A Randomized Basket Trial Design for Dose Optimization Based on Bayesian Model Averaging Using Spike-and-Slab Priors

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• The U.S. Food and Drug Administration (FDA) recommends dose optimization for cancer drugs, emphasizing dose-response studies.
• Precision medicine and cancer biology advancements drive the development of targeted therapies across various cancer types.
• Basket trials assess new treatments simultaneously across multiple indications, aligning with precision medicine trends.

Purpose of the Study:

• To propose a novel dose-ranging basket trial design.
• To integrate FDA's dose optimization principles with basket trial methodologies.
• To address treatment heterogeneity across different cancer indications and dose levels.

Main Methods:

• A Bayesian model-averaging approach was developed for dose-ranging basket trials.
• The design considers both efficacy and toxicity outcomes.
• Indications and dose levels were defined as baskets to analyze heterogeneity.

Main Results:

• The proposed Bayesian approach demonstrated superior performance in simulations compared to existing methods.
• It achieved higher statistical power and better control of Type I error rates.
• The method enabled precise optimal dose selection and significant sample size savings, especially with heterogeneous treatment effects.

Conclusions:

• The proposed dose-ranging basket trial design effectively optimizes cancer drug development.
• It offers a statistically robust framework for evaluating targeted therapies across diverse indications.
• This approach supports efficient clinical trial design by accounting for treatment heterogeneity and optimizing dose selection.