ABSTRACT Mendelian randomization (MR) is a method that utilizes genetic variants as instrumental variables to determine causal relationships between exposures and outcomes, thereby mitigating confounding bias and reverse causality inherent in observational studies. Rooted in Mendel's laws of inheritance, MR has undergone rapid development since Katan first introduced the concept in 1986, followed by the formalization of its methodology by George Davey Smith and Ebrahim in 2003. This review comprehensively summarizes the theoretical foundations, methodological innovations, and expanding applications of MR. This review discusses the three core assumptions underpinning MR methodologies—relevance, independence, and exclusion restriction—and examines advanced methodological extensions, two‐sample MR, multivariable MR, bidirectional MR, non‐linear MR, cis ‐MR, mediation MR, multi‐population MR, and cluster MR. It further provides an overview of key commonly used R packages, databases, and analytical workflows that facilitate the implementation of MR. Application domains spanning gene–environment interactions research, public health, complex diseases, drug target validation, and integration with other cutting‐edge technologies are highlighted through representative case studies demonstrating their translational potential. Lastly, we critically assess methodological limitations, including weak instrument bias, horizontal pleiotropy, population stratification, data quality heterogeneity, lack of benchmark validation and comparison across MR methods, as well as selection and collider bias, while proposing future directions for improving robustness and expanding the applicability of MR through integration with multi‐omics and cross‐ancestry analyses. Overall, MR serves as a cornerstone in modern causal inference research, bridging genetics, epidemiology, and precision medicine to advance our understanding of disease etiology and therapeutic innovation.
Shen et al. (Fri,) studied this question.