Background: Mammographic density (MD) is associated with increased breast cancer risk, yet limited data exist on an association between MD and breast cancer molecular subtypes.

Methods: Women ages 18 years and older with breast cancer and available mammograms between 2003 and 2012 were enrolled in a larger study on MD. MD was classified by the Breast Imaging Reporting and Data System (BI-RADS) classification and by volumetric breast percent density (Volpara Solutions). Subtype was assigned by hormone receptor status, tumor grade, and mitotic score (MS). Subtypes included: Luminal-A (ER/PR+ and grade = 1; ER/PR+ and grade = 2 and MS = 1; ER+/PR and grade = 1; n = 233); Luminal-B (ER+ and grade = 3 or MS = 3; ER+/PR and grade = 2; ER/PR+ and grade = 2 and MS = 2; n = 79); Her-2-neu+ (H2P; n = 59); triple-negative (ER/PR, Her-2; n = 86). Precancer factors including age, race, body mass index (kg/m2), family history of breast cancer, and history of lobular carcinoma in situ were analyzed.

Results: A total of 604 patients had invasive cancer; 457 had sufficient information for analysis. Women with H2P tumors were younger (P = 0.011) and had the highest volumetric percent density (P = 0.002) among subgroups. Multinomial logistic regression (LA = reference) demonstrated that although quantitative MD does not significantly differentiate between all subtypes (P = 0.123), the association between MD and H2P tumors is significant (OR = 1.06; confidence interval, 1.01–1.12). This association was not seen using BI-RADS classification in bivariable analysis but was statistically significant (P = 0.047) when controlling for other precancer factors.

Conclusions: Increased MD is more strongly associated with H2P tumors when compared with LA.

Impact: Delineating risk factors specific to breast cancer subtype may promote development of individualized risk prediction models and screening strategies. Cancer Epidemiol Biomarkers Prev; 26(10); 1487–92. ©2017 AACR.

Mammographic density (MD) is a known risk factor for developing breast cancer (1,3,). Several studies have examined the associations of MD with breast cancer subtype by hormone receptor status with variable results (4, 5,). These studies have predominantly examined the association between MD and estrogen receptor (ER) status without regard to HER-2-neu+ (H2P) or progesterone receptor (PR) status (4 ).

An abnormal finding on mammography is frequently the first indication that a breast cancer is present. However, increased MD is known to be associated with decreased mammographic sensitivity (6,). In addition, women with breast cancer detected by screening generally have a more favorable outcome than women whose cancers are detected based on physical exam findings or symptoms prompting evaluation (7,). Tumors detected through screening tend to have a higher rate of ER positivity, generally viewed to infer a more favorable diagnosis (8, 9,). Thus, breast cancer in women with dense breasts is less likely to be detected through conventional screening and may have more aggressive features (9 ).

In clinical practice, breast density is routinely classified according to the Breast Imaging Reporting and Data System (BI-RADS) which categorizes breast density into four categories (10,). This method relies on visual classification by a radiologist and has been shown to have significant intra- and inter-reader variability (11,13,). Automated volumetric density measurements have been shown to have higher reproducibility than visual assessment and eliminate subjectivity, but are used primarily in research rather than routine clinical practice (14, 15,). The purpose of this study was to examine the association between breast density and molecular subtype of breast cancer including H2P status using both BI-RADS classification and quantitative volumetric breast density measurements. Few studies have analyzed the association between MD and H2P tumor subtype as most have focused solely on ER receptor status. One study, using quantitative automated volumetric software to determine MD, identified a stronger association between increased MD and H2P tumors within a Korean population (16 ). Our study examines the associations between MD and four molecular subtypes of breast cancer in a North American population utilizing two different methods to ascribe MD.

All women 18 years of age and older diagnosed with breast cancer between January 2003 and December 2012 and enrolled in a larger, single-center study on MD and breast cancer risk (n = 839) were eligible for this retrospective case–only analysis. Risk factor information was obtained using a questionnaire given between May 2011 and December 2012 (mean, 4 years after diagnosis; range, 0–10 years). Due to the retrospective collection, menopausal status could not accurately be assessed. Exclusion criteria from the larger study included lack of bilateral digital mammography performed at our institution during the study period, initiation of systemic treatments for breast cancer prior to obtaining a digital mammogram, women with bilateral breast cancer at initial diagnosis or a new cancer diagnosis in the contralateral breast within 1 year of the initial diagnosis, as well as patients with a history of breast implants or reduction surgery. Patients were excluded from the currently reported study if cancer was only in situ or if pathologic information allowing assignment to molecular tumor subtype was incomplete. Patient demographics, family history of breast cancer in a first- or second-degree relative, and clinical information were collected through a combination of patient survey and retrospective chart review. Breast cancer detection was categorized as occurring through mammographic screening, palpable mass, or by other symptoms such as breast pain or nipple discharge.

