Abstract
Background
Much of the existing literature reporting associations of breast density with breast cancer risk has evaluated the density of the entire breast. However, density is not homogeneously distributed throughout the breast, and tumors are more likely to arise in areas of dense tissue. This study describes novel measures of local breast density and evaluates their association with participant characteristics and risk of breast cancer.
Methods
Among a cohort of women aged 40–65 years who underwent image-guided breast biopsy following an abnormal mammogram from 2007 to 2010 (n = 808), absolute fibroglandular volume (FGV, cm3) and percentage fibroglandular volume (%FGV) of the entire breast (global) and local to the biopsy site (lesional, perilesional) were assessed from pre-biopsy raw digital mammograms using Volpara® software. Associations of participant characteristics with density measures were evaluated using multivariable linear regression models. Associations of density measures with prevalent (baseline) breast cancer and risk of incident (during 10-years of follow-up) breast cancer were evaluated using multivariable logistic and Cox proportional hazard regression models respectively, adjusted for age at mammogram, body mass index, reason for mammogram and self-reported lump prior to mammogram.
Results
Global and local density measures shared similar patterns of associations with participant characteristics. %FGV was not associated with prevalent or risk of incident breast cancer. However, absolute FGV was positively associated with prevalent breast cancer (n = 177 cases; odds ratio (OR) per 0.25 standard deviation (SD) increase for global FGV: OR 2.74, 95% confidence interval (CI) 1.07–7.00; lesional FGV: OR 1.43, 95% CI 1.16–1.76; perilesional FGV: OR 1.11, 95% CI 1.01–1.23). Additionally, local absolute FGV was associated with increased risk of incident breast cancer (n = 31 cases) during follow-up, with a hazard ratio (HR) per 0.25 SD increase for lesional FGV of 2.05, 95% CI 1.53–2.76 and for perilesional FGV of 1.42, 95% CI 1.14–1.77.
Conclusion
Among women biopsied following an abnormal mammogram, higher global and local absolute FGV were associated with prevalent breast cancer. Additionally, elevated absolute FGV local to the biopsy site was associated with increased risk of breast cancer during follow-up. These findings may assist with risk stratification of women undergoing diagnostic breast biopsy.
Supplementary Information
The online version contains supplementary material available at 10.1186/s13058-025-02212-5.
Keywords: Mammographic breast density, Breast cancer risk, Breast pathology, Breast diseases, Benign breast disease
Background
Higher breast density, reflecting the fibroglandular tissue content of the breast, is a strong and independent breast cancer risk factor. Much of the existing literature evaluating breast density-associated breast cancer risk has focused on measurements encompassing the entire breast [1–5]. Dense tissue is heterogeneously distributed throughout the breast [6–8], with cancers frequently developing in dense areas [9–11]. However, a substantial percentage of cancers also develop among women without dense breasts [12]. Evaluating dense tissue at specific, local regions of the breast may provide a measure of risk that complements that of the entire breast. The development of automated, quantitative methods for assessing volumetric breast density from digital mammograms has created opportunities for reliable, reproducible, and high-throughput evaluation of mammographic features beyond the traditional visual assessment of breast density [1, 13, 14].
Prior work to measure local density has used single X-ray absorptiometry (SXA). These studies have shown that higher local density is associated with reduced terminal duct lobular unit (TDLU) involution [15] and increased collagen fiber density [16], tissue features associated with increased breast cancer risk [17–21]. In studies that have retrospectively evaluated local density in the prior mammograms of women diagnosed with breast cancer, local density was shown to be associated with interval cancer and risk of masking [7] and with short-term risk of breast cancer in the same location [9, 11]. However, at present there is a lack of prospective studies evaluating the etiological significance of local breast density and its role as a marker for longer-term breast cancer risk. Therefore, within a cohort of women who underwent image-guided breast biopsy following an abnormal mammogram, this study aimed to measure the local volumetric breast density proximal to the biopsy site (as shown in Fig. 1) using a novel, automated, quantitative approach, and assess its determinants and breast cancer risk associations compared with global measures.
Fig. 1.
A schematic diagram of a mammogram (left) indicating breast segmentation into inner and outer breast areas. The inner breast area is associated with the portion of the breast in contact with the compression paddle, whereas the outer breast area includes the peripheral tissue that extends to the skin line. Local density was assessed for regions (dark grey circles) or parts of regions located within the inner breast area. Among women undergoing clinically-indicated breast biopsy following an abnormal mammogram, we assessed the density at and surrounding the biopsy site (local density) and evaluated relationships with participant characteristics and breast cancer risk. The box (right) illustrates the construction of the lesion and perilesion regions for local volumetric density assessment using the coordinates and radius of the biopsy site. A 4 mm offset was added to the biopsy radius to create the lesion volume. The perilesion region was then defined as a ring-shaped volume with an annular radius of 2 mm surrounding the lesion region.
Methods
Study population
This study included women referred for diagnostic image-guided breast biopsy following an abnormal mammogram at the University of Vermont Medical Center from 2007 to 2010 and enrolled into the National Cancer Institute’s (NCI) Breast Radiology Evaluation and Study of Tissues (BREAST) Stamp Project, a molecular epidemiological study of breast density [3, 15, 22, 23]. Eligible women (n = 1,227) were women aged 40–65 years who did not have a personal history of breast cancer, had not had breast surgery within the previous year, had not had breast implants, and were not taking breast cancer chemoprevention. At enrollment, study participants (n = 1,011) provided detailed demographic and risk factor information [24]. Participants were followed for a maximum of 10 years from the date of study enrollment for incident in situ and invasive breast cancer cases, which were reported to the Vermont Cancer Registry and Vermont Breast Cancer Surveillance System pathology database [25]. The Institutional Review Boards at the NCI and the University of Vermont approved the protocol for this project for either active consenting or a waiver of consent to enroll participants, link data and perform analytical studies.
Analytic population
The analytic population was selected from those who had pre-biopsy, cranio-caudal (CC), raw (For Processing Digital Imaging and Communications in Medicine (DICOM) Presentation Intent Type) digital mammograms available for breast density assessment. Women were excluded if they did not undergo a biopsy (n = 12), if they had only medio-lateral oblique (MLO) mammogram views available (n = 44), if their raw mammographic images were magnification view and thus could not be assessed by the Volpara® software (n = 1) or if there were no raw mammographic images available (n = 146). The final analytic population consisted of 808 women who had a total of 973 biopsies at enrollment (baseline). Of these women, 159 had biopsies at multiple locations: 154 women had two biopsies, four women had three biopsies, and one woman had four biopsies. The mean (standard deviation; SD) time from pre-biopsy mammogram to baseline breast biopsy was 21 (38) days.
Exposure assessment
At enrollment, detailed information on demographic, reproductive and lifestyle factors and medical history was collected from participants via a questionnaire administered by a research nurse [3, 24]. Radiological and pathological factors relevant to the eligible pre-biopsy mammogram and subsequent biopsy or biopsies were captured from participants’ medical records by the study coordinator. Relevant covariates for this analysis included age at mammogram (categorized for analysis as 40–44, 45–49, 50–54, 55–59, 60–65 years), race and ethnicity (non-Hispanic White, all other racial and ethnic groups), education (< high school or high school graduate, college/graduate school degree), body mass index (BMI; categorized for analysis as < 25, 25- < 30, ≥ 30 kg/m2), age at first birth (nulliparous, < 30 years, ≥ 30 years), menopausal status (pre- vs. post-menopausal), age at menopause (≤ 45, 46–49, 50–52, ≥ 53 years), benign breast biopsy prior to enrollment (never, ever), family history of breast cancer in a first-degree female relative (no, yes), self-report of lump in the three months prior to mammogram (no, yes), reason for mammography (screening, follow-up or evaluation of breast problem), biopsy type (ultrasound-guided, stereotactic-guided, other), Breast Imaging Reporting and Data System (BI-RADS) 4th edition density assessment (entirely fat, scattered densities, heterogeneously dense, extremely dense), and overall most severe biopsy diagnosis (benign/non-proliferative, proliferative without atypia, proliferative with atypia, lobular carcinoma in situ (LCIS), carcinoma in situ (ductal carcinoma in situ (DCIS), combined DCIS/LCIS, unspecified carcinoma in situ), invasive carcinoma).
