Intraoperative Sentinel Lymph Node Imprint Cytology Diagnosis in Breast Cancer Patients by General Surgical Pathologists: A Single-Institution Experience of 4327 Cases.

Context: Sentinel lymph node (SLN) biopsy is the standard of care for breast cancer (BC) patient staging. The axillary node status can be evaluated by frozen section (FS) or imprint cytology (IC). Despite the major disadvantages of tissue loss and freezing artifact in FS, many pathologists consider SLN IC a more difficult task requiring special expertise and prefer FS to IC. Aims: To conduct a large cohort study to evaluate the results of intraoperative SLN IC, compare them with those of FS and permanent section (PS), and determine the best method for general pathologists. Settings and Design: A very large cohort study with many pathologists engaged was designed and the diagnostic results were compared. Materials and
Methods: With 12 pathologists engaged, we conducted the largest patient cohort study of SLN IC to date, including 4,327 consecutive BC patients undergoing SLN biopsy during the past 15 years. The touch imprints were stained using three different methods and evaluated. The PS was used as the gold standard for evaluation. Statistical Analysis: The false positivity, false negativity, accuracy, sensitivity, and specificity of the IC and FS for each pathologist were evaluated and compared.
Results: Our results showed an overall sensitivity of 82.7%, specificity of 99.3%, and an accuracy rate of 95.9% for SLN IC, which were comparable to those of FS. The accuracy rate did not correlate with the length of working experience and the staining method. Conclusions: Intraoperative hematoxylin and eosin (H&E)-stained IC preparation was found to be the best SLN examination method for general pathologists. Copyright:

INTRODUCTION

Sentinel lymph node (SLN) evaluation is now the preferred procedure for staging in breast cancer (BC) patients requiring a mastectomy.[1] SLN imprint cytology (IC) and frozen section (FS) examination are the two main ways to evaluate the SLN status during surgery. Most pathologists are more familiar with FS examination and consider SLN IC a more difficult task requiring special expertise. The primary goals of this largest cohort study were to evaluate the results of the intraoperative SLN IC examination in our institution, compare the results with those of FS, and determine the feasibility of promoting the method to general pathologists.

SUBJECTS AND METHODS

A series of 4,327 consecutive BC patients who underwent SLN biopsy for intraoperative IC (with or without concurrent FS) examination at our hospital between April 2005 and April 2020 were retrospectively enrolled in the study. The SLNs were identified by the surgeons using the standard methylene blue dye injection protocol.[2] All SLNs submitted for evaluation were carefully collected and each node was sectioned into 1- to 2-mm-thick slices parallel to the long axis. A set of three imprints were obtained by gently touching the cut surfaces of all slices of SLNs to the glass slides. One imprint from each set was immediately fixed in 95% ethanol for 3 min and then stained with hematoxylin and eosin (H&E). The second imprint from each set was air-dried and stained with Liu's (or Riu's) technique.[3] The third imprint from each set was fixed in 95% ethanol for 30 min and then stained with the Papanicolaou method.[4] The first two imprints were examined intraoperatively and the third was evaluated postoperatively for comparison [Figure 1]. Concurrent FS of the most suspicious or largest SLNs was performed for every patient during the first 8 years. After that, SLN FS was done only if necessary based on the judgment of on-duty pathologists. All slices of SLNs were submitted for permanent tissue sections after surgery and, if the postoperative H and E sections were negative for metastasis, an additional immunostain for cytokeratin AE1/AE3 was performed to detect occult tumor foci or isolated tumor cells (ITC) in the most suspicious or largest SLNs on permanent section. For each patient, one of the following intraoperative IC diagnostic options was given: negative (no malignant tumor cells found), atypia (favor negative), atypia (favor positive), suspicious for malignancy, and positive for metastasis. For the purposes of surgical treatment and this study, the diagnosis “atypia (favor negative)” was regarded as negative, whereas “atypia (favor positive)” and “suspicious for malignancy” were regarded as positive. A “positive” finding in intraoperative SLN assessment usually resulted in an immediate ALN dissection, whereas “atypia (favor positive)” and “suspicious for malignancy” were followed by a separate subsequent ALN dissection only when the final permanent section pathology report of the SLN was “positive for metastasis.”