MD was abstracted from existing imaging reports using the 4th edition BI-RADS category reported in the mammogram (screening or diagnostic) closest to the time of cancer diagnosis, within 1 year prior to diagnosis or within 3 months following diagnosis but prior to treatment. BI-RADS density definitions included (1) almost entirely fat (<25% glandular), (2) scattered fibroglandular densities (25%–50% glandular), (3) heterogeneously dense (51%–75% glandular), and (4) extremely dense (>75% glandular). In addition, volumetric percent breast density measurements were obtained for each patient using automated software (Volpara, Volpara Solutions). The mean breast density for all views of the contralateral breast was calculated for each patient because breast cancer would typically increase breast density measurements.

Pathology reports were reviewed to determine tumor type, invasive tumor size, histologic grade, mitotic score (MS), hormone receptor positivity, presence of ductal carcinoma in situ, and lymph node involvement. Lymph node positivity included both macro-metastases (N1 or higher) and micro-matastases (N1mic), but isolated tumor cells (N0i+) were classified as node negative. Molecular subtype was assigned by hormone receptor status, tumor grade, and MS. Subtype categories included: Luminal A (ER/PR+ and grade 1; ER/PR+ and grade 2 and MS = 1; ER+/PR and grade 1); Luminal B (ER+ and grade 3 or MS = 3; ER+/PR and grade 2; ER/PR+ and grade 2 and MS = 2); Her-2+ (ER+ or ER and Her-2+); triple-negative (TN; ER/PR, Her-2; ref. 17 ).

All continuous variables were evaluated by the Kruskal–Wallis test with values being reported as medians with provided interquartile ranges (IQR). Categorical variables were analyzed using Pearson X2 or Fisher exact test as appropriate. Multinomial logistic regression was used to evaluate the association between MD and molecular breast cancer subtype in conjunction with relevant precancer factors including patient age, race, body mass index (BMI), family history of breast cancer, and history of lobular carcinoma in situ (LCIS). Separate regression models were constructed to analyze this relationship using BI-RADS classification as well as continuous volumetric breast density. The presented ORs for volumetric measurements reflect a “per percentage point” increase in breast density. All statistical analyses were performed using SAS statistical software version 9.3 (SAS Institute) with P values <0.05 considered as being significant. Approval from the University of Virginia Institutional Review Board (#HSR 15885) was obtained prior to study initiation.

Of the 604 patients with invasive breast cancer included in the larger study, 462 met study criteria by having sufficient imaging and pathology information for analysis. Five patients were deemed unclassifiable, including 3 patients who were ER but PR+ and 2 patients who had TN cancers that were grade 1 and had an MS = 1.

Of the remaining 457 patients, there were 59 (12.9%) patients with H2P tumors, 233 (51.0%) with Luminal A (LA) tumors, 79 (17.3%) with Luminal B (LB) tumors, and 86 (18.8%) with TN tumors as delineated by the subtype classification scheme (Table 1). There was no significant difference in ER receptor status between those with fatty, scattered, heterogeneously dense, or extremely dense breasts with 62 (79.5%), 144 (77.4%), 117 (75.5%), and 26 (72.2%) having ER-positive tumors, respectively (P = 0.822). There was also no significant difference in the percent volumetric breast density by ER receptor status with the median for ER-negative tumors being 7.71% (IQR, 5.44%–12.61%) and ER-positive tumors being 7.71% (IQR, 5.17%–12.58%), P = 0.766. In addition, there was no significant difference in PR receptor status between BI-RADS MD categories [54 (69.2%) fatty, 119 (64.0%) scattered, 94 (60.7%) heterogeneously dense, and 22 (61.1%) extremely dense] with a P value = 0.624. There was also no significant difference in the percent volumetric density by PR receptor status with the median for PR-negative tumors being 7.69% (IQR, 5.49%–12.42%) and that for PR-positive tumors being 7.78% (IQR, 4.87%–12.86%), P = 0.672. When examining Her-2-neu receptor status by BI-RADS MD categories, the percentage of women with H2P tumors increased with increasing breast density [fatty = 5 (6.4%), scattered = 20 (10.8%), heterogeneously dense = 25 (16.1%), and extremely dense = 9 (25.0%), P = 0.021]. When using volumetric percent density measurements, this was again seen with Her-2-neu–negative tumors having a median percent density of 7.48% (IQR, 4.97%–12.02%) compared with H2P tumors with a median percent density of 10.25% (IQR, 5.96%–16.51%), P = 0.002.

Table 1.

