Risk-group-specific dose finding based on an average toxicity score.

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Optimizing drug dosage is crucial for maximizing efficacy and minimizing toxicity in clinical trials.
• Existing dose-finding methods often do not fully account for varying toxicity severities or patient-specific susceptibility.
• Heterogeneity in patient response necessitates adaptive trial designs.

Purpose of the Study:

• To develop a Bayesian dose-finding design that integrates toxicity severity and patient heterogeneity.
• To establish a robust framework for characterizing overall toxicity using an average toxicity score (ATS).
• To improve the precision and safety of dose escalation in early-phase clinical trials.

Main Methods:

• Utilized a Bayesian approach with a multinomial likelihood and Dirichlet prior to model toxicity scores.
• Defined severity scores for different toxicity levels.
• Incorporated patient risk stratification and Bayesian isotonic transformation for order-restricted inference on ATS.
• Simulated trial performance for a multiple myeloma study.

Main Results:

• The proposed design effectively models dose-dependent toxicity with varying severity.
• Accounting for patient heterogeneity improved the accuracy of dose selection.
• Simulations demonstrated the design's feasibility and performance in a realistic clinical trial setting.

Conclusions:

• The novel Bayesian design offers a more refined approach to dose-finding by considering toxicity severity and patient susceptibility.
• This method can lead to safer and more efficient dose selection in oncology drug development.
• The framework provides a valuable tool for adaptive clinical trial design.