Follow-up and breast cancer ascertainment
Following the baseline mammogram and biopsy (or biopsies), participants received either benign (n = 631) or breast cancer (prevalent cases; n = 177) diagnoses. Benign diagnoses included benign or non-proliferative breast disease, proliferative disease without atypia, proliferative disease with atypia, and LCIS. Breast cancer was defined as a diagnosis of invasive or in situ carcinoma. Participants with benign diagnoses at baseline were followed for a maximum of 10 years from the date of study enrollment to ascertain incident breast cancer diagnoses (incident cases; n = 31); recurrent or second primary breast cancers (n = 20), i.e., breast cancers diagnosed during follow-up among women who had a baseline cancer diagnosis, were not included. The median (range) time from baseline mammogram to breast cancer diagnosis among prevalent cases was 8 (0–216) days and among incident cases was 4.4 (0.6–9.1) years.
Assessment of global and local density
Global and local density were assessed on the raw, CC view mammograms of the ipsilateral (biopsied) breast using commercially available, automated volumetric breast density software (Volpara Imaging Software version 3.1, Volpara Health Technologies Ltd, Wellington, New Zealand). The Volpara software uses a relative physics-based model to generate a Density Map from the raw image. This Density Map represents the fibroglandular volume at each image pixel location and is computed by comparing the pixel intensity to a reference fat point (assumed to be composed entirely of adipose tissue), as described previously [27].
Global breast density was assessed as the absolute fibroglandular volume (FGV, cm3) and as the percent fibroglandular volume (%FGV, i.e., the absolute FGV divided by the total breast volume, expressed as a percentage) for the entire breast.
Local density was assessed by customizing the outputs from Volpara Density Maps to evaluate the fibroglandular volume (FGV, %FGV) of restricted lesional and perilesional regions based on the biopsy coordinates and diameter annotated on the ‘For Presentation’ pre-biopsy mammograms by the study radiologists (JMJ, SDH). As shown in Fig. 1, the lesional region was defined as a circular region centered on the biopsy coordinates and with a radius equal to that of the biopsy plus a 4 mm offset. The perilesional region was defined as an annulus (ring-shaped region) with a width of 2 mm surrounding the lesional region.
A key input into the Volpara algorithm is the compressed breast thickness, which is extracted from the DICOM image meta-data [27]. The algorithm segments the breast into inner and outer breast areas (Fig. 1). The inner area defines the region of the breast that is assumed to be in direct contact with the compression plates of the mammography equipment. The outer breast area indicates the periphery of the breast that is not in direct contact and where the breast thickness varies towards the skin line. For global measures, the Volpara algorithm models the average thickness of the outer breast, but as the algorithm does not output a pixel-by-pixel estimate of tissue thickness in this area for research use, local density cannot be calculated in the outer breast area from the Density Map. For this study, the assessment of local density in the outer breast was derived using the compressed thickness, which may have affected the robustness of these estimates. Therefore, we focused our primary analysis on local density data for the inner breast area, and we carried out sensitivity analyses using local density measures for the entire breast (encompassing both inner and outer areas).
Statistical analyses
The characteristics of the analytic population were described as frequencies and percentages. Statistical differences in participant characteristics were evaluated using Chi-squared tests. All other analyses were performed at the biopsy level. All density measures were square-root transformed to better approximate normal distributions. To evaluate and compare determinants of local density (lesional %FGV, lesional FGV (cm3), perilesional %FGV, perilesional FGV (cm3)) and global density (%FGV, FGV (cm3)) in this population, associations of participant characteristics with each density measure were assessed using generalized estimating equation (GEE) linear regression models with an independent correlation structure to account for within-woman correlations of multiple biopsies [28]. To inform the construction of the multivariable linear regression models, we first examined univariate associations of participant characteristics with local and global density measures. We evaluated associations of BMI with density measures overall and stratified by menopausal status. To formally test if associations of BMI with global and local density varied by menopausal status, we compared models with and without an interaction term for BMI and menopausal status. Fully-adjusted models included age at mammogram and BMI as ordinal trends, education, menopausal status, history of breast biopsy, self-report of lump in the three months prior to mammogram, and an interaction term for BMI and menopausal status. Beta coefficients (β), i.e., slope estimates, of all GEE linear models indicated the average change of the density measure per unit change of the independent variable. For nominal variables, we reported P-values from Chi-squared tests, and for ordinal variables, we reported P-values from Wald tests using ordinal categories of the variable. Analyses were stratified by baseline biopsy diagnosis (benign or cancer).
Odds ratios (ORs) and 95% confidence intervals (95% CIs) for associations of global and local density (%FGV, FGV (cm3)) with breast cancer at baseline (prevalent cases) were evaluated in multivariable logistic regression models. Hazard ratios (HRs) and 95% CIs for associations of density measures with risk of breast cancer during follow-up (incident cases) were evaluated in multivariable Cox proportional hazard regression models; in sensitivity analyses, we excluded incident cases diagnosed within two years of the baseline mammogram. Additionally, as we did not observe statistically significant heterogeneity of estimates comparing prevalent and incident cases in case-case analyses using logistic regression models, we evaluated associations of global and local density with risk of breast cancer for all cases combined using HRs and 95% CIs from Cox proportional hazards models, in which the time from baseline mammogram to baseline biopsy accounted for the follow-up period for prevalent cases. Based on univariate associations, all models were adjusted for age at mammogram in five-year groups as an ordinal trend, BMI as an ordinal trend, reason for mammogram and self-report of breast lump in the three months prior to the baseline mammogram. We additionally considered adjustment for lesion radius; however, its inclusion had minimal impact on risk estimates and so the results presented are from models not adjusted for lesion radius. To evaluate possible effect modification by age, we compared models using time since baseline mammogram as the scale and models using age as the scale (accounting for delayed entry), observing some differences in hazard ratio estimates. To explore this further and identify possible age interactions, we fit separate models to women younger or older than the median age at mammogram) using time since mammogram as the underlying scale, and found that for global density measures, associations were stronger among older women, and for local density measures, associations were stronger among younger women. Therefore, to best control for age, we present results from Cox proportional hazards models using age in years as the underlying time scale and additionally adjusted for age at mammogram as an ordinal trend. Models used an independent working correlation structure to account for within-woman correlations of multiple biopsies [28]. Global and local density measures were modeled separately as continuous (per 5% increase for %FGV and per 0.25 standard deviation increase for absolute FGV) or categorical variables. For categorical density measures, quartiles were constructed using the density distributions among women with benign diagnoses at baseline.
P-values testing for linear trends were corrected for multiple testing using a Bonferroni adjustment in accounting for the total number of tests per density measure for each outcome type. Analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, North Carolina, USA). All P-values were two-sided and considered statistically significant if p < 0.05.
