Bayesian Adaptive Designs in the Presence of Multiple Factors: Hyperbaric Oxygen Brain Injury Treatment (HOBIT) Phase II Trial

• Activity Number: 45
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Byron J. Gajewski* and Scott M. Berry and Gaylan L. Rockswold
• Companies: University of Kansas Medical Center and Berry Consultants and Hennepin County Medical Center
• Keywords: Clinical Trials ; Bayes ; Adaptive ; Accrual

Abstract

We are interested in optimizing Hyperbaric Oxygen (HBO2) treatment instituted in three dimensions using an exploratory clinical trial (phase II). Specifically we have three factors: pressure (3 levels), additional FiO2 (2 levels), and frequency (2 levels). After taking out three unpromising factor combinations we have 10 arms (with control). As a practical constraint we are limited to 200 patients with a six month endpoint. Response adaptive randomization (RAR) across the 10 regimens is of great interest because of its favorable clinical and statistical properties and the cell means methodology is well developed for a wide variety of models. A main effects factor model (no interactions) is very attractive in optimizing each dimension. In balance, we have designed a trial so adaptations are driven by cells model to identify a set of cells better than control. A follow-up main effects model identifies optimal pressure and the other factors. We present this novel clinical trial by reporting the operating characteristics for both the cells and main effects models for fixed, RAR, RAR with early stopping, and RAR with early stopping augmented by early (30-day) clinical information.