Fine-needle aspiration generally produces results that are not as good for lobular carcinoma as the results for ductal carcinoma of the breast. In this study, the authors evaluated their team's performance in cytologic diagnosis of lobular carcinoma over 11 years and analyzed the reasons for diagnostic failure.
Cytologic findings were analyzed in 555 consecutive fine-needle cytopuncture specimens from women with palpable, invasive lobular carcinoma of the breast. The authors also examined the influence of the cytologist's experience, the clinical tumor size, the histologic subtype, and the histologic grade on diagnostic performance. All negative samples were re-examined, along with all samples that had been obtained during the last year of the study, to refine the morphologic description of lobular carcinoma.
Malignancy was diagnosed in 68.8% of specimens overall. The individual pathologists diagnosed malignancy in from 44.4% to 81.1% of specimens, depending on their experience. Diagnostic performance was correlated with clinical tumor size, histologic grade, and histologic subtype, and correct diagnoses were significantly more frequent in pleomorphic subtypes than in “classic” types. Re-examination of all 32 negative specimens reduced the false-negative rate from 5.8% to 3.8%.
Fine-needle aspiration (FNA) is a rapid, safe, minimally invasive, and inexpensive way of diagnosing malignancy in women with palpable breast tumors. Coupled with physical examination and imaging, as part of the triple diagnosis, FNA is used by many experienced European and American teams with near-perfect sensitivity.1 However, FNA performs less well on certain types of cancer, including lobular carcinoma of the breast.
Invasive lobular carcinoma, after ductal carcinoma, is the most frequent type of breast cancer and accounts for approximately 5% to 15% of cases.2 Its incidence is increasing, especially because the recognition of new subtypes that previously were categorized as ductal carcinomas.3 Lobular carcinoma not only has different morphologic characteristics than ductal carcinoma, but it also has a different course, different clinical and radiologic features, and probably qualifies for specific therapeutic management. FNA fails more frequently in lobular carcinoma than in other types of breast cancer, with failure rates ranging from 4% to 39.5%, depending on the series.4, 5 This high rate usually is attributed to inadequate sampling or poor cellularity and to difficulties in interpretation because of the presence of only mild atypias and more frequent small cells. For the current study, we retrospectively analyzed a series of 555 fine-needle cytopunctures of invasive lobular carcinomas that were sampled at the Rene Huguenin Cancer Center over an 11-year period to check our team's performance in the diagnosis of lobular carcinoma, to identify correctable sources of failure, and to ensure cytologic description of these tumors.
MATERIALS AND METHODS
At Rene Huguenin Cancer Center, from January 1992 to December 2002, 555 patients with palpable, histologically proven, primary, invasive lobular breast carcinoma were investigated previously by fine-needle cytopuncture without aspiration (FNC) for diagnostic purposes.
The clinical size of the tumors ranged from 5 mm to 120 mm. According to the Tumor, Lymph Node, Metastasis (TNM) classification system, this series included 166 T1 tumors, 258 T2 tumors, 88 T3 tumors, and 43 T4 tumors.
Tissue sections from biopsy or surgical specimens were stained with hematoxylin-eosin-safran. Lobular carcinoma was classified based on World Health Organization criteria as follows: classic, solid, alveolar, tubulolobular, or pleomorphic. The presence of a ductal component (mixed forms) also was recorded. The tumors were graded with the Scarff-Bloom-Richardson (SBR) system, then classified with the modified SBR (MSBR) grading system.6 The MSBR takes into account only the 2 nuclear parameters of the SBR, nuclear pleomorphism and mitoses, and each is scored from 1 to 3: Total tumor scores of 2, 3, and 4 indicate low-grade tumors (MSBR 1), and scores of 5 or 6 indicate high-grade tumors (MSBR 2).
FNC was performed by 8 cytopathologists who interpreted the slides prepared from their own specimens. Samples were taken with a 23-gauge needle, without aspiration, at 3 different tumor sites, as described previously.7 Briefly, the needle was inserted into the lesion, rotated gently, moved back and forth in different directions, and withdrawn as soon as a droplet was observed in the hub. The slides were air dried and stained with May-Gruenwald-Giemsa. When a concurrent axillary lymph node was palpable, it was cytopunctured with the same technique. Since April 1997, at the clinicians' request and when possible, part of the sample has been included in a cytoblock, as described elsewhere for immunohistochemical studies of hormone receptors, Ki67, and HER2 and, more recently, E-cadherin expression, when the cytologic aspect was suggestive of lobular carcinoma.8
The initial cytologic diagnoses were classified in 5 categories: inadequate, benign, atypical, suspicious of malignancy, and malignant. To simplify the tables, we grouped together “suspicious” and “malignant” results.
