Mendelian randomization analysis

In this study, fixed and random effect inverse variance weighted (IVW), weighted median and MR-Egger regression methods were used for MR analysis (Burgess et al., 2013). If all SNPS selected are valid instrumental variables, IVW method can provide unbiased estimation (Burgess, et al., 2013). Assuming that no more than 50% of the estimated MR effect weights are derived from multi-effect SNPS, the weighted median method gives consistent effect estimates, where the weights are determined by the strength of their relationship to exposure (Bowden et al., 2016). The MR-Egger regression examines the horizontal pleiotropy of instrumental variables and provide estimates after correcting for pleiotropy (Bowden et al., 2015). If the intercept of the regression equation is 0 or the p-value of the intercept term is not significant, no horizontal pleiotropy is suggested. MR pleiotropy residual sum and outlier (MR-PRESSO) method was used to detect the presence of horizontal outliers. Cochran’s Q was used to verify the heterogeneity between the causal estimates of each SNPs in the IVW and MR-Egger methods. The leave-one-out method was used to exclude each SNP one by one and calculate the combined effect of remaining SNPs, so as to determine whether the causal association between exposure factors and outcome variables is caused by the main effect of an instrumental variable. R (Version 3.6.1) software with MR-PRESSO (Version 1.0) and Two-Sample MR (Version 0.5.5) packages were used for statistical analysis, and p < 0.05 was considered statistically significant.