All Times EDT
Keywords: Bayesian Additive Regression Trees; Individual patient data; Machine learning; Subgroup borrowing; Basket trials
Commonly used in the oncology field, basket trial designs are characterized by the underlying notion that the presence of a molecular marker can predict response to a targeted therapy. One of key objective of the basket trial is to evaluate the treatment effect of a targeted therapy in patients carrying the same genetic or molecular mutation. Various Bayesian hierarchical models have been proposed in these basket trial settings. However, these models rely on the underlying assumption that patients with similar characteristics will have a similar outcome to the same treatment. Patient populations within each subgroup must subjectively be deemed similar enough to borrow response information across subgroups. We propose utilizing the machine learning method of Bayesian Additive Regression Trees (BART) to provide a method for subgroup borrowing that does not rely on an underlying assumption of homogeneity between subgroups. BART is a data-driven approach that utilizes patient-level observations. The amount of borrowing between subgroups is automatically adjusted as BART learns the individual covariate-response relationships. Modeling patient-level data rather than treating the subgroup as a single unit minimizes any assumptions regarding homogeneity across subgroups. Having the amount of borrowing be analytically determined and controlled for based on the similarity of individual patient-level characteristics will allow for more objective decision making in the Basket trials.