Assessing Mediation in Cross-Sectional Stepped Wedge Cluster Randomized Trials.

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Epidemiology

Background:

• Mediation analysis is crucial for understanding treatment effect mechanisms.
• Existing methods are limited for correlated data in stepped wedge cluster randomized trials (SW-CRTs).
• Understanding mediation in SW-CRTs is vital for public health interventions.

Purpose of the Study:

• To develop and present novel regression-based methods for mediation analysis in SW-CRTs.
• To extend mediation analysis to handle correlated data structures common in SW-CRTs.
• To provide tools for estimating natural indirect effects and mediation proportions in complex SW-CRT designs.

Main Methods:

• Utilized linear and generalized linear mixed models for mediation analysis.
• Developed estimators for natural indirect effect and mediation proportion.
• Derived mediation expressions accounting for exposure-time treatment effect heterogeneity.
• Proposed methods applicable to continuous and binary mediators and outcomes.

Main Results:

• The developed mediation estimators demonstrate good performance across various data types and treatment effect structures.
• The methods successfully handle correlated data within SW-CRTs.
• Effectiveness shown for both typical and complex SW-CRT scenarios with time-varying effects.

Conclusions:

• The proposed regression-based methods offer a robust framework for mediation analysis in SW-CRTs.
• These methods enhance the understanding of treatment effect mechanisms in complex trial designs.
• The accompanying R package, mediateSWCRT, facilitates practical application and implementation.