위염

Background: Observational evidence has shown that smoking, alcohol consumption, type 2 diabetes, and body mass index (BMI) are risk factors for peptic ulcer disease (PUD), including gastric ulcer (GU) and duodenal ulcer (DU). However, the observed associations may be confounding factors. Herein, we use Mendelian randomization (MR) to examine causal associations such as smoking, alcohol, type 2 diabetes, BMI, and risks of PUD.

Methods: We used 8,17,41,325,82, 231, and 616 identified genetic variants as proxies for age of smoking initiation (AgeSmk), smoking cessation (SmkCes, current/former), number of cigarettes smoked per day (CigDay), smoking status (SmkIni, ever/never), alcohol consumption, type 2 diabetes, and BMI to obtain unconfounded effect estimates on the GU and DU levels among 452,264 participants from the Gene ATLAS. The causal relationship was estimated by using inverse-variance weighted (IVW) as the main method. Sensitivity analysis includes Cochran’s Q test, the MR-Egger test, MR pleiotropy residual sum and outlier (MR-PRESSO), and MR-robust adjusted profile score (MR-RAPS). In addition, secondary MR analysis was conducted within summary data using genetic risk scores (GRSs) as instrumental variables (IVs).

Results: In our two-sample MR analyses, genetic predisposition to smoking (SmkInit) and BMI were associated with an increased risk of GU. The beta values were 0.0035 (95% CI, 0.0021, 0.0049, p = 1.56E-06) for smoking (SmkInit) and 0.0021 (95% CI, 0.0009, 0.0033, p = 0.0008) for BMI. Genetic predisposition to smoking (SmkInit) and higher genetically predicted BMI were associated with an increased risk of DU. The beta values of DU were 0.0029 (95% CI, 0.0017, 0.0041, p = 2.43E-06) for smoking (SmkInit) and 0.0018 (95% CI, 0.0007, 0.0029, p = 0.001) for BMI. No other causal association between smoking (AgeSmk, CigDay, and SmkCes), alcohol consumption, type 2 diabetes, and GU or DU was observed. Consistent results were obtained in sensitivity analyses. Furthermore, the GRS approach showed similar results in the several MR methods.

Conclusion: These findings do not support a causal role of AgeSmk, CigDay, SmkCes, alcohol consumption, and type 2 diabetes in the development of GU and DU. However, it is confirmed that SmkInit and BMI have a causal part in the development of GU and DU.

3.1 Instrumental variables and their validity: A total of 231 independent SNPs for type 2 diabetes, 82 independent SNPs for alcohol consumption, 616 independent SNPs for BMI, and 8, 41, 17, and 325 SNPs for the AgeSmk, CigDay, SmkCes, and SmkInit of smoking phenotypes, respectively, were included as IVs (Supplementary Tables S1–S7).

3.2 Mendelian randomization / 3.2.1 MR results of smoking to GU and DU: Genetic predisposition to SmkInit was associated with an increased risk of GU and DU. These associations are consistent across multiple MR methods. The beta and corresponding 95% CIs of GU were 0.0035 (95% CI, 0.0021, 0.0049) for one SD increase for SmkInit and DU were 0.0029 (95% CI, 0.0017, 0.0041) for one SD increase for SmkInit (Table 1, Supplementary Figures S1–S2). Genetic predisposition to AgeSmk, CigDay, and SmkCes was not associated with GU and DU susceptibilities (Table 1, Supplementary Figures S1–S2).

3.2.2 MR results of alcohol consumption to GU and DU: Increased alcohol consumption had no discernible influence on the risk of GU and DU (beta = -0.0030, 95% CI = -0.0080, 0.0021, p = 0.25; beta = 0.0027, 95% CI = -0.0016, 0.0070, p = 0.21, respectively) (Table 1). The MR-Egger regression and Cochran’s Q test did not reveal any evidence that supported horizontal pleiotropy and heterogeneity. MR-PRESSO shows an SNP with horizontal pleiotropic, with one SNP identified as outliers (rs281379). When we removed rs281379, our results were stable, and neither horizontal pleiotropic nor heterogeneity was detected (Table 3). The results of all methods were consistent, and MR-PRESSO did not show the existence of pleiotropic SNPs (Table 1, Supplementary Figures S1–S2).

