Validation of nomograms to predict the risk of non-sentinels lymph node metastases in North African Tunisian breast cancer patients with sentinel node involvement

Abstract

IntroductionIn approximately half of patients with breast cancer and lymph node metastases, the sentinel node (SN) is the only involved axillary node. Scoring systems have been developed to predict probability of non-SN metastases among those with a positive SN. The goal of the present study was to determine whether the five models (Memorial Sloan-Kettering Cancer Center (MSKCC), Stanford, Tenon, Cambridge and the Turkish model) accurately predicted non-SN involvement in a North African Tunisian population. MethodsDuring a five years period, we identified 87 cases of invasive breast cancer which had a positive SN biopsy and complete axillary lymph node dissection (CALND). The MSKCC, Stanford, Tenon, Cambridge and Turkish models were tested. Results were compared using the area under the curve (AUC) of the receiver operating characteristics for each model. False negative and false positive rates were also calculated. ResultsThe AUC of the MSKCC, Stanford, Tenon, Cambridge and Turkish models was respectively 0.73 (95% CI 0.6–0.86), 0.76 (95% CI 0.65–0.87), 0.75 (95% CI 0.63–0.87), 0.67 (95% CI 0.53–0.82) and 0.75 (95% CI 0.63–0.88). The threshold for a 10% false negative of non-SN involvement was obtained with a cut off value of 10% for MSKCC, 25% for Stanford, a score of 3 for Tenon, 6% for Cambridge and 15% for the Turkish nomogram. ConclusionsMeaningfully applied to our population, although AUC values had overlapping of 95% confidence intervals but combined our data suggest that the Stanford nomogram may be the most accurate. Before prospective trials validate these nomograms, CALND remains the standard for patients who have SN metastases.