Optimal Sequential Predictive Probability Designs for Early-Phase Oncology Expansion Cohorts.

Area of Science:

• Clinical trial design
• Pharmacology
• Biostatistics

Background:

• Traditional 3+3 dose-escalation designs are often unsuitable for targeted therapies.
• Phase I dose-expansion cohorts are crucial for safety and efficacy assessment but often lack rigorous design.
• Existing methods may not optimally balance resource preservation with statistical accuracy.

Purpose of the Study:

• To introduce a novel approach for designing phase I dose-expansion cohorts using sequential predictive probability monitoring.
• To develop optimization criteria for early trial termination due to futility.
• To improve the efficiency and ethical considerations of early-phase clinical trials.

Main Methods:

• Proposed two optimization criteria for trial design: one for efficiency and one for accuracy.
• Utilized sequential predictive probability monitoring for adaptive trial adjustments.
• Demonstrated the design's utility through simulation and a case study of atezolizumab in urothelial carcinoma.

Main Results:

• The sequential predictive probability design outperformed existing methods like Simon's two-stage designs.
• Achieved increased statistical power and significantly reduced average sample size under the null hypothesis.
• The optimal efficiency design demonstrated desirable properties for resource preservation while maintaining error rate control.

Conclusions:

• The proposed optimal efficiency design enables early halting of dose-expansion cohorts when efficacy is not promising, preserving resources.
• Maintains traditional control of type I and type II error rates.
• Offers a more ethical and resource-conscious approach to early-phase clinical trial design for targeted therapies.