Optimization of guidelines for Risk Of Recurrence/Prosigna testing using a machine learning model: a Swedish...

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Gene expression profiling aids treatment decisions in adjuvant hormone receptor-positive, HER2-negative breast cancer.
• Existing optimization algorithms focus on RS/Oncotype Dx, not the ROR/Prosigna assay.
• Accurate patient selection for genomic assays is crucial for optimizing adjuvant therapy.

Purpose of the Study:

• To develop and validate a machine learning model for enhanced pre-selection of patients for ROR/Prosigna testing.
• To improve the accuracy of risk stratification in HR+/HER2- breast cancer.
• To reduce the number of patients requiring ROR/Prosigna testing through better pre-selection.

Main Methods:

• A machine learning model was developed using prognostic factors (tumor size, progesterone receptor expression, grade, Ki67) in 348 postmenopausal women.
• The model predicted ROR/Prosigna output for patients with resected HR+/HER2- node-negative breast cancer.
• Performance was compared to existing risk stratification schemes regarding over- and undertreatment.

Main Results:

• The machine learning model demonstrated strong predictive performance with AUCs of 0.77 (training) and 0.83 (validation) for predicting chemotherapy indication.
• Validated upper and lower cut-offs improved risk stratification accuracy for low, intermediate, and high-risk disease.
• The model significantly reduced the proportion of patients needing ROR/Prosigna testing compared to current methods.

Conclusions:

• Machine learning algorithms can effectively enhance patient selection for gene expression profiling.
• The developed model improves risk stratification and reduces unnecessary testing for ROR/Prosigna.
• Further external validation is recommended to confirm the generalizability of these findings.