Abstract:
|
Bayesian methodology to compose a prior distribution from exploratory trial data is often used to calculate probability of success (POS) for a confirmatory trial as a reference of business assessment. The POS for all trials is required in performing multiple confirmatory trials. However, if Bayesian POSs are simply multiplied for individual trial levels, then, the POS for all trials is significantly lower. In this paper, a Bayesian methodology is developed to determine the POS of 2 confirmatory trials in continuous outcomes based from the current data of a given exploratory trial. Prior distributions are correlated in 2 trials that are conditionally independent where prior probabilities are fixed. The 2 trials are described in 2 different ways. The first description is as identical trials (mirror trials). The second description is with different characteristics.
|