|
Abstract:
|
Bayesian methods allow utilizing the historical information into current clinical trials, especially on comparable control groups, so that the information from a current trial is augmented and the precision could be increased by the incorporation of prior information. We apply Bayesian modeling in historical data borrowing for binary endpoints in Phase 2 study design and evaluate the operating characteristics using dual Bayesian proof of concept (POC) criteria P(pi_T-pi_C>0?data)>p_1 and P(pi_T-pi_C>TD?data)>p_2, where TD stands for a clinically meaningful Target Difference. We illustrate how this approach could potentially reduce the number of patients to be exposed in the control group and also save time and cost. Intuitively, when there is little existing historical information, or much between-trial heterogeneity in historical trials, fewer historical sample should be borrowed in order to reduce the bias and Type I Error. Recommendations will be made to choose the parameters to calibrate the priors to borrow an optimal amount of strength from historical data.
|
