Can external and subjective information ever be used to reduce the size of randomised controlled trials?

Abstract

In the design of randomised trials in rare cancers, a Bayesian approach has been advocated, which allows for external and subjective information to be formally incorporated. We explore whether this can be extended more generally to allow for smaller trials to be conducted using a case study involving a trial of nasopharyngeal carcinoma. The external information available at various points during the trial is first summarised in the form of ‘prior distributions’. Each of these is then combined with the accumulated data from the trial at that point in time to form ‘posterior distributions’, from which conclusions are drawn. We have argued that such a framework for the design, analysis and interpretation of a randomised trial in the light of external evidence is particularly useful in situations such as trials in rare cancers. But more generally, it may potentially also allow for smaller trials to be conducted. Although, at this point in time, we are hesitant to recommend the full implementation of the Bayesian methodology to modify the (conventionally) planned trial size we submit that a formal synthesis of the external evidence bearing on the question of concern is a valuable exercise in itself.