Classic lobular neoplasia (LN) of the breast includes atypical lobular hyperplasia (ALH) and lobular carcinoma in situ with classic morphology (LCIS). Recently recognized lobular proliferations with unusual morphologic features, such as pleomorphic LCIS and LCIS with massive acinar distention, necrosis, and coarse calcifications are not part of classic LN.1 LN is associated with a high risk of subsequent invasive carcinoma in either breast, with higher risk for LCIS2 than for ALH.3 Although recent data indicate that LCIS represents a nonobligate precursor of invasive lobular carcinoma,4 LN continues to be managed clinically as a high-risk lesion.5
In theory, the presence of LN in tissue obtained at core-needle biopsy (CNB) could represent an incidental finding. It is possible to hypothesize that surgical excision is only warranted if the pathologic findings do not account for the imaging target. Currently, however, there is an open debate on the need to excise a breast target that yields LN as the highest risk lesion at CNB, because some series report significant rates of upgrade.6-9 Most studies addressing this issue are small and retrospective, with potential selection bias and limited information on imaging histologic correlation.
Since June 2004 at our center, all breast lesions that yield a finding of LN at CNB are referred for surgical excisional biopsy (EXB), regardless of whether the histologic findings provide a sufficient explanation for the imaging features and independent of other high-risk histology. We reviewed the surgical findings in these lesions to determine the rate of upgrade to carcinoma in patients with LN at CNB.
MATERIALS AND METHODS
Case Selection and Study Design
The study was approved by the Institutional Review Board of the Memorial Sloan-Kettering Cancer Center. We excluded all CNBs with a diagnosis of ductal carcinoma in situ (DCIS) and/or invasive carcinoma, atypical ductal hyperplasia, flat epithelial atypia, papilloma, radial scar/sclerosing lesions, extravasated mucin, and LCIS variants.
A search of our pathology database identified 85 consecutive CNBs performed at our center between June 2004 and May 2009 that had a diagnosis of LN as the highest risk lesion mandating excision. Prospective excision was recommended for 85 targets, but 5 lesions (6%) in 5 patients did not undergo EXB. Four patients, each with 1 lesion, were immediately lost to follow-up. The fifth patient received treatment for contralateral invasive cancer; her index lesion (mammographic calcification) was stable on imaging at 30 months of follow-up. Clinical and follow-up information was obtained through the e-medical records. In total, 80 consecutive lesions that had LN at CNB in patients who underwent subsequent EXB constituted the basis of this study.
Imaging Targets and Biopsy Technique
Each imaging target was classified according to the Breast Imaging Reporting and Data System (BI-RADS) lexicon as Category 4 (suspicious lesion, recommend biopsy) or Category 5 (highly suggestive of malignancy).10 The breast biopsy technique was tailored for the imaging target.
For calcifications and for mammographically evident, sonographically occult masses, patients underwent biopsy with stereotactic guidance (MultiCare Platinum prone stereotactic biopsy table; Hologic, Indianapolis Ind) with an 11-gauge vacuum-assisted biopsy (VAB) probe (Mammotome Breast Biopsy System; Ethicon Endo-Surgery, Cincinnati, Ohio), obtaining a median of 14 cores per lesion (range, 5-31 cores per lesion). For sonographically evident masses, patients underwent biopsy with ultrasound guidance (Pro-Mag 2.2′ [Manan Medical Products, Northbrook, Ill]; Magnum [Bard Biopsy Systems, Tempe, Ariz]; Achieve [Carefusion, San Diego, Calif]) with a 14-gauge automated core biopsy technique, obtaining a median of 5 cores per lesion (range, 4-6 cores per lesion). For MRI enhancement, patients underwent biopsy with MRI guidance (Signa; General Electric Medical Systems, Milwaukee, Wis) with a 9-gauge VAB probe (ATEC Biopsy Probe; Hologic), obtaining a median of 10 cores per lesion (range, 8-15 cores per lesion).
