Metabolic syndrome and mammographic density: The Study of Women's Health Across the Nation

Authors


  • Clinical Centers: University of Michigan, Ann Arbor, MI: MaryFran Sowers, Principal Investigator (PI); Massachusetts General Hospital, Boston, MA: Joel Finkelstein, PI, 1999 to present and Robert Neer, PI, 1994–1999; Rush University, Rush University Medical Center, Chicago, IL: Howard Kravitz, PI, 2009 to present and Lynda Powell, PI, 1994–2009; University of California, Davis/Kaiser, CA: Ellen Gold, PI; University of California, Los Angeles, CA: Gail Greendale, PI; Albert Einstein College of Medicine, Bronx, NY: Rachel Wildman, PI, 2010 and Nanette Santoro, PI 2004–2010; University of Medicine and Dentistry-New Jersey Medical School, Newark, NJ: Gerson Weiss, PI 1994–2004; and the University of Pittsburgh, Pittsburgh, PA: Karen Matthews, PI. NIH Program Office: National Institute on Aging, Bethesda, MD: Sherry Sherman, 1994 to present and Marcia Ory, 1994–2001; National Institute of Nursing Research, Bethesda, MD: Program Officer. Central Laboratory: University of Michigan, Ann Arbor, MI: Daniel McConnell; (Central Ligand Assay Satellite Services). Coordinating Center: University of Pittsburgh, Pittsburgh, PA: Kim Sutton-Tyrrell, PI, 2001 to present; New England Research Institutes, Watertown, MA: Sonja McKinlay, PI, 1995–2001. Steering Committee: Susan Johnson, Current Chair; Chris Gallagher, Former Chair.

Abstract

The metabolic syndrome (MetS) is associated with an increase in breast cancer risk. In our study, we evaluated whether the MetS was associated with an increase in percent mammographic density (MD), a breast cancer risk factor. We used linear regression and mixed models to examine the cross-sectional and longitudinal associations of the MetS and components of the MetS to percent MD in 790 premenopausal and early perimenopausal women enrolled in the Study of Women's Health Across the Nation (SWAN). In cross-sectional analyses adjusted for body mass index (BMI), modest inverse associations were observed between percent MD and the MetS [β = −2.5, standard error (SE) = 1.9, p = 0.19], abdominal adiposity (β = −4.8, SE = 1.9, p = 0.01) and raised glucose (β = −3.7, SE = 2.4, p = 0.12). In longitudinal models adjusted for covariates including age and BMI, abdominal adiposity (β = 0.34, SE = 0.17, p = 0.05) was significantly positively associated with slower annual decline in percent MD with time. In conclusion, our results do not support the hypothesis that the MetS increases breast cancer risk via a mechanism reflected by an increase in percent MD.

The metabolic syndrome (MetS) is a cluster of metabolic abnormalities. It is clinically defined as having at least three of the following disorders: abdominal adiposity, raised blood pressure, raised fasting plasma glucose, raised triglycerides and reduced high-density lipoprotein (HDL) cholesterol.1 Women with the MetS have an estimated 58% higher risk for postmenopausal breast cancer, compared to women without the MetS.2 The MetS is positively associated with triple-negative breast cancer,3 aggressive tumor phenotype4 and breast cancer recurrence5 and mortality.6 Approximately one-third of US women between 40 and 59 years of age are estimated to meet the criteria for the MetS.7 Characterizing metabolic markers of breast cancer risk among these at risk women is of major public health relevance.

Several possible biologic mechanisms exist by which the MetS may adversely influence breast carcinogenesis. The components of the MetS may cooperatively contribute to cancer initiation and progression via mechanisms involving hyperinsulinaemia, adipokine production, glucose utilization, chronic low-grade inflammation and oxidative stress.8 For example, the effects of abdominal adiposity on breast cancer risk may be mediated via insulin, insulin-like growth factors and leptin; all of these factors act as mitogens in breast epithelial cells9–11 and are pro-angiogenic in breast cancer animal models and human breast cell lines.12, 13 Neoplastic breast cells have elevated glucose metabolism,14 suggesting that a glucose enrich environment is associated with breast cancer development. Additionally, glucose levels and hypertension have been positively correlated with markers for systemic inflammation15 and oxidative stress,16, 17 both of which may increase breast cancer risk.

