|
Abstract:
|
Several methods exist for incorporating efficacy data from adult trials into pediatric trials. In a regulatory environment, methods for extrapolation must be pre-specified as part of the pediatric primary analysis plan. This motivates the idea of Bayesian dynamic borrowing methodologies, which borrow more heavily when pediatric trial results are similar to adult trial results and borrow less when results are discordant. In this talk, we contrast the dynamic borrowing of Bayesian hierarchical models to that of “test then pool” strategies, and to hybrid approaches such as Prior-Data Conflict Calibrated Power Priors (PDCCP). In a rare disease setting with adult and pediatric trials running concurrently, we demonstrate that Bayesian hierarchical models provide better performance and are better suited to address the uncertainty in the homogeneity of the populations. In addition, such methods naturally lead to estimates of treatment efficacy that can and should be included in a product label.
|
