C M Quinn, Consultant Histopathologist, Irish National Breast Screening Programme, St Vincent’s University Hospital, Elm Park, Dublin 4, Ireland. e-mail: email@example.com
Columnar cell lesions of the breast comprise a group of conditions characterized by dilation of terminal duct lobular units lined by columnar epithelial cells, ranging from one or two layers of benign epithelium to stratified epithelium with atypia. Although these lesions have been recognized by pathologists for many years, they have recently assumed a new significance due to their increased detection as mammographic calcification, the potential for overdiagnosis as ductal carcinoma in situ (DCIS) and their possible relationship to invasive carcinoma. This short overview aims to outline the pathological features and likely clinical significance of these lesions, with emphasis on diagnostic criteria, terminology and classification, and management strategies.
columnar cell alteration with prominent apical snouts and secretions
columnar cell change
columnar cell hyperplasia
columnar cell lesion
cystic hypersecretory hyperplasia
ductal carcinoma in situ
enlarged lobular units with columnar alteration
flat epithelial atypia
needle core biopsy
terminal duct lobular unit
The term ‘blunt duct adenosis’ was introduced in 1945 by Foote and Stewart to describe a condition affecting the terminal duct lobular units (TDLU) characterized by hypertrophy of the epithelial and myoepithelial cell layers with enlargement of lumina.1 The enlarged acini assumed blunt lateral outlines and were associated with a specialized type of stroma akin to that seen around ducts. The term ‘columnar metaplasia’ was used by Bonser, Dossett and Jull in 1961 to describe the same process.2 This term was never taken into common usage, but had the merit of recognizing that epithelial hypertrophy was the main feature of the condition and also implied that the alteration took place in a pre-existing, normal structure. Other terms used to describe this condition include ‘atypical cystic lobules’,3‘columnar alteration of lobules’,4‘metaplasia cylindrique’5 and ‘hyperplastic terminal groupings’. In the 1970’s, Azzopardi coined the term ‘clinging carcinoma’ to describe a form of ductal carcinoma in situ (DCIS) characterized by neoplastic cells confined to the periphery of the affected structures.6 He emphasized that there were two types of clinging carcinoma, one characterized by the presence of pleomorphic epithelial cells and representing an early stage of so-called comedo DCIS. The second variant of clinging carcinoma tended to affect TDLUs and was characterized by dilated acini lined by one or a few layers of neoplastic epithelial cells showing mild cytological atypia and containing inspissated densely staining material sometimes with stippled or other types of calcification. The luminal cells frequently displayed a form of blebbing possibly representing ‘apocrine snouts’. This form of clinging carcinoma was considered to be difficult to diagnose and required examination of breast biopsy tissue at increased scanning magnification.
Clinging carcinoma, as described by Azzopardi, bears close similarity to the lesion termed columnar alteration with prominent apical snouts and secretions (CAPSS) reported by Fraser et al. in 1998.7 The term CAPSS encompassed a spectrum of lesions affecting the TDLU, observed in breast biopsies performed for evaluation of microcalcification. Typically, the lesion was characterized by dilated acini lined by columnar epithelial cells with prominent apical cytoplasmic snouts, luminal secretions and varying degrees of nuclear atypia and architectural complexity. Low-grade DCIS was also present in some of the biopsy specimens, more commonly in association with CAPSS lesions showing atypia. Kasami et al. used the term enlarged lobular units with columnar alteration (ELUCA) to describe enlarged TDLUs with columnar cell alteration.8 Within the ELUCAs they observed a process that they termed ‘hypersecretory hyperplasia’ characterized by bubbly cytoplasm, nuclear protrusions and intracytoplasmic and extracellular secretions. The condition was analogous to exaggerated lactational change and in some cases showed cytological atypia of the luminal epithelial cells and accompanying atypical ductal hyperplasia (ADH) and DCIS. The authors noted the similarity with CAPSS but considered that this was a separate condition. The term ‘flat epithelial atypia’ (FEA) was introduced by the World Health Organization to replace ‘clinging carcinoma, monomorphous type’, ‘atypical cystic lobules’ and other terms, e.g. ‘atypical lobules type A’, ‘atypical columnar change’ and ‘ductal intraepithelial neoplasia 1A’.9
Classification and pathological features
It is clear from the above that these various terms apply to a process characterized by enlarged TDLUs with dilated acini lined by epithelial cells showing columnar cell morphology. Luminal secretions and associated calcification are common. In some cases there is accompanying atypia, predominantly cytological.
