Diagnostic Pitfalls in Breast Cancer Pathology With an Emphasis on Core Needle Biopsy Specimens

DCIS is a malignant intraductal proliferation of epithelial cells, which may be of low, intermediate, or high nuclear grade, with a variety of architectural patterns, including solid, cribriform, micropapillary, papillary, or clinging, and with or without associated necrosis. Given these aforementioned features and that intraductal proliferative lesions more broadly may be benign, atypical, or malignant, it stands to reason that the range of lesions mimicking DCIS is wide.

One of the major diagnostic challenges when evaluating DCIS, particularly intermediate-nuclear-grade DCIS, is distinguishing it from usual ductal hyperplasia (UDH), a potential pitfall with significant management ramifications. Other diagnostic pitfalls include high-nuclear-grade DCIS mimicking the pleomorphic variant of lobular carcinoma in situ (PLCIS) and vice versa, and the solid pattern of low-grade DCIS mimicking classic lobular carcinoma in situ (LCIS).

Characterizing intraductal proliferations on CNB is a routine exercise in breast pathology, but it is not always straightforward. Careful evaluation of both the cytology and the subtle architectural features facilitates correct classification, particularly for those lesions with a solid growth pattern where distinction between florid UDH and intermediate-nuclear grade DCIS is most problematic (Figure 1, A through F).

UDH is composed of a polymorphous population of epithelial cells. On low power the epithelial cells impart a relatively pinker appearance because of a low nucleus to cytoplasm ratio, whereas atypical proliferations look more blue because of an increase in nuclear size. Architecturally, UDH often has peripheral slitlike spaces. If there are bridges or arcades, the cells are streaming and overlapping without conferring rigidity to the structures. Because UDH is composed of a polymorphous population of epithelial cells, the nuclei and chromatin texture look different from one cell to the next.

As noted above, the main diagnostic pitfall is with solid-pattern, intermediate-nuclear-grade DCIS. Both proliferations appear to have a variable population of epithelial cells that lack architecture. However, in both, there may be architectural clues to the correct diagnosis: subtle areas of polarization or rosette formation in the case of DCIS, or parallel orientation of cells around any slitlike lumens that may be present in the case of UDH. The nuclear features and quality of the chromatin may be helpful too. In addition to the nuclear heterogeneity described above, the nuclei of UDH may also show grooves and/or eosinophilic pseudoinclusions.^1,2  In contrast, DCIS is composed of a clonal population of epithelial cells, and the chromatin quality and texture are similar in all nuclei (Figure 1, D through F). Necrosis, apoptosis, and mitoses should prompt consideration of DCIS,^1,3,4  but may occasionally be seen in UDH. Indeed, UDH with necrosis can be a diagnostic pitfall. In most cases when necrosis is present, either focal or comedo type, it is associated with DCIS. However, it is important to be aware that areas of necrosis may rarely be present in association with UDH¹  (Figure 1, A through C). It can be helpful when evaluating a solid intraductal proliferation with necrosis to mentally subtract the necrosis and focus on the epithelial proliferation to assess for the aforementioned features of UDH.

In problematic cases, estrogen receptor (ER) and cytokeratin 5/6 (CK5/6) immunostains may be a necessary aid in differentiating between UDH and atypia^5,6  (Figure 1, A through F). DCIS will show strong, diffuse staining with ER and an absence of expression with CK5/6. In contrast, UDH, being a heterogeneous proliferation, will have a variable or mosaic pattern of staining with both ER and CK5/6. Note that some high-nuclear-grade DCIS lesions are ER negative and may express basal cytokeratins,⁷  such as CK5/6, but the degree of nuclear pleomorphism invariably permits diagnosis without the need for immunostains for that particular differential diagnosis.

Finally, when evaluating ER as a predictive biomarker for patient management, always use this as an opportunity to confirm that the proliferation is DCIS. Both low- and intermediate-grade (and most high-grade) DCIS will demonstrate strong, diffuse ER expression.⁸  If the stain intensity is variable from cell to cell, rereview of the hematoxylin-eosin (H&E) slide is indicated to ensure the correct diagnosis has been rendered.

As will be emphasized throughout this article, the management ramifications are probably most pronounced at the time of CNB diagnosis, where an erroneous diagnosis of DCIS or invasive carcinoma sets the patient on a pathway of surgery and radiation therapy, with or without chemotherapy. A diagnosis of UDH requires no further management; thus, careful interpretation of the morphologic features, and, if necessary, use and interpretation of immunostains as described above, are critical.

Low-grade and high-grade DCIS are usually readily recognized. Low-grade DCIS has a very uniform appearance on low power, conferred by both the cytologic monomorphism and architectural atypia, which is often in the form of cribriform spaces, microacini, rigid bridges, arcades, and micropapillations. High-grade DCIS may have these same architectural features as well as a so-called clinging pattern, in which the involved space is lined by 1 or 2 layers of DCIS cells. Both may also have a solid pattern, and, as with intermediate-nuclear-grade DCIS, it is this pattern that poses the greatest diagnostic challenge, the risk here being misinterpreting low- or high-nuclear-grade solid-pattern DCIS as classic or nonclassic LCIS, respectively (Figure 2, A through F).

