Title: Leveraging auxiliary covariates to improve efficiency of inferences: A general framework and practical considerations

Abstract

Auxiliary covariates are routinely collected in randomized clinical trials. Although it has been long recognized in statistical literature that incorporating covariates can improve the efficiency of inferences and reduce chance imbalance, covariates adjustment has not been used as often as it should be in the primary analysis of clinical trials in practice. This is partly due to that many practitioners remain skeptical of its usefulness or have other concerns regarding model misspecifications. We try to address these concerns and discuss a general framework that allows one to adjust covariates robustly. That is, this framework can guarantee improvement in efficiency, if covariates are predictive of outcomes, and valid inferences regardless of whether the specified models for covariates are correct or not. Therefore, it can eliminate the concern over model misspecification. We further discuss practical considerations in terms of covariate adjustment, focusing on finite sample effects, stratified randomization, and what variables to adjust. We attempt to address the question of how and when to use covariate adjustment for randomized clinical trials in practice.