Imaging-Histologic Correlation: Assessing Concordance and Discordance
In all cases, the radiologist who performed the CNB correlated the images obtained before, during, and after the biopsy with the pathology report. The radiologist issued an addendum to the radiology report documenting the histologic findings and correlating the imaging findings with pathology. A case was classified concordant if the histologic findings provided a sufficient explanation for the imaging features and discordant otherwise, using previously published criteria, as discussed below.11, 12
At stereotactic and ultrasound (US)-guided biopsy, discordant lesions included: 1) calcific lesions in which no calcifications were identified on specimen radiography; 2) lesions that were suspicious for malignancy (ie, BI-RADS Category 4) and in which the histologic findings did not account for the imaging pattern (eg, a discrete mass lesion that yielded only benign breast tissue); or 3) lesions that were highly suggestive of malignancy (ie, BI-RADS Category 5) that were sampled with benign results. Examples of discordant lesions at MRI-guided biopsy include a BI-RADS Category 5 lesion that yielded benign results or BI-RADS Category 4 lesions in which the histology and imaging findings did not match based on published or institutional experience of the MRI patterns of different histologic entities.12 The impression of discordance at MRI-guided biopsy was based on correlating the pathology findings with the images obtained on the day of the biopsy (both before and after tissue acquisition), interpreted in conjunction with MRI performed before the biopsy date.12
In all cases, the radiologist who performed the biopsy commented on the concordance/discordance between the pathology findings and the imaging features and made specific management recommendations, documenting to whom they were communicated. For the purpose of the study, all pertinent images in lesions that yielded an LN result were reviewed by a dedicated breast radiologist (L.L.) to confirm the classification of concordance or discordance.
Slide Review and Immunohistochemical Studies
All CNB material was submitted in its entirety and routinely processed to obtain at least three 4-micron-thick, hematoxylin and eosin-stained, level sections from each tissue block. A monotonous proliferation of nonpolarized cells with low nuclear grade atypia filling and distending >50% of the acini of a lobule was classified as LCIS.2, 13 A less developed lesion with similar cytology was diagnosed as ALH.2, 13 A diagnostic “upgrade” consisted of either DCIS and/or invasive mammary carcinoma in the EXB specimen.11
All available slides of CNB and EXB specimens were reviewed by at least 2 pathologists (M.P.M. and C.L.) to confirm the morphologic diagnosis. A third pathologist (E.B.) reviewed selected representative slides of all cases. We performed immunoperoxidase studies for E-cadherin (MU390-UC antibody; Biogenex, San Ramon, Calif) and p120 (420M-15 antibody; Cell Marque, Rocklin, Calif) on all concordant cases with diagnostic upgrade.
Statistical analysis was performed with the Fisher exact test. The 95% confidence intervals on percentages were calculated with Geigy Scientific Tables.14
Findings at Core-Needle Biopsy in 80 Lesions With Prospective Excision
The 80 index lesions occurred in 77 women with a mean age of 54 years (range, 31-74 years). Imaging targets for the 80 index lesions were calcifications in 48 lesions (60%), MRI enhancement in 17 lesions (21%), and mass in 15 lesions (19%). The median size of the imaging target was 0.8 cm (range, 0.5-4.2 cm). CNB histology yielded LCIS in 46 of 80 lesions (58%) and ALH in 34 of 80 lesions (42%). CNB histology was concordant with imaging findings in 72 of 80 lesions (90%) and discordant in 8 of 80 lesions (10%).
Lesions With Imaging-Histologic Concordance
The 72 lesions with imaging-histologic concordance occurred in 69 women with a mean age of 53 years (range, 31-73 years), including 3 women who had 2 imaging targets each. Twenty-six of these 69 women (38%) had a personal history of breast cancer (25 contralateral and 1 ipsilateral), consisting of invasive carcinoma in 21 women (17 invasive ductal, 3 invasive lobular, and 1 invasive with mixed ductal and lobular features) and DCIS in 5 women.
