Microbiome mediation analysis: methods, assumptions, and practical considerations

Microbiome mediation analysis provides a principled framework for understanding how environmental, behavioral, or clinical exposures influence human health through microbiomemediated biological pathways. However, its application is complicated by the compositional, sparse, and high-dimensional nature of microbiome data. A growing body of methods has been developed to address these challenges, drawing on structural equation modeling, counterfactual causal inference, distance-based testing, Bayesian variable selection, and nonparametric approaches. This paper reviews methodological developments designed to address these challenges and enable valid and interpretable mediation analysis in microbiome studies with particular emphasis on their underlying assumptions, limitations, and appropriate contexts of use, while also highlighting existing gaps and outlining future research directions.