|
Abstract:
|
In this work, we proposed a Bayesian basket trial design with similarity-based information borrowing under local multisource exchangeability assumption. Our approach first partitioned potentially heterogeneous baskets (subtypes) into non-exchangeable clusters that restricted information borrowing to occur only locally, i.e., among similar baskets within the same cluster, then evaluated the amount of information for borrowing based on between-basket similarities. The number of clusters and cluster memberships were inferred from posterior probability of each partition. By simulation analysis for baskets with balanced and unbalanced sample sizes using single-stage or two-stage design, the proposed method has shown well-controlled family-wise error rate and desirable basket-wise power in comparison with the Multisource Exchangeability Model and Simon's two-stage design. In addition, our method is computationally efficient in that the posterior profiles of interest could be derived explicitly without the need for sampling algorithms.
|
