Background: Basket trials are used to study targeted treatments across a range of tumor sites. In these designs all patients have the same target altered and receive the same regimen but they are placed in baskets defined by the site of primary tumor. This design is particularly appealing since the prevalence of the target mutation is typically low in any given tumor site, and the drug sponsor is naturally interested in as broad an indication for use as possible. Given the multiple baskets, this design requires the investigation of whether the activity of the regimen applies uniformly to all baskets, i.e., homogeneity of treatment effect, and if not, which baskets show sufficient response. Many basket trial designs to date have resembled parallel phase II studies, using Simon's two-stage design separately for each basket. The inclusion of multiple baskets increases the overall false positive rate in the trial, an aspect that is commonly not taken into account in current designs. More importantly, the use of independent, parallel phase II trials fails to take advantage of the opportunity to combine baskets in the event that the interim results suggest common responsiveness across baskets, either common evidence of efficacy or common evidence of lack of efficacy.

Methods: We develop a novel, adaptive study design that makes use of an interim assessment of the heterogeneity of treatment effect across baskets to determine the most appropriate statistical analysis, while controlling the overall false positive rate in the trial. Our proposed design allows for aggregation of baskets in stage 2 when evidence points to a common effect of the drug, while retaining the option to maintain selected individual baskets in stage 2, if the interim evidence supports this strategy. Similar to conventional designs, our proposed design allows for early termination of baskets for lack of responsiveness. Using simulations we demonstrate the potential reduction in the number of patients required compared with the traditional strategy based on parallel Simon two-stage designs, while maintaining similar levels of Type I and II errors.

Citation Format: Kristen M. Cunanan, Alexia Iasonos, Ronglai Shen, David Hyman, Colin Begg, Mithat Gonen. Efficient basket trial designs. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr CT118.