Regression Adjustment in Covariate-Adaptive Randomized Experiments With Missing Covariates

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Covariate-adaptive randomization balances prognostic factors in clinical trials.
• Regression adjustments improve estimation efficiency.
• Missing covariate data poses challenges for analysis.

Purpose of the Study:

• To investigate asymptotic properties of average treatment effect estimators with missing covariates under covariate-adaptive randomization.
• To develop consistent variance estimators for valid statistical inference.
• To evaluate finite-sample performance of different methods.

Main Methods:

• Combining missing data handling procedures with regression adjustment.
• Asymptotic analysis of treatment effect estimators.
• Derivation of consistent variance estimators.
• Model-free analysis ensuring validity under misspecification.

Main Results:

• Established asymptotic properties for average treatment effect estimators.
• Developed consistent variance estimators for reliable inference.
• Numerical studies demonstrated performance across various scenarios.

Conclusions:

• The proposed methods provide a robust framework for analyzing clinical trials with missing covariate data under covariate-adaptive randomization.
• The model-free approach ensures validity even with potential regression model misspecification.
• Recommendations are provided for practical application based on simulation results.