Annals of Surgical Oncology

, Volume 17, Supplement 3, pp 211–218

Impact of Breast Density on the Presenting Features of Malignancy

Authors

  • Nimmi Arora
    • Breast Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center, Evelyn H. Lauder Breast Center
  • Tari A. King
    • Breast Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center, Evelyn H. Lauder Breast Center
  • Lindsay M. Jacks
    • Department of Epidemiology and BiostatisticsMemorial Sloan-Kettering Cancer Center
  • Michelle M. Stempel
    • Breast Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center, Evelyn H. Lauder Breast Center
  • Sujata Patil
    • Department of Epidemiology and BiostatisticsMemorial Sloan-Kettering Cancer Center
  • Elizabeth Morris
    • Department of RadiologyMemorial Sloan-Kettering Cancer Center
    • Breast Service, Department of SurgeryMemorial Sloan-Kettering Cancer Center, Evelyn H. Lauder Breast Center
American Society of Breast Surgeons

DOI: 10.1245/s10434-010-1237-3

Cite this article as:
Arora, N., King, T.A., Jacks, L.M. et al. Ann Surg Oncol (2010) 17: 211. doi:10.1245/s10434-010-1237-3

Abstract

Purpose

To determine the relationship between breast density, presenting features and molecular subtype of cancer, and surgical treatment received.

Methods

Retrospective review of a prospectively maintained database. Eligible patients had stage 1–3 cancer, were treated between 1/2005 and 6/2007, and had estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) measurements and films available for review. Density was classified at presentation as 1–4 using the Breast Imaging Reporting and Data System (BI-RADS) classification.

Results

1,323 patients were included. Significant differences across the four density groups were present in age, race, multicentricity/focality, and presence of an extensive intraductal component (EIC). When density was combined into two groups, after adjustment for age, only an EIC and mammographically occult cancer were significantly more common in the dense groups. Extremely dense breasts (BI-RADS density 4) more commonly had luminal A tumors (p = 0.05), lobular cancers (p = 0.03), multicentricity (p = 0.02), and occult tumors (p < 0.0001). Greater density was associated with increased mastectomy use, with 61% of the extremely dense group having mastectomy versus 43% of those of lesser density (p = 0.01).

Conclusions

Cancers in extremely dense breasts occur in younger women, are more often mammographically occult, and appear to be phenotypically different from those arising in other density groups. The more common use of mastectomy may be related to these features, although density itself is not a selection criterion for mastectomy.

Breast density is recognized as a risk factor for development of breast cancer, as well as a feature that is associated with failure to identify cancers with screening mammography.15

A substantial component of breast density is genetically inherited, and there is increasing interest in the idea that the genetic variants responsible for a high level of breast density may provide clues to the underlying mechanisms of breast carcinogenesis.69

The genetic variations responsible for breast density may also result in the development of different subtypes of breast cancer. A small number of studies have examined the relationship between breast density and tumor features, such as hormone receptor status and tumor size, with inconsistent results.10,11 In light of the recognition that breast cancer is not a single entity, but is composed of genetically distinct tumor subtypes, we sought to determine whether breasts of different density gave rise to breast cancers of different molecular subtypes, to determine if presenting features of breast cancer which might be important for selection of local therapy varied among breasts of different density, and to determine whether surgical therapy varies among women with different breast densities.12

Patients and Methods

After Institutional Review Board approval, data were obtained from a prospectively maintained, registered database. Patients with stage 1–3 invasive breast cancer were eligible for inclusion if they were surgically treated at Memorial Sloan–Kettering Cancer Center (MSKCC) between 1/2005 and 6/2007, and had a mammogram at the time of diagnosis reviewed by the MSKCC breast-imaging group. This time period was chosen because estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) were routinely obtained, the Breast Imaging Reporting and Data System (BI-RADS), classification of breast density was routinely included in the reporting of mammograms, and no major changes in imaging equipment occurred.13 Patients with history of prior breast cancer were excluded. Mammographic density was classified into four groups using BI-RADS definitions: 1 = almost entirely fatty (< 25% glandular); 2 = scattered fibroglandular densities (25–50% glandular); 3 = heterogeneously dense (51–75% glandular); or 4 = extremely dense (> 75% glandular). Density determination was made on the basis of films obtained at MSKCC if available. If not, the MSKCC radiologists’ classification of breast density for films obtained at outside institutions was used. Thus, both digital and film screen studies were used for density determination. The mammographic presentation of the cancer was recorded. For seven patients, mammograms obtained after surgical excision were the only available images for review; as a result, the presenting mammographic findings were not available. Tumor features, including histologic type, size, grade, ER and PR status, HER2/neu status, multifocality and multicentricity, lymphovascular invasion (LVI), and presence of an extensive intraductal component (EIC) were obtained from the pathology report of the definitive surgery. Multifocality and multicentricity were defined on pathologic examination as discontinuous growth of tumor in one quadrant of the breast or presence of tumor in multiple quadrants of the breast, respectively. An EIC was defined as presence of intraductal carcinoma both in the invasive tumor and in adjacent breast tissue which comprised > 25% of the tumor. Lymph node positivity was defined as presence of any tumor cells in a lymph node and included macrometastases, micrometastases, and isolated tumor cells. Tumor subtype was classified using immunohistochemical surrogates as luminal A (ER-positive and/or PR-positive and HER2-negative), luminal B (ER-positive and/or PR-positive and HER2-positive), HER2 (ER-negative, PR-negative, and HER2-positive), and basal (ER-negative, PR-negative, and HER2-negative).

