2.4. Considerations for Key Assumption of MR

The MR analysis requires that three assumptions be met to demonstrate causal effects [13]. First, the relevance assumption is that the genetic instrument should be strongly associated with the exposure of interest, and the assumption was met by utilizing the SNPs reaching genome-wide significant level association. Second, the independence association is that the genetic instrument should not be associated with confounders. We performed well-known MR sensitivity analysis available for summary-level MR, relaxing this assumption for some of the instruments [19,20]. Third, the exclusion-restriction assumption is that the causal effects should be through the exposure of interest and cannot be formally tested. However, median-based sensitivity MR methods relax this assumption in up to 50% of the genetic instruments and are thus considered a sensitivity analysis for this assumption [20]. Further, we tested the causal effects from a single variant that is biologically proven for its effect on the homocysteine metabolic pathway and the causal estimates from the single-variant MR, which would be minimally biased from a horizontal pleiotropic pathway.