Figure 1

Slides with three different staining methods (from left to right): H and E, Liu's (or Riu's), and Papanicolaou

The H and E permanent section and immunostaining results (permanent tissue pathology) were used as the gold standard to evaluate the overall false positivity, false negativity, accuracy, sensitivity, and specificity of the IC. For purposes of nodal classification, SLNs containing only ITC were not regarded as positive. However, when the IC diagnosis was positive and only ITC were found postoperatively, the IC result was considered correct. False positivity was defined as the percentage of positive IC results among patients with negative permanent tissue pathology. False negativity was defined as the percentage of negative IC results among patients with positive permanent tissue pathology. Accuracy was defined as the percentage of true positive and true negative IC results among all patients examined. Sensitivity was defined as the percentage of positive IC results among patients with positive permanent tissue pathology. Specificity was defined as the percentage of negative IC results among patients with negative permanent tissue pathology.

RESULTS

During the 15 years of case collection, 12 board-certified surgical pathologists (7 males and 5 females, 30–65 years of age) were engaged in the intraoperative diagnosis of SLN IC. The working experience of pathology consultants as a sign-out pathologist ranged from 0 to 33 years. The three pathologists with working experience of more than 20 years were put in the “more-experienced” group. All other nine pathology consultants had working experience of fewer than 10 years and were labeled “less-experienced.” Only one of the pathology consultants was subspecialized in breast pathology. All other pathology consultants were either subspecialists in other pathological fields or general surgical pathologists without any subspecialty training. SLN mapping and biopsy were performed in 4,327 BC patients and 13,676 lymph nodes were identified (average 3.2 nodes/patient). For each of the collected SLNs, a set of three imprints was made. In all, 1,288 patients had concurrent FS examination of the most suspicious or largest SLNs, and an additional postoperative cytokeratin AE1/AE3 immunostain was performed in the most suspicious or largest SLNs of 3,541 cases. The number of SLN IC cases in the first year (from April 2005 to April 2006) was 29. The number of cases increased steadily in the following 10 years [Table 1] until it reached a plateau with a peak of 512 cases from 2017 to 2018 [Figure 2]. The peak case number was 17.7 times that of the first year.

Table 1

Number of SLN IC cases and times of increase vs. service year

Service year	Case number	Time of increase
1st (2005-2006)	29	1.0
2nd (2006-2007)	59	2.0
3rd (2007-2008)	81	2.8
4th (2008-2009)	126	4.3
5th (2009-2010)	163	5.6
6th (2010-2011)	196	6.8
7th (2011-2012)	217	7.5
8th (2012-2013)	254	8.8
9th (2013-2014)	371	12.8
10th (2014-2015)	390	13.4
11th (2015-2016)	469	16.2
12th (2016-2017)	507	17.5
13th (2017-2018)	512	17.7
14th (2018-2019)	490	16.9
15th (2019-2020)	460	15.9

Figure 2

The number of SLN IC cases vs. service year

In our patients, positive SLNs were detected in 764 cases (17.7%) by intraoperative IC evaluation. Among these, 740 cases were true positive and 24 were false positive. Among the 3,563 IC-negative cases, 3,408 were true negative and 155 were false negative. The false-positive rates of our pathologists ranged from 0.0% to 2.5% (pooled 0.6%) and false-negative rates ranged from 1.6% to 5.7% (pooled 3.6%). Our pathology consultants’ IC examination had a sensitivity of 66.7 to 85.7% (pooled 82.7%), and a specificity of 96.9 to 100% (pooled 99.3%). The accuracy rates ranged from 92.9% to 97.6% (pooled 95.9%) [Table 2]. More than half (57.8% or 2,503 cases) of our intraoperative IC cases were evaluated by a single “key” pathologist (i.e., the breast subspecialist), and about one-third (14.4% or 621 cases) of the remainder (42.2% or 1,824 cases) were done by the “No. 2” pathologist. The IC workloads of the rest of our pathology consultants ranged from 0.4% (18 cases) to 5.2% (226 cases) [Table 2]. The sensitivity, specificity, and accuracy rate of the “key” pathologist were 85.7%, 99.9%, and 96.8% respectively; those of the “No. 2” pathologist were 74.6%, 97.0%, and 92.9%, respectively; and, for the rest of our pathology consultants, were 79.3%, 99.2%, and 95.4%, respectively. The “key” member had 24 years working experience as a sign-out pathologist; the “No. 2” member had 26 years of experience, and the experience of other pathology consultants ranged from 0 to 33 years. The “key” pathologist (the breast pathology subspecialist) had better sensitivity, specificity, and accuracy rate than other pathology consultants; however, the results of other pathologists were quite acceptable.

Table 2

Results of intraoperative imprint cytology examination of 4,327 cases by 12 pathologists (no. 1 to no. 12).

