Associations of Oral Contraceptives with Mammographic Breast Density in Premenopausal Women

Mammographic breast density is a well-established and strong predictor of breast cancer risk (1). Appearance of the breast on the mammogram is a reflection of the amount of fat, connective tissue, and epithelial tissue in the breast (2). Light (nonradiolucent) areas on the mammogram represent fibrous and glandular tissues (mammographically dense), whereas the dark (radiolucent) areas are primarily fat. Women with ≥75% density (proportion of the total breast area that appears dense on the mammogram) are at four- to six-fold greater risk of breast cancer compared with women with fatty breasts (1).

Majority of the epidemiologic studies on the association between oral contraceptive (OC) use and breast cancer risk have reported a positive association (3–8). A systematic review of 44 studies found an 8% increased risk of breast cancer in OC users as compared with nonusers [OR = 1.08; 95% confidence interval (CI), 1.00–1.17; ref. 3]. The positive associations were more evident among current/recent OC users relative to past or nonusers and the strength of the association varied by the type of OC formulation with an increased risk among high-dose estrogen OC users (5–7). Furthermore, these associations were more apparent for premenopausal rather than postmenopausal breast cancer (9–11). It has been suggested that the effects of OCs on breast cancer risk result from increased cellular proliferation and subsequently, increased risk of malignant transformation in the rapidly proliferating epithelium (12, 13). This increased amount of epithelial tissue could potentially be reflected on the woman's mammogram. However, the evidence on the association between OC use and mammographic breast density is extremely limited (14–18). The majority of the previous studies reported no associations of OCs with breast density, though a few noted positive associations with duration of OC use and inverse associations with the age at initiation of OC use. However, most of these studies were cross-sectional and had methodologic limitations potentially undermining validity of their findings.

In addition, it has been recently shown that other features within a mammogram may provide additional information for breast cancer risk prediction beyond what percent density is able to accomplish by capturing different attributes. In addition, there is a greater potential for reader error at low values of percent density (19). Novel algorithms for assessment of heterogeneity in patterns of mammographic breast density (referred to as texture) have shown that even within women with similar percent density, texture features, such as a novel image intensity variation (V) measure (V-measure), can further differentiate and predict women who are at an increased risk of breast cancer (20). Independent studies from United States and UK showed that various measures of breast tissue texture features are associated with breast cancer, independent of percent density (20–22). Previous studies also show only moderate correlation between V-measure and percent density (19, 23). In previous studies, V-measure has been associated with increase breast cancer risk in both pre- and postmenopausal women (19, 24, 25). In addition, consistent with what is observed for percent breast density, premenopausal status, younger age (<50 years), body mass index (BMI) < 25 kg/m² have been associated with greater V-measure while early-life body fatness measures (average body fatness at ages 5–10, average body fatness at ages 10–20, and BMI at age 18 years) were associated with lower V-measure, independent of current BMI and percent density (19, 23). Associations of OCs with V-measure have never been investigated.

In this study, we examined the associations of the use of OCs with percent breast density, absolute dense area, nondense area, and a V-measure in premenopausal women using prospective data from the Nurses' Health Study II (NHSII) cohort.

Women for this study were selected from a nested case–control study within the NHSII cohort, a prospective study that was established in 1989 and followed 116,430 female registered nurses in the United States who were 25 to 42 years old at enrollment. After administration of the initial questionnaire, information on breast cancer risk factors (BMI, reproductive history, and alcohol use) and any diagnoses of cancer or other diseases was updated through biennial questionnaires (2, 26).