The plethora of terminology and pathological descriptions has been synthesized and simplified into a practical classification system by Schnitt and Vincent-Salomon.10 The various changes are collectively known as ‘columnar cell lesions’ (CCLs) and divided into two broad categories: columnar cell change (CCC) and columnar cell hyperplasia (CCH), each further subclassified according to the absence or presence of cytological atypia.
CCC is characterized by TDLUs that show varying degrees of acinar dilation. The acini have an undulating outline and are lined by one or two layers of columnar epithelial cells with uniform, ovoid nuclei oriented perpendicular to the basement membrane with evenly dispersed chromatin, inconspicuous nucleoli and infrequent mitotic activity. The luminal cells may show apical snouts, and secretions may be seen in the lumina of dilated acini with or without associated calcification, corresponding to the CAPSS lesion described by Fraser.7
In CCH, the acini are lined by more than two layers of columnar type epithelial cells with similar cytological features to CCC. The proliferating cells may form small micropapillae. Apical snouts and luminal secretions with calcification, sometimes resembling psammoma bodies, are common.
Columnar cell lesions with cytological atypia
Flat epithelial atypia
The lesions of CCC and CCH may display cytological atypia of the epithelial lining cells termed CCC with atypia (Figure 3) and CCH with atypia (Figure 4), respectively, collectively known as FEA. At low magnification, TDLUs affected by FEA may appear dark due to the increased nuclear–cytoplasmic ratio of the lining cells. The dilated spaces tend to have a rigid contour and the intralobular stroma may contain a lymphoid infiltrate.11 In FEA the nuclei of the lining cells are typically round rather than ovoid, are not oriented perpendicular to the basement membrane and may have visible nucleoli and occasional mitotic figures. In some cases the cells lining the dilated acini may resemble the cells of tubular carcinoma. The group of lesions categorized as CCLs with cytological atypia encompass lesions previously described as low-grade clinging carcinoma. The authors of the proposed classification system have emphasized that high-grade cytological atypia is not a feature of CCLs and should be diagnosed as high-grade DCIS, even in single-layered lesions.
Columnar cell lesions with architectural atypia
The classification system described by Schnitt and Vincent-Salomon recognizes that some CCL-type lesions show complex architectural atypia, including well-developed micropapillae, arch and bridge formation, and early cribriform features. The authors advocate that these lesions are better categorized as ADH or DCIS depending on the severity of the findings, a view recently reiterated by Pinder et al.11 This appears to be a practical and sensible approach, in view of the already considerable interobserver variation in the diagnosis of ADH versus DCIS,12 which can only be exacerbated by the introduction of CCL with architectural atypia, another category of architecturally atypical ductal intraepithelial proliferations.
Saqi et al. have retrospectively reviewed fine-needle aspirates (FNAs) from 20 patients who subsequently had a histological diagnosis of CCL without ADH, DCIS or invasive carcinoma.13 Characteristic FNA findings included the presence of cohesive three-dimensional clusters of polygonal epithelial cells intermixed with myoepithelial cells. The cells in the centre of these clusters tended to display crowding, overlapping and some loss of polarity. Apical snouts were seen in some cases. Enlarged columnar cells, similar to those seen on histology, lined the periphery of the clusters with a palisaded effect. The presence of scattered, single, columnar cells with bland nuclear morphology was also a common finding. Eighteen of the 20 FNAs were interpreted as atypical, with only five showing atypia on subsequent excisional biopsy. On review, no notable differences were detected in the cytological appearances of FNAs obtained from CCLs with or without histological atypia. As such, FNA is not likely to prove useful in identifying patients with atypical CCLs in need of further diagnostic work-up.