Epidemiologic studies have reported moderate positive associations for breast cancer risk with individual components of the MetS.2, 18–20 For postmenopausal breast cancer, the presence of raised triglycerides or reduced HDL-cholesterol was associated with 60% higher risk,2 and abdominal adiposity was associated with 34% higher risk.18 Raised fasting glucose was associated with a 2-fold higher risk for premenopausal breast cancer.19 Only a few studies have been conducted on breast cancer and the MetS as a single entity and results have been inconsistent, with one study reporting a positive association with postmenopausal breast cancer2 and two reporting no association.21, 22

The proportion of the breast occupied by radiologically dense breast tissue, or percent mammographic density (MD), may be on the causal pathway between metabolic abnormalities and breast cancer development.23 Percent MD is associated with a 4- to 6-fold increase in breast cancer risk, comparing high (> 60%) versus low (< 10%) density.24 The elevated breast cancer risk may persist for an estimated 10 years after breast density assessment regardless of menopausal status or age.25 Percent MD may be a biological marker of cumulative exposure to hormones and other growth factors.23 Abdominal adiposity influences MD and breast cancer risk in opposite directions, with abdominal adiposity associated with low percent MD.26 Data on the effect of other metabolic abnormalities on MD require further study; for example, limited available data support a modest positive association with HDL-cholesterol27, 28 and a modest inverse association with triglycerides.28 No epidemiologic studies have been conducted to date that have evaluated the association between the MetS and percent MD.

We examined the cross-sectional and longitudinal associations of the MetS and components of the MetS to percent MD in a subset of women in the Study of Women's Health Across the Nation (SWAN), a community-based cohort of African American, Chinese, Japanese and Caucasian women. Specifically, we hypothesized that the MetS would be positively associated with percent MD and inversely associated with annual decline in percent MD.

Abbreviations

AHA: American Heart Association; BMI: body mass index; DBP: diastolic blood pressure; HDL: high-density lipoprotein; HOMA-IR: homeostasis model assessment of insulin resistance; MHT: menopausal hormone therapy; MD: mammographic density; MetS: metabolic syndrome; NCEP-ATPIII: National Cholesterol Education Program—Adult Treatment Panel III; SBP: systolic blood pressure; SD: standard deviation; SHS: secondhand smoke; SWAN: Study of Women's Health Across the Nation.

Material and Methods

Study population

SWAN, a multi-ethnic, community-based, longitudinal study of the menopausal transition, enrolled 3,302 women in 1996–1997 from seven clinical sites.29 The SWAN MD ancillary study included SWAN participants enrolled in three of the seven study sites: University of California Davis-Kaiser (Oakland), University of California Los Angeles (Los Angeles) and University of Pittsburgh (Pittsburgh). To identify community-based women, two sites (Los Angeles and Pittsburgh) used random digit dialing-sampling and one site selected randomly from a healthcare organization membership listing (Oakland). Eligible women were aged between 42 and 52 years, reported having had a menstrual period and no use of exogenous hormones within the 3 months before recruitment and identified their primary ethnic group as African-American (Pittsburgh), Chinese (Oakland), Japanese (Los Angeles), or Caucasian (all sites). Of the 1,248 women in follow-up at these three sites at the time of enrollment into the ancillary MD study (i.e., at visit 5 or 6), 1,005 (81%) agreed to participate and had at least one eligible mammogram for MD assessment.

For analyses presented here, women were excluded if they were not premenopausal or early perimenopausal (defined later) at their index mammogram (N = 205), were missing MD measures (N = 4) or were missing MetS information (N = 6). The remaining 790 women were included in these analyses. The study protocol for both the core SWAN study and the ancillary study was approved by the institutional review boards of the participating institutions, and all women provided signed, written informed consent for participation in the studies.

Mammographic density

Eligible mammograms were those taken as part of routine medical care from 2 years before the baseline SWAN visit through 2 years after the sixth visit and were restricted to those done on breasts without prior surgery. We defined the “index” mammogram as the earliest available mammogram within 2 years of a SWAN visit when the MetS was assessed. The SWAN visit closest in time to the index mammogram date was the baseline or visit 1 for the majority of women (84%). The average lag between the date of the index mammogram and the date of the closest SWAN visit was 0.6 years [standard deviation (SD) = 0.4; range, 0–2 years].

