Bayesian basket trial design with exchangeability monitoring

Abstract

Precision medicine endeavors to conform therapeutic interventions to the individuals being treated. Implicit to the concept of precision medicine is heterogeneity of treatment benefit among patients and patient subpopulations. Thus, precision medicine challenges conventional paradigms of clinical translational which have relied on estimates of population‐averaged effects to guide clinical practice. Basket trials comprise a class of experimental designs used to study solid malignancies that are devised to evaluate the effectiveness of a therapeutic strategy among patients defined by the presence of a particular drug target (often a genetic mutation) rather than a particular tumor histology. Acknowledging the potential for differential effectiveness on the basis of traditional criteria for cancer subtyping, evaluations of treatment effectiveness are conducted with respect to the “baskets” which collectively represent a partition of the targeted patient population consisting of discrete subtypes. Yet, designs of early basket trials have been criticized for their reliance on basketwise analysis strategies that suffered from limited power in the presence of imbalanced enrollment as well as failed to convey to the clinical community evidentiary measures for consistent effectiveness among the studied clinical subtypes. This article presents novel methodology for sequential basket trial design formulated with Bayesian monitoring rules. Interim analyses are based a novel hierarchical modeling strategy for sharing information among a collection of discrete potentially nonexchangeable subtypes. The methodology is demonstrated by analysis as well as permutation and simulation studies based on a recent basket trial designed to estimate the effectiveness of vemurafenib in BRAF^V600 mutant non‐melanoma among six primary disease sites and histologies.