Journal Pre-proof A Randomization Investigation

Background & Aims Obesity, type 2 diabetes, and lifestyle factors (cigarette smoking, alcohol drinking, and coffee consumption) have been associated with the risk of developing gallstone disease in observational studies, but whether these associations are causal is undetermined. We conducted a Mendelian randomization study to assess these associations. Methods Genetic instruments associated with the exposures at the genome-wide significance ( p <5×10 −8 ) level were selected from corresponding genome-wide associations studies (n=224 459 to 1 232 091 individuals). Summary-level data for gallstone disease were obtained from the UK Biobank (10 520 cases and 350 674 non-cases) and FinnGen consortium (11 675 cases and 121 348 non-cases). Univariable and multivariable Mendelian randomization analyses were conducted. Results from UK Biobank and FinnGen were combined using fixed-effects meta-analysis. Results The odds ratios were 1.63 (95% confidence interval (CI), 1.49, 1.79) for one standard deviation (SD) increase in body mass index, 1.81 (95% CI, 1.60, 2.05) for one SD increase in waist circumference, 1.13 (95% CI, 1.09, 1.17) for one unit increase in the log-odds ratio of type 2 diabetes and 1.25 (95% CI, 1.16, 1.34) for one SD increase in prevalence of smoking initiation. The associations for body mass index and type 2 diabetes persisted after mutual adjustment. Genetically predicted coffee consumption was inversely associated with gallstone disease after adjustment for body mass index and smoking (odds ratio per 50% increase 0.44 , 95% CI, 0.21, 0.91). There was no association with alcohol consumption. Conclusions This study supports independent causal roles of obesity, type 2 diabetes, and smoking in gallstone disease.


Background
Gallstone disease affects around 10% to 20% of the global adult population and is associated with the highest socioeconomic costs of all gastrointestinal conditions. [1][2][3] Serious complications, such as cholecystitis, cholangitis, and pancreatitis, as well as gallstone recurrence influence health status and quality of life of gallstone patients after cholecystectomy. 4 Obesity, 1, 2, 4 type 2 diabetes 5-7 and cigarette smoking 8 have been proposed as risk factors, and alcohol 4,9,10 and coffee 4,11,12 consumption as protective factors for gallstones in traditional observational studies. However, the causality of these observational findings cannot be determined as residual confounding may have biased the results. For example, obesity and type 2 diabetes are closely interrelated and their independent association with cholelithiasis is uncertain. Likewise, cigarette smoking, alcohol drinking, and coffee consumption are overlapping behaviors, 13 and this may introduce residual confounding in observational studies.
Mendelian randomization (MR) leverages the random allocation of genetic variants during conception as instrumental variables to estimate the causal association between an exposure (e.g., body mass index (BMI)) and health outcome. 14 Here, the MR design was used to assess the potential causal associations of overall obesity (measured as BMI), abdominal obesity (measured as waist circumference), type 2 diabetes, and lifestyle factors (cigarette smoking, alcohol drinking, and coffee consumption) with risk of gallstone disease.

Study design and data sources
Study design overview and the assumptions of a Mendelian randomization study are shown in Figure 1. Genetic instruments for the exposures were obtained from published genomewide association studies. [15][16][17][18][19] Data for gallstone disease were obtained from UK Biobank 20 J o u r n a l P r e -p r o o f and the FinnGen consortium 21 . Detailed information on used data sources is displayed in Table 1. All studies had been approved by a relevant ethical review board and participants had given informed consent.

Genetic instrument selection
Genetic instruments for BMI 15 , waist circumference (with and without adjustment for BMI) 16 , cigarette smoking 17 , alcohol drinking 17 , coffee consumption 18 , and type 2 diabetes 19 were selected at genome-wide significance threshold (p <5×10 −8 ) from corresponding genomewide association studies. A set of genetic instruments associated with smoking index 22 (taking into account smoking status as well as smoking duration, heaviness, and cessation in ever smokers) were used in a supplementary analysis. Linkage disequilibrium (LD) among single-nucleotide polymorphisms (SNPs) for each risk factor was calculated based on 1000 genomes LD reference panel (European population) using the PLINK clumping method.
SNPs in LD (r 2 >0.01 and clump window <10 kb) were excluded and the SNP with the lowest p-value was retained. In the analysis of coffee consumption, five SNPs were excluded due to pleiotropic associations with other potential risk factors (in particular BMI) at genome-wide significance (rs1260326 in the GCKR gene, rs574367 in SEC16B, rs10865548 in TMEM18, rs66723169 in MC4R, and rs34060476 in MLXIPL). 23

