. **HMGB1: a common biomarker and potential target for TBI, neuroinflammation, epilepsy, and cognitive dysfunction**. *Front Neurosci* (2018.0) **12** 628. PMID: 30271319 54. Ooi SZY, Spencer RJ, Hodgson M. **Interleukin-6 as a prognostic biomarker of clinical outcomes after traumatic brain injury: a systematic review**. *Neurosurg Rev* (2022.0) **45** 3035-3054. DOI: 10.1007/s10143-022-01827-y 55. Datsi A, Steinhoff M, Ahmad F. **Interleukin-31: the “itchy” cytokine in inflammation and therapy**. *Allergy* (2021.0) **76** 2982-2997. PMID: 33629401 56. Lane T, Gillmore JD, Wechalekar AD. **Therapeutic blockade of interleukin-6 by tocilizumab in the management of AA amyloidosis and chronic inflammatory disorders: a case series and review of the literature**. *Clin Exp Rheumatol* (2015.0) **33** S46-S53 57. Panigrahy N, Policarpio J, Ramanathan R.. **Multisystem inflammatory syndrome in children and SARS-CoV-2: a scoping review**. *J Pediatr Rehabil Med* (2020.0) **13** 301-316. PMID: 33252101 58. Ambrósio G, Kaufmann FN, Manosso L. **Depression and peripheral inflammatory profile of patients with obesity**. *Psychoneuroendocrinology* (2018.0) **91** 132-141. PMID: 29550676 59. Forcina L, Franceschi C, Musarò A. **The hormetic and hermetic role of IL-6**. *Ageing Res Rev* (2022.0) **80** 101697. PMID: 35850167
--- title: Incidence of Trigger Finger in Surgically and Nonsurgically Managed Carpal Tunnel Syndrome authors: - Lauren E. Wessel - Alex Gu - Paul Asadourian - Jeffrey G. Stepan - Duretti T. Fufa - Daniel A. Osei journal: Journal of Hand Surgery Global Online year: 2022 pmcid: PMC10039288 doi: 10.1016/j.jhsg.2022.10.017 license: CC BY 4.0 --- # Incidence of Trigger Finger in Surgically and Nonsurgically Managed Carpal Tunnel Syndrome ## Body There is a known relationship between carpal tunnel syndrome (CTS) and trigger finger (TF) or stenosing flexor tenosynovitis.1, 2, 3, 4, 5, 6 Previous studies have demonstrated an $11\%$ to $40\%$ coincidence of CTS and TF.1,7, 8, 9 However, whether one pathology precedes the other is unclear, and studies are further divided as to whether surgical treatment of CTS further increases the risk of TF development.1,3,6,7 Additionally, some surgeons believe that the increased postsurgical swelling from carpal tunnel release (CTR) may contribute to a heightened inflammatory cascade and the development of TF.1,10 Specifically, Lin et al10 noted that this risk was specifically increased in the first 6 months after carpal tunnel surgery and subsequently stabilized. The contribution of surgical intervention to the risk of TF development is of interest in providing presurgical counseling and its possible impact on management, such as whether to offer concomitant surgical release or injection of an early TF. Prior studies on this topic have been performed at single centers, where there may be limited practice variation.1,3,6 To address this limitation, our study used a large administrative database to gather epidemiologic information regarding the development of TF after CTR. Specifically, among a population of patients with diagnosed CTS, our primary aim was to determine whether extremities undergoing CTR had an increased rate of TF than conservatively managed CTS in a propensity-matched population. Our secondary aims were the following: [1] to determine whether CTR is associated with an increased rate of ipsilateral TF compared with the contralateral extremity, acting as an internal control, and [2] to determine the time course of TF development after CTR. We hypothesized that the extremities that undergo CTR for CTS would show an increased risk of developing TF during a 6-month postsurgical period compared with a nonsurgical cohort. ## Abstract ### Purpose The purpose of this study was to determine whether extremities undergoing carpal tunnel release (CTR) have an increased rate of trigger finger (TF) compared with conservatively managed carpal tunnel syndrome. ### Methods Data were collected from the Humana Insurance Database, and subjects were chosen on the basis of a history of CTR with propensity matching performed to develop a nonsurgical cohort. Following propensity matching, 16,768 patients were identified and equally split between surgical and nonsurgical treatments. Demographic information and medical comorbidities were recorded. Univariate and multivariate analyses were performed to identify risk factors for the development of TF within 6 months of carpal tunnel syndrome diagnosis. ### Results Patients in the surgical cohort were more likely to develop TF than those in the nonsurgical cohort whether in the ipsilateral or contralateral extremity. Whether managed surgically or nonsurgically, extremities with carpal tunnel syndrome demonstrated an increased prevalence of TF than their contralateral, unaffected extremity. ### Conclusions Surgeons should be aware of the association of TF and CTR both during the presurgical and postsurgical evaluations as they might impact patient management. With knowledge of these data, surgeons may be more attuned to detecting an early TF during the postsurgical period and offer more aggressive treatment of TF pathology during CTR. ### Type of study/level of evidence Prognostic III. ## Materials and Methods Data were collected from the Humana Insurance Database using the PearlDiver Patient Records Database11 from 2015 to 2017. The PearlDiver database contains records for more than 22 million patients, further describing hospital and physician billing records and procedural information. Subjects with CTS were identified using the International Statistical Classification of Diseases, Tenth Revision (ICD-10) codes. To maintain laterality, only ICD-10 codes were used. Relevant diagnostic and procedural codes are referenced in Appendix 1 (available on the Journal’s website at www.jhsgo.org). Demographic information and medical comorbidities were recorded. Using laterality-specific coding, the incidence of the TF diagnosis within 6 months of CTR was determined using the ICD-10 diagnosis codes. The incidence of TF diagnosis among the nonsurgical group was determined within 6 months of CTS diagnosis. The incidence of TF was determined by the concomitant, laterality-specific ICD-10 code and the Current Procedure T erminology code for TF injection or release, indicating a clinically relevant diagnosis of TF. The International Statistical Classification of Diseases, Tenth *Revision diagnosis* codes associated with unspecified limbs were excluded. Carpal tunnel release was determined on the basis of the presence of a CTR-related Current Procedure Terminology code (Appendix 1). Two patient cohorts were identified as follows: [1] surgical cohort: patients who underwent CTR for CTS, and [2] nonsurgical cohort: patients who were treated conservatively for CTS within our study period (Fig. 1). Patients were followed up for the development of TF in the ipsilateral or contralateral extremity. These 2 sets of breakpoints served as a proxy for the severity of disease during the study period (surgical vs nonsurgical management) and controlled for patients’ biology and the effects of surgery (comparison of ipsilateral or contralateral). Patients were required to have a continuous, active status in the database for inclusion in the study to ensure no loss of patients due to changes in insurance carrier status. Additionally, each patient’s contralateral extremity was used as an internal control for each of these cohorts. Patients were excluded if the CTS diagnosis existed bilaterally to maintain a non-CTS internal control for comparison. Additionally, patients were excluded if they had a prior history of TF as determined by ICD-10 attribution or CTS within the lifespan of the database, if laterality could not be determined, or if they had less than 6 months of follow-up. The control of comorbidities was conducted using the Charlson Comorbidity Index (CCI) and was used for propensity matching.12Figure 1The study population. CTR: Carpal Tunnel Release; CTS: Carpal Tunnel Syndrome; TF: Trigger Finger ## Propensity score matching To balance measured and unmeasured covariates, and thus, mitigate potential confounders, we used propensity score matching to create matched cohorts of surgical and nonsurgical patients with CTS. The propensity score was defined as the conditional probability of having undergone CTR based on age, sex, CCI, and diabetes mellitus status. These factors were chosen given their association with the development of TF.13,14 Matching was conducted using a 1:1 nearest neighbor matching by univariate analysis, with the caliper set at 0.02 of the standard deviation. Matching was conducted using demographics collected at the time of CTS diagnosis among the surgical and nonsurgical cohorts. Propensity score matching was conducted using R software provided by PearlDiver. ## Statistical analysis Data on the patients’ demographics, CCI, history of diabetes mellitus, and incidence of TF were analyzed using univariate and multivariate analyses on the R software provided within PearlDiver. Propensity matching was conducted to reduce potential confounders and provide a more homogeneous cohort. Univariate analysis was first performed using Pearson chi-square or analysis of variance. For the multivariate analyses, logistic regression analyses were performed to determine adjusted associations of risk factors for postsurgical TF. The results were reported as odds ratios (ORs) and $95\%$ confidence intervals (CIs). A P value of <.05 was used as the cutoff for significance. ## Results Following propensity matching, 16,768 patients with a unilateral CTS diagnosis were selected for the study. Among those, 8,384 patients who underwent CTR constituted the surgical cohort and 8,384 patients who underwent conservative management for CTS were selected as their closest propensity-matched controls. Baseline patient demographics and clinical characteristics are listed in the Table. *In* general, there were no differences between the surgical and nonsurgical cohorts with respect to gender, age, CCI, or diabetes mellitus status (Table). Among all extremities that developed TF, the middle finger was the most common digit affected, followed by the ring and thumb fingers (Fig. 2).TableDemographics and Medical Comorbidities of Patients Diagnosed With Carpal Tunnel SyndromeCategoryCTRNon-CTRP Value∗$$n = 8$$,384N = 8,384n%n%Sex.823 Male3,31239.53,31039.5 Female5,07260.55,07460.5Age (y).453 <505646.75666.8 50–591,08212.91,08012.9 60–692,25726.92,25826.9 70–793,02436.13,02936.1 80–891,28715.41,28615.3 >901702.01652.0CCI.754 02,90634.72,91334.7 11,58919.01,58518.9 21,07312.81,08813.0 390310.889010.6 4+1,91322.81,90822.8Diabetes mellitus6647.96738.0.542∗Significant P values are $P \leq .05.