Abstract
Background & Aims:
Adiposity has been consistently associated with gallstone disease risk. We aimed to characterize associations of anthropometric measures (body mass index [BMI], recent weight change, long-term weight change, waist circumference, and waist-to-hip ratio) with symptomatic gallstone disease according to strata of gallstone disease polygenic risk score (PRS).
Methods:
We conducted analysis among 34,626 participants with available genome-wide genetic data within three large, prospective, US cohorts – Nurses’ Health Study (NHS), Health Professionals Follow-up Study, and NHS II. We characterized joint associations of PRS and anthropometric measures and tested for interactions on the relative and absolute risk scales.
Results:
Women in the highest BMI and PRS categories (BMI ≥ 30 kg/m2 and PRS ≥ 1 SD above mean) had odds ratio (OR) for gallstone disease of 5.55 (95% CI, 5.29–5.81) compared with those in the lowest BMI and PRS categories (BMI < 25 kg/m2 and PRS < 1 SD below mean). The corresponding OR among men was 1.65 (95% CI, 1.02–2.29). Associations for BMI did not vary within strata of PRS on the relative risk scale. On the absolute risk scale, the incidence rate difference between obese and normal-weight individuals was 1086/100,000 PYs within the highest PRS category, compared to 666/100,000 PYs in the lowest PRS category, with strong evidence for interaction with the ABCG8 locus.
Conclusions:
While maintenance of a healthy body weight reduces gallstone disease risk among all individuals, risk reduction is higher among the subset with greater genetic susceptibility to gallstone disease.
Keywords: BMI, GWAS, Polygenic risk score, Gallbladder disease
Graphical Abstract
INTRODUCTION
Gallstone disease is a cause of significant morbidity and one of the most common causes of gastrointestinal-related hospitalizations in the United States.1 Previous studies have established numerous lifestyle risk factors for gallstone disease.2–5 In particular, Anthropometric measures are among the strongest risk factors of gallstone disease.2 Body mass index (BMI),2 weight fluctuation,6 waist circumference,7 and waist-to-hip ratio7 have all been shown to independently contribute to risk, and are thought to influence lipid metabolism and biliary supersaturation of cholesterol.8 The role of cholesterol supersaturation in gallstone formation was further supported by findings from our group and others of 6 SNPs associated with gallstone disease susceptibility discovered through a large genome wide association study (GWAS), 5 of which were mapped to cholesterol or bile acid metabolism.9,10
The complex interplay between genetic and environmental contributions to gallstones has been previously explored, mainly through familial clustering, twin studies, and Mendelian randomization studies.11–14 Yuan et al. investigated the causal association via Mendelian randomization and found that genetic predisposition to high BMI, high waist circumference, and type 2 diabetes were independently associated with a higher risk of gallstone disease.13 Gene-environment interactions (GxE) — how an individual’s genetic risk is modified by environmental determinants — are characterized on the relative risk (multiplicative) or the absolute risk (additive) scale.15–17 GxE multiplicative interactions assess whether the excess in ratio measure of the association between risk factors and disease varies between individuals with and without the genetic risk factor. In contrast, additive interactions assess whether the excess in difference measure of the association varies between individuals with and without genetic risk factor.
To our knowledge, no previous studies have comprehensively characterized joint associations of genetic and environmental factors and GxE in the risk of symptomatic gallstone disease. Thus, our objective was to characterize the joint associations of a polygenic risk score (PRS) composed of 6 GWAS-identified independent susceptibility SNPs (rs1260326 in GCKR, rs6471717 near CYP7A1, rs2547231 in SULT2A1, rs11887534 and rs4245791 in ABCG8, and rs9843304 in TM4SF4) and 5 anthropometric exposures (BMI, recent weight change, long-term weight change, waist circumference, and waist-to-hip ratio) in the risk of symptomatic gallstone disease. A characterization of joint associations and interactions of genetic and environmental risk factors may provide a better understanding of the mechanism and may help to identify subsets of the population in whom modifying lifestyle factors may result in the greatest risk reduction.
METHODS
Study Population
We examined joint associations of anthropometric exposures and polygenic risk scores in three large, prospective, US-based cohort studies – the Nurses’ Health Study (NHS), Health Professionals Follow-up Study (HPFS) and NHS II. The study population is described in detail in prior studies.18 In brief, the NHS began in 1976, with 121,700 female registered nurses aged 30–55 answering the baseline questionnaire. The NHS II began in 1989, recruiting 116,430 female registered nurses aged 25–42, and the HPFS began in 1986, enrolling 51,529 male health workers aged 40–75. Participants of these three cohorts have been followed by biennial questionnaires assessing medical conditions and lifestyle factors, with ≥90% follow-up rates.
(1). Complete Cohort
The “complete cohort” included 236,942 participants (89,583 in NHS; 100,128 in NHS II; 47,231 in HPFS), after excluding participants who were deceased or had cholecystectomy prior to the earliest questionnaire that queried about cholecystectomy.
(2). Genotyped Subcohort
Within the complete cohort, 37,664 participants had genome-wide genotyping data available.19 After baseline exclusions, 34,626 participants (16,289 in NHS; 7,832 in NHS II; 10,505 in HPFS) with genotyping information constituted the “genotyped sub-cohort.”
(3). Matched Case-Control Dataset
We then used a modified version of incidence density sampling approach20 to generate a nested case-control dataset. We matched each case (N= 4,785) to 3 controls based on cohort, age, follow-up period, race, DNA source, genotyping platform, the probability to be sampled for GWAS, and follow-up period. In the end, 4,761 cases were matched to 3 controls each and remaining cases had between 0 and 2 controls matched (NHS: 2,710 cases and 8,123 controls; NHS II: 1,178 cases and 3,532 controls; HPFS: 897 cases and 2,691 controls). These participants formed the “matched case-control dataset.”
Details of the participants who provided biospecimen, matching criteria for the matched case-control dataset, and calculation of the probability for being sampled for GWAS are described in the Supplementary Materials. The NHS and NHS II were approved by the Institutional Review Board at Brigham and Women’s Hospital and the HPFS was approved by the Harvard T.H. Chan School of Public Health.
