|
Abstract:
|
During drug evaluation in early phases, previous information may be available from clinical trial already conducted on a different population. In these cases, it might be desirable to share information using efficiently the available resources. It can lead to a better estimation of the dose-response curve. In this work, we propose a modification and an adaptation of power and commensurate priors in case of bridging studies. In particular, we focus in bridging studies between Caucasians and Asians with sequential adaptive allocation design under Bayesian inference. We weight the prior information in two steps: (1) the effective sample size notion is used to set the desirable amount of information to be shared from previous data at each stage of the trial; (2) in a sort of empirical Bayes, a commensurability parameter is chosen using a measure of distribution distance. This approach avoids elicitation issue and computational issues. We have performed an extensive simulation study evaluating the sensitivity to prior choices and robustness of our proposition. Our method is robust whatever the dose-response curves in the studied populations are very similar or not.
|