$Figure 2The distribution of trigger finger by cohort. ## Trigger finger development among surgical versus nonsurgical cohorts (propensity matching) In total, 752 ($9.0\%$) extremities in the surgical cohort and 401 ($4.9\%$) extremities in the nonsurgical cohort were treated for TF during the 6-month study period (OR: 1.9; $95\%$ CI: 1.4–2.2; $P \leq .001$; Fig. 1). There was increased treatment of the thumb ($23\%$ vs $19\%$; $P \leq .05$) and the middle ($33\%$ vs $31\%$; $P \leq .05$) finger in the surgical cohort compared with the nonsurgical cohort (Fig. 2). The average time between CTS diagnosis to TF treatment was approximately 8 weeks and was similar between the surgical and conservatively managed cohorts (52.3 vs 56.2 days, respectively, $$P \leq .102$$). The overall time from incidence of CTR to ipsilateral TF injection or first annular pulley release was 36.3 days. ## Trigger Finger development versus the contralateral extremity in the surgical cohort (internal control) Among patients with unilateral CTS who were treated with surgical intervention, 752 ($9.0\%$) ipsilateral extremities and 268 ($3.2\%$) contralateral extremities were treated for TF during the 6-month study period (OR: 2.8; $95\%$ CI: 2.4–3.2; $P \leq .01$; Fig. 1). There was increased treatment of the thumb ($23\%$ vs $20\%$; $P \leq .05$) and decreased treatment of the middle ($37\%$ vs $33\%$; $P \leq .05$) finger in the ipsilateral cohort compared with the contralateral cohort (Fig. 2). ## Trigger finger development versus the contralateral extremity in the nonsurgical cohort (internal control) Patients who were diagnosed with CTS and treated conservatively also had an increased rate of TF of the ipsilateral compared with the contralateral side. Among 401 ipsilateral extremities, $4.8\%$ were treated for TF during the 6-month study, although there were only 225 ($2.7\%$) contralateral extremities that were treated for TF in the same period (OR: 1.8; $95\%$ CI: 1.4–2.2; $$P \leq .025$$; Fig. 1). There was no difference in the frequency of TF between various fingers in the ipsilateral extremity compared with the contralateral extremity. ## Discussion The relationship between CTS and TF has been investigated in recent studies in the hand literature.1, 2, 3, 4, 5, 6, 7,10,15, 16, 17 Yet, the literature does not clearly address whether an association exists between surgical intervention for CTS and the development of TF. We used the recent switch to ICD-10 coding, and thus, the introduction of laterality-specific coding to study a large administrative database of patients diagnosed with CTS. We studied the rates of TF in the CTS-affected extremities and compared the surgical and nonsurgical CTS cohorts. Additionally, we studied the rate of TF development against the respective contralateral extremities of each cohort. In our propensity-matched cohort analysis, we observed an association between surgical treatment for CTS and the increased rate of development of TF in the ipsilateral extremity compared with extremities with CTS that were managed nonsurgically. In both groups, the development of TF was similar at 8 weeks. In the internal control portion of the study, extremities with CTS diagnosis, irrespective of conservative versus surgical management, demonstrated an increased likelihood of TF development compared with the contralateral upper extremity. As a result, we caution against using these findings to suggest that CTR causes TF. However, they further illuminate the relationship between CTS and TF to raise the level of scrutiny for TF in patients with CTS and counsel patients. Trigger finger occurs at higher rates in extremities affected by CTS, rheumatoid arthritis, and hypothyroidism, supporting the belief that an inflammatory process may play a role in the development of TF.18, 19, 20 Moreover, a recent retrospective study by Zhang et al6 showed a lateral association of CTR and TF, as new-onset TF is 2.5 times more likely to develop in the surgical hand than the contralateral nonsurgical hand within the first postsurgical year. These findings are similar to our reported OR of 2.8 for this cohort. Given the co-occurrence of TF and the systemic inflammatory processes as well as postsurgical states,21 we hypothesized that patients who underwent surgical management for CTS would exhibit higher rates of TF than for those managed conservatively. Prior data on this association in the literature is varied. Although Zhang et al6 showed a lateral association between TF and CTR, they did not find a temporal association between CTR and the development of postsurgical TF in their single-center study of retrospectively collected data. Their data indicated that new-onset TF was $50\%$ less likely to develop in the surgical hand during the postsurgical year than the year before CTR. Conversely, many studies have highlighted the presence of the temporal course of postsurgical TF development.7,10,15,17 Furthermore, Lee et al17 studied the anatomic underpinnings behind this relationship. The authors used ultrasound to determine that patients who developed TF after CTR tended to have significantly increased volar migration of their flexor tendons relative to those who did not develop TF. They hypothesized that this volar migration was made possible by CTR. Our data regarding the comparison of a propensity-matched cohort and the comparison to the contralateral upper extremity support the association between CTR and the development of TF. However, based on these data, we could not determine the cause of the association. Whether this association is secondary to the increased severity of CTS pathology or surgical release remains uncertain. In our analysis, we determined that the most commonly affected digit in TF following CTR was the middle digit, followed by the ring and thumb digits (Fig. 2). This finding is distinct from the current literature, which most commonly cites the thumb as having the greatest involvement in the setting of CTS.1,3,6,17 However, our data support an increase in the involvement of the thumb and the middle digit in the surgical cohort compared with their nonsurgical counterparts, which may account for some discrepancies with prior literature. Additionally, the predilection for median nerve-innervated fingers may indicate that patients had early symptoms in these digits, which may have been masked or underappreciated because of concomitant median neuropathy. This point further underscores the importance of attention to this association in the presurgical patient physical examination and counseling, particularly in that increased inflammation in the postsurgical state may further contribute to the development of CTS. Regarding the timing of TF onset after CTS diagnosis, we found no difference between the time from CTS diagnosis to TF development between patients whose CTS was managed surgically versus those who underwent conservative management. These data are similar to those published by Zhang et al6, which failed to demonstrate a temporal relationship between CTR and new-onset TF. However, whereas Zhang et al6 used the period before CTR as a control for TF development, our study used a propensity-matched control cohort of conservatively managed patients with CTS to understand this difference. We believe that our comparison with the ipsilateral upper extremity and a cohort of propensity-matched patients with conservatively managed CTS appropriately controls for the natural history of TF development in our cohort. This study has several limitations. First, as this retrospective study used Humana insurance records, the study is subject to inaccuracies during the billing process. To limit the impact of these inaccuracies, we eliminated subjects with incomplete or unspecified billing codes and those with less than 6 months of follow-up within the insurance database. Our use of procedure codes in combination with ICD-10 diagnosis of TF focuses the study on clinically relevant TF for which intervention is sought and ensures closer alignment with coding specificity, as providers may be inconsistent in applying ICD-10 codes but are much more accurate at coding in combination with procedural codes. Additionally, our results showing increased TF in CTS extremities treated surgically may be biased by patients treated surgically who were under the care of a surgeon whose experience and threshold to treat TF may have been different than those patients treated nonsurgically and may have been in the care of generalists or specialists. We used a follow-up period of 6 months to determine TF development, which may exclude patients who received treatment for TF outside of the 6-month window. However, we believed that the 6-month follow-up period highlighted cases of TF development associated with CTS diagnosis and that a longer follow-up timeframe may reflect the natural history of TF development. This is reflected in our analysis in that TF treatment, on average, occurs within 2 months of CTS diagnosis in each group. Given the large administrative database, we could not standardize the indications for intervention for CTS or TF. Therefore, heterogeneity in the severity of pathology for any specific intervention may exist within our cohort. Additionally, given that the purpose of this study was to examine the development of a primary TF in the setting of CTS, we cannot comment on the frequency of recurrent or worsening TF. This may be an area of interest for future study. Lastly, we were unable to determine causality with this administrative database. We acknowledge and even hypothesize that the detected differences between the surgical and nonsurgical cohorts might be secondary to the postsurgical state, a generalized increased inflammatory state, or secondary to more extreme CTS severity in surgically managed patients. Further studies examining these factors may better elucidate these factors. In conclusion, in this study, we demonstrated an association between CTS and TF, which is increased in patients managed surgically for CTS compared with those managed nonsurgically during a 6-month follow-up period. This study used a laterality-specific administrative database to understand these relationships. Additionally, enabled by the number of patients available in such a database, our study is distinct in its use of propensity matching to compare surgically managed patients with CTS with conservatively managed patients. Further studies are required to determine the cause of this association. Surgeons should be aware of these associations during presurgical and postsurgical evaluations as they might impact patient management. Considering these data, surgeons who detect an early TF may offer more aggressive treatment of TF pathology during CTR. Similarly, surgeons may be more attuned to detecting an early TF during postsurgical assessment with the knowledge of these data. ## Supplementary Data Appendix 1 ## References 1. Hayashi M., Uchiyama S., Toriumi H., Nakagawa H., Kamimura M., Miyasaka T.. **Carpal tunnel syndrome and development of trigger digit**. *J Clin Neurosci* (2005.0) **12** 39-41. PMID: 15639409 2. Kim J.H., Gong H.S., Lee H.J., Lee Y.H., Rhee S.H., Baek G.H.. **Pre- and post-operative comorbidities in idiopathic carpal tunnel syndrome: cervical arthritis, basal joint arthritis of the thumb, and trigger digit**. *J Hand Surg Eur Vol* (2013.0) **38** 50-56. PMID: 22553311 3. King B.A., Stern P.J., Kiefhaber T.R.. **The incidence of trigger finger or de Quervain’s tendinitis after carpal tunnel release**. *J Hand Surg Eur Vol* (2013.0) **38** 82-83. PMID: 22791612 4. Kumar P., Chakrabarti I.. **Idiopathic carpal tunnel syndrome and trigger finger: is there an association?**. *J Hand Surg Eur Vol* (2009.0) **34** 58-59. PMID: 18936127 5. Wessel L.E., Fufa D.T., Boyer M.I., Calfee R.P.. **Epidemiology of carpal tunnel syndrome in patients with single versus multiple trigger digits**. *J Hand Surg Am* (2013.0) **38** 49-55. PMID: 23200219 6. Zhang D., Collins J., Earp B.E., Blazar P.. **Relationship of carpal tunnel release and new onset trigger finger**. *J Hand Surg Am* (2019.0) **44** 28-34. PMID: 30409506 7. Goshtasby P.H., Wheeler D.R., Moy O.J.. **Risk factors for trigger finger occurrence after carpal tunnel release**. *Hand Surg* (2010.0) **15** 81-87. PMID: 20672394 8. Harada K., Nakashima H., Teramoto K., Nagai T., Hoshino S., Yonemitsu H.. **Trigger digits-associated carpal tunnel syndrome: relationship between carpal tunnel and trigger digits**. *Hand Surg* (2005.0) **10** 205-208. PMID: 16568515 9. Rottgers S.A., Lewis D., Wollstein R.A.. **Concomitant presentation of carpal tunnel syndrome and trigger finger**. *J Brachial Plex Peripher Nerve Inj* (2009.0) **4** 13. PMID: 19706185 10. Lin F.Y., Manrique O.J., Lin C.L., Cheng H.T.. **Incidence of trigger digits following carpal tunnel release: a nationwide, population-based retrospective cohort study**. *Medicine (Baltimore)* (2017.0) **96** 11. 