Intraoperative Lymph Node Assessment (Touch Preparation Only) for Metastatic Breast Carcinoma in Neoadjuvant and Non-neoadjuvant Settings Open Access

This study was approved by the institutional review board of University of Massachusetts Chan Medical School-Baystate Medical Center, Springfield, Massachusetts (institutional review board No. 1584367). A search of the pathology CoPath database was performed from 2016 to 2019 for all surgical pathology reports of patients who had intraoperative sentinel/nonsentinel lymph node assessment for MBC. Patients with any subtype of invasive breast carcinoma and a small subset with ductal carcinoma in situ (DCIS) and DCIS with microinvasion treated with primary mastectomy were included. Clinicopathologic parameters including patient age, type of surgical procedure, tumor size, size of nodal metastasis (isolated tumor cells [≤0.2 mm], micrometastasis [>0.2 mm to ≤2 mm], or macrometastasis [>2 mm]), histologic tumor type, histologic grade, estrogen receptor (ER) status, human epidermal growth factor receptor 2 (HER2) status (if applicable), tumor focality, lymphovascular invasion (LVI), pathologic tumor stage, pathologic nodal stage, and the result of prior axillary sampling if present (fine-needle aspiration or core biopsy) were obtained through chart review and the CoPath database. Radiologic tumor size was obtained from radiologic reports of the patients who had mass-forming lesion(s) on ultrasonography and/or magnetic resonance imaging. Follow-up data for axillary management of patients with false-negative intraoperative TP diagnosis was obtained through chart review and the CoPath database.

Our standard operating procedure for intraoperative lymph node assessment to detect MBC was used in all cases included in this study. Briefly, lymph nodes were serially sectioned into 2- to 3-mm-thick cross sections; TP(s) were performed by imprinting the whole cut surface on 1 side of all cross sections, with at least 2 TPs performed on most of the lymph nodes; and tissue scraping of grossly suspicious areas was performed in some cases. TP slides were stained with routine hematoxylin-eosin stain. The TP(s) were interpreted by a primary pathologist and a second pathologist, at least one of whom was required to be a cytopathologist. Although the frozen section was not part of the standard operating procedure in our institution, some pathologists preferred to perform frozen sections in addition to TPs. The cases in which the primary pathologist opted to perform an intraoperative frozen section in addition to TP (6 lymph nodes from 4 patients in non-neoadjuvant setting [NNAS] and 4 lymph nodes from 3 patients in NAS) were excluded from this study. For the final diagnosis, hematoxylin-eosin–stained 3-step levels of the entire lymph node were assessed. Of note, keratin immunohistochemistry was not part of the routine practice for sentinel lymph node evaluation in patients with breast cancer; however, it would be requested if needed by the pathologist for the final diagnosis. The results of intraoperative and final diagnoses were obtained from the final pathology reports. The lymph nodes with intraoperative TP diagnosis of “atypical” (8 of 207 lymph nodes) were considered to be negative for statistical calculations owing to the indistinct nature of this category for clinical interpretation of MBC. To reveal the underlying reasons of discrepancy, intraoperative TP slides of the discordant cases (intraoperative versus final diagnosis) were reviewed retrospectively by 1 surgical pathologist (G.M.C.), 1 cytopathologist (Q.J.C.), and 2 pathology residents (E.E., M.E.). Discrepancy reasons were categorized under undersampling due to small tumor deposit (≤2 mm), poor TP quality secondary to neoadjuvant therapy–related histomorphologic changes, or interpretation challenge (eg, lobular histotype). Among the discordant cases, the TP slides of 1 non-neoadjuvant case were not available for retrospective review.

Accuracy statistics (sensitivity, specificity, and negative predictive value with 95% CIs) were calculated for NAS and NNAS samples, comparing intraoperative TP diagnosis to final diagnosis. Because a given patient may have multiple lymph nodes evaluated intraoperatively, the statistics were calculated in 2 different manners, on the level of individual lymph node and on the patient level by accounting for all TP results of the lymph nodes obtained from that particular patient. As an exploratory analysis, we described factors associated with either (1) lymph node metastasis on final diagnosis or (2) discordant results (intraoperative TP versus final diagnosis). For these analyses, we present descriptive statistics (frequencies and percentages) along with Fisher exact test, odds ratios (ORs), and 95% CIs (calculated by using unadjusted logistic regression) to describe possible associations. Categories of variables with zero cells were excluded from OR calculations. Cases with DCIS and DCIS with microinvasion were excluded from the exploratory analysis in NNAS. Since our sample was small and underpowered, we used P values < .20 as suggestive of possible associations. In addition, ORs and CI width were used to assist in the interpretation of possible associations and a range of estimates compatible with our data.

