A simulation study of methods for selecting subgroup-specific doses in phase 1 trials.

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Patient heterogeneity can complicate dose-finding in phase 1 clinical trials when dose-toxicity relationships vary across subgroups.
• Separate trials for each subgroup may result in insufficient sample sizes, particularly for low-prevalence groups.
• Conducting a single trial while accounting for heterogeneity presents significant challenges.

Purpose of the Study:

• To develop and evaluate a novel dose-finding design for phase 1 clinical trials that accommodates patient heterogeneity.
• To propose a hierarchical Bayesian dose-toxicity model that allows for borrowing strength across exchangeable subgroups.
• To provide a framework for subgroup-specific dose selection and safety rules within a single trial.

Main Methods:

• A generalized continual reassessment method (CRM) based on a hierarchical Bayesian dose-toxicity model was developed.
• The model assumes exchangeability between subgroups, enabling data sharing to improve estimation.
• A simulation study compared the proposed hierarchical method against three non-hierarchical approaches.

Main Results:

• The hierarchical model-based method demonstrated superior performance in settings where dose-toxicity curves are exchangeable across subgroups.
• The proposed design effectively manages subgroup-specific dose selection and safety rules.
• Simulations confirmed the advantages of the hierarchical approach over non-hierarchical alternatives.

Conclusions:

• The hierarchical Bayesian dose-finding method is recommended for phase 1 trials with exchangeable patient subgroups.
• This approach offers a more efficient and robust alternative to traditional methods when dealing with heterogeneity.
• Practical guidelines and computational tools are provided for implementation in clinical trial design and conduct.