Informative priors or noninformative priors? A Bayesian re-analysis of binary data from Macugen phase III clinical trials

It has became more popular in the recent statistical literature to see Bayesian approaches for clinical trials, such as assurance calculations for study designs and the use of posterior probability for data analyses. When applying Bayesian analysis to clinical trial data, one common question is to use informative priors or noninformative priors. In order to explore this question, we looked for existing clinical trial data with a simple structure and a simple clinical endpoint so that the Bayesian re-analyses can be easily performed. We came across the published Phase III Macugen® data that were suitable for this exploration. In this manuscript, the Macugen Phase III development program was described, the primary data for the two Phase III pivotal studies were re-analyzed using Bayesian applications with informative priors and noninformative priors. These re-analysis results were summarized, compared, and discussed.