A Bayesian approach in Proof-of-Concept study design: a case study
*Michael Lee, Janssen R&D 


Proof-of-Concept (PoC) study is a key step in drug development. Conventionally, PoC is accomplished by showing efficacy or expected pharmacological activity in at least one dose of the drug being studied. Because of the primary objective of PoC studies, the dose studied often is the maximum tolerated dose. Efficacy in other dose level is seldom studied. There are circumstances where knowledge of other dose level can help designing subsequent studies. For example, if we know a dose level that does not meet target effect size, this dose level can be the lower bound of the subsequent dose-finding study. An adaptive design can offer the possibility of studying more than 1 dose. In this presentation we will share an adaptive PoC study design that potentially allows studying more than one dose. A challenge in this design is that treatment period is long compared to enrollment period. To allow adaptation before all subjects are randomized, a prediction of outcome at the end of treatment based on early signals becomes necessary. The prediction is made by a statistical model and Bayesian approach so that data from historical studies as well as data accumulated from the current study thus far can be utilized. Characteristics of this study design will be presented.