Triple-negative breast cancer (TNBC) is distinct from other breast cancers, because the tumor cells lack estrogen and progesterone receptors (hormone receptors) and also are negative for human epidermal growth factor receptor 2 (HER2). They comprise a heterogeneous group of tumors with various histologic features and clinical behaviors. High-grade, invasive ductal carcinoma not otherwise specified is the most frequent type, and a substantial fraction of TNBCs belongs to the basal-like tumor type. The purpose of this study was to determine whether some cytologic features could predict the triple phenotype of breast carcinoma.
Fine-needle aspiration cytology samples of 62 TNBCs were compared with samples of 82 hormone receptor-positive, high-grade, invasive carcinomas (HRBC) and with samples of 33 hormone receptor-negative, HER2 positive, invasive carcinomas (HER2BC) for the following cytomorphologic features: cellularity, necrosis, lymphocytes, syncytial clusters, tubular/ductal-like clusters, large bare nuclei, streaming within the clusters, and calcifications. Moreover, single cell features, such as cellular borders, cytoplasm, cytoplasmic vacuoles, nuclear pleomorphism, nucleoli, and type of chromatin pattern, were evaluated. Descriptive analyses and 2 multivariate regression models were performed to compare TNBC, HRBC, and HER2BC and to identify the cytologic factors that were associated with tumor type.
TNBCs were more likely to have an abundant necrotic background, many lymphocytes, many syncytial clusters, and ill defined cell borders than non-TNBCs. A tubular/ductal pattern was observed only rarely in TNBCs. Multivariate logistic analysis indicated a 90.8% probability of identifying TNBC versus HRBC by the following cytologic variables: lymphocytes, ill defined cell borders and syncytial clusters, tubular/ductal clusters, cytoplasmic vacuoles, and cellular pleomorphism; whereas there was a 77.5% probability of identifying TNBC rather than HER2BC by the following variables: cellularity, ill defined cellular borders and syncytial clusters, and tubular/ductal clusters.
Triple-negative breast cancer (TNBC) is distinct from other breast cancers, because the tumor cells in TNBC lack estrogen receptor (ER) and progesterone receptor (PR) (together, hormone receptors [HRs]) and also are negative for human epidermal growth factor receptor 2 (HER2). These features mean that hormone therapy is ineffective and that patients who have triple-negative disease are not candidates for anti-HER2 treatment with trastuzumab. Approximately 10% to 17% of all invasive breast cancer diagnoses are characterized as triple negative, and TNBC appears to be more common among young women, and particularly young African American women.1, 2
In the last decade, analyses of gene expression arrays have resulted in the recognition of 5 distinct molecular subtypes of breast cancer with clinical, biologic, and therapeutic implications. The luminal subtypes of breast cancers express ER, ER-responsive genes, and other genes that encode characteristic proteins of luminal epithelial cells. There appear to be at least 2 groups of ER-positive breast cancers, commonly defined as luminal A and luminal B tumors, depending on expression levels of the characteristic genes and of other genes that pertain either to the proliferation cluster and/or to HER-2. The second broad group (ER-negative tumors) is subdivided into 3 groups: HER2-positive tumors, basal-like tumors, and so-called normal breast-like tumors. HER2-positive tumors express high levels of genes located in the HER2 amplicon on region 21 of the long arm of chromosome 17 (17q21) and lack ER expression. Basal-like tumors lack ER and HER2 and express genes that are characteristic of basal epithelial cells. The normal breast-like group resembles normal breast tissues with relatively high expression of many genes that are characteristic of adipose cells and other nonepithelial cell types and with low-level expression of luminal epithelial cell genes. These breast carcinoma subtypes are associated with different clinical outcomes, from the relatively good prognosis for patients with luminal A tumors to the worst prognosis for those with basal-like and HER2 tumors.3, 4
