Validation and comparison of nomograms and scores to predict non-sentinel lymph node status in breast cancer patients with metastatic sentinel lymph node: a prospective multicentric study.

Abstract

Abstract Abstract #203 Background
 Five tools have been developed to predict non-SN status in breast cancer patients with SN metastasis: two nomograms (the Memorial Sloan-Kettering Cancer Center nomogram (MSKCC nomogram) and the nomogram developed by Degnim) and three scoring systems (the Tenon score, the score from the M.D. Anderson Cancer Center (MDA score), and the score developed by Saidi). Our purpose was to assess the validity of these tools in a prospective multicentric study of 561 SN-positive patients.
 Methods
 For 561 SN-positive patients who underwent ALND, we tested the accuracy of the five mathematical models. We calculated the false-negative (FN) rates and the areas under the receiver operating characteristics curves (AUC) as a measure of discriminatory capacity. We also tested the accuracy of each model in the subgroup of 246 patients with micrometastasis or isolated tumor cells in the SN; and in the subgroup of 165 patients with histological tumor size measuring more than 2 cm. Results were compared to the optimal logistic regression model that could be developed on the 561 patient cohort.
 Results
 At least one non-SN was positive in 147 patients (26.2%). The Tenon score, MSKCC nomogram, MDA score, nomogram of Degnim and score of Saidi applied to all 561 patients achieved AUCs of 0.81, 0.78, 0.73, 0.73, and 0.65 respectively. The FN rate were 4.4% (12/273), 6.5% (13/201), 5.7% (8/140), 12.5% (2/16) and 15% (21/140) respectively. In the subgroup of patients with only SN micrometastases, the FN rates were 2.9% (6/209) and 3.5% (5/143) for the Tenon score and the MSKCC nomogram respectively. The AUCs were 0.81 and 0.72 respectively
 In the subgroup of patients with histological tumor size measuring more than 2 cm, the FN rates were 0% (0/10) and 28.6% (4/14) for the Tenon score and the MSKCC nomogram respectively. The AUCs were 0.69 and 0.66 respectively. For comparison, a logistic regression model developed on this multicentric cohort yielded AUCs of 0.83, 0.82, and 0.75 for the whole cohort, patients with micrometastasis or isolated tumor cells in the SN and in patients with tumor > 2 cm, respectively.
 Conclusion
 In this multicenter study, the Tenon score outperformed other scoring system. It was also more discriminant than other scores in patients with micrometastasis or isolated tumor cells in the SN and in patients with tumor > 2 cm. Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 203.