To identify sources of erroneous interpretation and inadequate sampling, we examined the influence of the cytopathologist's experience, the clinical size of the tumor, the histologic subtype, and the MSBR grade.
All FNC specimens in the “benign” group were reviewed subsequently by consensus by 3 of the cytopathologists (E.M., V.B., and M.B.). Moreover, all malignant specimens that were diagnosed in 2002, which was the last year of the study, were reviewed to ensure the main cytologic features of lobular carcinoma. The SAS statistical package (SAS Institute, Inc., Cary, NC) was used for statistical analyses.
The cytologic results are summarized in Table 1. In 70 samples (12.6%), the material was considered inadequate for cytologic diagnosis. Benignity was diagnosed in 32 samples (5.8%) and corresponded to false-negative results. Seventy-one samples (12.8%) were considered “atypical” either because of poor cellularity or because atypias were too mild and infrequent to show malignancy. Malignancy was diagnosed in 382 samples (68.8%). All samples but 1, which was misdiagnosed as lymphoma, were diagnosed as carcinoma. A diagnosis of lobular carcinoma was suggested in 64 of these samples (16.7%). Malignancy was suspected in only 110 of these 382 samples because of only rare atypia and/or small nuclei.
Table 1. Cytologic Diagnosis of the 555 Lobular Carcinomas
No of Patients.
The value in parentheses indicates the percentage of malignant tumors in which a diagnosis of lobular carcinoma was suggested.
Regarding the cytologic results in 60 patients who were recruited in 2002, there were 42 diagnoses of malignancy (70%), including 34 diagnoses (56.7%) that were certain and 8 diagnoses that were “suspicious.” A diagnosis of lobular carcinoma was suggested in 12 of these patients (28.6%). A lobular type was suggested more frequently in 2002 than in the remaining group (12 vs 30 patients and 52 vs 288 patients, respectively; P = .03).
Among the 101 FNC specimens of axillary lymph nodes, 40 specimens were considered metastatic. The rate of correct cytologic diagnosis of malignancy was increased to 69.4% by these findings, indicating malignancy in 3 specimens in which FNC material from the breast tumor had been “insufficient” (1 specimen) or only “atypical” (2 specimens). Between April 1997 and late 2002, 46 cytoblocks that were prepared from breast tumor FNC specimens and, in a few instances, from a concurrent metastatic lymph node were considered satisfactory for immunohistochemical studies.
Influence of the Cytopathologist's Experience
The performance of the 8 cytopathologists with respect to the diagnosis of malignancy ranged from 44.4% to 81.1%. The most experienced cytopathologist performed 201 cytopunctures, and the least experienced performed 10. The most experienced cytopathologist performed best, as expected.
To avoid a bias linked to the respective percentage of small tumors sampled by each cytopathologist, for each team member, we compared the percentage of “inadequate” results for T1 tumors with the percentage of inadequate results for all tumors. The 2 cytopathologists with the best overall results had very similar percentages of inadequate samples for T1 tumors and for all tumors (4% for 1 and 8% in the other), whereas the other team members had higher percentages of “inadequate” results for T1 tumors than for all tumors.
Influence of Clinical Size
The rate of correct diagnosis rose significantly with the tumor size from 56.6% for T1 tumors to 86% for T4 tumors (P = .0002) (Table 2). In contrast, the frequency of “inadequate” results that were correlated negatively with tumor size fell from 22.3% for T1 tumors, to approximately 9% for T2 and T3 tumors, and to 2.3% for T4 tumors (P = .0001). A similar trend was observed for false-negative results (P = .082; nonsignificant).
Table 2. Cytologic Diagnosis With Tumor Size, Histologic Subtype, and Histologic Nuclear Grade
MSBR indicates the modified Scarff-Bloom-Richardson grading system.
Values in parentheses indicate the number of malignant tumors for which a diagnosis of lobular carcinoma was suggested.
Influence of the Histologic Subtype
Forty-nine lobular carcinomas were pleomorphic (Table 2). The proportion of pleomorphic tumors diagnosed as malignant was 89.8%. Cytopuncture was significantly more accurate for pleomorphic subtype (44 vs 5 tumors) than for other subtypes (338 vs 168 tumors; P = .00091). In contrast, the lobular type was suggested less often for these pleomorphic subtypes than for other subtypes, especially the classic type (6.8% vs 18%).
