No Association Between 25-Hydroxyvitamin D and Mammographic Density

Vitamin D is antiproliferative and proapoptotic in vitro (1, 2) and a small number of epidemiologic studies support a potential protective effect against breast cancer (3-5). Several studies have examined evidence for an association between dietary and supplemental vitamin D intake and mammographic density (6-8). Bérubé et al. (6) found an inverse association between dietary intake of vitamin D and mammographic density after adjustment for potential confounders (odds ratio, 0.24; 95% confidence interval, 0.11-0.53, for ≥200 versus <50 IU/d), with evidence for a stronger association with lower density in those with high intake of both vitamin D and calcium. This finding was replicated for premenopausal women only in a separate study with information on both dietary and supplemental vitamin D (7). In contrast, a previous analysis of the Minnesota Breast Cancer Family Study found no association between dietary and supplemental vitamin D intake and mammographic density, but analyses were not stratified by calcium intake (8).

The conversion from 25-hydroxyvitamin D (25OHD) to the biologically active form is tightly regulated and the circulating level of 25OHD better represents total vitamin D exposure (9, 10). A number of tissues, including breast, can locally convert 25OHD to the active form, making 25OHD etiologically relevant (11, 12). In recent results from the Nurses' Health Study, prediagnostic plasma 25OHD was significantly lower in breast cancer cases compared with controls (5). We examined the association between circulating 25OHD and the intermediate phenotype of mammographic density (13) within a sample of women from the Minnesota Breast Cancer Family Study who had both blood samples and mammograms available.

Details of the baseline (14) and first follow-up (15) phases of the Minnesota Breast Cancer Family Study have previously been described. Briefly, a family study of breast cancer was initiated in 1944 at the University of Minnesota. Probands were ascertained at the Tumor Clinic of the University of Minnesota Hospital between 1944 and 1952 (n = 544). From 1990 to 1996, 426 (78%) families of the 544 probands were updated (first follow-up); each proband's first- and second-degree female relatives and spouses of male relatives were contacted, and extensive risk factor data were collected by telephone interview on 6,194 women (94.6% of those eligible). Females of ages >40 years were asked to provide a recent mammogram. If one had not been taken in the previous year (2 years if under age 50 years), they were instructed to obtain a new one through their personal physician. At the time of first follow-up, usual dietary intake and vitamin supplement use over the past year was also assessed with a semiquantitative Food Frequency Questionnaire adapted from Willett et al. (16); 3,598 (57.9%) women returned the questionnaire.

In 2001, a second follow-up of these families was conducted through a mailed questionnaire to obtain additional risk factor and cancer information and authorization for release of prior mammograms taken over the past 10 years (17). Of the 6,194 eligible women who had completed first follow-up, a total of 3,743 women completed this questionnaire (77.1% of those contacted and competent and 60.4% overall).

Blood as a source of DNA and serum was obtained on 1,328 women in the family cohort study for purposes of several substudies, including a nested breast cancer case-control study, linkage analysis of breast density, and examination of penetrance of major breast cancer genes.

Mammograms for the study were obtained at time of both first follow-up (n = 2,491) and second follow-up (n = 2,070); the goal of the additional mammogram retrieval at second follow-up was to ascertain serial mammograms on participants. The participation rate for the mammogram component at first follow-up was 48% of women ages ≥40 years; at second follow-up, 87% of eligible women provided authorization to retrieve their mammograms but we ascertained images on 69% (n = 2,070); the others were either unavailable (n = 295) or not attempted (n = 539). Women with breast cancer before the mammogram, however, were excluded (n = 97).

Both the left craniocaudal and mediolateral oblique views were digitized on a Lumiscan 75 scanner with 12-bit grayscale depth. The pixel size was 0.130 × 0.130 mm² for both the 18 × 24-cm² and 24 × 30-cm² films. Percent mammographic density [(dense area / total area) × 100] and absolute dense area (cm²) were estimated for each view using a computer-assisted thresholding program (Cumulus; ref. 18). All images were read by a single trained technician. Although we could not explicitly distinguish screening mammograms from those taken for diagnostic purposes, as the mammograms were all standard screening four-view (two views per breast) and women with breast cancer were excluded, it is likely that most or all of the mammograms were for screening purposes.