In a similar review of FNAs from 10 patients with biopsy-proven CCLs, Jensen and Kong observed flat sheets of cells with features reminiscent of apocrine metaplasia to be the most common pattern.14 Typically, the cells had round nuclei and distinct cell borders, but with moderate amounts of finely granular cytoplasm, in contrast to the abundant coarsely granular cytoplasm typical of apocrine cells. Myoepithelial cells were inconspicuous. Cytological atypia, ranging from minimal to severe, was reported in seven of the 10 patients. However, only one patient, in whom malignancy had been erroneously diagnosed on FNA, had atypia on excisional biopsy. These authors observed that the FNA appearances of CCLs show considerable overlap with well-differentiated breast carcinoma and emphasized the potential for overdiagnosis of malignancy on FNA.
The differential diagnosis of CCLs without atypia includes apocrine metaplasia, pregnancy-like (pseudolactational) hyperplasia (PLH) and cystic hypersecretory hyperplasia (CHH).
In common with CCLs, apocrine metaplasia typically involves cystically dilated acini and frequently displays apical snouting. The latter is less prominent in apocrine metaplasia, which is also distinguished from CCLs by the abundant granular eosinophilic cytoplasm, absence of luminal secretions and calcification and negative oestrogen receptor (ER) status.15
PLH is also characterized by dilated TDLUs with cytoplasmic blebbing, reminiscent of apical snouts. Discriminating features include the type of luminal secretions, the character of the lining epithelial cells, which are enlarged with abundant pale-clear, vacuolated or finely granular cytoplasm and the characteristic laminated calcification.16
CHH comprises dilated cystic acini lined by flat columnar epithelial cells, with abundant luminal homogeneous colloid-like material. Calcification is infrequent.16,17
The differential diagnosis of CCLs with atypia includes ADH18 (Figure 5) and low-grade DCIS. As outlined above, we agree that CCLs with cytological and complex architectural atypia should be categorized as ADH or DCIS, depending on the severity and extent of change, as proposed by Schnitt and Vincent-Salomon.10
Evidence is gathering that CCLs may be biologically significant, possibly representing a very early stage in the evolution of low-grade DCIS and invasive carcinoma. This information comes from observational studies reporting the coexistence of CCLs with more advanced lesions, a small number of longitudinal patient follow-up studies and immunohistochemical and molecular genetic profile studies.
Observational studies demonstrating coexistence of CCLs with more advanced lesions
CCL-type lesions have been observed in association with lobular neoplasia (LN), DCIS and invasive carcinoma.
In 1995 Weidner reported personal observation of a lesion characterized by ectatic ducts lined by a monotonous population of small hyperchromatic cells, some with apocrine snouts, and noted their frequent association with, and cytological similarity to, invasive tubular carcinoma.19 Rosen has also reported the frequent occurrence of this type of lesion in the breast tissue adjacent to tubular carcinoma and for many years has used the term pretubular hyperplasia to describe this change.20,21
In a comprehensive review of 253 breast lesions (32 tubular carcinoma, 121 invasive ductal carcinoma, 40 invasive lobular carcinoma, 20 low-grade DCIS, 40 cystic change), Goldstein and O’Malley have identified ‘small ectatic ducts lined by atypical cells with apocrine snouts’ in 17 of the lesions—14 tubular carcinomas and three grade 1 invasive ductal carcinomas.22 In all but one of the 17 cases, low-grade DCIS was also present. As in other studies, the authors observed that the cells lining the ectatic ducts were morphologically identical to those of the coexistent invasive carcinoma and interpreted the findings as a form of lobular cancerization by low-grade DCIS.
