Online Program

Thursday, October 20
Knowledge
Community
Influence
Thu, Oct 20, 5:15 PM - 5:50 PM
Carolina Ballroom
Poster Session 1 & Opening Mixer
Sponsored by Bank of America

Impact of Time Trend Confounding in a Response Adaptive Randomized Clinical Trial with a Binary Outcome (303521)

Valerie Lynn Durkalski, Medical University of South Carolina 
*Yunyun Jiang, Medical University of South Carolina 
Wenle Zhao, Medical University of South Carolina 

With response adaptive randomization (RAR), the shift in patients’ disease risk over time could potentially result in differences in the distribution of population characteristics between treatments. The traditional test statistics is no longer valid for making causal inference for the treatment effect and tends to yield large type I error rate. This study focuses on addressing the issue of time trend in phase 3 Bayesian response adaptive randomized clinical trial study with a binary outcome. Specifically, we evaluated three currently available Bayesian response adaptive randomization formulas with respect to their impact on the trial operating characteristics. In addition, we aim to propose a model-based randomization approach to randomize patients based on adjusted treatment effect while accounting for the trend effect based on Bayesian state-space structural model at each allocation update. The new method is expected to control biased treatment estimation and preserve type I error rate of the trial.