A Bayesian re-assessment of two Phase II trials of gemcitabine in metastatic nasopharyngeal cancer.

The Simon two-stage minimax design is a popular statistical design used in Phase II clinical trials. The analysis of the data arising from the design typically involves the use of frequentist statistics. This paper presents an alternative, Bayesian, approach to the design and analysis of Phase II clinical trials. In particular, we consider how a Bayesian approach could have affected the design, analysis and interpretation of two parallel Phase II trials of the National Cancer Centre Singapore, on the activity of gemcitabine in chemotherapy-naïve and in previously treated patients with metastatic nasopharyngeal carcinoma. We begin by explaining the Bayesian methodology and contrasting it with the frequentist approach. We then carry out a Bayesian analysis of the trial results. The conclusions drawn using the Bayesian approach were in general agreement with those obtained from the frequentist analysis. However they had the advantage of allowing for different and potentially more useful interpretations to be made regarding the trial results, as well as for the incorporation of external sources of information. In particular, using a Bayesian trial design, we were able to take into account the results of the parallel trial results when deciding whether to continue each trial beyond the interim stage.