Information borrowing in phase II basket trials: a comparison of different designs

Traditional phase II oncology clinical trials are focussed on a single treatment for a subgroup of patients with specific histological characteristics. Recently, the advances in cancer biology and genomic medicine have led to the flourish of new molecularly targeted therapies and to a re-thinking of the design of phase II trials assessing clinical activity. In this work, we evaluate the potential benefit of information borrowing across tumor subgroups in basket trials by comparing the operating characteristics of the models presented in the literature. In addition, we suggest a modification of the model proposed in [2] which exhibits good performances in terms of both the average number of patients treated and the ability to preserve the type I error rate.