Approximate Bayesian computation design for phase I clinical trials.

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Phase I clinical trials are crucial for determining the maximum tolerated dose (MTD) in new cancer treatments.
• Existing model-based designs risk misspecification, while algorithm-based designs are less efficient in data utilization.
• There is a need for efficient and robust phase I trial designs that avoid model assumptions.

Purpose of the Study:

• To introduce a novel approximate Bayesian computation (ABC) approach for phase I clinical trial design.
• To develop a design that is free of dose-toxicity curve assumptions and efficiently uses all accrued data.
• To evaluate the robustness and efficiency of the proposed ABC design compared to existing methods.

Main Methods:

• Proposed a new approximate Bayesian computation (ABC) method for phase I clinical trial dose escalation.
• The ABC design utilizes all available data without requiring a predefined dose-toxicity model.
• Conducted extensive simulation studies to compare the ABC design with traditional model-based and algorithm-based designs.

Main Results:

• The ABC design demonstrated robustness and efficiency in simulations compared to other phase I trial designs.
• The approach effectively aggregates all accrued trial information for dose assignment.
• Application to a selumetinib trial showed satisfactory performance of the ABC design.

Conclusions:

• The proposed approximate Bayesian computation design offers a simple, efficient, and robust alternative for phase I clinical trials.
• This method addresses limitations of existing designs by avoiding model assumptions and improving data utilization.
• The ABC design is a valuable addition to the repertoire of phase I clinical trial methodologies for cancer drug development.