Film-screen mammograms were sent periodically in batches for assessment by Martine Salane, an established expert in the techniques of measuring MD.30–32 Unknown to Salane, 10% of the films for our study were sent for re-review. The initial and repeat readings had excellent concordance (the intraclass correlation coefficient was 0.96 for percent MD and 0.98 for area of dense breast). Quantitative assessment was obtained by measuring the total area of the breast and the area of dense breast tissue with a compensating polar planimeter (LASICO, Los Angeles, CA) on the craniocaudal view of the right breast. Mammograms from the left breast were used for density assessments when a woman reported surgery in the right breast (e.g., biopsy, breast augmentation, reduction or reconstruction) or when films from the right breast were unavailable.

Metabolic syndrome

At baseline and annual follow-up visits, all participants completed a standard protocol that included self- and interviewer-administered questionnaires, phlebotomy and anthropometry, and this information was used to determine the MetS status. Phlebotomy was performed in the morning after an overnight fast. Participants were scheduled for venipuncture on days 2 to 5 of a spontaneous menstrual cycle within 60 days of their previous annual visit or if this timed blood draw was not possible within 2 months of the anticipated annual visit and then whenever was convenient. Blood pressure measurements were taken using a standard mercury sphygmomanometer when seated and after a minimum of 5 min of rest, and two sequential measurements were averaged. Waist circumference was measured with the respondents in nonrestrictive undergarments. Biochemical assays for triglycerides,33 HDL-cholesterol,34 glucose35 and insulin35 were performed on the serum samples obtained at baseline and at visits 1, 3–7.

The MetS was defined in accordance with the National Cholesterol Education Program—Adult Treatment Panel III (NCEP-ATPIII)36 with race/ethnic-specific cut points for waist circumference.1 Any participant with at least three of the following components were considered to have the MetS: (a) waist circumference (≥ 88 cm for African Americans and Caucasians and ≥ 80 cm for Chinese and Japanese); (b) triglycerides level ≥ 150 md/dL; (c) HDL-cholesterol levels < 50 mg/dL; (d) systolic blood pressure (SBP) ≥ 130 mm Hg, diastolic blood pressure (DBP) ≥ 85 mm Hg or self-reported use of antihypertensive medications; and (e) fasting glucose levels ≥ 110 mg/dL or self-reported use of antidiabetic medication (insulin or oral agents).

Covariates

Self-reported race/ethnicity, education, age at menarche, parity, age when first child was born, family history of breast cancer, prior oral contraceptive use and prior exogenous hormone use (e.g., fertility drugs, estrogens or progestins, hormone patches or creams and hormone injections or postmenopausal hormones) were obtained at the baseline examination. Age was calculated based on the date of the index mammogram. Time-dependent covariate data, such as body mass index (BMI, in kg/m2), menopausal status, diabetes status, insulin resistance, physical activity, alcohol consumption, smoking status and annual household income were obtained from the SWAN visit closest in time to the mammogram date.

BMI was calculated from measurements of weight and height that were obtained with a calibrated scale and a stadiometer at each visit. Menopausal status was based on response to questions on date of last menstrual bleeding, changes in regularity of bleeding, menopausal hormone therapy (MHT) use and hysterectomy and oophorectomy at each visit as follows: (i) premenopause: a menstrual period within the past 3 months with no change in regularity; (ii) early perimenopause: a menstrual period within the past 3 months but with a change in cycle lengths; (iii) late perimenopause: no menstrual bleeding for 3 to 11 months; (iv) postmenopause: no menstrual bleeding for at least 12 months; (v) surgical menopause: hysterectomy and/or bilateral oophorectomy; and (vi) undetermined: use of MHT before 12 months of amenorrhea.

Diabetes status and insulin resistance were based on fasting serum samples obtained at baseline and at visits 1, 3–7. Diabetes status was determined by glucose levels ≥ 126 mg/dL or self-reported use of antidiabetic medication (insulin or oral agents). Insulin resistance was calculated by the homeostasis model assessment of insulin resistance (HOMA-IR).37

Physical activity was assessed using the Kaiser Physical Activity Survey38 at baseline and at visits 3, 5 and 6. Physical activity information associated with sports/exercise activities was used in our study because it more strongly correlated with the MetS than other activity domains (e.g., routine activities and household/child care) and moderate-to-vigorous activity had been shown to prevent the MetS.39 The sports/exercise activity index was based on participant responses about frequency, duration, perceived physical exertion and an activity-specific intensity code. The values ranged from 1 (lowest) to 5 (highest) and represented relative ranking of individuals.

