Accounting for correlation and compliance in cluster randomized trials

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Cluster randomized trials (CRTs) are susceptible to post-randomization exposures.
• Differential compliance between treatment arms can bias causal inference in CRTs.
• Survival data analysis in CRTs presents unique challenges due to clustering.

Purpose of the Study:

• To develop methods for causal inference with survival data from CRTs.
• To address the impact of post-randomization exposures and selective compliance in CRTs.
• To adapt structural models to account for clustering in survival analysis.

Main Methods:

• Utilized structural models to account for post-randomization exposures.
• Adapted structural estimators using marginal modeling to handle clustering.
• Employed random effects models to incorporate clustering in survival data analysis.
• Applied methods to survival data from a cluster randomized trial.

Main Results:

• Demonstrated that structural models can be adapted to account for clustering in CRTs.
• Showcased how marginal and random effects models adjust structural estimators for clustered survival data.
• Provided an empirical illustration using a vitamin A trial for infant mortality prevention.

Conclusions:

• Post-randomization exposures and selective compliance are critical considerations in CRTs.
• Marginal and random effects models offer valid approaches for causal inference with clustered survival data.
• The proposed methods enhance the reliability of causal effect estimation in CRTs with survival outcomes.