Pathologists	No. 1	No. 2	No. 3	No. 4	No. 5	No. 6	No. 7	No. 8	No. 9	No. 10	No. 11	No. 12
Cases	2,503	63	621	130	81	205	212	124	119	226	25	18
TP	475 (19.0%)	9 (14.3%)	85 (13.7%)	23 (17.7%)	9 (11.1%)	28 (13.7%)	31 (14.6%)	11 (8.9%)	18 (15.1%)	43 (19.0%)	6 (24.0%)	2 (11.1%)
TN	1948 (77.8%)	51 (81.0%)	492 (79.2%)	100 (76.9%)	69 (85.2%)	169 (82.4%)	168 (79.2%)	110 (88.7%)	94 (79.0%)	174 (77.0%)	18 (72.0%)	15 (83.3%)
FP	1 (0.0%)	0 (0.0%)	15 (2.4%)	2 (1.5%)	0 (0.0%)	0 (0.0%)	1 (0.5%)	1 (0.8%)	3 (2.5%)	1 (0.4%)	0 (0.0%)	0 (0.0%)
FN	79 (3.2%)	3 (4.8%)	29 (4.7%)	5 (3.8%)	3 (3.7%)	8 (3.9%)	12 (5.7%)	2 (1.6%)	4 (3.4%)	8 (3.5%)	1 (4.0%)	1 (5.6%)
Sensitivity	85.7%	75.0%	74.6%	82.1%	75.0%	77.8%	72.1%	84.6%	81.8%	84.3%	85.7%	66.7%
Specificity	99.9%	100.0%	97.0%	98.0%	100.0%	100.0%	99.4%	99.1%	96.9%	99.4%	100.0%	100.0%
Accuracy	96.8%	95.2%	92.9%	94.6%	96.3%	96.1%	93.9%	97.6%	94.1%	96.0%	96.0%	94.4%

TP: true positive; TN: true negative; FP: false positive; FN: false negative

For all our patients, the results of IC evaluation by H and E stain alone, H and E plus Liu's stains, and H and E plus Liu's plus Papanicolaou stains were virtually identical [Table 3]. For cases with concurrent FS examination, the IC and FS results were rather similar [Table 4]. The turn-around time of the SLN IC examination was less than 20 min in more than 90% of cases when the number of SLN submitted was 2 or less [Table 5].

Table 3

IC evaluation results by different staining methods

Diagnosis	H&E	H&E + Liu’s	H&E + Liu’s + Pap
Negative	3,540	3,534	3,538
Atypia (favor negative)	25	29	26
Atypia (favor positive)	10	9	10
Suspicious	11	13	12
Positive	741	742	741

Table 4

Results of concurrent IC and FS examination

Diagnosis	IC	FS
Negative	1,050	1,073
Atypia (favor negative)	9
Atypia (favor positive)	5
Suspicious	6
Positive	218	215

Table 5

Number of SLN IC examined vs. number (and percentage) of cases with turn-around time (TAT) < 20 min

SLN number	Cases	Cases with TAT <20 min
1	737	737 (100.0%)
2	1,078	1,016 (94.2%)
3	1,048	924 (88.2%)
4	641	510 (79.6%)
5 or more	820	335 (40.9%)

DISCUSSION

An accurate intraoperative SLN examination can greatly reduce the reoperation rate in BC patients requiring an axillary lymph node (ALN) dissection. In the past, patients with negative SLNs would not need a further operation, whereas those with a positive SLN should undergo ALN dissection. However, based on the results of the American College of Surgeons Oncology Group (ACOSOG) Z0011 trial,[5] the 2014 American Society of Clinical Oncology (ASCO) guideline update advised obviation of ALN dissection in patients with one to two metastatic SLNs undergoing whole-breast radiotherapy.[6] The ACOSOG Z0011 trial also suggested that SLNs harboring micrometastases or isolated tumor cells had little or no prognostic significance in current clinical practice.[5]

There are two main ways to evaluate the SLN status during surgery: frozen section (FS) histology and touch imprint cytology (IC). Most pathologists are more familiar with intraoperative FS examination; however, it is more time-consuming and expensive than IC and has the major disadvantages of nodal tissue loss and freezing artifact.[7] Many pathologists consider cytological SLN appraisal a more difficult task requiring special expertise and find it difficult to confirm cytological findings in permanent histological sections.[7] Reports of FS examination and meta-analysis of IC of SLNs in BC patients showed that FS had better sensitivity than IC in most direct comparisons but the specificities of the two methods were comparable.[8] The FS examination had a sensitivity of 44 to 100% (pooled 76%) and a specificity of 99 to 100% (pooled 99%), whereas IC had a sensitivity of 34 to 96% (pooled 63%) and a specificity of 94 to 100% (pooled 99%).[8]

