A BRCA1/2 mutation, high breast density and prominent pushing margins of a tumor independently contribute to a frequent false-negative mammography
Version of Record online: 6 SEP 2002
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 102, Issue 1, pages 91–95, 1 November 2002
How to Cite
Tilanus-Linthorst, M., Verhoog, L., Obdeijn, I.-M., Bartels, K., Menke-Pluymers, M., Eggermont, A., Klijn, J., Meijers-Heijboer, H., van der Kwast, T. and Brekelmans, C. (2002), A BRCA1/2 mutation, high breast density and prominent pushing margins of a tumor independently contribute to a frequent false-negative mammography. Int. J. Cancer, 102: 91–95. doi: 10.1002/ijc.10666
- Issue online: 25 SEP 2002
- Version of Record online: 6 SEP 2002
- Manuscript Accepted: 25 JUL 2002
- Manuscript Revised: 23 JUL 2002
- Manuscript Received: 15 APR 2002
- breast cancer;
- young patients;
Female BRCA1/2 mutation carriers develop in up to 50% breast cancer (BC) before age 50 years. We investigated whether the specific histologic features of BRCA1/2-associated breast cancer influence imaging. We correlated the mammographic results with the histology of 34 BC in BRCA1/2 mutation carriers and 34 sporadic cancers in patients, matched for age and year of diagnosis. Mammography was significantly more frequently false-negative in carriers than controls (62% vs. 29% p = 0.01), despite comparable tumor size (mean ⊘ 1.51 vs. 1.75) and breast density (high 41% vs. 53%). The image in carriers was significantly less as spiculated mass (6 vs. 18 p = 0.01). Cancers of BRCA1/2 mutation carriers had frequently higher mitotic counts (p < 0.0001) and prominent pushing margins around the tumor (p = 0.08) (p = 0.05 for 32 BRCA1). We also observed that prominent “pushing margins” correlated significantly with a false-negative mammography (p = 0.005) and with a mammographic image of a smooth, not a spiculated, mass (p = 0.01). False-negative mammography correlated independently with: BRCA1/2 mutation (p = 0.02), prominent pushing margins (p = 0.03) and high breast density (p = 0.01). MRI was carried out in 12 carriers, had 100% sensitivity and detected 5 cancers, still occult at physical examination and mammography. A BRCA1/2 mutation and high breast density at mammography contribute independently to false-negative mammography results. In mutation carriers any mammographic mass must be regarded with suspicion. Pushing margins of the tumor partly explain these results. For early BC detection in mutation carriers additional methods like MRI may be needed. This may not be necessary in other young women with breast symptoms. © 2002 Wiley-Liss, Inc.
Female carriers of a BRCA1 or BRCA2 mutation are at an increased risk of developing breast cancer (BC) at a young age. Risks of 3% at 30 years of age and up to 50% at 50 years have been described.1, 2 Carriers of these mutations often choose surveillance at a young age, consisting generally of regular clinical breast examination and mammography, with the addition of breast magnetic resonance imaging (MRI) in some centers. Preliminary reports on the performance of mammography in the detection of breast cancer in mutation carriers, however, show false-negative results of up to 62%.3, 4, 5 A breast tumor is best recognized at mammography as a malignancy, when it is depicted as a spiculated or ill-defined mass, malignant calcifications, an asymmetric opacity or an architectural distortion. A well-defined mass at mammography indicates a malignancy in <1%.6, 7
One of the causes for disappointing mammography results may be the specific histologic phenotype of BRCA1/2-associated breast cancers. This histologic phenotype is distinguishable from non-hereditary tumors with more frequent high grade features like high mitotic count and less tubule formation. They more often show features of medullary carcinoma like continuous pushing margins (well-defined pushing edge, caused by a continuous front of tumor cells not separated by connective tissue) around >75% of the tumor.8 BRCA1-associated tumors show more prominent lymphocytic infiltrate and a less extensive intraductal component has been described.8, 9, 10, 11, 12, 13 This may possibly influence the imaging in BRCA1/2 carriers.
We investigated whether the specific histologic features of breast cancer in BRCA1/2 gene mutation carriers influence their mammographic appearance and cause a false negative mammographic result.
We correlated the histopathologic characteristics of the breast cancers of BRCA1 and 2 mutation carriers with the mammographic result and presentation. To correct for the improvement of the mammography over 2 decades and for the influence of young age, we compared the BRCA1/2-associated cases with sporadic cancers in patients matched for year of diagnosis and age at onset. Further we checked whether the histologic features in the 2 groups showed the differences described in the literature.
