Causal effect estimation for competing risk data in randomized trial: adjusting covariates to gain efficiency.

Area of Science:

• Biostatistics and Clinical Trials
• Causal Inference
• Epidemiology

Background:

• Double-blinded randomized trials are the gold standard for estimating average causal effects (ACE).
• While naive estimators are consistent, covariate adjustment can improve efficiency and balance treatment groups.
• Prior work demonstrated efficiency gains with covariate adjustment in linear regression models.

Purpose of the Study:

• To extend the benefits of covariate adjustment to the competing risks data setting.
• To demonstrate that adjusted estimators maintain convergence rates and offer efficiency gains in competing risks analysis.
• To illustrate the proposed method using real-world clinical trial data.

Main Methods:

• Extension of covariate adjustment techniques to the competing risks framework.
• Utilizing augmented inverse probability censoring weighting (AIPCW) for adjusted estimation.
• Validation through extensive simulations and application to the Women's Health Initiative (WHI) trial.

Main Results:

• The AIPCW-based adjusted estimator demonstrates the same convergence rate as unadjusted estimators.
• Significant efficiency gains are observed with the adjusted estimator compared to the naive estimator in finite samples.
• The method is successfully applied to analyze the effect of dietary modification on cardiovascular disease mortality.

Conclusions:

• Covariate adjustment is beneficial for improving efficiency in randomized trials with competing risks.
• The proposed AIPCW-based method provides a robust approach for causal effect estimation in such settings.
• Findings support the use of adjusted estimators for more precise insights into treatment effects on mortality.