MethodsUnderstanding Interventional Effects: A More Natural Approach to Mediation Analysis?Moreno-Betancur, Margaritaa,b; Carlin, John B.a,bAuthor Information From the aClinical Epidemiology and Biostatistics Unit, Murdoch Children’s Research Institute, Melbourne, Australia bCentre for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, University of Melbourne, Melbourne, Australia. Submitted October 3, 2017; accepted May 28, 2018. Availability of data and code for replication: The R code required to reproduce the numerical example in the eAppendix can be accessed at the first author’s GitHub repository (https://github.com/moreno-betancur/Compare_mediation_effects/archive/master.zip). Supported by a Centre of Research Excellence grant from the Australian National Health & Medical Research Council, ID#1035261, to the Victorian Centre for Biostatistics (ViCBiostat). The Murdoch Children’s Research Institute is supported by the Victorian Government’s Operational Infrastructure Support Program. The authors report no conflicts of interest. Supplemental digital content is available through direct URL citations in the HTML and PDF version of this article (www.epidem.com). Correspondence: Margarita Moreno-Betancur, Clinical Epidemiology and Biostatistics Unit, Murdoch Children’s Research Institute, 50 Flemington Road, Parkville, Victoria 3052, Australia. E-mail: [email protected]. Epidemiology: September 2018 - Volume 29 - Issue 5 - p 614-617 doi: 10.1097/EDE.0000000000000866 Buy SDC Metrics Abstract The causal mediation literature has mainly focused on “natural effects” as measures of mediation, but these have been criticized for their reliance on empirically unverifiable assumptions. They are also impossible to estimate without additional untestable assumptions in the common situation of exposure-induced mediator–outcome confounding. “Interventional effects” have been proposed as alternative measures that overcome these limitations, and 2 versions have been described for the exposure-induced confounding problem. We aim to provide insight into the interpretation of these effects, particularly by describing randomized controlled trials that could hypothetically be conducted to estimate them. In contrast with natural effects, which are defined in terms of individual-level interventions, the definitions of interventional effects rely on population-level interventions. This distinction underpins the previously described advantages of interventional effects, and reflects a shift from individual effects to more tangible population-average effects. We discuss the conceptual and practical implications for the conduct of mediation analysis. See video abstract at, http://links.lww.com/EDE/B383. Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.