In this retrospective study, we identified 207 sentinel/nonsentinel lymph nodes from 103 patients, which were evaluated via intraoperative TP for lymph node metastasis of breast carcinoma. Table 1 summarizes patient characteristics of this study.

A total of 108 lymph nodes from 56 patients, ages ranging from 29 to 89 years (mean age, 59.2 years), were identified. The lymph nodes were part of the operative procedures including mastectomies, breast-conserving surgeries (eg, lumpectomy, seed- or wire-localized excisional biopsies), and sentinel lymph node excision only following upstaging to invasive carcinoma at the time of primary surgery (Table 1). The final pathologic diagnoses were DCIS, DCIS with microinvasion, invasive ductal carcinoma, invasive lobular carcinoma, mixed ductal and lobular carcinoma, and other (eg, mucinous carcinoma) (Table 1). More than three-quarters (84 of 108, 77.8%) of the lymph nodes were evaluated by at least 2 TPs per lymph node. Three of 56 patients had axillary lymph node sampling before surgery, and all results were negative.

On the level of individual lymph node evaluation, the sensitivity, specificity, and negative predictive value of TP for MBC including isolated tumor cells were 37.5%, 100%, and 90.2%, respectively (Table 2). Among 10 false-negative results, macrometastasis, micrometastasis, and isolated tumor cells were identified in 10%, 70%, and 20%, respectively.

On the patient level, the sensitivity, specificity, and negative predictive value of TP for MBC including isolated tumor cells were 36.3%, 100%, and 86.5%, respectively. Among 7 patients with false-negative results, macrometastasis, micrometastasis, and isolated tumor cells were identified in 14%, 58%, and 28%, respectively.

Among the factors evaluated, discrepancy between intraoperative and final diagnosis was mostly due to undersampling secondary to small metastatic tumor deposits (≤2 mm) (6 of 7, 86%). The only patient with macrometastasis and false-negative intraoperative TP result was diagnosed with invasive ductal carcinoma, grade 1; unfortunately, TP slides of this patient were not available for retrospective review.

On univariable analysis, presence of LVI was found to be associated with lymph node metastasis on final diagnosis (OR, 8.85; 95% CI, 2.56–30.58). Pathologic infiltrative tumor size was found to be associated with lymph node metastasis on final diagnosis (eg, 11–20 mm compared to >20 mm; OR, 0.59; 95% CI, 0.19–1.86). None of the patients with pathologic tumor size of 10 mm or smaller had lymph node metastasis, whereas one of the patients with radiologic tumor size of 10 mm or smaller had macrometastasis on the final diagnosis. Radiologically, this patient had developing asymmetry around the mass-forming lesion and was diagnosed with mixed ductal and lobular carcinoma. Higher histologic grade was found to have possible association with lymph node metastasis (eg, grade 1 compared to grade 3; OR, 0.46; 95% CI, 0.10–2.10). No meaningful associations were observed for the other parameters including histologic type, ER and HER2 status, and focality in NNAS (Table 3).

LVI was found more in discordant cases than in concordant ones (OR, 5.60; 95% CI, 1.34–23.42). None of the patients with tumor smaller than 10 mm (pathologic or radiologic measurement) had discrepancy between the intraoperative and final diagnosis. For the other parameters evaluated, no meaningful differences were observed between the discordant and concordant cases in NNAS (Table 4).

On follow-up, 2 of 7 non-neoadjuvant patients with false-negative TP diagnosis underwent second surgery for completion ALND, including one with macrometastasis and the other with micrometastasis.

