Background
An improved procedure that allows accurate detection of negative sentinel lymph node (SLN) and of SLN macrometastases during surgery would be highly desirable in order to protect patients from further surgery and to avoid unnecessary costs. We evaluated the accuracy of an intraoperative procedure that combines touch imprint cytology (TIC) and subsequent frozen section (FS) analysis. 2276 SLNs from 1072 patients with clinical node-negative early breast cancer were evaluated during surgery using TIC. Only cytologically-positive SLN were subsequently analysed with a single FS, preserving cytologically-negative SLN for the final postoperative histological diagnosis. Sensitivity, specificity and the accuracy of this approach were analysed by comparing the results from intra- and postoperative SLN and axillary node evaluation. This intraoperative method displayed 100% specificity for SLN metastases and was significantly more sensitive for prognostically relevant macrometastases (85%) than for micrometastases (10%). Sensitivity was highest for patients with two or more positive LNs (96%) than for those with only one (72%). 98% of the patients with final pN2a-pN3a were already identified during surgery. Patients who received primary axillary lymph node dissection had significantly more frequent metastases in further LNs (44.6%). Sensitivity was highest for patients with luminal-B, HER2+ and triple negative breast cancer and for any subtype if Ki-67 > 40%. TIC and subsequent FS of cytologically-positive SLNs is highly reliable for detection of SLN macrometastases, and allows accurate identification of patients with a high risk of extended axillary involvement during surgery, as well as accurate histological diagnosis of negative SLN.