Molecular subtype classification by hormone receptor status, tumor grade, and MS

LALBH2PTN
ER+/PR+ and grade 1 ER+ and grade 3 or MS = 3 Her-2-neu+ ER/PR and Her-2-neu 
ER+/PR+ and grade 2 and MS = 1 ER+/PR and grade 2   
ER+/PR and grade 1 ER+/PR+ and grade 2 and MS = 2   
LALBH2PTN
ER+/PR+ and grade 1 ER+ and grade 3 or MS = 3 Her-2-neu+ ER/PR and Her-2-neu 
ER+/PR+ and grade 2 and MS = 1 ER+/PR and grade 2   
ER+/PR and grade 1 ER+/PR+ and grade 2 and MS = 2   

Patients with H2P tumors were significantly younger with a median age of 54 (IQR, 46–70) compared with women with LA, LB, and TN tumors with median ages of 61 (IQR, 54–70), 58 (IQR, 50–67), and 59 (48–67), respectively (Table 2). They were also more likely to have concomitant invasive and in situ disease (81.4%) compared with those with LA (64%), LB (70.9%), and TN tumors (45.3%; P < 0.001). TN breast cancer was associated with an increased proportion of African American women (26.7% of patients with TN tumors) compared with those with LA (9.9%), LB (16.5%), and H2P (10.2%) tumors (P = 0.002). TN breast cancer was also less commonly associated with a history of LCIS (5.8%) compared with those with LA (20.6%), LB (10.1%), and H2P (13.6%) tumors (P = 0.004). Women with LA tumors were significantly more likely to have had their breast cancer diagnosed by mammography (67%) as opposed to having a palpable mass or other symptoms when compared with women with LB (46.8%), H2P (47.5%), and TN (41.9%) tumors (P = 0.001).

Table 2.

Association between patient factors and molecular breast cancer subtype

Variable (n = 457)Luminal A (n = 233, 51.0%)Luminal B (n = 79, 17.3%)Her-2-Neu (n = 59, 12.9%)Triple-negative (n = 86, 18.8%)Global P value
Age, years (median, IQR) 61 (54–70) 58 (50–67) 54 (46–70) 59 (48–67) 0.006 
BMI, kg/m2 (median, IQR) 27.12 (22.86–30.47) 25.65 (23.03–30.04) 26.00 (22.66–32.56) 28.07 (24.29–32.61) 0.173 
BI-RADS breast density classification (n, %)     0.183 
 Fatty 42 (18.0%) 15 (19.0%) 5 (8.5%) 16 (18.6%)  
 Scattered 103 (44.2%) 27 (34.2%) 20 (33.9%) 36 (41.9%)  
 Heterogeneous 74 (31.8%) 30 (38.0%) 25 (42.4%) 26 (30.2%)  
 Extreme 14 (6.0%) 6 (7.6%) 9 (15.3%) 7 (8.1%)  
Percent volumetric breast density (median, IQR) 7.18 (4.74–11.25) 8.68 (5.68–14.34) 10.25 (5.96–16.51) 7.00 (4.97–11.89) 0.002 
Race (n, %)a     0.002 
 Caucasian 202 (86.7%) 65 (82.3%) 51 (86.4%) 61 (70.9%)  
 African American 23 (9.9%) 13 (16.5%) 6 (10.2%) 23 (26.7%)  
 Other 2 (0.9%) 2 (3.4%)  
Detection method (n, %)a     0.001 
 Mammogram 156 (67.0%) 37 (46.8%) 28 (47.5%) 36 (41.9%)  
 Palpable tumor 58 (24.9%) 31 (39.2%) 23 (39.0%) 40 (46.5%)  
 Other 13 (5.6%) 9 (11.4%) 6 (10.2%) 6 (7.0%)  
Concomitant in situ and invasive disease (n, %) 149 (64.0%) 56 (70.9%) 48 (81.4%) 39 (45.3%) <0.001 
History of LCIS (n, %) 48 (20.6%) 8 (10.1%) 8 (13.6%) 5 (5.8%) 0.004 
History of atypical ductal hyperplasia (n, %) 27 (11.6%) 5 (6.3%) 3 (5.1%) 5 (5.8%) 0.235 
Family history of breast cancer (n, %) 107 (45.9%) 34 (43.0%) 21 (35.6%) 39 (45.3%) 0.625 
Variable (n = 457)Luminal A (n = 233, 51.0%)Luminal B (n = 79, 17.3%)Her-2-Neu (n = 59, 12.9%)Triple-negative (n = 86, 18.8%)Global P value
Age, years (median, IQR) 61 (54–70) 58 (50–67) 54 (46–70) 59 (48–67) 0.006 
BMI, kg/m2 (median, IQR) 27.12 (22.86–30.47) 25.65 (23.03–30.04) 26.00 (22.66–32.56) 28.07 (24.29–32.61) 0.173 
BI-RADS breast density classification (n, %)     0.183 
 Fatty 42 (18.0%) 15 (19.0%) 5 (8.5%) 16 (18.6%)  
 Scattered 103 (44.2%) 27 (34.2%) 20 (33.9%) 36 (41.9%)  
 Heterogeneous 74 (31.8%) 30 (38.0%) 25 (42.4%) 26 (30.2%)  
 Extreme 14 (6.0%) 6 (7.6%) 9 (15.3%) 7 (8.1%)  
Percent volumetric breast density (median, IQR) 7.18 (4.74–11.25) 8.68 (5.68–14.34) 10.25 (5.96–16.51) 7.00 (4.97–11.89) 0.002 
Race (n, %)a     0.002 
 Caucasian 202 (86.7%) 65 (82.3%) 51 (86.4%) 61 (70.9%)  
 African American 23 (9.9%) 13 (16.5%) 6 (10.2%) 23 (26.7%)  
 Other 2 (0.9%) 2 (3.4%)  
Detection method (n, %)a     0.001 
 Mammogram 156 (67.0%) 37 (46.8%) 28 (47.5%) 36 (41.9%)  
 Palpable tumor 58 (24.9%) 31 (39.2%) 23 (39.0%) 40 (46.5%)  
 Other 13 (5.6%) 9 (11.4%) 6 (10.2%) 6 (7.0%)  
Concomitant in situ and invasive disease (n, %) 149 (64.0%) 56 (70.9%) 48 (81.4%) 39 (45.3%) <0.001 
History of LCIS (n, %) 48 (20.6%) 8 (10.1%) 8 (13.6%) 5 (5.8%) 0.004 
History of atypical ductal hyperplasia (n, %) 27 (11.6%) 5 (6.3%) 3 (5.1%) 5 (5.8%) 0.235 
Family history of breast cancer (n, %) 107 (45.9%) 34 (43.0%) 21 (35.6%) 39 (45.3%) 0.625 