A nested case–control approach was originally used as an efficient sampling design to examine the association between selected biomarkers and breast cancer risk within the NHSII cohort (27). Using incidence density sampling, women without any type of cancer (other than nonmelanoma skin cancer) at the time of the case's cancer diagnosis (controls) were matched 1:2 with women diagnosed with in situ or invasive breast cancer (cases) on age at the time of blood collection, menopausal status and postmenopausal hormone use (current vs. not current) at blood draw, day/time of blood draw, and race/ethnicity and day in the luteal phase (28). Our analysis included the controls only from this nested case–control study as well as additional eligible women within this cohort (without a history of any cancer other than nonmelanoma skin) who were not included in the original nested breast cancer case–control study These additional women did not provide a blood sample but donated a cheek swab in 2006. We attempted to obtain mammograms closest to the time of blood collection (or approximately 1997 for those who did not provide blood samples). From all eligible women, 1,292 premenopausal women provided consent and had a usable mammogram for density estimation. Of these women, 1,233 had data on OC use and covariates and were included in the analysis; of these women, 736 were controls from the nested case–control study and 497 were additional controls from cheek swab subcohort. The study protocol was approved by the institutional review boards of the Brigham and Women's Hospital (Boston, MA) and Harvard T.H. Chan School of Public Health (Boston, MA), and those of participating registries as required. Consent was obtained or implied by return of questionnaires.

Details of OC use assessment in NHSII were reported previously (29, 30). Briefly, at baseline, women were asked to report if they had ever used OCs and if they were currently using OCs. For each year of age, women were asked to report if they had used OCs for at least 2 months in that year, if they had used OCs for at least 10+ months in that year, and the brand of OC used during that year. At each follow-up questionnaire cycle, women reported if they were currently using OCs, if they had used OCs since the last questionnaire, their duration of OC use since the last questionnaire in prespecified categories (≤1, 2–4, 5–9, 10–14, 15–19, and ≥20 months), and the OC brand they had used for the longest during the last 2 years. To reduce recall inaccuracy, each questionnaire was accompanied with a booklet containing names and color photographs of all OC brands available during the relevant time period (30).

The time from baseline (1989) to mammogram date was on average 9.6 years (range 1–20 years). We assessed life-long OC exposure using all questionnaires available from before the mammogram date. The information on OC use was updated from most recent questionnaire preceding the mammogram and the time between most recent questionnaire update and the mammogram date was on average 0.5 years (range 0–1 years). In our sample, based on the mammogram dates, only 1 woman had her OC use information taken only from baseline and for the majority of the women, the information came from all follow-up cycle updates through 2011, on average six questionnaire cycles per woman.

Mammographic density was assessed in three batches approximately 2 to 3 years apart. To quantify mammographic density, craniocaudal views of both breasts for first two batches of mammograms in the NHSII were digitized at 261 μmol/L per pixel with a Lumisys 85 laser film scanner (Lumisys). The third batch of NHSII mammograms was digitized using a VIDAR CAD PRO Advantage scanner (VIDAR Systems Corporation) and comparable resolution of 150 dots per inch and 12 bit depth. Cumulus software (University of Toronto, Toronto, Canada) was used for computer-assisted determination of the absolute dense area, nondense area, and percent breast density on all mammograms (2, 31). All NHSII images were read by a single reader. Although within batch reproducibility was high (intraclass correlation coefficient ≥ 0.90; ref. 32), density measures varied across the NHSII batches. The density measures from the second and third batches of NHSII mammograms were adjusted to account for the batch effect (whether due to intrareader variability or scanner), as previously described (33).

Percent breast density was measured as percentage of the total area occupied by epithelial/stromal tissue (absolute dense area) divided by the total breast area. Because breast density of the right and left breasts for a given woman are strongly correlated (31), the average density of both breasts was used in this analysis.

Additionally, we used the V-measure, a novel automated measure that captures the grey-scale variation within a mammogram. The V-measure for each woman was estimated using a previously described method (19). Briefly, the breast area on the craniocaudal mammogram view was first automatically segmented from the background. Then, unwanted spatial variation was reduced by eliminating a portion of the breast area corresponding to where the breast was not uniformly compressed during the image acquisition; the breast area was eroded by 25% along a radial direction (34). This erosion step (approximately) removed the regions that could potentially interfere with the V-measure (35). In the final step, the V-measure was calculated as the SD of the pixel values within the eroded breast region for each mammogram. The original V-measure was standardized to the lowest resolution to account for three different resolutions used originally for mammogram digitalization (high resolution: 171 μm; medium resolution: 232 μm; and low resolution: 300 μm). Because the V-measures of the left and right breasts for an individual participant were strongly correlated (r = 0.87, P < 0.0001), the average V measure of both breasts were used in this study.

