Sensitivity Analysis For Causal Mediation Analysis With Mendelian Randomization

The goal of the mediator analysis is to identify whether the effect of an exposure on an outcome are influenced by some mediators.Traditional mediation analysis experiences a number of challenges,including bias due to measurement error and unmeasured confounders between an exposure,mediator,and outcome.Mendelian randomization(MR)is widely used in causal mediation analysis to control unmeasured confounding effects,which is valid under some strong assumptions.However,the MR assumptions are typically hard to satisfy in real studies.It is thus of great interest to assess the impact of violation of these MR assumptions through sensitivity analysis.Sensitivity analyses have been conducted for simple MR-based causal average effect analyses,but they are not available for MR-based mediation analysis studies.This paper aims to fill this gap,and two sensitivity parameters are proposed to quantify the effect due to the deviation of the MR assumptions.With these two sensitivity parameters,consistent indirect causal effect estimators are derived for both individual data and summary data,together with their asymptotic properties.Moreover,counterparts for the direct causal effect.These theoretical results can be used in MR-based mediation analysis to study the impact of violation of MR assumptions.The finite sample performance of the proposed method is illustrated through simulation studies,sensitivity analysis,and the application to a real genome-wide association study.