Women with a family history of breast cancer may be at higher genetic risk for breast cancer because of either recognized deleterious mutations or as yet unidentified low-penetrance alleles. Such women are likely to be especially motivated to modify their lifestyle to reduce their risk for breast cancer. Diet modification is one possible strategy, but few previous studies evaluating diet and breast cancer have focused on women at high genetic risk 12.

Breast density, the percentage of total breast area with a mammographically dense appearance, is an informative marker for breast cancer risk because of its strong association with breast cancer 345. Breast density has been shown to predict risk among women with a genetic predisposition for breast cancer 6, including women carrying BRCA1 or BRCA2 mutations 7. Among the few studies on diet and breast density 8910111213141516, only one specifically included women with a family history of the disease 17. The objective of the present study was to evaluate associations between dietary intake and breast density among women at potentially high genetic risk for breast cancer. We focused on dietary factors associated with breast density in previous studies conducted among average-risk samples 89101213141518, as well as on suspected risk factors or protective factors for breast cancer 19

The 157 women in the analysis had a mean ± standard deviation age of 50 ± 8 years; 96% were white, and 73% had at least a college degree. Among the mammographic images included in the analyses, 12.7% (n = 20) were classified as entirely fatty, 28.0% (n = 44) as having scattered fibroglandular tissue, 34.4% (n = 54) as heterogeneously dense, and 24.8% (n = 39) as extremely dense. Having denser breasts was associated with younger age, lower body mass index, older age at menarche, smaller number of live births, older age at first live birth, and a history of having used hormone replacement therapy (Table 1). Menopausal status was inversely associated with breast density, but not significantly after adjustment for age. Level of education, current smoking, and a hereditary family history of breast cancer showed nonsignificant positive associations with breast density. We observed no associations with breast density for breastfeeding, history of oral contraceptive use, or physical activity.

Breast density was significantly inversely associated with intake of vitamin D (Table 2); women in the highest tertile of vitamin D intake were one-half as likely to have higher breast density as women in the lowest tertile (odds ratio, 0.5; 95% confidence interval, 0.2–1.0). We also observed a suggestive but nonsignificant inverse association for folate (odds ratio, 0.5 for highest tertile versus lowest tertile; 95% confidence interval, 0.2–1.1; trend P = 0.09). Breast density was not associated with any other nutrients of interest (Table 2) or with any foods of interest (meats, fruits and vegetables, tofu, alcoholic beverages – results not shown) among all participants combined.

In stratified analyses, we found evidence for effect modification by family history category (Table 3). Whereas protein intake and animal protein intake were not significantly associated with breast density among women with a hereditary pattern of family history of breast cancer, these nutrients were positively associated with breast density among women with sporadic or other familial histories. The odds ratios for women with intake above the median were 3.0 (95% confidence interval, 1.3–6.9) and 4.3 (95% confidence interval, 1.8–10.3) for protein and animal protein, respectively. These results were not meaningfully different when we controlled for number of live births and age at first live birth in the 120 women with complete data (results not shown).

In this broad exploration of dietary factors, we found that only vitamin D intake was inversely associated with breast density in a sample of women enrolled in a high-risk program for breast cancer. Protein and animal protein were positively associated with breast density, but only among women not classified as having the hereditary familial cancer patterns.

Our results are consistent with recent work showing an inverse association between vitamin D and breast density 1418, as well as studies suggesting a protective effect of vitamin D against breast cancer 27. A previous study observed an inverse association of vitamin D with breast density only among premenopausal women 18, and primarily among premenopausal women with higher levels of insulin-like growth factor 1 or of insulin-like growth factor binding protein 3 28. In the present study, we saw no difference in effect by menopausal status. Vitamin D may lower breast density through the antiproliferative and proapoptic effects of its biologically active form, 1,25-dihydroxyvitamin D, or through modulation of the immune system 2729. Notably, most tissues and cells, including those in the breast, have vitamin D receptors and are also able to metabolize circulating 25-hydroxyvitamin D to 1,25-dihydroxyvitamin D, allowing for local, cellular effects in the breast tissue 29.

Whether the inverse association we observed is truly attributable to a protective effect of dietary vitamin D, however, remains unclear. We did not assess sunlight exposure, another important determinant of vitamin D status, nor did we have serum samples available to assess the relationship of breast density either with 25-hydroxyvitamin D as an indicator of vitamin D status or with 1,25-dihydroxyvitamin D. Of two previous studies that examined 25-hydroxyvitamin D in relation to breast density, one study saw a protective effect 30 while the other study did not 30. In additional analyses we explored whether breast density might be more strongly associated with major dietary sources of vitamin D, such as dairy or fish, but we observed no clear associations (results not shown).