Like low-nuclear-grade DCIS, classic LCIS is also composed of low-grade monotonous epithelial cells, sometimes with intracytoplasmic vacuoles, but lobular neoplasia lacks architectural atypia. In addition, LCIS cells are characterized by cytologic dyshesion due to the loss or inactivation of the cell adhesion molecule E-cadherin.⁹  However, when the cells are filling and distending an acinus, it may be hard to appreciate this dyshesive appearance (Figure 2, A and B). In such cases, looking for subtle architectural features, such as formation of microacini or rosettes or peripheral polarization, can be a clue that the proliferation is DCIS (Figure 2, C and D). In many diagnostic scenarios, evaluation of the background breast parenchyma can provide additional clues as to how to classify the lesion in question, but both classic LCIS and low-nuclear-grade DCIS belong to the same low-grade breast neoplasia family, and as such often have the same putative precursor lesions in the surrounding breast tissue,^10–12  that is, atypical ductal hyperplasia (ADH), atypical lobular hyperplasia, and flat epithelial atypia. That said, if there are more readily recognizable foci of ADH or atypical lobular hyperplasia exhibiting the characteristic architectural and cytologic features of atypical ductal and lobular lesions discussed above with which the ambiguous solid intraductal proliferation can be compared, this may be helpful. If diagnostic uncertainty remains, immunostains can be of value in distinguishing ductal from lobular proliferations (see discussion below).

It bears mentioning that the distinction of low-grade DCIS from classic LCIS has important management ramifications: in the CNB setting, excision is not required for radiologic-pathologic concordant classic LCIS, whereas DCIS is surgically excised; in excision specimens, margin evaluation is required for DCIS, but not for classic LCIS; and radiation therapy may be indicated for patients with DCIS. Of note, if a solid-pattern low-nuclear-grade intraductal proliferation of limited extent is present in a CNB, that is, ADH versus DCIS, it is the recommendation of the expert editorial panel of the World Health Organization (WHO) Classification of Tumors of the Breast²  that these cases be categorized as ADH or as an atypical intraductal proliferation, rather than as definitive low-grade DCIS, given the limited sampling in this setting.

For high-grade intraductal proliferations with comedo necrosis, the first diagnostic consideration is usually DCIS. The potential pitfall here is with PLCIS, an uncommon lesion that can, on occasion, be difficult to distinguish from DCIS. Both are composed of ducts or acini expanded by large cells with pleomorphic nuclei, and often central necrosis with or without associated calcifications (Figure 2, E and F). Both lesions may have apocrine features. Again, looking for subtle morphologic clues such as microacini and areas of cellular polarization, in the case of DCIS, or cytologic dyshesion and intracytoplasmic vacuoles or concomitant classic lobular neoplasia, in the case of PLCIS, can aid in making the diagnosis.¹³ 

From a management perspective, both high-grade DCIS and PLCIS require excision following diagnosis on CNB.^14–17  Margin evaluation should be provided for both entities in excision specimens^2,18–21 ; however, although most surgeons would re-excise if margins were positive for PLCIS,^19,20  the decision to irradiate the breast of patients with a diagnosis of PLCIS varies among oncologists.^22–25 

Immunostains are a useful tool in diagnostically challenging or problematic cases. In LCIS, alterations in the CDH1 gene, usually somatic mutations or promoter hypermethylation, are present, whereas the gene is intact in cells with a ductal phenotype.^26,27  This can be exploited by immunostaining for E-cadherin, as well as p120 catenin and/or β-catenin⁹  (Figure 2, A through F). Cells with a ductal phenotype exhibit a strong membranous pattern of expression with all 3 antibodies. Cells that have a lobular phenotype lack this cell membrane staining; both E-cadherin and β-catenin demonstrate an absence of expression, whereas with p120 catenin, the LCIS cells have cytoplasmic expression because of dissociation of the molecule away from the cell membrane and into the cytoplasm. Aberrant staining with E-cadherin has been reported, often incomplete/noncircumferential and granular, or of variable/decreased intensity; in such cases comparison with the background ductal epithelium can be helpful.^9,28 

On occasion, extensive lymphovascular space invasion (LVI) can mimic DCIS. In this situation, the nests of tumor within the lymphovascular spaces are often large and may have comedo necrosis. The endothelial cells are often attenuated and are readily misinterpreted as myoepithelial cells. Clues to the correct diagnosis include distribution of the LVI around normal terminal duct lobular units and/or adjacent to vascular bundles. Myoepithelial cell immunostains and the lymphatic endothelial cell marker D240 can be helpful in resolving this diagnostic dilemma, with the former being positive in DCIS and the latter being positive in LVI. Note that D240 may cross-react with myoepithelial cells, hence the importance of including a marker such as p63 or smooth muscle myosin heavy chain (SMMHC).

Mimics of cribriform pattern DCIS include collagenous spherulosis, particularly when involved by LCIS, invasive cribriform carcinoma, and adenoid cystic carcinoma. For further discussion of adenoid cystic carcinoma, see the article “Salivary Gland-Like Tumors of the Breast: A Comparison of Clinicopathologic Features and Molecular Pathogenesis with Their Salivary Gland Counterparts” in this journal.

