Serum Vitamin D and Breast Density in Breast Cancer Survivors

Radiologically dense breast tissue is a risk factor for breast cancer (1-3). Tissue appearing dense on mammography has more fibroglandular tissue and actively proliferating epithelial and stromal cells relative to fatty tissue (3). Proliferating cells are more vulnerable to genetic damage, thereby increasing the risk of neoplastic events (2, 3). These factors are important to understand for breast cancer survivors because those women are at risk of recurrence or new primary tumors.

Breast density can be modified by diet and body weight (4, 5). Among dietary factors, vitamin D has received recent interest as a potential modifier of mammographic density. Vitamin D receptors are present throughout the human body, including the breast. Much recent evidence suggests that adequate vitamin D status may protect against breast cancer (6-8). Biological mechanisms to support these associations include the strong antiproliferative and proapoptotic properties of vitamin D (9-12) The possibility that vitamin D may exert its effect on breast cancer risk through decreasing mammographic density has been investigated in several observational studies in healthy women, but results have been inconsistent (13-17). Even less well understood is whether vitamin D influences mammographic density in cancer survivors. This issue is particularly important in breast cancer survivors who are at risk of recurrence. Further, because clinical guidelines for survivors include at least yearly mammograms, understanding factors that influence breast density is important. To further investigate this issue, we examined the associations of serum concentrations of serum 25-hydroxyvitamin D [25(OH)D] with breast density in an ethnically diverse sample of breast cancer survivors.

The Health, Eating, Activity, and Lifestyle (HEAL) study is a population-based, multicenter, multiethnic prospective cohort study of 1,183 breast cancer patients investigating whether weight, physical activity, diet, and sex hormones affect prognosis and survival. Details of the study design and procedures are published (18). Briefly, we used the Surveillance, Epidemiology, End Results (SEER) registries in New Mexico, Los Angeles County (CA), and western Washington for study recruitment. In New Mexico, we recruited 615 women, ages 18 y or older, diagnosed with in situ to stage IIIA breast cancer between July 1996 and March 1999, and living in three counties. In Washington, we recruited 202 women, between the ages of 40 to 64 y, diagnosed with in situ to stage IIIA breast cancer between September 1997 and September 1998, and living in a three-county area. In Los Angeles, we recruited 366 African-American women with in situ to IIIA primary breast cancer, who had previously participated in one of two breast cancer case-control studies. The Los Angeles eligible participants are a subset of women diagnosed with breast cancer between May 1995 and May 1998, were ages 35 to 64 y at diagnosis, English speaking, and U.S. born. Procedures were approved by the Institutional Review Boards of the participating centers and all women signed informed consent.

HEAL participants completed examinations at baseline (within their first year postdiagnosis; on average, 7.5 mo postdiagnosis) and 24 mo later. Of the 1,223 women initially enrolled, we excluded 39 (3.2%) women who were subsequently found to have had a prior diagnosis of breast cancer and 1 woman (<1.0%) with metastatic disease at initial diagnosis because these women no longer met the enrollment criteria. Of the remaining 1,183 women, 944 (79.8%) completed the 24-mo follow-up interview, which included detailed questions on health, menopausal status, diet, dietary supplement use, physical activity, tamoxifen use, and alcohol and tobacco use. The staff also measured height and weight, and collected fasting bloods. Reasons for nonparticipation in the 24-mo follow-up were death (n = 44), illness (n = 2), refusal (n = 105), moved (n = 16), and unable to contact (n = 72). Analyses in this report are restricted to postmenopausal women (n = 426) due to the marked changes in breast density at menopause and because we do not have menstrual cycle data for premenopausal women in relation to mammogram dates; the menstrual cycle influences breast density measures (19).

Data on breast cancer stage of disease were obtained from the SEER registries. Participants were classified as having in situ, stage I, or stage II to IIIA breast cancer based on American Joint Committee on Cancer stage of disease classifications as recorded by SEER. Additionally, treatment and other clinical data were obtained from a medical record review that provided more detailed information on chemotherapy and hormonal therapy than was available within SEER.

A fasting blood sample was collected at the 24-mo interview. Blood was processed within 3 h of collection and was stored at −70° to −80°C until analysis. Although the biologically active form of vitamin D is [1,25(OH)D], it is not considered a good biomarker due to its short half-life and tight homeostatic control (20). [25(OH)D] is an excellent biomarker of vitamin D status and it represents both cutaneous synthesis and intake from foods and supplements (21, 22). Serum [25(OH)D] was assayed at the Medical University of South Carolina using a radioimmunoabsorbant assay (DiaSorin, Inc.). We included blinded duplicates in each assay run to evaluate assay precision and the coefficient of variation for each of the three batches analyzed was 3.7%. The personnel performing the assays were blinded to patient identity and all personal and medical characteristics.