Serum samples from 488 women with a craniocaudal view and with sufficient serum available were assayed for 25OHD. Serum 25OHD was measured with the Diasorin RIA kit (Stillwater, MN), with which our laboratory holds a certificate of proficiency from the Vitamin D External Quality Assessment Scheme (Charing Cross Hospital, London, United Kingdom). Between-assay coefficient of variation for various levels of quality-control sera was <10%. One sample did not have detectable 25OHD. The observed range in values was 14 to 174 nmol/L. Of the 487 women included in this analysis, 356 (73%) had their blood drawn within 1 year of the mammogram used in this analysis and 131 (27%) had their blood drawn more than 1 year after the mammogram, but within 4 years. One woman had her blood taken ∼8 years after the mammogram.

We compared mean values of mammographic density and circulating levels of vitamin D across categories of covariates using ANOVA. Four density measures were considered: percent density and dense area from the craniocaudal view, and percent density and dense area from the mediolateral oblique view. If more than one mammogram was available, the one with at least a craniocaudal view taken nearest the blood sample was selected for the analysis. The following potential risk factors were considered as covariates: age at interview, season of blood sample receipt (October-March versus April-September), family history (first or second degree), age at menarche (<12, 12, 13, >13 years), menopausal status at interview, parity and age at first birth (nulliparous, <3 at age 20 years or earlier, <3 after 20 years, 3+ at age 20 years or earlier, 3+ after 20 years), hormone replacement therapy (never, former ≤5 years, former >5 years, current ≤5 years, current >5 years), body mass index (BMI; <20, 20-24.9, 25-29.9, ≥30 kg/m²), physical activity, education (less than high school, high school, some college, college graduate), smoking status (never, former, current), alcohol usual frequency of ≥1 drink (daily, weekly, monthly, never), exposure to sunlight (less than average, average, more than average), skin type (fair, medium, dark), calorie intake (quartiles), intake of calcium (quartiles), and use of dietary supplements (current use, yes/no). Physical activity was derived from questions on frequency of moderate and vigorous activities and defined as three levels (low, moderate, and high). A woman was considered postmenopausal if she responded with a “no” when asked whether she had had a menstrual period within the last year, not including periods brought on by taking hormones. Any woman responding “yes” to this question was considered premenopausal unless she reported her periods had stopped because of pregnancy or breast-feeding. All covariates were derived from information collected either by telephone interview at first follow-up or by mailed questionnaire at second follow-up as described above.

Comparisons of the four breast density measures across quartiles of serum 25OHD were carried out after adjustment for the effects of potential confounding variables using analysis of covariance (ANCOVA). Two sets of models were fit: one adjusting for age and BMI, and one adjusting for age, BMI, and other covariates. Covariates associated with both circulating levels of 25OHD and any of the four breast density measures (P < 0.10), after accounting for the effects of age and BMI, were considered potential confounding variables and included in the final ANCOVA model. These covariates included parity, age at first birth, and physical activity. Least squares means and corresponding SEs were used to describe the covariate-adjusted associations between breast density and vitamin D levels.

We also determined if season of blood sample receipt, menopausal status, or calcium intake modified the breast density-vitamin D associations. For these analyses, calcium intake was classified as the upper 25% (>1,385 mg) versus the lower 75% (≤1,385 mg). We assessed effect modification using a formal test for interaction, adjusted for the same potential confounding variables as outlined above.

All hypothesis tests were two sided, and all analyses were carried out with the SAS software system (SAS Institute, Inc., Cary, NC). Primary analyses assumed independence across all observations. However, we also conducted analyses using general linear mixed models to account for correlation among individuals within the same family. For all such models, family-specific correlations were modeled using a compound symmetry covariance matrix.

There were 487 women in the study who did not have breast cancer, had sufficient serum samples available, had 25OHD results, and also had estimates of mammographic density from at least a craniocaudal view. The mean age was 56.4 years (range, 27-85 years) and the mean BMI was 27.2 kg/m² (range, 18.0-64.3 kg/m²). Among the 356 women with dietary information available, the mean calorie intake was 1,974 (range, 673-4,210) and the mean calcium intake was 1,075 mg (range, 153-3,418 mg). Table 1 shows the relationship between a number of covariates and 25OHD, percent density (craniocaudal view), and dense area (craniocaudal view). As the results for the mediolateral oblique view were very similar to the craniocaudal view, only the results for craniocaudal view are shown. As expected, unadjusted mammographic density was inversely associated with age, parity, menopausal status, and BMI, and positively associated with age at first birth. There was also some evidence of positive associations with alcohol consumption and physical activity. Vitamin D was strongly inversely related to BMI and positively related to physical activity and there was some evidence of a relationship with parity and age at first birth. As expected, 25OHD was positively related to sun exposure.

