Predicting pathologic complete response to neoadjuvant chemotherapy in breast cancer using a machine learning...

Area of Science:

• Oncology
• Machine Learning in Medicine
• Clinical Decision Support

Background:

• Existing prediction models for neoadjuvant chemotherapy (NACT) response in breast cancer often rely on standard statistical methods or complex data like imaging and gene expression.
• There is a need for accessible machine learning models utilizing solely clinicopathological features for predicting pathologic complete response (pCR).
• Such models can aid clinical decision-making across various healthcare settings.

Purpose of the Study:

• To develop and validate a robust machine learning model for predicting pCR in breast cancer patients undergoing NACT.
• To evaluate the model's performance using only quantitative clinicopathological features.
• To assess the clinical utility and potential impact on chemotherapy use.

Main Methods:

• Development and validation of logistic regression and machine learning models using the National Cancer Data Base (NCDB) and an external cohort.
• Comparison of model performance with and without the incorporation of quantitative clinicopathological features.
• Decision curve analysis to determine the clinical utility of the best-performing model.

Main Results:

• The machine learning model incorporating quantitative clinicopathological features achieved the highest discrimination (AUC: 0.785) and calibration.
• The model demonstrated superior performance in hormone receptor-positive/human epidermal growth factor receptor 2-negative (HR+/HER2-) breast cancer (AUC: 0.817).
• In an external cohort, the model maintained robust performance (AUC: 0.711 overall, 0.810 in HR+/HER2- subgroup), suggesting potential for reduced chemotherapy use.

Conclusions:

• A validated machine learning model can accurately predict pCR in breast cancer patients receiving NACT using clinicopathological data.
• The model shows particular promise for identifying HR+/HER2- patients less likely to respond to chemotherapy, enabling alternative treatment considerations.
• This model serves as a valuable baseline for future research integrating additional granular features.