On Group Sequential Enrichment Design for Basket Trial

ABSTRACTCancer is becoming a collection of “niche diseases” defined by the molecular subtypes. Such understanding forms the basis of the basket trial, which pools together multiple cohorts of patients with the same molecular subtype across different tumor indications and thus facilitates the development of targeted therapies. Efficient pruning is critical in basket designs as it ensures the internal consistency of the selected indications and improves the probability of success on the selected pool of indications. In this article, we consider pruning both inactive and extremely active indications and propose a group sequential enrichment design with testing procedures to both control the family-wise error rate (FWER) in the weak sense and maintain the power for basket trials with such pruning strategy. We compare the proposed design with a relevant design which prunes only unresponsive subgroups. The total misclassification rates and misclassification rates among truly unresponsive indications are smaller...