11PearlDiver. Humana Insurance Database. Accessed September 7, 2020. www.pearldiverinc.com 12. Sundararajan V., Henderson T., Perry C., Muggivan A., Quan H., Ghali W.A.. **New ICD-10 version of the Charlson comorbidity index predicted in-hospital mortality**. *J Clin Epidemiol* (2004.0) **57** 1288-1294. PMID: 15617955 13. Moore J.S.. **Flexor tendon entrapment of the digits (trigger finger and trigger thumb)**. *J Occup Environ Med* (2000.0) **42** 526-545. PMID: 10824306 14. Saldana M.J.. **Trigger digits: diagnosis and treatment**. *J Am Acad Orthop Surg* (2001.0) **9** 246-252. PMID: 11476534 15. Hombal J.W., Owen R.. **Carpal tunnel decompression and trigger digits**. *Hand* (1970.0) **2** 192-196. PMID: 5520745 16. Karalezli N., Kütahya H., Güleç A., Toker S., Karabörk H., Ogun T.C.. **Transverse Carpal ligament and forearm fascia release for the treatment of carpal tunnel syndrome change the entrance angle of flexor tendons to the a1 pulley: the relationship between carpal tunnel surgery and trigger finger occurrence**. *ScientificWorldJournal* (2013.0) **2013** 17. Lee S.K., Bae K.W., Choy W.S.. **The relationship of trigger finger and flexor tendon volar migration after carpal tunnel release**. *J Hand Surg Eur Vol* (2014.0) **39** 694-698. PMID: 23442341 18. Cakir M., Samanci N., Balci N., Balci M.K.. **Musculoskeletal manifestations in patients with thyroid disease**. *Clin Endocrinol (Oxf)* (2003.0) **59** 162-167. PMID: 12864792 19. Chammas M., Bousquet P., Renard E., Poirier J.L., Jaffiol C., Allieu Y.. **Dupuytren’s disease, carpal tunnel syndrome, trigger finger, and diabetes mellitus**. *J Hand Surg Am* (1995.0) **20** 109-114. PMID: 7722249 20. Clark D.D., Ricker J.H., MacCollum M.S.. **The efficacy of local steroid injection in the treatment of stenosing tenovaginitis**. *Plast Reconstr Surg* (1973.0) **51** 179-180. PMID: 4683941 21. Wessel L.E., Gu A., Coxe F., Stepan J.G., Fufa D.T., Osei D.A.. **Brief communication: operative fixation for distal radius fracture increases rates of trigger finger over a 6-month post-operative period**. *Orthop Res J* (2020.0) **6**
--- title: Awareness of Stroke Risk Factors, Warning Signs, and Preventive Behaviour Among Diabetic Patients in Al-Ahsa, Saudi Arabia journal: Cureus year: 2023 pmcid: PMC10039371 doi: 10.7759/cureus.35337 license: CC BY 3.0 --- # Awareness of Stroke Risk Factors, Warning Signs, and Preventive Behaviour Among Diabetic Patients in Al-Ahsa, Saudi Arabia ## Abstract Objectives This study aims to measure the level of awareness about stroke symptoms, risk factors, and preventive health practices that could be taken to reduce the risk of stroke among diabetic patients in Al-Ahsa, Saudi Arabia. Methods A cross-sectional study was conducted in Al-Ahsa, Saudi Arabia in 2020. The sample included a total of 202 male and female Saudi adults aged 18-65 years, with either type 1 or type 2 diabetes mellitus, and living in Al Ahsa, Saudi Arabia. The information was collected randomly through an online questionnaire distributed among patients after getting their contact information from relevant governmental and private diabetes clinics and after signing the informed consent. For awareness and knowledge items, each correct answer was scored one point and the total summation of the discrete scores of the different items was calculated. A diabetic patient with a score less than $60\%$ of the total score was considered to have poor awareness while a score of $60\%$ or more of the total score was considered a good level of awareness. Results A total of 87 ($43.1\%$) participants had an overall good awareness level, while 115 ($56.9\%$) had poor awareness levels. Around $40.6\%$ of the study patients had heard about stroke, $61.9\%$ knew that stroke affects the brain, and $24.3\%$ reported that stroke is higher among males. As for factors associated with stroke, the most reported was high blood pressure ($71.8\%$), followed by diabetes mellitus ($69.3\%$). Exactly $65.8\%$ of participants knew about the mechanism of ischemic stroke and $42.6\%$ reported hemorrhagic stroke. A high percentage of patients ($73.1\%$) realize that they could reduce their risk of stroke. Conclusion The findings of the current study showed that less than half ($43.1\%$) of the Saudi patients with DM had a good awareness level regarding stroke and its related risk factors and warning signs. Older patients (aged 50-65 years) with high social levels (high education and income) and those with a family history of stroke had significantly higher awareness levels. Hypertension, DM, and smoking are the highest reported known risk factors of stroke, and speech disorders are the highest known stroke presentation to the respondents. ## Introduction Diabetes mellitus (DM) is a metabolic disorder characterized by either insulin resistance or the inability of the body to produce a proper amount of insulin or both, resulting in hyperglycemia [1,2]. The prevalence of DM is estimated to be $34\%$ in the Saudi population, while the global prevalence of DM is about $9.3\%$. The Kingdom of Saudi Arabia has the seventh highest DM prevalence [3-5]. DM is considered an independent risk factor for stroke due to the fact that DM causes vascular aging, which leads to microvascular and macrovascular complications. Stroke is a neurological condition, which occurs abruptly due to a pathology in the vessels of the brain or in the vessels from or to the brain. DM is a chronic disease that does not only need medical attention but also requires self-care and a patient's knowledge about the potential secondary illnesses that might be caused by it [6-11]. DM is considered a major modifiable risk factor for stroke. Moreover, DM is highly associated with pathophysiological changes including vascular endothelial dysfunction, arterial stiffness, thickening of the capillary basal membrane, and systemic inflammation. Stroke is one of the highlighted macrovascular complications of DM and it is more likely to be associated with type-2 DM [6-11]. Furthermore, diabetic patients have twice the risk of developing both ischemic and hemorrhagic stroke compared with those who are non-diabetic, which makes DM accountable for almost one-quarter of all stroke cases [6-11]. In addition, diabetic patients who developed stroke have less favorable outcomes [9]. Worldwide, stroke and other cerebrovascular accidents are major health issues with a huge burden at the individual, family, and social levels [8]. Also, the financial burden is a significant issue with an estimated direct medical cost of $273-$818 billion between 2010 and 2030 in the United States alone [8]. Stroke is one of the most prevalent diseases in Saudi Arabia with a prevalence rate of 43.8 per 100,000 in Riyadh and 40 per 100,000 in the Eastern province. In Saudi Arabia, stroke is the third leading cause of death after ischemic heart disease and road injuries [12-13]. A previous study assessed the knowledge of stroke risk factors and warning symptoms among the *Saudi* general population and found that $63.8\%$ of the participants had a poor knowledge level [14]. Another study conducted among stroke patients at King Abdulaziz Medical City, Riyadh, Saudi Arabia, has shown that more than $50\%$ of the patients were unaware that they were having a stroke. Most of them sought medical care late because of failing to notice signs and symptoms [15]. Therefore, this study aimed to measure awareness levels about stroke symptoms, risk factors, and preventive health practices among patients with diabetes. Subsequently, this will lead to a reduction in the risk of stroke among these patients. The results of the study would be of help to the local health authorities in planning effective educational programs to increase the awareness of these patients. Eventually, this may reduce the burdens and costs caused by stroke. ## Materials and methods A cross-sectional questionnaire-based study was conducted in Al-Ahsa, Saudi Arabia, from January 1, 2022, to December 31, 2022. The population included both male and female diabetic Saudi adults, 18-65 years of age, living in Al Ahsa, Saudi Arabia. All types of DM were included. The study was approved by the Research Committee of King Fahad Hospital, Hofuf, AI Ahsa, Kingdom of Saudi Arabia (Approval number: IRB KFHH (H05-HS-065) 46-33-2020). The calculated sample size was 385 participants, determined by the Richard Geiger equation, with a margin of error determined as $5\%$, a confidence level of $95\%$, the population as 1,041,863, and $50\%$ for response distribution. The information was collected randomly through an online questionnaire distributed among patients after getting their contact information from relevant governmental and private DM clinics. All patients signed approval to share in the study after being informed of the study details and its rationale. We obtained responses only from 202 participants. After data were extracted, they were revised, coded, and analyzed using IBM SPSS Statistics for Windows, Version 22.0 (Released 2013; IBM Corp., Armonk, New York, United States). All statistical analysis was done using two-tailed tests. A p-value less than 0.05 was considered statistically significant. For awareness and knowledge items, each correct answer was scored one point and the total summation of the discrete scores of the different items was calculated. A diabetic patient with a score less than $60\%$ of the total score was considered to have poor awareness while a score of $60\%$ or more of the total score was considered good awareness. Descriptive analysis based on frequency and percent distribution was done for all variables including patients’ demographic data, medical and family history, and smoking. Also, patients’ awareness regarding stroke, risk factors, warning signs, and prevention with preferred sources of information were tabulated and graphed. Cross tabulation was used to assess the distribution of knowledge levels according to patients' personal data, medical data, and risk factors. Relations were tested using the Pearson chi-square test and exact probability test for small frequency distributions. ## Results A total of 202 diabetic patients completed the study questionnaire. Patients were aged 18 to 65 years with a mean age of 45.1 ± 16.6 years old. A total of 101 ($50\%$) patients were females and $50\%$ were males. A total of 137 ($67.8\%$) patients were married and 73 ($36.1\%$) were university educated while 66 ($32.7\%$) had a secondary level of education. Considering work, 102 ($50.5\%$) were non-healthcare workers, healthcare workers represented only $5\%$ of the samples, and the rest are students or not working; a monthly income of less than 3000 SR was reported by 77 ($38.1\%$) participants (Table 1). **Table 1** | Socio-demographic data | No | % | | --- | --- | --- | | Age in years | | | | 18-24 | 34.0 | 16.8% | | 25-34 | 25.0 | 12.4% | | 35-49 | 57.0 | 28.2% | | 50-65 | 86.0 | 42.6% | | Gender | | | | Male | 101.0 | 50.0% | | Female | 101.0 | 50.0% | | Marital status | | | | Single | 65.0 | 32.2% | | Married | 137.0 | 67.8% | | Educational level | | | | Below secondary | 63.0 | 31.2% | | Secondary | 66.0 | 32.7% | | University | 73.0 | 36.1% | | Work | | | | Not working / student | 90.0 | 44.6% | | Non-health care worker | 102.0 | 50.5% | | Health care worker | 10.0 | 5.0% | | Monthly income | | | | < 3000 SR | 77.0 | 38.1% | | 3000-5000 SR | 33.0 | 16.3% | | 5000-10000 SR | 52.0 | 25.7% | | > 10000 SR | 40.0 | 19.8% | Exactly 95 ($47\%$) patients had hypercholesterolemia, 88 ($43.6\%$) were hypertensive, 46 ($22.8\%$) had cardiovascular disease, and 32 ($15.8\%$) were smokers for more than one year. A total of 74 ($36.8\%$) patients had a family history of stroke (Table 2). **Table 2** | Medical & family history | No | % | | --- | --- | --- | | Co-morbidities | | | | HTN | 88.0 | 43.6% | | CVD | 46.0 | 22.8% | | Hypercholesterolemia | 95.0 | 47.0% | | Smoking for more than 1 year | 32.0 | 15.8% | | Do you have anyone in your family who has or had a stroke | | | | Yes | 74.0 | 36.8% | | No | 103.0 | 51.2% | | Don’t know | 24.0 | 11.9% | | Have you had a stroke before? | | | | Yes | 11.0 | 5.5% | | No | 182.0 | 90.5% | | Don’t know | 8.0 | 4.0% | | If yes, how many times you had it? | | | | 1 time | 9.0 | 81.8% | | > 3 times | 2.0 | 18.2% | Generally, $40.6\%$ of the study patients heard about stroke, $61.9\%$ knew that stroke affected the brain, and $24.3\%$ reported that stroke was higher among males. As for factors associated with stroke, the most reported was high blood pressure ($71.8\%$), followed by DM ($69.3\%$), cigarette smoking ($64.9\%$), and high cholesterol level ($57.4\%$). A total of $65.8\%$ knew about the mechanism of ischemic stroke. Regarding clinical presentation, $70.8\%$ mentioned speech disorders, and $69.8\%$ knew of weakness or disability to move one-half of the body. A total of $73.1\%$ of the patients knew that they could reduce the risk of stroke (Table 3). A total of 87 ($43.1\%$) had an overall good awareness level while 115 ($56.9\%$) had a poor awareness level (Figure 1). Exactly $52.3\%$ of the older patients (aged 50-65 yers) had a good awareness level compared to $47.1\%$ of the younger age group, which was was statistically significant (P-value=0.021). Also, $67.5\%$ of patients with high income showed a good awareness level (P-value=0.004). A total of $50.5\%$ of patients with hypercholesterolemia had a good awareness of stroke (P-value=0.044). Additionally, $51.4\%$ of patients with a family history of stroke had a good awareness of the disease (P-value=0.012) (Table 4). The most preferred source was healthcare personnel ($53.2\%$), followed by the internet ($51.7\%$), family/friends ($26.9\%$), books ($19.4\%$), and television and radio ($18.9\%$) (Figure 2). ## Discussion The current study aimed to assess diabetic patients’ awareness regarding stroke risk factors, warning signs, and prevention in Al-Ahsa, Saudi Arabia. The study showed that less than half ($43.1\%$) of the patients had a good awareness level regarding stroke and its related risk factors and warning signs. More specifically, less than half ($40.6\%$) of the study patients had heard about stroke. About two-thirds ($61.9\%$) of the patients knew that stroke affects the brain, but only one-fourth of them reported that stroke is higher among males. The literature showed that the prevalence of stroke is higher among men till the age of 80 years, after that it is higher in women. Most of the studies concluded that the case fatality rate is higher in female than in male stroke patients; there is also some evidence, albeit relatively weak, indicating a better functional outcome in men. As for factors associated with stroke, the most reported was high blood pressure, DM, cigarette smoking, and high cholesterol level [16-19]. Regarding knowing stroke different clinical presentations by the study group, $70.8\%$ reported knowing about speech disorders, $69.8\%$ knew about weakness or disability to move one-half of the body, $64.4\%$ knew about decreased vision, and $62.9\%$ reported knowing about decreased sensation or inability to feel things. A total of $73.1\%$ of the patients knew that they could reduce the risk of stroke. The study also revealed that older patients (aged 50-65 years) with high social levels (high education and income) and those with a family history of stroke had significantly higher awareness levels. Arisegi et al. showed that $70.3\%$ of diabetic patients had good knowledge of stroke, organs or parts of the body affected by stroke ($89.1\%$), signs or symptoms of stroke ($87.0\%$), stroke risk factors ($86.6\%$), and stroke prevention ($90.8\%$) [20]. Formal education was the sole predictor of good knowledge of the signs or symptoms of stroke. Studies in Nigeria [21,22], the United States [23], and Australia [24] also showed that the majority of the participants had good knowledge of stroke as a disease of the blood vessels in the brain. Regarding warning signs and clinical presentation, a study in Ghana reported numbness or paralysis as the most common stroke warning sign known to participants [25]. While it also concurs with the findings in studies conducted in Osogbo, Nigeria [22], Benin [26], and Nigeria [27]. It differs from the findings in studies conducted in Australia [24] and Ireland [28] that reported visual problems and slurred speech, respectively, as the most common stroke signs identified. In Saudi Arabia, Alhazzani et al. found that $63.6\%$ and $43.7\%$ of primary health center patients correctly identified thrombosis and hemorrhage, respectively, as types of strokes. The most reported risk factors were hypertension ($55.8\%$), dyslipidemia ($45.8\%$), and smoking ($41.9\%$). Sudden severe headache ($54.1\%$), dizziness ($51.0\%$), and difficulty in speaking ($44.3\%$) were the most frequently recognized symptoms [29]. Another study revealed that the mean knowledge of stroke risk among hypertensive patients was 10.73 ±3.53 while the mean knowledge of warning signs was 9.276±2.99 [30]. The main limitation was the smaller than desired sample size. ## Conclusions The findings of the current study showed that less than half ($43.1\%$) of the patients had a good awareness level regarding stroke and its related risk factors and warning signs. Older patients with high social levels (high education and income) and those with a family history of stroke had significantly higher awareness levels. As for factors associated with stroke, the most reported was high blood pressure, DM, cigarette smoking, and high cholesterol level. Regarding the awareness of stroke clinical presentations, $70.8\%$ knew about speech disorders, $69.8\%$ knew about weakness or disability to move one-half of the body, $64.4\%$ know about decreased vision, and $62.9\%$ knew about decreased sensation or inability to feel things. Finally, formal education was the sole predictor of good knowledge of the signs or symptoms of a stroke. So, further emphasis on the importance of formal education is required. ## References 1. **Diagnosis and classification of diabetes mellitus**. *Diabetes Care* (2010) **33** 0-9 2. Al-Ramadhan HJ, Alzuhair HA, Al-Aithan HA, Abdrabalnabi A, Aljumaiah HA. **Perception of cardiovascular risks among type-2 diabetic patient population in Al-Ahsa, Saudi Arabia**. *IJMRP* (2018) **4** 24-28 3. Aldossari KK, Aldiab A, Al-Zahrani JM. **Prevalence of prediabetes, diabetes, and its associated risk factors among males in saudi arabia: a population-based survey**. *J Diabetes Res* (2018) **2018** 2194604. PMID: 29850599 4. Saeedi P, Petersohn I, Salpea P. **Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: R\results from the International Diabetes Federation Diabetes Atlas, 9th edition**. *Diabetes Res Clin Pract* (2019) **157** 107843. PMID: 31518657 5. Al-Rubeaan K, Al-Hussain F, Youssef AM, Subhani SN, Al-Sharqawi AH, Ibrahim HM. **Ischemic stroke and its risk factors in a registry-based large cross-sectional diabetic cohort in a country facing a diabetes epidemic**. *J Diabetes Res* (2016) **2016** 4132589. PMID: 26989695 6. Alloubani A, Saleh A, Abdelhafiz I. **Hypertension and diabetes mellitus as a predictive risk factors for stroke**. *Diabetes Metab Syndr* (2018) **12** 577-584. PMID: 29571978 7. Liao CC, Shih CC, Yeh CC, Chang YC, Hu CJ, Lin JG, Chen TL. **Impact of diabetes on stroke risk and outcomes: two nationwide retrospective cohort studies**. *Medicine (Baltimore)* (2015) **94** 0 8. Chen R, Ovbiagele B, Feng W. **Diabetes and stroke: epidemiology, pathophysiology, pharmaceuticals and outcomes**. *Am J Med Sci* (2016) **351** 380-386. PMID: 27079344 9. Mostafa M, Mohamed N. **Effect of glycemic control on the severity and outcome of stroke in Saudi Arabia**. *Egypt J Neurol Psychiatr Neurosurg* (2015) **52** 228-210 10. Rammal SA, Almekhlafi MA. **Diabetes mellitus and stroke in the Arab world**. *J Taibah Univ Med Sci* (2016) **11** 295-300 11. Hill MD. **Stroke and diabetes mellitus**. *Handb Clin Neurol* (2014) **126** 167-174. PMID: 25410221 12. Alaqeel A, AlAmmari A, AlSyefi N, Al-Hussain F, Mohammad Y. **Stroke awareness in the Saudi community living in Riyadh: prompt public health measures must be implemented**. *J Stroke Cerebrovasc Dis* (2014) **23** 500-504. PMID: 23680680 13. **Global Health - Saudi Arabia**. (2023) **5** 14. Alreshidi FM, Alrashidi AS, Alshammari FN, Qadi AB, Alrashidi AG, Alghaythi SM. **Knowledge, attitude and practicetowards stroke risk factors and warning symptoms in Saudi Arabia, 2017**. *Egypt J Hosp Med* (2017) **69** 2082-2087 15. Al Khathaami AM, Mohammad YO, Alibrahim FS, Jradi HA. **Factors associated with late arrival of acute stroke patients to emergency department in Saudi Arabia**. *SAGE Open Med* (2018) **6** 2050312118776719. PMID: 29844910 16. Thomas Q, Crespy V, Duloquin G, Ndiaye M, Sauvant M, Béjot Y, Giroud M. **Stroke in women: when gender matters**. *Rev Neurol (Paris)* (2021) **177** 881-889. PMID: 34172293 17. Katsiki N, Ntaios G, Vemmos K. **Stroke, obesity and gender: a review of the literature**. *Maturitas* (2011) **69** 239-243. PMID: 21602002 18. Park TH, Ko Y, Lee SJ. **Gender differences in the age-stratified prevalence of risk factors in Korean ischemic stroke patients: a nationwide stroke registry-based cross-sectional study**. *Int J Stroke* (2014) **9** 759-765. PMID: 24148312 19. Wyller TB. **Stroke and gender**. *J Gend Specif Med* (1999) **5** 21-25 20. Arisegi SA, Awosan KJ, Oche MO, Sabir AA, Ibrahim MT. **Knowledge and practices related to stroke prevention among hypertensive and diabetic patients attending Specialist Hospital, Sokoto, Nigeria**. *Pan Afr Med J* (2018) **29** 63. PMID: 29875944 21. Njoku CH, Aduloju AB. **Stroke in Sokoto, Nigeria: a five year retrospective study**. *Ann Afr Med* (2004) **3** 73-76 22. Ayanniyi O, Akande O, Mustapha AF. **Knowledge and perception of stroke among adults in Osogbo, Nigeria**. *Afr J Med Med Sci* (2006) **35** 447-452. PMID: 17722811 23. Jones SP, Jenkinson AJ, Leathley MJ, Watkins CL. **Stroke knowledge and awareness: an integrative review of the evidence**. *Age Ageing* (2010) **39** 11-22. PMID: 19897540 24. Sug Yoon S, Heller RF, Levi C, Wiggers J. **Knowledge and perception about stroke among an Australian urban population**. *BMC Public Health* (2001) **1** 14. PMID: 11734071 25. Donkor ES, Owolabi MO, Bampoh P, Aspelund T, Gudnason V. **Community awareness of stroke in Accra, Ghana**. *BMC Public Health* (2014) **14** 196. PMID: 24559414 26. Cossi MJ, Preux PM, Chabriat H, Gobron C, Houinato D. **Knowledge of stroke among an urban population in Cotonou (Benin)**. *Neuroepidemiology* (2012) **38** 172-178. PMID: 22472485 27. Wahab KW, Okokhere PO, Ugheoke AJ, Oziegbe O, Asalu AF, Salami TA. **Awareness of warning signs among suburban Nigerians at high risk for stroke is poor: a cross-sectional study**. *BMC Neurol* (2008) **8** 18. PMID: 18513399 28. Hickey A, O'Hanlon A, McGee H, Donnellan C, Shelley E, Horgan F, O'Neill D. **Stroke awareness in the general population: knowledge of stroke risk factors and warning signs in older adults**. *BMC Geriatr* (2009) **9** 35. PMID: 19656359 29. Alhazzani AA, Mahfouz AA, Abolyazid AY, Awadalla NJ, Ahmed RA, Siddiqui AF, Khalil SN. **Awareness of stroke among patients attending primary healthcare services in Abha, Southwestern Saudi Arabia**. *Neurosciences (Riyadh)* (2019) **24** 214-220. PMID: 31380821 30. Mersal FA, Tork HM. **Stroke risk perception and its awareness among hypertensive patients in Qassim region Saudi Arabia**. *Majmaah J Health Sci* (2020) **8** 9-22
