Estimands and Doubly Robust Estimation for Cluster-Randomized Trials With Survival Outcomes

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Cluster-randomized trials (CRTs) involve randomizing groups, necessitating specialized analysis.
• Estimating treatment effects in CRTs requires differentiating cluster-level and individual-level impacts.
• Survival outcomes in CRTs present unique analytical challenges, especially with right-censoring.

Purpose of the Study:

• To formally define cluster-level and individual-level treatment effect estimands for CRTs with right-censored survival data.
• To propose novel, doubly robust estimators for these estimands.
• To provide robust statistical methods for analyzing survival outcomes in CRTs.

Main Methods:

• Developed doubly robust estimators for cluster- and individual-level treatment effects.
• Addressed dependent censoring on baseline covariates, ensuring consistency if either outcome or censoring model is correct.
• Employed various modeling strategies for censoring and outcome distributions.
• Utilized a deletion-based jackknife method for variance and interval estimation.

Main Results:

• Proposed estimators demonstrated consistency under dependent censoring.
• Simulation studies confirmed adequate finite sample performance of the methods.
• The methods were successfully applied to a real-world CRT with survival endpoints.

Conclusions:

• The developed doubly robust estimators provide a reliable approach for analyzing survival data in CRTs.
• These methods enhance causal inference by accurately distinguishing treatment effects at different levels.
• The findings offer valuable tools for researchers conducting CRTs with survival outcomes.