After adjustment for age and BMI, there was no significant relationship between 25OHD and mammographic density, although there was a nonsignificant trend of increasing density with increasing 25OHD (Table 2). This trend was attenuated after further adjustment for parity and age at first birth as well as physical activity. As the time between sample and mammogram ranged up to 4 years (with one outlier at 8.4 years), we also did analyses including only the 356 (73%) of women who had blood drawn within 1 year of their mammogram, but we observed no difference (data not shown).

As season of blood draw was highly related to 25OHD with significantly lower levels in winter, we looked for potential differences in the relationship between 25OHD and mammographic density by season (Table 3). There was no evidence for any relationship in either season. There was also no evidence for differences by menopausal status.

We examined the relationship between 25OHD and mammographic density separately for those in the upper quartile of calcium intake and those with lower intakes among the 356 women who had this information available (Table 3). There was marginal evidence for an interaction between 25OHD and calcium for percent density (P = 0.05), but there was no clear trend among those with high calcium intake. There was no evidence for interaction for dense area (P = 0.23). However, the lowest observed values for both percent density and dense area occurred among those women in both the highest quartile for calcium intake as well as the highest quartile of 25OHD.

Analyses accounting for possible familial correlations did not appreciably change results (data not shown).

We did not find evidence for a relationship between mammographic density and serum 25OHD. Two previous studies of the association between dietary vitamin D (6) or dietary and supplemental vitamin D (7) and mammographic density found an inverse relationship, but a previous analysis of the Minnesota Breast Cancer Family Study (n = 1,694 women) found no association between dietary and supplemental vitamin D and mammographic density (8). Our current findings with a serum biomarker of vitamin D confirm the previous findings in this cohort.

There are a number of possible reasons that could explain our lack of association. First, there might truly be no association between serum 25OHD and mammographic density. Alternatively, it is possible that vitamin D exposure earlier in life is more relevant to both mammographic density and breast cancer. Bérubé et al. (7) found a relationship between current intake of vitamin D and density primarily among premenopausal women. However, we observed no evidence for an association in either premenopausal or postmenopausal women, although our definition of menopausal status was fairly crude. The number of premenopausal women was quite small (n = 133) and it is possible that we had insufficient statistical power to detect a relationship, although no trends were evident. In addition, the time between blood sample and mammogram varied and the measure of vitamin D may not always have reflected the level at the time the mammogram was taken, although when we restricted the time interval to within 1 year, the results did not change. Vitamin D levels can vary considerably by season as sun exposure is an important source, but we did not find evidence of a relationship with density in either winter or summer. Long-term stability of 25OHD is less clear, but intra-woman variability in 25OHD measurements of between 13% and 19% has been observed over a 5-year period (19).

In the two studies of Bérubé et al., the association between vitamin D intake and mammographic density was most pronounced among those with the highest levels of dietary or total calcium intake, with a reduction in mean density of 8.5% in premenopausal women with a daily total intake of 400 IU of vitamin D and 1,000 mg of calcium (6, 7). In the present study, we investigated a potential interaction between vitamin D and calcium in an attempt to replicate the previous studies. There was marginal evidence for an interaction between calcium intake and serum 25OHD with respect to percent density, but not dense area, in the subset where this information was available. There were no clear trends, but it should be noted that the lowest levels of density were observed in those with the highest levels of 25OHD and the highest intake of calcium. However, the marginal interaction observed with percent density could simply have occurred by chance. Vitamin D is involved in calcium regulation, but the exact mechanism of an interaction with respect to breast density is not clear.

Although we did not confirm the previous association of dietary vitamin D and mammographic density, future studies should focus on premenopausal women, as this might be the time period of greatest importance. More recent work also suggests that insulin-like growth factor-I and insulin-like growth factor–binding protein-3 levels may also influence the relationship between vitamin D and mammographic density in premenopausal women (20). The inverse association between vitamin D and/or calcium and mammographic density was observed primarily in premenopausal women with insulin-like growth factor-I or insulin-like growth factor–binding protein-3 levels above the median. These findings suggest that the relationship between diet and mammographic density is complex. In general, the inconsistencies of associations seen between dietary factors likely reflect this complexity (6, 21, 22).

Strengths of the study include the first examination of a serum biomarker of vitamin D with mammographic density. In addition, we used a well-characterized assessment of mammographic density and a measure of serum vitamin D that better reflects total exposure than self-reported intake. However, our study had limitations including dietary data available on a limited subset of women and mammograms that did not correspond exactly with the timing of blood draw.

In conclusion, the data on vitamin D and breast cancer risk consistently show an inverse association. However, the biological basis for this association is not completely understood. Our results suggest that this association is not mediated by an effect on mammographic density, at least in postmenopausal women.