Results
Characteristics of participants at baseline
The median (range) age of participants at mammogram was 50 (40–65) years (Table 1). Women with benign diagnoses were younger than those with in situ or invasive cancer diagnosed at baseline (median (range) ages: 49 (40–65) years and 54 (40–65) years, respectively; p < 0.001). Most participants were Non-Hispanic White (91.5%) and had a college or graduate degree (79.8%). Approximately half (50.9%) of the population had a BMI < 25 kg/m2. 29.6% reported a breast biopsy prior to study enrollment and 21.5% had a first-degree family history of breast cancer. Most participants (79.6%) reported screening as the reason for mammography, while 13% of participants reported a palpable lump in the three months prior to their mammogram. At baseline, 40.3% of women with benign diagnoses had heterogeneously or extremely dense breasts compared to 29.4% of women with cancer diagnoses (p = 0.045), possibly due to the younger age distribution of women with benign diagnoses. Benign diagnoses on biopsy (benign/non-proliferative, proliferative without atypia, proliferative with atypia, LCIS) were reported in 78.1% of participants. The remaining women (22%) had biopsy diagnoses of invasive or in situ carcinoma and were thus classed as prevalent cases.
Table 1.
Characteristics of women who underwent diagnostic biopsy and had mammographic images eligible for breast density assessment by Volpara®, overall and by baseline biopsy diagnosisa
| Characteristic | Total (n = 808) | Baseline biopsy diagnosis | P-valuee | |
|---|---|---|---|---|
| Benignb (n = 631) | Cancerc (n = 177) | |||
| n (%)d | n (%)d | n (%)d | ||
| Age at mammogram (years) | < 0.001 | |||
| 40–44 | 164 (20.3) | 145 (23) | 19 (10.7) | |
| 45–49 | 203 (25.1) | 171 (27.1) | 32 (18.1) | |
| 50–54 | 193 (23.9) | 153 (24.2) | 40 (22.6) | |
| 55–59 | 137 (17.0) | 92 (14.6) | 45 (25.4) | |
| 60–65 | 111 (13.7) | 70 (11.1) | 41 (23.2) | |
| Median (Range) | 50 (40–65) | 49 (40–65) | 54 (40–65) | |
| Race | 0.210 | |||
| White, non-Hispanic | 739 (91.5) | 573 (90.8) | 166 (93.8) | |
| Other | 69 (8.5) | 58 (9.2) | 11 (6.2) | |
| Education level | 0.373 | |||
| ≤ High school | 140 (17.3) | 105 (16.6) | 35 (19.8) | |
| College or graduate degree | 645 (79.8) | 506 (80.2) | 139 (78.5) | |
| Body mass index (BMI), kg/m2 | 0.247 | |||
| < 25 | 411 (50.9) | 325 (51.5) | 86 (48.6) | |
| 25- < 30 | 197 (24.4) | 157 (24.9) | 40 (22.6) | |
| > 30 | 193 (23.9) | 142 (22.5) | 51 (28.8) | |
| Median (Range) | 24.9 (15.9–62.3) | 24.8 (15.9–62.3) | 25.4 (18.0–50.3) | |
| Age at first birth (years) | 0.255 | |||
| Nulliparous | 177 (21.9) | 137 (21.7) | 40 (22.6) | |
| < 30 | 450 (55.7) | 360 (57.1) | 90 (50.8) | |
| ≥ 30 | 177 (21.9) | 131 (20.8) | 46 (26.0) | |
| Menopausal status | < 0.001 | |||
| Premenopausal | 450 (55.7) | 383 (60.7) | 67 (37.9) | |
| Postmenopausal | 327 (40.5) | 226 (35.8) | 101 (57.1) | |
| Breast biopsy prior to enrollment | 0.096 | |||
| Never | 560 (69.3) | 427 (67.7) | 133 (75.1) | |
| Ever | 239 (29.6) | 195 (30.9) | 44 (24.9) | |
| Family history of breast cancer in a first-degree female relative | 0.263 | |||
| 0 | 606 (75.0) | 477 (75.6) | 129 (72.9) | |
| ≥ 1 | 174 (21.5) | 130 (20.6) | 44 (24.9) | |
| Self-reported lump in 3 months prior to mammogram | < 0.001 | |||
| No | 695 (86.0) | 556 (88.1) | 139 (78.5) | |
| Yes | 105 (13.0) | 68 (10.8) | 37 (20.9) | |
| Reason for mammography | 0.016 | |||
| Screening | 643 (79.6) | 514 (81.5) | 129 (72.9) | |
| Follow-up or evaluation of breast problem | 160 (19.8) | 113 (17.9) | 47 (26.6) | |
| BI-RADS density assessment | 0.045 | |||
| Entirely fat | 125 (15.5) | 96 (15.2) | 29 (16.4) | |
| Sattered densities | 361 (44.7) | 269 (42.6) | 92 (52.0) | |
| Heterogeneously dense | 256 (31.7) | 210 (33.3) | 46 (26.0) | |
| Extremely dense | 50 (6.2) | 44 (7.0) | 6 (3.4) | |
| Overall worst diagnosis at baselinef | ||||
| Benign/non-proliferative | 266 (32.9) | 266 (42.2) | ||
| Proliferative without atypia | 305 (37.8) | 305 (48.3) | ||
| Proliferative with atypia | 52 (6.4) | 52 (8.2) | ||
| LCIS | 8 (1.0) | 8 (1.3) | ||
| In situ (DCIS, mixed DCIS/LCIS, unspecified) | 61 (7.6) | 61 (34.5) | ||
| Invasive carcinoma | 116 (14.4) | 116 (65.5) | ||
aEligible mammographic images were pre-biopsy, cranio-caudal (CC), non-magnification view, raw (For Processing DICOM Presentation Intent Type) digital mammograms of the biopsied breast. bBenign diagnoses comprised the pathological findings benign/non-proliferative, proliferative without atypia, proliferative with atypia, and lobular carcinoma in situ (LCIS). cCancer diagnoses comprised the pathological findings in situ (ductal carcinoma in situ (DCIS), mixed DCIS/LCIS, in situ unspecified) and invasive breast carcinoma. dPercentages may not sum to 100 due to rounding. eP-values obtained from Chi-squared tests excluding missing categories. P-values < 0.05 were considered significant and are indicated in the table in bold font. fn = 159 women (19.7%) had more than one biopsy.
Distributions of global and local density measures
The distributions of the Volpara global density measures and the local density measures from the inner breast area are shown in Fig. 2, and the distributions of the local density measures from the entire breast (i.e., both inner and outer areas) are shown in Supplemental Fig. 1. All distributions were positively skewed.
Fig. 2.
Distributions of global (density from entire breast area) and local (regions or portions of regions within inner breast area only) Volpara density measures among the analytic population (n = 973 biopsies in 808 women).
Associations of participant characteristics with % fibroglandular global and local density measures
As shown in Table 2, age was inversely associated with all %FGV measures among women with benign lesions at baseline (all p < 0.001); however, there was no clear pattern of association among women diagnosed with cancer at baseline. All %FGV measures decreased with increasing BMI, irrespective of baseline biopsy diagnosis (all p < 0.001, except for lesional %FGV pcancer = 0.006). Among women with benign lesions, a history of breast biopsy showed positive associations with global and lesional %FGV measures (global %FGV pbenign = 0.002; lesional %FGV pbenign = 0.017; perilesional %FGV pbenign = 0.039), findings that were not observed among women with cancer at baseline. Global breast density, measured as BI-RADS density (all p < 0.001) or as Volpara global %FGV (all p < 0.001) and absolute FGV (all p ≤ 0.001, except lesional FGV pcancer = 0.006) was positively associated with all local %FGV measures (Table 2).
Table 2.