Collagenous spherulosis is recognized as a cribriform-appearing process in which the rounded lumens within an intraductal proliferation are surrounded by myoepithelial cells that produce basement membrane material. Thus, the nuclei of the cells surrounding the lumens are more spindled and lie parallel to the spaces, which contrasts with cribriform-pattern DCIS, in which the epithelial cells surrounding the lumens are polarized toward the spaces. Greater diagnostic difficulty may arise, however, when collagenous spherulosis is involved by classic LCIS. In this situation, the intraductal proliferation has a very monomorphic appearance; that, compounded by the cribriform architecture, creates the morphologic impression of low-grade DCIS (Figure 3, A). The key to distinguishing collagenous spherulosis from DCIS is the presence of the basement membrane material in the lumens lined by compressed myoepithelial cells in the former. This amorphous material may look like eosinophilic spherules or may have a more amorphous appearance that can be either eosinophilic or basophilic. Immunostains for myoepithelial cells are useful to make the distinction (Figure 3, B). Lack of staining by E-cadherin can confirm that the process is involved by lobular neoplasia (Figure 3, C). Of note, the myoepithelial cells will stain for E-cadherin.²⁹  As noted above, the cribriform pattern of adenoid cystic carcinoma, another epithelial-myoepithelial lesion, is a mimic for collagenous spherulosis and DCIS. Collagenous spherulosis is benign and will not have an infiltrative pattern, whereas adenoid cystic carcinoma will.^1,29  Adenoid cystic carcinoma will label with MYB by immunohistochemistry (IHC) or MYB by fluorescence in situ hybridization.³⁰ 

DCIS can also have a papillary pattern, and therefore the many papillary lesions of the breast enter into the differential diagnosis; furthermore, the diagnostic difficulty here is not limited to CNB specimens. An algorithmic approach is of particular value with papillary lesions (Figure 4). Upon recognition that a proliferation is papillary through the identification of fibrovascular cores, evaluation for the presence or absence of myoepithelial cells at the periphery of the proliferation and running along the papillae is a useful starting place. Once this has been established, the epithelial component should be evaluated for the presence or absence of atypia according to the same morphologic criteria used for assessment of nonpapillary intraductal proliferations (see discussion above). On occasion, immunostains will be needed, either to highlight the presence of the myoepithelial cell layer or to aid in differentiating florid UDH from DCIS using CK5/6 and ER as described above.^5,6 

Papillary DCIS is an intraductal proliferation characterized by fibrovascular cores surmounted by malignant epithelial cells.^2,31–33  The epithelial cells tend to be columnar in arrangement, organized perpendicular to the fibrovascular cores. The nuclei can be of low, intermediate, or high nuclear grade. Although the peripheral myoepithelial cell layer is retained, there is an absence of myoepithelial cells running along the fibrovascular cores in papillary DCIS²  (Figure 5, A and B).

DCIS can also have a solid papillary growth pattern. This entity is characterized by multiple large, round nests of tumor cells with a fine, almost inconspicuous, fibrovascular network^2,31–33  (Figure 5, C). The nuclei are often of intermediate nuclear grade and may have neuroendocrine features, that is, fine, even chromatin.^34,35  The particular challenge with solid papillary carcinoma is that the tumor cell nests, despite appearing well circumscribed, may lack a peripheral myoepithelial cell layer.³⁶  Despite this, these tumors typically have an indolent behavior; as such, if the nests are well rounded and the growth pattern is noninfiltrative, the WHO expert editorial panel advises that these tumors be staged and managed as in situ lesions.² 

Encapsulated papillary carcinoma (EPC) is another category of papillary carcinoma. This is a well-circumscribed papillary lesion, surrounded by a fibrous capsule^2,37  (Figure 5, D). The thin/delicate fibrovascular cores are lined by low- or intermediate-grade nuclei. Myoepithelial cells are typically absent around the periphery of the lesion, as well as being absent along the fibrovascular cores. EPCs are generally ER positive and HER2 negative, and have an indolent behavior. Despite the absence of a peripheral myoepithelial cell layer, it is recommended that EPCs be staged and managed as pTis tumors. EPCs with high-nuclear-grade atypia and/or that are triple negative or HER2 positive are rare and should be graded, staged, and managed as invasive carcinomas.^2,38  On CNB it is often difficult and may be impractical to tease apart papillary DCIS from solid papillary carcinoma from EPC. In the absence of a component of conventional invasive carcinoma, we render a diagnosis of “papillary carcinoma, at least in situ,” with a note deferring definitive characterization to the excision specimen.

A rarely encountered, but nonetheless important, potential pitfall in the setting of a malignant papillary proliferation is misinterpreting metastatic serous carcinoma of the gynecologic tract as primary papillary carcinoma of the breast, either in situ or invasive.^39,40  Clinical history of prior ovarian/fallopian tube or peritoneal carcinoma will facilitate appropriate interpretation of the breast biopsy specimen. However, it is not uncommon for that history to not have been provided or for presentation in the breast or axillary lymph nodes to be the first site of identification of serous carcinoma, which is notoriously occult, even at an advanced stage.

