Consecutive, consenting, clinically node negative women with unifocal, invasive breast cancer undergoing surgical treatment for invasive breast cancer between December 2004 and December 2006 were studied. All patients underwent synchronous excision of the primary breast cancer (either by wide local excision or mastectomy) and sentinel node biopsy by one of two consultant surgeons (TC or RGW). All women were counselled and fully consented regarding the risk of requiring formal axillary clearance if their sentinel node was found to contain metastases either on frozen section examination or on subsequent conventional analysis. The degree of risk communicated was dependent mostly on the patient's tumour size (no patient had palpable lymphadenopathy). Under the study protocol, any evidence of nodal metastases in the sentinel node (either by frozen section or on subsequent formal pathological examination, including micrometastases (i.e. deposits <2 mm in diameter), were taken to mandate subsequent full axillary clearance at a second operation. This was decided, despite the recent WHO classification of isolated tumour cells as conferring a node-negative status, because of both the experimental nature of the technique in our hands and the ongoing controversy of whether such cells actually are prognostically important4.

Our method of sentinel node identification incorporates the preoperative injection of both radioisotope and blue dye as previously described5. In brief, radioisotope tracer (99m-Tc-labelled colloidal rhenium sulphide-Nanocis, Cis Biointernational) was injected subcutaneously peri-tumourly one day prior to surgery. In cases of an impalpable tumour, fine wire localization (FWL) in combination with radio-isotope occult lesion localization (ROLL) methods was implemented6. Immediately preoperatively, after induction of general anesthesia, 3–5 mls of isosulphan blue dye (lymphazurin^®, Ben Venue Laboratories Inc., Bedford OH) was also injected peritumorally. The SLN was then identified intra-operatively by visual inspection for blue dye in combination with searching for radiation counts with a hand held gamma-ray detection probe (Neoprobe^®, Neoprobe Corporation). While the nodal tissue was being processed by frozen section (see below), the primary tumour was resected. The excised breast specimens were examined as usual to ascribe the type and grade (Bloom Richardson grading of hematoxylin and eosin-stained specimens) of the invasive component of the primary tumor, as well as its estrogen (ER) and progesterone receptor (PR) status.

The same consultant pathologist (RL) examined all the specimens. Once localized and excised, the SLN(s) was sent immediately to the laboratory for frozen section examination. On average, this involved examination of three sections per node were (using OCT compound and freezing spray) unless obvious macroscopic metastases were apparent on the first slice examined. Patients found to have tumour cells in their SLN biopsy on frozen section analysis underwent immediate level II axillary lymph node dissection with all resected tissue being sent for standard pathological processing and review. Any remaining sentinel nodal tissue after frozen sectioning were fixed in formalin and embedded in paraffin with subsequent examination by H&E staining (again an average of 3 levels per node with a fixed distance between levels of 100 microns.) and, if any uncertainty pertained, immunohistochemistry. With regard to the nodes excised as part of an axillary clearance, smaller nodes were bisected and the whole node examined as per the sentinel node. Larger nodes were bisected and one half examined similarly. If no macroscopic metastases were evident, the other half was then also examined.

The results of the immediate, intraoperative FS assessment of the SLN were then compared with that of their delayed formal pathological examination with regard to nodal oncological status. ANOVA testing was used to examine for statistically significant differences between the groups with the Bonferroni correction used post hoc because of small numbers involved.

In total, 94 women were included in the study. Their mean age was 60 years (range 35–81, 73 being older than 50). Twenty eight patients had breast-conserving surgery (two required radiological guidance to localise the primary) while the remainder underwent mastectomy due either to high tumor:breast size, central location of the tumour or the patient's own preference. On final pathological analysis of the resected primary, 37 patients were found to have T1 tumours (2 were T1a, 9 were T1b and 26 were T1c) while 43 were T2 and 14 T3. Seventy six cancers were ductal adenocarcinoma while 18 were lobular in type. Nine were Grade I by Bloom Richardson scoring on H&E examintion while 50 were Grade II and 35 Grade III. Seventy eight of cancers were ER positive with 12 being both ER and PR positive.

There were no failed lymphatic mappings. The median number of SLN identified was 2 (mean 2.1 range 1–6). Intraoperative frozen section of the sentinel nodes identified tumour cells in 23 women with the subsequent formal H&E examination confirming the presence of lymphatic metastases in every one. Each of these patients underwent immediate level II axillary clearance on receipt of the frozen section report. The mean number of nodes subsequently harvested at axillary clearance was 15.3 (range 3–39). Sixteen patients had further nodal metastases identified in their non-sentinel nodes.