Calcifications were the imaging target in 46 cases (64%), mass was the target in 9 cases (12%), and MRI enhancement was the target in 17 cases (24%). CNB histology yielded LCIS in 42 of 72 lesions (58%) and ALH in 30 of 72 lesions (42%). Microscopic calcifications were identified at CNB in all 46 lesions and were evident as mammographic calcifications, and small focal calcifications were associated with LN (9 LCIS and 4 ALH) in 13 of 46 calcific lesions (28%). CNB histologic findings in 17 lesions, evident as MRI enhancement, included stromal fibrosis, sclerosing adenosis, fibrocystic changes, and pseudoangiomatous stromal hyperplasia. The histologic findings at CNB of 9 mass lesions included fibroadenoma, ectatic ducts, fibroadenomatoid, and fibrocystic changes. The mean interval from CNB to EXB was 4 weeks (range, from 4 days to 22 weeks).
EXB yielded carcinoma in 2 of 72 imaging-histologic concordant lesions (3%; 95% confidence interval, 0%-9%) (Table 1). In both lesions, stereotactic, 11-gauge VAB of calcifications yielded focal ALH and minute calcifications in benign epithelium. The morphologic interpretation of ALH was confirmed by the results from E-cadherin and p120 immunoperoxidase staining in both the CNB specimen and the EXB specimen.
Table 1. Imaging and Histologic Findings in 2 Upgraded Cases
|Ca++||ALH; Ca++ in benign glands||Yes||High (strong family history)||DCIS LG, 2 mma; Ca++ in DCIS and benign glands|
|Ca++||ALH; Ca++ in benign glands||Yes||No known risk factors||IDC MSBR grade I/III, 2 mma; DCIS LG, Ca++ in benign glands|
Upgrades occurred in 2 patients (ages 56 years and 60 years) who had no personal history of breast carcinoma or LN. One patient had a strong family history of breast carcinoma and a personal history of serous borderline tumor of the ovary. The other patient reported no family history of breast cancer. The time interval from CNB to EXB for these 2 patients was 6 weeks and 9 weeks, respectively.
Surgical excision of both lesions that were upgraded yielded focal, low-grade carcinoma. In one lesion, the upgrade was a 0.2-cm focus of low-grade DCIS; calcifications were present in DCIS and benign epithelium. The other lesion that was upgraded consisted of a 0.2-cm invasive ductal carcinoma (IDC) (Modified Scarff-Bloom-Richardson grade 1 + 1+1 = 3) and focal, low-grade DCIS; calcifications were observed in benign epithelium. In both lesions, biopsy site changes were present in a tissue section that did not reveal DCIS/IDC, raising the possibility that the carcinoma may not have been the imaging target.
Lesions With Imaging-Histologic Discordance
Eight lesions that yielded a finding of LN at CNB had discordant imaging-histologic findings. These occurred in 8 patients with a mean age of 56 years (range, 43-74 years). Six patients (75%) had a prior history of breast cancer, which was invasive carcinoma in all patients (4 contralateral, 1 ipsilateral, and 1 bilateral).
Imaging targets were calcifications in 2 patients (25%), mass in 5 patients (63%), and calcified mass in 1 patients (12%). The BI-RADS category was 5 (highly suggestive of malignancy) in 1 lesion and 4 (suspicious, recommend biopsy) in 7 lesions. CNB failed to sample the target calcifications in 2 lesions. In 6 mass lesions (including 1 with calcifications), the CNB histology was not considered sufficient to explain the imaging features. CNB yielded LCIS in 4 lesions (50%) and ALH in 4 lesions (50%). The mean interval from CNB to EXB was 4 weeks (range, 3-7 weeks).
