A Bayesian Proportional Hazards Model to Predict Patient Recruitment in Multicenter Clinical Trials

In most clinical trials, patient recruitment rates are overestimated, leading to several problems, including delays, additional costs, lost revenue, and even trial discontinuation. The accurate prediction of the time and resources needed to complete recruitment is therefore of crucial importance. Two questions often arise. First, given the current status of an ongoing trial, how long will it take to recruit the remaining patients needed? Second, how many additional centers should be opened in order to ensure completion within the required timelines? To address these very practical questions, a proportional hazards model related to classical ideas in the statistical renewal theory is useful for modeling the randomization dates of future patients. Such a model is most conveniently fit using a Bayesian framework and is incorporated seamlessly with other time-to-event distributions used in the trial design. We illustrate the proposed Bayesian predictive methodology using two real datasets from actual multicenter clinical trials. We also review other Bayesian approaches that have appeared in the literature and illustrate their connections to ours. We also offer a simulation study designed to assess the model’s predictive performance across a range of true underlying scenarios. Overall, our results suggest that Bayesian proportional hazards models can be highly effective tools for modeling multicenter clinical trial recruitment.