Statistical analysis was performed using SAS version 9.1 (SAS Institute, Cary, NC). Differences in clinicopathologic and demographic variables among the four groups were univariately tested using the chi-square test for binary variables and the Kruskal–Wallis test for continuous variables. The four density groups were then divided into two categories: almost entirely fatty and scattered densities (BI-RADS density 1 and 2), and heterogeneously dense and extremely dense (BI-RADS density 3 and 4). Logistic regression analysis was performed to evaluate unadjusted and age-adjusted associations between breast density and patient and tumor characteristics. Similar analyses were done to evaluate association between clinicopathologic and demographic variables and density when density was divided into the following categories: extremely dense (BI-RADS density 4), and others (BI-RADS density 1–3). p-value < 0.05 was considered to be statistically significant.

Results

A total of 1,323 patients were included in the study. Median patient age was 56 years (range 27–91 years), and median tumor size was 1.5 cm (range 0.1–11 cm). Table 1 shows the patient distribution by density category and the presenting tumor characteristics for each density group. Breasts that were mostly fatty or extremely dense accounted for only 5% and 10% of cases, respectively. Significant differences in age, race, mammographic presentation, presence of multifocality/multicentricity, and an EIC were noted among the four groups. Increasing patient age was significantly associated with less-dense breasts (Fig. 1), while Caucasian patients were more likely to have dense breasts than those of other races.
Table 1

Association of breast density and clinicopathologic variables

Clinicopathologic variables

BI-RADS breast density group

p-valuea

1

2

3

4

n

%

n

%

n

%

n

%

N

72

5%

313

24%

807

61%

131

10%

Age, years

 Median (range)

62 (30–83)

62 (31–84)

54 (29–91)

47 (27–78)

<0.0001

Ethnicity

 Caucasian

55

76%

265

85%

682

85%

116

89%

0.003

 Black

13

18%

36

12%

63

8%

6

5%

 

 Asian

3

4%

9

3%

52

6%

8

6%

 

 Other

1

0%

4

0%

 

 Unknown

1

1%

2

1%

6

1%

1

1%

 

Mammographic findings

 Normal

1

1%

6

2%

82

10%

54

41%

<0.0001

 Distortion

2

3%

20

6%

81

10%

13

10%

 

 Calcifications

3

4%

34

11%

163

20%

23

18%

 

 Mass

66

92%

252

81%

475

59%

41

31%

 

 Unknown

1

0%

6

1%

 

Tumor size (cm)

 N

72

308

790

128

 

 Median (range)

1.6 (0.2–8.0)

1.4 (0.1–11.0)

1.5 (0.1–11.0)

1.5 (0.1–6.8)

0.58

Subtype

 Lum A

50

69%

237

76%

579

72%

102

78%

0.26

 Lum B

9

13%

19

6%

58

7%

13

10%

 

 HER2

3

4%

14

4%

48

6%

6

5%

 

 Basal

10

14%

43

14%

122

15%

10

8%

 

Tumor type

 Ductal/mixed

63

88%

265

85%

702

87%

107

82%

0.71

 Lobular

7

10%

37

12%

83

10%

20

15%

 

 Special

2

3%

11

4%

22

3%

4

3%

 

Extensive intraductal component

 Yes

6

8%

34

11%

147

18%

21

16%

0.006

 No

66

92%

279

89%

660

82%

110

84%

 

