A surface-free design for phase I dual-agent combination trials

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Combination therapies are increasingly common in oncology.
• Existing Phase I dose-escalation designs often rely on restrictive parametric models.
• There is a need for flexible and robust dose-finding designs for combination treatments.

Purpose of the Study:

• To propose a novel model-free dose-finding design for Phase I oncology combination trials.
• To evaluate the performance of the proposed design against existing model-based and model-free methods.
• To demonstrate the design's ability to identify target combinations with reduced toxicity.

Main Methods:

• Developed a model-free design based on the assumption of monotonicity for each agent.
• Parametrized ratios between neighboring combinations using independent Beta distributions.
• Compared the proposed design with the continual reassessment method and the product of independent beta design via simulation.

Main Results:

• The proposed design achieved comparable or superior proportions of correct target combination selection.
• The design resulted in an equal or lower average number of toxic responses.
• Demonstrated robustness without requiring pre-specified parametric models or knowledge of toxicity ordering.

Conclusions:

• The novel model-free design offers an effective and safer alternative for Phase I oncology combination trials.
• This approach provides flexibility by avoiding restrictive model assumptions.
• The design shows promise for optimizing drug combinations in clinical settings.