Machine Learning-Based Prediction of Distant Recurrence Risk and Ribociclib Treatment Effect in HR+/HER2- Early...

Area of Science:

• Oncology
• Machine Learning in Medicine
• Clinical Decision Support

Background:

• Distant recurrence remains a significant concern for patients with HR+/HER2- EBC despite standard endocrine therapy.
• Personalized risk assessment is crucial for optimizing treatment strategies in early breast cancer.
• Machine learning offers potential for identifying complex risk factors and improving recurrence prediction.

Purpose of the Study:

• To develop and validate machine learning models for predicting distant recurrence risk in HR+/HER2- EBC.
• To identify key predictor variables associated with distant recurrence in this patient population.
• To aid clinical decision-making by providing individualized recurrence risk assessments.

Main Methods:

• Utilized a large, real-world dataset of HR+/HER2- EBC patients from the Flatiron Health Research Database.
• Employed gradient boosting for predictor variable identification and elastic net-penalized Cox models for prediction.
• Validated models internally with real-world data and externally using data from the NATALEE trial.

Main Results:

• The developed model demonstrated accurate distant recurrence prediction in the real-world cohort (C-index: 0.85) with sustained performance over 10 years.
• External validation with the NATALEE NSAI arm showed discriminative performance (C-index: 0.66), improved by training on NATALEE data (C-index: 0.70).
• The NATALEE-trained model predicted a 3.2% reduction in distant recurrence at 48 months with ribociclib treatment in the real-world cohort.

Conclusions:

• A robust machine learning model accurately predicts distant recurrence in HR+/HER2- EBC.
• Identified predictor variables and developed models can support personalized, risk-based treatment decisions.
• This approach holds promise for improving outcomes in early breast cancer management.