Associations of global and local Volpara % fibroglandular volume with participant characteristics, by baseline biopsy diagnosis (n = 973 biopsies in 808 women)a
| Characteristic | Global % fibroglandular volume | Lesional % fibroglandular volumeb | Perilesional % fibroglandular volumeb | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Benign (n = 736) | Cancer (n = 237) | Benign (n = 665) | Cancer (n = 214) | Benign (n = 693) | Cancer (n = 220) | |||||||||||||
| β | SE | P-valuec | β | SE | P-value c | β | SE | P-valuec | β | SE | P-value c | β | SE | P-valuec | β | SE | P-valuec | |
| Age at mammogram, years | < 0.001 | 1.000 | < 0.001 | 1.000 | < 0.001 | 1.000 | ||||||||||||
| 40–44 | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| 45–49 | − 0.12 | 0.10 | 0.03 | 0.24 | − 0.27 | 0.16 | 0.40 | 0.38 | − 0.23 | 0.16 | 0.30 | 0.39 | ||||||
| 50–54 | − 0.30 | 0.11 | 0.29 | 0.21 | − 0.44 | 0.16 | 0.86 | 0.35 | − 0.42 | 0.16 | 0.72 | 0.35 | ||||||
| 55–59 | − 0.62 | 0.14 | 0.11 | 0.24 | − 0.86 | 0.20 | 0.20 | 0.41 | − 0.89 | 0.20 | 0.33 | 0.40 | ||||||
| 60–65 | − 0.91 | 0.14 | − 0.02 | 0.23 | − 1.35 | 0.22 | − 0.12 | 0.42 | − 1.17 | 0.22 | 0.09 | 0.39 | ||||||
| Education level | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.190 | ||||||||||||
| ≤ High school | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| College or graduate degree | 0.12 | 0.07 | 0.22 | 0.12 | 0.20 | 0.12 | 0.36 | 0.18 | 0.20 | 0.13 | 0.50 | 0.17 | ||||||
| Body mass index (BMI), kg/m2 | < 0.001 | < 0.001 | < 0.001 | 0.006 | < 0.001 | < 0.001 | ||||||||||||
| < 25 | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| 25– < 30 | − 0.95 | 0.08 | − 0.88 | 0.16 | − 1.17 | 0.13 | − 0.81 | 0.28 | − 1.18 | 0.13 | − 1.05 | 0.26 | ||||||
| > 30 | − 1.56 | 0.10 | − 1.26 | 0.19 | − 1.82 | 0.17 | − 1.23 | 0.33 | − 1.85 | 0.17 | − 1.44 | 0.32 | ||||||
| Age at first birth, years | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| Nulliparous | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| < 30 | − 0.06 | 0.09 | − 0.13 | 0.16 | − 0.03 | 0.14 | − 0.33 | 0.27 | − 0.05 | 0.13 | − 0.28 | 0.26 | ||||||
| ≥ 30 | 0.14 | 0.11 | − 0.11 | 0.18 | 0.20 | 0.16 | − 0.43 | 0.29 | 0.16 | 0.16 | − 0.39 | 0.29 | ||||||
| Menopausal status | 0.490 | 0.036 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| Premenopausal | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Postmenopausal | − 0.31 | 0.13 | − 0.73 | 0.22 | − 0.32 | 0.18 | − 0.21 | 0.34 | − 0.35 | 0.19 | − 0.56 | 0.32 | ||||||
| Breast biopsy prior to enrollment | 0.002 | 1.000 | 0.017 | 1.000 | 0.039 | 1.000 | ||||||||||||
| Never | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Ever | 0.30 | 0.08 | 0.14 | 0.14 | 0.39 | 0.11 | 0.06 | 0.20 | 0.36 | 0.11 | 0.11 | 0.20 | ||||||
| Family history of breast cancer in a 1st degree female relative | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| No | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Yes | 0.03 | 0.08 | 0.12 | 0.15 | − 0.09 | 0.13 | − 0.19 | 0.23 | − 0.08 | 0.13 | − 0.20 | 0.21 | ||||||
| Self-reported lump in 3 months prior to mammogram | 0.010 | 0.029 | 0.039 | 0.017 | 0.003 | 0.080 | ||||||||||||
| No | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Yes | 0.42 | 0.11 | 0.48 | 0.13 | 0.77 | 0.17 | 0.87 | 0.23 | 0.73 | 0.17 | 0.75 | 0.23 | ||||||
| Reason for mammogram | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| Screening | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Breast symptoms or follow-up visit | 0.02 | 0.10 | − 0.18 | 0.20 | 0.03 | 0.17 | 0.07 | 0.27 | 0.08 | 0.17 | − 0.11 | 0.25 | ||||||
| BI-RADS density assessment | < 0.001 | < 0.001 | < 0.001 | < 0.001 | < 0.001 | < 0.001 | ||||||||||||
| Entirely fat | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Scattered fibroglandular densities | 0.54 | 0.06 | 0.42 | 0.08 | 0.89 | 0.13 | 0.76 | 0.20 | 0.88 | 0.11 | 0.73 | 0.18 | ||||||
| Heterogeneously dense | 1.36 | 0.08 | 1.35 | 0.12 | 2.01 | 0.16 | 1.78 | 0.26 | 2.09 | 0.14 | 1.94 | 0.23 | ||||||
| Extremely dense | 2.28 | 0.14 | 2.21 | 0.27 | 2.49 | 0.23 | 2.15 | 0.29 | 2.51 | 0.22 | 2.18 | 0.29 | ||||||
| Volpara global density | ||||||||||||||||||
| %FGV | 1.13 | 0.04 | < 0.001 | 1.09 | 0.09 | < 0.001 | 1.16 | 0.04 | < 0.001 | 1.14 | 0.08 | < 0.001 | ||||||
| FGV (cm3) | 0.18 | 0.02 | < 0.001 | 0.16 | 0.04 | 0.006 | 0.20 | 0.02 | < 0.001 | 0.17 | 0.04 | 0.001 | ||||||
aCoefficients are from multivariable linear regression models adjusted for age at baseline mammogram, BMI, education, menopausal status, history of breast biopsies, self-reported lump in 3 months prior to mammogram and BMI-menopausal status interaction. Density measures were square-root transformed. N values refer to the number of biopsies; some women had multiple biopsies. Correlations due to multiple biopsies for the same woman were accounted for in the variance calculation. b% fibroglandular volume was calculated by dividing the fibroglandular volume (cm3) by the overall volume (cm3) of the lesion or perilesion region and expressing as a percentage. cFor nominal variables, reported p-values are from χ2 test. For ordinal variables, reported p-values are from Wald tests using ordinal categories of the variable. All p-values are Bonferroni corrected for multiple testing by multiplying model p-values by the number of tests. P-values < 0.05 were considered significant and are indicated in the table in bold font.
Associations of participant characteristics with absolute fibroglandular global and local density measures
Among women with benign diagnoses, age was inversely associated with all absolute FGV measures (global FGV pbenign = 0.024; lesional FGV pbenign = 0.002; perilesional pbenign < 0.001; Table 3), and no associations were observed among women with cancer at baseline. Among women with benign biopsies, BMI was inversely associated with perilesional FGV (p < 0.001), but not lesional or global FGV. BI-RADS global density was positively associated with lesional FGV among women with benign diagnoses at baseline and with perilesional FGV regardless of baseline biopsy diagnosis (all p < 0.001). Volpara global %FGV (all p < 0.001, except lesional FGV pcancer = 0.017) or absolute global FGV (all p < 0.001, except lesional FGV pcancer = 0.048) were positively associated with all local FGV measures regardless of baseline biopsy diagnosis (Table 3).
Table 3.