Assessment of Symptomatic Gallstone Disease
Consistent with prior studies,9,21 we defined symptomatic gallstone disease as self-reported cholecystectomy on a biennial questionnaire. Self-reports of gallstones without cholecystectomy were included as non-cases due to risk of cholecystectomy in the future. Further, a selection of gallstone disease with symptoms that lead to cholecystectomy may be clinically more relevant.9 Participants were asked if they underwent a cholecystectomy biennially from 1982 to 2004 and again in 2012 in NHS, from 1991 to 2013 in NHS II, and from baseline to 2012 in HPFS. Right censoring occurred at following cutoffs, whichever came first: the time of cholecystectomy, death, loss to follow-up, or last questionnaire (2012 in NHS and HPFS; 2013 in NHS II). Self-reports of cholecystectomy were validated in previous studies through review of medical records.21 Leitzmann et al. reviewed randomly selected 50 NHS participants who reported a cholecystectomy in 1982: among 43 nurses who responded, all affirmed the report and among 36 medical records that were obtained, all confirmed cholecystectomy.21
Anthropometric and Polygenic Risk Score Assessment
We prospectively obtained anthropometric exposures from questionnaires. Details on assessment and categorization of anthropometric measures and calculation and categorization of PRS for marginal association and interaction analyses are described in Supplementary Materials. PRS was computed as a weighted sum of number of risk alleles from 6 SNPs, followed by standardization to mean=0 and standard deviation=1.
Statistical Analyses
Association of BMI, recent weight change, long-term weight change, waist circumference, and waist-to-hip ratio was assessed in three different ways using (1) complete cohort (N=236,942), (2) genotyped sub-cohort (N=34,626), and (3) matched case-control dataset (N=19,131).
(1). Complete Cohort
Complete cohort analysis was performed using Cox proportional hazards regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Marginal associations were assessed in complete cohort, adjusting for covariates (calorie intake, Alternative Healthy Eating Index score, physical activity, smoking, coffee intake, hyperlipidemia with or without statin use, hypertension with or without thiazide use, diabetes, current aspirin use, current use of other NSAID, postmenopausal hormone use). See Supplementary Materials for covariate assessment. Ptrend was calculated as linear trend by assigning cohort-specific median values to all participants in each exposure category. Ptrend was not computed for recent weight change because prior studies have observed an increased association for both recent weight loss and recent weight gain categories.6
(2). Genotyped sub-cohort
Genotyped sub-cohort analysis was performed using Cox proportional hazards regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Analysis of the genotyped sub-cohort data was inverse probability weighted26 based on the probability of being selected for genotyping. Marginal associations were assessed, adjusting for covariates described above.
(3). Matched Case-Control Dataset
Matched case-control analysis was performed using conditional logistic regression, conditional on matching variables (see Study Population), to estimate odds ratios (ORs) and 95% CIs. Marginal associations were also assessed, adjusting for covariates described above. We tested for multiplicative interactions using likelihood ratios that compare ORs from two conditional logistic regression models with and without product terms of PRS and anthropometric exposures, and with and without adjusting for covariates. We tested for additive interactions by computing two measures: relative excess risk due to interaction (RERI) and synergy index (SI).23,27
Assuming 60 independent hypothesis tests (6 SNPs, 5 anthropometric exposures and tests for multiplicative and additive interactions), we conducted multiple comparisons correction using the Bonferroni method and set the corresponding α threshold to 8.33×10−4. Marginal associations were tested using Statistical Analysis Software 9.4 (SAS Institute Inc, Cary, NC). Interactions were tested using R, version 3.5.0.
All authors had access to the study data and reviewed and approved the final manuscript.
RESULTS
Complete Cohort
In the complete cohort, over a total follow-up period of 4.8 million person-years, we documented 28,524 incident cholecystectomies – 12,548 cases in NHS, 12,933 cases in NHS II, and 3,043 cases in HPFS. Table 1 shows age-standardized descriptive characteristics at mid follow-up (year 2000 in NHS and HPFS; 2001 in NHS II) of 67,115 NHS, 34,260 HPFS, 84,853 NHS II participants. Risk patterns for anthropometric markers were consistent across the three cohorts. As expected, participants with a greater BMI had a greater weight gain since early adulthood, greater waist circumference, and greater waist-to-hip ratio.
Table 1.
Age-standardized characteristics according to body mass index (kg/m2) categories among all participants in 2000–2001
| Nurses’ Health Study 2000 | Nurses’ Health Study II 2001 | Health Professionals Follow-Up Study 2000 | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Body Mass Index, kg/m2 (n=number of participants) | <25 (n=30,287) |
25–29.9 (n=22,959) |
≥30 (n=13,783) |
<25 (n=42,907) |
25–29.9 (n=23,304) |
≥30 (n=18,642) |
<25 (n=13,527) |