Information on breast cancer risk factors was obtained from the biennial questionnaires closest to the date of the mammogram. For exclusion from this analysis, women were considered to be postmenopausal if they reported: (i) no menstrual periods within the 12 months before blood collection with natural menopause, (ii) bilateral oophorectomy, or (iii) hysterectomy with one or both ovaries retained, and in addition were 54 years or older for ever smokers or 56 years or older for never smokers (36, 37). Height and weight at age 18 years were reported on the 1989 questionnaire.

We used generalized linear regression to examine the associations of OC use with percent density, absolute dense area, absolute nondense area, and V-measure, while taking into account the correlation between matched controls (38). Percent density, absolute dense, and nondense area measures were square root–transformed to improve normality of the error distribution in all the regression analyses. The regression estimates were adjusted for the following covariates at the mammogram date: age(continuous), BMI (continuous), age at menarche (<12, 12–13, >13 years), parity and age at first child's birth (nulliparous, parous with age at first birth <25, parous with age at first birth ≥25), a confirmed history of benign breast disease (yes, no), a family history of breast cancer (yes, no), and alcohol consumption (0, 1–<5, ≥5 g/day, unknown).

OC use was categorized with previously used approaches (7, 39) as follows: OC use status (never, past use, current use with duration >0–8 years, and current use with duration ≥8 years), duration of OC use (never/<1 year, 1–<5, 5–<10, and ≥10 years), time since last OC use (≤4, >4–<10, 10–<15, and ≥15 years), age at first OC use (<20, ≥20–24, 25–29, and ≥30 years), and age at last OC use (continuous, years). The duration of OC use was analyzed for all participants and separately among past users only.

A two-sided test for trend was performed modeling relevant contraceptive use variables as ordinal variables and using the median level in each category. Statistical significance in all the analyses was assessed at 0.05 level. The analyses were performed using SAS software (version 9.4, SAS Institute, Cary, North Carolina).

Of the 1,233 premenopausal control women, 184 (15%) were never, 952 (77%) were past, and 97 (8%) were current OC users. The average age of the study population was 44.7 years (range 30–56). The age-adjusted characteristics of the study population by OC use status are presented in Table 1. The distribution of breast density measures was similar across OC use status categories (percent density: 39.7%, 40.5%, and 40.6% for never users, past and current OC users, respectively; absolute dense area: 42.9 cm², 44.5 cm², and 40.3 cm², respectively; absolute nondense area: 73.9 cm², 73.3 cm², and 62.0 cm², respectively; V-measure: 324.9, 341.1, and 330.6, respectively). The density measures were only moderately correlated with each other (V and percent density: correlation coefficient: r = 0.42, P < 0.0001; V and absolute dense area: r = 0.35, P < 0.0001; V and nondense area: r = −0.29, P < 0.0001).

In multivariable analyses, OC use status was not associated with percent density (current vs. never: β = −0.06; 95% CI, −0.37–0.24; past vs. never: β = 0.10; 95% CI, −0.09–0.29; P_trend = 0.99), absolute dense area (current vs. never: β = -0.20; 95% CI, -0.59–0.18; past vs. never: β = 0.13; 95% CI, −0.12–0.39; P_trend = 0.66), and absolute nondense area (current vs. never: β = -0.19; 95% CI, −0.56–0.18; past vs. never: β = −0.01; 95% CI, −0.28–0.25; P_trend = 0.43; Table 2). Although not statistically significant, there was a trend of a higher V-measure in past (β = 22.12), and current (β = 11.25) users as compared with never users (P_trend = 0.27).

