ABSTRACT
The approaches based on average power and assurance in Chapters 2 and 3 were based on the expectation of the study power with respect to the prior distribution of the treatment effect. This approach has been the standard hybrid Bayesian/frequentist approach used since the introduction of the concept into clinical trials planning in the 1980s. However, as Spiegelhalter, Abrams and Myles in their book Bayesian Approaches to Clinical Trials and Health-Care Evaluation in 2004 observed, “.… if we express uncertainty over the parameters as a prior distribution, then the power can be considered as an unknown quantity with a distribution induced by the prior”. In other words, based on the assumptions underlying these approaches, we are considering it is possible to take an alternative course and to consider the whole prior distribution of the study power, or alternatively the whole prior distribution of the sample size per arm. This is the topic of this chapter in which we consider the median (prior) power and look to determine the prior probability that the study power lies in any pre-determined interval.