aPercentages may not add up to 100% due to rounding or missing data.

When utilizing the BI-RADS classification scheme for MD, there was no statistically significant difference in MD category among the four different molecular subtypes in bivariable analysis (Table 2; P = 0.183); however, 15.3% of patients with H2P tumors had extremely dense breasts by BI-RADS classification, whereas patients with extreme MD accounted for only 6.0% to 8.1% of patients with other breast cancer subtypes. When using volumetric measurements of MD, the mean volumetric breast density for patients with H2P, LA, LB, and TN tumors was 10.25 (IQR, 5.96–16.51), 7.18 (IQR, 4.74–11.25), 8.68 (IQR, 5.68–14.34), and 7.00 (IQR, 4.97–11.89), respectively (P = 0.002). There was no significant difference in patient BMI, history of atypical ductal hyperplasia, or family history of breast cancer in a first- or second-degree relative between molecular subtypes.

Multinomial logistic regression controlling for precancer factors including age, BMI, race, history of LCIS, and family history of breast cancer in a first- or second-degree relative demonstrated that although mean volumetric MD does not significantly differ across all molecular subtypes when compared with LA tumors (P = 0.123), the association between MD and H2P tumors was significantly stronger than the association of MD with LA tumors [OR, 1.06; confidence interval (CI), 1.01–1.12; P = 0.022; Table 3]. TN tumors were found to be significantly associated with African American race when compared with Caucasian patients (OR, 2.90; CI, 1.47–5.70; P = 0.002) and were less likely to be associated with LCIS than LA tumors (OR, 0.25; CI, 0.09–0.67; P = 0.006). In a second multinomial logistic regression utilizing categorical BI-RADS MD where scattered fibroglandular densities (25%–50% glandular) were used as the reference category for MD, the association between MD and H2P tumors when compared with the reference category of LA tumors is again seen among women with extremely dense breasts (OR, 2.98; CI, 1.14–7.83; P = 0.042) but not heterogeneously dense breasts (OR, 1.74; CI, 0.92–3.30; P = 0.273; Table 4).

Table 3.