Multicentric/focal

 Yes

16

22%

79

25%

227

28%

51

39%

0.02

 No

56

78%

234

75%

580

72%

80

61%

 

Histologic gradeb

 I

2

3%

13

4%

31

4%

1

1%

0.42

 II

14

19%

54

17%

152

19%

26

20%

 

 III

48

67%

200

64%

519

64%

77

59%

 

 N/A

6

8%

35

11%

71

9%

19

15%

 

 Unknown

2

3%

11

4%

34

4%

8

6%

 

Lymphovascular invasion

 Yes

21

29%

91

29%

292

36%

51

39%

0.07

 No

51

71%

222

71%

515

64%

80

61%

 

No. of nodes positive

 0

44

61%

171

55%

449

56%

68

52%

0.85

 1–3

20

28%

105

34%

249

31%

44

34%

 

 ≥4

8

11%

37

12%

109

14%

19

15%

 

HER2 human epidermal growth factor receptor 2, ILC infiltrating lobular carcinoma, BI-RADS Breast Imaging Reporting and Data System

Percentages and statistical tests based on available data

aChi-square test for binary variables; Kruskal–Wallis test for continuous variables

bILC patients were not graded (histologic grade = N/A)

https://static-content.springer.com/image/art%3A10.1245%2Fs10434-010-1237-3/MediaObjects/10434_2010_1237_Fig1_HTML.gif
Fig. 1

Distribution of breast density by age. Patients younger than 40 years of age were significantly more likely to be in the higher BI-RADS density categories than their older counterparts. BI-RADS Breast Imaging Reporting and Data System

Subsequent analyses were carried out by collapsing the population into two groups. Twenty-nine percent had almost entirely fatty breasts or breasts with scattered density (BI-RADS density 1 and 2; group A), and 71% of patients were classified as having heterogeneously or extremely dense breasts (BI-RADS density 3 and 4; group B). Patients in group A had median age of 62 years (range 30–84 years), while those in group B had median age of 53 years (range 27–91 years; p < 0.0001). Black women had an age-adjusted odds ratio (OR) of 0.57 for extremely or heterogeneously dense breasts compared with Caucasian women (p = 0.002).

The mammographic presentation of cancer differed significantly on the basis of density. Patients in group B were significantly more likely to have mammographically occult tumors than those in group A (14% versus 2%). While a mass was the most common mammographic presentation in both groups, cancers presenting as calcifications and architectural distortion were more frequently observed in group B. Differences in mammographic presentation on the basis of density were highly significant after adjustment for age (p < 0.0001) and are consistent with the observation that patients in group B were significantly more likely than those in group A to have tumors with an EIC (18% versus 10%, age-adjusted OR 0.63 for no EIC in group B, p = 0.02). The distribution of molecular subtypes of cancer did not vary with density in this analysis. After adjustment for age, basal tumors accounted for 14% of tumors in both the fatty and the dense groups, while luminal A tumors comprised 75% of the fatty group and 73% of the dense group (p = 0.85). There were no differences in tumor histology or prognostic features between group A and group B. In the subset of patients with lymph node metastasis, there was also no difference seen between the two groups in the number of positive nodes (p = 0.93).

We also compared patients with extremely dense breasts (BI-RADS density 4) with those in all other density groups (Table 2). Patients in the BI-RADS density 4 group (n = 131) had median age of 47 years compared with 58 years in the other density group (p < 0001), and 41% (n = 54) had mammographically occult tumors. After adjustment for age, luminal A tumors were significantly more common in the BI-RADS density 4 group (p = 0.05). This difference is reflected in the distribution of histologic subtypes, with lobular cancers occurring with an odds ratio of 2.04 in patients with extremely dense breasts compared with those of other densities (p = 0.03). Multifocal/centric cancers were also noted to be more frequent in the extremely dense group (p = 0.02). Prognostic features such as tumor size, nodal positivity, and histologic grade (after exclusion of the lobular group, which was not routinely graded) did not differ significantly.
Table 2

Clinicopathologic characteristics of extremely dense breasts compared with all other density groups

Clinicopathologic variables

Breast density

Logistic regression (extremely dense versus other)

Less than extremely dense

Extremely dense

Unadjusted

Age-adjusted

n

%

n

%

N

1,192

90%

131

10%

OR (95% CI)

p-valuea

OR (95% CI)

p-valueb

Age, years

 Median (range)

58 (29–91)

47 (27–78)

0.94 (0.93–0.96)

 <0.0001

  

