Variable selection for covariate-adjusted semiparametric inference in randomized clinical trials.

Area of Science:

• Biostatistics
• Clinical Trials Methodology

Background:

• Baseline covariates in randomized clinical trials (RCTs) can enhance treatment effect inference.
• Current covariate adjustment methods face challenges: post hoc selection risks bias, while pre-specification may omit key variables.

Purpose of the Study:

• To introduce objective covariate selection methods for maximizing efficiency in RCTs.
• To address the controversy surrounding covariate adjustment in clinical trial analysis.

Main Methods:

• Utilized modern variable selection techniques within a semiparametric framework.
• Separated variable selection for covariate-outcome relationships from treatment effect estimation.
• Proposed a novel approach for estimating uncertainty (e.g., standard errors) in finite samples.

Main Results:

• Demonstrated that objective variable selection identifies key covariates.
• Showcased unbiased and efficient inference on the treatment effect.
• The proposed uncertainty estimation method outperformed existing approaches.

Conclusions:

• Objective covariate selection using modern methods in a semiparametric framework resolves bias issues in RCTs.
• The proposed methods enhance the efficiency and validity of treatment effect estimation and uncertainty quantification.