25–29.9 (n=16,334) |
≥30 (n=4,399) |
| Age, years | 66.9 (7.2) | 66.2 (7.0) | 65.0 (6.8) | 46.3 (4.6) | 47.1 (4.6) | 47.3 (4.6) | 67.8 (9.5) | 65.7 (8.5) | 64.3 (7.7) |
| Weight change in 2 years, lbsa | −0.9 (8.0) | 1.0 (9.7) | 3.5 (15.1) | 0.3 (7.4) | 2.9 (10.7) | 5.9 (15.8) | −1.3 (9.5) | 1.1 (7.7) | 4.0 (12.3) |
| Weight change since age 18 or 21, lbs | 11.5 (16.7) | 35.5 (17.0) | 66.6 (27.8) | 13.4 (14.0) | 35.8 (17.0) | 71.0 (31.3) | 8.7 (16.5) | 25.7 (17.9) | 51.9 (25.3) |
| Waist circumference, inches | 30.8 (3.6) | 35.1 (4.0) | 39.8 (5.2) | 28.1 (2.8) | 31.3 (3.6) | 36.6 (5.6) | 35.9 (2.8) | 39.1 (3.1) | 43.9 (4.1) |
| Waist-to-hip ratio | 0.81 (0.09) | 0.85 (0.10) | 0.87 (0.10) | 0.76 (0.07) | 0.79 (0.08) | 0.82 (0.09) | 0.93 (0.06) | 0.96 (0.07) | 0.98 (0.07) |
| Total energy intake, kcal/day | 1729.6 (518.3) |
1732.9 (526.0) |
1740.5 (542.1) |
1803.6 (541.2) |
1818.2 (551.8) |
1857.5 (572.5) |
2004.4 (596.2) |
1986.6 (607.8) |
2042.3 (650.2) |
| AHEI score | 55.2 (10.9) | 54.0 (10.6) | 52.5 (10.1) | 48.9 (11.1) | 47.3 (10.6) | 45.8 (10.4) | 58.1 (11.6) | 55.8 (11.0) | 53.8 (10.8) |
| Exercise, MET-hrs/wk | 21.0 (23.9) | 16.6 (20.4) | 11.5 (17.4) | 25.5 (30.8) | 19.0 (23.7) | 13.3 (18.5) | 37.3 (30.7) | 26.8 (27.9) | 19.5 (23.6) |
| Non-Caucasian race, % | 5.9 | 7.3 | 7.8 | 5.0 | 5.5 | 5.9 | 5.3 | 4.3 | 4.1 |
| Smoking | |||||||||
| Never, % | 43.9 | 44.7 | 45.0 | 65.7 | 63.3 | 65.0 | 51.1 | 43.2 | 37.4 |
| Past, % | 48.0 | 49.6 | 50.6 | 28.0 | 29.9 | 28.8 | 45.9 | 52.7 | 59.7 |
| Current, % | 8.2 | 5.7 | 4.4 | 6.3 | 6.9 | 6.2 | 3.0 | 3.1 | 2.9 |
| Coffee intake, servings/day | 1.3 (1.4) | 1.3 (1.4) | 1.2 (1.4) | 1.3 (1.4) | 1.3 (1.5) | 1.1 (1.4) | 1.1 (1.4) | 1.3 (1.5) | 1.4 (1.5) |
| Hyperlipidemia, % | 54.4 | 64.1 | 66.8 | 22.2 | 33.6 | 43.3 | 48.0 | 54.7 | 57.2 |
| Hypertension, % | 36.7 | 50.2 | 66.4 | 8.4 | 17.1 | 34.4 | 32.9 | 44.3 | 60.9 |
| Diabetes Mellitus, % | 2.3 | 5.9 | 13.7 | 0.3 | 1.0 | 5.6 | 4.4 | 6.2 | 12.1 |
| Current statin use, % | 14.8 | 22.2 | 24.0 | 2.1 | 5.0 | 8.0 | 17.2 | 22.5 | 23.0 |
| Current thiazide use, % | 7.9 | 11.9 | 17.2 | 1.8 | 3.8 | 8.4 | 3.7 | 5.2 | 9.1 |
| Current aspirin use, % | 42.8 | 44.7 | 43.4 | 9.3 | 10.4 | 11.6 | 47.0 | 48.0 | 48.4 |
| Current NSAID use, % | 15.9 | 19.6 | 21.0 | 37.3 | 42.3 | 47.7 | 5.1 | 7.1 | 10.0 |
| Postmenopausal, % | 98.8 | 98.8 | 98.6 | 26.4 | 29.3 | 31.2 | N/A | N/A | N/A |
| Past/current PMH use, % | 76.3 | 73.1 | 66.8 | 19.1 | 21.1 | 21.6 | N/A | N/A | N/A |
Values are means (SD) or percentages and are standardized to the age distribution of the study population.
Weight change over 1998–2000 for HPFS and NHS, and 1999–2001 for NHS II.
Consistent with prior studies,2 in multivariable analyses of anthropometric measures among complete cohort participants (Table 2 and Supplementary Tables 1–4), those with BMI ≥35.0 kg/m2 (HR=3.41; 95% CI, 32.21–3.63), weight loss >5 lbs over past 2 years (OR=1.36; 95% CI, 1.32–1.41), weight gain ≥30 lbs since age 18 or 21 (OR=1.62; 95% CI, 1.55–1.70), waist circumference 5th quintile (OR=1.53; 95% CI, 1.44–1.62), and waist-to-hip ratio 5th quintile (OR=1.33; 95% CI, 1.26–1.40) had a higher risk of cholecystectomy than their respective counterparts in the lowest quantile. The association of BMI (BMI ≥35.0 kg/m2 vs BMI <22) with gallstone disease risk was stronger among women (HR=3.52; 95% CI, 3.31–3.75), compared to men (HR=1.92; 95% CI, 1.47–2.51) (Table 2). Supplementary Table 5 shows that among the anthropometric indicators, BMI is the most important risk factor associated with gallstone risk. Adult weight gain is independently associated among women and men. Additionally, waist circumference is associated in men after adjusting for other anthropometric markers.
Table 2.
Association of body mass index with symptomatic gallstone disease (kg/m2) in male and female participants
| Body mass index, kg/m2 | |||||||
|---|---|---|---|---|---|---|---|
| <22 | 22–24.9 | 25–29.9 | 30–34.9 | ≥35 | P trend a | ||
| All participants (NHS, NHS II, and HPFS) | |||||||
| Complete b | Cases/Person-years | 3019/1,047,003 | 5806/1,405,915 | 9984/1,553,000 | 5693/539,996 | 3979/268,644 | |
| Age-adjusted, HR (95% CI) | 1 (Ref.) | 1.51 (1.45, 1.58) | 2.49 (2.39, 2.60) | 3.81 (3.65, 3.99) | 4.99 (4.75, 5.24) | <10−16 | |
| MV-adjustedc, HR (95% CI) | 1 (Ref.) | 1.39 (1.33, 1.46) | 2.00 (1.90, 2.10) | 2.79 (2.64, 2.96) | 3.41 (3.21, 3.63) | <10−16 | |
| Genotyped d | Cases/Person-years | 482/153,208 | 1067/240,431 | 1734/277,632 | 966/87,613 | 536/36,934 | |
| Age-adjusted, HR (95% CI) | 1 (Ref.) | 1.51 (1.43, 1.59) | 2.32 (2.21, 2.44) | 4.09 (3.87, 4.32) | 4.96 (4.67, 5.28) | <10−16 | |
| MV-adjustedc, HR (95% CI) | 1 (Ref.) | 1.46 (1.38, 1.55) | 1.88 (1.77, 2.00) | 2.95 (2.75, 3.17) | 3.42 (3.16, 3.70) | <10−16 | |
| Matched e | Cases/Controls | 482/2901 | 1067/4134 | 1734/4852 | 966/1676 | 536/783 | |
| Unadjusted, OR (95% CI) | 1 (Ref.) | 1.59 (1.41, 1.79) | 2.26 (2.02, 2.53) | 3.75 (3.29, 4.26) | 4.57 (3.93, 5.32) | <10−16 | |
| MV-adjustedc, OR (95% CI) | 1 (Ref.) | 1.45 (1.28, 1.65) | 1.81 (1.58, 2.07) | 2.75 (2.35, 3.22) | 3.23 (2.69, 3.87) | <10−16 | |
| Female participants (NHS and NHS II) | |||||||
| Complete b | Cases/Person-years | 2840/965,398 | 4986/1,097,047 | 8461/1,124,571 | 5282/455,091 | 3880/251,015 | |
| Age-adjusted, HR (95% CI) | 1 (Ref.) | 1.53 (1.46, 1.61) | 2.56 (2.45, 2.67) | 3.93 (3.75, 4.12) | 5.10 (4.85, 5.36) | <10−16 | |