Ethnicity

 Caucasian

1002

84%

116

89%

1.00

0.20

1.00

0.12

 Black

112

9%

6

5%

0.46 (0.20–1.08)

 

0.42 (0.18–0.99)

 

 Asian/other

69

6%

8

6%

1.00 (0.47–2.14)

 

0.77 (0.35–1.66)

 

 Unknown

9

1%

1

1%

    

Mammographic findings

 Normal

89

7%

54

41%

1.00

<0.0001

1.00

<0.0001

 Distortion

103

9%

13

10%

0.21 (0.11–0.41)

 

0.22 (0.11–0.43)

 

 Calcifications

200

17%

23

18%

0.19 (0.11–0.33)

 

0.20 (0.11–0.35)

 

 Mass

793

67%

41

31%

0.09 (0.05–0.14)

 

0.11 (0.07–0.17)

 

 Unknown

7

1%

    

Tumor size (cm)

 N

1170

128

    

 Median (range)

1.5 (0.1–11.0)

1.5 (0.1–6.8)

1.05 (0.93–1.19)

0.45

1.00 (0.87–1.14)

0.97

Subtype

 Lum A

866

73%

102

78%

1.00

0.13

1.00

0.05

 Lum B

86

7%

13

10%

1.28 (0.69–2.38)

 

0.95 (0.50–1.80)

 

 HER2

65

5%

6

5%

0.78 (0.33–1.85)

 

0.66 (0.27–1.60)

 

 Basal

175

15%

10

8%

0.49 (0.25–0.95)

 

0.39 (0.20–0.77)

 

Tumor type

 Ductal/mixed

1030

86%

107

82%

1.00

0.28

1.00

0.03

 Lobular

127

11%

20

15%

1.52 (0.92–2.53)

 

2.04 (1.19–3.48)

 

 Special

35

3%

4

3%

1.10 (0.38–3.16)

 

1.17 (0.40–3.46)

 

Extensive intraductal component

 Yes

187

16%

21

16%

1.00

0.92

1.00

0.34

 No

1005

84%

110

84%

0.97 (0.60–1.59)

 

1.28 (0.77–2.13)

 

Multicentric/focal

 Yes

322

27%

51

39%

1.00

0.004

1.00

0.02

 No

870

73%

80

61%

0.58 (0.40–0.84)

 

0.64 (0.44–0.94)

 

Histologic grade

 I

46

4%

1

1%

1.00

0.10

1.00

0.004

 II

220

18%

26

20%

5.44 (0.72–41.07)

5.48 (0.72–41.87)

 

 III

767

64%

77

59%

4.62 (0.63–33.94)

3.58 (0.48–26.62)

 

 N/A

112

9%

19

15%

7.80 (1.02–59.99)

9.19 (1.18–71.66)

 

 Unknown

47

4%

8

6%

 

 

 

 

Lymphovascular invasion

 Yes

404

34%

51

39%

1.00

0.25

1.00

0.98

 No

788

66%

80

61%

0.80 (0.56–1.17)

 

1.00 (0.68–1.46)

 

No. of nodes positive

 0

664

56%

68

52%

1.00

0.70

1.00

0.99

 1–3

374

31%

44

34%

1.15 (0.77–1.71)

 

1.02 (0.67–1.53)

 

 ≥4

154

13%

19

15%

1.21 (0.70–2.06)

 

0.97 (0.56–1.69)

 

OR odds ratio, CI confidence interval, HER2 human epidermal growth factor receptor 2

Percentages and statistical tests are based on available data

aUnadjusted logistic regression (extremely dense versus other)

bBivariate logistic regression (extremely dense versus other), adjusted for age at surgery

cILC patients were not graded (histologic grade = N/A)

Use of breast-conserving surgery (BCS) was 68% in patients in the BI-RADS density 1 group, 61% in the BI-RADS density 2 group, 55% in the BI-RADS density 3 group, and 39% in the BI-RADS density 4 group (p < 0.0001). After age adjustment, there was no difference in use of BCS when patients were grouped into two density groups, with 62% of the BI-RADS density 1 and 2 patients undergoing BCS compared with 53% of the BI-RADS density 3 and 4 patients (p = 0.11). However, when the BI-RADS density 4 patients were examined separately, after adjustment for age, a significant decrease in use of BCS was noted, with 39% of patients in this group having BCS compared with 57% in the less-dense group (p = 0.01).