Carcinoma was present at EXB in 3 of 8 discordant lesions (38%; 95% confidence interval, 9%-76%) and consisted of DCIS in 2 lesions and IDC in 1 lesion. Imaging of a lesion that was upgraded to DCIS had mammographic calcifications, but no calcifications were identified on CNB specimen radiographs, and the CNB yielded a finding of LCIS. Imaging of the other lesion that was upgraded to DCIS was a discrete calcified mass, but the US-guided CNB yielded only LCIS and ill-defined fibroadenomatoid changes. The third lesion that was upgraded to IDC at surgery was a dense, spiculated mass at mammography and sonography, but the US-guided CNB yielded only fragments of benign breast tissue with stromal fibrosis and ALH. (Table 2)
Table 2. Radiology and Pathology Findings in 8 Discordant Core-Needle Biopsies
|Ca++||ALH, FCC||ADH||Few Ca++ on specimen radiographs; no Ca++ at pathology|
|Ca++||LCIS, cysts, Ca++||DCIS, flat type, intermediate nuclear grade||No Ca++ on specimen radiographs|
|Mass with Ca++||LCIS and FA||DCIS, cribriform & papillary types, intermediate nuclear grade||Review of images before, during, and after CNB suggested that target was not sampled|
|Mass||ALH, SF||IDC, 1.6 cm MSBR grade II/III||Spiculated mass (BI-RADS 5); scant tissue cores obtained judged to be poor quality|
|Mass||ALH, SF||LCIS and ALH in FA||Insufficient explanation for mass|
|Mass||LCIS in SA||LCIS in SA, papillomas||Insufficient explanation for mass|
|Mass||LCIS, SF, SA||LCIS, SA, FA||Insufficient explanation for mass|
|Mass||ALH, SF||ALH, SF, cysts||Insufficient explanation for mass|
The rate of upgrade to carcinoma at EXB was significantly greater for lesions that had discordant imaging-histologic findings compared with lesions that had concordant findings (3 of 8 lesions [38%] vs 2 of 72 lesions [3%]; P < .01).
No consensus exists regarding the optimal management of patients with LN at CNB. The limitations of published series include retrospective design, small numbers, and incomplete or absent information regarding imaging-histologic correlation. Only a subset of patients who had lesions that yielded a finding of LN at CNB underwent EXB in most series, and the rationale for surgical excision in those patients often is incompletely reported. In 10 series (including the current study) that assessed 478 lesions yielding LN at CNB in patients who underwent subsequent EXB, surgery yielded cancer in 8% of these lesions (range, 3%-35%), 67% of lesions were invasive, and 33% of lesions were DCIS (Table 3).
Table 3. Findings at Surgical Excision of Lobular Neoplasia Diagnosed at Core-Needle Biopsy: Summary of the Literaturea
|Shin & Rosen, 20027b||13||5||8||NA||No||2||1||1||2/13 (15)||—|
|Middleton,20038b||17||6||9||2||No||6||0||6||6/17 (35)||All mass lesions|
|Renshaw, 200630b||92||NA||NA||92||No||7||1||6||7/92 (8)||Carcinoma away from CNB site in 5 cases; 1 case was not cLN at CNB|
|Cangiarella, 200828b||38||18||20||NA||No||3||1||2||3/38 (8)||Two mass lesions|
|Elsheikh & Silverman, 200522c||33||20||13||NA||Yes||9||4||5||9/33 (27)||Two mass lesions; 1 case was not cLN at CNB|
|Liberman, 19996b||13||9||4||NA||Yes||2||1||1||2/13 (15)||All non-cLN|
|Hwang, 200821b||87||48||39||NA||Yes||4||1||3||4/87 (5)||Three cases were not cLN at CNB|
|Nagi, 200820b||45||NA||NA||45||Yes||2||1||1||2/45 (4)||—|
|Rendi, 201127c||68||20||48||NA||Yes||2||2||0||3/68 (4)||All in high-risk women; 1 mass lesion|
|Current studyc||72||42||30||NA||Yes||2||1||1||2/72 (3)||One in a high-risk woman|
In examining CNB outcomes, it is critical to assess concordance or discordance between imaging features and CNB histology. Discordance suggests that the imaging target may not have been accurately sampled. At stereotactic and US-guided biopsy, discordance has been reported in 0% to 6% of all percutaneous breast biopsies; among discordant lesions in the published literature, surgical excision yielded carcinoma in 18% (range, 3%-63%).11 In a prior study from our center, we reported discordance in 3% of stereotactic or US-guided biopsies, and cancer was identified at EXB in 24% of these discordant lesions.11 The upgrade