[Comparison of three models to predict non-sentinel lymph node status in breast cancer patients with metastatic sentinel lymph nodes: a retrospective study of 77 patients]

Various statistic predictive models have been developed to predict the status of the non-sentinel lymph nodes (NSLN) when the sentinel one was invaded by tumor cells in the case of breast invasive carcinoma. The objective of this study was to compare the accuracy of three of these statistic models: the MSKCC nomogram, the Tenon score and the MDA score. The study was conducted at Croix Rousse Hospital, Lyon, France.When sentinel lymph node is invaded by metastasis, complete lymph node dissection of the axilla can be avoided using statistic predictive models especially when talking about micrometastases and moreover about isolated tumor cells in the sentinel lymph node. Over the different existing models, we tested the three most used ones to determine the model that is most adapted to our everyday practice.The study population consisted of 77 women with an invasive breast cancer treated by complete axillary lymph node dissection for metastatic sentinel lymph node. Over the nine models already published, we focused on three of them due to their ease to use as well as their good preliminary results: the Memorial Sloan-Kettering Cancer Center nomogram (MSKCC), the Tenon score and the M.D Anderson score (MDA). Different criteria are used by these models, but all of them take in consideration: tumor size, multifocality, tumor type, lymphovascular invasion, hormonal receptors, number of sentinel lymph node resected, degree of sentinel lymph node invasion (i.e. macrometastasis, micrometastasis or isolated tumor cells) and histologic method used to diagnose sentinel lymph node invasion. These scores are validated depending on their false negative ratio as well as the proportion of patient selected as having a low risk of non-sentinel lymph node invasion. The major criterion for the MSKCC score is the calibration that compares the predicted likelihood of invasion and the histologically proved one.The rate of non-sentinel lymph node invasion was 37.6%. The discrimination of the three models was good with an AUC of 0.74 for both MSKCC nomogram and Tenon score and of 0.72% for MDA score. MSKCC nomogram was well calibrated (P=0.23). The false negative ratio was 0 (CI 95%: 0-18.1%) for MSKCC nomogram, 16.7% (CI 95%: 8.6%-27.8%) for Tenon score and 16% (CI 95%: 6.8-31%) for MDA score. The percentage of patients considered as having a low risk of invasion of the non-sentinel lymph node was 20.5% for MSKCC nomogram, 34.7% for MDA score and 47.4% for Tenon score.MSKCC nomogram seems to be the statistic model that is most adapted to the population we studied. To validate the Tenon and MDA score, we should have a most important population.