Comparison of models to predict nonsentinel lymph node status in breast cancer patients with metastatic sentinel lymph nodes: a prospective multicenter study
Related Experiment Video
Updated: Jun 24, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Charles Coutant1, Camille Olivier, Eric Lambaudie
1Department of Obstetrics and Gynecology, Hôpital Tenon, 4 rue de la Chine, Paris 75020, France. charles.coutant@tnn.aphp.fr
Purpose:
Several models have been developed to predict nonsentinel lymph node (non-SN) status in patients with breast cancer with sentinel lymph node (SN) metastasis. The purpose of our investigation was to compare available models in a prospective, multicenter study.
Patients And Methods:
In a cohort of 561 positive-SN patients who underwent axillary lymph node dissection, we evaluated the areas under the receiver operating characteristic curves (AUCs), calibration, rates of false negatives (FN), and number of patients in the group at low risk for non-SN calculated from nine models. We also evaluated these parameters in the subgroup of patients with micrometastasis or isolated tumor cells (ITC) in the SN.
Results:
At least one non-SN was metastatic in 147 patients (26.2%). Only two of nine models had an AUC greater than 0.75. Three models were well calibrated. Two models yielded an FN rate less than 5%. Three models were able to assign more than a third of patients in the low-risk group. Overall, the Memorial Sloan-Kettering Cancer Center nomogram and Tenon score outperform other methods for all patients, including the subgroup of patients with only SN micrometastases or ITC.
Conclusion:
Our study suggests that all models do not perform equally, especially for the subgroup of patients with only micrometastasis or ITC in the SN. We point out available evaluation methods to assess their performance and provide guidance for clinical practice.