Morphologic clues that should give pause when evaluating a papillary carcinoma on CNB include a tumor of high nuclear grade with frequent mitoses. Although EPCs with a high nuclear grade have been described, these are rather unusual, representing only about 3% of encapsulated carcinomas.³⁸  The presence of numerous psammomatous calcifications can also be indicative of serous carcinoma (Figure 5, E). It should be noted that the breast biomarker panel (ER, progesterone receptor [PR], and HER2) will not necessarily be a cue to the diagnostic error, as high-grade serous carcinomas are frequently ER positive. Mammaglobin and gross cystic disease fluid protein 15 (GCDFP-15), which may be used to support diagnosis of a tumor of breast origin, may be positive in almost 20% of cases of serous carcinoma; and furthermore, negativity does not exclude either diagnosis.⁴¹  Differential immunostains include WT1 and PAX8, which are positive in serous carcinoma and rarely so in breast carcinoma, and GATA3, for which the converse is found.^40–44  As with so many of the pitfalls discussed in this article, attention to subtle deviations from the characteristic pathology is the means to prevention of diagnostic error.

An additional papillary pitfall is nodular hidradenoma, a benign primary cutaneous neoplasm that arises from the apocrine sweat glands.^45–47  When these lesions occur in the breast, they have mainly solid and cystic architecture with a variable cellular composition (Figure 5, F). The cells comprising the solid component include clear cells, eosinophilic cells, epidermoid cells, and transitional cells.⁴⁸  Sometimes ductal differentiation can be appreciated in the form of intracytoplasmic lumina, which can be highlighted by carcinoembryonic antigen (CEA). These lesions have associated hyalinized stroma throughout that may be vascularized, which gives the appearance of fibrovascular cores. Thus, when they occur in the breast, the main differential diagnosis is with a papillary lesion. Apocrine or squamous metaplasia may be present, which can also be seen in some papillary lesions. Although overt cytologic atypia in terms of pleomorphism is not present, the uniformity of the cells imparts a low-grade appearance and may be mistaken for low- or intermediate-nuclear grade DCIS with a papillary pattern. These lesions generally express low- and high-molecular-weight cytokeratins, p63, and GATA3.⁴⁷  The unusual morphology and staining profile are clues to the correct diagnosis. These lesions may also be in the differential diagnosis for other primary breast lesions, including a mucoepidermoid carcinoma or adenomyoepithelioma.^47,49 

For a more in-depth discussion of the diagnostic features of the principal papillary lesions of the breast, the reader is referred to many excellent reviews or book chapters on the topic.^31–33,37 

The morphologic spectrum and special histologic types of invasive carcinoma of the breast are even greater than for DCIS. Tumors can be of no special type (NST) and of low, intermediate, or high grade (grades 1–3); combinations of ER, PR, and HER2 positive or negative; and/or of a special histologic subtype, of which the WHO Classification of Tumors of the Breast²  lists more than 20. For reasons of scope, and to highlight those entities for which management ramifications are of greatest significance, this section will discuss the pitfalls associated with ER-positive invasive carcinomas, NST, and benign sclerosing lesions and adenosis; triple-negative breast cancers of NST; and the spindle cell metaplastic carcinomas.

Radial scars (RSs) and complex sclerosing lesions (CSLs) are benign sclerosing lesions with glandular structures entrapped in a fibroelastotic stroma. Associated epithelial proliferative lesions, such as UDH, papillomas, and apocrine metaplasia, may be present.²  RSs are smaller, with a central nidus and a stellate pattern. CSLs are larger and do not have as organized an architecture, a distinction that may not be easily made on CNB. Sclerosing adenosis, a benign proliferation of small glands and nests, tends to be less problematic, except when involved by carcinoma in situ. On imaging studies, the stellate appearance of RS/CSL can be a mimic for carcinoma,^50,51  particularly on tomosynthesis,^50,52,53  and as such these lesions are often targeted for biopsy; smaller lesions may be incidentally identified on CNB or excision.⁵⁰ 

Benign sclerosing lesions also mimic invasive carcinoma histologically, specifically grade 1 or grade 2 carcinoma, and rarely low-grade adenosquamous carcinoma (LGASC; see below). On low-power microscopic examination, the epithelial elements are seen to have a lobulocentric architecture radiating out from the central fibroelastotic core. However, these features may not be as evident (or may be overlooked) when the CNB specimen captures only a portion of the lesion. In such instances, the epithelial nests may appear to have an infiltrative and haphazard growth pattern in association with dense sclerotic stroma, as can be seen in some invasive carcinomas. Recognizing the lobulocentricity of an RS, with its central nidus, is key to the correct diagnosis (Figure 6, A). CSLs have a more irregular pattern, distorting ducts and lobules often involved by florid UDH, again creating an impression of irregular, infiltrative nests similar to those seen in DCIS or invasive carcinoma (Figure 6, B). Myoepithelial cells can be markedly attenuated and difficult to appreciate on H&E-stained sections of benign sclerosing lesions and sclerosing adenosis, prompting consideration of ancillary IHC studies if confirmation of an intraductal process is needed (see discussion below).⁵⁴ 

