Strong versus weak prior for predicting patient recruitment at the start of clinical trials in the context of drug availability issues

Designing efficient drug distribution planning represents a critical aspect of clinical trials. Inadequate supply as well as overproduction can be terribly costly for pharmaceutical firms. Therefore, being able to anticipate any issues regarding initial planning is crucial in order to ensure that the appropriate quantities of drugs are produced and delivered to each center. In this thesis, we investigate a Bayesian method to predict the most accurately patient recruitment in ongoing trials to improve their monitoring. More specifically, we were interested in the beginning of clinical trials, when only few data are available and information is uncertain. More particularly, we aimed at evaluating the performance of several priors in their ability to prevent drug availability issues in medical centers. The priors represented different levels of information and were evaluated through the use of simulations. We were interested in priors that represent robust predictions to avoid shortage. Our results suggest that, while situations where it is relevant to introduce strong prior information exist, overall, weak priors still represented the safest decision when providing information into a model.