Active smoke exposure was annually assessed using a self-administered questionnaire adapted from the American Thoracic Society.40 Secondhand smoke (SHS) exposure for never smokers, defined as more than 1 hr of cumulative tobacco smoke exposure while at home, work, or other places within the past 7 days,41 was assessed at baseline and at visit 3.

Statistical analyses

The characteristics of the study population by metabolic status were compared using χ2 tests for categorical variables and Student's t-test for continuous variables. General linear models were used to estimate the association between metabolic abnormalities and percent MD (index mammogram). Our primary analyses were focused on percent MD because previous studies reported stronger associations for breast cancer risk with percent MD than the area of dense breast tissue, after adjusting for body size.42 We built two models with different primary independent variables: the first used presence or absence of the MetS and the second used each of the five components of the MetS. We included the following covariates in all adjusted models because these factors have been associated with MD in previous analyses43 of this subcohort: age (continuous), BMI (continuous), race/ethnicity-study site (Caucasian-Oakland, Chinese-Oakland, Caucasian-Los Angeles, Japanese-Los Angeles, Caucasian-Pittsburgh and African American-Pittsburgh), parity (0, 1–2, ≥3), menopausal status (premenopausal and early perimenopausal), past use of oral contraceptive hormones (never and ever), smoking (never active/without SHS, never active/with SHS, former active and current active) and physical activity (continuous). Other covariates were then added to the MetS models individually, including annual household income (<$50 K, $50 K–$75 K, ≥$75 K), age at menarche (<12, 12, 13, ≥14 years), prior use of exogenous hormones not including oral contraceptive use (never and ever) and alcohol consumption (abstainer, ≥1 drink/week). The addition of these covariates did not change the beta coefficient value for the MetS in minimally adjusted models by at least 10% and, therefore, were not included in the final adjusted model.

We performed a number of additional analyses. First, we examined interactions, e.g., whether the association between percent MD and the MetS varied by race/ethnicity (Caucasian and other), menopausal status (pre and early peri) or BMI (< 27.5, ≥ 27.5 kg/m2) using stratified analyses and the inclusion of interaction terms. Second, we performed two sensitivity analyses. One excluded women with diabetes at baseline, and the other used the American Heart Association (AHA)'s revised version of the NCEP-ATPIII definition of the MetS that included a lower threshold for raised fasting glucose of ≥100 mg/dL,1 compared to ≥ 110 mg/dL.36 Third, we modeled area of dense breast tissue and area of nondense breast tissue as the outcome of interest; the variables were log transformed to normalize their distributions.

Longitudial analyses were performed among a subset of women who had at least two mammograms (N = 728). We used mixed-effects linear models to estimate the annual change in percent MD and how the MetS or the components of the MetS assessed at the index mammogram modified that change; time was modeled as years since the index mammogram. The average number of mammograms per woman was 5.3 (SD = 1.8; range, 2–10 mammograms). The average follow-up from the index to last mammogram was 5.7 years (SD = 1.7; range, 0.9–9.3 years). In addition to the covariates included in the cross-sectional models above, within-woman changes for BMI and menopausal status were calculated as the difference between the value at a subsequent assessment and the index value collected at the SWAN visit closest to the index mammogram and modeled as interactions with time. The change variable for menopausal status was defined as (i) no change in menopausal status, (ii) premenopausal to early perimenopausal, (iii) premenopausal to late perimenopausal, (iv) early perimenopausal to late perimenopausal, (v) perimenopausal to naturally postmenopausal, (vi) early perimenopausal to naturally postmenopausal and (vii) became postmenopausal due to hysterectomy and/or bilateral oophorectomy or became undetermined due to use of MHT before 12 months of amenorrhea. The change by time interaction variables was evaluated only in the longitudinal models.

Statistically significant interactions in the final adjusted models were determined by p-values < 0.05. Statistical computing was conducted using SAS version 9.2 (SAS Insitute, Cary, NC).

Results

In the cohort of 790 women, approximately 50% were Caucasian, the average age was 46.4 years (SD = 3.0) and the median BMI was 24.1 kg/m2 (IQR = 6.4) at the time of the index mammogram. A majority of the women were parous (82%), had used oral contraceptives (75%), abstained from alcohol (71%) and never smoked (68%) (Table 1). Women with the MetS were more likely to be overweight or obese (BMI ≥ 25 kg/m2), have an earlier age at menarche, abstain from alcohol, be less physically active, be diabetic and have insulin resistance, compared to women without the MetS (Table 1).

