Bayesian Hierarchical Models in Adaptive Clinical Trials Aiming to Generate Pediatric Efficacy Data
View Presentation View Presentation
*Roger Lewis, UCLA Medical Center 


Challenge: It is common in clinical practice for pharmaceutical agents and medical devices to be used “off-label” in pediatric patients—with the best of intentions—despite little or no data to support efficacy or safety. Barriers to the generation of valid data, supporting or contradicting the wisdom of such use, include: (1) the relative rarity of many target conditions in children; (2) low event rates in children, further limiting the power of pediatric trials; (3) the primary interest, in many settings, of generating data to support efficacy and safety in adults; (4) heterogeneity within pediatric populations; and (5) parental desire to only allow participation in studies that are likely to benefit their child. Potential Solutions: Many of the challenges noted above can potentially be addressed, at least in part, by using Bayesian hierarchical models (BHMs) as the inferential foundation for adaptive clinical trials that include children. The speaker will discuss the structure and advantages of BHMs, especially when one needs to rationally combine information from multiple heterogeneous sources (e.g., data from adult and pediatric populations) but obtain separate estimates of treatment effects. He will also present illustrative trial designs, including adaptive trial designs that include response-adaptive randomization to partially address the goal of benefit to individual subjects, that address each of the challenges listed above.