Data sources for gallstone disease
Summary-level genetic data for gallstone disease (cholelithiasis, defined by the International Classification of Diseases 10th Revision code K80) were obtained from the UK Biobank cohort 20 and the FinnGen consortium 21 . The second wave of analyses of UK Biobank from Neale lab was used in the present study. Data from Neale lab included 361 194 UK Biobank participants after exclusion of individuals of non-European ancestry, closely related J o u r n a l P r e -p r o o f individuals (or at least one of a related pair of individuals), individuals with sex chromosome aneuploidies, and individuals who had withdrawn consent from the UK Biobank study.
Association tests had been adjusted for age, sex and up to 20 principal components. The R3 release of the FinnGen consortium data was used. In that dataset, individuals with ambiguous gender, high genotype missingness (>5%), excess heterozygosity (±4 standard deviation) and non-Finnish ancestry had been excluded. Association tests had been adjusted for age, sex, 10 genetic principal components and genotyping batch.

Statistical analysis
The multiplicative random-effects inverse-variance weighted model was used to obtain the MR estimates. Derived estimates based on UK Biobank and the FinnGen consortium were combined using fixed-effects meta-analysis. Multivariable random-effects inverse-variance weighted model was used to adjust for BMI in the analysis of type 2 diabetes and vice versa.
We also conducted a two-step multivariable MR analysis in the analysis of coffee consumption to adjust for BMI and smoking initiation. Likewise, multivariable MR analysis was performed to adjust for smoking initiation in the analysis of alcohol drinking as these two traits are genetically correlated (r g =0.34) 17 . Three other methods, including the weighted median, MR-Egger regression and MR-PRESSO methods, were used as sensitivity analyses. The weighted median model provides consistent estimates on the condition that ≥50% of the weight in the analysis comes from valid instrumental variables. 24 MR-Egger regression analysis can detect and correct for directional pleiotropy whereas it compromises power. 25 The p value for the MR-Egger intercept was used to indicate directional pleiotropy. The MR-PRESSO approach can detect outliers and generate estimates after outliers removing. 26 The MR-PRESSO distortion test aims at examining the differences between the estimates before and after outlier correction and a p<0.05 of distortion test indicates a significant difference in J o u r n a l P r e -p r o o f estimates before and after outlier correction. 26 Cochrane's Q value was estimated to assess the heterogeneity among used SNPs for each risk factor. Odds ratios (ORs) and corresponding confidence intervals (CIs) of gallstone disease were scaled to one standard deviation (SD) increase in BMI and waist circumference, one unit increase in log-odds ratio of type 2 diabetes, one SD increase in prevalence of smoking initiation, one SD increase of lifetime smoking index, one SD increase of log-transformed alcoholic drinks/week, and 50% increase in coffee consumption, respectively. All analyses were performed using the TwoSampleMR 27 , Mendelianrandomization 28 and MR-PRESSO 26 packages in R Software 3.6.0.

Results
Genetically predicted higher BMI, waist circumference, and liability to type 2 diabetes were associated with elevated risk of gallstone disease in UK Biobank, FinnGen and meta-analysis (p<0.001) (Figure 2 and Figure 3). The combined ORs of gallstones were 1.63 (95% CI, 1.49, 1.79) per one SD increase in BMI, 1.81 (95% CI, 1.60, 2.05) per one SD increase in waist circumference, and 1.13 (95% CI, 1.09, 1.17) for one unit increase in log-odds ratio of type 2 diabetes. The associations were consistent in all sensitivity analyses, though pleiotropy was detected in the analysis of type 2 diabetes in UK Biobank by the MR-Egger regression.
After removing outliers in the MR-PRESSO analysis, the association between type 2 diabetes and cholelithiasis persisted and the p value for the distortion test were above 0.05 ( Table 2).
Associations for BMI and type 2 diabetes persisted and attenuated slightly after mutual adjustment (Figure 2). The association between genetically predicted waist circumference and gallstone disease did not remain in UK Biobank, and attenuated greatly in FinnGen after adjusting for BMI (Figure 3).