Associations of global and local Volpara absolute fibroglandular volume with participant characteristics, by baseline biopsy diagnosis (n = 973 biopsies in 808 women)a
| Characteristic | Global fibroglandular volume (cm3) | Lesional fibroglandular volume (cm3) | Perilesional fibroglandular volume (cm3) | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Benign (n = 736) | Cancer (n = 237) | Benign (n = 665) | Cancer (n = 214) | Benign (n = 693) | Cancer (n = 220) | |||||||||||||
| β | SE | P-valueb | β | SE | P-valueb | β | SE | P-valueb | β | SE | P-valueb | β | SE | P-valueb | β | SE | P-valueb | |
| Age at mammogram, years | 0.024 | 1.000 | 0.002 | 1.000 | < 0.001 | 1.000 | ||||||||||||
| 40–44 | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| 45–49 | − 0.51 | 0.27 | 1.00 | 0.80 | − 0.22 | 0.10 | 0.54 | 0.26 | − 0.08 | 0.04 | 0.28 | 0.11 | ||||||
| 50–54 | − 0.40 | 0.38 | 1.08 | 0.84 | − 0.21 | 0.11 | 0.26 | 0.26 | − 0.07 | 0.05 | 0.17 | 0.11 | ||||||
| 55–59 | − 1.21 | 0.40 | 1.66 | 0.79 | − 0.33 | 0.14 | 0.27 | 0.28 | − 0.16 | 0.06 | 0.23 | 0.12 | ||||||
| 60–65 | − 1.74 | 0.43 | 1.14 | 0.82 | − 0.64 | 0.13 | 0.13 | 0.28 | − 0.30 | 0.06 | 0.20 | 0.12 | ||||||
| Education level | 1.000 | 0.310 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| ≤ High school | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| College or graduate degree | 0.06 | 0.21 | 0.99 | 0.37 | 0.04 | 0.08 | 0.07 | 0.14 | 0.04 | 0.04 | 0.04 | 0.06 | ||||||
| Body mass index (BMI), kg/m2 | 0.300 | 1.000 | 1.000 | 1.000 | < 0.001 | 1.000 | ||||||||||||
| < 25 | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| 25- < 30 | 0.67 | 0.25 | 0.33 | 0.54 | − 0.07 | 0.09 | 0.16 | 0.19 | − 0.13 | 0.04 | 0.01 | 0.08 | ||||||
| > 30 | 0.87 | 0.35 | 1.17 | 0.72 | − 0.21 | 0.11 | 0.06 | 0.23 | − 0.21 | 0.05 | − 0.03 | 0.11 | ||||||
| Age at first birth, years | 1.000 | 1.000 | 1.000 | 0.500 | 1.000 | 1.000 | ||||||||||||
| Nulliparous | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| < 30 | − 0.32 | 0.23 | − 0.51 | 0.50 | 0.07 | 0.08 | − 0.40 | 0.20 | 0.00 | 0.03 | − 0.13 | 0.07 | ||||||
| ≥ 30 | 0.28 | 0.31 | − 0.66 | 0.53 | 0.14 | 0.10 | − 0.49 | 0.21 | 0.06 | 0.04 | − 0.15 | 0.08 | ||||||
| Menopausal status | 1.000 | 0.050 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| Premenopausal | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Postmenopausal | − 0.55 | 0.39 | − 2.20 | 0.63 | − 0.03 | 0.11 | − 0.09 | 0.24 | − 0.03 | 0.05 | − 0.08 | 0.10 | ||||||
| Breast biopsy prior to enrollment | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| Never | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Ever | 0.31 | 0.21 | − 0.01 | 0.39 | 0.07 | 0.07 | − 0.20 | 0.14 | 0.05 | 0.03 | − 0.01 | 0.06 | ||||||
| Family history of breast cancer in a 1st degree female relative | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| No | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Yes | 0.11 | 0.20 | 0.75 | 0.47 | − 0.01 | 0.07 | − 0.02 | 0.20 | 0.00 | 0.03 | − 0.06 | 0.08 | ||||||
| Self-reported lump in 3 months prior to mammogram | 1.000 | 0.530 | 0.170 | 0.031 | 0.110 | 0.750 | ||||||||||||
| No | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Yes | 0.87 | 0.42 | 1.12 | 0.45 | 0.45 | 0.15 | 0.58 | 0.16 | 0.19 | 0.06 | 0.16 | 0.07 | ||||||
| Reason for mammogram | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | ||||||||||||
| Screening | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Breast symptoms or follow-up visit | − 0.25 | 0.34 | − 0.72 | 0.66 | 0.03 | 0.13 | 0.52 | 0.34 | − 0.02 | 0.05 | 0.06 | 0.12 | ||||||
| BI-RADS density assessment | < 0.001 | < 0.001 | < 0.001 | 0.42 | < 0.001 | < 0.001 | ||||||||||||
| Entirely fat | Ref | Ref | Ref | Ref | Ref | Ref | ||||||||||||
| Scattered fibroglandular densities | 1.40 | 0.22 | 1.29 | 0.34 | 0.28 | 0.09 | 0.37 | 0.15 | 0.18 | 0.04 | 0.20 | 0.06 | ||||||
| Heterogeneously dense | 2.65 | 0.27 | 3.25 | 0.52 | 0.56 | 0.10 | 0.46 | 0.20 | 0.40 | 0.04 | 0.35 | 0.08 | ||||||
| Extremely dense | 3.83 | 0.66 | 4.41 | 1.05 | 0.60 | 0.17 | 1.00 | 0.54 | 0.39 | 0.07 | 0.66 | 0.19 | ||||||
| Volpara global density | ||||||||||||||||||
| %FGV | 0.30 | 0.04 | < 0.001 | 0.39 | 0.11 | 0.017 | 0.19 | 0.02 | < 0.001 | 0.23 | 0.04 | < 0.001 | ||||||
| FGV (cm3) | 0.12 | 0.02 | < 0.001 | 0.12 | 0.04 | 0.048 | 0.06 | 0.01 | < 0.001 | 0.07 | 0.01 | < 0.001 | ||||||
aCoefficients are from multivariable linear regression models adjusted for age at baseline mammogram, BMI, education, menopausal status, history of breast biopsies, self-reported lump in 3 months prior to mammogram and BMI-menopausal status interaction. Density measures were square-root transformed. N values refer to the number of biopsies; some women had multiple biopsies. Correlations due to multiple biopsies for the same woman were accounted for in the variance calculation. bFor nominal variables, reported p-values are from χ2 test. For ordinal variables, reported p-values are from Wald tests using ordinal categories of the variable. All p-values are Bonferroni corrected for multiple testing by multiplying model p-values by the number of tests. P-values < 0.05 were considered significant and are indicated in the table in bold font.
Association of global and local density measures with prevalent breast cancer
No statistically significant associations were observed for global, lesional or perilesional %FGV with prevalent breast cancer (Fig. 3a).
Fig. 3.
a Association of global and local Volpara % fibroglandular volume (%FGV) with risk of breast cancer at baseline and over 10 years of study follow-up among women biopsied following an abnormal mammogram (n = 973 biopsies in n = 808 women)a. b: Association of global and local Volpara fibroglandular volume (FGV; cm3) with risk of breast cancer at baseline and over 10 years of study follow-up among women biopsied following an abnormal mammogram (n = 973 biopsies in 808 women)a.
a Risk estimates are presented as odds ratios (ORs) with 95% confidence intervals (CIs) from multivariable logistic regression models (prevalent cases) and hazard ratios (HRs) with 95% CIs from multivariable Cox proportional hazards regression models (incident cases and all cases). All models were adjusted for age at baseline mammogram, BMI, reason for mammogram, and self-report of breast lump in the 3 months prior to baseline mammogram. Cox proportional hazards models used age as the timescale. Density measures were square-root transformed. Correlations due to multiple biopsies for the same woman were accounted for in the variance calculation.