| MV-adjustedc, HR (95% CI) | 1 (Ref.) | 1.41 (1.34, 1.48) | 2.07 (1.96, 2.18) | 2.91 (2.75, 3.09) | 3.52 (3.31, 3.75) | <10−16 | |
| Genotyped d | Cases/Person-years | 432/133,133 | 835/163,686 | 1283/167,273 | 834/65,521 | 504/31,947 | |
| Age-adjusted, HR (95% CI) | 1 (Ref.) | 1.55 (1.47, 1.64) | 2.42 (2.29, 2.55) | 4.29 (4.06, 4.54) | 5.13 (4.82, 5.45) | <10−16 | |
| MV-adjustedc, HR (95% CI) | 1 (Ref.) | 1.52 (1.43, 1.61) | 1.97 (1.85, 2.11) | 3.15 (2.93, 3.39) | 3.57 (3.30, 3.87) | <10−16 | |
| Matched e | Cases/Controls | 432/2656 | 835/3257 | 1283/3576 | 834/1435 | 504/731 | |
| Unadjusted, OR (95% CI) | 1 (Ref.) | 1.61 (1.42, 1.83) | 2.33 (2.06, 2.64) | 3.88 (3.38, 4.45) | 4.73 (4.03, 5.54) | <10−16 | |
| MV-adjustedc, OR (95% CI) | 1 (Ref.) | 1.48 (1.29, 1.70) | 1.83 (1.58, 2.13) | 2.84 (2.39, 3.38) | 3.35 (2.75, 4.07) | <10−16 | |
| Male participants (HPFS) | |||||||
| Complete b | Cases/Person-years | 179/81,605 | 820/308,868 | 1523/428,429 | 411/84,905 | 99/17,629 | |
| Age-adjusted, HR (95% CI) | 1 (Ref.) | 1.27 (1.08, 1.50) | 1.74 (1.49, 2.04) | 2.42 (2.02, 2.89) | 2.90 (2.26, 3.73) | <10−16 | |
| MV-adjustedc, HR (95% CI) | 1 (Ref.) | 1.18 (1.00, 1.39) | 1.42 (1.20, 1.68) | 1.73 (1.42, 2.12) | 1.92 (1.47, 2.51) | 4.31×10−11 | |
| Genotyped d | Cases/Person-years | 50/20,075 | 232/76,745 | 451/110,359 | 132/22,092 | 32/4987 | |
| Age-adjusted, HR (95% CI) | 1 (Ref.) | 1.16 (0.97, 1.37) | 1.61 (1.37, 1.90) | 2.38 (1.97, 2.87) | 2.47 (1.84, 3.31) | <10−16 | |
| MV-adjustedc, HR (95% CI) | 1 (Ref.) | 1.06 (0.89, 1.27) | 1.33 (1.11, 1.59) | 1.69 (1.22, 2.33) | 1.69 (1.22, 2.33) | 1.72×10−9 | |
| Matched e | Cases/Controls | 50/245 | 232/877 | 451/1276 | 132/241 | 32/52 | |
| Unadjusted, OR (95% CI) | 1 (Ref.) | 1.32 (0.94, 1.85) | 1.77 (1.28, 2.46) | 2.81 (1.93, 4.09) | 3.24 (1.89, 5.55) | 2.41×10−12 | |
| MV-adjustedc, OR (95% CI) | 1 (Ref.) | 1.25 (0.88, 1.78) | 1.48 (1.03, 2.11) | 2.08 (1.36, 3.20) | 2.14 (1.18, 3.88) | 2.48×10−4 | |
Fixed effects models.
Complete cohort: all participants regardless of the genetic data availability.
Adjusted for long-term weight change, waist-to-hip ratio, calorie intake, Alternative Healthy Eating Index score, physical activity, smoking, coffee intake, hyperlipidemia, statin use, hypertension, thiazide use, diabetes, aspirin use, other NSAID use, and postmenopausal hormone use.
Genotyped sub-cohort: inverse proportional weighted model among participants with genetic data.
Matched case-control: matched based on age, sex, race, DNA source, propensity score, genotyping platform, and time period.
Genotyped Subcohort
In the genotyped subcohort, we identified 4,785 incident cases of cholecystectomy over a cumulative follow-up of 795,818 person-years. Supplementary Tables 6 shows descriptive characteristics of the genotyped subcohort. The characteristics of the genotyped subcohort was not observed to be substantially different from the complete cohort at mid-follow up. The magnitude and directions of associations of anthropometric measures (Table 2 and Supplementary Tables 1–4) were similar in the genotyped subcohort compared to complete cohort.
Matched Case-Control Dataset
Supplementary Table 7 shows descriptive characteristics of the matched case-control dataset according to BMI. Supplementary Tables 8–11 show that descriptive characteristics of relevant covariates did not change substantially by genotype status. Supplementary Table 12 shows associations for gallstone disease predisposition SNPs in our dataset. Supplementary Table 13 shows associations for quintiles of PRS and risk of cholecystectomy. PRS also showed stronger associations with cholecystectomy among women (Quintile 5 vs. Quintile 1 OR = 1.93; 95% CI, 1.72–2.17) compared to men (Quintile 5 vs. Quintile 1 OR = 1.45; 95% CI, 1.14–1.85).
Supplementary Table 14 shows age-adjusted BMI associations within strata of PRS, conditional on matching factors. Among participants with low genetic predisposition (PRS >1 SD below mean), OR for obese vs. normal-weight individuals was 3.03 (95% CI, 2.79–3.27). The corresponding magnitude of association among participants with average genetic predisposition (PRS ±1 SD of mean) was OR 3.03 (95% CI, 2.88–3.17) and among high genetic predisposition (PRS >1 SD above mean) was OR 2.62 (95% CI, 2.49–2.74). Overall, we did not observe evidence for heterogeneity in trends, suggesting BMI associations did not substantially vary within strata of genetic risk on the relative risk scale. Supplementary Table 15 showed BMI associations within strata of PRS after adjusting for covariates.
Table 3 reports combined associations of BMI and PRS after adjusting for covariates. Both BMI and PRS contributed to the increase in gallstone disease risk – participants in the lowest BMI and highest PRS categories had an OR 1.91 (95% CI, 1.71–2.12), and participants in the highest BMI and lowest PRS categories had an OR 2.71 (95% CI 2.46–2.95). The risk of gallstone disease proportionately increased among individuals in the intermediate categories and those in the highest categories of BMI and PRS. OR for gallstone disease risk for those in the highest BMI and PRS categories vs. lowest BMI and PRS categories was observed to be 4.62 (95% CI, 4.39–4.86). These associations were stronger among women (OR=5.55; 95% CI, 5.29–5.81) compared to men (OR=1.65; 95% CI, 1.02–2.29).
Table 3.