Because it is not always feasible to obtain gene expression array information, a clinicopathologic classification recently was proposed that categorizes the definitions of intrinsic subtypes of breast cancer according to an immunohistochemical analysis of ER and PR status, overexpression and/or amplification of HER2, and the Ki-67 labeling index. The 5 subtypes—luminal A, 2 luminal B (HER2-negative and HER2-positive), HER2-positive (nonluminal), and triple-negative (ductal)—are similar, but not identical, to intrinsic breast cancer classification subtypes.5
TNBCs comprise a heterogeneous group of tumors with various histologic features and clinical behaviors. High-grade, invasive ductal carcinomas not otherwise specified are the most frequent. Other carcinomas that usually have a triple-negative phenotype include low-grade and high-grade metaplastic, medullary, apocrine, adenoid cystic, and juvenile secretory carcinomas.5, 6
Fine-needle aspiration cytology (FNAC) is widely used in the initial assessment of patients who have symptomatic or screen-detected breast lesions. This technique is safe, simple, inexpensive, and can accurately diagnose carcinoma, thus reducing the need for open surgical biopsy with its higher cost for the health care system and potentially greater morbidity. Definitive treatment can be performed on the basis of a positive FNAC diagnosis, including preoperative therapy that will enable subsequent conservative therapy.7-9 The objective of the current study was to determine whether some cytologic features could predict the triple-negative phenotype of breast carcinoma, using as gold standard a series of high-grade, infiltrating breast carcinomas classified according to the immunohistochemical classification as TNBC, HER2-positive breast cancer (HER2BC), and HR-positive breast cancer (HRBC).
MATERIALS AND METHODS
One hundred seventy-seven consecutive, histologically high-grade, invasive breast carcinomas with preoperative FNAC material available from patients who were diagnosed between January 2003 and December 2009 were identified from the database of the Department of Pathology of S. Chiara Hospital of Trento, Italy. For all patients, histologic diagnosis, ER and PR status, and HER2 expression were known.
Three or 4 conventionally hematoxylin and eosin-stained or Papanicolaou-stained cytologic smears for each case were available for review. A whole range of morphologic features was scored independently by 2 pathologists (L.M. and E.M.B.) who were blinded to the HR and HER2 status of the study group at time of analysis.
The following cytologic features were scored: cellularity (low, +; moderate, ++; or high, +++), necrosis (absent, 0; low, +; or abundant, ++), the presence of lymphocytes (absent, 0; few, +; or many, ++), the presence of syncytial clusters (defined as groups of cells with inconspicuous, intercellular membranes and nuclear overlapping: absent, 0; few, +; or many, ++), the presence of tubular/ductal-like clusters (defined as groups of polarized cells forming tubules or gland-like clusters: absent, 0; few, +; or many, ++), the presence of large bare nuclei (defined as single, high-grade, malignant nuclei without cytoplasm: absent, 0; few, +; or many, ++), the presence of streaming within the clusters (absent, 0; low, +; or abundant, ++), and the presence of calcifications (absent, 0; or present, +). Moreover, single-cell features, such as cellular borders (well defined or ill defined), type of cytoplasm (abundant or scant; clear or dense), the presence of cytoplasmic vacuoles (absent, 0; low, +; or many, ++), nuclear pleomorphism, (slight or marked), the presence of prominent nucleoli (absent, 0; or present, +), and the type of chromatin pattern (finely hypercromic or clumped) also were evaluated.
Histopathologic findings from excisional biopsies, breast-conserving surgery, or mastectomy specimens were used as the reference standard. Histologic grade, histologic type, pathologic tumor (pT) classification, and pathologic lymph node (pN) status were analyzed according to the World Health Organization classification10 Moreover, other histopathologic parameters, such as pushing margins, tumor lymphoid infiltrate, and the presence of necrosis, as described in basal-like breast cancer,11 were reviewed.
