2.3 Statistical analysis
We used the “TwoSampleMR” and “MRPRESSO”, “mr.raps” packages of
the R software (version 4.2.2) to perform MR analysis. Multiple MR
approaches were utilized to explore
causal relationships between exposure and outcome. Causal effect
estimates from the IVW was considered the criterion for establishing a
causal relationship between exposure and outcome(33). It is suggested
that if the IVs meet the three major assumption of MR and the selected
SNPs are all valid IVs, result from IVW model provides an efficient and
reliable estimate for the causal effect(34). As an extension to IVW,
MR-PRESSO approach attempts to remove outliers based on their
contributions to heterogeneity, while other MR Methods were regarded as
a supplement and reference(35). After MR analysis, we assessed the
quality of the harmonization and whether the variants were difficult to
harmonize by performing Cochran’s Q test and MR Egger regression
analysis, as well as leave-one-out method. The IVW model was applied to
the heterogeneity analysis, where Cochran’s Q statistic provided
evidence for heterogeneity and invalid instruments, with p value
greater than 0.05 indicating the absence of heterogeneity(34). The
intercept of MR-Egger regression was used to explore and account for the
impact of horizontal pleiotropy, and p value greater than 0.05 is
evidence for credible instrumental variables(36). To further assess the
reliability of the MR results, we evaluated the impact of one single SNP
for estimates by performing the leave-one-out method, in which new MR
results was recalculated after leaving out each SNP in turn. Finally, we
visualized the main results of the MR analysis using scatter plots,
forest plots, and funnel plots. Figure 1B shows a detailed flowchart of
this MR study.
Results