2.4 Mendelian randomization analysis and sensitivity analysis
MR-Weighted mode, weighted median, inverse variance weighted (IVW), and MR-Egger regression were adopted to evaluate the significant causal association between gut microbiota and the risk of MN. Because IVW combined the Wald ratio estimates of all relevant instrumental variables, we selected IVW as the main analysis method if the results of these four methods were inconsistent [6].
Cochran’s Q statistic was performed to estimate the potential heterogeneity. IVW method was used to decrease the effect of heterogeneity on the causal effect if there was remarkable heterogeneity [16]. P > 0.05 was regarded as no heterogeneity. Moreover, the MR-Egger regression test was employed to detect horizontal pleiotropy, and the results with P < 0.05 were excluded [17]. In addition, the MR-PRESSO test was also performed to find out the outliers and adjust for the heterogeneity. Finally, we conducted the leave-one-out analysis to identify and remove the potential outliers that might influence the results.