J o u r n a l P r e -p r o o f
Among lifestyle factors, genetic predisposition to smoking initiation was associated with gallstones in the meta-analysis of data from UK Biobank and FinnGen (OR 1.25; 95% CI, 1.16, 1.34, for one SD increase in smoking prevalence) (Figure 3 and Table 2). This positive association was replicated in the sensitivity analysis of lifetime smoking index.
There were no associations of alcohol or coffee consumption with gallstones in the primary analysis (Figure 3) or the sensitivity analyses of each data source ( Table 2). After adjustment for BMI, coffee consumption showed an inverse association with gallstones (OR 0.50; 95% CI, 0.29, 0.88, per 50% increase in coffee consumption). This association remained after additional adjustment for smoking initiation (OR 0.44; 95% CI, 0.21, 0.91). In meta-analysis of estimates from UK Biobank and FinnGen, coffee consumption showed an association with gallstones in analysis based on the weighted median method with OR of 0.78 (95% CI, 0.63, 0.97). Genetically predicted alcohol drinking (the combined OR 1.06; 95% CI, 0.77, 1.47) was not associated with gallstones in the multivariable MR analysis adjusting for smoking initiation.

Discussion
The present MR study supports that both obesity and type 2 diabetes are independently and causally associated with the risk of gallstone disease. Moreover, our results provide support for a causal association between smoking and risk of gallstone disease. A possible causal association is observed between high coffee consumption and decreased risk of gallstone disease. There is no evidence that alcohol drinking is causally associated with gallstones.
BMI, representing overall obesity, has been associated with gallstones in observational studies 4,29 and in a one-sample MR study in 77 679 Danish adults, 30 which was corroborated by the present two-sample MR analysis in British and Finnish populations. The association between waist circumference and gallstones did not remain after adjustment for J o u r n a l P r e -p r o o f BMI, which indicated that central obesity is not a more important risk factor than overall fat mass for gallstone disease. The association between obesity and gallstones was slightly attenuated in the multivariable MR analysis with adjustment for genetically predicted type 2 diabetes liability, which may suggest that type 2 diabetes partly mediates this association.
Several mechanisms, including alterations of metabolic factors, hepatic secretion of supersaturated bile, dyslipidemia, intestinal and gallbladder hypomotility, gallbladder stasis, decreased secretion of bile acids, cholesterol crystallization and precipitation, and supersaturated gallbladder bile, may underlie the increased risk of gallstones among obese individuals. 31 Even though obesity plays a causal role in the development of gallstones, both bariatric and a low-calorie diet have been shown to increase the risk of asymptomatic and symptomatic gallstones. 32,33 Such increased risk might be related to a rapid weight loss. 34 Studies of the association of type 2 diabetes with risk of gallstones have not been entirely consistent. A meta-analysis of 10 cohort studies showed an overall 56% increased risk of gallstones among diabetes patients compared with individuals without diabetes 7 . An increased risk of gallstones in type 2 diabetes patients was also found in a cross-sectional study with more than 4 million diabetes patients 6 , but not in several other case-control studies 5,35 . In one of those case-control studies, type 2 diabetes was associated with gallstones in the univariable (crude) analysis, but the association was attenuated after adjustment for obesity and other risk factors 35 . The present MR study based on a large number of gallstone disease cases revealed a possibly causal association between type 2 diabetes and gallstones and the association was likely independent of BMI. Several possible mechanisms may explain the association between type 2 diabetes and gallstone disease. For example, hepatic insulin resistance has been shown to directly promote the formation of cholesterol gallstones. 36 Gallbladder hypomotility in diabetes might be an additional predisposing factor for cholesterol gallstone formation. 3