Multiplicative interaction between polygenic risk score and body mass index in male and female participants
| BMI < 25 kg/m2 | BMI 25–29.9 kg/m2 | BMI ≥ 30 kg/m2 | Ptrend | ||
|---|---|---|---|---|---|
| All matched participants (NHS, NHS II, and HPFS) | |||||
| PRSa < −1 | Cases/Controls | 195/1105 | 183/774 | 188/380 | <10−16 |
| ORb (95% CI) | 1.0Ref | 1.34 (1.11, 1.57) | 2.71 (2.46, 2.95) | ||
| Incidence rate/100,000 PYsc | 320 | 428 | 986 | ||
| -1 ≤ PRSa < 1 | Cases/Controls | 1073/5098 | 1230/3499 | 1071/1790 | <10−16 |
| ORb (95% CI) | 1.20 (1.03, 1.37) | 1.96 (1.79, 2.13) | 3.27 (3.10, 3.45) | ||
| Incidence rate/100,000 PYsc | 378 | 615 | 1179 | ||
| PRSa ≥ 1 | Cases/Controls | 281/832 | 321/579 | 243/289 | <10−16 |
| ORb (95% CI) | 1.91 (1.71, 2.12) | 3.14 (2.93, 3.35) | 4.62 (4.39, 4.86) | ||
| Incidence rate/100,000 PYsc | 576 | 920 | 1662 | ||
| P Interaction-3×3 d | 0.313 | ||||
| P Interaction-3×1 e | 0.896 | ||||
| Among females (NHS and NHS II) | |||||
| PRSa < −1 | Cases/Controls | 155/928 | 140/585 | 164/353 | <10−16 |
| ORb (95% CI) | 1.0Ref | 1.45 (1.20, 1.71) | 2.74 (2.48, 3.00) | ||
| Incidence rate/100,000 PYsc | 334 | 525 | 1047 | ||
| -1 ≤ PRSa < 1 | Cases/Controls | 867/4305 | 911/2568 | 952/1573 | <10−16 |
| ORb (95% CI) | 1.21 (1.02, 1.39) | 2.11 (1.91, 2.30) | 3.56 (3.36, 3.75) | ||
| Incidence rate/100,000 PYsc | 405 | 757 | 1345 | ||
| PRSa ≥ 1 | Cases/Controls | 245/680 | 232/423 | 222/240 | <10−16 |
| ORb (95% CI) | 2.16 (1.93, 2.38) | 3.32 (3.08, 3.56) | 5.55 (5.29, 5.81) |
||
| Incidence rate/100,000 PYsc | 680 | 1144 | 2012 | ||
| P Interaction-3×3 d | 0.606 | ||||
| P Interaction-3×1 e | 0.655 | ||||
| Among males (HPFS) | |||||
| PRSa < −1 | Cases/Controls | 40/177 | 43/189 | 24/27 | 1.95×10−4 |
| ORb (95% CI) | 1.0Ref | 0.97 (0.47, 1.46) | 3.58 (2.91, 4.24) | ||
| Incidence rate/100,000 PYsc | 277 | 268 | 705 | ||
| -1 ≤ PRSa < 1 | Cases/Controls | 206/793 | 319/931 | 119/217 | 9.05×10−7 |
| ORb (95% CI) | 1.21 (0.83, 1.60) | 1.55 (1.17, 1.93) | 2.29 (1.86, 2.73) | ||
| Incidence rate/100,000 PYsc | 296 | 400 | 592 | ||
| PRSa ≥ 1 | Cases/Controls | 36/152 | 89/156 | 21/49 | 3.30×10−3 |
| ORb (95% CI) | 1.06 (0.54, 1.57) | 2.60 (2.16, 3.05) | 1.65 (1.02, 2.29) | ||
| Incidence rate/100,000 PYsc | 283 | 609 | 585 | ||
| P Interaction-3×3 d | 6.24×10−4 | ||||
| P Interaction-3×1 e | 0.162 | ||||
PRS < −1 refers to PRS more than 1 SD below the mean. −1 ≤ PRS < 1 refers to PRS within 1 SD from the mean. 1 ≤ PRS refers to more than or equal to 1 SD above the mean.
Odds ratios are unadjusted and conditional on matching factors.
Incidence rates were calculated using all genotyped participants as number of incident cases per 100,000 person-years of follow-up.
P-value for interaction between 3 categories of BMI (<25, 25–29.9, ≥30 kg/m2) and 3 categories of PRS (< −1, ≤ and <1, ≥1).
P-value for interaction between continuous BMI and 3 categories of PRS (< −1, ≤ and <1, ≥1).
Table 3 also reports the absolute incidence rates of gallstone disease among strata of BMI and PRS in the overall cohort with genetic data. Among participants with low genetic predisposition, incidence rates increased from 320 to 986/100,000 person-years, reflecting an absolute increase of 666/100,000 person-years. This absolute increase was greater among participants who had a stronger genetic predisposition for gallstone disease – among average genetic risk individuals it was 801/100,000 person-years and among high genetic risk individuals it was 1086/100,000 person-years. This heterogeneity in the increase in incidence rates of gallstone disease due to BMI within strata of PRS on the absolute risk scale is indicative of BMI-PRS additive interaction. The absolute excess risk due to interaction in participants with average genetic predisposition was 135/100,000 person-years, and in participants with high genetic predisposition was 420/100,000 person-years. The excess risk due to BMI-PRS interaction was primarily observed among females. Supplementary Table 16 shows that the 10-year cumulative probability of incident cholecystectomy was 18.39% among women within the highest BMI and PRS categories, compared to 6.60% among those with healthy weight status and highest PRS category. The corresponding absolute risks among men with high genetic liability are 5.70% and 2.79% respectively. Thus, maintaining a healthy weight status offsets detrimental genetic influence.
Table 4 reports tests for BMI-PRS additive interactions, each considered dichotomous. On the absolute risk scale, the RERI was estimated to be 0.35 (95% CI, 0.11–0.60; P=4.85 × 10−3) and the corresponding SI was 1.23 (95% CI, 1.07–1.38; P=8.59×10−3). These were not statistically significant after correction for multiple comparisons. However, additive interaction tests between BMI and individual SNPs suggested the ABCG8 SNP rs4245791 was the likely driver of the excess risk in joint associations. Table 5 shows interactions between rs4245791 in ABCG8 and BMI. The estimated rise in the absolute risk of cholecystectomy among participants with the TT/TC genotype and BMI ≥25 kg/m2 was greater than the sum of excess risks associated with main effects of the genotype or BMI (RERI=0.39; 95% Cl, 0.18–0.61; P=4.16×10−4). When analyses were stratified by sex, the association was statistically significantly stronger for female participants with at least one T allele at rs4245791 (RERI=0.53; 95% CI, 0.27–0.79; P=6.05×10−5), compared to carriers with the wildtype C/C genotype. In contrast, we did not observe this interaction among male participants (RERI=0.04; 95% CI, −0.36–0.44; P=0.838).
Table 4.
