An extension of Bayesian predictive sample size selection designs for monitoring efficacy and safety

Most exploratory clinical trials in cancer are designed as single-arm trials using a binary efficacy outcome with or without interim monitoring. In this context, we have proposed a Bayesian adaptive design denoted as predictive sample size selection design (PSSD), which considered a binary efficacy outcome associated with early futility stopping (Statistics in Medicine 2012; 31: 4243-4254). As a matter of course, it would be more ethical and informative to evaluate safety as well as efficacy during the course of a trial. However, in most of the trials, only major adverse events are taken into account for early termination of the trial, and safety itself is used as a secondary endpoint. In this paper, we propose an extension of the PSSD to monitor efficacy, take into consideration the sample size adaptation during the trial and add continuous monitoring of safety to the trial design. This method is developed in the Bayesian framework, in which a decision to stop for reasons of safety can be made based on the posterior probability or predictive probability, not necessarily at the time of pre-specified monitoring for efficacy. We investigate the operating characteristics of the proposed method through simulation studies and show that the posterior probability-based method with less informative prior to monitor safety has more reasonable performance.