The other histologic variants, which were less numerous in this series, had little influence on the cytologic results. Ten carcinomas were tubulolobular (6 malignant, 1 atypical, 1 benign, and 2 inadequate), 4 were solid and/or alveolar (3 malignant and 1 inadequate), and 4 were mixed (both lobular and ductal: 2 malignant, 1 atypical, and 1 inadequate). Furthermore, 5 tumors were associated with a ductal carcinoma in situ component, and all were diagnosed as malignant by cytopuncture.
Influence of the Histologic Nuclear Grade (MSBR)
Five hundred forty-nine carcinomas could be graded, and 6 could not. The MSBR grade was 1 in 493 tumors (89.8%) and 2 in 56 tumors (10.2%).
Comparison of the cytologic diagnoses with the MSBR histologic grade (Table 2) indicated that correct cytologic diagnoses were significantly more frequent for high-grade tumors (MSBR 2) than for low-grade tumors (MSBR 1; (49 of 56 tumors vs 327 of 493 tumors, respectively; P = .0012). The percentages of “inadequate,” “benign,” and “atypical” diagnoses clearly were higher for low-grade tumors (13.4%, 6.5%, and 13.8%, respectively) than for high-grade tumors (7.1%, 0%, and 5.4%, respectively). Among the 49 pleomorphic carcinomas, 32 tumors were MSBR grade 2 (65.3%), and 17 were MSBR grade 1 (34.7%)
Review of False-negative Samples
The percentage of false-negative samples initially was 5.8% overall and ranged from 0% to 15% (median, 4.7%), depending on the cytopathologist. A slide review of the 32 false-negative samples confirmed the absence of suspicious or malignant cells in 21 samples and rectified the diagnosis in 11 samples. Among these 11 samples, suspicious cells were observed in 8 samples, 1 sample was considered “suspicious,” and the other 2 samples were reclassified as “malignant.” The pitfalls noted during this slide review included poor cellularity, very fragile samples (often crushed), the presence of only mild atypia, small nuclei, or intermingling with benign-like material.
Reviewing slides from the 42 malignant samples that were diagnosed in 2002, we observed that the background often was thick, eosinophilic, and sometimes had crushed material interspersed with fatty vacuoles and small, fibrous stromal fragments (Fig. 1). The cells were isolated and dispersed or were grouped together in small and poorly cohesive clusters with occasional single-cell alignments (Fig. 2). Intracytoplasmic vacuoles with a targetoid appearance sometimes compressed and indented the nucleus (Fig. 3). The shape of the nuclei was rather characteristic, with irregular angular, triangular, indented, and occasionally budding nuclei (Fig. 4).
In the literature, FNA reportedly diagnoses malignancy correctly in 63% to 94.8% of all breast cancers (median, 84%), depending on the team.9, 10 This high rate is accounted for mainly by ductal carcinoma, which is the most common type of breast carcinoma, whereas sensitivity is lower for some other types, including lobular carcinoma. Few studies have focused on FNA diagnosis of lobular carcinoma.
According to the literature, the cytologic features of the “classic” form is well defined: It has small, monomorphous cells with mild atypia and often has intracytoplasmic vacuoles with targetoid appearance, usually isolated or arranged in single-cell files or in small, loose clusters.11, 12 The diagnosis of lobular carcinoma often is hindered by the absence or marginality of these criteria, especially on poorly abundant material.
The rate of false cytologic diagnosis of benignity in lobular carcinoma varies from 1 team to another between 4% and 32%.4, 13–19 Our rate of 5.8%, therefore, is relatively low. The false-negative results corresponded either to “true cytologic-negative” samples, which contained only benign epithelial cells because of possible missampling; or to the presence of benign cells linked to associated fibrocystic change; or to misinterpretation of tumor cells with rare and mild atypia. All 32 “benign” samples in our series were reviewed by 3 members of the team, who confirmed only 21 samples, reducing the false-negative rate from 5.8% to 3.8%. Other research groups also have reported a reduction in the false-negative rate after a slide review.17, 18
Overall, the percentage of “inadequate” samples was approximately 12.8% in our series. Other authors have reported an over-representation of lobular carcinoma among “inadequate” samples, especially because of poor cellularity and mild cytonuclear atypia.20, 21 In our team, the performance in sampling improved over the years, and the proportion of “inadequate” samples fell to only 11.7% in 2002.
