Bayesian Borrowing across Pediatric Indications: An FDA CID Case Study in Pediatric Epilepsy

Oregon Convention Center 

We discuss a randomized, double-blind, placebo-controlled trial design for children and adolescents with epilepsy with myoclonic-atonic seizures (EMAS). The primary endpoint is EMAS-associated seizure frequency over the treatment period. Based on the need for therapeutic treatments in the EMAS population, the difficulty of enrollment in this rare population, and the consistent treatment effect across related indications for the investigational therapy, Bayesian methods are used to formally incorporate previous trial results from related populations into the primary analysis of the proposed study in the EMAS population. A Bayesian hierarchical model is specified for the treatment effects across populations that induces dynamic borrowing of the data from the historical studies. In addition, the design incorporates an adaptive sample size via Goldilocks methodology using Bayesian predictive probabilities. We discuss the process of trial design in the context of the FDA Complex and Innovative Design (CID) program, and the key simulations used to explore and evaluate the innovative trial design.