Total duration of use, time since last use, age at first use, and age at last use were not associated with percent density, absolute dense, and nondense areas (Table 2). We observed an inverse association between age at first OC use and the V-measure (<20 years vs. never: β = 26.88; 95% CI, 3.18–50.58; 20–24 years vs. never: β = 20.23; 95% CI, −4.24–44.71; 25–29 years vs. never: β = 2.61, 95% CI, −29.00–34.23; ≥30 years vs. never: β = 0.28; 95% CI, −34.16–34.72; P_trend = 0.03). Total duration of use, time since last use, and age at last use were not associated with V-measure.

In this study, we investigated the associations of OC use with various measures of mammographic breast density (percent breast density, absolute dense area, nondense area, and V-measure) in cancer-free premenopausal women (controls). We found no associations of any of the OC exposure variables, except the age at first use, with any of the density measures. The age at first OC use was inversely associated with V-measure.

The evidence regarding the possible association between OC use and mammographic breast density remains very limited but our findings of no association between OCs and breast density are consistent with those from previous studies. A cross-sectional study of 366 cancer-free women from United Arab Emirates by Albeshan and colleagues found no association between ever use of OC and mammographic breast density defined using American College of Radiology's Breast Density classification system [BI-RADS; odds ratio (OR), 1.25; 95% CI, 0.50–3.16] but the odds of OC use for ≥3 years were 511% higher among women with high dense breasts compared with those with low density (OR, 6.11; 95% CI, 1.41–26.57; ref. 14). However, these associations were univariate and the estimates were not adjusted for important breast cancer risk factors thus making these results questionable. Additionally, the information on duration of OC use was available only for 49 women. A cross-sectional study by Jeon and colleagues compared OC use in women with dense (BI-RADS II and IV) and fatty (BI-RADS I and II) breasts among 516 Korean women, although it was not clear if women with a history of breast cancer were excluded. This study found no associations (OR, 1.16, 95% CI, 0.51–2.64; ref. 16). Similarly, Ahmadinejad and colleagues found no associations of OCs with breast density among 728 Iranian women; however, this study utilized both diagnostic and screening mammograms and did not exclude women with a personal history of breast cancer (17). Higher breast density was reported for OC users among postmenopausal women participating in a population-based mammography registry in Croatia. This study used the BI-RADS breast density measures from 52,752 mammograms and was originally designed to examine associations of various risk factors with breast cancer risk (18). The study found a greater proportion of women who ever used OCs among women with greater breast density [27.3% for scattered fibroglandular densities (BI-RADS II), 29.6% for heterogeneously dense (BI-RADS III), and 33.2 for extremely dense (BI-RADS IV) vs. 24.5 for entirely fat (BI-RADS I); P < 0.001). The risk estimates in this study were adjusted for the same covariates as in our analysis, except alcohol, and were also additionally adjusted for postmenopausal hormone use. Finally, Dorgan and colleagues reported an inverse association of age at OC use initiation and positive associations of OC use duration with percent dense breast volume (%DBV) and absolute dense breast volume (ADBV) in a cross-sectional study of 176 healthy young women (age range 25–29 years). This study used MRI for breast density estimation. Further, generalizability of the findings from this study is limited as all study participants had elevated low-density lipoprotein cholesterol and met several additional eligibility criteria as part of the original randomized control trial (15). Finally, previous studies varied with respect to the included generational cohorts (from women born as early as 1920s to those born in 1980s); thus, the findings could potentially be affected by the changes in composition of OCs over time.