OR for a specific molecular subtype of breast cancer when assigning mammographic density with automated volumetric software

VariableMolecular subtypetb3fn1OR (95% CI)Overall P valueSubtype P value
Volumetric breast densitya LAb 1.00b  — 
 LB 1.04 (0.99–1.09)  0.162 
 H2P 1.06 (1.01–1.12) 0.123 0.022 
 TN 1.02 (0.97–1.07)  0.489 
Age LAb 1.00b  — 
 LB 0.99 (0.97–1.01)  0.342 
 H2P 0.98 (0.95–1.01) 0.378 0.117 
 TN 0.99 (9.97–1.01)  0.294 
BMI LAb 1.00b  — 
 LB 0.99 (0.94–1.04)  0.650 
 H2P 1.-2 (0.96–1.08) 0.441 0.519 
 TN 1.03 (0.99–1.08)  0.196 
African American race LAb 1.00b  — 
 LB 1.55 (0.71–3.39)  0.269 
 H2P 0.86 (0.32–2.30) 0.010 0.757 
 TN 2.90 (1.47–5.70)  0.002 
History of LCIS LAb 1.00b  — 
 LB 0.42 (0.19–0.95)  0.038 
 H2P 0.60 (0.26–1.39) 0.013 0.232 
 TN 0.25 (0.09–0.67)  0.006 
Family history of breast cancer LAb 1.00b  — 
 LB 0.95 (0.56–1.62)  0.851 
 H2P 0.69 (0.37–1.28) 0.639 0.241 
 TN 1.08 (0.63–1.83)  0.787 
VariableMolecular subtypetb3fn1OR (95% CI)Overall P valueSubtype P value
Volumetric breast densitya LAb 1.00b  — 
 LB 1.04 (0.99–1.09)  0.162 
 H2P 1.06 (1.01–1.12) 0.123 0.022 
 TN 1.02 (0.97–1.07)  0.489 
Age LAb 1.00b  — 
 LB 0.99 (0.97–1.01)  0.342 
 H2P 0.98 (0.95–1.01) 0.378 0.117 
 TN 0.99 (9.97–1.01)  0.294 
BMI LAb 1.00b  — 
 LB 0.99 (0.94–1.04)  0.650 
 H2P 1.-2 (0.96–1.08) 0.441 0.519 
 TN 1.03 (0.99–1.08)  0.196 
African American race LAb 1.00b  — 
 LB 1.55 (0.71–3.39)  0.269 
 H2P 0.86 (0.32–2.30) 0.010 0.757 
 TN 2.90 (1.47–5.70)  0.002 
History of LCIS LAb 1.00b  — 
 LB 0.42 (0.19–0.95)  0.038 
 H2P 0.60 (0.26–1.39) 0.013 0.232 
 TN 0.25 (0.09–0.67)  0.006 
Family history of breast cancer LAb 1.00b  — 
 LB 0.95 (0.56–1.62)  0.851 
 H2P 0.69 (0.37–1.28) 0.639 0.241 
 TN 1.08 (0.63–1.83)  0.787 

NOTE: Adjusted for age, race, volumetric breast density, history of LCIS, and family history of breast cancer in a first- or second-degree relative.

aOR reflects 1 percentage point increase in MD.

bReference category.

Table 4.

OR for a specific molecular subtype of breast cancer when assigning mammographic density with BI-RADS classification scheme

VariableMolecular subtypeOR (95% CI)Overall P valueSubtype P value
BI-RADS classification LAa 1.00a  — 
Fatty (<25%) LB 1.33 (0.68–2.62)  0.405 
 H2P 0.70 (0.29–1.73)  0.444 
 TN 0.73 (0.35–1.50)  0.392 
Scattered (25%–50%) a 1.00a  — 
Heterogeneous (51%–75%) LB 1.21 (0.68–2.14) 0.314 0.516 
 H2P 1.74a (092–3.30)  0.089 
 TN 1.16 (0.65–2.07)  0.605 
Extremely dense (>75%) LB 1.1 (0.36–2.81)  0.985 
 H2P 2.98 (1.14–7.83)  0.027 
 TN 1.54 (0.57–4.16)  0.390 
History of LCIS LAa 1.00a  — 
 LB 0.48 (0.23–1.00)  0.049 
 H2P 0.53 (0.243–1.15) 0.009 0.109 
 TN 0.27 (0.11–0.662)  0.004 
Family history of breast cancer LAa 1.00a  — 
 LB 0.93 (0.58–1.50)  0.760 
 H2P 0.82 (0.48–1.41) 0.909 0.470 
 TN 0.94 (0.58–1.53)  0.797 
African American race LAa 1.00a  — 
 LB 2.03 (1.07–3.86)  0.030 
 H2P 1.29 (0.60–2.80) 0.004 0.515 
 TN 2.89 (1.58–5.28)  0.001 
VariableMolecular subtypeOR (95% CI)Overall P valueSubtype P value
BI-RADS classification LAa 1.00a  — 
Fatty (<25%) LB 1.33 (0.68–2.62)  0.405 
 H2P 0.70 (0.29–1.73)  0.444 
 TN 0.73 (0.35–1.50)  0.392 
Scattered (25%–50%) a 1.00a  — 
Heterogeneous (51%–75%) LB 1.21 (0.68–2.14) 0.314 0.516 
 H2P 1.74a (092–3.30)  0.089 
 TN 1.16 (0.65–2.07)  0.605 
Extremely dense (>75%) LB 1.1 (0.36–2.81)  0.985 
 H2P 2.98 (1.14–7.83)  0.027 
 TN 1.54 (0.57–4.16)  0.390 
History of LCIS LAa 1.00a  — 
 LB 0.48 (0.23–1.00)  0.049 
 H2P 0.53 (0.243–1.15) 0.009 0.109 
 TN 0.27 (0.11–0.662)  0.004 
Family history of breast cancer LAa 1.00a  — 
 LB 0.93 (0.58–1.50)  0.760 
 H2P 0.82 (0.48–1.41) 0.909 0.470 
 TN 0.94 (0.58–1.53)  0.797 
African American race LAa 1.00a  — 
 LB 2.03 (1.07–3.86)  0.030 
 H2P 1.29 (0.60–2.80) 0.004 0.515 
 TN 2.89 (1.58–5.28)  0.001 

