Combining biomarkers to optimize patient treatment recommendations

Area of Science:

• Biostatistics
• Genomic Medicine
• Oncology

Background:

• Predictive markers can personalize cancer treatment, improving patient outcomes.
• The OncotypeDX RecurrenceScore aids chemotherapy decisions in estrogen-receptor-positive breast cancer.
• Existing methods for combining markers are often limited to predicting outcomes under a single treatment.

Purpose of the Study:

• To develop and evaluate a novel method for combining multiple genetic markers for optimal cancer treatment selection.
• To model treatment effect as a function of markers to identify patients most likely to benefit from specific therapies.
• To compare the performance of the proposed marker combination method against existing approaches.

Main Methods:

• A boosting approach is proposed, iteratively fitting models by upweighting subjects misclassified regarding treatment benefit.
• The method models treatment effect as a function of genetic markers.
• A simulation study compares the boosting approach with existing methods using expected outcome changes.

Main Results:

• The proposed boosting approach demonstrates improved performance in some settings and comparable performance in others compared to existing methods.
• The simulation study provides insights into the relative strengths and weaknesses of different marker combination methodologies.
• Application to breast cancer data suggests potential for improved treatment selection marker combinations.

Conclusions:

• The novel boosting method offers a promising strategy for developing optimized marker combinations for personalized cancer treatment selection.
• This approach enhances the ability to identify patients who will benefit most from specific therapies, moving beyond single-treatment outcome prediction.
• Further application and validation of this method could refine treatment decisions in oncology, particularly for hormone-therapy-treated breast cancer.