PATIENTS AND METHODS
BRCA1 and BRCA2-associated cases
In this cross-sectional study we included all consecutive cases of breast cancer from January 1980–September 2001 in patients with a disease-causing BRCA1/2 germline mutation, whose mammography at time of diagnosis and histopathologic report could be obtained in the Daniel den Hoed Cancer Center. Carriers were identified through the registry of the Family Cancer Clinic of the Erasmus University Medical Center Rotterdam. We included invasive and in situ cancers and both primary and contralateral breast cancer if the first treatment was surgery. We excluded all ipsilateral recurrences. In this way, 43 cancers were identified. No imaging was available of the detection of 9 cancers, so the analyses were carried out with 34 cancers.
Selection of control cases
The group of 34 BRCA1/2-associated BC cases was matched for age at onset and year of diagnosis (the same year or as close as possible within 5 years) with 34 BC cases in sporadic patients without a family history of breast cancer detected in the outpatient breast clinic of our hospital. The controls were not tested for mutations in BRCA1 or BRCA2. Controls were further excluded if the mammography carried out at detection of BC or histologic slides were not available in our clinic.
All mammograms were reviewed by an experienced radiologist who was informed of the clinical data provided at the initial reading but blinded for carriership and the initial mammographic result. Appearance of the tumor was described and classification was carried out according to the protocol of the American College of Radiology.14 True-positive were mammograms with the classifications of suspicious for malignancy or malignant. False-negative were mammograms with the classifications of no abnormality, benign or probably benign.
Breast cancer staging was carried out according to the TNM classification.15 The histologic slides were reviewed by 2 pathologists (L.V. and Th.vdK) together and in consensus. They were uninformed of both the BRCA1/2 mutation status and radiological appearance of the tumors. Cases and controls were eligible for all the characteristics when invasive BC was present, (in in situ carcinoma only presence of DCIS could be scored).
The slides with the largest diameter of the tumor were used for evaluation. The presence of so-called “pushing margins” (smooth well-defined edge), lymphoplasmacytic infiltrate and ductal carcinoma in situ (DCIS) were scored semi-quantitatively.9, 10 A tumor with <25% of pushing margins was given 1 point, 25–75% 2 points and >75% 3 points. If DCIS was absent it was given 1 point, if present, 2 points and extensive, 3 points. The amount of stroma was scored as the estimated percentage of the area within the circumference of the tumor that was not occupied by tumor cells. Less than 25% of stroma was 1 point, 25–75% 2 points and >75% 3 points. Because the quality of the stroma could account for a range of distinct appearances (i.e., densely fibrotic, highly cellular, myxoid stroma), the degree of fibrosis was similarly scored.
The percentage of tubule formation was scored according to the Nottingham modification of the Bloom-Richardson system; >75% (high) 1, 10–75% (medium) 2 and <10% (low) 3.16
Differences in patient and tumor characteristics between carriers and sporadic cases were tested by t-tests for continuous variables and χ2 tests for categorical variables. Fisher's exact test was used if the count in at least 1 cell was <5. Odds ratios (that can be interpreted as relative risks) to investigate the simultaneous effect of several variables on mammographic presentation were computed by conditional logistic regression. A p-value of <0.05 was considered significant. All analyses were carried out with SPSS version 9.0.
Imaging at detection was available of 34 breast cancers (2 in situ), detected between January 1, 1983–Sept 1, 2001 in 26 BRCA1 and 2 BRCA2 (2 cancers) mutation carriers. These cases were matched for age at onset and year of detection to 34 sporadic cancers in 33 patients.
Review of mammography and histopathology
Mammography was significantly less often suspicious for malignancy in carriers as compared to controls, 38% vs. 71% respectively at review (p = 0.01) (Table I). Mammographic review did not considerably change the percentage of false-negative reports in both groups. The difference in true-positive results of mammography was also significant in subgroup analyses with:
- 1Only the 56 cases with all histologic variables complete, with mammography suspect in carriers in (10/28 [36%] vs. controls 20/28 [71%] p = 0.02).
- 2Only BRCA1 and their matched controls (12/32 [41%] vs. 24/32 [75%] p = 0.01).
- 3Exclusion of 3 carriers with no histology available and their matched controls (12/31 [39%] vs. 22/31 [71%] p = 0.02).
