Borrowing information across subgroups in phase II trials: is it useful?

Area of Science:

• Oncology
• Biostatistics
• Clinical Trial Design

Background:

• Human tumors exhibit heterogeneity, leading to distinct patient subgroups with varying molecular/histologic features.
• Designing phase II cancer trials faces challenges in balancing subgroup-specific evaluations with resource constraints.
• Ignoring patient heterogeneity in pooled analyses risks missing treatments effective only in certain subgroups.

Purpose of the Study:

• To evaluate the benefits of a hierarchical Bayesian approach versus simple pooling for phase II cancer trials.
• To assess the efficacy of information borrowing across patient subgroups in clinical trial settings.

Main Methods:

• Simulations were conducted to compare the hierarchical Bayesian approach with a simpler pooling method.
• The study focused on phase II trial designs for screening anticancer agents.

Main Results:

• The hierarchical Bayesian approach demonstrated suboptimal performance in the simulated phase II settings.
• Insufficient outcome data was available for the hierarchical Bayesian model to appropriately determine information borrowing across subgroups.
• Simple pooling approaches may be advantageous when uniform anticancer agent activity across subgroups is anticipated.

Conclusions:

• The hierarchical Bayesian approach, as implemented, may not be suitable for phase II cancer trials due to data limitations in determining information borrowing.
• Simple pooling strategies can be effective in phase II trials when there is a strong rationale for expecting consistent drug activity across all patient subgroups.