A Unified Method for Balancing Continuous and Categorical Baseline Covariates in Randomized Clinical Trials

• Activity Number: 35
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yunzhi Lin*+ and Zheng Su
• Companies: University of Wisconsin-Madison and Genentech
• Keywords: Clinical Trial ; Randomization ; Baseline Covariates ; Imbalance ; Cumulative Distribution Function ; Minimization

Abstract

Covariate adaptive randomization is often adopted in randomized clinical trials to maintain the balance of baseline covariates. Whilst various techniques have been developed, most methods require that the prognostic covariates to be balanced be categorical. Continuous covariates are either excluded or need to be categorized, which is not desirable in many clinical settings. In this paper, we propose a novel approach that maintains the balance for multiple continuous and categorical covariates. The method uses the normalized area between the empirical cumulative distribution functions of the observed covariates as the imbalance metric, which is shown to be a unified measure for continuous and categorical covariates. Performance of the proposed method is evaluated from three aspects: the amount of balance achieved, the loss of information due to imbalance, and the impact on the inference. Numerical results show that the new approach produces more accurate treatment effect estimates and leads to more powerful trials than existing methods for trials with binary, continuous, and time-to-event outcomes. We illustrate the utility of our method by applying it to a clinical trial data set.