Dose-escalation strategies which use subgroup information.

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Dose-escalation trials typically assume a uniform patient population, which can mask treatment benefits in specific subgroups.
• Excluding potentially responsive subgroups or diluting treatment effects in homogeneous models can lead to missed therapeutic opportunities.

Purpose of the Study:

• To extend standard Bayesian dose-escalation methods to incorporate subgroup effects.
• To improve the identification of efficacious doses for diverse patient populations.

Main Methods:

• A Bayesian model-based dose-escalation framework was enhanced by including covariates for subgroup membership in the dose-toxicity model.
• Spike and slab priors for variable selection were proposed to continuously assess subgroup effects, addressing data inefficiency and low power of traditional hypothesis tests.

Main Results:

• The proposed method efficiently utilizes trial data throughout the dose-escalation process.
• It enables continuous assessment and identification of subgroup effects and recommended doses.

Conclusions:

• The novel Bayesian approach for dose-escalation, incorporating subgroup effects via spike and slab priors, is promising and feasible.
• This method enhances the ability to detect and utilize treatment efficacy across different patient subgroups.