Discussion

Mammographic density is a reflection of the proportion of epithelial and stromal tissue relative to fat on a mammogram. Increased mammographic density increases the risk of breast cancer development 3- to 5-fold over that of women with fatty breasts.13,14 It has long been recognized that increased density of breast parenchyma is associated with cancers which are not detected by screening mammography.4,15,16 Here we confirm that, despite advances in mammographic technique, a significant number of cancers (41%) arising in dense breasts are still mammographically occult. In our study, patients with heterogeneously or extremely dense breast tissue were more likely to present with calcifications or architectural distortion than their counterparts with fatty breasts. This is consistent with the work of Porter et al., who found that in 455 patients with screen-detected cancers, those with dense breast parenchyma were significantly more likely to present with architectural distortion and less likely to present with a mass compared with patients with fatty breast tissue.17 However, they did not find any difference in presentation with calcifications between groups, while calcifications were twice as frequent in extremely or heterogeneously dense breasts in our study. This difference is most likely a reflection of differences in the patient populations studied. Porter et al. studied patients with screen-detected cancers, while we included patients with clinical presentations and those not undergoing routine mammography in addition to those with cancers detected by screening. In this study, we also confirmed the relationship, previously described by others, between Caucasian ethnicity and greater breast density.18,19

A substantial component of mammographic density, approximately 60–70%, is thought to be the result of a genetic inheritance pattern, with the remainder determined by weight, parity, and menopausal status.6,20,21 The molecular profiles of dense and fatty stroma differ, and multiple candidate genes primarily from pathways related to steroid hormone regulation and proliferation have been examined to determine if polymorphisms in these genes are responsible for the variations in density observed in the population.7 The increasing recognition of the importance of host factors in cancer outcome, coupled with the association between hormonal factors and breast density, led us to examine the possibility of an association between breast density and molecular subtypes of breast cancer. We have previously shown that clinical features of breast cancer at presentation vary significantly between the molecular subtypes.22 In this study, we did not observe any correlation between density and development of a particular subtype of cancer in an unadjusted analysis or when density was dichotomized into groups of greater and lesser density. However, a significant increase in the proportion of luminal A cancers was noted when extremely dense breasts were compared with all others. This finding, although counterintuitive, is internally consistent with the observation that infiltrating lobular cancers were also most frequent in this group. Others have shown that classical infiltrating lobular carcinomas almost all have gene expression profiles consistent with the luminal A phenotype, and Fasching et al. have shown increased lobular cancer incidence in patients with dense breasts.23,24 In a case–control study examining the relationship between density and breast cancer subtype, Ma et al. found increased breast density to be a risk factor for both luminal A and basal cancers, but that the strength of association did not differ on the basis of subtype.25 In contrast, we observed both an increased incidence of luminal A cancers and a decreased incidence of basal cancers in patients with extremely dense breasts.

Other studies have examined the relationship between ER status or HER2 expression and breast density. Seo et al. found no correlation between density and HER2 status in 498 patients, consistent with our finding of no difference in incidence of the HER2 subtype on the basis of density.26 A number of studies have examined the relationship between mammographic density and hormone receptor expression, and a consistent relationship has not been identified.11,25,27,28 Our study builds upon the existing literature by examining all four of the distinct molecular subtypes of breast cancer which have immunohistochemical surrogates.

We did not find any correlation between breast density and some of the more common tumor prognostic features, such as size, grade, and lymph node involvement. Table 3 summarizes seven recent studies that have attempted to identify associations between density and tumor characteristics. Four of these studies found a positive correlation between increasing density and tumor size.5,10,11,17 Two studies found no correlation.28,29 One study found a negative correlation.24 We also observed no correlation between tumor size and breast density. Despite the failure to detect 41% of cancers in extremely dense breasts mammographically, median tumor size in this group was 1.5 cm, the same median tumor size seen in the group with lower breast density. Differences in the literature on the impact of density on tumor size are likely to reflect differences between pure screening populations and those that include a more heterogeneous group of patients such as our study.
Table 3

Recent studies investigating the relationship between breast density and invasive cancer

Author

Year

Study size

Density measure

N for highest density group

Significant findingsa

Sala5

2000

747

Wolfe grade

90

Tumor size, LN status (+)

Roubidoux11

2004

121

BI-RADS

16

Tumor size, stage (+); ER status (−)

Aiello10

2005

546

BI-RADS

23

Tumor size, LN status, LVI (+); grade and mitotic index (−)

Morishita28

2005

163

BI-RADS

NA

Association between ER–negative/dense breasts and distant mets: none

Fasching24

2006

434

BI-RADS

97

Tumor size (−); invasive lobular type (+)

