L or M1—Critical Challenges in Mediation Analysis

Methods for causal mediation analysis have developed dramatically over the past few decades.1–7 In the causal mediation literature, several causal quantities—or estimands—have been proposed, including natural direct and indirect effects, interventional direct and indirect effects, and separable direct and indirect effects. As another possible causal estimand, Chen and Lin8 proposed separable path-specific effects, which is an extension of the separable effects framework to cases that involve multiple ordered mediators. In this commentary, I briefly discuss the newly proposed method from a broader perspective on causal mediation analysis. For readers less familiar with common causal mediation approaches, please see related literature.1–3,9–11

This is a non-final version of an article published in final form in Suzuki, Etsujia. L or M1—Critical Challenges in Mediation Analysis. Epidemiology 36(5):p 686-689, September 2025. | DOI: 10.1097/EDE.0000000000001888.

This fulltext file will be available in Sep. 2026.