660 – Probability of Success in Clinical Trials and Decision Rule
A Bayesian Predictive Approach to Go/No-Go Decision
Kaihong Jiang
Sanofi
Yan Zheng
Sanofi
Phase II go/no-go decisions are typically based on observed phase II data and statistics like p-values, mean differences, and associated confidence intervals. These statistics, however, can be ambiguous for decision making as they do not inform about the risks of failure or the success probability of the subsequent phase III trials. Probability of success (POS), which has grown in some popularity recently in biostatistics literature, is a more useful and possibly necessary statistics for facilitating a risk-informed decision making process. A Bayesian statistic, POS incorporates both the observed phase II trial data and the design parameters of the planned phase III trials. Jiang (2011) developed a POS function in a closed form for a simple two-parallel-group setting where response variables are normally distributed. In this paper, we extend the results to binary data and derive the POS function for response variables that follow a binomial distribution. Applications are shown in go/no-go decisions and portfolio management.