A curve-free Bayesian decision-theoretic design for two-agent Phase I trials

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Bayesian statistical methods offer a natural framework for incorporating prior information in medical research.
• Complexity of traditional Bayesian methods has hindered their adoption in clinical trials.
• Phase I trials are crucial for determining safe drug dosages.

Purpose of the Study:

• To propose an accessible Bayesian design for two-agent Phase I trials.
• To simplify the implementation and reduce the computational burden of Bayesian methods in this context.
• To facilitate the use of prior toxicity data from single-agent trials.

Main Methods:

• A working model with conjugate priors was selected for analytical posterior distributions.
• A strategy for specifying priors based on single-agent trial toxicity data was developed.
• The proposed design aims for ease of understanding and implementation by clinicians.

Main Results:

• The proposed Bayesian design performs comparably to more complex methods.
• It offers a significant reduction in sample size without compromising performance.
• The dose-finding algorithm is conservative, ensuring a relatively safe approach.

Conclusions:

• The proposed Bayesian design enhances the adoption of these methods in Phase I trials.
• It provides a practical and efficient approach to dose-finding.
• The method balances statistical rigor with clinical applicability and safety.