Fraser et al., in a review of 100 consecutive breast biopsies performed for microcalcification, identified the lesion they termed CAPSS in 42 of the biopsy specimens.7 The frequency of LN, ADH and DCIS was similar in specimens with and without CAPSS, but there was a higher incidence of low-grade DCIS in those showing CAPSS with atypia. CAPSS and DCIS tended to involve the same or adjacent TDLUs.
Oyama et al. observed ‘atypical cystic lobules’ (ACLs), characterized by proliferation of luminal epithelial cells showing low-grade cytological atypia without architectural atypia, in association with 16 of 44 (36%) low-grade DCIS lesions and nine of 31 (29%) LN lesions.23 They noted striking cytological and immunohistochemical similarities between this entity and low-grade DCIS, a similar pattern of calcification and frequent merging of the two lesions. They also reported the presence of ACLs in otherwise benign biopsy specimens and have postulated that this entity represents a forerunner of DCIS rather than its advancing edge, acknowledging its similarity to clinging carcinoma as described by Azzopardi.6
In a detailed morphological study of putative precursor lesions and low-grade invasive carcinomas, Abdel-Fatah et al. have documented the presence of CCLs in 53 of 56 (95%) tubular carcinomas, in 12 of 14 (86%) tubulo-lobular carcinomas and in 34 of 57 (60%) invasive lobular carcinomas, with striking similarity of morphological features and geographical location.24 The majority of CCLs showed flat epithelial atypia, suggesting that the acquisition of nuclear atypia may be a requirement for progression to invasive carcinoma. LN was also present in 16% of tubular carcinomas, an association observed by other authors.21,25,26 In this study, LN with tubular carcinoma always occurred in association with CCLs and ADH/DCIS, leading the authors to speculate on a shared developmental lineage and a possible interactive precursor pathway in the development of this type of invasive carcinoma. LN was present and co-localized in 91% of invasive lobular carcinomas, with coexistent CCLs in 60%, suggesting a possible relationship between LN and CCLs in the genesis of invasive lobular carcinoma also. The coexistence of FEA and LN in 87% of a series of 111 breast biopsy specimens that did not contain DCIS or invasive carcinoma lends further support to the theory of a biological relationship between CCLs and LN.27 The observation of focal loss of E-cadherin immunoreactivity in CCLs is interesting in this regard.28
Longitudinal patient follow-up studies
Information on the likely natural history of CCL-type lesions is scarce. An isolated case study has reported tubular carcinoma in a woman who had columnar cell alteration on needle core biopsy from the same site 6 years previously.25 In a retrospective histological review of 4371 breast biopsy specimens previously reported as benign, Eusebi et al. have identified 21 cases of clinging carcinoma as described by Azzopardi and obtained long-term (16.7 years) patient follow-up.29 Two of the 21 patients developed invasive carcinoma and subsequently died from their disease. In the same study, seven patients with previously undiagnosed conventional DCIS were also identified, one of whom later developed invasive breast carcinoma, and the authors have suggested that patients with clinging carcinoma are at similar risk of developing invasive carcinoma as those with conventional DCIS. However, in this study the authors did not subdivide clinging carcinoma into the high- and low-grade (CCL-like) subtypes as originally described by Azzopardi.
In a subsequent publication, Eusebi and colleagues have reported long-term follow-up of 80 patients with previously undiagnosed DCIS, including the 28 patients in the above study.30 Forty-one cases of pure clinging carcinoma were identified, and in this study were subdivided into those with pleomorphic (P) (n = 9) and monomorphic (M) nuclei (n = 32). Two patients with M nuclei developed recurrent DCIS and were successfully treated by conservation surgery, with no evidence of recurrent disease 19 and 26 years after first biopsy, respectively. Three patients with P nuclei developed invasive carcinoma, 6, 6 and 10.5 years following biopsy and subsequently died from their disease. In the period of follow-up (17.5 years), no patient with pure clinging carcinoma of M nuclear type developed invasive breast carcinoma.
