Abstract:
|
In clinical trials, a Bayesian Meta Analytic Predictive (MAP) approach has been successfully implemented to guide a dose escalation decision making in Phase I studies and to minimize the sample size needed in the placebo group of Phase II studies. In the latter case, due to ethical consideration, enrolling patients to the placebo group should be minimized if there are historical controls, which can be used to borrow information. A robust prior component is usually included in the MAP to address a prior-data conflict. Recently, we applied the MAP approach with a robust prior component to investigate a power gain and an inflation of type I error as compared to the MAP with non-robust, and uniform priors. In our application, we borrowed information from previous studies with a dose 1. The dose 2 has never been studies before and a 2:1 randomization ratio was used to randomize as much as twice patients in dose 2 compared to dose 1. The power needed to detect a meaningful difference between the two doses was calculated using a MAP approach. A power gain of 5%-10% with a 1-5% inflation type I error were found in the MAP with the robust component compared to the MAP with uniform prior.
|