Formal pathological examination identified nodal tumour deposits in seven patients whom the initial frozen section deemed not to have lymphatic dissemination. The frequency of false negative cases among those judged to be negative by frozen section analysis was therefore 9.9% while the false negative rate was 23%. Three of these patients actually had micrometastatic disease in their sentinel node (one of these patients required further immunohistochemical staining to confirm the presence of malignant cells on the formal specimen). All patients with involved nodes underwent level II axillary clearance at a second operation as per protocol. The mean number of nodes cleared at this operation was 11.5 (range 6–22). In two patients the sentinel node(s) were the only involved nodes while the other five women (included two of whom had only micrometastases evident in their sentinel node) had non-sentinel nodal metastases evident on full examination of the residual lymphatic basin.

The sensitivity and specificity rates as well as positive and negative predictive values both overall and by final pathological T stage are shown in Table 1. The accuracy of frozen section analysis of sentinel nodes according to tumour size was 97% for T1 tumours, 91% for T2 tumours and 86% for those with T3 disease. For those with Grade III cancers, it was 91%. The characteristics of those patients with false negative frozen sections in contrast to those with true negative and true positive frozen section results are shown in Table 2. As is evident from the table, patients with false negative sentinel node analysis by frozen section tended to have less sentinel nodes resected. They were also more likely to have larger tumours that were ER negative and were less likely to have Grade I tumours. On formal statistical analysis, however, no significant difference was demonstrable between the groups.

While there remains some dispute on the role of routine axillary clearance in selected patients with lymphatic metastases, most authorities recommend lymph basin dissection for node positive women in order to both fully debulk and stage the disease. Therefore, women found to have axillary disease after conventional pathological examination of their SLN require a second, separate operation to clear their (now scarred) axilla. A high likelihood of nodal disease undermines confidence in the potential advantages of SLN biopsy as any benefits to be derived from this technique are quickly outweighed by the physical risks and psychological detriment of a second reoperative procedure. Therefore many proponents of lymphatic mapping advocate that the technique be only implemented in patients with "early" breast cancer (that is, those with T1 cancers and a clinically negative axilla). Recent reports however have shown that the technique can function reliably in more advanced T-stages78 although clearly such patients have a higher risk of nodal metastases and therefore a greater risk of needing a second operation if convention histological examination of the nodes is employed9.

The ability to precisely measure tumours before their resection is however limited and clinical examination of the axilla is itself known to lack sufficient specificity and sensitivity to confidently guide appropriate patient selection for this technique10. How best preoperatively available measures of the primary tumor characteristics11 can be used to select patients for SLN prior to definitive surgery is also unclear1213. Furthermore, as selection criteria become increasingly stringent, a diminishing proportion of patients who could benefit from sentinel node biopsy are selected for the technique5. Additionally, it is uncertain whether recommendations based on T-stage alone can even be applied to symptomatic patients.

An accurate means of intra-operative analysis of sentinel nodes has the potential to allow the completion of surgical treatment for a patient with breast cancer in a single session. While touch imprint cytology has been suggested as being a useful means of such analysis1415, currently, it is limited by its availability as well as variable sensitivity rates and, perhaps, most worryingly specificity rates that are less than 100% (therefore still risking a chance of condemning women without nodal disease to axillary clearance). Data on frozen section assessment however is emerging showing sufficient1617, and perhaps superior18, intraoperative evaluation of nodal tumour burden, despite initial concerns that it consumes more of the specimen than cytological assessment. Analysis of published data to date shows that the accuracy of frozen section analysis with a combination of H&E staining and immunohistochemistry on sentinel lymph nodes lie between 73 to 96% 192021222324252627282930.

Overall the use of intraoperative frozen section analysis of sentinel node allowed 87/94 patients to have the axillary component of their surgery completed in a single step. Patients could also be informed of their status with this degree of confidence on awaking from anaesthesia. As has been suggested previously24, the facility for intraoperative analysis of sentinel node status was particularly advantageous for those with larger tumors. Furthermore, the accurate, intraoperative selection of 23 of the 30 patients with axillary metastases allowed definitive axillary operation to be concluded at the index operation with confidence of therapeutic benefit. A second operation rate of 7% overall (23% in node positive patients) seems justifiable in the light of being able to maximize the numbers of node-negative women who can and should benefit from SLN. Rather than 45 T1 patients undergoing this procedure with five requiring a second operation for unexpected lymphatic disease, incorporation of frozen section analysis allowed all seventy one node-negative women in this series the benefits of minimally invasive determination of lymph node status at the expense of seven second procedures. The major undermining factor for frozen section analysis was in the detection of micrometastatic disease. While much attention is currently being focussed on micrometastatic detection in sentinel nodes, the clinical and prognostic significance of the finding of just a few malignant cells has not yet been fully elucidated.