Porter17

2007

759

BI-RADS

71

Tumor size (+)

Ghosh29

2008

286

Percentage and area density

NA

Tumor grade (−)

Current study

2010

1323

BI-RADS

131

Luminal A, invasive lobular type, multifocal/multicentric (+)

LN lymph node, BI-RADS Breast Imaging Reporting and Data System, ER estrogen receptor, LVI lymphovascular invasion, mets metastasis, NA not available

aStatistically significant findings after controlling for confounding variables; positive correlation denoted by (+), negative correlation by (−)

While presence of mammographically dense breasts alone is not considered a contraindication to BCS, we did observe significantly more mastectomies in the extremely dense group in comparison with the other densities after adjustment for age.30 The reasons for this are not entirely clear. Morrow et al. have shown that the mammographically occult tumors that are more common in dense tissue are not associated with an increased rate of failure of BCS.31 However, many patients and their physicians have concerns about difficulty in detection of local recurrence in tumors that are mammographically occult. Our finding that multifocality, multicentricity, and infiltrating lobular carcinomas were more frequent in extremely dense breasts may help to explain the increased use of mastectomy in this group. Multicentricity is a contraindication to BCS, while multifocality and infiltrating lobular carcinoma have been associated with a higher likelihood of positive margins in some studies.30,3234 Limited and conflicting information on the impact of breast density on local recurrence has been reported. Park et al. performed a nested case–control study of 136 women with invasive cancer treated with BCS and radiotherapy and observed a hazard ratio of 4.3 (95% confidence interval, 0.88–21.00) for local recurrence among those with greater than 75% density compared with those with less than 25% density.35 In contrast, Cil et al. found an association between breast density and local recurrence only among patients who did not receive RT.36 If, as we observed, extremely dense breast tissue is associated with a greater incidence of luminal A tumors, this subset of patients has the lowest risk of local recurrence after BCS and treatment with radiotherapy and tamoxifen.37,38 The possible relationship between density and local recurrence is a subject worthy of further investigation, but our study does not identify any features of tumors occurring in dense breasts which would be associated with an increased risk of local recurrence other than young patient age.

There are several potential limitations to this study. First, classifications based on ER, PR, and HER2 status are only approximations of the molecular subtypes of breast cancer. The definition of luminal B used in this study (ER- and/or PR-positive, HER2-positive) does not identify all luminal B tumors because only 30–50% are HER2 positive.39 Similarly, not all basal tumors are triple negative, and not all triple-negative tumors are basal. However, immunohistochemistry (IHC) profiles that use readily available clinical receptors are cost effective and practical, and have been successfully used as surrogates for gene expression profiling by others. This study used the BI-RADS classification of density assigned at time of clinical management for density determination. Other studies have measured the percentage of the breast which is composed of dense tissue, or acquired density measurements directly from digital mammograms, and found this to be a more accurate method of density determination than the BI-RADS classification.19 However, the distinction between extremely dense breasts, the subgroup of importance in this study, and other density groups is rarely difficult, and it is unlikely that an alternative method of density measurement would change our results. It is possible that the increased use of mastectomy in the extremely dense group, by providing the pathologist with more breast tissue to examine, resulted in an increased detection rate of subclinical multifocal or multicentric disease. Since serial subgross sectioning of mastectomies was not performed in this study, we believe the mastectomy rate is unlikely to have had a major impact on the rate of multicentricity/focality. The increased rate of this feature is internally consistent with the finding of more infiltrating lobular carcinoma—a lesion well documented to be multifocal—in the extremely dense group. Similarly, it is possible that magnetic resonance imaging was more frequently employed in the extremely dense group, potentially resulting in a higher mastectomy rate.40 This was not evaluated in this study, but a more detailed study of variations in treatment factors on the basis of density is ongoing. Finally, although our overall sample size was large, the number of patients in the BI-RADS 4 density group was relatively small, and confirmation of these findings in a larger group of patients is appropriate.

In summary, this study confirms that the association between increased mammographic breast density, young patient age, and occult tumors persists despite improvements in mammographic technique. We identified an association between extremely dense breasts and the luminal A molecular subtype of breast cancer, as well as between extremely dense breasts and cancers of lobular histology. Consistent with these associations, we did not find any correlation between density and tumor prognostic factors, calling into question the preferential use of mastectomy in women with dense breasts.

Copyright information

© Society of Surgical Oncology 2010