Prediction of axillary lymph node pathological complete response to neoadjuvant therapy using nomogram and machine...

Area of Science:

• Oncology
• Medical Imaging
• Biostatistics

Background:

• Accurate prediction of axillary lymph node pathological complete response (apCR) is crucial for tailoring neoadjuvant therapy (NAT) in early breast cancer (eBC).
• Current methods for assessing treatment response in axillary lymph nodes can be limited.
• Nomogram and machine learning approaches offer potential for improved prediction accuracy.

Purpose of the Study:

• To evaluate the feasibility of using nomogram and machine learning models to predict apCR in patients with eBC undergoing NAT.
• To compare the predictive performance of these two distinct methodologies.

Main Methods:

• Retrospective analysis of 247 eBC patients who received NAT.
• Calculation of maximum diameter change of the primary lesion (MDCPL) and lymph node score (LNS) from pre- and post-NAT ultrasound data.
• Development of a nomogram using logistic regression and a random forest (RF) machine learning model for apCR prediction.

Main Results:

• MDCPL, LNS changes, N stage, and HER2 status were identified as independent predictors of apCR.
• The nomogram achieved an area under the curve (AUC) of 0.74 for the training set and 0.76 for the validation set.
• The RF model demonstrated a higher AUC of 0.85 in the final validation set, incorporating various clinicopathological factors.

Conclusions:

• Both nomogram and machine learning models demonstrate strong predictive capabilities for apCR.
• Nomograms offer simplicity and practicality, while machine learning effectively utilizes comprehensive clinicopathological data.
• These predictive models can significantly assist surgeons in optimizing axillary surgery decisions for eBC patients.