Exploring Sensitive Biomarkers Associated With Short-Term Responses and Long-Term Outcomes Using Bayesian Additive...

Area of Science:

• Biostatistics
• Precision Medicine
• Clinical Trial Analysis

Background:

• Optimizing patient selection for new treatments is crucial in precision medicine.
• Early biomarker responses can guide treatment decisions, but their predictive value for long-term outcomes needs robust statistical investigation.

Purpose of the Study:

• To propose and evaluate a two-stage subgroup analysis method for assessing the predictive value of short-term biomarker responses on long-term treatment benefits.
• To identify patient subpopulations who are most likely to benefit from a new treatment based on early response indicators.

Main Methods:

• Utilized Bayesian Additive Regression Trees (BART) for counterfactual modeling to derive predictive conditional treatment effects (PCTE).
• Employed a two-stage subgroup analysis: first, deriving PCTE based on short-term biomarker response, and second, analyzing long-term outcomes within identified subgroups.
• Conducted extensive simulation studies to validate the method's performance.

Main Results:

• The proposed method demonstrated the predictive value of short-term biomarker responses for long-term treatment outcomes.
• Simulation studies confirmed the operating characteristics of the statistical approach.
• Application to a breast cancer trial illustrated the identification of a sensitive subpopulation.

Conclusions:

• Short-term biomarker responses can effectively predict long-term treatment benefits.
• The developed statistical approach provides a valuable tool for precision medicine, enabling the identification of patient subgroups likely to benefit from novel therapies.