NOTE: Adjusted for age, race, BI-RADS MD category, history of LCIS, and family history of breast cancer in a first- or second-degree relative.

aReference category.

This study examines the association between MD using BI-RADS and automated volumetric measurement with molecular subtype of breast cancer utilizing hormone receptor status, tumor grade, MS, and Her-2-neu status to define subtype. Although our study did not demonstrate a statistically significant difference in MD between all four defined molecular subtypes on bivariate analysis, multivariable analysis controlling for other precancer risk factors demonstrated a significant association between H2P tumors and dense breasts when compared with the reference category of LA tumors. Our study also found that TN tumors were associated with African American race and were less likely to be associated with LCIS than LA tumors. There was no significant difference in these findings between BI-RADS versus automated MD classification techniques. Although not the objective of this study, the negative association noted between history of LCIS and TN tumors may reflect the association of LCIS with slower low-grade pathways of tumor development (18 ).

Previous studies examining the association between MD and molecular subtype or receptor status have demonstrated conflicting results. We found no difference between ER or PR hormone receptor status among different breast densities similar to several other studies found in the literature (19,23,). Alternatively, studies by Ding and colleagues and Conroy and colleagues found an association between increased MD and ER+ tumors, whereas a study by Yaghjyan and colleagues found an association between increased MD and ER tumors (9, 24, 25,). In addition, a study by Sartor and colleagues found an association between increased MD and ER tumors, though only among clinically detected cancers instead of those identified on screening (26 ).

Our study did reveal an association between increased MD and Her-2-neu positivity. Although multiple studies did not assess differences in Her-2-neu status, a study by Park and colleagues showed a strong association between increased volumetric breast density and Her-2-neu–positive breast cancer in postmenopausal women but not in premenopausal women (16,). In the study by Seo and colleagues, women with dense breasts more frequently had tumors that overexpressed Her-2-neu than those with less dense breasts, but these results were not statistically significant (27,). Alternatively, studies by Fasching and colleagues and Yaghjyan and colleagues found no difference in Her2 status among MD categories (9, 21,). A meta-analysis of MD and breast cancer according to receptor status expectedly revealed an increased risk of breast cancer in women with increased MD but found no difference among receptor status and MD, though this study is limited by significant heterogeneity between included studies (4 ). These data highlight that high MD appears to be a risk factor for all types of breast cancer without drastic impact on subtype.

Notably, most prior studies have relied on BI-RADS density category classification, which is known to be subjective with only moderate inter- and intra-reader agreement (11,13,). In our study, BI-RADS categories were not associated with breast cancer subtype prior to multivariable analysis. Quantitative automated density software, which has been used in our study and in that of Park and colleagues, demonstrated a significant association between increasing MD and Her-2–positive disease in the bivariable analysis (16 ). An association between Her-2-neu+ tumors and MD measured quantitatively has now been shown in two different populations in the United States and Korea.

Study limitations include using hormone receptor status, grade, and MS as approximate indicators for molecular subtype. Ki67, which has been included in some studies examining molecular subtypes of breast cancer, was not consistently available for our analysis. In addition, having reliable menopausal status data would strengthen the analysis. Strengths of this study include the large number of patients enrolled as well as the utilization of both BI-RADS, which is in widespread clinical use, and quantitative volumetric software, which is more objective and reproducible, to assign MD. This allowed us to assess not only the association between breast density and molecular subtypes of breast cancer but also differences between these methods of ascribing MD. This highlights that the results did not significantly change in our analysis when BI-RADS classification was used, a relevant finding in that BI-RADS classification is much more readily used in contemporary clinical practice.

