PS54: Have Predictive Probabilities Improved Clinical Trials?

Clinical trial resources are scarce. Every decision about sample size, arm selection, and subgroup identification within a trial impacts future patients and future trials, as we attempt to be efficient stewards of those limited resources. With uncertainty looming with every decision, we need to predict which actions are most likely to lead to successful treatments for patients, whether these actions are to stop an arm, stop a population, stop an entire study and investigate another question, or invest more resources into a confirmatory study. The better our predictions, the better our decisions, and we become more efficient in translating scientific discoveries and hypotheses into actionable treatments. Predictive probabilities have been used commonly for over a decade for these purposes. This session reviews current practice involving predictive probability based decision rules, assesses their contribution to the respective designs, and discusses lessons learned that may be applied to future trials. The session aims to highlight settings where predictive probabilities are advantageous, discuss efficient methods for computing predictive probabilities, and share practical advice from clinical trials employing this technique. Overall, this session seeks to bridge the gap between advanced statistical methods and practical applications in clinical trials. By focusing on predictive probabilities, it aims to showcase how these tools can be employed to make more informed, nuanced decisions in clinical research, ultimately enhancing the efficiency and effectiveness of clinical trials.