A Bayesian Two-stage Phase II Design with Historical Controls

• Activity Number: 121
• Type: Contributed
• Date/Time: Monday, August 9, 2004 : 12:00 PM to 1:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Michael W. Sill*+ and Roger L. Priore
• Companies: Roswell Park Cancer Institute and Roswell Park Cancer Institute
• Keywords: Bayesian ; design ; phase II ; clinical trials ; sequential ; two-stage

Abstract

Heitjan (1997) proposed a useful concept called "Bayesian Persuasion Probabilities" for use in phase II clinical trials. Essentially, it involves two competing views of a study agent's effectiveness. Prior to observing the data, the "skeptic" has a tendency to believe in the null hypothesis that a new drug is ineffective while the "enthusiast" has a tendency to believe in the alternative hypothesis (the drug is effective). According to the author, data from a good study should be able to lead both views to the same conclusion. In practice, however, it would take large sample sizes to reach this consensus with high probability. In the proposed design, we will require the data to persuade the enthusiast to abandon the alternative hypothesis and accept the null hypothesis as plausible before a study can close early. On the other hand, the data at the end of the trial must persuade the skeptic to abandon the null hypothesis and accept the alternative as plausible before the design will recommend further study in a phase III trial. This approach seems to yield designs with better operating characteristics for typical phase II sample sizes.