--- title: 'Association of blood pressure with incident diabetic microvascular complications among diabetic patients: Longitudinal findings from the UK Biobank' authors: - Cong Li - Honghua Yu - Zhuoting Zhu - Xianwen Shang - Yu Huang - Charumathi Sabanayagam - Xiaohong Yang - Lei Liu journal: Journal of Global Health year: 2023 pmcid: PMC10039372 doi: 10.7189/jogh.13.04027 license: CC BY 4.0 --- # Association of blood pressure with incident diabetic microvascular complications among diabetic patients: Longitudinal findings from the UK Biobank ## Body Diabetic microvascular complications (DMCs) characterised by diabetic retinopathy (DR), diabetic kidney disease (DKD), and diabetic neuropathy (DN), are highly prevalent among populations with diabetes. Meta-analyses based on clinical trials suggested that lowering blood pressure (BP) can effectively reduce the risk of diabetes and its microvascular and macrovascular complications [1,2]. In the past few years, the UK Prospective Diabetes Study (UKPDS) showed that the incidence of clinical DMCs (predominantly retinal photocoagulation) was significantly associated with raised systolic BP (SBP) [3]. Moreover, several previous studies evaluated the associations between high BP variability and adverse microvascular outcomes in patients with type 2 diabetes [4-10]. However, most of them focused only on nephropathy [5-7,10], with inconclusive outcomes for retinopathy [8,9]. Furthermore, a long-term prospective cohort study addressing the influence of BP on all components of DMCs is still needed, as this is an essential strategy for disease prevention. Notably, compared to the phenotype, the genetic risk score (GRS) represents the congenital risk of a disease, which is less influenced by environmental or other systemic confounding factors. Hence, investigating a disease-trait association from both phenotypic and genotypic aspects is now widely accepted [11,12]. However, the association between BP GRSs and DMCs remains unclear. There is vast evidence on the role of hypertension in the occurrence and progression of DMCs [13,14]. In 2017, the American College of Cardiology (ACC) and the American Heart Association (AHA) released an updated guideline, which changed the definition of hypertension, lowering the cut-off for defining hypertension to SBP/diastolic blood pressure (DBP)≥$\frac{130}{80}$ mm Hg [15], while the upper end of pre-hypertension based on the seventh report of the Joint National Committee (JNC 7) [16] was reclassified as stage 1 hypertension. In contrast to the updated American guideline, the 2018 European Society of Cardiology (ESC)/European Society of Hypertension (ESH) BP guideline defined hypertension based on a threshold of ≥$\frac{140}{90}$ mm Hg [17], which is the same as for JNC 7. With the recent guidelines gradually becoming popular, exploring the impacts of stricter definitions for related disease prevalence, treatment, and control to reduce the disease burden worldwide is crucial. However, no report has determined the difference in DMCs incidence using the 2017 ACC/AHA rule, and compared this result with the JNC 7 rule. We used data from a large-scale UK Biobank population-based cohort over more than 10 years of follow-up, in order to examine the association between BP levels and incidence of DMCs, investigate the associations of BP GRSs with longitudinal DMCs, and determine any differences in incidence of DMCs in relation to hypertension, according to the JNC 7 and 2017 ACC/AHA guidelines. ## Abstract ### Background Evidence suggests a correlation of blood pressure (BP) level with presence of diabetic microvascular complications (DMCs), but the effect of BP on DMCs incidence is not well-established. We aimed to explore the associations between BP and DMCs (diabetic retinopathy, diabetic kidney disease, and diabetic neuropathy) risk in participants with diabetes. ### Methods This study included 23 030 participants, free of any DMCs at baseline, from the UK Biobank. We applied multivariable-adjusted Cox regression models to estimate BP-DMCs association and constructed BP genetic risk scores (GRSs) to test their association with DMCs phenotypes. Differences in incidences of DMCs were also compared between the 2017 ACC/AHA and JNC 7 guidelines (traditional criteria) of hypertension. ### Results Compared to systolic blood pressure (SBP)<120 mm Hg, participants with SBP≥160 mm Hg had a hazard ratio (HR) of 1.50 ($95\%$ confidence interval (CI) = 1.09, 2.06) for DMCs. Similarly, DMCs risk increased by $9\%$ for every 10 mm Hg of higher SBP at baseline ($95\%$ CI = 1.04, 1.13). The highest tercile SBP GRS was associated with $32\%$ higher DMCs risk ($95\%$ CI = 1.11, 1.56) compared to the lowest tercile. We found no significant differences in DMCs incidence between JNC 7 and 2017 ACC/AHA guidelines. ### Conclusions Genetic and epidemiological evidence suggests participants with higher SBP had an increased risk of DMCs, but hypertension defined by 2017 ACC/AHA guidelines may not impact DMCs incidence compared with JNC 7 criteria, contributing to the care and prevention of DMCs. ## Study design and population We used data from the UK Biobank, which is available from a public, open-access repository; its profile and detailed methods have been described elsewhere [18]. Briefly, a total of 502 628 participants (aged 40-69 years) from the general population were recruited between 2006 and 2010 at one of 22 assessment centres in Scotland, England, or Wales. We included 30 262 individuals with diabetes at recruitment. Diabetes cases were defined as those who had self-reported or doctor-diagnosed diabetes mellitus, were taking anti-hyperglycaemic medications or using insulin, or glycosylated haemoglobin (HbA1c)>48 mmol/mol [19]. Finally, we included data from 23 030 participants with diabetes in the main analysis after excluding participants with any type of DMCs at baseline [4199] or missing values on BP [3033] (Figure 1). **Figure 1:** *Flowchart for population selection from UK Biobank.* ## Blood pressure measurement Trained nurses measured BP (mmHg) twice at one-minute intervals using a digital sphygmomanometer (Omron 705 IT; OMRON Healthcare Europe B.V., Hoofddorp, Netherlands) after the participant had been at rest for at least five minutes in the seated position. We used the average of the two repeated measurements in the analysis. We diagnosed hypertension using two different guidelines: the JNC 7 (SBP/DBP≥$\frac{140}{90}$ mm Hg) and the 2017 ACC/AHA (SBP/DBP≥$\frac{130}{80}$ mm Hg) (Table S1 in the Online Supplementary Document). Furthermore, hypertension was classified into three categories: isolated diastolic hypertension (IDH) (SBP<140 mm Hg and DBP≥90 mm Hg for JNC 7; SBP<130 mm Hg and DBP≥80 mm Hg for 2017 ACC/AHA), isolated systolic hypertension (ISH) (SBP≥140 mm Hg and DBP<90 mm Hg for JNC 7; SBP≥130 mm Hg and DBP<80 mm Hg for 2017 ACC/AHA), and systolic-diastolic hypertension (SDH) (SBP≥140 mm Hg and DBP≥90 mm Hg for JNC 7; SBP≥130 mm Hg and DBP≥80 mm Hg for 2017 ACC/AHA) [20]. ## Ascertaining incident diabetic microvascular complication cases We identified the DMCs (DR, DKD, and DN) using the algorithms provided by UK Biobank, which were based on self-reported and data from electronic health records according to ICD-9 and ICD-10 codes (Table S2 in the Online Supplementary Document). Our analysis of DMCs incidence excluded individuals with any diagnosed or self-reported DMCs prior to the baseline assessment. We calculated the follow-up time as the duration between the date of baseline assessment and censored at the date of DMCs incidence, date of death, date lost to follow-up, or the end of follow-up, whichever occurred first. ## Data collection and assessment of covariates Extensive phenotypic and genotypic data were collected at recruitment. All participants completed touch-screen questionnaires collecting information on socio-demographics, habitual diet, lifestyle factors, and medical history, underwent physical examinations on anthropometric measurements, and provided biological samples including of blood, urine, and saliva. The confounders in our analysis included age, sex, body mass index, Townsend index (an area-based proxy measure for socioeconomic status), smoking status (recorded as current/previous and never), alcohol assumption (recorded as current/previous and never), HbA1c, duration of diabetes, use of anti-hyperglycaemic and anti-hypertension medication, estimated glomerular filtration rate (eGFR), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and triglycerides (TG). ## Genetic risk score for blood pressure BiLEVE Axiom array, or the UK Biobank Axiom array was used for genotyping by Affymetrix, and approximately 450 000 UK Biobank participants were genotyped. The genotype imputation using the Haplotype Reference Consortium reference panel was conducted by the UK Biobank researchers before data were released, followed by extensive quality control [21]. Genetic variants associated with BP were selected based on a recent genome-wide association study conducted among over one million people of European ancestry by Evangelou et al [22]. Following their method, a total number of 885 single nucleotide polymorphisms (SNPs) associated with BP were used for GRS calculation. For each SNP, the magnitude of its association (beta coefficient) with BP was used as the weighting factor and, for each participant, the dosage of the risk allele times the weight was calculated, and the sum across all the SNPs was considered as the GRS. In PLINK 2.0, the–score function was used for GRS generation. The above formula was used to generate GRS, where k is the number of independent genetic variants associated with BP, βi is the effect estimate, and *Ni is* the number of risk alleles for each locus. ## Ethics The UK Biobank study was previously given ethical approval by the National Information Governance Board for Health and Social Care and the NHS North West Multicentre Research Ethics Committee (11/NW/0382). All participants completed written informed consent before enrolment. This study was conducted under application number 62489 of the UK Biobank resource. ## Statistical analysis We reported the data as mean and standard deviation (SD) for normally distributed variables, median and interquartile range (IQR) for skewed variables, and number and percentage for categorical variables. We compared the baseline characteristics of participants from UK Biobank of groups with and without DMCs by Mann-Whitney or unpaired t-tests for continuous data, and Pearson χ2 or Fisher exact tests for categorical data. To indicate the relationship between BP and the risk of DMCs incidence, we modelled BP as restricted quadratic splines to provide a smooth, yet flexible description of the dose-response relationship. We estimated the hazard ratios (HRs) and $95\%$ confidence intervals (CIs) of DMCs among different levels and categories of BP by Cox proportional hazards model. We conducted subgroup analyses to test whether the association between BP and DMCs incidence differed between groups of sex, HbA1c, and use of anti-hyperglycaemic and anti-hypertensive medication. Furthermore, we tested how the BP GRSs were associated with DMCs risk by estimating HR and $95\%$ CIs across GRS terciles (i.e., top and bottom terciles represent high and low genetic risk, respectively); we examined linear trends using the GRS as a continuous variable. Additionally, we calculated the incidence rate with $95\%$ CI of diabetic adults with newly diagnosed DMCs based respectively on the 2017 ACC/AHA and the JNC 7 guidelines. We performed statistical analyses with Stata version 16 (StataCorp LLC, College Station, Texas USA) and R software (www.r-project.org, version 4.1.2). We calculated the P-value for trends for BP categories. We imputed missing data using multiple imputation by chained equations in Stata. All P values were two-sided with significance set at $P \leq 0.05.