The results of this study indicate that there is a cross-sectional relation between serum vitamin D and percent breast density in postmenopausal women, yet this association is eliminated by adjustment for BMI. We found no evidence that any of the examined molecules of the vitamin D pathway are independently associated with percent breast density after adjustment for BMI, and little evidence that they modify the relation between vitamin D and percent breast density.
These findings add to growing evidence suggesting no strong relation between vitamin D and mammographic breast density in postmenopausal women. A handful of studies have examined dietary intake of vitamin D in relation to breast density. With a couple of exceptions [32
], the majority of these studies reported null associations in postmenopausal women [34
]. In contrast, an inverse association between dietary intake of vitamin D and breast density has been more consistently observed among premenopausal women [32
]. To our knowledge, only three previous studies have examined breast density in relation to circulating vitamin D levels in postmenopausal women, each of which reported null associations [13
]. Inverse [14
] and null [17
] findings have been reported in studies of premenopausal women.
To our knowledge, this is the first study to examine circulating PTH in relation to breast density. PTH can stimulate the proliferation of quiescent MCF7 breast cancer cells in vitro [40
], and a high percentage of human mammary cancers and hyperplastic mammary epithelial cells express the PTH receptor [9
]. A number of medical record linkage studies have reported that hyperparathyroidism is a risk factor for breast cancer [41
]. However, we found that there was no independent association between PTH and percent breast density in our study.
Previous studies have observed a positive association between circulating IGF-1 levels and breast cancer risk, though this has largely been limited to premenopausal women (reviewed in [10
]). We observed no independent associations between breast density and IGF-1, IGFBP-3, or the IGF-1:IGFBP-3 molar ratio. These results are consistent with previous studies which observed null relations between breast density and IGF-1, IGFBP-3, and their molar ratio in postmenopausal women [44
]. In contrast, significant associations have been observed in premenopausal women [44
We found no evidence for an influence of either calcium or retinol on the relation between vitamin D and breast density. We had also hypothesized that the effect of vitamin D on breast tissue may vary according to sex hormone levels. For instance, the influence of vitamin D may be limited in women with high estrogen levels, such as those who are obese. We observed some suggestion that the relation between vitamin D and percent breast density varied according to BMI (Pinteraction = 0.05); the association was positive among women with low BMI and negative among women with high BMI. However, this association did not reach statistical significance in any specific BMI category. Direct evaluation of effect modification by serum estradiol did not reveal a strong interaction, though there was the suggestion of an inverse association between vitamin D and breast density in women with estradiol levels above the median.
Our results provide further evidence against a strong relation between the molecules of the vitamin D pathway and breast density in postmenopausal women. Interpretation of these results should be balanced by consideration of the study’s limitations. Measurements of serum molecules and breast density were made at one single point in time in this cross-sectional study. The half-life of 25(OH)D in the blood is only 2-4 weeks [48
], thus variation in past exposure levels are not captured by a single measurement. The resulting misclassification may attenuate any association between vitamin D and breast density, and we cannot evaluate the temporality of the relation. However, blood levels of vitamin D generally remain consistent over time in individuals. In the Nurses Health Study an intraclass correlation coefficient of 0.72 was observed for 25(OH)D measures taken two to three years apart among postmenopausal women [50
]. In the general population, moderate correlation (r = 0.5) has been observed at up to 14 years between measurements [51
]. Additionally, very little intra-individual diurnal variation in 25(OH)D levels has been observed [48
Breast density was assessed from digital mammograms using Cumulus software. Cumulus was designed for use on film images that have been digitized. Use of digital images may have introduced measurement error which could obscure a relation between breast density and vitamin D or other circulating molecules. However, we were able to detect the expected associations between breast density and established risk factors, such as age, body mass index, and parity, suggesting that any measurement error for breast density was within reasonable limits.
Multiple imputation was used to impute the small numbers of missing laboratory and covariate data. Sensitivity analyses in which subjects with missing laboratory values were excluded revealed a negligible impact upon the results.
While the study was sufficiently powered to detected clinically relevant differences in breast density, the relatively small sample size of the study limited the power to detect smaller differences in breast density, and we are unable to rule out potential interactions between circulating molecules. Since all women were recruited from UW Health Clinics in Madison, Wisconsin, the generalizability of the results is somewhat limited. The limited geographic range of participants likely restricted the observed variation in vitamin D exposure. Ethnic diversity in our sample was minimal, as ~97% of subjects reported white race (reflecting the clinic patient demographics). Additionally, eligibility was restricted to postmenopausal women, aged 55-70, who had never used postmenopausal hormones. While this provided an optimal opportunity to detect an effect of vitamin D on density (i.e. in a low endogenous hormone background), the impact of vitamin D in premenopausal women or users of postmenopausal hormones could not be evaluated.
In summary, we found little evidence to support an independent association between molecules of the vitamin D pathway and breast density in postmenopausal Caucasian women. While it remains possible that vitamin D could influence breast cancer risk, our results suggest that such an effect would be mediated through pathways other than breast density. Continued investigation into the influence of vitamin D on breast tissue will be needed to understand the potential mechanisms by which this molecule may reduce breast cancer risk.