It is especially challenging when a benign sclerosing lesion or sclerosing adenosis is involved by DCIS or LCIS, as the cytologic atypia and stromal sclerosis or the numerous cytologically atypical small glands and nests, respectively, will prompt consideration of invasive carcinoma (Figure 6, C and D). Careful attention to the lobulocentric configuration on low-power examination and appreciation of the stromal changes (fibrotic or hyalinized) more characteristic of a benign sclerosing lesion, as opposed to desmoplastic, may help prevent erroneous diagnosis as invasive carcinoma. Immunostains for myoepithelial cells, such as p63, SMMHC, and calponin, will help confirm that the atypical process is in situ, though these must be interpreted with caution as there may be reduced or absent expression with one or more of these immunostains in this setting⁵⁴  (Figure 6, E and F). ER and CK5/6, if needed, can support interpretation of an intraductal proliferation as DCIS versus florid UDH, and E-cadherin and p120 catenin immunostains can confirm a lobular phenotype in the case of LCIS involving sclerosing adenosis or a benign sclerosing lesion.^5,6,9  Without an awareness of the histologic pitfalls, an erroneous diagnosis is all too easy to render, with significant management ramifications.

As noted above, when evaluating hormone receptors, and HER2 in a putative case of invasive carcinoma, there is an opportunity, or rather an obligation, to confirm the invasive nature of the lesion. If the ER is variably positive, it is important to consider the possibility that the proliferation may represent UDH in a sclerosing lesion and not an invasive carcinoma. As the breast biomarkers are invariably first performed on CNBs, this second check can spare the patient an unnecessary surgical excision with/without lymph node sampling. Note that this safeguard will not apply when carcinoma in situ involves a benign sclerosing lesion or sclerosing adenosis.

LGASCs are composed of a mixture of epithelial glands/tubules admixed with squamous cell nests, all set in a spindled stroma. Here, the diagnostic pitfalls include not appreciating the presence of the tumor at all, which can at times be quite subtle, or, when identification of an adenosquamous proliferation is made, not recognizing it as a LGASC, but rather considering it to be part of the reactive stroma of a benign sclerosing lesion.⁵⁵  This latter situation is particularly challenging, as LGASCs, although rare tumors in and of themselves, often arise in association with benign sclerosing lesions such as CLSs and sclerosing papillomas.⁵⁶  Lymphoid aggregates may be present in association with LGASC, and can be a helpful clue. In CNB specimens, in which LGASC is a serious consideration, a descriptive diagnosis with recommendation for excision is most prudent, as definitive characterization can be very challenging on CNB even for experienced breast pathologists. Immunostains are not especially helpful in resolving the issue.⁵⁵ 

During any needling procedure, there is the opportunity for some of the epithelial elements of the targeted lesion to be displaced into the adjacent stroma; this phenomenon is particularly prone to occur following biopsy of papillary proliferations.⁵⁷  Failure to recognize the phenomenon for what it is and misclassifying it as invasive carcinoma has significant consequences, including unnecessary surgical re-excision, sentinel lymph node sampling, and radiation therapy, not to mention the emotional trauma for the patient.

There are morphologic clues, however, that can aid in preventing diagnostic error. First, the displaced epithelium is present within the granulation tissue of the biopsy site. The nests and single cells are often scattered in a haphazard manner within this reactive stroma. Fat necrosis, histocytes, hemorrhage, and hemosiderin are indicative of this being the biopsy site. The epithelial cells themselves are usually rather degenerated in appearance, sometimes having a somewhat squamoid appearance. These features are sufficiently characteristic that a diagnosis of biopsy site change with displaced epithelial elements can be included in the pathology report (secondary to the diagnosis of the target lesion). It is important to note that the myoepithelial cells are often absent (ie, not displaced with the epithelial elements), so immunostains for p63, SMMHC, and calponin, for example, will invariably be negative, which may compound any initial interpretative error. Caution should be exercised in the absence of a preexisting diagnosis of carcinoma or in the absence of conventional invasive carcinoma away from the biopsy site. Cases of DCIS can be more problematic, but careful evaluation using the above morphologic features will prevent misdiagnosis.

Triple-negative tumors present a particularly high-risk situation, principally because of the management implications associated with a diagnosis of ER-, PR-, and HER2-negative breast cancer (ie, triple-negative breast carcinoma [TNBC]) diagnosed on CNB, that being the likelihood the patient will receive preoperative (neoadjuvant) chemotherapy. Although there are ramifications to diagnostic errors for the lesions discussed above, it could be argued that the impact of diagnostic error is greatest for this group of tumors. Thus, it is incumbent on the pathologist signing out the breast predictive biomarker studies to rereview the H&E slide to ensure the morphologic features are compatible with a diagnosis of TNBC.