We found an inverse association between age at OC use initiation and the V-measure, while no associations were observed with percent density, absolute dense, and nondense areas. Dorgan and colleagues also found an inverse association between the age at OC initiation and mean dense breast volume (15). OCs can potentially increase breast density by inducing epithelial proliferation and thus increasing the area of the breast occupied by fibroglandular tissue that appears “dense” on the mammogram (15, 40–44). However, OCs vary greatly with respect to their formulations and indications (for example, treatment of endometriosis and menstrual disorders) which have changed significantly over the years (45, 46). Previous studies have demonstrated that addition of progestogens to estrogens in combined OCs (containing estrogens and progestogen) as well as in combined hormone replacement therapy induces epithelial proliferation in the breast tissue to a greater degree than use of estrogens alone (12, 13, 47, 48). While the evidence on OCs and breast density is very limited, prior studies on hormone replacement therapy and breast density consistently demonstrated an increase in breast density among hormone therapy users which was more pronounced in combined therapy users than in users of estrogen-alone preparations or never users (48). Moreover, some studies in postmenopausal women showed that in combined postmenopausal hormone therapy users, breast proliferation was localized to the terminal duct-lobular units, where most breast cancers originate (47), thus further suggesting potential mutagenic properties of progestogens. In addition to an increase in epithelial proliferation, it has also been hypothesized that addition of progesterone may cause stromal edema and since variations in stroma explain large proportion of the variation in breast density (approximately 29%; ref. 49), this mechanisms could also contribute to more pronounced breast density changes in women using combined estrogen + progesterone preparations. The observed higher density among women with younger age at initiation of OC use could potentially result from a longer exposure to the hormones. The time before the first full-term pregnancy represents one of the few windows of increased susceptibility of breast tissue to carcinogenic influences, including hormones, and thus earlier initiation of OCs in this window of susceptibility might result in more prominent effects of these exogenous hormones on the breast tissue that could later be reflected in adult breast density (50–53). However, future studies are needed to confirm our findings and to elucidate underlying biological mechanisms.

Our study is the largest study to investigate the association of OCs with various measures of mammographic density among cancer-free premenopausal women. The analysis used data from the NHSII, an established cohort with more than 25 years of follow-up, ascertainment of disease status, and comprehensive information on breast cancer risk factors and breast density.

Our study has a few limitations. The examined associations were based on density measures from a single mammogram rather than the woman's life-long density patterns. However, previous studies have suggested that a single breast density measure can predict breast cancer risk for up to 10 years in both pre- and postmenopausal women (54, 55) and that breast density measures of a woman over a long period of time are highly correlated (56). Despite the prospective nature of the cohort, potential errors in OC use recall are possible; however, previous validation studies in this cohort demonstrated high accuracy in the self-report of different aspects of OC use (57). Next, our study included only cancer-free women; however, when women with breast cancer (cases) were included in the sensitivity analysis, the findings did not change. Finally, we used square root–transformed percent breast density, absolute dense, and nondense areas, which makes interpretation of the magnitude of the associations challenging. However, the approach of square-root transformation of breast density measures has been widely used in previous studies (58–70) as a robust approach, and facilitates comparison across studies.

In conclusion, we investigated the associations of OC use with percent breast density, absolute dense, and nondense areas, and V-measure in women. Our results suggest that women with earlier age at initiation of OC use could have greater mammographic breast density. Further studies are warranted to confirm our findings.

J. Heine reports patents for US 6,310,967 BI, US 7,664,604, US 10,007,982 B2, US 10,134,148 B2, and US 10,846,856 B2 issued. G.A. Colditz reports grants from NIH during the conduct of the study. B. Rosner reports grants from NIH during the conduct of the study. R.M. Tamimi reports grants from NIH/NCI during the conduct of the study. No disclosures were reported by the other authors.

L. Yaghjyan: Conceptualization, statistical analysis, writing–original draft, writing–review and editing. C. Smotherman: Writing–original draft, writing–review and editing. J. Heine: Investigation, methodology, writing–review and editing. G.A. Colditz: Resources, methodology, writing–review and editing. B. Rosner: Writing–review and editing. R.M. Tamimi: Conceptualization, funding acquisition, methodology, writing–review and editing.

This work was supported by the NCI at the NIH (CA131332 and CA175080 to R.M. Tamimi, UM1 CA186107 and P01 CA087969 to M.S., U01 CA176726 to W.W, U01 CA200464 to J. Heine), Avon Foundation for Women, Susan G. Komen for the Cure, and Breast Cancer Research Foundation. We would like to thank the participants and staff of the NHS and NHSII for their valuable contributions as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, and WY. The authors assume full responsibility for analyses and interpretation of these data.

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.