Table 1. Study population characteristics by MetS status, SWAN, 1996–2004, N = 790
inline image

Lower mean percent MD was observed for women with the MetS and each component of the MetS, compared to those without the metabolic abnormality (Table 2). Abdominal adiposity and reduced HDL-cholesterol were the most prevalent metabolic abnormalities among women with the MetS (Table 2). Lower mean percent MD was observed among women with abdominal adiposity, regardless of their MetS status (Table 2). For women with raised glucose, lower mean percent MD was only observed among those who had the MetS, although the interaction between raised glucose and the MetS was not statistically significant (Table 2).

Table 2. Mean percent MD by prevalence of metabolic abnormalities, SWAN, 1996–2004, N = 790
inline image

In multivariable cross-sectional analyses, lower percent MD was associated with the MetS, abdominal adiposity and raised glucose (Table 3, Model 1). All associations were attenuated after adjustment for BMI (Table 3, Model 2). Only the association with abdominal adiposity remained statistically significant. Similar results for all metabolic abnormalities were obtained with percent MD after excluding women with type 2 diabetes (N = 25) (data not shown). An additional 13 women were classified with the MetS when we used a lower threshold of ≥100 mg/dL for determining raised glucose, in accordance with the AHA's revised NCEP-ATPIII definition.1 The association between the MetS and the percent MD did not change with the less conservative definition [β = −2.4, standard error (SE) = 1.8, p = 0.19]. No statistically significant interactions were observed between the MetS and race/ethnicity (p = 0.18), menopausal status (p = 0.98) or BMI (p = 0.36) in its relation with percent MD.

Table 3. Multivariable cross-sectional analyses of percent MD1 by metabolic abnormalities, regression coefficients and standard errors, SWAN, 1996–2004, N = 790
inline image

We evaluated associations for nondense and dense breast tissue area separately in multivariable models adjusted for BMI. Abdominal adiposity was positively associated with nondense breast tissue area (β = 0.24, SE = 0.06, p < 0.001), but not with dense breast tissue area (β = −0.01, SE = 0.08, p = 0.93). Raised glucose was not associated with nondense (β = 0.08, SE = 0.08, p = 0.31) or dense breast area (β = −0.13, SE = 0.10, p = 0.19).

Our longitudinal analyses revealed an unadjusted mean annual decline of 1.1 (SD = 0.1) in percent MD. In analyses unadjusted for BMI, the MetS, abdominal adiposity and raised glucose were associated with annual decline in percent MD (Table 4, Model 1). These associations were attenuated and lost statistical significance after adjusting for BMI (Table 4, Model 2). Abdominal adiposity was associated with a 0.34 point slower annual decline in percent MD, although the association had borderline statistical significance.

Table 4. Multivariable longitudinal analyses of percent MD1 by metabolic abnormalities, regression coefficients and standard errors, SWAN, 1996–2004, N = 726
inline image

Discussion

Using data from a multi-ethnic cohort of premenopausal and early perimenopausal women, we found a moderate inverse association between the MetS and percent MD; however, after adjusting for BMI, the strength of this association was greatly attenuated, and it was no longer statistically significant. Examination of the relations between percent MD and each of the five components of the MetS revealed a statistically significant inverse association between percent MD and raised glucose and abdominal adiposity. The inverse association between abdominal adiposity and lower percent MD remained statistically significant after controlling for BMI, although the effect size was roughly halved with the adjustment. In longitudinal models adjusted for BMI, abdominal adiposity lessened the annual decline in percent MD.

The inverse association between abdominal adiposity and percent MD after adjustment for BMI is consistent with findings from other studies.26, 44 This inverse association is paradoxical to the positive association between abdominal adiposity and breast cancer, independent of BMI.18, 45 One explanation for the paradox is that the effects of abdominal adiposity on breast cancer risk are not mediated by MD but via alternative pathways that are not represented by MD.46 This notion is supported by our observation, as well as others,44 that abdominal adiposity was positively associated with nondense breast area.

