COCA: a randomized Bayesian design integrating dose optimization and component contribution assessment for...

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Developing effective cancer combination therapies requires optimizing drug doses and understanding each component's contribution.
• Traditional methods necessitate large sample sizes in early-phase trials, posing challenges for drug developers.

Purpose of the Study:

• To propose a novel two-stage randomized phase II design for integrated combination dose optimization and component contribution assessment.
• To enhance trial efficiency and reduce sample size through adaptive data combination.

Main Methods:

• A two-stage randomized phase II design integrating dose optimization and contribution assessment.
• Stage 1: Optimal combination dose selection based on risk-benefit tradeoff.
• Stage 2: Multi-arm randomization to evaluate component contributions, with adaptive Bayesian logistic regression (spike-and-slab prior) for data pooling.

Main Results:

• The proposed design successfully achieves dual objectives of dose optimization and component contribution assessment.
• Extensive simulations demonstrate substantial sample size savings compared to existing designs.
• A novel calibration procedure ensures desired operating characteristics for sample size and decision cutoffs.

Conclusions:

• The proposed adaptive two-stage design offers an efficient approach for developing cancer combination therapies.
• This methodology addresses the critical need for optimizing doses and evaluating drug contributions while minimizing sample size.
• The design provides a valuable tool for accelerating drug development and improving the risk-benefit profile of combination treatments.