As shown in Fig. 3b, global, lesional and perilesional absolute FGV (per 0.25 SD increase) were significantly associated with prevalent cancer (global FGV: OR 2.74, 95% CI 1.07–7.00; lesional FGV: OR 1.43, 95% CI 1.16–1.76; perilesional FGV: OR 1.11, 95% CI 1.01–1.23). Lesional absolute FGV was also positively associated with prevalent cancer for the highest compared to the lowest quartile of FGV (OR Q4vsQ1 2.08, 95% CI 1.44–3.48; Bonferroni corrected ptrend = 0.03; Fig. 3b).
Association of global and local density measures with risk of incident breast cancer
We did not observe statistically significant associations for global, lesional or perilesional %FGV with risk of incident breast cancer (Fig. 3a).
For absolute FGV (Fig. 3b), elevated lesional and perilesional measures were statistically significantly associated with increased risk of incident cancer for the highest compared with the lowest quartile of FGV (lesional FGV: HRQ4vsQ1 4.08, 95% CI 1.50–11.05, Bonferroni corrected ptrend = 0.04; perilesional HRQ4vsQ1 4.28, 95% CI 1.44–12.71, Bonferroni corrected ptrend = 0.03; Fig. 3b) and per 0.25 SD increase (lesional FGV: HR 2.05, 95% CI 1.53–2.76; perilesional FGV: HR 1.42, 95% CI 1.14–1.77; Fig. 3b). Global absolute FGV was not associated with risk of incident cancer. In sensitivity analyses excluding 4 cases (5 biopsies) diagnosed within the first two years following the baseline mammogram, HRs were similar in magnitude to those observed for all incident cases (data not shown).
Association of global and local density measures with risk of breast cancer (all cases)
We did not observe statistically significant associations for global, lesional or perilesional %FGV with risk of any breast cancer (all cases; Fig. 3a).
As shown in Fig. 3b, lesional and perilesional absolute FGV were positively associated with risk of any breast cancer for the highest compared with the lowest quartile of FGV (lesional FGV: HRQ4vsQ1 2.26, 95% CI 1.42–3.61, Bonferroni corrected ptrend = 0.04; perilesional HRQ4vsQ1 1.70, 95% CI 1.06–2.72) and per 0.25 SD increase (lesional FGV: HR 1.42, 95% CI 1.20–1.68; perilesional FGV: HR 1.11, 95% CI 1.02–1.22).
Sensitivity analyses
As described above, the Volpara algorithm segments the breast into inner and outer breast areas to distinguish between tissue assumed to be in contact with the compression paddle, and thus at a reasonably constant thickness (i.e., the inner breast) or in the periphery, where breast thickness decreases towards the skin line (i.e., the outer breast). This segmentation may result in partial or total truncation of local density regions that are located close to, or entirely within, the outer breast area (Fig. 1). To examine the potential impact of this truncation, we expanded our analyses of local density to the entire breast in sensitivity analyses.
Using local %FGV and absolute FGV measured in the entire breast (Supplemental Tables 1 and 2), we observed similar patterns of associations for all participant characteristics as in primary analyses. Similar to the main findings, no association for any %FGV measures from the entire breast with risk of prevalent or incident breast cancer was observed (Supplemental Figure S2a). For local absolute FGV for the entire breast, similar although attenuated estimates for the associations with risk of prevalent and incident breast cancer were observed (Supplemental Figure S2b). These data suggest that despite the truncation of some local density regions as a result of breast segmentation, the associations observed for local density with participant characteristics and risk of breast cancer using inner breast area data are robust. Moreover, and as noted above, we examined the effect of adjusting risk models for lesion radius and observed no material change in risk estimates, further underscoring our findings that associations of local density with breast cancer risk are not driven by lesion size.
Discussion
Our analysis of global density and novel, automated, quantitative measures of density at biopsy targets and surrounding parenchyma suggests that these measurements share common epidemiological determinants. We found that among women undergoing biopsy following an abnormal mammogram, absolute FGV, but not %FGV, was associated with breast cancer. Further, among women with BBD at baseline, elevated absolute FGV proximal to the biopsy site was significantly associated with risk of breast cancer during study follow-up. These findings suggest opportunities for developing improved risk assessment among women with abnormal mammograms.
We evaluated breast density using Volpara density software, an automated, quantitative, pixel-wise method of assessing fibroglandular volume from the two-dimensional mammographic representation of the breast, which has been shown to be highly reproducible [29]. In most applications, a global measure of density is reported, but a customization of the algorithm gave us the opportunity to additionally evaluate novel measures of local breast density at and around the biopsy site.
Our findings of associations for age, BMI, and history of benign breast disease (BBD) with global density measures among women without cancer at baseline are consistent with prior studies evaluating global breast density using Volpara software [1, 14]. However, we did not replicate positive associations for education [14], age at first birth and family history [1, 14] with either global %FGV or absolute FGV, possibly reflecting the particular characteristics of our cohort of predominantly white and higher socioeconomic status women undergoing breast biopsy. However, we found that global %FGV and absolute FGV were inversely associated with age among women with a benign diagnosis, but not among women with a cancer diagnosis at baseline. Associations of participant characteristics with local density largely mirrored those for global density.
We reported that global and local absolute FGV were associated with a breast cancer diagnosis at baseline. In this study and others [30, 31] approximately 80% of women whose mammograms are abnormal and who are then referred for biopsy will not be diagnosed with breast cancer. In addition to evaluating global breast density, assessing the absolute FGV at and around a region of abnormality prior to biopsy may assist with the clinical management of these women by providing information about lesions that are most likely to be malignant. We also showed that among women with benign diagnoses at baseline, local absolute FGV was associated with risk of future breast cancer. We hypothesize that increased absolute FGV proximal to a region of abnormality may represent tissue features indicative of a pro-tumorigenic environment.
Previous studies examining the relationship between breast density, BBD and breast cancer risk have focused on global breast density measurements [32–37]. However, the biologic mechanisms underpinning density-associated breast cancer risk remain unclear. Radiologically dense fibroglandular tissue represents a combination of epithelial tissue and non-fatty stroma [37], and disentangling the relationship of these tissue components may provide insights into the heterogeneous breast cancer risk observed among women with apparently similar global breast densities. Our prior studies have highlighted the contribution of non-fatty stroma to global and local density [23], and the association of collagen fiber density [16] and terminal duct lobular unit (TDLU) involution measures [15] with local, but not global, density. These findings support the hypothesis that local density may represent pro-tumorigenic tissue features that are not captured radiologically when evaluating global density. Further studies evaluating histological correlates of local density and their relationship with breast cancer risk are warranted and may provide clues to the etiopathology of breast cancer, with implications for improved risk stratification of the increasing population of women undergoing biopsy and subsequently being diagnosed with BBD.
We did not observe an association of global or local %FGV with risk of breast cancer, at baseline or during follow-up. %FGV is known to be strongly confounded by BMI [26, 38], and indeed, a previous report showed that the strength of association of global %FGV with breast cancer risk increased with increasing BMI [39]. Approximately half of our study population had a BMI in the normal range (< 25 kg/m2), which may have limited our ability to detect an association of %FGV with breast cancer risk.
Strengths of this study include the detailed information on participant characteristics and baseline biopsy findings of a contemporary cohort of women undergoing digital mammography, in which we evaluated a novel measure of local breast density at and surrounding an area of abnormality using an automated, quantitative method of breast density assessment that is reliable and reproducible [1, 5, 40]. This allowed us to examine global and local density measures among women with benign and cancerous lesions who were generally representative of those at the University of Vermont Medical Center being referred to undergo diagnostic breast biopsy after an abnormal breast imaging exam.