3.2.3 MR results of type 2 diabetes to GU and DU: Genetic predisposition to type 2 diabetes was not associated with an increased risk of GU and DU. These associations are consistent across multiple MR methods. The beta and corresponding 95% CIs of GU were 0.0002 (95% CI, -0.0003, 0.0007) for type 2 diabetes and DU were 0.0002 (95% CI, -0.0002, 0.0007) for type 2 diabetes (Table 1, Supplementary Figures S1–S2). The MR-Egger regression intercept provided no support for directional pleiotropy (Table 3). The Cochran’s Q test did show some heterogeneity among individual SNPs effect estimates in type 2 diabetes on DU; therefore, a random-effects IVW model would be implemented. Random-effects IVW model results were consistent with other methods (fixed-effects IVW model weighted median estimator, simple median estimator, MR-RAPS, and MR-PRESSO) (Table 1, Supplementary Figures S1–S2).

3.2.4 MR results of BMI to GU and DU: There was a link between a higher risk of GU and DU and a genetic tendency to BMI. The beta and corresponding 95% CIs of GU were 0.0021 (95% CI, 0.0009, 0.0033) for one SD increase for BMI and DU were 0.0018 (95% CI, 0.0007, 0.0029) for one SD increase for BMI (Table 1, Supplementary Figures S1–S2). The Cochran’s Q test and MR-Egger regression intercept for the relationship between BMI and GU did not reveal any evidence of directional pleiotropy or heterogeneity, and the Cochran’s Q test and MR-Egger regression intercept did show heterogeneity and directional pleiotropy among individual SNPs effect estimates in BMI on DU (Table 3). However, no SNPs were identified as an outlier in the MR-PRESSO method in the BMI to DU MR analysis. In addition, random-effects IVW model results were consistent with other methods (fixed-effects IVW model, weighted median estimator, simple median estimator, MR-RAPS, and MR-PRESSO) (Table 1, Supplementary Figures S1–S2).

3.3 GRSsmoking, GRSalcohol, GRStype 2 diabetes, and GRSBMI with GU and DU: Consistent with the MR results of smoking (AgeSmk, CigDay, SmkCes, and Smklnit), alcohol consumption, type 2 diabetes, and BMI to GU and DU, the GRS revealed no causal effect of smoking (AgeSmk, CigDay, and SmkCes), alcohol consumption, and type 2 diabetes on GU and DU risks (Table 2). The GRS revealed significant causal effect of smoking (Smklnit) and BMI on GU and DU (Table 2). The beta and corresponding 95% CIs of GU were 0.0035 (95% CI, 0.0022, 0.0048) for one SD increase for Smklnit and DU were 0.0029 (95% CI, 0.0018, 0.0041) for one SD increase for Smklnit (Table 2, Supplementary Figure S3). The beta and corresponding 95% CIs of GU were 0.0021 (95% CI, 0.0009, 0.0033) for one SD increase for BMI and DU were 0.0018 (95% CI, 0.0008, 0.0028) for one SD increase for BMI (Table 2, Supplementary Figure S3).

3.4 Heterogeneity and sensitivity analysis: Cochran’s Q tests revealed that there was no heterogeneity between the smoking, alcohol consumption, type 2 diabetes, and BMI IVs in the study of smoking, alcohol consumption, type 2 diabetes, BMI, and GU MR (Table 3). In smoking and alcohol consumption to DU MR analysis, the results of Cochran’s Q tests showed that there was no heterogeneity among the smoking and alcohol consumption-related IVs (Table 3). In type 2 diabetes and BMI to DU MR analysis, the Cochran’s Q tests revealed some heterogeneity between both the IVs for type 2 diabetes (Q = 284.7, p = 0.01) and BMI (Q = 689.6, p = 0.02) (Table 3); however, random-effects IVW model results were consistent with other methods (fixed-effects IVW model, weighted median estimator, simple median estimator, MR-RAPS, and MR-PRESSO). Therefore, our results are still very reliable and this little heterogeneity has little effect on our results.

The MR-Egger regression analysis revealed that alcohol consumption, type 2 diabetes, smoking, and BMI to GU MR did not indicate horizontal pleiotropy of the IVs (Table 3), and the horizontal pleiotropy of the IVs was not present for alcohol consumption, smoking (CigDay, SmkCes, and Smklnit), and type 2 diabetes to DU MR and it was present for smoking (AgeSmk, MR-Egger p = 0.01) and BMI (MR-Egger p = 0.049) (Table 3). The MR-PRESSO global test p was >0.05 in smoking (AgeSmk) and global test p was <0.05 in BMI to DU MR, with no SNPs identified as outliers (Table 3).

3.5 Excluding associations between individual exposure factors: To exclude associations between individual exposure factors, we searched the phenoscanner website for SNPs of each exposure factor. SNPs that were strongly correlated with other exposure factors were eliminated. Specific screening results of SNPs for each exposure factor are presented in Supplementary Table S8. After removing these SNPs, which were strongly correlated with other exposure factors, we performed MR analysis again, and we got the same results as before (Supplementary Table S9). Therefore, our results can be considered very reliable.