Mammographic films, corresponding to ∼1.5 y after diagnosis, were retrieved from the individual providers. Each film was digitized using either an Epsom 1680 scanner (Epson America, Inc.; used in Washington state), or Cobrascan CX-812 M large format 12-bit X-ray scanner (Radiographic Digital Imaging; used in New Mexico and Los Angeles). We measured the craniocaudal view contralateral to the breast diagnosed with breast cancer for mammographic percentage density and total dense area (measured in thousands of pixels and converted to square millimeters by multiplying by 0.0676). The density readings were conducted by a trained technician using Cumulus 108, a computer-assisted mammogram-measurement program developed at the University of Toronto (Ontario, Canada). This method has been described in detail elsewhere (5). Briefly, the reader uses a sliding scale to outline the breast edge and then the dense breast area is based on pixel brightness. Percent density is the proportion of dense breast area relative to the total area of the breast. The reader was unaware of any participant clinical characteristics.

Participants wore light clothing and weight was measured to the nearest 0.1 kg using a balance-beam laboratory scale (New Mexico and Washington) or portable scale (Los Angeles). Height was measured to the nearest 0.1 cm using a stadiometer. All measurements were done and recorded twice in succession, then were averaged for a final value for analyses. Body mass index (BMI) was computed as [weight (kg)/height (m²)]; BMI was then categorized as: normal, ≤25.0 kg/m²; overweight, 25.0-29.9 kg/m²; and obese, ≥30.0 kg/m² (23).

Standardized questionnaire information was collected on medical history, family history of breast and other cancers, smoking, physical activity, demographic characteristics, and race/ethnicity. None of the women in this analysis used postmenopausal hormone therapy. Dietary vitamin D was assessed at the 24-mo interview with a self-administered 121-item food frequency questionnaire (24). Use of vitamin D–containing dietary supplements was estimated by self-report and was coded as a binary variable because dosage data were not available. Season of mammogram (fall, summer, winter, and spring) was used as a surrogate of UV-B sunlight exposure (22, 25).

Our analytic objective was to determine whether the primary exposure of interest (circulating vitamin D) was associated with measures of mammographic density. Serum concentrations of [25(OH)D] were categorized using clinical cutpoints for deficient (<16.0 ng/mL), marginal (16.0-31.9 ng/mL), or sufficient (≥32.0 ng/mL; refs. 20, 21). Multivariate linear regression tested for differences in breast density measures across the categories of vitamin D status. Based on our previous work as well as on regression diagnostics for this study, several covariates were considered a priori as potential confounding factors and were included in all models: age (continuous), BMI (continuous), tamoxifen use (yes/no), stage of disease, breast cancer treatment (surgery, radiation, chemotherapy, combinations), season of mammogram and blood draw (fall, winter, spring, or summer), physical activity [Metabolic equivalents (METs)/wk continuous], and number of intact ovaries (0, 1, or 2; refs. 5, 26). A race/ethnicity/study site variable was created to adjust for confounding because race/ethnicity and study site were highly correlated: non-Hispanic white in New Mexico, non-Hispanic white in Seattle, Hispanic, and African-American. Dietary vitamin D from the food frequency questionnaire and vitamin D–containing dietary supplements (single supplements and known combinations, such as calcium + vitamin D) were examined as potential confounders but were not used in the final models because they were neither statistically significant nor influential on parameter estimates. For analytic purposes, serum [25(OH)D] concentrations were log transformed to achieve a more normal distribution; geometric means with 95% confidence intervals are presented. All analyses were conducted with SAS version 9.1 (SAS Institute, Inc.).

Table 1 provides descriptive data for the 426 postmenopausal HEAL participants with mammographic density data. The mean age of participants was 61.0 years and nearly two-thirds were overweight or obese (BMI ≥ 25.0 kg/m²). The sample was racially diverse—22.8% were African-American, 62.8% were non-Hispanic white, and 11.3% were Hispanic. Only 99 (23.2%) of HEAL participants had circulating [25(OH)D] concentrations considered clinically adequate (≥ 32.0 ng/mL). Pearson correlations of serum [25(OH)D] with breast density were weak, but statistically significant; r = 0.11 (P = 0.02) for breast dense area and r = 0.15 (P = 0.002) for percent density (data not shown).

Across the three serum [25(OH)D] categories, percent breast density did not vary for all participants or for racial/ethnic subgroups (Table 2). In general, non-Hispanic white women from New Mexico had slightly higher percent breast density across the vitamin D categories than other women. Formal tests for multiplicative interaction revealed no effect modification of the vitamin D–breast density association by race/ethnicity.