Classic TNBCs are characterized by sheets of tumor cells with large pleomorphic nuclei, numerous mitoses, and, often, large zones of necrosis and abundant tumor-infiltrating lymphocytes. In many cases there is no, or only limited, associated DCIS.^58,59  Given these high-grade morphologic features and the triple-negative immunophenotype, several other entities may enter into the differential diagnosis, including epithelioid angiosarcoma, high-grade B-cell lymphomas, and metastases from other sites, such as metastatic malignant melanoma. Without considering the possibility of these other diagnoses, it is all too easy to render an incorrect diagnosis of breast carcinoma, NST, in this situation.

First and foremost, knowledge of any prior history of breast or other cancers is especially crucial in this scenario. If there is a prior history of breast cancer treated with excision and radiation therapy, angiosarcoma may be a consideration. In the setting of history of another type of cancer with the potential for high-grade epithelioid morphology, the breast CNB should be assessed and, if indicated, further worked up with the specific differential diagnosis in mind. Comparison with the primary tumor, if available, is a prudent starting point.

In addition to patient history, there are morphologic clues that should prompt consideration of diagnoses other than TNBC. Any unusual features, such as the presence of abundant red blood cells or a vaguely vasoformative architecture; pigment, plasmacytoid-looking cells, or nuclei with bright eosinophilic nucleoli; or frankly, any feature that deviates from the typical in an ER-, PR-, and HER2-negative tumor may be reason to perform additional immunostains. The presence of high-nuclear-grade DCIS is supportive of the diagnosis of TNBC; its absence does not exclude it.

Routine use of “breast markers” such as GCDFP-15, mammaglobin, and GATA3 is not recommended, as the vast majority of cancers occurring in the breast are readily recognizable as primary breast carcinomas and conform to the expected hormone receptor and HER2 pattern. However, triple-negative cancers in the breast merit special consideration and workup with additional immunostains, first to rule in carcinoma (with keratin immunostains), followed by confirmation of breast origin with the aforementioned breast markers if indicated (Table). SOX10 and TRPS1 (trichorhinophalangeal syndrome type 1) are markers of additional value in this particular situation, being of greater sensitivity and specificity in the identification of TNBCs.^41,60–63  Note, though, that SOX10 does not rule out metastatic melanoma (see discussion below) and should be used in combination with a keratin immunostain, as stated above.

Angiosarcoma of the breast is uncommon, and primary angiosarcomas particularly so. Radiation-induced (secondary) angiosarcomas are more frequently encountered. These are typically located in the skin and subcutaneous tissue of the breast, rather than the breast parenchyma itself, which is the usual location of primary angiosarcoma of the breast.

The specific diagnostic pitfall here is with high-grade epithelioid angiosarcoma encountered in a CNB specimen mimicking primary breast carcinoma.⁶⁴  In addition to a prior history of breast irradiation, secondary angiosarcomas are usually associated with clinical signs on the skin of the breast: violaceous plaques or papules, or skin changes resembling a bruise, which if noted on the pathology requisition can be an invaluable clue to the correct diagnosis. Knowledge that the tumor is more superficially located rather than deep in the breast may serve as an additional clue.

As alluded to above, microscopic examination may reveal areas of hemorrhage or vaguely vasoformative foci within the solid sheets of tumor cells. On occasion, the nuclei of angiosarcoma are rather bizarre, and almost “too atypical” for invasive ductal carcinoma, NST (Figure 7, A). This constellation of clinicopathologic features merits additional IHC workup with both keratin and vascular markers to support the diagnostic impression. Note should be made of the fact that angiosarcoma, and epithelioid angiosarcoma in particular, can demonstrate positivity with keratins (such as AE1/AE3 and CAM 5.2)⁶⁵ ; thus, use of a panel of immunostains is recommended rather than a single “rule-in/rule-out” keratin stain. Vascular markers ERG and CD31 show strong diffuse staining; variable expression is seen with CD34 and FLI1.^2,66–68  Immunoreactivity is seen with MYC in secondary angiosarcoma, with concomitant c-myc amplification on fluorescence in situ hybridization.^69–71 

Once again, patient history is critical to avoiding the diagnostic pitfall. In the absence of this, keeping an open mind when evaluating high-grade tumors of the breast is imperative. Cytologically, there may be little to distinguish TNBC from metastatic melanoma with an epithelioid morphology, with the possible exception, as noted above, of bright eosinophilic nucleoli and more plasmacytoid features, if present (Figure 7, B).⁷²  The identification of pigment should prompt further scrutiny to determine whether it represents melanin pigment (black or brown coarse granules) or hemosiderin (golden brown globules).

Again, IHC workup will be needed to facilitate distinction of this differential diagnosis. Be aware that some melanoma markers, for example S100 protein and SOX10, can stain carcinomas⁷³ ; indeed, the latter is often used to support breast origin in TNBCs. To compound the difficulty, malignant melanoma, particularly the epithelioid variant, may rarely stain with some keratin immunostains. A panel that includes a keratin cocktail, S100 protein, SOX10, and Melan-A represents a reasonable option (Table).

Primary breast lymphoma is exceptionally rare, representing less than 0.5% of all malignant breast neoplasms.^2,74,75  Secondary involvement of the breast by systemic lymphoma is also unusual.^2,76  Diffuse large B-cell lymphoma arising in or involving the breast is the most common of this rare group, accounting for between 45% and 70% of primary breast lymphomas.⁷⁷  It has a bimodal distribution, peaking in ages 30 to 35 years and 55 to 60 years.⁷⁸  That young patients are affected bears relevance to this discussion, as a greater proportion of younger patients have TNBC.