$ ## Baseline characteristics We included 23 030 individuals (median age 61.00 years, $40.08\%$ females) for the final analysis. Over the mean 11.95 (IQR = 11.16-12.71) years of the follow-up period, a total of 1060 ($4.60\%$; $95\%$ CI = $4.34\%$, $4.88\%$) diabetic individuals developed DMCs, of which 762 ($3.31\%$; $95\%$ CI = $3.08\%$, $3.55\%$) suffered from DR, 102 ($0.44\%$; $95\%$ CI = $0.36\%$, $0.54\%$) suffered from DKD, and 321 ($1.39\%$; $95\%$ CI = $1.25\%$, $1.55\%$) suffered from DN. The clinical characteristics of the participants with and without incident DMCs are shown in Table 1. Participants with incident DMCs were mostly older, of male sex, used anti-hyperglycaemic medication, and had higher BMI, higher Townsend index, longer diabetic duration, higher SBP, and higher HbA1c levels, but lower DBP, TC, LDL-C, and were more often current/former drinkers than participants without incident DMCs. **Table 1** | Variables | Total | With microvascular complications | Without microvascular complications | P-value* | | --- | --- | --- | --- | --- | | N | 23 030 | 1 060 | 21 970 | | | Age in years, median (IQR) | 61.00 (54.00, 65.00) | 62.00 (55.00, 66.00) | 61.00 (54.00, 65.00) | <0.001 | | Sex, n (%) | | | | 0.007 | | Females | 9230 (40.08) | 383 (36.13) | 8847 (40.27) | | | Males | 13 800 (59.92) | 677 (63.87) | 13 123 (59.73) | | | Ethnicity, n (%) | | | | 0.938 | | White | 19 883 (86.34) | 916 (86.42) | 18 967 (86.33) | | | Others | 3147 (13.66) | 144 (13.58) | 3003 (13.67) | | | BMI in kg/m2, mean (SD) | 31.32 (5.77) | 31.94 (5.93) | 31.29 (5.77) | <0.001 | | Townsend index, median (IQR) | -1.32 (-3.21, 1.93) | -0.80 (-3.07, 2.78) | -1.34 (-3.22, 1.88) | <0.001 | | Education level, n (%) | | | | 0.385 | | College or university degree | 17472 (75.87) | 816 (76.98) | 16 656 (75.81) | | | Others | 5558 (24.13) | 244 (23.02) | 5314 (24.19) | | | Smoking status, mean (%) | | | | 0.163 | | Never | 10 736 (46.62) | 472 (44.53) | 10 264 (46.72) | | | Former/current | 12 294 (53.38) | 588 (55.47) | 11 706 (53.28) | | | Alcohol consumption, mean (%) | | | | 0.017 | | Never | 2089 (9.07) | 118 (11.13) | 1971(8.97) | | | Former/current | 20 941 (90.93) | 942 (88.87) | 19 999 (91.03) | | | Duration of diabetes in years, median (IQR) | 5.00 (2.17, 9.00) | 8.00 (4.00, 14.00) | 5.00 (2.00, 8.80) | <0.001 | | SBP in mmHg, mean (SD) | 141.47 (17.18) | 143.43 (18.04) | 141.37 (17.13) | <0.001 | | DBP in mmHg, mean (SD) | 82.53 (9.34) | 81.57 (10.01) | 82.57 (9.31) | 0.001 | | HbA1c in mmol/mol, mean (SD) | 52.64 (14.70) | 60.22 (17.50) | 52.28 (14.40) | <0.001 | | TG in mmol/L, median (IQR) | 1.89 (1.31, 2.70) | 1.94 (1.32, 2.75) | 1.88 (1.31, 2.70) | 0.294 | | TC in mmol/L, mean (SD) | 4.71 (1.14) | 4.52 (1.09) | 4.72 (1.15) | <0.001 | | LDL-C in mmol/L, mean (SD) | 2.86 (0.86) | 2.71 (0.80) | 2.86 (0.85) | <0.001 | | HDL-C in mmol/L, mean (SD) | 1.21 (0.32) | 1.19 (0.35) | 1.21 (0.32) | 0.107 | | eGFR in mL/(min ×1.73 m2), median (IQR) | 93.78 (84.15, 100.76) | 93.30 (83.03, 100.95) | 93.81 (84.20, 100.75) | 0.370 | | Anti-hyperglycaemic medication, n (%) | | | | <0.001 | | No | 9724 (42.22) | 188 (17.74) | 9536 (43.40) | | | Yes | 13 306 (57.78) | 872 (82.26) | 12 434 (56.60) | | | Anti-hypertension medication, n (%) | | | | 0.354 | | No | 18 249 (79.24) | 828 (78.11) | 17 421 (79.29) | | | Yes | 4781 (20.67) | 232 (21.89) | 4549 (20.71) | | ## Blood pressure and diabetic microvascular complications The frequency distribution of DMCs for different BP levels is shown in Table 2. The trends for incidence of overall DMCs, DR, and DKD significantly rose with increasing SBP, but not for individuals with DN. With the increase of DBP, the change trends for incidence of overall DMCs and DR were significant, but not for DKD and DN. **Table 2** | Blood pressure level (mmHg) | Overall | Overall.1 | DR | DR.1 | DKD | DKD.1 | DN | DN.1 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | No | Incident (95% CI) | No | Incident (95% CI) | No | Incident (95% CI) | No | Incident (95% CI) | | Systolic BP | | | | | | | | | | <120 | 83 | 4.05 (3.24, 5.00) | 54 | 2.64 (1.99, 3.43) | 9 | 0.44 (0.20, 0.83) | 28 | 1.37 (0.91, 1.97) | | 120-129 | 155 | 4.05 (3.45, 4.72) | 111 | 2.90 (2.39, 3.48) | 8 | 0.21 (0.09, 0.41) | 51 | 1.33 (0.99, 1.75) | | 130-139 | 247 | 4.60 (4.05, 5.19) | 162 | 3.01 (2.57, 3.51) | 18 | 0.33 (0.24, 0.61) | 89 | 1.66 (1.33, 2.03) | | 140-149 | 215 | 4.28 (3.74, 4.88) | 161 | 3.20 (2.74, 3.73) | 20 | 0.40 (0.24, 0.61) | 59 | 1.17 (0.90, 1.51) | | 150-159 | 182 | 5.17 (4.47, 5.96) | 142 | 4.04 (3.42, 4.82) | 19 | 0.54 (0.33, 0.84) | 48 | 1.36 (1.01, 1.81) | | ≥160 | 178 | 5.50 (4.74, 6.34) | 132 | 4.08 (3.42, 4.82) | 28 | 0.86 (0.58, 1.25) | 46 | 1.42 (1.04, 1.89) | | P-value for trend* | 0.001 | 0.001 | <0.001 | <0.001 | <0.001 | <0.001 | 0.794 | 0.794 | | Diastolic BP | | | | | | | | | | <80 | 474 | 5.18 (4.73, 5.65) | 346 | 3.78 (3.40, 4.19) | 43 | 0.47 (0.34, 0.63) | 137 | 1.50 (1.26, 1.77) | | 80-89 | 385 | 4.38 (3.96, 4.83) | 269 | 3.06 (2.71, 3.44) | 36 | 0.41 (0.29, 0.57) | 122 | 1.39 (1.15, 1.66) | | 90-99 | 158 | 3.77 (3.21, 4.39) | 113 | 2.70 (2.23, 3.23) | 17 | 0.41 (0.24, 0.65) | 50 | 1.19 (0.89, 1.57) | | ≥100 | 43 | 4.81 (3.50, 6.42) | 34 | 3.80 (2.65, 5.27) | 6 | 0.67 (0.25, 1.46) | 12 | 1.34 (0.70, 2.33) | | P-value for trend* | 0.002 | 0.002 | 0.009 | 0.009 | 0.993 | 0.993 | 0.214 | 0.214 | Cox proportional hazards models were used to investigate the relationship between BP and incident DMCs (Table 3). Compared to participants with SBP<120 mm Hg, the HR for DMCs was 1.50 ($95\%$ CI = 1.09, 2.06) for participants with SBP≥160 mm Hg, after adjusting for confounding variables. For DBP, it was not significantly associated with DMCs incidence after adjusting for confounders (all $P \leq 0.05$). Similarly, further analysis of BP as a continuous variable suggested that the HR for DMCs was 1.09 ($95\%$ CI = 1.04, 1.13) with every 10 mm Hg higher SBP at baseline after adjusting for covariates, but the relationship was not significant regarding the increase of DBP. **Table 3** | Blood pressure level (mmHg) | Crude HR (95% CI) | P-value† | Adjusted HR (95% CI) | P-value†.1 | | --- | --- | --- | --- | --- | | Systolic blood pressure | | | | | | <120 | Reference | | Reference | | | 120-129 | 0.99 (0.76, 1.29) | 0.942 | 1.01 (0.73, 1.38) | 0.953 | | 130-139 | 1.12 (0.88, 1.44) | 0.364 | 1.21 (0.90, 1.63) | 0.217 | | 140-149 | 1.04 (0.81, 1.34) | 0.741 | 1.11 (0.82, 1.51) | 0.496 | | 150-159 | 1.28 (0.99, 1.66) | 0.064 | 1.33 (0.97, 1.83) | 0.074 | | ≥160 | 1.36 (1.05, 1.77) | 0.020 | 1.50 (1.09, 2.06) | 0.012 | | P-value for trend† | 0.001 | 0.001 | 0.001 | 0.001 | | Per 10 mm Hg higher at baseline | 1.07 (1.03, 1.11) | <0.001 | 1.09 (1.04, 1.13) | <0.001 | | Diastolic blood pressure | | | | | | <80 | Reference | | Reference | | | 80-89 | 0.83 (0.73, 0.95) | 0.008 | 0.90 (0.77, 1.05) | 0.164 | | 90-99 | 0.71 (0.59, 0.85) | <0.001 | 0.84 (0.68, 1.03) | 0.089 | | ≥100 | 0.92 (0.67, 1.25) | 0.578 | 1.34 (0.96, 1.88) | 0.085 | | P-value for trend† | 0.001 | 0.001 | 0.600 | 0.600 | | Per 10 mm Hg higher at baseline | 0.88 (0.83, 0.94) | <0.001 | 0.97 (0.90, 1.05) | 0.453 | We further used Cox proportional hazards models to determine the association between BP and incidence of DR, DKD, and DN (Table S3-S5 in the Online Supplementary Document, respectively). After adjusting for confounding factors, the HRs for DR were 1.73 ($95\%$ CI = 1.17, 2.56) for participants with SBP 150-159 and 1.84 ($95\%$ CI = 1.23, 2.73) for those with ≥160 mm Hg, compared with SBP<120 mm Hg. Multivariable analysis showed a higher risk of DR incidence for participants with DBP≥100 mm Hg compared with <80 mm Hg (HR = 1.60; $95\%$ CI = 1.09, 2.35). For BP as a continuous variable, the HRs for DR and DKD were 1.12 ($95\%$ CI = 1.06, 1.17) and 1.31 ($95\%$ CI = 1.17, 1.47), respectively, with every 10 mm Hg higher SBP at baseline after adjusting for confounders, but the association was not significant for the increase of DBP. The SBP was associated with the risk of DMCs with a nonlinear dose-response relationship (Figure 2, panels A-D). Dose-response relationships between DBP and DMCs incidence among diabetes participants showed approximately J-shaped curves (Figure 2, panels E-H), indicating that the effect of DBP on DMCs incidence also tended to be nonlinear. **Figure 2:** *Dose-response relationship between (A-D) SBP and (E-H) DBP and diabetic microvascular complications among patients with diabetes. Panel A. Dose-response relationship between SBP and overall diabetic microvascular complications. Panel B. Dose-response relationship between SBP and DR. Panel C. Dose-response relationship between SBP and DKD. Panel D. Dose-response relationship between SBP and DN. Panel E. Dose-response relationship between DBP and overall diabetic microvascular complications. Panel F. Dose-response relationship between DBP and DRD Panel G. dose-response relationship between DBP and DKD. Panel H. Dose-response relationship between DBP and DN. DR – diabetic retinopathy, DKD – diabetic kidney disease, DN – diabetic neuropathy, SBP – systolic blood pressure, DBP – diastolic blood pressure, HR – hazard ratio, CI – confidence interval.* ## Blood pressure GRS and diabetic microvascular complications After adjusting for confounders, those with high genetically defined SBP (estimated mean SBP = 160 mm Hg) among the UK population showed $32\%$ higher risk of DMCs incidence ($95\%$ CI = 1.11, 1.56) than participants at low genetic risk. With the increase of SBP genetic risk, the trends for DMCs incidence significantly increased. Similar results were found for DR and DKD incidence. However, no relationship between SBP GRS and DMCs was found when the GRS for SBP was considered a continuous variable (Table 4). The DBP GRS did not show a significant association with DMCs (data not shown). **Table 4** | Systolic blood pressure GRS | Crude HR (95% CI) | P-value† | Adjusted HR (95% CI) | P-value†.1 | | --- | --- | --- | --- | --- | | DMCs | | | | | | Low GRS | Reference | | Reference | | | Intermediate GRS | 1.12 (0.98, 1.31) | 0.146 | 1.10 (0.92, 1.30) | 0.300 | | High GRS | 1.28 (1.11, 1.50) | 0.001 | 1.32 (1.11, 1.56) | 0.002 | | P-value for trend† | 0.001 | 0.001 | 0.002 | 0.002 | | Per 10 mm Hg higher | 1.00 (0.85, 1.18) | 0.989 | 0.94 (0.78, 1.12) | 0.489 | | DR | | | | | | Low GRS | Reference | | Reference | | | Intermediate GRS | 1.15 (0.95, 1.38) | 0.151 | 1.17 (0.95, 1.44) | 0.143 | | High GRS | 1.43 (1.20, 1.71) | <0.001 | 1.48 (1.21, 1.82) | <0.001 | | P-value for trend† | <0.001 | | <0.001 | | | Per 10 mm Hg higher | 0.98 (0.81, 1.19) | 0.876 | 0.94 (0.76, 1.17) | 0.574 | | DKD | | | | | | Low GRS | Reference | | Reference | | | Intermediate GRS | 1.42 (0.80, 2.54) | 0.231 | 1.35 (0.71, 2.55) | 0.357 | | High GRS | 2.51 (1.49, 4.23) | 0.001 | 2.80 (1.57, 4.98) | <0.001 | | P-value for trend† | <0.001 | | <0.001 | | | Per 10 mm Hg higher | 1.99 (1.18, 3.37) | 0.010 | 2.47 (1.39, 4.40) | 0.002 | | DN | | | | | | Low GRS | Reference | | Reference | | | Intermediate GRS | 0.98 (0.75, 1.30) | 0.914 | 0.94 (0.70, 1.28) | 0.705 | | High GRS | 0.97 (0.74, 1.28) | 0.849 | 1.04 (0.77, 1.42) | 0.777 | | P-value for trend† | 0.849 | | 0.776 | | | Per 10 mm Hg higher | 1.04 (0.77, 1.39) | 0.819 | 0.90 (0.65, 1.25) | 0.537 | ## Blood pressure classification and diabetic microvascular complications The characteristics of the participants with diabetes and the DMCs incidence for JNC 7 and 2017 ACC/AHA guidelines for hypertension are presented in Table 5. The DMCs incidences significantly differed among diabetes participants of different age group, sex, BMI level, HbA1c level, TC level, and drinking status (all $P \leq 0.05$). There was no significant difference in DMCs incidence among participants defined by the JNC 7 or the 2017 ACC/AHA guidelines. **Table 5** | Variables | Overall diabetes | Overall diabetes.1 | JNC 7 defined hypertension | JNC 7 defined hypertension.1 | 2017 ACC/AHA defined hypertension | 2017 ACC/AHA defined hypertension.1 | Difference, % (95% CI) | P-value | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | No | Incident (95% CI) | No | Incident (95% CI) | No | Incident (95% CI) | | | | Age in years | | | | | | | | | | 40-44 | 39 | 3.45 (2.47, 4.69) | 21 | 5.20 (3.25, 7.84) | 33 | 4.39 (3.04, 6.12) | -0.81 (-3.40, 1.81) | 0.537 | | 45-54 | 197 | 4.11 (3.57, 4.72) | 103 | 4.71 (3.86, 5.68) | 163 | 4.38 (3.74, 5.09) | -0.33 (-1.43, 0.78) | 0.559 | | 55-64 | 476 | 4.44 (4.06, 4.85) | 278 | 4.64 (4.12, 5.21) | 392 | 4.44 (3.74, 5.09) | -0.20 (-0.89, 0.48) | 0.56 | | ≥65 | 348 | 5.44 (4.90, 6.03) | 207 | 5.16 (4.49, 5.89) | 293 | 5.33 (4.75, 5.96) | 0.17 (-0.73, 1.08) | 0.707 | | P-value* | 0.001 | 0.001 | | | | | | | | Sex | | | | | | | | | | Female | 383 | 4.15 (3.75, 4.58) | 209 | 4.48 (3.91, 5.12) | 318 | 4.39 (3.93, 4.89) | -0.09 (-0.85, 0.67) | 0.817 | | Male | 677 | 4.91 (4.55, 5.28) | 400 | 5.04 (4.57, 5.55) | 563 | 4.87 (4.48, 5.28) | -0.17 (-0.79, 0.45) | 0.586 | | P-value* | 0.007 | 0.007 | | | | | | | | Ethnicity | | | | | | | | | | White | 916 | 4.61 (4.32, 4.91) | 524 | 4.74 (4.35, 5.15) | 763 | 4.67 (4.35, 5.00) | -0.07 (-0.58, 0.44) | 0.782 | | Others | 144 | 4.58 (3.87, 5.37) | 85 | 5.51 (4.42, 6.77) | 118 | 4.80 (3.99, 5.73) | -0.71 (-2.12, 0.71) | 0.323 | | P-value* | 0.938 | 0.938 | | | | | | | | BMI in kg/m2 | | | | | | | | | | <25 | 106 | 4.09 (3.36, 4.92) | 51 | 4.63 (3.46, 6.04) | 80 | 4.49 (3.58, 5.56) | -0.14 (-1.71, 1.43) | 0.862 | | 25-29.9 | 334 | 4.20 (3.77, 4.66) | 190 | 4.37 (3.79, 5.03) | 277 | 4.31 (3.83, 4.84) | -0.06 (-0.85, 0.73) | 0.88 | | ≥30 | 601 | 4.89 (4.52, 5.29) | 358 | 5.07 (4.57, 5.61) | 507 | 4.86 (4.45, 5.29) | -0.21 (-0.87, 0.44) | 0.521 | | P-value* | 0.033 | 0.033 | | | | | | | | HbA1c in mmol/mol | | | | | | | | | | <48 | 242 | 2.86 (2.51, 3.24) | 130 | 3.00 (2.51, 3.55) | 192 | 2.90 (2.51, 3.33) | -0.10 (-0.75, 0.55) | 0.756 | | ≥48 | 818 | 5.62 (5.25, 6.00) | 479 | 5.80 (5.30, 6.32) | 689 | 5.66 (5.26, 6.08) | -0.14 (-0.79, 0.51) | 0.679 | | P-value* | <0.001 | <0.001 | | | | | | | | TG in mmol/L | | | | | | | | | | <2.3 | 607 | 4.40 (4.06, 4.75) | 347 | 4.83 (4.35, 5.35) | 505 | 4.59 (4.21, 5.00) | -0.24 (-0.87, 0.40) | 0.461 | | ≥2.3 | 453 | 4.91 (4.48, 5.37) | 262 | 4.84 (4.28, 5.45) | 376 | 4.82 (4.35, 5.32) | -0.02 (-0.77, 0.72) | 0.949 | | P-value* | 0.070 | 0.070 | | | | | | | | TC in mmol/L | | | | | | | | | | <5.17 | 735 | 4.87 (4.53, 5.23) | 413 | 5.23 (4.75, 5.75) | 607 | 5.02 (4.64, 5.43) | -0.21 (-0.84, 0.42) | 0.51 | | ≥5.17 | 325 | 4.09 (3.67, 4.55) | 196 | 4.17 (3.62, 4.78) | 274 | 4.08 (3.62, 4.58) | -0.09 (-0.83, 0.66) | 0.819 | | P-value* | 0.007 | 0.007 | | | | | | | | LDL-C, mmol/L | | | | | | | | | | <4.1 | 892 | 4.59 (4.30, 4.90) | 514 | 4.94 (4.53, 5.37) | 741 | 4.71 (4.39, 5.06) | -0.23 (-0.76, 0.31) | 0.406 | | ≥4.1 | 168 | 4.65 (3.99, 5.39) | 95 | 4.35 (3.53, 5.29) | 140 | 4.55 (3.84, 5.35) | 0.20 (-0.92, 1.33) | 0.724 | | P-value* | 0.883 | 0.883 | | | | | | | | HDL-C, mmol/L | | | | | | | | | | <1.0 | 254 | 4.55 (4.25, 4.87) | 143 | 5.07 (4.29, 5.95) | 202 | 4.69 (4.08, 5.36) | -0.38 (-1.41, 0.64) | 0.459 | | ≥1.0 | 806 | 4.76 (4.21, 5.37) | 466 | 4.77 (4.35, 5.21) | 679 | 4.69 (4.35, 5.04) | -0.08 (-0.63, 0.46) | 0.772 | | P-value* | 0.527 | 0.527 | | | | | | | | Smoking status | | | | | | | | | | Never | 472 | 4.40 (4.02, 4.80) | 267 | 4.66 (4.13, 5.24) | 408 | 4.69 (4.25, 5.16) | 0.03 (-0.68, 0.73) | 0.937 | | Former/current | 588 | 4.78 (4.41, 5.17) | 342 | 4.98 (4.48, 5.52) | 473 | 4.68 (4.28, 5.11) | -0.30 (-0.96, 0.36) | 0.376 | | P-value* | 0.163 | 0.163 | | | | | | | | Alcohol consumption | | | | | | | | | | Never | 118 | 5.65 (4.70, 6.73) | 65 | 6.11 (4.75, 7.72) | 95 | 5.78 (4.70, 7.02) | -0.33 (-2.16, 1.50) | 0.722 | | Former/current | 942 | 4.50 (4.22, 4.79) | 544 | 4.72 (4.34, 5.12) | 786 | 4.58 (4.27, 4.91) | -0.14 (-0.63, 0.36) | 0.592 | | P-value* | 0.017* | 0.017* | | | | | | | We further explored the association between BP classification defined by the JNC7 and 2017 ACC/AHA guidelines and incidence of DMCs, DR, DKD, and DN using Cox proportional hazards models (Tables S6-S9 in the Online Supplementary Document and Figure 3, panels A-H). Participants with higher BP classification had a higher risk of incidence of DMCs and DR, compared with normal stage. With the increase in BP classification, the risk trends for incidence of DR and DKD significantly increased, but not for incidence of DN. For the different categories of hypertension, participants with ISH had higher risk of incidence of DMCs, DR, and DKD. **Figure 3:** *Incidence rate and Cox proportional hazards models for (A-B) DMCs, (C-D) DR, (E, F) DKD, and (G, H) DN among different blood pressure categories defined by JNC7 (A, C, E, G) and 2017 ACC/AHA (B, D, F, H) guidelines. Panel A. Incidence rate and Cox proportional hazards models for DMCs among different blood pressure categories defined by JNC7. Panel B. Incidence rate and Cox proportional hazards models for DMCs among different blood pressure categories defined by 2017 ACC/AHA. Panel C. Incidence rate and Cox proportional hazards models for DR among different blood pressure categories defined by JNC7. Panel D. Incidence rate and Cox proportional hazards models for DR among different blood pressure categories defined by 2017 ACC/AHA. Panel E. Incidence rate and Cox proportional hazards models for DKD among different blood pressure categories defined by JNC7. Panel F. Incidence rate and Cox proportional hazards models for DKD among different blood pressure categories defined by 2017 ACC/AHA. Panel G. Incidence rate and Cox proportional hazards models for DN among different blood pressure categories defined by JNC7. Panel H. Incidence rate and Cox proportional hazards models for DN among different blood pressure categories defined by 2017 ACC/AHA; Adjusted for age, sex, body mass index, ethnicity, Townsend index, smoking status, alcohol consumption, glycosylated haemoglobin, low-density lipoprotein cholesterol, duration of diabetes, anti-hyperglycaemic medication, anti-hypertensive medication, and estimated glomerular filtration rate. DMCs – diabetic microvascular complications, DR – diabetic retinopathy, DKD – diabetic kidney disease, DN – diabetic neuropathy, IDH – isolated diastolic hypertension, ISH – isolated systolic hypertension, SDH – systolic-diastolic hypertension.* Additionally, we conducted subgroup analyses stratified by sex, HbA1c level, and use of anti-hyperglycaemic and anti-hypertensive medication to examine the relationship between BP and DMCs incidence using multivariable Cox models (Figure S1-S5 in the Online Supplementary Document). In the female subgroup, participants with SBP 130-139, 150-159, and ≥160 mm Hg had higher risk of DMCs, compared to SBP<120 mm Hg. In the subgroup of HbA1c ≥48 mmol/mol, participants with SBP 150-159 and ≥160 mm Hg had higher risk of DMCs. No significant difference was observed for subgroup of HbA1c <48 mmol/mol. For the subgroup of using medication, the association between SBP and risk of DMCs incidence was only found in the participants with anti-hyperglycaemic or anti-hypertensive medication separately or combine. ## Sensitivity analysis To explore the cumulative effect of different types of DMCs, we compared the risk of incidence of single and multiple DMCs with the increase in BP (Figure S6 in the Online Supplementary Document). The associations of SBP were stronger for those with two or more DMCs compared with those with only one DMCs. Considering that insulin therapy was an independent risk factor for diabetic complications [23], and patients with insulin use are usually sicker and develop complications earlier, we examined the association of different BP levels and BP stage by adjusting for use of insulin rather than of anti-hyperglycaemic medication (Table S10 in the Online Supplementary Document). However, the results were similar to those gained after adjusting for anti-hyperglycaemic medication. Additionally, a similar pathological basis existed in DKD and DR, which differed from DN [24]. Therefore, we further tested the association between BP and DMCs incidence after excluding patients with DN (Tables S11 and S12 in the Online Supplementary Document). The exclusion had little influence on the above associations. ## DISCUSSION In this large cohort study, higher levels of SBP were significantly associated with an increased risk of DMCs among patients with diabetes in a dose-response relationship. Moreover, there was no difference in DMCs incidence using either the newly established 2017 AHA/ACC or the JNC 7 guidelines. Although a relationship between lowering BP and BP variability and the risk of DN has been reported among diabetic populations in several prospective studies [3-5,7,10], retinopathic and neuropathic evidence is either inconsistent or limited [7,8]. In our study, a higher level of SBP was significantly associated with incidence of DMCs, DR, and DKD. Notably, the relationship between BP and DR incidence is controversial. Some studies revealed that higher BP was associated with development of DR [8,25-28], while others found no significant association between BP and DR [9,29-31]. Some possible mechanisms have been proposed by which hypertension could affect DR via haemodynamic changes and vascular endothelial growth factor-dependent pathways [32]. To minimise the influence of confounders, we constructed the GRS for BP, and these results further supported the significant relationship between SBP and DMCs and DR. The dose-response analysis showed nonlinear effects of BP on incidence of DMCs. Furthermore, the more intensive BP control in patients with hypertension and diabetes has benefits in prevention and treatment of DMCs [33]. Therefore, exploring the BP-controlling targets can help in increasing benefits and reducing microvascular damage from diabetes. Additionally, the effectiveness of glycaemic control and BP management to decrease DMCs remains controversial. The UKPDS showed that, after adjusting confounders, intensive BP control could significantly reduce the risk of DMCs [13]. In contrast, the Appropriate Blood Pressure Control in Diabetes Trial demonstrated no difference in DMCs progression over five years between the intensive and moderate BP control groups [33]. A systematic review of 15 randomised trials supported the hypothesis that BP control could reduce the incidence, but not the progression of DR [34]. Similarly, some studies have suggested that intensive glycaemic control reduced DN incidence [35,36], while others found no significant decrease in DN incidence after intensive glycaemic control for type 2 diabetes mellitus [37]. Our subgroup analyses showed that the association between BP and DMCs was more evident in patients using anti-hyperglycaemic or anti-hypertensive medication or a combination, which might result from the severity of the underlying disease rather than the consequence of treatment status. Further studies should explore the efficacy of BP and glycaemic control and the extent of reduction needed for beneficial effects. Moreover, we found that the relationship between SBP and DMCs was significant only in females. Previous studies suggested that the risk for vascular complications appear to be greater for diabetic females than males; the genetic, sex hormones, and sex-specific risk factors might explain the sex differences of DMCs [38,39]. Nonetheless, the reasons for this different impact between sex remain unclear and deserve further investigation. According to the 2017 AHA/ACC guideline, lowering the cut-off to define hypertension may increase the number of populations with hypertension, and is estimated to classify approximately $46\%$ of the US adult population as having hypertension [40]. However, whether the stricter definition will affect the incidence of DMCs in relation to hypertension remains unclear. Our results suggested that the new guideline on hypertension may not affect the incidence of DMCs in the UK population with diabetes. Future studies on other racial populations are needed to verify this impact. Our study showed that