Computationally Efficient Approach to Operational Prior Specification in Phase I Clinical Trials.

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Phase I clinical trials increasingly combine drug agents and schedules, necessitating flexible model-based designs.
• Model-based designs require pre-specified parameters, often optimized via computationally intensive simulation studies.
• Current parameter calibration methods, like grid search, are inefficient and costly.

Purpose of the Study:

• To develop efficient and systematic methods for calibrating design parameters in model-based Phase I clinical trials.
• To reduce the computational burden associated with parameter optimization and scenario evaluation.
• To maintain the operational characteristics of calibration methods while improving efficiency.

Main Methods:

• Introduced a novel 'cyclic calibration' method to replace computationally expensive 'grid search' approaches.
• Proposed a scenario reduction technique based on scenario complexity to streamline toxicity probability assessments.
• Evaluated the proposed methods for efficiency and performance in optimizing model-based trial designs.

Main Results:

• Cyclic calibration reduces computational requirements from multiplicative to additive, saving significant time.
• Scenario reduction methods decrease computation by over 500-fold.
• The proposed methods maintain similar operational characteristics to traditional grid search while drastically improving efficiency.

Conclusions:

• Efficient parameter calibration is crucial for improving model performance and saving computational resources in Phase I trials.
• Cyclic calibration and scenario reduction offer a computationally efficient alternative to conventional methods.
• These advancements facilitate more flexible and robust model-based designs for complex drug combination studies.