However, this study population included primarily non-Hispanic white women, limiting the generalizability of our findings to similar populations. Barnard and colleagues have shown that the distributions of volumetric density-associated breast cancer risk factors vary across racial and ethnic groups, highlighting the importance of diverse populations in understanding the etiology of breast density [13]. Strand and colleagues reported that local regions of dense tissue were associated with interval cancers and large tumors and suggested that local density may mediate at least some of its association with future breast cancer risk through a masking effect [7]. While we cannot fully account for this in the current study, all participants underwent biopsy at baseline to evaluate regions of mammographic abnormality, reducing the likelihood that regions of local density masked a latent tumor. Finally, the number of incident breast cancer cases in our cohort was low, limiting the power to detect some more modest associations. Further studies in larger cohorts of patients undergoing breast biopsy, with longer follow up, are needed to fully elucidate the association of local breast density with risk of future breast cancer.
Conclusion
In summary, we evaluated a novel measure of local breast density proximal to a mammographically-detected abnormality and showed that among women biopsied following an abnormal mammogram, absolute FGV, but not %FGV, was associated with breast cancer. Increased local absolute FGV was associated with risk of breast cancer during study follow-up among women with biopsy-confirmed BBD. Our findings suggest that in addition to global FGV, local FGV may serve as a complementary radiological marker of breast cancer risk, which may assist with the clinical management or risk stratification of this population.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
The authors are indebted to the participants in the BREAST Stamp Project for their outstanding cooperation and to the physicians, pathologists, nurses, technologists, and interviewers for their efforts in the field. The authors thank Clair Bove, Patricia Lutton, Ellen Young, Laura Linville, Daphne Papathomas, Jeff Wang, and Jaime Howes for research assistance. We also thank Janet Lawler-Heaver and Kerry Grace Morrissey from Westat for study management support, and Jane Demuth at Information Management Services for data support and analysis.
Abbreviations
- %FGV
Percent fibroglandular volume
- BBD
Benign breast disease
- BI-RADS
Breast imaging reporting and data system
- BMI
Body mass index
- BREAST Stamp Project
Breast Radiology Evaluation and Study of Tissues Stamp Project
- CC
Cranio-caudal
- CI
Confidence interval
- DCIS
Ductal carcinoma in situ
- DICOM
Digital imaging and communications in medicine
- FGV
Fibroglandular volume
- GEE
Generalized estimating equation
- HR
Hazard ratio
- IRB
Institutional review board
- LCIS
Lobular carcinoma in situ
- MLO
Medio-lateral oblique
- NCI
National Cancer Institute
- OR
Odds ratio
- SD
Standard deviation
- TDLU
Terminal duct lobular unit
Author contributions
Conceptualization: AB, GLG, MM; Methodology: AB, SF, GLG, MM, RMP; Resources: GLG, SDH, BLS, PMV, DLW; Funding acquisition: GLG, MM; Project administration: AB, GLG, MM; Investigation: AB, SF, RMP, AHLC, GLG, MLH, RPH, MM, AT; Formal analysis: AB, SF, RMP; Supervision: GLG, MM; Visualization: AB, GLG, MM, RMP; Writing—original draft: AB, MM; Writing—review and editing: all authors.
Funding
AB and MM are supported with funding from the Health Research Board (HRB) of Ireland under grant EIA-2019–012. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. SF, RMP, ARH and GLG are supported by Intramural Research Program Funds of the Division of Cancer Epidemiology and Genetics (DCEG) of the National Cancer Institute, Department of Health and Human Services, USA. This research was supported in part by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH author(s) were made as part of their official duties as NIH federal employees, are in compliance with agency policy requirements, and are considered Works of the United States Government. However, the findings and conclusions presented in this paper are those of the author(s) and do not necessarily reflect the views of the NIH or the U.S. Department of Health and Human Services. BLS, PMV and DLW were supported in part by the National Cancer Institute (HHSN261201800043C, R01CA282725). The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. MES receives research support from Exact Sciences, unrelated to this work.
Data availability
The datasets used and analyzed during the current study are not publicly available due to data privacy of participants. The authors will make the data available upon reasonable request.
Declarations
Ethics approval and consent to participate
Participants provided written informed consent in accordance with approvals from the NCI Special Studies Institutional Review Board (IRB) and the University of Vermont IRB.
Consent for publication
Not applicable.
Competing interests
The authors declare they have no competing interests (AB, SF, RMP, ARH, PMV, DLW, JMJ, SDH, BLS, MES, MM, GLG). At the time of investigation, AHLC, NK, and RPH were employees of Volpara Health Technologies Ltd. AT is an employee of Lunit International Ltd. MLH is an employee of Lunit Canda, Inc.
Footnotes
Maeve Mullooly and Gretchen L. Gierach: Jointly supervised this work.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Brand JS, Czene K, Shepherd JA, Leifland K, Heddson B, Sundbom A, et al. Automated measurement of volumetric mammographic density: a tool for widespread breast cancer risk assessment. Cancer Epidemiol Biomarkers Prev. 2014;23(9):1764–72. [DOI] [PubMed] [Google Scholar]
- 2.Eng A, Gallant Z, Shepherd J, McCormack V, Li J, Dowsett M, et al. Digital mammographic density and breast cancer risk: a case-control study of six alternative density assessment methods. Breast Cancer Res. 2014;16(5):439. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Gierach GL, Geller BM, Shepherd JA, Patel DA, Vacek PM, Weaver DL, et al. Comparison of mammographic density assessed as volumes and areas among women undergoing diagnostic image-guided breast biopsy. Cancer Epidemiol Biomarkers Prev. 2014;23(11):2338–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.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(6):1159–69. [DOI] [PubMed] [Google Scholar]
- 5.Wanders JOP, Holland K, Karssemeijer N, Peeters PHM, Veldhuis WB, Mann RM, et al. The effect of volumetric breast density on the risk of screen-detected and interval breast cancers: a cohort study. Breast Cancer Res. 2017;19(1):67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mahmoudzadeh AP, Malkov S, Hinton B, Sprague B, Kerlikowske K, Shepherd J. Do Women with Low Breast Density Have Regionally High Breast Density? Cham: Springer International Publishing; 2016. p. 548–53. [Google Scholar]
- 7.Strand F, Azavedo E, Hellgren R, Humphreys K, Eriksson M, Shepherd J, et al. Localized mammographic density is associated with interval cancer and large breast cancer: a nested case-control study. Breast Cancer Res. 2019;21(1):8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Wolfe JN. Risk for breast cancer development determined by mammographic parenchymal pattern. Cancer. 1976;37(5):2486–92. [DOI] [PubMed] [Google Scholar]
- 9.Pinto Pereira SM, McCormack VA, Hipwell JH, Record C, Wilkinson LS, Moss SM, et al. Localized fibroglandular tissue as a predictor of future tumor location within the breast. Cancer Epidemiol Biomarkers Prev. 2011;20(8):1718–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ursin G, Hovanessian-Larsen L, Parisky YR, Pike MC, Wu AH. Greatly increased occurrence of breast cancers in areas of mammographically dense tissue. Breast Cancer Res. 2005;7(5):R605–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Otsuka M, Harkness E, Chen X, Moschidis E, Bydder M, Gadde S, et al. Local mammographic density as a predictor of breast cancer: SPIE; 2015.
