Bayesian dose-finding designs for combination of molecularly targeted agents assuming partial stochastic ordering

Area of Science:

• Oncology
• Biostatistics
• Clinical Trial Design

Background:

• Molecularly targeted agent (MTA) combination therapy is a developing field.
• Current dose-finding trials for MTA combinations face challenges in optimizing both toxicity and efficacy.
• Existing designs often rely on parametric models, which may not fully capture complex dose-response relationships.

Purpose of the Study:

• To propose novel Bayesian dose-finding designs for identifying optimal biological dose combinations (OBDCs) in MTA therapy.
• To move beyond parametric model assumptions in dose-finding for MTA combinations.
• To develop a computationally efficient method for identifying OBDCs in early-phase clinical trials.

Main Methods:

• Developed Bayesian dose-finding designs based on partial stochastic ordering assumptions.
• Utilized isotonic regression to estimate marginal posterior distributions of efficacy and toxicity probabilities.
• Created a dose-combination-finding algorithm to identify OBDCs.

Main Results:

• The proposed method effectively accounts for partial ordering constraints and potential plateaus in dose-response surfaces.
• Simulations demonstrated desirable operating characteristics compared to alternative designs.
• The approach is computationally efficient and robust.

Conclusions:

• The proposed Bayesian designs offer a flexible and efficient approach for identifying OBDCs in MTA combination trials.
• These methods provide a valuable alternative to parametric models, particularly when dose-response relationships are complex.
• The findings support the application of these designs in early-phase oncology clinical trials.