rate was significantly higher for lesions that were classified as BI-RADS Category 5 versus Category 4 (44% vs 14%, P < .04).11 At MRI-guided VAB, we previously reported discordance in 7% of all CNBs, with EXB yielding carcinoma in 30%.12 The higher frequency of discordance among biopsies performed under MRI (7%) versus stereotactic or US guidance (3%) probably reflects the patient population undergoing breast MRI (that more often is at high risk and/or has synchronous breast cancer) as well as the more limited published experience with breast MRI-guided biopsy, both factors may lower the threshold for a “discordant” classification.
The most common scenario for imaging-histologic discordance is a lesion that is evident as calcifications at mammography for which the patient undergoes stereotactic biopsy, but no calcifications are identified at specimen radiography. For mass lesions at stereotactic or US-guided biopsy or for MRI enhancement at MRI-guided biopsy, the recognition of discordance is more challenging because of the lack of a reliable marker to confirm tissue retrieval. Specifically, specimen radiography, which is valuable to confirm lesion retrieval for calcific lesions, is not useful for masses or for MRI lesions.12, 15 Recognition of discordance for masses or MRI enhancement depends on familiarity with the imaging patterns of specific histologies and careful correlation of images obtained before, during, and after biopsy to determine the extent of sampling or the imaging target.
The issue of imaging-histologic discordance is of particular importance when considering appropriate management of lesions that yield a finding of LN at CNB. The classic teaching is that LN is an incidental finding in a biopsy performed for another reason. More recently, there has been increasing recognition that lobular lesions, particularly LCIS, may contain mammographically evident calcifications.16, 17 Some investigators have suggested that LCIS may be evident as enhancement at MRI,18, 19 an issue that warrants further investigation.
Lobular lesions generally do not present as masses at mammography or US; therefore, a diagnosis of LN generally does not provide sufficient explanation for a mammographic or US mass. Furthermore, a diagnosis of LN in a BI-RADS Category 5 lesion is considered a discordant result warranting surgical excision because of the high frequency of cancer at EXB in Category 5 lesions.11
In previously published studies that addressed LN at CNB, limited data exist regarding the concordance or discordance of imaging findings with histology. Middleton et al examined the findings in 17 lesions with LN at CNB. All lesions underwent excision, and carcinoma was present in 6 lesions (35%).8 The article contained no information on imaging-histologic concordance or histologic findings at CNB, but the authors comment that all 6 lesions with upgrade were detected as a mass on imaging and that there were no upgrades when calcifications were the imaging target.8
Nagi et al reported the findings at EXB from 45 women with LN on CNB.20 It appears that EXB was not prospective, and the authors did not specify why it was performed. Microinvasive lobular carcinoma was identified in 1 lesion, a small focus of DCIS was identified in another, and the overall upgrade rate was 4% (2 of 45 lesions).20
A retrospective series by Hwang et al included 87 lesions with LN.21 After excluding imaging-histologic discordant cases and cases with nonclassic LCIS morphology, the upgrade rate was approximately 1% (1 of 84 lesion) because of a 0.4-cm focus of DCIS with low nuclear grade, which was called “incidental” by the authors.21
Few studies have reported the findings at prospective EXB of LN. Among the 33 lesions reported by Elsheikh and Silverman, 18 lesions had prospective EXB which yielded cancer in 27%.22 Some of the lesions in that series yielded nonclassic LCIS at CNB. We excluded from our series any nonclassic variant of LCIS that was routinely referred for EXB,23-26 because our objective was to specifically evaluate the rate of upgrade of LN with classic morphology.
