Title:Inverse Probability Weighting-based Mediation Analysis for Microbiome Data

Abstract:Mediation analysis is an important tool for studying causal associations in biomedical and other scientific areas and has recently gained attention in microbiome studies. Using a microbiome study of acute myeloid leukemia (AML) patients, we investigate whether the effect of induction chemotherapy intensity levels on infection status is mediated by microbial taxa abundance. The unique characteristics of the microbial mediators -- high-dimensionality, zero-inflation, and dependence -- call for new methodological developments in mediation analysis. The presence of an exposure-induced mediator-outcome confounder, antibiotic use, further requires a delicate treatment in the analysis. To address these unique challenges in our motivating AML microbiome study, we propose a novel nonparametric identification formula for the interventional indirect effect (IIE), a recently developed measure for assessing mediation effects. We develop a corresponding estimation algorithm using the inverse probability weighting method. We also test the presence of mediation effects via constructing the standard normal bootstrap confidence intervals. Simulation studies demonstrate that the proposed method has good finite-sample performance in terms of IIE estimation accuracy and the type-I error rate and power of the corresponding tests. In the AML microbiome study, our findings suggest that the effect of induction chemotherapy intensity levels on infection is mainly mediated by patients' gut microbiome.