|
Abstract:
|
Bayesian Adaptive designs allow for key study design specifications to be modified as information is collected from observed data during a clinical trial. When carried out in the proper circumstances, adaptive designs can lead to more efficient allocation of resources and shorter study spans. Bayesian Adaptive Randomization, specifically, establishes a well-defined framework for updating allocation ratios based on dose performance, allowing to increase the probability of treatment allocation to more promising doses of a study drug, while also preserving the benefits of randomization. To study potential advantages of adaptive randomization and its statistical properties, we simulated data for a planned Phase II dose-ranging study of an experimental treatment for pain due to osteoarthritis. Bayesian interim analysis was conducted to implement adaptive randomization using the following three efficacy dose-response models for comparison: the Emax model, normal dynamic linear model (NDLM), and analysis of variance (ANOVA). Adaptive randomization was conducted based on predictive treatment efficacy and safety, and with the intention of avoiding exposure to unnecessary risk when high dose safety is uncertain. Through our simulation, we showed increased efficiency by our proposed approach to adaptive randomization that enables optimization of patient allocation balancing between efficacy and safety, as well as decreased enrollment of patients to unnecessary high doses.
|
