Using an Interaction Parameter in Model-Based Phase I Trials for Combination Treatments? A Simulation Study.

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Phase I dose-finding studies increasingly explore multiple agents simultaneously.
• Model-based designs are common for guiding dose escalation/de-escalation in combination trials.
• Estimating complex combination-toxicity relationships with limited Phase I data (20-60 patients) presents challenges.

Purpose of the Study:

• To evaluate the impact of model complexity on the accuracy of combination dose selection in Phase I trials.
• To compare two variants of a 4-parameter logistic model with reduced parameters.
• To investigate the effect of specific modelling assumptions, such as interaction parameters.

Main Methods:

• Extensive simulation study comparing two 4-parameter logistic models.
• Development of a framework for calibrating prior distributions for fair model comparison.
• Analysis of selection accuracy and patient allocation under different model assumptions.

Main Results:

• The inclusion of an interaction parameter between two compounds did not improve the average accuracy of selecting the optimal dose combination.
• Models with the interaction parameter allocated fewer patients to the target combination during the trial.
• Reduced parameter models were studied to assess the trade-off between flexibility and estimation challenges.

Conclusions:

• Simplifying combination-toxicity models by omitting interaction parameters may be beneficial in Phase I trials.
• While more parameters offer flexibility, they do not necessarily enhance selection accuracy and can reduce patient allocation to the target dose.
• Careful consideration of model assumptions is crucial for efficient and accurate dose-finding in early-phase combination studies.