We next examined associations of breast dense area (cm²) with serum [25(OH)D] for all participants and for each racial/ethnic group in the cohort (Table 2). We observed no association overall of serum [25(OH)D] with breast dense area after multivariate adjustment. Stratification by race/ethnicity revealed a marginal, positive correlation (P = 0.05) between serum [25(OH)D] and mammographic breast density, but only among non-Hispanic white women in Seattle. Serum [25(OH)D] was not related to mammographic breast density among non-Hispanic whites in New Mexico, African-American, or Hispanic women. Additional analyses modeled vitamin D from food + supplements, but there was no association of these exposures with mammographic density (data not shown).

Understanding factors that influence breast density in breast cancer survivors may be an important clinical component of reducing the risk of recurrence or new primary tumors among survivors. In this study of postmenopausal breast cancer survivors, we found little evidence for an association of circulating [25(OH)D] with mammographic breast density; results were similar using both percent breast density and dense area as outcomes. We hypothesized that serum concentrations of [25(OH)D] would be inversely associated with breast density. This rationale was based on evidence supporting an antiproliferative role for vitamin D in many tissues in the body, including the breast (7-12).

To our knowledge, this is the first report of serum [25(OH)D] and mammographic breast density in breast cancer survivors. Other reports examined healthy women recruited from screening mammography programs (13) or those enrolled in a surveillance study of women with family history of breast cancer (16). Knight (16) observed no association of serum [25(OH)D] with mammographic density in 487 women who had no evidence of breast cancer. Conversely, Brisson (13) reported that among 741 healthy, premenopausal women, breast density varied by season; mean [25(OH)D] (by month) was inversely associated with smoothed mean breast density after allowing a 135-day lag time for [25(OH)D] to affect the breast tissue (r = −0.90). Other studies of vitamin D and mammographic breast density have used self-reported diet and/or supplemental intake of vitamin D, instead of a biomarker, and results have been inconsistent. Berube (14) reported that among healthy women, those with higher (≥200 IU/d) versus lower (≤49 IU/d) intake of vitamin D had a 0.24 relative odds (95% confidence interval, 0.11-0.53) for dense (versus nondense) breast tissue, which was no longer statistically significant after adjustment for calcium (odds ratio, 0.40; 95% confidence interval, 0.14-1.13). A subsequent study revealed that vitamin D (food + supplements) was inversely associated with breast density among premenopausal, but not postmenopausal women (each 100 IU vitamin D = 1.4% decrease in breast density; P = 0.004; ref. 15). Taken together, the results presented in this report plus relatively weak findings from other studies do not offer strong evidence for an association of vitamin D with breast density, regardless of whether self-reported diet and supplement use or a biomarker is used to assess vitamin D exposure.

Among the strengths of this study is that we measured serum concentrations of [25(OH)D], which is the most stable measure of vitamin D status in the body (18-20). Blood samples were drawn at the 24-month follow-up examination when breast cancer treatment other than tamoxifen, if prescribed, was finished. Thus, [25(OH)D] is unlikely to be affected by concurrent chemotherapy or radiation in this study. All mammograms were evaluated with a computer-assisted method and the reader was unaware of each participant's vitamin D status. There are also study limitations. Most HEAL participants (76.8%) had vitamin D insufficiency or frank deficiency (26); the lack of heterogeneity in our primary exposure may have prevented us from detecting an association. Another limitation is that our analyses by racial/ethnic subgroups resulted in small cell sizes for those groups. Although some studies have shown differences in mammographic density by race/ethnicity and the potential for diet to explain these differences (17, 27-30), we likely had insufficient power to detect meaningful differences across race. The blood draws and mammograms did not occur on the same date, but we did adjust for both of these dates by season in the models. Finally, as with all observational studies, we cannot rule out the possibility of residual confounding.

In conclusion, results from this cohort of multiethnic breast cancer survivors do not support the hypothesis that vitamin D status, as assessed by serum [25(OH)D], is associated with mammographic breast density. Despite the strong association of mammographic density with breast cancer risk (3), the mechanism is not likely to be mediated by vitamin D.

No potential conflicts of interest were disclosed.

Grant Support: National Cancer Institute contracts N01-CN-75036-20, N01-CN-05228, N01-PC-67010/N01-PC35139, N01-PC-67007/N01-PC-35138, and N01-PC-67009/N01-PC-35142 and NIH training grant T32 CA09661. A portion of this work was conducted through the Clinical Research Center at the University of Washington and supported by NIH grant M01-RR-00037. Data collection for the Women's Contraceptive and Reproductive Experiences Study (CARE) at the University of Southern California was supported by the National Institute of Child Health and Human Development contract N01-HD-33175. Patient identification was supported in part by the California Department of Health Services grant 050Q-8709-S1528.

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.