Keyboard design for phase I drug-combination trials

Area of Science:

• Clinical Trial Design
• Biostatistics
• Pharmacology

Background:

• The Keyboard design is a successful model-assisted method for single-agent dose-finding trials.
• Existing methods for dual-agent trials can be complex and difficult to implement.

Purpose of the Study:

• To extend the Keyboard design for dual-agent dose-finding trials.
• To maintain the simplicity and improve the performance of model-assisted dose-finding methods.

Main Methods:

• Extension of the single-agent Keyboard design to a dual-agent setting.
• Development of pre-tabulated dose escalation and deescalation rules.
• Evaluation using extensive simulations with a novel 2D dose-toxicity scenario generator.

Main Results:

• The Keyboard combination design demonstrates desirable theoretical properties, including optimal decision rules, coherent dose transitions, and convergence to the target dose.
• Simulation results confirm competitive operating characteristics compared to other methods.
• An R Shiny application is available for practical implementation.

Conclusions:

• The Keyboard combination design is a simple, effective, and theoretically sound method for dual-agent dose-finding.
• This design offers a practical and efficient approach for clinical trials involving drug combinations.