孟德尔随机化敏感性分析：如何运用MR-Egger与MR-PRESSO检测多效性偏倚

We employed multiple methods for sensitivity analysis to assess the robustness of causal effect estimates and detect potential pleiotropic bias. First, we evaluated directional pleiotropy using the MR-Egger intercept test; a significant intercept term (P < 0.05) would indicate the presence of pleiotropy. Second, we conducted leave-one-out analysis by sequentially removing each single nucleotide polymorphism (SNP) and re-running the inverse variance weighted (IVW) analysis to determine whether the causal effect is driven by a single SNP, thereby assessing the stability of the results. Third, we performed single-SNP analyses to visualize the independent contribution of each SNP to the causal effect. Finally, we used MR-PRESSO to detect horizontal pleiotropy and outlier SNPs; this method can identify and exclude SNPs that may introduce bias and provide corrected causal effect estimates. Through the comprehensive evaluation of these sensitivity analyses, we further validated the reliability of the causal associations.