diabetic patients with higher SBP levels had a higher risk of DMCs. This transition requires clinical and preventative strategies for BP control among patients with diabetes. If confirmed by replication, our findings may have implications for DMCs prevention strategies that target improving and maintaining BP measurements among patients with diabetes. They also contribute to the scientific basis for the development of intervention studies for future DMCs prevention among patients with diabetes. A major strength of our study is that the UK Biobank study is a prospective long-term cohort study which has collected extensive phenotypic and genotypic data. Therefore, we made meticulous adjustments for a wide range of potential confounding factors. However, our study also has several potential limitations. First, the ICD code may be insufficient for detecting cases in an early stage or to classifying DR, DKD, and DN case subtypes. Second, we used the baseline measurement of BP levels in the UK Biobank, so we did not capture changes in BP levels during follow-up, which may lead to non-differential misclassification bias. Third, self-reported lifestyle factors and medical history data were subject to measurement error, which may lead to misclassification bias. Fourth, our study was limited to the UK population with diabetes who were aged 40-69 years, thus, our findings may not be directly generalisable to other populations. Further studies with a wider age range are needed. Additionally, the sample sizes for DKD and DN were relatively small, which might have limited power for non-significant associations. Finally, six different hypertension management guidelines have been published by American (JNC7, 2017 ACC/AHA), European ($\frac{2013}{2018}$ ESH/ESC), and international organisations (WHO/International Society of Hypertension (ISH) 2003, ISH 2020); we only compared the DMCs differences between JNC7 and 2017 AHA/ACC, because the 2017 AHA/ACC is the only one that does not define hypertension as $\frac{140}{90}$ mm Hg. ## CONCLUSIONS In this long-term follow-up, large-scale, prospective cohort study of a population with diabetes, both genetic and epidemiological evidence suggested that higher levels of SBP were significantly associated with an increased risk of DMCs. Furthermore, high BP, defined as SBP/DBP of at least $\frac{130}{80}$ mm Hg using the 2017 AHA/ACC guidelines, did not influence the incidence of DMCs. The findings suggest that BP control among patients with diabetes may be useful in DMCs care and prevention. ## References 1. Nazarzadeh M, Bidel Z, Canoy D, Copland E, Wamil M, Majert J. **Blood pressure lowering and risk of new-onset type 2 diabetes: an individual participant data meta-analysis.**. *Lancet* (2021) **398** 1803-10. DOI: 10.1016/S0140-6736(21)01920-6 2. Emdin CA, Rahimi K, Neal B, Callender T, Perkovic V, Patel A. **Blood pressure lowering in type 2 diabetes: a systematic review and meta-analysis.**. *JAMA* (2015) **313** 603-15. DOI: 10.1001/jama.2014.18574 3. Adler AI, Stratton IM, Neil HA, Yudkin JS, Matthews DR, Cull CA. **Association of systolic blood pressure with macrovascular and microvascular complications of type 2 diabetes (UKPDS 36): prospective observational study.**. *BMJ* (2000) **321** 412-9. DOI: 10.1136/bmj.321.7258.412 4. Hata J, Arima H, Rothwell PM, Woodward M, Zoungas S, Anderson C. **Effects of visit-to-visit variability in systolic blood pressure on macrovascular and microvascular complications in patients with type 2 diabetes mellitus: the ADVANCE trial.**. *Circulation* (2013) **128** 1325-34. DOI: 10.1161/CIRCULATIONAHA.113.002717 5. Okada H, Fukui M, Tanaka M, Matsumoto S, Mineoka Y, Nakanishi N. **Visit-to-visit blood pressure variability is a novel risk factor for the development and progression of diabetic nephropathy in patients with type 2 diabetes.**. *Diabetes Care* (2013) **36** 1908-12. DOI: 10.2337/dc12-2087 6. McMullan CJ, Lambers Heerspink HJ, Parving H-H, Dwyer JP, Forman JP, de Zeeuw D. **Visit-to-visit variability in blood pressure and kidney and cardiovascular outcomes in patients with type 2 diabetes and nephropathy: a post hoc analysis from the RENAAL study and the Irbesartan Diabetic Nephropathy Trial.**. *Am J Kidney Dis* (2014) **64** 714-22. DOI: 10.1053/j.ajkd.2014.06.008 7. Noshad S, Mousavizadeh M, Mozafari M, Nakhjavani M, Esteghamati A. **Visit-to-visit blood pressure variability is related to albuminuria variability and progression in patients with type 2 diabetes.**. *J Hum Hypertens* (2014) **28** 37-43. DOI: 10.1038/jhh.2013.36 8. Sohn M-W, Epstein N, Huang ES, Huo Z, Emanuele N, Stukenborg G. **Visit-to-visit systolic blood pressure variability and microvascular complications among patients with diabetes.**. *J Diabetes Complications* (2017) **31** 195-201. DOI: 10.1016/j.jdiacomp.2016.09.003 9. Takao T, Matsuyama Y, Yanagisawa H, Kikuchi M, Kawazu S. **Visit-to-visit variability in systolic blood pressure predicts development and progression of diabetic nephropathy, but not retinopathy, in patients with type 2 diabetes.**. *J Diabetes Complications* (2014) **28** 185-90. DOI: 10.1016/j.jdiacomp.2013.11.003 10. Viazzi F, Bonino B, Mirijello A, Fioretto P, Giorda C, Ceriello A. **Long-term blood pressure variability and development of chronic kidney disease in type 2 diabetes.**. *J Hypertens* (2019) **37** 805-13. DOI: 10.1097/HJH.0000000000001950 11. Nierenberg JL, Anderson AH, He J, Parsa A, Srivastava A, Cohen JB. **Association of Blood Pressure Genetic Risk Score with Cardiovascular Disease and CKD Progression: Findings from the CRIC Study.**. *Kidney360* (2021) **2** 1251-60. DOI: 10.34067/KID.0007632020 12. Cho YA, Lee J, Oh JH, Chang HJ, Sohn DK, Shin A. **Genetic Risk Score, Combined Lifestyle Factors and Risk of Colorectal Cancer.**. *Cancer Res Treat* (2019) **51** 1033-40. DOI: 10.4143/crt.2018.447 13. **Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes: UKPDS 38. UK Prospective Diabetes Study Group.**. *BMJ* (1998) **317** 703-13. DOI: 10.1136/bmj.317.7160.703 14. Yamazaki D, Hitomi H, Nishiyama A. **Hypertension with diabetes mellitus complications.**. *Hypertens Res* (2018) **41** 147-56. DOI: 10.1038/s41440-017-0008-y 15. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: Executive Summary: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines.**. *Hypertension* (2018) **71** 1269-324. DOI: 10.1161/HYP.0000000000000066 16. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL. **Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure.**. *Hypertension* (2003) **42** 1206-52. DOI: 10.1161/01.HYP.0000107251.49515.c2 17. Williams B, Mancia G, Spiering W, Agabiti Rosei E, Azizi M, Burnier M. **2018 ESC/ESH Guidelines for the management of arterial hypertension.**. *Eur Heart J* (2018) **39** 3021-104. DOI: 10.1093/eurheartj/ehy339 18. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J. **UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age.**. *PLoS Med* (2015) **12**. DOI: 10.1371/journal.pmed.1001779 19. Zhu Z, Shi D, Liao H, Ha J, Shang X, Huang Y. **Visual Impairment and Risk of Dementia: The UK Biobank Study.**. *Am J Ophthalmol* (2022) **235** 7-14. DOI: 10.1016/j.ajo.2021.08.010 20. Cesena FHY, Nary FC, Santos RD, Bittencourt MS. **The contribution of the systolic and diastolic components for the diagnosis of arterial hypertension under the 2017 ACC/AHA Guideline and metabolic heterogeneity among individuals with Stage 1 hypertension.**. *J Clin Hypertens (Greenwich)* (2020) **22** 1192-9. DOI: 10.1111/jch.13907 21. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K. **The UK Biobank resource with deep phenotyping and genomic data.**. *Nature* (2018) **562** 203-9. DOI: 10.1038/s41586-018-0579-z 22. Evangelou E, Warren HR, Mosen-Ansorena D, Mifsud B, Pazoki R, Gao H. **Genetic analysis of over 1 million people identifies 535 new loci associated with blood pressure traits.**. *Nat Genet* (2018) **50** 1412-25. DOI: 10.1038/s41588-018-0205-x 23. Gurung RL, FitzGerald LM, Liu E, McComish BJ, Kaidonis G, Ridge B. **The effect of insulin on response to intravitreal anti-VEGF injection in diabetic macular edema in type 2 diabetes mellitus.**. *BMC Ophthalmol* (2022) **22** 94. DOI: 10.1186/s12886-022-02325-x 24. Saini DC, Kochar A, Poonia R. **Clinical correlation of diabetic retinopathy with nephropathy and neuropathy.**. *Indian J Ophthalmol* (2021) **69** 3364-8. DOI: 10.4103/ijo.IJO_1237_21 25. Leske MC, Wu S-Y, Hennis A, Hyman L, Nemesure B, Yang L. **Hyperglycemia, blood pressure, and the 9-year incidence of diabetic retinopathy: the Barbados Eye Studies.**. *Ophthalmology* (2005) **112** 799-805. DOI: 10.1016/j.ophtha.2004.11.054 26. Knowler WC, Bennett PH, Ballintine EJ. **Increased incidence of retinopathy in diabetics with elevated blood pressure. A six-year follow-up study in Pima Indians.**. *N Engl J Med* (1980) **302** 645-50. DOI: 10.1056/NEJM198003203021201 27. Teuscher A, Schnell H, Wilson PW. **Incidence of diabetic retinopathy and relationship to baseline plasma glucose and blood pressure.**. *Diabetes Care* (1988) **11** 246-51. DOI: 10.2337/diacare.11.3.246 28. Liu L, Quang ND, Banu R, Kumar H, Tham Y-C, Cheng C-Y. **Hypertension, blood pressure control and diabetic retinopathy in a large population-based study.**. *PLoS One* (2020) **15**. DOI: 10.1371/journal.pone.0229665 29. Janghorbani M, Amini M, Ghanbari H, Safaiee H. **Incidence of and risk factors for diabetic retinopathy in Isfahan, Iran.**. *Ophthalmic Epidemiol* (2003) **10** 81-95. DOI: 10.1076/opep.10.2.81.13893 30. Yanko L, Goldbourt U, Michaelson IC, Shapiro A, Yaari S. **Prevalence and 15-year incidence of retinopathy and associated characteristics in middle-aged and elderly diabetic men.**. *Br J Ophthalmol* (1983) **67** 759-65. DOI: 10.1136/bjo.67.11.759 31. Mitchell P, Smith W, Wang JJ, Attebo K. **Prevalence of diabetic retinopathy in an older community. The Blue Mountains Eye Study.**. *Ophthalmology* (1998) **105** 406-11. DOI: 10.1016/S0161-6420(98)93019-6 32. Srivastava BK, Rema M. **Does hypertension play a role in diabetic retinopathy?**. *J Assoc Physicians India* (2005) **53** 803-8. PMID: 16334627 33. Estacio RO, Jeffers BW, Gifford N, Schrier RW. **Effect of blood pressure control on diabetic microvascular complications in patients with hypertension and type 2 diabetes.**. *Diabetes Care* (2000) **23** B54-64. PMID: 10860192 34. Do DV, Wang X, Vedula SS, Marrone M, Sleilati G, Hawkins BS. **Blood pressure control for diabetic retinopathy.**. *Cochrane Database Syst Rev* (2015) **1**. PMID: 25637717 35. Nathan DM, Genuth S, Lachin J, Cleary P, Crofford O, Davis M. **The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.**. *N Engl J Med* (1993) **329** 977-86. DOI: 10.1056/NEJM199309303291401 36. Ohkubo Y, Kishikawa H, Araki E, Miyata T, Isami S, Motoyoshi S. **Intensive insulin therapy prevents the progression of diabetic microvascular complications in Japanese patients with non-insulin-dependent diabetes mellitus: a randomized prospective 6-year study.**. *Diabetes Res Clin Pract* (1995) **28** 103-17. DOI: 10.1016/0168-8227(95)01064-K 37. Callaghan BC, Little AA, Feldman EL, Hughes RAC. **Enhanced glucose control for preventing and treating diabetic neuropathy.**. *Cochrane Database Syst Rev* (2012). DOI: 10.1002/14651858.CD007543.pub2 38. Maric-Bilkan C. **Sex differences in micro- and macro-vascular complications of diabetes mellitus.**. *Clin Sci (Lond)* (2017) **131** 833-46. DOI: 10.1042/CS20160998 39. Seghieri G, Policardo L, Anichini R, Franconi F, Campesi I, Cherchi S. **The Effect of Sex and Gender on Diabetic Complications.**. *Curr Diabetes Rev* (2017) **13** 148-60. DOI: 10.2174/1573399812666160517115756 40. Muntner P, Carey RM, Gidding S, Jones DW, Taler SJ, Wright JT. **Potential U.S. Population Impact of the 2017 ACC/AHA High Blood Pressure Guideline.**. *J Am Coll Cardiol* (2018) **71** 109-18. DOI: 10.1016/j.jacc.2017.10.073