Additive interaction between polygenic risk score and body mass index in male and female participants
| BMI < 25 kg/m2 | BMI ≥ 25 kg/m2 | Interaction P-value | ||
|---|---|---|---|---|
| All participants (HPFS, NHS and NHS II) | ||||
| PRSa < 0 | Cases/Controls | 723/3891 | 1554/4059 | |
| ORb (95% CI) | 1.0Ref | 2.13 (2.02, 2.24) | ||
| PRSa ≥ 0 | Cases/Controls | 826/3144 | 1682/3252 | |
| ORb (95% CI) | 1.42 (1.32, 1.52) | 2.90 (2.76, 3.04) | ||
| Combined Association | ||||
| Additive interaction: RERI for (95% Cl); H0c: RERI = 0 | 0.35 (0.11, 0.60) | 4.85×10−3 | ||
| Additive interaction: SI (95% CI); H0c: SI = 1 | 1.23 (1.07, 1.38) | 8.59×10−3 | ||
| Multiplicative interaction: IOR (95% CI); H0c: IOR = 1 | 0.96 (0.82, 1.10) | 5.70×10−1 | ||
| Among females (NHS and NHS II) | ||||
| PRSa < 0 | Cases/Controls | 586/3283 | 1252/3197 | |
| ORb (95% CI) | 1.0Ref | 2.29 (2.17, 2.41) | ||
| PRSa ≥ 0 | Cases/Controls | 681/2630 | 1369/2545 | |
| ORb (95% CI) | 1.45 (1.34, 1.56) | 3.15 (3.00, 3.30) | ||
| Combined Association | ||||
| Additive interaction: RERI for (95% Cl); H0c: RERI = 0 | 0.41 (0.12, 0.70) | 5.61×10−3 | ||
| Additive interaction: SI (95% CI); H0c: SI = 1 | 1.24 (1.08, 1.39) | 9.34×10−3 | ||
| Multiplicative interaction: IOR (95% CI); H0c: IOR = 1 | 0.95 (0.80, 1.10) | 4.97×10−1 | ||
| Among males (HPFS) | ||||
| PRSa < 0 | Cases/Controls | 137/608 | 302/862 | |
| ORb (95% CI) | 1.0Ref | 1.56 (1.30, 1.82) | ||
| PRSa ≥ 0 | Cases/Controls | 145/514 | 313/707 | |
| ORb (95% CI) | 1.25 (1.03, 1.48) | 2.01 (1.69, 2.33) | ||
| Combined Association | ||||
| Additive interaction: RERI for (95% Cl); H0c: RERI = 0 | 0.20 (−0.24, 0.64) | 3.77×10−1 | ||
| Additive interaction: SI (95% CI); H0c: SI = 1 | 1.24 (0.70, 1.78) | 4.30×10−1 | ||
| Multiplicative interaction: IOR (95% CI); H0c: IOR = 1 | 1.03 (0.71, 1.35) | 8.64×10−1 | ||
PRS < 0 refers to PRS below the mean. PRS ≥ 0 refers to PRS above or equal to the mean.
Odds ratios are unadjusted and conditional on matching factors.
H0 = the null hypotheses tests were for RERI = 0; Synergy Index = 1 and interaction odds ratio = 1
Table 5.
Additive interaction between ABCG8 rs4245791 and body mass index in male and female participants
| BMI < 25 kg/m2 | BMI ≥ 25 kg/m2 | Interaction P-value | ||
|---|---|---|---|---|
| All participants (HPFS, NHS and NHS II) | ||||
| rs4245791 | Cases/Controls | 791/3843 | 1615/4088 | |
| CC | ORa (95% CI) | 1.0Ref | 1.99 (1.88, 2.10) | |
| rs4245791 | Cases/Controls | 758/3192 | 1621/3223 | |
| TT+TC | ORa (95% CI) | 1.16 (1.06, 1.26) | 2.55 (2.41, 2.68) | |
| Combined Association | ||||
| Additive interaction: RERI for (95% Cl); H0b: RERI = 0 | 0.39 (0.18, 0.61) | 4.16×10−4 | ||
| Additive interaction: SI (95% CI); H0b: SI = 1 | 1.34 (1.16, 1.52) | 1.63×10−3 | ||
| Multiplicative interaction: IOR (95% CI); H0b: IOR = 1 | 1.10 (0.96, 1.24) | 1.71×10−1 | ||
| Among females (NHS and NHS II) | ||||
| rs4245791 | Cases/Controls | 641/3235 | 1288/3233 | |
| CC | ORa (95% CI) | 1.0Ref | 2.10 (1.98, 2.22) | |
| rs4245791 | Cases/Controls | 626/2678 | 1333/2509 | |
| TT+TC | ORa (95% CI) | 1.18 (1.07, 1.29) | 2.82 (2.66, 2.97) | |
| Combined Association | ||||
| Additive interaction: RERI for (95% Cl); H0b: RERI = 0 | 0.53 (0.27, 0.79) | 6.05×10−5 | ||
| Additive interaction: SI (95% CI); H0b: SI = 1 | 1.41 (1.22, 1.60) | 3.62×10−4 | ||
| Multiplicative interaction: IOR (95% CI); H0b: IOR = 1 | 1.13 (0.98, 1.29) | 1.14×10−1 | ||
| Among males (HPFS) | ||||
| rs4245791 | Cases/Controls | 150/608 | 327/855 | |
| CC | ORa (95% CI) | 1.0Ref | 1.57 (1.30, 1.83) | |
| rs4245791 | Cases/Controls | 132/514 | 288/714 | |
| TT+TC | ORa (95% CI) | 1.05 (0.82, 1.27) | 1.65 (1.33–1.98) | |
| Combined Association | ||||
| Additive interaction: RERI for (95% Cl); H0b: RERI = 0 | 0.04 (−0.36, 0.44) | 8.38×10−1 | ||
| Additive interaction: SI (95% CI); H0b: SI = 1 | 1.07 (0.41, 1.72) | 8.44×10−1 | ||
| Multiplicative interaction: IOR (95% CI); H0b: IOR = 1 | 1.01 (0.69, 1.33) | 9.55×10−1 | ||
Odds ratios are unadjusted and conditional on matching factors.
H0 = the null hypotheses tests were for RERI = 0; Synergy Index = 1 and interaction odds ratio = 1
DISCUSSION
In this study of three large prospective US cohorts of healthcare professionals which included 28,524 incident cholecystectomies over a combined follow-up period of greater than 4.1 million person-years, we confirmed associations2,6 for body mass index and other anthropometric measures with risk of symptomatic gallstone disease (defined by cholecystectomy). In the genotyped subset with a combined follow-up of 795,818 PYs, we documented 4,785 cholecystectomy cases. We assessed joint associations between 5 anthropometric markers and PRS in the cholecystectomy risk. We observed evidence that associations between BMI and risk of cholecystectomy was stronger in women compared to men but did not vary according to strata of genetic susceptibility on the relative risk scale. However, the absolute risk difference between obese and normal-weight individuals was higher in participants in the highest PRS category, compared to lowest PRS category, and these differences were stronger among women. This excess risk due to interaction was most pronounced among female carriers of at least one T allele at the ABCG8 SNP rs4245791 and having a BMI in the overweight or obese range.28 We did not observe a corresponding excess risk among male participants.