- 4Only palpable tumors (9/24 [38%] vs. 23/32 [72%] p = 0.01).
|Mean age (range)||39.4 (23–52)||39.8 (24–53)||0.85|
|Suspicious/malignant||13/34 (38%)||23/34 (68%)||0.03|
|Suspicious/malignant||13/34 (38%)||24/34 (71%)||0.01|
|High breast density||14/34 (41%)||18/34 (53%)||0.47|
|Appearance at review|
|(Partly) smooth mass||13||4||0.01|
|MRI malignant||12/12 (100%)||6/7 (86%)||0.37|
|Diameter all tumors||1.51 cm||1.75||0.30|
The frequency of high breast density at mammography 14/34 (41%) vs. 18/34 (53%) (p = 0.47) and tumor size (mean ⊘ 1.5 cm vs. 1.75 cm respectively; p = 0.30) was no different between carriers vs. controls.
The mammographic images differed significantly between carriers and sporadic cases: a “spiculated/ill-defined mass” was reported significantly less in carriers vs. controls (6 vs. 18) a (partially) smooth mass more often (13 vs. 4) (p = 0.01) (Table I).
MRI detected 5 cancers in carriers that were still occult at physical examination and mammography.
Histopathologic differences between carriers and controls
The histologic slides of 3 carriers were not available (including 1 DCIS) and we excluded their 3 matched controls. Only absence or presence of DCIS could be scored in 1 in situ carcinoma and in 2 of the 4 Stage T1a tumors in carriers. Of their 3 matched controls only DCIS was scored. Breast cancers of BRCA1/2 mutation carriers showed more frequently than controls a high mitotic count (p < 0.0001), a low score for tubule formation (p = 0.007) and more lymphocytic infiltrate (p = 0.02) (Table II). A greater proportion of the tumors with continuous pushing margins were seen more frequently although nonsignificantly (p = 0.08), but reached just significance when comparing (conform subanalysis 2) 32 cancers of BRCA1 carriers with their matched controls.
|Group||BRCA1/2 carrier||Sporadic ca||p-value for trend|
|Tis||2/34 (6)||0/34 (0)|
|T1||27/34 (79)||27/34 (79)||1.00|
|≥ T2||5/34 (15)||7/34 (21)|
|N02||23/32 (72)||22/34 (65)||0.38|
|Low||1/28 (4)||17/28 (61)|
|Medium||6/28 (21)||5/28 (18)||0.0001|
|High||21/28 (75)||6/28 (21)|
|High||1/28 (4)||8/28 (29)|
|Medium||4/28 (14)||6/28 (21)||0.007|
|Low||23/28 (82)||14/28 (50)|
|< 25||18/28 (64)||23/28 (82)|
|25–75||6/28 (21)||5/28 (18)||0.08|
|> 75||4/28 (14)||0/28|
|Grade 1||8/28 (29)||19/28 (68)|
|Grade 2||11/28 (39)||4/28 (14)||0.02|
|Grade 3||9/28 (32)||5/28 (18)|
|Absent||13/31 (42)||9/31 (29)|
|Present||11/31 (35)||17/31 (55)||0.60|
|Extensive||7/31 (23)||5/31 (16)|
|Grade 1||12/28 (43)||6/28 (21)|
|Grade 2||7/28 (25)||14/28 (50)||0.40|
|Grade 3||9/28 (32)||8/28 (29)|
Correlation between pathology and imaging
Prominent vs. absent or <25% of continuous pushing margins at histopathologic examination correlated significantly with a false-negative vs. true-positive (suspicious/malignant) result at mammography (p = 0.005) (Table III). The correlation with negative result of mammography was nonsignificant for lymphocytic infiltrate (p = 0.10), stroma (p = 0.194), fibrosis(p = 0.13) and absence of DCIS (p = 0.22) (Table III).
|Mammography||True positive1||False-negative2||p-value for trend|
|Number at mammography||37 (13/24)3||31 (21/10)3|
|Histology available4||34 (12/22)3||28 (19/9)3|
Absent or <25% of continuous pushing margins at histology correlated also significantly with an image of a spiculated/ill-defined mass at mammography (p = 0.01) (Table IV). With pushing margins >75% of the tumor (3 carriers) no spiculated mass was seen at mammography, whereas pushing margins scored mostly absent or low (score 1) if mammography showed a spiculated mass (3 carriers and 18 controls). A spiculated vs. a smooth mass also correlated significantly with a high score for fibrosis (p = 0.004) and stroma (p = 0.02) but not with lymphocytic infiltrate (p = 0.10) or presence of DCIS (p = 0.20) (Table IV).
