2.1. Study Design and Sample
The study sample comprised 1316 individuals from Northern Germany, recruited by the PopGen biobank in Kiel, Germany. Of them, 747 participants were derived from a random sample of the general population of Kiel, and 569 individuals were blood donors recruited at the University Hospital Schleswig-Holstein in Kiel (Germany), between 2005 and 2007 [
13]. A total of 952 individuals agreed to participate in the first follow-up examination (between 2010 and 2012) where blood sampling and a medical examination was performed. Furthermore, participants filled-out standardized questionnaires on demographics and various health-related characteristics, including smoking status, medical history, dietary intake, and physical activity (please see below for details) [
13,
14,
15]. Ultrasound examinations of the upper abdomen were performed in a subsample of 846 participants (please see below for details). Participants with missing information on ultrasound examination (
n = 4) or diagnosis of gallbladder sludge (
n = 6) were excluded, leaving 836 participants with ultrasound information regarding gallstone disease. In 582 of these participants, circulating vitamin E (α- and γ-tocopherol) concentrations were available. The study was approved by the Ethics Committee of the Medical Faculty of the Christian-Albrechts University of Kiel (Project identification code A 156/03). All participants provided written informed consent.
2.4. Definition of Covariates Assessment
Prevalent hypertension was defined as systolic blood pressure ≥140 mmHg, or diastolic blood pressure ≥90 mmHg, or use of antihypertensive medication, or self-reported hypertension history. Type 2 diabetes was defined as use of anti-diabetic medication or glycated hemoglobin (HbA1c) ≥6.5% (48 mmol/mol) or fasting serum glucose ≥126 mg/dL, or self-reported physician diagnosis. Liver disease was defined as self-reported hepatitis A, B, C, or D virus infection, hemochromatosis, autoimmune liver disease, or liver cirrhosis diagnosis.
In order to quantify the amount of physical activity, participants were asked to report their time spent walking, cycling, “do-it-yourself” activities, gardening, sports, and household chores during the past 12 months, and the average number of stairs climbed per day [
18]. The duration of each physical activity was then multiplied by the corresponding metabolic equivalent task (MET)-values and summed over all activities [
19,
20]. To determine smoking status, participants were classified into three categories: no-smokers (never smoked), former smokers (smoked in the past and quit smoking more than 1 year ago), and current smokers (currently smoking 1 or more cigarettes per day). According to their level of education, participants were categorized into three categories: low (≤9 years), middle (10 years), or high (≥11 years).
2.6. Statistical Analyses
Missing covariate values (
n = 13) were imputed as follows: missing categorical variables were imputed by the most commonly observed category of that respective variable (
n = 9), and missing values of normally distributed continuous variables were replaced by the respective mean, whereas skewed variables were imputed by the sex-specific median (
n = 4). When values of γ-tocopherol were below the detection limit (
n = 14), they were imputed by the lowest γ-tocopherol concentration measured in our study sample. Because vitamin E circulates in the blood bound to lipoproteins [
22,
23], circulating vitamin E (α- and γ-tocopherol) levels (μmol/L) were divided by circulating total cholesterol levels (mmol/L), and used as α-tocopherol/total cholesterol ratio (μmol/mmol) and γ-tocopherol/total cholesterol ratio (μmol/mmol) as primary exposure variables in the statistical analyses.
The following analyses were performed: first, for descriptive purposes, differences in the characteristics of the participants with and without gallstone disease were tested for statistical significance using the chi-squared test or the Fisher’s exact test, as appropriate, for categorical variables, and the Wilcoxon’s rank-sum test for continuous variables. Second, logistic regression models were conducted to estimate the associations of the α- and γ-tocopherol/cholesterol ratio (each ratio considered separately and modeled in tertiles) with gallstone disease. Age- and sex-adjusted, as well as multivariable-adjusted logistic regression models, were performed. The multivariable-adjusted model controlled for age (continuous in years), sex (female, male), education (low, medium, high), physical activity (continuous in MET-hour/week), smoking status (never, current, former), vitamin E supplementation (yes, no), alcohol intake (continuous in g/day), total energy intake (continuous in kJ/day), and BMI (continuous in kg/m2). Categorical variables with more than 2 categories were included as indicator variables.
Third, we performed several sensitivity analyses: we repeated the above mentioned logistic regression models, after (a) excluding vitamin E supplement users (n = 44, prevalent gallstone cases = 3); (b) excluding individuals with self-reported liver disease (hepatitis A, B, C, or D virus infection, hemochromatosis, autoimmune liver disease, or liver cirrhosis) (n = 30, prevalent gallstone cases = 1) from our analyses; and (c) excluding individuals with type 2 diabetes (n = 61, prevalent gallstone cases = 7). In further sensitivity analyses, we used α- and γ-tocopherol (without adjustment for cholesterol) as the exposure variables, and assessed the associations of these biomarkers with gallstone disease. Finally, we excluded all individuals with any missing values, and performed a complete case analysis (n = 558, prevalent gallstone cases = 45).
Analyses were performed with SAS 9.4 (SAS Institute, Cary, NC, USA). Statistical tests were 2-sided, and p values <0.05 were considered statistically significant.
Sabina Waniek 1
, 
