Prediction Models for Sentinel Lymph Node Metastasis in Clinically Node-Negative Breast Cancer: Validation of...

Area of Science:

• Oncology
• Surgical Oncology
• Biostatistics

Background:

• Axillary management in early breast cancer (EBC) is shifting towards surgical de-escalation.
• Sentinel lymph node biopsy (SLNB) reduces morbidity but carries risks and may be unnecessary for low-risk patients.
• Accurate prediction of sentinel lymph node (SLN) metastasis is crucial for risk-adapted omission strategies.

Purpose of the Study:

• To validate existing prediction models for SLN macrometastasis.
• To develop a new cohort-derived prediction model.
• To evaluate ensemble modeling for improved SLN macrometastasis prediction in clinically node-negative (cN0) EBCs.

Main Methods:

• Retrospective cohort study of 1080 women with cN0 EBCs (2012-2020).
• External validation of Memorial Sloan Kettering Cancer Center Nomogram (MSKCCN) and MD Anderson Cancer Center Nomogram (MDACCN).
• Development of a cohort-derived generalized linear model (GLM) and two ensemble models; performance assessed by AUC, calibration, and Brier score.

Main Results:

• 17.3% of patients had nodal macrometastases.
• The MDACCN had an AUC of 0.78; the cohort-derived GLM achieved an AUC of 0.83.
• An ensemble model showed an AUC of 0.82 and the lowest Brier score (0.103), identifying a larger low-risk group (37%) with a 2.5% observed macrometastasis rate.

Conclusions:

• Routinely available clinicopathological variables can identify cN0 EBC patients at very low risk of SLN macrometastasis.
• Ensemble modeling enhances classification performance and expands the identification of low-risk patients.
• Risk-adapted prediction tools can support selective axillary de-escalation in EBC management.