Causal Mediation Analysis with Multiple Causally Ordered and Non-ordered Mediators based on Summarized Genetic Data

Abstract Causal mediation analysis aims to investigate the mechanism linking an exposure and an outcome. Dealing with the impact of unobserved confounders among the exposure, mediator and outcome has always been an issue of great concern. Moreover, when multiple mediators exist, this causal pathway intertwines with other causal pathways, making it more difficult to estimate of path-specific effects (PSEs). In this article, we propose a method (PSE-MR) to identify and estimate PSEs of an exposure on an outcome through multiple causally ordered and non-ordered mediators using Mendelian Randomization, when there are unmeasured confounders among the exposure, mediators and outcome. Additionally, PSE-MR can be used when pleiotropy exists, and can be implemented using only summarized genetic data. We also conducted simulations to evaluate the finite sample performances of our proposed estimators in different scenarios. The results show that the causal estimates of PSEs are almost unbiased with good coverage and Type I error properties. We illustrate the utility of our method through a study of exploring the mediation effects of lipids in the causal pathways from body mass index to cardiovascular disease. Author summary A new method (PSE-MR) is proposed to identify and estimate PSEs of an exposure on an outcome through multiple causally ordered and non-ordered mediators using summarized genetic data, when there are unmeasured confounders among the exposure, mediators and outcome. Lipids play important roles in the causal pathways from body mass index to cardiovascular disease