In a comprehensive review of the histopathological characteristics of 863 of the 1010 patients with DCIS randomized to surgical excision with or without radiotherapy in the European Organization for Research and Treatment of Cancer trial 10853, Bikjer et al.31 categorised DCIS into three subgroups in accordance with the Holland classification system.32 These authors found that DCIS recurrence was significantly related to histological grade, being least common in the well-differentiated/low-grade subgroup. Within this latter subgroup, significantly different DCIS recurrence rates were observed in lesions with clinging, micropapillary and cribriform architectural patterns (0, 7 and 18%, respectively). In view of the fact that none of the 59 patients with low-grade clinging DCIS developed recurrent DCIS, the authors have suggested that this lesion, which others would categorize as ‘columnar alteration with prominent apical snouts and secretion’, be recognized and patients offered local excision and close follow-up only.
Thus, the available evidence is very limited, but suggests that CCL-type lesions are biologically low grade with a low potential for either recurrence or progression to invasive carcinoma.
Immunohistochemical and molecular genetic profile studies
CCLs typically show immunopositivity for with antibodies to ER, PR, Bcl-2 and luminal cytokeratin (CK) 19.10,23,28,33–36 Cyclin D1 positivity has been documented.37 Reports on Ki67 positivity are variable, with some groups reporting little or no reactivity,33,38 whereas others have demonstrated increased levels of positivity compared with normal epithelium.28,39 The lesions are usually negative for Her-2/neu, p53 and basal CK5/6 and CK14.34,40 This immunohistochemical profile resembles that seen in ADH and low-grade DCIS and contrasts with that seen in benign epithelial hyperplasia, delineating CCL as a clonal proliferation and possibly a precursor lesion of low-grade malignancy.
This thesis is supported by the elegant studies of Simpson et al., who have demonstrated a pattern of chromosomal change in CCLs analogous to that seen in low-grade DCIS and well-differentiated invasive carcinoma.34 Using comparative genomic hybridization, a range of recurrent chromosomal abnormalities was identified, including loss on 16q, 17p and X and gain on 15q, 16p and 19. In addition to the four categories of CCL in the Schnitt and Vincent-Salomon classification system, Simpson et al. have studied CCLs with architectural atypia and those with both cytological and architectural atypia and have documented increasing genetic instability with increasing morphological abnormality. In a detailed study of three cases, each displaying the entire spectrum of CCLs through to DCIS and including foci of invasive carcinoma, Dabbs et al. have similarly observed progressive accumulation of allelic damage with increasing severity of lesion, targeted at 9q, 10q, 17p and 17q.41 Other authors have reported the same pattern of genetic alteration in flat epithelial atypia as in the adjacent in situ and invasive carcinoma, compared with the normal breast epithelium, in which no genetic abnormality was detected.42
CCL on needle core biopsy
The optimum management of patients with CCLs identified on needle core biopsy (NCB) is unknown. In an attempt to answer this question, Guerra-Wallace et al. have evaluated the incidence of malignancy on initial NCB or in subsequent excision biopsies in 195 patients after NCB diagnosis of lesions described as CAPSS.43 There was a higher incidence of coexistent malignancy in patients who had CAPSS with atypia [11/60 (18%)] than in those with CAPSS without atypia [10/135 (7%)]. Furthermore, in over one-third of patients who had CAPSS with atypia on NCB and coexistent malignancy, the malignancy was diagnosed only on subsequent excision biopsy, performed in 37 patients.
Brogi et al. have correlated the finding of CCC on NCB with the subsequent excision biopsy pathology.44 Based on architectural features, 162 cases of CCC on NCB were classified as mild, moderate or severe. Twenty-three patients proceeded to excision biopsy, seven of whom had a diagnosis of DCIS and/or invasive ductal carcinoma (30%). In all seven patients the CCC had been classified as moderate or severe.