Estrogen Receptor, Progesterone Receptor, and Human Epidermal Growth Factor Receptor2 Analysis
Hormone receptor analysis was performed on surgical specimens by immunohistochemical evaluation of ER (clone SP1; Neomarkers, Lab Vision Corporation, United Kingdom; dilution 1:100) and PR (clone SP2; Neomarkers Lab Vision Corporation; dilution 1:50). The UltraVision LP Detection System HRP Polymer (Thermo Fisher Scientific Anatomic Pathology Division, Fremont, Calif) was used to obtain better signal amplification. The percentage (0%-100%) of neoplastic cells with positive nuclear staining and staining intensity (0, negative staining; +, weak staining; ++, intermediate staining; or +++ strong staining) was recorded for each case. All samples with negative staining and those that had <10% stained nuclei were regarded as negative; and, for the purpose of statistical analysis, all samples with >10% stained cells were regarded as positive.
HER2 status was assessed using a US Food and Drug Administration-approved assay kit (HercepTest; DakoCytomation, Copenhagen, Denmark) according to the manufacturer's instructions. Staining results were scored as follows: 0 indicated samples with no membrane staining; +, partial membrane staining in >10% of tumor cells with no complete circumferential staining; ++, circumferential membrane staining in >10% of tumor cells; and +++, circumferential membrane staining in >30% of tumor cells, all in accordance with expert consensus guideline recommendations for HER2 testing in breast cancer.
Cases scored as ++ were investigated further with fluorescence in situ hybridization (FISH) using the HER2 FISH pharmDx assay kit (DakoCytomation). The technique employs a ready-to-use FISH probe mix that consists of a mixture of Texas Red-labeled DNA probes that cover a 218-kb region, including the HER2 gene on chromosome 17, and a mixture of fluorescein-labeled peptide nucleic acid probes targeted at the centromeric region of chromosome 17. The assay was performed according to the manufacturer's instructions. The slides were then viewed by fluorescence microscopy using an Axioplan (Zeiss, Oberkochen, Germany) with appropriate filters for fluorescein and Texas-Red fluorochromes. The signals were recorded with a charged-coupled device camera (Image Pro-Plus; Media Cybernetics, Inc., Bethesda, Md). In each breast cancer specimen, on average, 60 nuclei were enumerated in the invasive part of the tumor. The HER2 gene copy number, the chromosome 17 copy number, and the average HER2 gene-to-chromosome-17-signal ratio were evaluated according to American Society of Clinical Oncology guideline recommendations.12 Samples with a ratio value >2.2 were considered to be amplified. Polysomy for chromosome 17 was defined as the occurrence of ≥3 copy numbers of centromeres for chromosome 17 per cell according to Salido et al.13
Descriptive analyses included calculation of the observed frequencies and the respective percentages for each categorical variable, whereas means and standard deviations were computed for age (as a continuous variable). The Fisher exact test was used to compare tumor types (TNBC vs HRBC and TNBC vs HER2BC) for each cytologic variable. To compare patients' mean age, pT classification, and pN status by tumor type, a t test was used. P values ≤ .05 were considered statistically significant for both statistical tests.
Tumor type (TNBC vs HRBC [1 vs 0] and TNBC vs HER2 [1 vs 0]) was used as a dichotomous response variable in 2 multivariate, stepwise logistic regression analyses (forward selection). The objective of these analyses was to identify the cytologic factors associated with tumor type.
In a stepwise selection analysis, any significant variables (P ≤ .05) were inserted into the model as covariates, but an attempt was made to remove any insignificant variables from the model before a new, significant variable was added to the model. Each addition or deletion of a variable to or from a model was listed as a separate step; and, at each step, a new model was fitted. In this report, only the final model is presented. Results are provided in terms of odds ratios (ORs) adjusted by the other variables that were tested in the model. To measure the goodness-of-fit of the models, the overall rate of correct classification was estimated.14 Descriptive and multivariate analyses were performed with the SAS statistical software package (version 9.1.3; SAS Institute, Inc., Cary, NC).