A total of 99 lymph nodes from 47 patients, ages ranging from 28 to 79 years (mean age, 53.6 years), were identified. Forty-four of 47 patients were treated with neoadjuvant chemotherapy and 3 of 47 patients were treated with neoadjuvant aromatase inhibitors (eg, anastrozole). The lymph nodes were part of the operative procedures including mastectomies and breast-conserving surgeries (eg, lumpectomy, radioactive seed– or wire-localized excisional biopsies) (Table 1). The final pathologic diagnoses were invasive ductal carcinoma, invasive lobular carcinoma, mixed ductal and lobular carcinoma, and other (eg, mucinous carcinoma) (Table 1). Almost three-quarters of the lymph nodes (74 of 99, 74.7%) were evaluated by using at least 2 TPs per lymph node. Twenty-seven of 47 patients had axillary lymph node sampling before the neoadjuvant therapy and of these, 26 had biopsy-proven axillary lymph node metastasis.

On the level of individual lymph node evaluation, the sensitivity, specificity, and negative predictive value of TP for MBC including isolated tumor cells were 34.2%, 100%, and 70.9%, respectively (Table 5). Among 25 false-negative results, macrometastasis, micrometastasis, and isolated tumor cells were identified in 56%, 32%, and 12%, respectively. Of note, the final tumor deposit measurement was based on the largest contiguous focus of viable residual tumor cells, not including fibrosis of treatment effect.

On the patient level, the sensitivity, specificity, and negative predictive value of TP including isolated tumor cells were 47.6%, 100%, and 70.2%, respectively. Among 11 patients with false-negative results, the percentage of macrometastasis, micrometastasis, and isolated tumor cells were identified in 54.5%, 36.3%, and 9%, respectively.

Although there was overlap for the reasons of discrepancy in NAS, the main reasons were interpretation challenge due to lobular histotype (2 of 11, 18.2%), poor quality of TP due to therapy-induced histomorphologic changes (4 of 11, 36.3%), and undersampling secondary to small tumor deposits (≤2 mm) (5 of 11, 45.5%). The patients with macrometastasis and false-negative intraoperative TP result had lobular histotype (Figure 1, A through D) or therapy-induced histomorphologic changes affecting the overall quality of TP (Figure 2, A and B).

Lymph node metastasis was found to be associated with the following tumor parameters in NAS: presence of LVI (OR, 4.90; 95% CI, 1.68–14.28), ER positivity (OR, 11.70; 95% CI, 3.90–34.32), HER2 negativity (OR, 22.39; 95% CI, 2.88–174.42), histologic grade 2 (compared to grade 3: OR, 7.45; 95% CI, 2.81–19.75), invasive lobular carcinoma (compared to invasive ductal carcinoma: OR, 18.00; 95% CI, 2.16–150.12), and pathologic tumor size greater than 10 mm (Table 6).

Discordant cases were more likely to be found with the following tumor parameters: invasive lobular carcinoma (compared to invasive ductal carcinoma: OR, 16.00; 95% CI, 3.07–83.27), histologic grade 2 (compared to grade 3: OR, 9.70; 95% CI, 2.66–35.35), ER positivity (OR, 15.09; 95% CI, 3.31–68.76), HER2 negativity (OR, 10.82; 95% CI, 1.38–84.93), multifocality (OR, 4.29; 95% CI, 1.46–12.61), and pathologic tumor size greater than 10 mm on final diagnosis (Table 7). Twenty-one of 25 lymph nodes (84%) with discordant results on the individual lymph node level were from the patients who had biopsy-proven lymph node metastasis before neoadjuvant therapy. For the other parameters evaluated, no meaningful differences were observed between the discordant and concordant cases in NAS.

On follow-up, 8 of 11 neoadjuvant patients with false-negative TP diagnosis underwent second surgery for completion of ALND, including 6 with macrometastasis and 2 with micrometastasis. One of 11 patients had an intraoperative TP diagnosis of “atypical,” and ALND was performed at the time of primary operation.