The clinical presentation of a young woman with a rapidly developing breast mass is a high-anxiety situation often accompanied by a request for rush processing and interpretation of the CNB. The microscopic appearance of sheets of large tumor cells with fairly pleomorphic nuclei and prominent nucleoli, as well as numerous mitotic figures and admixed lymphocytes, is superficially compatible with a diagnosis of TNBC⁷⁹  (Figure 7, C). As emphasized above, it takes consideration of the possibility of alternate diagnoses supported by immunostains to prevent diagnostic error, especially when under pressure to render a diagnosis quickly. A keratin cocktail and CD20 will allow for appropriate triage of the case and further hematopathologic workup if CD20 positivity in the tumor cells is confirmed.

At the other end of the spectrum are low-grade triple-negative tumors. This group includes lesions categorized as benign, of low malignant potential, and frankly malignant. Here the imperative is to convey to the treating physician that the (malignant) tumor is not a TNBC of NST, which would typically be managed with neoadjuvant chemotherapy as discussed above, but rather that it is a tumor of special histologic subtype and generally of indolent behavior for which chemotherapy is not indicated.

The triple-negative carcinomas of special histologic subtype include adenoid cystic carcinoma, acinic cell carcinoma, secretory carcinoma, and tall cell carcinoma with reversed polarity (TCCRP). Each of these is characterized by specific histopathologic features that are fairly readily recognizable when the diagnosis is considered, though admittedly more challenging in the setting of CNB. The first 3 are discussed in detail in a companion article in this special section. The latter, TCCRP, is a more recently defined entity with which some may not be familiar.^80–84 

TCCRP is a low-grade triple-negative (or low-ER-positive) breast carcinoma characterized by rounded nests of tumor cells with fibrovascular cores, which often contain histocytes.²  The epithelial cells within the nests are tall, with oval-shaped nuclei and nuclear grooves. Strikingly, the nuclei within the tumor cells are located at the apical rather than the basal aspect of the cell, imparting the reversed polarity of the tumor’s name (Figure 7, D). The nests have an infiltrative pattern of growth and lack a myoepithelial cell layer. Calcifications may be seen in association with these tumors. As mentioned, the morphology is so characteristic that if it is present in full, an IHC workup is not required. In less well-developed examples, ancillary studies may be needed to support the diagnostic impression. Unusually for a low-grade invasive carcinoma, these tumors express CK5/6 in addition to CK8/18.⁸²  Myoepithelial cell markers are negative. ER is negative or low positive, and HER2 is negative.⁸⁰  Interestingly, a large majority (>75%) of TCCRPs harbor a hotspot mutation in IDH2 that can be leveraged using the IDH2 R172 IHC antibody, which is both sensitive and specific for this diagnosis.^84,85 

The diagnostic pitfalls here could be not providing the specific histologic subtype and categorizing the lesion as a low-grade invasive ductal carcinoma, NST (recall, that ER, PR, and HER2 negativity in a low-grade tumor should prompt reconsideration) or mistaking the nested growth pattern and presence of fibrovascular cores and oval-shaped nuclei with nuclear grooves for UDH or papillomatosis, an error that could be compounded by CK5/6 positivity and low ER positivity. Appreciating the reversed cell polarity is really the key to correct diagnosis. Metastatic thyroid carcinoma may also be a consideration, but markers of thyroid differentiation are negative.^81,82 

Adenomyoepithelioma is a breast tumor newly recognized as being of low malignant potential.²  These tumors are characterized by a biphasic population of tumor cells, namely epithelial and myoepithelial cells, arranged in a variety of patterns, most commonly as a lobulated, well-circumscribed mass. As these tumors are infrequently encountered, and because some of the architectural features may be absent or less obvious in CNB specimens, there is risk of diagnostic error, which will be discussed in brief here (see salivary gland tumor article in this issue for fuller discussion of this entity).

The pitfall most commonly encountered with this entity is to misinterpret the accompanying collagenous or myxoid matrix within the tumor as desmoplasia and to ascribe an infiltrative growth pattern to the glands or nests within this stroma, overlooking the lobulated low-power appearance and the presence of a biphasic cell population. If an error is made at the time of H&E diagnosis, it should be appreciated at the time of receptor review, as the ER profile (negative or low positive) would not be compatible with the histologic appearance of an apparently low-grade invasive ductal carcinoma. Rereview of the H&E slide with deployment of myoepithelial cell immunostains will assist in establishing the correct diagnosis, as the latter will decorate the numerous myoepithelial cells present within the tumor.⁸⁶ 

Evaluation of spindle cell lesions is challenging and can be even more so on CNB. The diagnostic considerations are broad, ranging from reactive lesions to high-grade malignant lesions, each with numerous mimics. Some metaplastic carcinomas can be deceptively bland, and some benign processes can mimic malignant lesions. The most commonly encountered spindle cell proliferation in the breast is a scar, so clinical and radiologic history is important during microscopic evaluation.