The presence of BMI-PRS interactions on the absolute risk scale suggests that an intervention to gradually reduce body weight to prevent symptomatic gallstone disease is likely to be more successful if it is targeted to those genetically predisposed to symptomatic gallstone disease risk, particularly among female participants. Among females in highest genetic susceptibility category, the 10-year cumulative risk of gallstone disease for an average female in the obese BMI range was 18.4% compared to 10.9% for an average female in the overweight BMI range. The corresponding risk reduction among women in the lowest genetic predisposition category would be expected to be from 10% to 5.1%.
To our knowledge, no previous study has comprehensively examined the joint associations of gallstone disease PRS and anthropometric measures in symptomatic gallstone disease risk. However, Buch et al. examined interactions between BMI and rs11887534 in ABCG8 as a secondary analysis in their discovery GWAS of gallstone disease and did not observe a significant interaction on the relative risk scale, which is consistent with our findings.10
The biological mechanism underlying the associations between BMI and gallstone disease and between ABCG5/8 and gallstone disease has been explored in previous studies; however, the mechanism underlying the interaction between BMI and ABCG5/8 is unclear. Mendelian randomization studies by Stender et al. and Yuan et al. found that genetic predisposition to high BMI were associated with a risk of gallstone disease, suggesting BMI is likely a causal risk factor.13,14 This association was more pronounced in female participants.14 Additionally, Yuan et al. reported a causal role for central obesity in the risk of gallstone disease.13 Variants at ABCG5/8 regulate the intestinal and hepatic efflux of cholesterol, and gain-of-function variation is thought to decrease levels of plasma cholesterol and increase levels of biliary cholesterol.29–32 Further research is warranted to examine whether evidence of statistical interaction on the absolute risk scale also implies the biological interaction between BMI and ABCG5/8 in gallstone disease risk. In clinical practice, our finding points to a subset of population that will most benefit from healthy weight, in prevention of symptomatic gallstone disease.
Strengths of our study include a large sample size, which provided the power to detect gene-environment interactions. In addition, prospective collection of exposure variables helped to minimize study bias, including recall bias, and comprehensive questionnaires with high validity allowed us to adjust for the broad range of possible confounders. This study also has notable limitations. First, our questionnaires did not distinguish between cholesterol and pigmented gallstones which have distinct etiological mechanisms. Second, we only included symptomatic gallstones that resulted in cholecystectomy. However, in the absence of universal screening for gallstones, it is not possible to capture all participants with asymptomatic gallstones. Moreover, self-reported gallstones without cholecystectomy was a very small minority (<5%) of all gallstone cases. This group is heterogeneous because they may include participants with incidental gallstone diagnosis as well as participants who were ineligible or refused surgery. Therefore, cholecystectomy is the more clinically relevant endpoint and suitable proxy for symptomatic gallstone disease.
In conclusion, greater BMI may confer higher excess risk of gallstone disease among individuals who are already genetically susceptible to gallstone disease risk, especially among females. These results suggest that maintenance of a healthy body weight may be effective in preventing cases of gallstone disease among all individuals, but the benefit may be higher among females and individuals with high genetic susceptibility to gallstone disease.
Supplementary Material
WHAT YOU NEED TO KNOW:
Background
Anthropometric measures have been associated with gallstone disease; however, whether the associations differ within the strata of genetic susceptibility remains largely unknown.
Findings
In three large, prospective US based cohorts, we found that the absolute risk difference between obese and normal-weight individual was higher among those with greater genetic susceptibility.
Implications for patient care
Maintenance of a healthy body weight may reduce the risk of gallstone disease primarily among the subset of the population with greater genetic susceptibility.
Grant support:
This work was supported by the National Institute of Health grants (UM1 CA186107, U01 CA176726, U01 CA167552, R01 CA49449, R01 CA67262), NIDDK K01-DK110267 to ADJ. and NIDDK K24-DK098311 to ATC.
Abbreviations:
- BMI
body mass index
- CI
confidence interval
- GxE
Gene-environment interactions
- GWAS
genome wide association study
- HPFS
Health Professionals Follow-up Study
- HR
hazard ratio
- IOR
interaction odds ratio
- NHS
Nurses’ Health Study
- OR
odds ratio
- PRS
polygenic risk score
- PY
person-year
- RERI
relative excess risk due to interaction
- SD
standard deviation
- SI
synergy index
- SNP
single-nucleotide polymorphism
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Potential competing interests: All authors declare no conflict of interest.
Data transparency statement: Individual participant data will not be shared.
REFERENCES
- 1.Shaffer EA. Epidemiology of gallbladder stone disease. Best Pract Res Clin Gastroenterol. 2006;20(6):981–996. doi: 10.1016/j.bpg.2006.05.004 [DOI] [PubMed] [Google Scholar]
- 2.Field AE, Coakley EH, Must A, et al. Impact of overweight on the risk of developing common chronic diseases during a 10-year period. Arch Intern Med. 2001;161(13):1581–1586. [DOI] [PubMed] [Google Scholar]
- 3.Leitzmann MF, Tsai C-J, Stampfer MJ, et al. Alcohol consumption in relation to risk of cholecystectomy in women. Am J Clin Nutr. 2003;78(2):339–347. [DOI] [PubMed] [Google Scholar]
- 4.Aune D, Vatten LJ, Boffetta P. Tobacco smoking and the risk of gallbladder disease. Eur J Epidemiol. 2016;31(7):643–653. doi: 10.1007/s10654-016-0124-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Shabanzadeh DM, Sørensen LT, Jørgensen T. Determinants for gallstone formation - a new data cohort study and a systematic review with meta-analysis. Scand J Gastroenterol. 2016;51(10):1239–1248. doi: 10.1080/00365521.2016.1182583 [DOI] [PubMed] [Google Scholar]