The sensitivity and specificity of intraoperative imprint cytology/TP varies among different studies in NAS, ranging from 38.2% to 100% for sensitivity and from 94.9% to 100% for specificity.^21–25  In some studies, imprint cytology/TP was found to have comparable sensitivity and specificity in NAS compared to NNAS^21,22  and was defined as a reliable method for intraoperative axillary lymph node assessment after neoadjuvant chemotherapy.²⁵  However, Elliott et al²³  reported 38.6% sensitivity for intraoperative TP in NAS and concluded not to use TP alone for intraoperative sentinel lymph node assessment owing to overall low sensitivity, especially for patients with invasive lobular carcinoma and those who had clinically positive axilla before initiation of neoadjuvant therapy. In our study, we found comparable sensitivity and specificity between neoadjuvant and non-neoadjuvant settings; however, more lymph nodes with macrometastasis were missed via intraoperative TP in NAS. In an early meta-analysis, sentinel node imprint cytology had 81% sensitivity for macrometastasis¹⁰ ; in this study, it was around 48% in NAS versus 83% in NNAS.

Lobular histotype has been previously described as one of the pitfalls on TPs owing to the dyshesive nature of the tumor cells,^23,26  which we observed in 18.2% of the lymph nodes in NAS as a discrepancy reason related to interpretation challenge. In NNAS, most of the false-negative results were related to undersampling secondary to the small size of the tumor deposits (≤2 mm), which has been well established before.^10,11,14  Among neoadjuvant patients with macrometastasis who had false-negative TP result, 5 of 6 had biopsy-proven axillary lymph node metastasis before the neoadjuvant therapy. Overall, our findings suggest that intraoperative TP is not reliable in NAS owing to poor sensitivity for macrometastasis, especially in patients with lobular histotype and biopsy-proven axillary lymph node metastasis before neoadjuvant therapy.

Lymphovascular involvement is known to predict lymph node metastasis for breast cancer.^27–29  In this study, lymphovascular involvement was found to have association with lymph node metastasis in both neoadjuvant and non-neoadjuvant settings (Tables 3 and 6). Tumor size has been described as another independent predictor of lymph node metastasis.^28,30  We noted that none of the non-neoadjuvant therapy patients with pathologic tumor size of 10 mm or smaller had lymph node metastasis. For NAS, tumors greater than 10 mm were more likely to have lymph node metastasis than tumors 10 mm or smaller (Table 6). In this study, other parameters associated with lymph node metastasis for NAS were ER positivity, HER2 negativity, and invasive lobular carcinoma (compared to invasive ductal carcinoma) (Table 6). Although lobular histotype has a mildly higher incidence of lymph node metastasis on univariate analysis, this can be secondary to confounding factors that interact with lymph node involvement on multivariate analysis as shown by Vandorpe et al.³¹  High-histologic-grade breast cancers (grade 3 versus grade 1 and 2) have better response to neoadjuvant chemotherapy and higher negative conversion of axillary lymph nodes.^32,33  In our study, histologic grade 2 was more likely to have lymph node metastasis than grade 3 in NAS on the final pathologic assessment. This can be explained by a better treatment response in histologic grade 3 carcinomas than in histologic grade 2. Hormone receptor–positive/HER2-negative breast cancers have higher odds of demonstrating LVI than triple-negative tumors.³⁴  In a study of intraoperative lymph node assessment via frozen section, discordant results were more likely to have ER-positive/HER2-negative status in NAS.²⁰  Even though our study included TP-only cases, we observed a similar finding that the neoadjuvant cases with discrepancy were more likely to have ER positivity and HER2 negativity. Interestingly, histologic subtype was not found to have any association with discrepancy in the aforementioned study, which would highlight the interpretation challenge due to the lobular histotype on TPs, especially in the setting of neoadjuvant therapy–induced histomorphologic changes (eg, histiocytic infiltration) (Figure 1, A through D).

In conclusion, intraoperative TP should be used very cautiously in NAS owing to a high false-negative rate for macrometastasis when compared to NNAS, especially for patients with invasive lobular carcinoma and known axillary lymph node metastasis before their neoadjuvant therapy. Small size of metastatic tumor deposits (≤2 mm) is one of the major discrepancy reasons, but may not necessarily change the axillary management of non-neoadjuvant patients if only limited lymph node involvement is present. However, low-volume sentinel lymph node disease remains an indication for ALND after neoadjuvant chemotherapy.³⁵