Keywords: borrowing, Bayesian hierarchical model, indications
Designing clinical trials with multiple indications could can be challenging. Rare indications lead to very small sample size and outcome usually may vary across indications. In addition, model based approach facilitates the use of interim analysis. In this setting, maximizing information by appropriately borrowing of information across different indications leads to more effective decision making. The study considered is an open label, multi-arm, phase II study. Patients will be enrolled in five different selected indications. The primary endpoint is Observed Response Rate (ORR). In this study, a robust Bayesian hierarchical model allows dynamic borrowing of information between groups such that more borrowing occurs across the groups that have similar ORR and less borrowing between groups, which differ. In this way, the model is a compromise between the two alternate extremes of either a completely pooled analysis or stratified analysis. Simulations are performed under different scenarios to assess the model using R and JAGS. Simulation based operating characteristics are provided.