To our knowledge, our study is the largest patient cohort study of SLN IC to date, including 4,327 consecutive BC patients undergoing SLN biopsy during the past 15 years, and 12 board-certified surgical pathologists were engaged in the study. The aims of the study were to evaluate the sensitivity, specificity, and accuracy rate of intraoperative SLN IC in a large consecutive cohort of BC patients, compare the results of IC and FS examination, determine the relationship between pathological experience and IC interpretation, and finally, determine the feasibility of promoting the best and easiest IC method to pathologists without subspecialty training in breast pathology and cytopathology. This study showed a very low false-positive rate (0.0–0.8%) in 9 (2 more-experienced and 7 less-experienced) of our pathology consultants and low false-positive rate (1.5–2.5%) in the remaining 3 (1 more-experienced and 2 less-experienced) pathologists, giving an overall specificity of 99.3%. The possible reasons for false-positive misinterpretation of the test by the pathologists included histiocytes, endothelial cells, large lymphoid cells, dendritic cells, fibroblasts, plasma cells with Russell bodies, necrotic or ischemic cells, other inflammatory cells, cells with benign inclusions, thick cytologic preparation or cell overlapping, and contamination of tumor cells. The first three of the above (histiocytes, endothelial cells, and large lymphoid cells) [Figure 3a-c] were by far the most common sources of false-positive misinterpretation. Aggregates of a large number of histiocytes were easily misinterpreted as cancer cells. Active-looking endothelial cells were also commonly mistaken for carcinoma. Large lymphoid cells, especially when their nuclei appeared atypical, were a common source of misinterpretation. More than a dozen of such false-positive errors in our cases could be avoided if these cells with atypical or worrisome features were correctly interpreted. It is therefore very likely that the group of pathologists with a 1.5 to 2.5% false-positive rate can achieve as low as 0.0 to 0.8% false-positive rate if they correct themselves through the regular review of such cases to look for the possible cause of misinterpretation, or take a second opinion, even with the slightest doubt.

Figure 3

Most common causes of cell misinterpretation in SLN imprint cytology: (a) histiocytes, (b) endothelial cells, (c) large lymphoid cells, (d) low-grade carcinoma, (e) lobular carcinoma, (f) micrometastasis

The sensitivity for SLN IC was previously reported to vary widely from 34% to 96%.[5] The SLN IC interpretation by our pathology consultants had a sensitivity of 66.7 to 85.7% (pooled: 82.7%). This study showed a low false-negative rate (1.6–5.7%, pooled 3.6%) in our pathology consultants. The possible causes of false-negative misinterpretation by the pathologists included tumor cells missed, low suspicion index of malignancy, low self-confidence of interpretation, tumor cells not in “examined” field, presence of the only individual, isolated, or tiny clusters of tumor cells (i.e., micrometastases or isolated tumor cells), low-grade carcinoma, lobular carcinoma, spindled tumor cells, tissue necrosis, thick cytologic preparation or cell overlapping, and tumor cells obscured by blood. Low-grade carcinoma, lobular carcinoma, and carcinoma with only micrometastases or isolated tumor cells [Figure 3d-3f] were the most frequent causes of missed diagnoses in our study. The former two were also reported to be more difficult to identify in IC preparation when compared to ductal carcinoma[9] but not by others.[1011] Metastatic lobular carcinoma and low-grade ductal carcinoma, especially when tumor cells were dispersed in the lymphoid tissue, were quite easily missed because of their small bland-looking nuclei. False-negative errors due to micrometastases or isolated tumor cells were usually hard to avoid because there was often no tumor cell on the imprint slides or only very tiny clusters of cancer cells.

