Subgroup-specific dose finding for phase I-II trials using Bayesian clustering | JoVE Visualize

Area of Science:

Biostatistics
Clinical Trial Design
Pharmacometrics

Background:

Traditional dose-finding trials often assume patient homogeneity, which can be inefficient and lead to suboptimal dosing in heterogeneous populations.
Identifying the optimal dose for immunotherapeutic oncology treatments requires balancing both efficacy and toxicity outcomes.
Existing methods for handling subgroup heterogeneity in dose-finding may impose restrictive assumptions like subgroup exchangeability or ordering.

Purpose of the Study:

To develop and evaluate an adaptive Bayesian clustering method for dose-finding in clinical trials with heterogeneous patient populations.
To accommodate situations where both efficacy and toxicity data are crucial for dose optimization.
To provide a flexible approach that does not require a priori assumptions about subgroup relationships or ordering.

Main Methods:

An adaptive Bayesian clustering algorithm was developed to borrow strength among similar subgroups and cluster homogeneous ones.
The methodology allows for the incorporation of subgroup-specific prior information if available.
Operating characteristics were compared against Bayesian hierarchical models for subgroups through simulation studies.

Main Results:

Both the proposed clustering method and Bayesian hierarchical models outperformed separate subgroup-specific models when all subgroups shared similar dose-efficacy and dose-toxicity curves.
The adaptive Bayesian clustering method demonstrated superior performance compared to hierarchical models when one subgroup exhibited distinct dose-efficacy or dose-toxicity curves.

Conclusions:

The proposed adaptive Bayesian clustering method offers an effective approach for dose-finding in heterogeneous populations, particularly when subgroup responses vary.
This method provides a more robust alternative to traditional homogeneity assumptions and existing subgroup methodologies.
It enhances the precision of optimal dose identification in complex clinical trial settings, such as immunotherapies.