Frequentist Operating Characteristics of Bayesian Posterior Probability Designs for Medical Device Trials That Include a Single, Late-Information-Time, Interim Analysis

• Activity Number: 611
• Type: Topic Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Greg Maislin*+
• Companies: Biomedical Statistical Consulting
• Keywords: Bayesian design ; posterior probability ; interim analysis ; frequentist operating characteristics ; informative prior ; orthopedic device

Abstract

In medical device trials, a single, late-information-time interim analysis can substantially reduce time to regulatory filing. Moreover, candidate control devices are often the subject of their own recent regulatory study, if not of multiple studies. Entirely ignoring this information at the time of analysis is increasingly difficult to justify. In this paper, Bayesian posterior probability designs are proposed that address these issues. Design goals include retention of some randomization to address hidden bias while allowing a substantial reduction in the control sample size through the use of a properly calibrated informative Bayesian prior. The FDA Guidance on Bayesian trials (February 2010) stipulates the need to explore frequentist operating characteristics of such designs. The numbers of draws needed to achieve acceptably low rates of simulation error will be evaluated for the case of non-inferiority trials utilizing a composite clinical success endpoint. Characteristics of designs with and without using a control informative prior will be compared and assessed as functions of sample size and how informative the control prior is assumed to be.