Causal Mediation Analysis: A Summary-Data Mendelian Randomization Approach.

ABSTRACT: Summary-data Mendelian randomization (MR), a widely used approach in causal inference, has recently attracted attention for improving causal mediation analysis. Two existing methods corresponding to the difference method and product method of linear mediation analysis have been developed to perform MR-based mediation analysis using the inverse-variance weighted method (MR-IVW). Despite these developments, there is still a need for more rigorous, efficient, and precise MR-based mediation methodologies. In this study, we develop summary-data MR-based frameworks for causal mediation analysis. We improve the accuracy, statistical efficiency and robustness of the existing MR-based mediation analysis by implementing novel variance estimators for the mediation effects, deriving rigorous procedure

SUBMITTER: Lin SC 

PROVIDER: S-EPMC11799828 | biostudies-literature | 2025 Feb

REPOSITORIES: biostudies-literature