CMStatistics 2017: Start Registration
View Submission - CMStatistics
B0187
Title: Explaining the total effect in the presence of multiple mediators and interactions Authors:  Linda Valeri - Harvard Medical School and McLean Hospital (United States) [presenting]
Abstract: Mediation analysis allows decomposing a total effect into a direct effect of the exposure on the outcome and an indirect effect operating through a number of possible hypothesized pathways. A recent study has provided formal definitions of direct and indirect effects when multiple mediators are of interested. Parametric and semi-parametric methods to estimate path-specific effects have also been described. Investigating direct and indirect effects with multiple mediators can be challenging in the presence of multiple exposure-mediator and mediator-mediator interactions. Three main contributions are provided: 1) we obtain counterfactual definitions of interaction terms when more than one mediator is present; 2) we derive a decomposition of the total effect that unifies mediation and interaction when multiple mediators are present; and 3) we illustrate the connection between our decomposition and the 4-way decomposition of the total effect introduced in the context of a single mediator. The framework applies to continuous, categorical or time-to-event outcomes. We illustrate the properties of the proposed framework for multiple mediators and interactions, in a secondary analysis of a pragmatic trial for the treatment of schizophrenia. We employ the decomposition to investigate the complex interplay of side-effects and psychiatric symptoms trajectories in explaining the effect of antipsychotics on social functioning in schizophrenia patients.