Subgroup identification in dose-finding trials via model-based recursive partitioning

Area of Science:

• Pharmacometrics
• Clinical Trial Design
• Biostatistics

Background:

• Identifying patient subgroups that respond differently to experimental treatments is vital in early-phase drug development.
• Existing subgroup identification methods are primarily designed for two-arm randomized clinical trials, with less focus on dose-finding trials.

Purpose of the Study:

• To propose and evaluate new strategies for subgroup analyses specifically within dose-finding clinical trials.
• To address the unique challenges presented by subgroup identification in settings with multiple dose randomization.

Main Methods:

• Utilizing model-based recursive partitioning, a method previously applied to two-arm trials.
• Assessing the viability of this method through simulations and a real-world clinical trial example.

Main Results:

• Model-based recursive partitioning effectively identifies patient subgroups with distinct dose-response curves.
• This approach enhances the estimation of treatment effects and minimum effective doses when patient heterogeneity is present.

Conclusions:

• Model-based recursive partitioning is a viable and effective method for subgroup analysis in dose-finding trials.
• This technique improves upon models that do not account for patient heterogeneity, leading to more precise drug development insights.