Rendi and colleagues studied 68 lesions with LN at CNB and prospective EXB. One of the lesions that was upgraded had discordant imaging-histologic findings. After excluding the discordant case, there were 2 of 67 lesions (3%) that were upgraded at EXB. In both lesions, the imaging target was an MRI nonmass-like enhancement, and the upgrade consisted of DCIS.27
All patients who were diagnosed at our center with LN at CNB have been referred for prospective EXB of the target lesion since June 2004, regardless of imaging-histologic concordance. We reviewed the experience with lesions that were biopsied at our center and had concordant imaging-histologic findings to obtain an unbiased estimate of the upgrade rate of LN at CNB.
Among 72 concordant lesions in our study, the upgrade rate at prospective EXB was 3% (95% confidence interval, 0%-9%), which was low relative to most prior retrospective studies7, 8, 28 but similar to the upgrade rate observed in recent series with imaging-histologic correlation21, 27 and prospective excision.27 In our series, both concordant lesions with upgrade consisted of minute, low-grade carcinomas: a 0.2-cm IDC with associated low-grade DCIS in one case and a 0.2-cm low-grade DCIS in the other. Similarly, the imaging-histologic concordant LN lesions with upgrade in the series by Hwang et al21 and Rendi et al27 were focal low-grade DCIS (total, 3 cases).
When prior results21, 27 are analyzed in combination with ours, a total of 5 minute, low-grade carcinomas were identified at excision of 223 concordant lesions that yielded a finding of classic LN at CNB, for an overall upgrade rate of just over 2%. Although there is no question that low-grade carcinomas are malignant, some of these minute lesions may be regarded as “incidental.” If undetected, these small carcinomas likely would have very minimal, if any, impact on patient survival, especially in patients with ALH/LCIS who are subject to close follow-up.
In the study by Rendi et al, all patients who had a lesion upgraded were women at high risk for breast carcinoma, and no upgrades occurred among women in a routine mammographic screening group that had no increased risk of breast cancer.27 Those authors suggested that excision of lesions that yielded LN at CNB is appropriate in women at high risk of breast cancer, but close follow-up may be sufficient if the same diagnosis is obtained at CNB of screening-detected mammographic calcifications. One of the 2 patients with upgrade in our study had a strong family history of breast carcinoma; but the other patient was not at high risk of breast cancer, and her index CNB targeted screen-detected calcifications. By using the algorithm suggested by Rendi et al for our study population, only the patient who had a strong family history would have been recommended for excision; the upgrade rate would have been 1% (1 of 72 lesions), and 1 of the 2 upgrades to low-grade cancer would have been missed.
Our findings, interpreted in conjunction with the results from recent series that included only cases of classic LN and with careful imaging-histologic correlation, indicate that the upgrade rate at EXB of LN identified at CNB is very low, just above the 2% threshold routinely used by radiologists to recommend surgical excision.29 In addition, the number of carcinomas that would have been missed if there had been no excision for concordant cases was extremely small and consisted mostly of microscopic DCIS with low nuclear grade.
Lesions that yield LN at CNB with imaging-histologic discordance warrant surgical excision because of the high rate of cancer in these lesions (38% in our series; 95% confidence interval, 9%-76%). Among lesions that yield LN at CNB with imaging-histologic concordance, the cancer rate was significantly lower (3%; 95% confidence interval, 0%-9%), and both lesions that were upgraded at surgery were small, low-grade cancers (1 DCIS and 1 IDC). This information can help radiologists, pathologists, and surgeons assess the risks/benefits of excision in cases with imaging-histologic concordance.