- 12.Kerlikowske K, Sprague BL, Tosteson ANA, Wernli KJ, Rauscher GH, Johnson D, et al. Strategies to identify women at high risk of advanced breast cancer during routine screening for discussion of supplemental imaging. JAMA Intern Med. 2019;179(9):1230–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Barnard ME, DuPré NC, Heine JJ, Fowler EE, Murthy DJ, Nelleke RL, et al. Reproductive risk factors for breast cancer and association with novel breast density measurements among Hispanic, Black, and White women. Breast Cancer Res Treat. 2024;204(2):309–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hjerkind KV, Ellingjord-Dale M, Johansson ALV, Aase HS, Hoff SR, Hofvind S, et al. Volumetric mammographic density, age-related decline, and breast cancer risk factors in a National Breast Cancer Screening Program. Cancer Epidemiol Biomarkers Prev. 2018;27(9):1065–74. [DOI] [PubMed] [Google Scholar]
- 15.Gierach GL, Patel DA, Pfeiffer RM, Figueroa JD, Linville L, Papathomas D, et al. Relationship of terminal duct lobular unit involution of the breast with area and volume mammographic densities. Cancer Prev Res (Phila). 2016;9(2):149–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Bodelon C, Mullooly M, Pfeiffer RM, Fan S, Abubakar M, Lenz P, et al. Mammary collagen architecture and its association with mammographic density and lesion severity among women undergoing image-guided breast biopsy. Breast Cancer Res. 2021;23(1):105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Figueroa JD, Pfeiffer RM, Brinton LA, Palakal MM, Degnim AC, Radisky D, et al. Standardized measures of lobular involution and subsequent breast cancer risk among women with benign breast disease: a nested case-control study. Breast Cancer Res Treat. 2016;159(1):163–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ghosh K, Vachon CM, Pankratz VS, Vierkant RA, Anderson SS, Brandt KR, et al. Independent association of lobular involution and mammographic breast density with breast cancer risk. J Nat Cancer Inst. 2010;102(22):1716–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ghosh K, Vierkant RA, Frank RD, Winham S, Visscher DW, Pankratz VS, et al. Association between mammographic breast density and histologic features of benign breast disease. Breast Cancer Res. 2017;19(1):134. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Northey JJ, Barrett AS, Acerbi I, Hayward MK, Talamantes S, Dean IS, et al. Stiff stroma increases breast cancer risk by inducing the oncogene ZNF217. J Clin Invest. 2020;130(11):5721–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Sprague BL, Vacek PM, Mulrow SE, Evans MF, Trentham-Dietz A, Herschorn SD, et al. Collagen organization in relation to ductal carcinoma in situ pathology and outcomes. Cancer Epidemiol Biomarkers Prev. 2021;30(1):80–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Felix AS, Lenz P, Pfeiffer RM, Hewitt SM, Morris J, Patel DA, et al. Relationships between mammographic density, tissue microvessel density, and breast biopsy diagnosis. Breast Cancer Research. 2016;18(1):88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Mullooly M, Ehteshami Bejnordi B, Pfeiffer RM, Fan S, Palakal M, Hada M, et al. Application of convolutional neural networks to breast biopsies to delineate tissue correlates of mammographic breast density. NPJ Breast Cancer. 2019;5:43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Breast Cancer Surveillance Consortium. Data Elements and Questionnaires [23/10/24]. Available from: https://www.bcsc-research.org/index.php/documentation/elements.
- 25.Sprague BL, Nowak SA, Ahern TP, Herschorn SD, Kaufman PA, Odde C, et al. Long-term mammography screening trends and predictors of return to screening after the COVID-19 pandemic: results from a statewide registry. Radiol: Imaging Cancer. 2024;6(3):e230161. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kerlikowske K, Bissell MCS, Sprague BL, Tice JA, Tossas KY, Bowles EJA, et al. Impact of BMI on Prevalence of Dense Breasts by Race and Ethnicity. Cancer Epidemiol, Biomarkers Prev. 2023; 32 (11): 1524-1530. [DOI] [PMC free article] [PubMed]
- 27.van Engeland S, Snoeren PR, Huisman H, Boetes C, Karssemeijer N. Volumetric breast density estimation from full-field digital mammograms. IEEE Trans Med Imaging. 2006;25(3):273–82. [DOI] [PubMed] [Google Scholar]
- 28.Wei LJ, Lin DY, Weissfeld L. Regression analysis of multivariate incomplete failure time data by modeling marginal distributions. J American Statistic Assoc. 1989;84(408):1065–73. [Google Scholar]
- 29.Alonzo-Proulx O, Mawdsley GE, Patrie JT, Yaffe MJ, Harvey JA. Reliability of automated breast density measurements. Radiology. 2015;275(2):366–76. [DOI] [PubMed] [Google Scholar]
- 30.Allison KH, Abraham LA, Weaver DL, Tosteson ANA, Nelson HD, Onega T, et al. Trends in breast biopsy pathology diagnoses among women undergoing mammography in the United States: a report from the Breast Cancer Surveillance Consortium. Cancer. 2015;121(9):1369–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Lilleborge M, Falk RS, Sørlie T, Ursin G, Hofvind S. Can breast cancer be stopped? Modifiable risk factors of breast cancer among women with a prior benign or premalignant lesion. Int J Cancer. 2021;149(6):1247–56. [DOI] [PubMed] [Google Scholar]
- 32.Byrne C, Schairer C, Brinton LA, Wolfe J, Parekh N, Salane M, et al. Effects of mammographic density and benign breast disease on breast cancer risk (United States). Cancer Causes Control. 2001;12(2):103–10. [DOI] [PubMed] [Google Scholar]
- 33.Román M, Louro J, Posso M, Alcántara R, Peñalva L, Sala M, et al. Breast density, benign breast disease, and risk of breast cancer over time. Eur Radiol. 2021. 10.1007/s00330-020-07490-5. [DOI] [PubMed] [Google Scholar]
- 34.Tice JA, O’Meara ES, Weaver DL, Vachon C, Ballard-Barbash R, Kerlikowske K. Benign breast disease, mammographic breast density, and the risk of breast cancer. J Natl Cancer Inst. 2013;105(14):1043–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Vierkant RA, Degnim AC, Radisky DC, Visscher DW, Heinzen EP, Frank RD, et al. Mammographic breast density and risk of breast cancer in women with atypical hyperplasia: an observational cohort study from the Mayo Clinic Benign Breast Disease (BBD) cohort. BMC Cancer. 2017;17(1):84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Yaghjyan L, Colditz GA, Rosner B, Tamimi RM. Mammographic breast density and breast cancer risk: interactions of percent density, absolute dense, and non-dense areas with breast cancer risk factors. Breast Cancer Res Treat. 2015;150(1):181–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Abubakar M, Fan S, Bowles EA, Widemann L, Duggan MA, Pfeiffer RM, et al. Relation of quantitative histologic and radiologic breast tissue composition metrics with invasive breast cancer risk. JNCI Cancer Spectr. 2021;5(3):pkab015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Hudson S, Vik Hjerkind K, Vinnicombe S, Allen S, Trewin C, Ursin G, et al. Adjusting for BMI in analyses of volumetric mammographic density and breast cancer risk. Breast Cancer Res. 2018;20(1):156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Engmann NJ, Scott CG, Jensen MR, Winham S, Miglioretti DL, Ma L, et al. Combined effect of volumetric breast density and body mass index on breast cancer risk. Breast Cancer Res Treat. 2019;177(1):165–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Engler C, Nogueira MS. Analysis of the relationship between global breast density and maximum points of breast density in a sample of Brazilian women. Appl Radiat Isot. 2023;194:110703. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and analyzed during the current study are not publicly available due to data privacy of participants. The authors will make the data available upon reasonable request.