Generally, spindle cell lesions and their differential diagnostic considerations are grouped as either bland appearing or atypical. The most important lesion to consider when evaluating any spindle cell proliferation is a metaplastic carcinoma. The morphologic impression should be used to create a differential diagnosis with a panel of IHC to support the workup and diagnosis.

This section will review some of the main diagnostic challenges and pitfalls when faced with a spindle cell metaplastic carcinoma (specifically fibromatosis-like metaplastic carcinoma [FLMC] and spindle cell metaplastic carcinoma) of the breast on CNB. For greater discussion of the general pathologic features of metaplastic carcinomas, the interested reader is referred to other reviews on the diagnosis of breast spindle cell lesions.^87–91 

Metaplastic carcinomas are rare, representing 0.2% to 1% of all invasive breast carcinomas.²  They comprise a heterogeneous group of TNBCs, classified into several different subtypes by the WHO Classification of Tumors of the Breast: LGASC, FLMC, squamous cell carcinoma, spindle cell carcinoma, and metaplastic carcinoma with heterologous mesenchymal differentiation.² 

In pure spindle cell lesions, evidence of epithelial differentiation is essential for the diagnosis of metaplastic carcinoma, either recognized morphologically or demonstrated by IHC. The presence of DCIS can be a clue to the diagnosis. Metaplastic carcinomas stain with keratin immunoantibodies, especially high-molecular-weight keratin, as well as with p63 or p40. Variability and heterogeneity of expression are seen; therefore, including multiple keratins in the panel of immunostains is necessary to confirm epithelial differentiation, or, more accurately stated, to ensure it is not missed.

As the name implies, this subtype of metaplastic carcinoma has morphologic similarities with desmoid fibromatosis (DF). Both lesions are composed of bland spindle cells arranged in long, sweeping fascicles, although more variability is seen in the fascicular pattern of FLMC than in DF. Both tumors infiltrate into the adjacent breast parenchyma or surrounding structures (Figure 8, A through D). Without careful search, nuclear atypia is not obvious; cytologic atypia is best detected at the periphery of FLMC, which may not have been sampled during a CNB procedure. Mitotic figures, too, can be difficult to identify without careful scrutiny. A key to reaching the correct diagnosis is identification of subtle areas of epithelial or squamous differentiation of the spindle cells, best appreciated as condensation or coalescence of the tumor cells into more epithelioid-appearing foci. IHC to rule in a metaplastic carcinoma and rule out a DF is essential for diagnosis. Lymphoid aggregates may be located at the periphery of DFs, a feature that would favor the diagnosis on H&E. DF shows nuclear staining for β-catenin in 76% to 100% of cases and is negative for p63 and keratins, ruling out FLMC.²  In CNB, some circumspection may be prudent in a pure spindle cell lesion lacking definitive immunophenotypic features of one or the other of these differential diagnoses.

Spindle cell metaplastic carcinomas can have a spectrum of nuclear atypia and a mixture of growth patterns. As the name implies, the predominant pattern is of spindle cells, but areas of more epithelioid differentiation can often be identified (Figure 8, E and F). Spindle cell metaplastic carcinomas can have some morphologic overlap with the lower-grade and less-aggressive FLMC, but the nuclear atypia should be intermediate or high grade and mitotic figures should be readily identified. The diagnostic pitfall for these cases is the resemblance of spindle cell metaplastic carcinoma to the stromal component of a malignant phyllodes tumor, particularly in the setting of a pure malignant spindle cell proliferation on a CNB.^2,92,93 

The stromal cells of malignant phyllodes tumor are cytologically atypical with brisk mitotic activity. If only the stromal component is sampled, there are no definitive morphologic features that can reliably distinguish these 2 diagnoses from one another. Additional levels may reveal some ducts or strips of benign epithelium on the tips of the core biopsy fragments, favoring phyllodes tumor. In the absence of demonstrating a coexistent benign epithelial component, the usual panel of keratin and p63 immunostains should be used, and if this is strongly and diffusely positive, a diagnosis of spindle cell metaplastic carcinoma is highly likely. However, it is important to note that malignant phyllodes tumors can demonstrate focal, weak keratin and/or p63 or p40 staining in the stromal cells.⁹⁴  GATA3 and SOX10 have been reported to be of value in this differential diagnosis, with approximately half of spindle cell metaplastic carcinomas staining with these antibodies.^95,96  In cases that are not readily categorized, a comment that the lesion is an atypical or malignant spindle cell proliferation with a differential diagnosis that includes a malignant (or perhaps borderline) phyllodes tumor and metaplastic carcinoma is recommended. The final categorization can await excision.

Breast pathology is challenging, with many mimics and potential pitfalls on the pathway to the correct diagnosis. This is particularly so for CNB specimens given the limited tissue available for evaluation. Although IHC is a useful adjunct to help support the morphologic impression and to rule out mimics, with awareness of the pitfalls and careful morphologic evaluation, many cases will not need IHC. Finally, it is important to remember that if the diagnosis on CNB is not straightforward or definitively resolved by IHC, a descriptive diagnosis with a recommendation for excision for further characterization is appropriate.