In a study designed to identify factors that predict the underdiagnosis of malignancy in patients with atypical lesions on NCB, Bonnet et al. have reviewed 167 patients, included nine who had a CCL with atypia.45 Excision biopsy revealed DCIS in two of the nine patients (one high-grade and one low-grade micropapillary) and ADH in three patients. In a study of 24 patients, Lim et al. have reported upgrading to DCIS on excision specimen in 40% of patients with ADH on NCB, compared with 20% in those with an atypical CCL on NCB.46 Further similar studies have also reported the presence of a more significant lesion on excision biopsy in approximately 20% of patients who have atypical CCL/FEA on NCB.38,47–49
In general, these studies are limited by small numbers, and many patients with a diagnosis of CCL without atypia on NCB did not proceed to surgical excision biopsy. There is therefore no real control group for comparison with atypical CCL-associated pathology. Despite this, it is generally recommended that patients with a diagnosis of CCL without atypia on NCB require no further intervention. In view of the finding of a more significant lesion on excision in 20% of patients with atypical CCL/FEA on NCB, these patients should be advised to undergo surgical excision biopsy for full histological evaluation of the mammographic abnormality.50
CCL in excision breast specimens
The success of breast-conserving therapy in the management of DCIS is dependent on the disease extent and particularly on the adequate excision of the lesion with a clear margin of surrounding breast tissue. It has been suggested that clinging carcinoma, a lesion now considered to be equivalent to CCL with atypia (FEA), may account for some of the 20% of DCIS local recurrences that occur despite so-called negative margins.42 Some recommend that clinging carcinoma be regarded as conferring a risk equivalent to that of ADH based on published rates of local recurrence and development of invasive breast cancer in association with this lesion.51 The lack of consensus, and the need for guidelines in managing these lesions, was highlighted by a recent survey, which found that 21% of pathologists would recommend re-excision where multiple ducts showing FEA were identified at a margin in a patient with coexisting low-grade DCIS, the latter > 5 mm from the margin.52 Although the evidence is limited, atypical CCLs appear to be associated with a very low risk of progression to invasive carcinoma, and the prevailing view is that these lesions should not be taken into consideration in determining the extent of DCIS or in evaluating margin status in therapeutic excision specimens.53 Similarly, it is recommended that patients with an atypical CCL in a diagnostic excision specimen without evidence of established DCIS are best treated by clinical and mammographic surveillance.
In the past, CCLs were usually observed as an incidental background finding in breast tissue removed for evaluation of a palpable lump and did not pose diagnostic difficulty or a management dilemma. With the widespread use of mammographic screening, CCLs are now encountered as a primary pathological entity due to their tendency to calcify.54 The biological significance of these lesions is uncertain, but the cumulative evidence from observational studies demonstrating coexistence with more advanced lesions, as well as immunohistochemical and molecular genetic studies, suggests that CCLs represent a very early form of malignant change. CCLs consistently demonstrate ER, PR and Bcl-2 positivity, but Her-2/neu and p53 negativity, and frequently show 16q loss. This profile suggests that CCLs are the forerunner lesion of low-grade DCIS and well-differentiated invasive carcinoma.28,55 Despite this, it appears that the actual risk of progression to invasive carcinoma is low, albeit from the limited information on clinical outcome in patients with pure CCLs. Our lack of knowledge of the true significance of this lesion is exacerbated by the use of varied terminology. The classification system proposed by Schnitt and Vincent-Salomon represents a significant advance in unification of terminology. Early studies suggest that the distinction of atypical CCLs from those without atypia is highly reproducible with the use of available diagnostic criteria56, although recognition of atypia requires scrutiny of CCLs at medium to high magnification.57 Appreciation of CCLs as a cause of mammographic calcification is important in the evaluation of screen-detected pathology. The management guidelines presented by Schnitt and colleagues, as outlined above, reflect the available clinical and scientific evidence. The use of standardized terminology, unified diagnostic criteria and multicentre collaboration in the follow-up of patients with CCLs would greatly enhance our understanding of this interesting and challenging entity.