Background: Sentinel node (SN) biopsy has become accepted as a reliable method of predicting the state of the axilla in breast cancer. The key issue, however, is the accuracy of the pathological evaluation of the biopsied node, which should be done intraoperatively whenever possible.
Methods: In our initial experience on 192 patients using a conventional intraoperative frozen section method, the false-negative rate was 6.3%, and the negative predictive value was 93.7%. We devised a new and exhaustive intraoperative method, requiring about 40 minutes, in which pairs of sections are taken every 50 microm for the first 15 sections and every 100 microm thereafter, sampling the entire node. Sentinel node metastases were found in 143 of the 376 T1N0 cases examined (38%).
Results: Metastases were always identified on hematoxylin and eosin sections, although in 4% of cases, cytokeratin immunostaining on adjacent sections was useful for confirming malignancy. In 233 patients the SNs were disease-free; of these patients, 222 had metastasis-free axillary nodes, and 11 (4.7%) had another metastatic node.
Conclusion: Extensive intraoperative examination of frozen sentinel nodes correctly predicts an uninvolved axilla in 95.3% of cases (negative predictive value). This method is, therefore, suitable for identifying patients in whom axillary dissection can be avoided.