Our study did not find a significant difference in the association between MD and ER/PR receptor status, though we did detect an association between increased MD and H2P tumors when compared with LA tumors. Our study used both qualitative and quantitative methods for assigning MD. Further studies are needed to assess this relationship as delineating risk factors specific to molecular subtype may promote the future development-individualized risk prediction models and screening strategies.

J.A. Harvey reports receiving commercial research support from, and has an ownership interest in, Volpara Analytics, LLC, and Hologic, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.A. Atkins, G.J. Stukenborg, W.M. Novicoff, W.F. Cohn, J.A. Harvey, A.T. Schroen

Development of methodology: B.L. Edwards, G.J. Stukenborg, W.M. Novicoff, W.F. Cohn, J.A. Harvey, A.T. Schroen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.A. Atkins, W.M. Novicoff, K.N. Larson, W.F. Cohn, J.A. Harvey

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.L. Edwards, K.A. Atkins, G.J. Stukenborg, W.M. Novicoff, K.N. Larson, J.A. Harvey, A.T. Schroen

Writing, review, and/or revision of the manuscript: B.L. Edwards, K.A. Atkins, G.J. Stukenborg, W.M. Novicoff, J.A. Harvey, A.T. Schroen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.L. Edwards, G.J. Stukenborg, K.N. Larson, W.F. Cohn

