Bayesian Adaptive Designs for FDA Required Safety Studies
View Presentation View Presentation
Donald A Berry, MD Anderson Cancer Center  *Scott M Berry, Berry Consultants  Jason T Connor, Berry Consultants 

Keywords: bayesian, adaptive, safety, diabetes

FDA required evidence of drugs' safety 25 years before evidence of efficacy. After the recall of rofecoxib and black box warning on rosiglitazone we're seeing a new focus on premarket studies of drug safety. This includes FDA requiring pharmaceutical companies to demonstrate their new drugs are heart-safe by showing the confidence interval for the risk ratio for CV events in the drug vs. a suitable control is less than a constant (e.g. 1.8 or 2). Designing such trials and ensuring adequate power is difficult since the CV event rate in populations without underlying heart conditions is poorly known and because the CV event rate continues to decrease with improvements in cardiac medicine. We demonstrate an adaptive design to provide sufficient sample size/power that also offers futility stopping for trials aiming to prove a new drugs' CV safety when the a priori event rate is uncertain.