Sixty-two tumors were identified triple-negative after immunohistochemical and/or FISH analysis. The patients with those tumors ranged in age from 34 years to 89 years (mean age, 41.6 years). Thirty-seven patients had pT1 disease (59.7%; 5.4% pT1a, 27% pT1b, and 67.6% pT1c), 21 patients had pT2 tumors (33.9%), 2 patients had pT3 tumors, (3.2%), and 2 patients had pT4 tumors (3.2%). Forty-one patients had pN0 disease (66.2%), 17 patients had pN-positive disease (27.4%), and 4 patients had pNX disease (6.4%).
Thirty-three HR-negative tumors had HER2 overexpression. The patients with those tumors ranged in age from 30 years to 86 years (mean age, 58.6 years). Seventeen patients had pT1 tumors (51.5%; 6% pT1a, 6% pT1b, and 88% pT1c), 11 patient shad pT2 tumors (33.3%), 4 patients had pT3 tumors (12.1%), and 1 patient had a pT4 tumor (3%). Twelve patients had pN0 disease (36.4%), 19 patients had pN-positive disease (57.6%), and 2 patients had pNX disease (6%).
There were 82 HR-positive, high-grade carcinomas. The patients with those tumors ranged in age from 31 years to 86 years (mean age, 59 years). Forty-three patients had pT1 tumors (52.4%; 2% pT1a, 21% pT1b, and 77% pT1c), 36 patients had pT2 tumors (43.9%), and 3 patients had pT4 tumors (3.7%). Thirty-one patients had pN0 disease (37.8%), 49 patients had pN-positive disease (59.8%), and 2 patients had pNX disease (2.4%). TNBCs were pN0 more frequently than HRBCs and HER2BCs, and the differences were statistically significant (P < .0001 and P < .006, respectively), whereas no statistically significant difference was observed for patient age or pT/pN status.
Compared with HRBC and HER2BC, TNBC had several statistically significant differences (P < .05; Fisher exact test): TNBCs were more likely to have an abundant necrotic background (Fig. 1), many lymphocytes, many syncytial clusters, and ill defined cell borders (Fig. 2) than non-TNBCs. Moreover a tubular/ductal pattern was observed only rarely in TNBC. Other differences, as presented in Table 1, were observed in cellularity, cellular borders, cytoplasmic vacuoles, cellular pleomorphism (Fig. 3), chromatin pattern, large bare nuclei (Fig. 4), and streaming. The significance of some of these differences disappeared in a multivariate analysis that was adjusted for all of the other variables.
Table 1. Frequency Distribution (%) According to Cytologic Variables and Tumor Type: Triple-Negative Breast Cancer Versus Hormone Receptor-Positive Breast Cancer and Versus Human Epidermal Growth Factor Receptor 2-Positive Breast Cancer (Fisher Exact Test)
Abbreviations: HER2BC, estrogen and progesterone receptor (hormone receptor)-negative, human epidermal growth factor receptor 2 (HER2)-positive breast cancer; HRBC, estrogen and progesterone receptor (hormone receptor)-positive breast cancer; TNBC, triple-negative breast cancer (negative for estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2).
Table 2 provides the results from a multivariate logistic regression analysis for the probability of identifying TNBC versus HRBC according to cytologic variables. Compared with HRBC, TNBC exhibited a significantly higher probability of presenting with a lymphocyte score of ++ (vs 0), a tubular/ductal clusters score of 0 (vs ++), a cytoplasmic vacuoles score of 0 (vs +), marked cellular pleomorphism (vs slight pleomorphism [a score of 1 vs 0]), and an ill defined cellular borders and syncytial clusters score of ++ (vs 0) (Fig. 5). An other significant result was observed when lymphocytes were scored + compared with a score of 0, because the odds of TNBC when lymphocytes are scored + is almost 75% less than the odds of TNBC when lymphocytes are scored 0.
