A Bayesian response-adaptive covariate-balanced randomization design with application to a leukemia clinical trial

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Optimizing treatment allocation in clinical trials is crucial for identifying effective therapies.
• Balancing patient covariates across treatment arms is essential for unbiased comparisons.
• Existing randomization methods may not fully address both efficacy-driven allocation and covariate balance simultaneously.

Purpose of the Study:

• To propose a novel Bayesian response-adaptive covariate-balanced (RC) randomization design.
• To enhance the efficiency and reliability of multiple-arm comparative clinical trials.
• To skew treatment allocation towards more efficacious arms while maintaining covariate balance.

Main Methods:

• Developed a new covariate-adaptive randomization (CA) method using a prognostic score.
• The CA method accommodates continuous and categorical factors, assigning covariate importance weights.
• Integrated the CA design into a group sequential response-adaptive randomization (RA) scheme.

Main Results:

• The proposed RC design effectively balances covariates across treatment arms.
• The design successfully skews allocation probabilities towards more efficacious treatment options.
• Demonstrated the design's utility in a phase II leukemia clinical trial via simulations.

Conclusions:

• The Bayesian RC randomization design offers a robust approach for clinical trials.
• This design integrates the benefits of covariate-adaptive and response-adaptive randomization.
• The method shows promise for improving the statistical power and ethical considerations in comparative trials.