Improving Power and Precision in Randomized Trials Using Covariate Adjustment

Music City Center 

Room: CC-110A 

Covariate adjustment in randomized trials allows investigators to utilize information collected prior to randomization in pre-planned analyses, leading to potential improvements in statistical power and precision. These improvements can result in more ethical, efficient, and cost-effective research, benefiting research participants, the population under study, and research sponsors. Covariate adjustment in randomized trials is broadly supported by both American (FDA) and European (EMA) regulators but remains underutilized in practice.

First, the conceptual basis for covariate adjustment is introduced with a survey of the literature, focusing on marginal (average) treatment effects. Participants will learn how covariate adjustment works, when these methods are most beneficial, how to assess their benefits, and the assumptions underlying these analyses. Recommended methods are robust to model misspecification for outcomes or missingness, do not introduce additional assumptions, and are at least as efficient as unadjusted analyses.

Next, covariate adjusted analyses are illustrated step-by-step with example code and realistic data based on actual trials. Participants will learn about software for analyzing continuous, binary, ordinal, and time-to-event outcomes.

Finally, covariate adjustment is discussed for study designs allowing early stopping for success or futility. This can lead to more ethical, efficient trials, that are robust to misspecification in design calculations.