J o u r n a l P r e -p r o o f
Epidemiological data on the association between cigarette smoking and gallstones are inconsistent with positive 8,37 and null findings 38 reported. Our study provided evidence to support a causal role of cigarette smoking in the development of gallstone and further that the association is likely independent of BMI. The different effect sizes of smoking on gallstones in UK Biobank and FinnGen might be related to differences in smoking patterns. Metaanalyses on alcohol consumption in relation to gallstones revealed a dose-response inverse relationship 9, 10 and a J-shaped association (the risk of gallstones decreased with increasing alcohol consumption up to 30 g/day and thereafter plateaued) 10 . However, our study did not confirm the inverse association 4, 9 , but could not rule out a possible weak non-linear association of alcohol drinking with gallstones.
With regard to coffee consumption, an inverse association with risk of overall and symptomatic gallstone disease has been found in observational studies 4,11,12 and in a onesample MR study based on 114 220 Danish individuals (including 7294 cases) 39 . The present MR study supported a possible association between high coffee consumption and lower risk of gallstones. Potential mechanisms underlying this inverse association might be that components in coffee stimulate cholecystokinin release, enhance gallbladder motility, inhibit gallbladder fluid absorption, decrease cholesterol crystallization in bile, down-regulate the hepatic low density lipoprotein receptor and decrease 3-hydroxyl-3-methylglutaryl Co A reductase activity. 40 The major strengths of the present study are the MR design and the large number of gallstone disease cases. The similar results in two independent populations supported the robustness and reliability of our results for the associations of obesity and type 2 diabetes with gallstones. Genetic associations with the International Classification of Diseases-defined gallstone endpoint in UK Biobank has been replicated in both Icelandic and Danish cohorts, 41,42 which supported a high validity of gallstone definition in our study. Population J o u r n a l P r e -p r o o f stratification bias is unlikely to have influenced our results as the majority of recruited participants in the genome-wide association studies were of European origin and adjustment was made for population structure through genetic principal components. However, this population confinement might also limit the generalizability of our findings to other populations.
A limitation of the present study is that we could not distinguish between cholesterol and pigment gallstones, which have different etiologies. Nevertheless, the vast majority of gallstones in the Western world are cholesterol gallstones.
A potential concern in any MR study is pleiotropy. There are two types of pleiotropy in MR studies -vertical and horizontal pleiotropy. 43,44 In this study, genetic instruments for BMI influenced the risk of gallstone disease partly via type 2 diabetes, which would represent vertical pleiotropy (or mediated pleiotropy) as BMI was a causal risk factor for type 2 diabetes. 45 Vertical pleiotropy is not a barrier in causal inference of MR studies. 43,44 The genetic instruments for coffee consumption had moderate correlations with BMI and smoking. 18 Given that coffee consumption is less likely to cause obesity and smoking initiation, this phenomenon called horizontal pleiotropy would bias causal estimation. We found an inverse association between genetically predicted coffee consumption and gallstones after removal of pleiotropy by excluding pleiotropic genetic variants and after adjusting for BMI and smoking through multivariable MR analysis. 43,44 Considering the consistent associations of genetic predisposition to type 2 diabetes and smoking initiation with gallstone disease across the two data sources and different MR models, these results are unlikely to be driven by horizontal pleiotropy.
In conclusion, the present study provides MR evidence supporting causal roles of obesity, type 2 diabetes and cigarette smoking in gallstones. The possible association between coffee consumption and gallstone disease needs verification.  BMI, body mass index; IVW, inverse-variance weighted; LD, linkage disequilibrium; SNP, single-nucleotide polymorphisms; T2DM, type 2 diabetes mellitus; WC, waist circumference. Assumption 1 indicates that the genetic variants proposed as instrumental variables should be robustly associated with the risk factor of interest; assumption 2 indicates that the used genetic variants should not be associated with potential confounders, and assumption 3 indicates that the selected genetic variants should affect the risk of the outcome merely through the risk factor, not via alternative pathways. The Mendelian randomization design can reduce residual confounding and reverse causality, thereby reinforcing the causal inference of an exposure-outcome association. The basis of this is that genetic variants, selected as instrumental variable for studying the effect of modifying the exposure, are randomly allocated at conception and are therefore less vulnerable to confounding from environmental factors and reverse causation.  Associations of genetically predicted waist circumference and modifiable lifestyle factors with gallstone disease*. BMI indicates body mass index; CI, confidence interval; OR, odds ratio; UKBB, UK Biobank. Estimates for coffee consumption with adjustment for BMI were derived from the multivariable inverse-variance weighted model with random effects. The ORs of gallstone disease were scaled to one standard deviation increase in waist circumference, one standard deviation increase in prevalence of smoking initiation, one standard deviation increase of lifetime smoking index, one standard deviation increase of logtransformed alcoholic drinks/week and 50% increase in coffee consumption. *Estimates were derived using the multiplicative random-effects inverse-variance weighted Mendelian randomization method and combined using fixed-effects meta-analysis.  Obesity, type 2 diabetes, and lifestyle factors have been associated with the risk of developing gallstone disease in observational studies, but whether these associations are causal is undetermined.

Findings
Genetic predisposition to obesity, type 2 diabetes and smoking initiation was associated with an increased risk of gallstone disease. Genetically predicted coffee consumption but not alcohol consumption was associated with a reduced risk of gallstone disease.

Implications for patient care
Preventing obesity and diabetes and modifying lifestyle factors may act as prevention strategies for gallstone disease.