We observed a sensitivity for malignancy of 68.8% overall (382 samples). This rate rose to 69.4% when concurrent FNC of palpable metastatic axillary lymph nodes was taken into account. Our results are similar to those obtained in other series, in which sensitivity for malignancy ranged from 62% to 89%.14, 15 However, we must point out that we took into account only palpable tumors that had been sampled without sonographic guidance, whereas this is not always mentioned in other series.15, 17, 18 The proportion of our unequivocally malignant diagnoses rose to 56.7% in 2002 compared with 48.1% for the remaining previous years. Similarly, regarding malignant cytologic diagnoses, a diagnosis of lobular type was suggested more frequently in 2002 (28.6%) than for the remaining series (15.3%). This improvement probably can be attributed to the cytopathologists' increasing experience.
The rate of correct diagnosis was influenced by the cytopathologist's experience, as expected, and ranged from 44.4% to 81.1%. Lack of experience generally is reported as an important source of false-negative results.22–24 The importance of the cytopathologist's experience was confirmed in our study by the low and very stable rate of “inadequate” results obtained by the 2 most experienced team members, whatever the size of the tumor, whereas their colleagues' accuracy was influenced by tumor size.
However, the performance of cytopuncture depends globally on the clinical size of the tumor. The rate of “inadequate” results in our series was 22.3% for T1 tumors but only 2.3% for T4 tumors. These results are consistent with those published elsewhere.16
Similar to what was reported by Hwang et al,4 in most patients who had variants of lobular carcinoma, we usually obtained sufficient material for an unequivocal diagnosis of carcinoma. Among these different histologic subtypes of lobular carcinoma, the pleomorphic type differs from the classic type. Pleomorphic lobular carcinoma was recognized first approximately 15 years ago. On the basis of histopathologic criteria, pleomorphic lobular carcinoma is defined as an invasive carcinoma with the architecture of invasive lobular carcinoma but with very particular cytologic characteristics (marked atypia, high mitotic activity, and high nuclear grade). It is important to identify this form correctly, because it is more aggressive and carries a worse prognosis than the classic type.25, 26 Diagnosis of malignancy is easier in pleomorphic lobular carcinoma because of the presence of more marked atypia and because cellularity generally is higher in cytopuncture samples, as reflected in our series by the far lower percentage of “inadequate” results (6.1%) relative to classic forms (13.2%). Other authors also have reported higher rates of correct diagnosis of malignancy with this subtype of lobular carcinoma.4, 11 In contrast, it is more difficult to recognize the lobular type of these tumors, because the cytologic aspect is more suggestive of poorly differentiated ductal carcinoma.11 This pitfall emerged clearly in our series, because a diagnosis of lobular carcinoma was raised less frequently for the pleomorphic type (6.8%) than for “classic” forms (18%).
Nearly 90% of the lobular carcinomas in our series had low nuclear grade, and only 10.2% were high-grade tumors. Carcinomas with high nuclear grade corresponded mainly to variants of lobular carcinoma and especially to the pleomorphic type4; the latter represented 57% of high-grade tumors in our series, and >65% of the pleomorphic lobular carcinomas were MSBR grade 2. Furthermore, low-grade carcinoma appeared to be over-represented among cytologically “benign” and “inadequate” samples, both in our series (94% and 100%, respectively) and in the literature.4
Overall, our results confirm that the cytologic diagnosis of malignancy is more difficult in lobular carcinoma than in ductal carcinoma, even in the hands of experienced cytopathologists. Unfortunately, in the case of lobular carcinoma, physical examination and imaging studies also are less accurate than for most other carcinomas.19
In addition to diagnosing malignancy, it is important to attempt to identify the lobular type of these carcinomas. In our global series, lobular carcinoma was suggested in 16.7% of “malignant” samples based solely on the morphologic features. However, cytologic characteristics are not fully specific for lobular carcinoma, because they sometimes are observed (especially small nuclei) in other types of breast carcinoma, such as ductal carcinoma.27 Thus, some typical features, such as intracytoplasmic vacuoles with a targetoid appearance, in fact may correspond histologically to ductal carcinoma (Fig. 5); in our experience, this confusion arises in nearly 20% of samples (data not shown). The lobular type of breast carcinomas can be identified more accurately by using cytoblocks when the samples are sufficiently cellular, particularly with anti-E-cadherin immunohistochemistry.8 When the tumor cells are negative for E-cadherin (Fig. 6), then a diagnosis of lobular carcinoma can be made with confidence. This not only improves the sensitivity of cytologic diagnosis in some patients but also opens the way to more appropriate patient management. Indeed, lobular carcinoma differs from ductal carcinoma by its higher frequency of bilaterality, its clinical course, and its poor chemosensitivity (especially to neoadjuvant chemotherapy) and probably requires a different therapeutic strategy.28–30