|Mammography||Smooth mass||Spic1/ill-defined mass||p-value for trend|
|Number at mammography||17 (13/4)2||24 (6/18)2|
Multivariate analysis of influences on false negative mammography results
In Table V multivariate analysis is carried out in the 56 carriers + controls, in whom all mammographic and histologic characteristics could be scored, of the factors that potentially influence the false-negative results at mammography: size of the tumor, breast density, BRCA1/2 mutation carriership, prominent pushing tumor margins at histology and lymphocytic infiltrate.
|Breast density at mammography||6.860||(1.581–29.775)||0.01|
Factors that independently showed a significant correlation with false-negative mammographic results were carriership (p = 0.02), mammographic breast density (p = 0.01) and prominent pushing margins at histology (p = 0.03).
In our series, the poor sensitivity of mammography in our BRCA1/2 mutation carriers in comparison to controls was not caused by a difference in breast density, tumor size or young age.
The failure of mammography to recognize the cancers in BRCA1/2 mutation carriers has been described in smaller series.3, 4, 19 These could not, however, differentiate between young age, high breast density or the specific histology of carriers as an explanation because they had no age-matched controls or carried out no histology review.
Our observation that “prominent pushing margins” of cancers contribute to a false-negative mammography is new (Fig. 1). Our findings that cancers with absent or low “pushing margins” presented significantly more often as a spiculated mass at mammography (p = 0.01) and were more easily recognized as malignant (p = 0.008) are also new. This presentation was seen mostly in sporadic cancers. We demonstrated that in mutation carriers any mass at mammography must be regarded with suspicion, although benign tumors do occur in mutation carriers with normal frequency.18
Prominent pushing margins, a histologic characteristic of medullary carcinoma, is often described in the literature on the histology of breast cancer in BRCA1 and BRCA 2 mutation carriers. Breast cancers in our carriers and young control group showed the histologic differences in high grade and medullary features as described in the literature and seem representative for their group.8, 9, 10, 11, 12, 13 When pushing margins are prominent the fibrotic reaction of the connective tissue adjacent to the tumor, that causes the spiculated mass and architectural distortion at mammography characteristic of malignancy, is absent. Lack of these features can result in a false-negative mammography. This is also supported by the significantly higher score for fibrosis in our study (p = 0.004) when mammography showed a spiculated/ill-defined mass. Lakhani et al.8, 9 suggested that pushing margins, a continuous front of tumor cells not separated by connective tissue, could result from a reduced potential for stromal infiltration by individual or small groups of tumor cells. Not all the tumors in BRCA1/2 mutation carriers have medullary features like prominent pushing margins. Our multivariate analysis showed that carriership also has negative influence on imaging results independent of pushing margins and breast density.
Several pathologic features have been described for mammographic occult breast cancers apart from tumor size.17 The difference in density between the fibrous stroma produced by the tumor mass and the surrounding fibroglandular tissue can be invisible in dense breasts. This was in our series in carriers and controls of equal influence. Breast density (high: 41% in carriers vs. 53% in controls) by age conformed to the literature.
The desmoplastic reaction produced by the tumor can be poor (i.e., lobular carcinoma) whereas intraductal carcinoma is only recognized when it produces malignant microcalcifications.7, 17 Consistent with the literature, DCIS around the invasive tumors was seen less often in our carriers compared to controls.9, 13 The result, however, was not significant nor was the difference in presentation with microcalcifications at mammography. The presence of DCIS did not strongly influence our imaging results.
Our pathologic review could not further elucidate the mechanism by which BRCA1/2 mutation carriership leads to false-negative mammography results.
Our finding that in BRCA1/2 mutation carriers breast cancer is frequently missed at mammography must be taken into account when defining the optimal imaging strategy for carriers with symptoms or under surveillance. Additional detection methods such as breast-MRI may be needed in these mutation carriers,3, 4, 19, 20. This may not be necessary in other young women with breast complaints.14
We thank M. Voskuil, A. van Eekelen and E. Crepin for logistic support, M. Westerhout-Kersten for assisting literature search, and A. Moerman for assistance with the figure.
- 14Breast imaging reporting and data system. Reston, Virginia: American College of Radiology, 1993., , .
- 15UICC. TNM Classification of malignant tumors, 5th ed. New York: John Wiley & Sons Inc., 1997., .