Information Borrowing in Basket Trials

Bayesian methods allow for the incorporation of prior knowledge, in terms of either expert opinion from clinicians or historical data, in statistical inference. In a Bayesian framework, borrowing from historical data is equivalent to considering informative priors. Information borrowing is often desirable in the analysis of basket trials and trials for which external information is available.

In a recent webinar in our Innovations in Bayesian Clinical Trials Virtual Symposium, Tianjian Zhou, Assistant Professor in the Department of Statistics at Colorado State University, introduces some of his recent work on these topics. Tianjian proposes a formal Bayesian hypothesis testing framework for data analysis and decision-making in basket trials using historical and real world data.

A basket trial studies one drug in multiple cancer subtypes where each of these subtypes is called a basket. With recent advances in precision medicine, there has been a growth in personalized treatments which target a common genetic mutation instead of a cancer type. Many different types of cancer can possess the same genetic mutation and as a result the drug might be effective for all these cancer types.

It is expected that the Treatment effect of a new drug in one specific basket may provide information for treatment effects of the same drug in other baskets. Hence, the approach of information borrowing can be used for the analysis of Basket trials. Ideally, by using information borrowing, we can improve the power of detecting a treatment effect and speed up drug development by enrolling a wider patient population (e.g., those with rare cancers).

Existing methods for information borrowing: Bayesian hierarchical model

Bayesian hierarchical model (BHM) proposed by Berry et al.¹, adaptively borrows information across baskets to improve the statistical power of basket trials. It is a strong design for addressing possibly differential effects in different groups. In this model, the shrinkage parameter, which controls the strength of information borrowing, is treated as an unknown parameter following a noninformative prior. The data is allowed to determine how much information should be borrowed across tumor subgroups.

However, some research shows that Bayesian hierarchical models cannot appropriately determine the degree of information borrowing and may lead to substantially inflated type I error rates.

RoBoT: a robust Bayesian hypothesis testing method for basket trials

Tianjian Zhou and Yuan Ji, Professor of Biostatistics (with tenure) at The University of Chicago, propose a novel Bayesian method, RoBoT (Robust Bayesian Hypothesis Testing), for the data analysis and decision-making in phase II basket trials. In contrast to most existing methods that use posterior credible intervals to determine the efficacy of the new treatment, RoBoT builds upon a formal Bayesian hypothesis testing framework that leads to interpretable and robust inference. It is assumed that the baskets belong to several latent subgroups, and within each subgroup, the treatment has similar probabilities of being more efficacious than controls, historical, or concurrent. The number of latent subgroups and subgroup memberships are inferred by the data through a Dirichlet process mixture model. Such model specification helps avoid type I error inflation caused by excessive shrinkage under typical hierarchical models. The operating characteristics of RoBoT are assessed through computer simulations and are compared with existing methods.

Click the button to access the on demand webinar and register for other webinars in Cytel’s Bayesian symposium.