Table 2. Multivariate Logistic Analysis of the Probability of Identifying a Triple-Negative Breast Cancer Versus a Hormone Receptor-Positive Breast Cancer According to Cytologic Variablesa
The model was controlled for cellularity, necrosis, lymphocytes, syncytial clusters, tubular/ductal clusters, nucleoli, cellular borders, cytoplasm, cytoplasmic vacuoles, cellular pleomorphism, cromatin pattern, streaming, and calcification.
A significant interaction between cellular borders and syncytial clusters was observed: Presenting with ill defined cellular borders and syncytial clusters (a score of ++ vs 0) increased the probability of identifying TNBC by >5 times (OR, 5.23; P = .001). This means that the 2 variables together (multiplicative effect) explain the probability of identifying TNBC.
In terms of an adjusted OR, the probability of identifying TNBC more than doubled when lymphocytes were scored ++ versus 0 (OR, 2.24; P = .027). Presenting with tubular/ductal clusters that scored 0 versus ++ increased the probability of TNBC by 2.7 times. Finally, a cytoplasmic vacuoles score of 0 versus + and marked cellular pleomorphism, compared with slight cellular pleomorphism, increased the probability of identifying TNBC rather than HRBC by 4.86 and 4.15 times, respectively. The overall rate of correct classification according to the variables entered into the logistic regression model, was very high (90.8%). The total variance explained by the covariates was equal to 48% (R2 = 0.48).
Table 3 presents the results from a comparison between TNBC and HER2BC using multivariate logistic analysis. Compared with HER2BC, TNBC had a significantly higher probability of presenting with a cellularity score of ++ (vs +++), a tubular/ductal clusters score of 0 (vs ++), and an ill defined borders and syncytial clusters score of + or ++ (vs 0) (Fig. 6). Also in this analysis, the interaction between cellular borders and syncytial clusters was statistically significant: Ill defined borders together with a syncytial clusters score of + (vs 0) increased the probability of identifying TNBC rather than HER2BC by 5.3 times, and ill defined borders together with syncytial clusters (++ vs 0) increased the same probability by 4.3 times. In terms of an adjusted OR, the probability of identify TNBC rather than HER2BC increased by 6 times in samples that had a cellularity score of ++ versus +++ (OR, 6.08; P = .005); otherwise, the probability decreased by >75% in samples with a cellularity score of + versus +++ (OR, 0.23; P = .04). A tubular/ductal clusters score of 0, compared with ++, tripled the probability of identifying TNBC rather than HER2BC (OR, 3.00; P = .023). The overall rate of correct classification was 77.5%, (R2 = 0.27).
Table 3. Multivariate Logistic Analysis of the Probability of Identifying a Triple-Negative Breast Cancer Versus a Human Epidermal Growth Factor Receptor 2-Positive Breast Cancer According to Cytologic Variablesa
The model was controlled for cellularity, necrosis, lymphocytes, syncytial clusters, tubular/ductal clusters, nucleoli, cellular borders, cytoplasm, cytoplasmic vacuoles, cellular pleomorphism, cromatin pattern, streaming, and calcification.
Among the TNBCs, 43 tumors were classified histologically as grade 3 ductal not otherwise specified, 4 were classified as medullary, 7 were classified as atypical medullary, 6 were classified as metaplastic (4 spindle, 2 squamous), 2 were classified as small cell, and 1 was classified as pleomorphic. Basal-like features were observed in 21 tumors. Among the HER2BCs, 31 tumors were classified as grade 3 ductal, 1 was classified as medullary, and 1 was classified as atypical medullary. Among the HRBCs, 2 tumors were classified as grade 3 ductal not otherwise specified, 1 was classified as medullary, and 1 was classified as atypical medullary.
The current study demonstrates that there are significant cytologic differences between TNBC and non-TNBC. The most of statistically significant differences were observed between TNBC and high-grade HRBC. Our multivariate logistic analysis demonstrated that the probability of identifying TNBC versus high-grade HRBC by evaluating lymphocytes, ill defined cell borders and syncytial clusters, tubular/ductal clusters, cytoplasmic vacuoles, and cellular pleomorphism achieved 90.8%.
