Phase II trial design with Bayesian adaptive randomization and predictive probability.

Area of Science:

• Clinical Trial Design
• Biostatistics
• Bayesian Inference

Background:

• Traditional phase II clinical trials often lack efficiency in patient allocation and early stopping criteria.
• Frequentist group sequential designs can be rigid and may not fully leverage accumulating data for adaptive decision-making.

Purpose of the Study:

• To propose a novel Bayesian adaptive randomization and predictive probability monitoring design for phase II clinical trials.
• To enhance patient outcomes by directing more participants to efficacious treatments and enabling early trial termination.
• To provide a robust Bayesian alternative to frequentist group sequential designs.

Main Methods:

• Utilized Bayesian adaptive randomization to dynamically assign patients to more effective treatment arms based on posterior probabilities.
• Implemented predictive probability monitoring for continuous trial oversight and early stopping decisions.
• Developed two methods for calculating predictive probability, accounting for future sample size uncertainty.

Main Results:

• The proposed design effectively allocates more patients to superior treatments.
• Early stopping for overwhelming treatment superiority or equivalence was demonstrated.
• Simulation studies confirmed favorable operating characteristics, including control of Type I and Type II errors.

Conclusions:

• The coupled Bayesian adaptive randomization and predictive probability approach optimizes phase II trial efficiency and patient benefit.
• This design offers a flexible and statistically sound method for concluding treatment superiority or equivalence.
• The proposed Bayesian methodology provides a valuable alternative to existing frequentist trial designs.