Empirical likelihood inference in randomized clinical trials.

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Statistical Inference

Background:

• Randomized controlled trials (RCTs) often collect pre-randomization covariate data.
• Covariate adjustment in regression models is commonly used to increase the precision of average treatment effect (ATE) estimates.
• Concerns exist regarding potential compromises to objective inference in RCTs due to covariate adjustment.

Purpose of the Study:

• To propose an empirical likelihood approach for covariate adjustment in RCTs.
• To develop unbiased estimating functions that separate ATE estimation from covariate-outcome modeling.
• To evaluate the efficiency and robustness of the proposed method compared to existing techniques.

Main Methods:

• Developed an empirical likelihood framework for covariate adjustment.
• Proposed two unbiased estimating functions decoupling ATE and covariate-outcome relationships.
• Conducted a simulation study to compare finite sample performance.
• Analyzed data from an HIV clinical trial.

Main Results:

• The empirical likelihood estimator achieves efficiency comparable to existing adjusted estimators when regression models are correctly specified.
• The proposed estimator remains at least as efficient as existing methods regardless of working model specification.
• Simulation results show the empirical likelihood approach is more efficient and powerful when covariate-outcome relationships are misspecified.

Conclusions:

• The proposed empirical likelihood method offers a robust and efficient approach to covariate adjustment in RCTs.
• This method enhances statistical power and precision, particularly when regression models are misspecified.
• The approach provides reliable inference for average treatment effects in the presence of baseline covariates.