Bayesian adaptive dose-escalation designs for simultaneously estimating the optimal and maximum safe dose based on...

Area of Science:

• Clinical Trial Design
• Biostatistics
• Pharmacology

Background:

• Dose-escalation trials are crucial for identifying safe and effective drug doses for further studies.
• Traditional methods may not fully integrate both safety (dose-limiting toxicities) and efficacy signals.
• Accurate identification of optimal and maximum tolerated doses is essential for novel therapeutics.

Purpose of the Study:

• To introduce novel Bayesian adaptive dose-escalation designs incorporating dose-limiting toxicities (DLTs) and efficacy responses.
• To develop and compare 'single objective' and 'dual objective' dose-escalation strategies.
• To evaluate the performance of these strategies in identifying maximum tolerated dose (MTD) and optimal biological dose (OBD).

Main Methods:

• Employed a flexible nonparametric model for continuous efficacy responses and a logistic model for binary DLTs.
• Developed two Bayesian adaptive strategies: single objective (recommends one dose) and dual objective (estimates MTD and OBD).
• Utilized a gain function combining DLT probabilities and expected efficacy for dose escalation decisions.

Main Results:

• The nonparametric model effectively estimated efficacy responses across various true shapes.
• Dual-objective dose-escalation designs demonstrated superior performance in accurately identifying both MTD and OBD compared to single-objective designs.
• Simulations based on a type 2 diabetes trial example evaluated strategy performance and the impact of stopping rules.

Conclusions:

• The proposed Bayesian adaptive dose-escalation strategies effectively integrate safety and efficacy data.
• Dual-objective designs offer improved accuracy in identifying key target doses for drug development.
• These methods provide valuable information on drug safety and efficacy profiles to guide subsequent clinical studies.