TL5: Strategies to Mitigate Against Operational Bias in Adaptive Designs
*Greg Maislin, Biomedical Statistical Consulting 

Keywords: adaptive design, operational bias

The potential for improving clinical trial efficiency through adaptive design is widely recognized. There are a number of approaches to incorporating adaptive features that use data collected during the conduct of the study to modify or refine the design prior to its completion. These include group sequential methods, sample size re-estimation methods, and Bayesian adaptive design methods. If the data used in such adaptations involve unblinded determinations of effect size, there is concern that the process of implementing adaptive design elements may make information available to clinical trial staff that lead to behavioral changes. These behavioral changes can influence the conduct of the trial resulting in "operational bias" when estimating treatment effects and testing hypotheses. This roundtable session will address the following questions, "What strategies exist for reducing operational bias without introducing prohibitive operational complexity? Do such strategies differ between medical device trials involving unblinded treatments compared to double-blind clinical trials of pharmaceuticals?