A Review of Mendelian Randomization in the Presence of Weak Instrumental Variables; Statistical Methods and Challenges

Abstract

Introduction: Mendelian randomization (MR) assesses the causal effect of risk factors by using genetic variations as instrumental variables (IV) in nonexperimental data. IV strategies are one of a few available methods for determining causal effects in the absence of specific knowledge of all confounders in the exposure-outcome relationship. To use an IV as a legitimate instrumental variable, it must meet the following criteria: relevance, exchangeability, and exclusion restriction. A weak instrument is a circumstance in which there is a piece of weak statistical evidence for an association between IV and exposure. Weak instruments cause significant issues, including (i) insufficient statistical power to hypothesis testing, (ii) increasing bias with deviation from IV assumptions, and (iii) asymptotic estimation of standard errors and confidence intervals. Therefore, in this study, we intend to introduce the Mendelian randomization method, weak instrumental bias, and statistical remedy methods used in this bias. Methods: Current study was conducted by using Medline/PubMed, Scopus, Web of Sciences and Google Scholar. Results: This review provides a comprehensive description of the principles of MR, and a guide to basic MR methodology. To deal with these challenges, the bulk of the review considered statistical remedies. The review ends with a section that details the practical limitations, and recommendations regarding MR and the weak instruments. Conclusion: Depending on the type of data, several solutions can be used in one and multiple IV. Moreover, it can be used in solutions in the design and analysis phase to minimize the effects of weak instruments. © 2024 Tehran University of Medical Sciences.