Non-Sentinel Lymph Node Metastases in Breast Cancer Patients with a Positive Sentinel Lymph Node: Validation of Five Nomograms and Development of a New Predictive Model
| Autor: | Augusto Lombardi, Stefano Maggi, Marzia Lo Russo, Francesco Scopinaro, Domenica Di Stefano, Maria Grazia Pittau, Simone Tiberi, Claudio Amanti |
|---|---|
| Rok vydání: | 2011 |
| Předmět: |
Adult
Cancer Research Breast Neoplasms Unnecessary Procedures Sensitivity and Specificity nomogram non-sentinel lymph node 03 medical and health sciences breast cancer sentinel lymph node 0302 clinical medicine Predictive Value of Tests Humans 030212 general & internal medicine Aged Retrospective Studies Models Statistical Sentinel Lymph Node Biopsy General Medicine Middle Aged Nomograms Italy ROC Curve Oncology Lymphatic Metastasis 030220 oncology & carcinogenesis Axilla Lymph Node Excision Female Lymph Nodes |
| Zdroj: | Tumori Journal. 97:749-755 |
| ISSN: | 2038-2529 0300-8916 |
| Popis: | Aims and Background Discordance of intraoperative analysis with definitive histology of the sentinel lymph node in breast cancer leads to completion axillary lymph node dissection, which only in 35–50% shows additional nodal metastases. The aim of the study was to identify individual patient risk for non-sentinel lymph node metastases by validating several statistical methods present in the recent literature and by developing a new tool with the final goal of avoiding unnecessary completion axillary lymph node dissection. Methods We retrospectively evaluated 593 primary breast cancer patients. Completion axillary lymph node dissection was performed in 139 with a positive sentinel lymph node. The predictive accuracy of five published nomograms (MSKCC, Tenon, Cambridge, Stanford and Gur) was measured by the area under the receiver operating characteristic curve. We then developed a new logistic regression model to compare performance. Our model was validated by the leave-one-out cross-validation method. Results In 53 cases (38%), we found at least one metastatic non-sentinel lymph node. All the selected nomograms showed values greater than the 0.70 threshold, and our model reported a value of 0.77 (confidence interval = 0.69–0.86 and error rate = 0.28) and 0.72 (confidence interval = 0.63–0.81, error rate = 0.28) after the validation. With a 5% cutoff value, sensitivity was 98% and specificity 9%, for a cutoff of 10%, 96% and 2%, respectively. Conclusions All the nomograms were good discriminators, but the alternative developed model showed the best predictive accuracy in this Italian breast cancer sample. We still confirm that these models, very accurate in the institution of origin, require a new validation if used on other populations of patients. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|