The sensitivity, specificity, and accuracy rate of SLN IC interpretation among our pathology consultants varied, but each varied within an acceptable range. Our pathology consultants had a sensitivity of 66.7 to 85.7%, a specificity of 96.9 to 100%, and an accuracy rate of 92.9 to 97.6%. The length of working experience in pathology did not correlate with the accuracy rate of the IC interpretation. The pathologists with working experience of more than 20 years had an accuracy rate of 92.9 to 96.8% (3,268 cases in total), whereas those with working experience of fewer than 10 years had an accuracy rate of 93.9 to 97.6% (810 cases in total). The IC workload was also not related to the accuracy rate of the interpretation. The “key” pathologist having the workload of 57.8% (2,503 cases) of all SLN IC cases had an accurate interpretation rate of 96.8%. The “No. 2” pathologist had a workload of 14.4% (621 cases) of all IC cases and an accuracy rate of 92.9%. None of the other pathology consultants had a workload of more than 230 cases, and these pathologists had a pooled accuracy rate of 95.4% (range: 93.9%–97.6%). The fact that there was no positive correlation between the accuracy rate of SLN IC interpretation and the length of working experience in pathology highly suggested that intraoperative SLN IC was a low-threshold examination and even an inexperienced general pathologist without subspecialty training in breast pathology and cytopathology could do the job well. The phenomenon of a decrease in accurate interpretation rate from low workload to intermediate workload (decrease from 95.4% to 92.9%), and an increase in the accuracy rate from intermediate workload to high workload (an increase from 92.9% to 96.8%) could be partly explained by (when the workload was intermediate) the increased chance of encountering conditions that hindered a correct diagnosis, and (when the workload was high) increased knowledge and experience of dealing with such difficult differential diagnosis. Other important factors that might greatly influence the accuracy rate included awareness of the cytologic features of cancer, carefulness in finding abnormalities and interpreting them, and strict demand for high-quality IC preparation.

Three different staining methods were used in our study: H and E (wet fixed), Liu's (air-dried), and Papanicolaou (wet-fixed) stains. The former two methods were performed during surgery for immediate diagnosis, and the last method was done postoperatively for comparison. With regard to sensitivity, specificity, and accuracy rate of SLN IC interpretation, our results showed virtually no difference in these three staining methods. Every pathologist had their preferences and none of the methods was found to be much superior to the other two. However, more pathologists preferred H and E and Papanicolaou stains over Liu's stain as the former two provided better nuclear features and fewer air-drying artifacts. Because H and E stain is currently the most widely used stain in medical diagnosis and all pathologists are most familiar with it, this method should be considered as an easy, rapid, and reliable diagnostic tool that can be applied on a regular basis in intraoperative SLN IC samples for immediate diagnosis. Although Papanicolaou stain is a widely used stain in cytology, it is not suitable for rapid intraoperative SLN IC diagnosis because of its longer fixation time.[4] Ultrafast Papanicolaou stain[12] or modified ultrafast Papanicolaou stain[13] may be used as an alternative preparation to minimize the turn-around time.

When compared with the FS examination, our SLN IC interpretation had equivalent sensitivity, specificity, and accuracy rate. In the study by Petropoulou et al.,[7] IC appeared to be marginally more sensitive than FS in detecting SLN metastases; however, only a very small number (60) of patients with invasive BC were registered in that study. For every single case, the cost of SLN IC at our institution was NT$800 (about US$28) versus NT$6740 (about US$236) for frozen sectioning. In terms of turn-around time, the procedures of SLN IC could be completed within approximately the same time as FS. To be specific, when there were only one or two SLNs submitted, the IC diagnosis could be returned to the surgeon within 20 min in more than 90% of the cases.

Our case number per year increased in a fashion similar to that of cell growth. The case number increased steadily in the first 10 years and then reached a plateau with a peak. The highest case number from 2017 to 2018 was 17.7 times that of the first year (2005–2006). The continuous and substantial increase in our number of SLN IC cases was strong evidence of a high degree of satisfaction among surgeons. In the study, concurrent FS of SLNs were performed only for 8 years. In the following and last 7 years, virtually only routine SLN IC examinations were done during surgery. Intraoperative FS of the SLNs was exercised very occasionally and, as it were, had been replaced by IC.

In summary, intraoperative SLN diagnosis by imprint cytology is, in many aspects, comparable to that obtained from FS examination. SLN IC is a simple and easily performed diagnostic procedure that can be done in every pathology laboratory. The diagnosis of SLN IC can be given to the surgeon in the same way as FS and the turn-around time is similar; however, IC examination is much less expensive than FS and has the advantages of no nodal tissue loss and freezing artifact. Although a set of three imprints were obtained from every SLN received in our study, our results highly suggested that a single IC with H and E stain would suffice. Our analysis also concludes that intraoperative SLN IC is a low-threshold examination and even an inexperienced board-certified general surgical pathologist without subspecialty training in breast pathology and cytopathology can do the job quite well. The most important factors influencing the accuracy rate in intraoperative SLN IC are a high-quality IC slide preparation, awareness of the cytologic diagnosis pitfalls, and carefulness in finding cytologic abnormalities and interpreting them.

Key Messages

Imprint cytology preparation with H and E stain alone is the best (simple, fast, easily performed, inexpensive, and highly accurate) way of intraoperative sentinel lymph node diagnosis in BC patients even for the inexperienced general surgical pathologists without subspecialty training in breast pathology and cytopathology.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.