Many cytologic features of HER2BC and TNBC were overlapping. In fact, fewer statistically significant differences were observed between HER2BCs and TNBCs than between HER2BCs and HRBCs. Multivariate logistic analysis of the probability of identifying TNBC versus HR2BC by evaluating cellularity, ill defined cell borders and syncytial clusters, and tubular/ductal clusters achieved 77.5%. To our knowledge, this is the first study to date highlighting the cytology of TNBC. Four recent studies have discussed the cytologic features of basal-like breast cancer, which frequently is both HR-negative and HER2-negative and has a large molecular overlap with TNBC; however, those breast cancers are not the same entity even if a substantial fraction of TNBCs belongs to the basal-like tumor type.15
Dufloth et al16 observed that the presence of necrosis, prominent nucleoli, and higher cellularity were associated with basal-like breast cancer. Alvarez-Rodrigues et al17 retrospectively studied cytologic features of 13 patients with basal-like breast carcinoma and concluded that, in addition to the features reported by Dufloth et al, the presence of foamy histiocytes and naked, highly atypical nuclei were strongly associated with the basal phenotype and were observed only rarely in poorly differentiated, nonbasal tumors. Akashi et al18 investigated cytologic samples from 17 patients with triple-negative basal-like breast cancer and reported that lymphocyte infiltration, squamous metaplasia, and nuclear findings, such as high mitotic index, naked or large nuclei, irregular nuclear margins, and nucleoli, were significantly more frequent in basal-like tumors than in the luminal A and B subtypes.
In the current study, we observed that necrosis was significantly more frequent in TNBCs than in HRBCs, and many highly atypical, naked nuclei were significantly more frequent in TNBCs compared with HER2BCs. In addition, we observed that syncytial clusters of cells with ill defined borders were present more consistently in FNAC of TNBC compared with FNAC of high-grade non-TNBC. This characteristic was reported previously by Ishihara et al.19 In their study, which included the cytologic specimens from 28 basal-like breast cancers, moderate or marked syncytial status was observed in 92.9% of tumors.
It is noteworthy that many of these features have been previously described in FNAC of medullary and atypical medullary breast cancer.20 However, in our study, only 11 of 62 TNBCs were the medullary or atypical medullary type.
It has been reported that TNBC has characteristic imaging features compared with HER2BC and HRBC. On mammography, TNBC is most frequently observed as a round, oval, or lobular mass with indistinct margins. Moreover, compared with HER2BC and HRBC, TNBC less frequently has an irregular shape, spiculated margins, and calcifications.21, 22 On magnetic resonance imaging,23 unifocality, mass lesion, smooth mass margins, rim enhancement, and very high intratumoral signal intensity on T2-weighted images are associated significantly associated with TNBC.
Over the last decades, many studies have described the cytologic features of specific types of breast cancers and the importance of prognostic information provided by FNAC.24 Moreover, material procured from FNAC can be evaluated by ancillary studies for prognostic factors, including HR status25, 26 and HER2/neu expression.27, 28 Because TNBCs usually are highly aggressive tumors that can benefit from neoadjuvant chemotherapy,29-31 an accurate preoperative diagnosis can be crucial.
In conclusion, although TNBCs embrace a heterogeneous group of tumors, in our current study, they exhibited some common cytologic features that can help to distinguish them from other high-grade breast carcinomas in daily practice. However, despite these promising features, we encourage supplementing the diagnosis with HR and HER2 immunohistochemical studies, which can be performed easily on cytologic material. Fine-needle cytology associated with imaging and clinical patterns should suggest triple-negative cancer preoperatively and should prompt HR and HER2 characterization to achieve better therapeutic planning. If possible, this could save time, effort, and resources in this selected group of patients.