An optimal Bayesian predictive probability design for phase II clinical trials with simple and complicated endpoints

Area of Science:

• Clinical trial design
• Bayesian statistics
• Oncology drug development

Background:

• Traditional phase II clinical trial designs primarily use binary tumor response endpoints, which are insufficient for novel therapies.
• Emerging treatments like targeted agents and immunotherapy introduce complex endpoints (ordinal, nested, coprimary), complicating trial design.
• Existing methods struggle to unify the analysis of diverse endpoint types in phase II trials.

Purpose of the Study:

• To propose a unified and flexible Bayesian optimal phase II predictive probability (OPP) design.
• To accommodate various endpoint types, including binary, ordinal, nested, and coprimary endpoints.
• To provide a robust framework for decision-making in phase II clinical trials with novel therapeutics.

Main Methods:

• The proposed design utilizes a Bayesian approach with a Dirichlet-multinomial model to handle diverse endpoint types.
• At interim analyses, the Bayesian predictive probability of success is calculated based on observed data.
• Go/no-go decisions are made using this predictive probability, guiding trial continuation or termination.

Main Results:

• The Bayesian OPP design unifies the handling of binary and complex endpoints in phase II trials.
• Simulation studies demonstrate that the OPP design exhibits satisfactory operating characteristics.
• The design effectively controls type I error rates and optimizes power or sample size.

Conclusions:

• The Bayesian OPP design offers a simple, flexible, and unified approach for phase II clinical trials.
• This methodology is well-suited for trials involving novel therapies with complex endpoints.
• The OPP design facilitates efficient and statistically sound decision-making throughout the trial process.