This research was supported by NIH training grant number CA163177 and Department of Defense for “Building a Better Model: A Personalized Breast Cancer Risk Model Incorporating Breast Density to Stratify Risk and Improve Application of Resources” (Principal Investigator: J.A. Harvey, University of Virginia, Charlottesville, Virginia, Proposal Number BC100474, Award Number W81XWH-11-1-0545, HRPO Log Number A-17074).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
McCormack
VA
,
dos Santos Silva
I
. 
Breast density and parenchymal patterns as markers of breast cancer risk: a meta-analysis
.
Cancer Epidemiol biomarkers Prev
2006
;
15
:
1159
69
.
2.
Boyd
NF
,
Guo
H
,
Martin
LJ
,
Sun
L
,
Stone
J
,
Fishell
E
, et al
Mammographic density and the risk and detection of breast cancer
.
N Engl J Med
2007
;
356
:
227
36
.
3.
Harvey
JA
,
Bovbjerg
VE
. 
Quantitative assessment of mammographic breast density: relationship with breast cancer risk
.
Radiology
2004
;
230
:
29
41
.
4.
Antoni
S
,
Sasco
AJ
,
Dos Santos Silva
I
,
McCormack
V
. 
Is mammographic density differentially associated with breast cancer according to receptor status? A meta-analysis
.
Breast Cancer Res Treat
2013
;
137
:
337
47
.
5.
Arora
N
,
King
TA
,
Jacks
LM
,
Stempel
MM
,
Patil
S
,
Morris
E
, et al
Impact of breast density on the presenting features of malignancy
.
Ann Surg Oncol
2010
;
17
Suppl 3
(
April
):
211
8
.
6.
Wang
AT
,
Vachon
CM
,
Brandt
KR
,
Ghosh
K
. 
Breast density and breast cancer risk: a practical review
.
Mayo Clin Proc
2014
;
89
:
548
57
.
7.
Porter
PL
,
El-Bastawissi
AY
,
Mandelson
MT
,
Lin
MG
,
Khalid
N
,
Watney
EA
, et al
Breast tumor characteristics as predictors of mammographic detection: comparison of interval- and screen-detected cancers
.
J Natl Cancer Inst
1999
;
91
:
2020
8
.
8.
Sihto
H
,
Lundin
J
,
Lehtimäki
T
,
Sarlomo-Rikala
M
,
Bützow
R
,
Holli
K
, et al
Molecular subtypes of breast cancers detected in mammography screening and outside of screening
.
Clin Cancer Res
2008
;
14
:
4103
10
.
9.
Yaghjyan
L
,
Colditz
GA
,
Collins
LC
,
Schnitt
SJ
,
Rosner
B
,
Vachon
C
, et al
Mammographic breast density and subsequent risk of breast cancer in postmenopausal women according to tumor characteristics
.
J Natl Cancer Inst
2011
;
103
:
1179
89
.
10.
American College of Radiology
.
Breast Imaging Reporting and Data System (BI-RADS)
.
Reston, VA
:
American College of Radiology
; 
2003
.
11.
Nicholson
BT
,
LoRusso
AP
,
Smolkin
M
,
Bovbjerg
VE
,
Petroni
GR
,
Harvey
JA
. 
Accuracy of assigned BI-RADS breast density category definitions
.
Acad Radiol
2006
;
13
:
1143
9
.
12.
Ciatto
S
,
Houssami
N
,
Apruzzese
A
,
Bassetti
E
,
Brancato
B
,
Carozzi
F
, et al
Categorizing breast mammographic density: Intra- and interobserver reproducibility of BI-RADS density categories
.
Breast
2005
;
14
:
269
75
.
13.
Freer
PE
. 
Mammographic breast density: impact on breast cancer risk and implications for screening
.
Radiographics
2015
;
35
:
302
15
.
14.
Tagliafico
A
,
Tagliafico
G
,
Tosto
S
,
Chiesa
F
,
Martinoli
C
,
Derchi
LE
, et al
Mammographic density estimation: comparison among BI-RADS categories, a semi-automated software and a fully automated one
.
Breast
2009
;
18
:
35
40
.
15.
Alonzo-proulx
O
,
Mawdsley
GE
,
Patrie
JT
,
Yaffe
MJ
,
Harvey
JA
. 
Reliability of automated breast density measurements
.
Radiology
2015
;
275
:
366
76
.
16.
Park
IH
,
Ko
K
,
Joo
J
,
Park
B
,
Jung
SY
,
Lee
S
, et al
High volumetric breast density predicts risk for breast cancer in postmenopausal, but not premenopausal, Korean Women
.
Ann Surg Oncol
2014
;
21
:
4124
32
.
17.
Gruver
A
,
Portier
B
,
Tubbs
R
. 
Molecular pathology of breast cancer: the journey from traditional practice toward embracing the complexity of molecular classification
.
Arch Pathol Lab Med
2011
;
135
:
544
57
.
18.
Aulmann
S
,
Braun
L
,
Mietzsch
F
,
Longerich
T
,
Penzel
R
,
Schirmacher
P
, et al
Transitions between flat epithelial atypia and low-grade ductal carcinoma in situ of the breast
.
Am J Surg Pathol
2012
;
36
:
1247
52
.
19.
Ma
H
,
Luo
J
,
Press
MF
,
Wang
Y
,
Bernstein
L
,
Ursin
G
. 
Is there a difference in the association between percent mammographic density and subtypes of breast cancer? luminal a and triple-negative breast cancer
.
Cancer Epidemiol Biomarkers Prev
2009
;
18
:
479
85
.
20.
Phipps
AI
,
Buist
DSM
,
Malone
KE
,
Barlow
WE
,
Porter
PL
,
Kerlikowske
K
, et al
Breast density, body mass index, and risk of tumor marker-defined subtypes of breast cancer
.
Ann Epidemiol
2012
;
22
:
340
8
.
21.
Fasching
PA
,
Heusinger
K
,
Loehberg
CR
,
Wenkel
E
,
Lux
MP
,
Schrauder
M
, et al
Influence of mammographic density on the diagnostic accuracy of tumor size assessment and association with breast cancer tumor characteristics
.
Eur J Radiol
2006
;
60
:
398
404
.
22.
Razzaghi
H
,
Troester
MA
,
Gierach
GL
,
Olshan
AF
,
Yankaskas
BC
,
Millikan
RC
. 
Association between mammographic density and basal-like and luminal A breast cancer subtypes
.
Breast Cancer Res
2013
;
15
:
R76
.
23.
Aiello
E
,
Buist
D
,
White
E
,
Porter
P
. 
Association between mammographic breast density and breast cancer tumor characteristics
.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
662
8
. Available from: http://cebp.aacrjournals.org/cgi/content/abstract/14/3/662.
24.
Ding
J
,
Warren
R
,
Girling
A
,
Thompson
D
,
Easton
D
. 
Mammographic density, estrogen receptor status and other breast cancer tumor characteristics
.
Breast J
2010
;
16
:
279
89
.
25.
Conroy
SM
,
Pagano
I
,
Kolonel
LN
,
Maskarinec
G
. 
Mammographic density and hormone receptor expression in breast cancer: the Multiethnic Cohort Study
.
Cancer Epidemiol
2011
;
35
:
448
52
.
26.
Sartor
H
,
Zackrisson
S
,
Elebro
K
,
Hartman
L
,
Borgquist
S
. 
Mammographic density in relation to tumor biomarkers, molecular subtypes, and mode of detection in breast cancer
.
Cancer Causes Control
2015
;
26
:
931
9
.
27.
Seo
BK
,
Pisano
ED
,
Kuzimak
CM
,
Koomen
M
,
Pavic
D
,
Lee
Y
, et al
Correlation of HER-2/neu overexpression with mammography and age distribution in primary breast carcinomas
.
Acad Radiol
2006
;
13
:
1211
8
.