We found a modest inverse association between percent MD and raised glucose that did not reach statistical significance. There is evidence that metabolic conditions are associated with an increase in MD. Diabetic fibrous mastopathy is a rare complication of women with longstanding and often poorly controlled type I diabetes and characterized by stromal fibrosis and, therefore, increased MD.47 To our knowledge, no previous study examined high plasma glucose and MD; however, our inverse association is consistent with some48 but not all studies of MD in relation to diabetes.49 Almost half (46%) of the women in our study with high glucose (≥110 mg/dL) were also classified as diabetic. A cross-sectional study found diabetes to be associated with lower MD patterns (BIRADS, Breast Imaging Reporting and Data System) in premenopausal but not postmenopausal Native American women.48 In contrast, a cross-sectional analysis using data from the large Minnesota Breast Cancer Family Study did not reveal a difference in percent MD by diabetes status in BMI-adjusted models in premenopausal or postmenopausal Caucasian women.49 Findings from these studies, as well as our study, suggest that although raised glucose may be positively related to breast cancer risk,19 the effects of glucose on breast cancer risk are not mediated by an increase in percent MD.

The lack of association between percent MD and raised triglycerides or low HDL-cholesterol observed in our study is consistent with a study in postmenopausal Caucasian women, which adjusted for abdominal adiposity,50 but conflicts with studies in premenopausal women.27, 28 In a small sample of premenopausal Caucasian women (N = 273), Boyd et al. found an inverse association between triglyceride levels and percent MD after adjustment for age and BMI.28 Boyd et al. also found a positive association for percent MD with HDL-cholesterol levels,28 which was consistent with another study in premenopausal Caucasian women conducted by Furberg et al.27 Reasons for the discrepant findings across these studies may include residual confounding by abdominal adiposity due to inadequate adjustment by using BMI as a measure of overall body size.

Our study was conducted among a multi-ethnic, community-based cohort of midlife women and, therefore, is likely to be more generalizable to the United States population than previous studies of primarily Caucasian women or among clinic-based populations. The strengths of our study included the longitudinal measurement of the MetS, MD, and covariates by a specified protocol.29 The interviewers and MD assessor were blind to MetS status of the women, thereby limiting differential misclassification of exposure, outcome and important confounders, such as weight and height for the calculation of BMI. Fasting measurements were based on plasma samples taken on days 2 to 5 of a spontaneous menstrual cycle among 82% of our cohort, thus reducing variability in measurements of reproductive hormones and glucose.27 In addition, MD measurements were made using a planimeter assessment by a single expert with high reproducibility and accuracy.30, 31 She has been considered as the standard expert against which computer-based methods have been evaluated.30, 31

Our study also had limitations. Only a small percentage (15%) of our study population had the MetS, limiting our power to detect a small effect of the MetS or components of the MetS on MD, if an effect truly existed. For instance, we only had 54 women with high plasma glucose levels, which may account for the lack of statistical significance of our observed modest association on percent MD. Also, only a few women (N = 9) with normal BMI (BMI < 25 kg/m2) had the MetS, raising the question of whether results would be different in women with normal BMI. The later, however, is less of a concern given that our findings for the MetS did not change when analyses were limited to women with BMI < 27.5 kg/m2 (data not shown).

It is possible that we were unable to observe the hypothesized positive association between the MetS or components of the MetS and percent MD because we did not evaluate the most relevant etiologic time period. There is more consistent epidemiologic evidence for the positive association between the MetS and postmenopausal breast cancer than with premenopausal breast cancer.2, 6 Although we did not observe an association between the MetS and postmenopausal MD (data not shown), we only had these data among 271 women. Thus, it remains a possibility that the MetS is related to an increase in breast cancer risk through a mechanism involving MD, but that we did not have enough statistical power to observe the positive association, if in fact it exists.

Contrary to our hypotheses, we did not observe a positive association between the MetS nor any of the components of the MetS and percent MD. In conclusion, our results do not support the hypotheses that metabolic abnormalities affect breast cancer risk through a mechanism that involves an increase in percent MD or amount of dense breast tissue among premenopausal and early perimenopausal women.

Acknowledgements

The Study of Women's Health Across the Nation (SWAN) has grant support from the National Institutes of Health (NIH), DHHS, through the National Institute on Aging (NIA), the National Institute of Nursing Research (NINR) and the NIH Office of Research on Women's Health (ORWH) (Grants NR004061; AG012505, AG012535, AG012531, AG012539, AG012546, AG012553, AG012554, AG012495). The mammographic density ancillary study was supported by National Cancer Institute grant R01CA89552. Dr. Conroy was supported by a postdoctoral fellowship on grant R25 CA 90956. The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the NIA, NINR, ORWH or the NIH. The authors thank the study staff at each site and all the women who participated in SWAN.

Ancillary