A nonparametric Bayesian basket trial design

Area of Science:

• Oncology
• Biostatistics
• Genomics

Background:

• Genomic aberration analysis guides targeted cancer therapies, improving prognosis and treatment outcomes.
• Matching patients to therapies based on specific genomic alterations is a key strategy in cancer management.
• Identifying patient subpopulations who benefit most from targeted therapies across diverse cancer types is essential.

Purpose of the Study:

• To propose an adaptive Bayesian clinical trial design for effective patient allocation and subpopulation identification.
• To develop a decision-theoretic framework incorporating utility functions and probability models for subpopulation analysis.
• To enable adaptive allocation of patients to superior treatments based on identified subpopulations.

Main Methods:

• Utilizing a flexible nonparametric Bayesian survival regression with random covariate-dependent patient partitioning.
• Employing a decision-theoretic approach with a flexible utility function reflecting clinical requirements.
• Implementing adaptive patient allocation to optimize treatment assignment within identified subpopulations.

Main Results:

• The proposed adaptive Bayesian design demonstrates desirable operating characteristics through extensive simulations.
• The method effectively identifies patient subpopulations likely to benefit from specific targeted therapies.
• Simulation studies show favorable comparisons against alternative clinical trial designs.

Conclusions:

• The developed adaptive Bayesian clinical trial design offers a robust method for patient allocation and subpopulation identification in targeted cancer therapy.
• This approach facilitates personalized medicine by matching patients to the most effective aberration-specific treatments.
• The findings support the advancement of precision oncology through optimized clinical trial strategies.