- 6.Tsai C-J, Leitzmann MF, Willett WC, Giovannucci EL. Weight cycling and risk of gallstone disease in men. Arch Intern Med. 2006;166(21):2369–2374. doi: 10.1001/archinte.166.21.2369 [DOI] [PubMed] [Google Scholar]
- 7.Tsai C-J, Leitzmann MF, Willett WC, Giovannucci EL. Central adiposity, regional fat distribution, and the risk of cholecystectomy in women. Gut. 2006;55(5):708–714. doi: 10.1136/gut.2005.076133 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Klass DM, Bührmann K, Sauter G, et al. Biliary lipids, cholesterol and bile synthesis: different adaptive mechanisms to dietary cholesterol in lean and obese subjects. Aliment Pharmacol Ther. 2006;23(7):895–905. doi: 10.1111/j.1365-2036.2006.02836.x [DOI] [PubMed] [Google Scholar]
- 9.Joshi AD, Andersson C, Buch S, et al. Four Susceptibility Loci for Gallstone Disease Identified in a Meta-analysis of Genome-Wide Association Studies. Gastroenterology. 2016;151(2):351–363.e28. doi: 10.1053/j.gastro.2016.04.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Buch S, Schafmayer C, Völzke H, et al. A genome-wide association scan identifies the hepatic cholesterol transporter ABCG8 as a susceptibility factor for human gallstone disease. Nat Genet. 2007;39(8):995–999. doi: 10.1038/ng2101 [DOI] [PubMed] [Google Scholar]
- 11.Sarin SK, Negi VS, Dewan R, Sasan S, Saraya A. High familial prevalence of gallstones in the first-degree relatives of gallstone patients. Hepatol Baltim Md. 1995;22(1):138–141. [PubMed] [Google Scholar]
- 12.Katsika D, Grjibovski A, Einarsson C, Lammert F, Lichtenstein P, Marschall H-U. Genetic and environmental influences on symptomatic gallstone disease: a Swedish study of 43,141 twin pairs. Hepatol Baltim Md. 2005;41(5):1138–1143. doi: 10.1002/hep.20654 [DOI] [PubMed] [Google Scholar]
- 13.Yuan S, Gill D, Giovannucci EL, Larsson SC. Obesity, Type 2 Diabetes, Lifestyle Factors, and Risk of Gallstone Disease: A Mendelian Randomization Investigation. Clin Gastroenterol Hepatol Off Clin Pract J Am Gastroenterol Assoc. Published online January 6, 2021. doi: 10.1016/j.cgh.2020.12.034 [DOI] [PubMed] [Google Scholar]
- 14.Stender S, Nordestgaard BG, Tybjærg-Hansen A. Elevated body mass index as a causal risk factor for symptomatic gallstone disease: A Mendelian randomization study. Hepatology. 2013;58(6):2133–2141. doi: 10.1002/hep.26563 [DOI] [PubMed] [Google Scholar]
- 15.Rothman KJ. Synergy and antagonism in cause-effect relationships. Am J Epidemiol. 1974;99(6):385–388. [DOI] [PubMed] [Google Scholar]
- 16.Siemiatycki J, Thomas DC. Biological models and statistical interactions: an example from multistage carcinogenesis. Int J Epidemiol. 1981;10(4):383–387. [DOI] [PubMed] [Google Scholar]
- 17.Knol MJ, Egger M, Scott P, Geerlings MI, Vandenbroucke JP. When one depends on the other: reporting of interaction in case-control and cohort studies. Epidemiol Camb Mass. 2009;20(2):161–166. doi: 10.1097/EDE.0b013e31818f6651 [DOI] [PubMed] [Google Scholar]
- 18.Bao Y, Bertoia ML, Lenart EB, et al. Origin, Methods, and Evolution of the Three Nurses’ Health Studies. Am J Public Health. 2016;106(9):1573–1581. doi: 10.2105/AJPH.2016.303338 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lindström S, Loomis S, Turman C, et al. A comprehensive survey of genetic variation in 20,691 subjects from four large cohorts. PLoS ONE. 2017;12(3). doi: 10.1371/journal.pone.0173997 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Richardson DB. An incidence density sampling program for nested case-control analyses. Occup Environ Med. 2004;61(12):e59. doi: 10.1136/oem.2004.014472 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Leitzmann MF, Rimm EB, Willett WC, et al. Recreational physical activity and the risk of cholecystectomy in women. N Engl J Med. 1999;341(11):777–784. doi: 10.1056/NEJM199909093411101 [DOI] [PubMed] [Google Scholar]
- 22.Thaler RH. Anomalies: The Winner’s Curse. J Econ Perspect. 1988;2(1):191–202. doi: 10.1257/jep.2.1.191 [DOI] [Google Scholar]
- 23.Joshi AD, Lindström S, Hüsing A, et al. Additive Interactions Between Susceptibility Single-Nucleotide Polymorphisms Identified in Genome-Wide Association Studies and Breast Cancer Risk Factors in the Breast and Prostate Cancer Cohort Consortium. Am J Epidemiol. 2014;180(10):1018–1027. doi: 10.1093/aje/kwu214 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Maclure KM, Hayes KC, Colditz GA, Stampfer MJ, Speizer FE, Willett WC. Weight, Diet, and the Risk of Symptomatic Gallstones in Middle-Aged Women. N Engl J Med. 1989;321(9):563–569. doi: 10.1056/NEJM198908313210902 [DOI] [PubMed] [Google Scholar]
- 25.Stampfer MJ, Maclure KM, Colditz GA, Manson JE, Willett WC. Risk of symptomatic gallstones in women with severe obesity. Am J Clin Nutr. 1992;55(3):652–658. doi: 10.1093/ajcn/55.3.652 [DOI] [PubMed] [Google Scholar]
- 26.Cole SR, Hernan MA. Constructing Inverse Probability Weights for Marginal Structural Models. Am J Epidemiol. 2008;168(6):656–664. doi: 10.1093/aje/kwn164 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Han SS, Rosenberg PS, Garcia-Closas M, et al. Likelihood Ratio Test for Detecting Gene (G)-Environment (E) Interactions Under an Additive Risk Model Exploiting G-E Independence for Case-Control Data. Am J Epidemiol. 2012;176(11):1060–1067. doi: 10.1093/aje/kws166 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Knol MJ, VanderWeele TJ, Groenwold RHH, Klungel OH, Rovers MM, Grobbee DE. Estimating measures of interaction on an additive scale for preventive exposures. Eur J Epidemiol. 2011;26(6):433–438. doi: 10.1007/s10654-011-9554-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Stender S, Frikke-Schmidt R, Nordestgaard BG, Tybjærg-Hansen A. The ABCG5/8 Cholesterol Transporter and Myocardial Infarction Versus Gallstone Disease. J Am Coll Cardiol. 2014;63(20):2121–2128. doi: 10.1016/j.jacc.2013.12.055 [DOI] [PubMed] [Google Scholar]
- 30.Graf GA, Yu L, Li W-P, et al. ABCG5 and ABCG8 are obligate heterodimers for protein trafficking and biliary cholesterol excretion. J Biol Chem. 2003;278(48):48275–48282. doi: 10.1074/jbc.M310223200 [DOI] [PubMed] [Google Scholar]
- 31.Yu L, Gupta S, Xu F, et al. Expression of ABCG5 and ABCG8 Is Required for Regulation of Biliary Cholesterol Secretion. J Biol Chem. 2005;280(10):8742–8747. doi: 10.1074/jbc.M411080200 [DOI] [PubMed] [Google Scholar]
- 32.Grünhage F, Acalovschi M, Tirziu S, et al. Increased gallstone risk in humans conferred by common variant of hepatic ATP-binding cassette transporter for cholesterol. Hepatology. 2007;46(3):793–801. doi: 10.1002/hep.21847 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.