ABSTRACT
In this chapter, we turn to Bayesian design and analysis of phase III studies. These are typically randomized controlled multicenter trials on large patient groups (300-3,000 or more, depending upon the disease/medical condition studied) aimed at being the definitive assessment of how effective the drug is, in comparison with current “gold standard” treatment. Such trials are often called confirmatory trials. The approach of the chapter is to focus on what is different for a Bayesian
statistician in a confirmatory trial. The development here is not meant as a general reference for designing or analyzing confirmatory trials, but rather as a description and exemplification of the features and challenges for a Bayesian. Especially early in the chapter, we rely heavily on demonstrating these features and challenges through a running “example trial” that we construct for pedagogical purposes. This chapter is not meant as a precise recipe for doing every possible Bayesian confirmatory trial. In fact, we believe such an attempt would be counterproductive, both to the Bayesian approach and to science in general. After all, an important aspect of the Bayesian approach is its flexibility and synthetic nature. The ability to create a unique design specific to the challenges of each problem, guided by Bayesian principles and philosophy, is one of the strengths of the approach. Any attempt to describe in overly “cookbook-y” steps how this should be done is likely to stifle its effectiveness.
