Selection of instrumental variables

In the two-sample MR analysis of the present study, the MR approach we used must meet the following three assumptions [25]: (a) genetic variants must be closely linked to the exposure (PA); (b) genetic variants wouldn’t be influenced by other factors; (c) genetic variants must only be related to the risk of outcomes (OA) through exposures (PA), not through other pathways. Schematic diagram of Mendelian randomization study is shown in Fig. 1.

Schematic diagram of Mendelian randomization study. This design is under the hypothesis that genetic instrumental variables are associated with exposure, but not with confounders, and the genetic instrumental variable affect outcome only through exposure. Genetic instrumental variables represent single nucleotide polymorphisms

SNPs (P < 5 × 10^-8) associated with a corresponding exposure at the genome-wide significance threshold were extracted in our current study. To ensure independence among each selected SNP constituting each genetic instrument, a Linkage disequilibrium analysis of the corresponding SNPs for each instrumental variable (r2 = 0.001 and kb = 10,000 kb) was applied. Then, the PhenoScanncer GWAS database was used to assess and delete SNPs correlated to potential confounding factors (including smoking, drinking, and BMI) and outcomes.