Our results are also consistent with previous studies that found positive associations of breast density with protein or meat intake 111231. Animal protein intake may increase breast density by increasing circulating levels of insulin-like growth factor-1 3233, which has been linked to higher breast density 34 and higher breast cancer risk 353637 among premenopausal women. In the present study, the associations for protein and animal protein were limited to women without a hereditary family history of breast cancer. Classification of a hereditary family history of breast cancer in our sample primarily represented the presence of a BRCA1 mutation or a BRCA2 mutation. While this effect modification may be a chance finding, it may also suggest that, in such women, the same genetic factors that predispose to higher breast cancer risk increase the risk for dense breasts as well 38, and that genetic effects can overwhelm the effects of reduced dietary protein. For example, if BRCA1 exerts downstream effects on transcription, expression, or circulating levels of members of the insulin-like growth factor system as suggested by recent work 394041, then breast tissue may be denser among BRCA1 mutation carriers 38 regardless of their level of protein intake.

Other dietary factors may affect breast density by affecting estrogen levels or by directly affecting breast tissue morphology 81014 – but only limited, indirect evidence exists to support these mechanisms, and previous study findings with respect to other dietary factors have been inconsistent. In cross-sectional analyses, breast density has been both positively associated 891131 and inversely associated 1317 with total and saturated fat intake. Findings for soy and isoflavones have been similarly mixed 1315164243. Among other nutrients, higher breast density has been related to alcohol intake 174445 and inversely related to carotenoid, fiber, and calcium 8914184546.

Previous studies have observed a protective effect of folate against breast cancer 47, possibly because of its role in DNA methylation, synthesis, and repair 48, although breast density studies have not supported this 1745. In our sample we observed only a suggestive, nonsignificant inverse association. A protective effect of folate may be more pronounced among women with high alcohol intake 47, and alcohol intake in our sample was generally low (median 1.1 g/day alcohol). While too few women reported even moderate alcohol intake to examine folate in this group separately, when we excluded nine women with alcohol intake >15 g/day the inverse association for folate was slightly attenuated (trend P = 0.27).

Difficulties in assessing breast density may contribute to the inconsistent results of studies on diet and breast density. Most studies, including the current analysis, defined breast density in categories of percentage density. While the method is more subjective than a computer-assisted method 49, the Breast Imaging Reporting and Data System categories are nevertheless a strong predictor of breast cancer risk 50. Some workers have suggested that absolute size of the area of density is the more relevant measure 51, but only four previous studies presented results for dense area size 11151642 – three of which focused specifically on soy or isoflavones 151642.

The relevant time period of exposure is also unclear. Women randomized to a low-fat, high-carbohydrate diet for 2 years as part of the Canadian Diet and Breast Cancer Prevention Study Group experienced a greater decrease in area of breast density relative to a nonintervention group 1011, suggesting that recent, relatively short-term dietary exposures can affect breast density and possibly breast cancer risk, at least in premenopausal women or perimenopausal women. In contrast, a 2-year soy intervention had no effect on breast density change over time, although soy consumption in early life and adulthood did appear to influence density change 16.

A limitation of the present study is its relatively small sample size, which limited our ability to detect significant associations and effect modification by menopausal status. Our sample is unique, however, in that all women had a family history of breast cancer. Moreover, over 40% of them had a family history indicating the presence of a high-penetrance cancer predisposition gene, or had been tested to confirm it.

Examining only the recent diet may additionally have limited our ability to evaluate the role of dietary intake in relation to breast density. Findings were not different when we used the average value of dietary intake reported at enrollment into the FRAP and that reported for the current breast density study, with a median time interval between the two administrations of 4.5 years, but no other dietary information was available to assess effects of longer-term dietary intake on breast density. Data were incomplete for the number of live births and for age at first live birth; but despite their significant associations with breast density, these two variables were not important confounders among the 120 women not missing this information.

Our findings may be biased if a woman's knowledge of her breast density, or some characteristic related to her breast density, influenced her reporting of her dietary intake, or changed her intake so that her current report was not reflective of her long-term diet. For example, women with higher breast density may have been more likely to perceive themselves at higher risk for breast cancer and to modify their diets accordingly. We saw no differences, however, in either breast density or dietary intake by perception of risk for breast cancer reported at baseline (results not shown).

Because of the large number of dietary factors explored in this analysis, our findings should be interpreted with caution, warrant further investigation into the etiologic mechanism and warrant confirmation in larger hypothesis-based and pathway-based studies.

Finally, our findings may be limited in their ability to be generalized to other populations. Participants were drawn from participants in a high-risk program, and, even among these women, those with a lesser family history of breast cancer were underrepresented in our sample. Nevertheless, the availability of detailed pedigrees is a significant strength of the study as it allowed us to classify women on the basis of their family history of breast cancer in more detail than has been conducted previously, and hence to distinguish women at the highest genetic risk for breast cancer. Our study offers suggestive findings that should be confirmed in a larger sample representing a range of genetic and familial risk for breast cancer, in order to clarify the extent to which diet differences might be related to breast density, and whether effects differ by genetic susceptibility. Future studies might extend these findings to see whether changes in dietary intake would be associated with a change in breast density and a subsequent change in breast cancer risk.

In summary, in the present broad exploration of dietary factors in women enrolled in a high-risk program for breast cancer, we found that only vitamin D intake was inversely associated with breast density. Protein and animal protein intake were positively associated with breast density, but only among women without a hereditary family history of breast cancer. Our latter, tentative finding suggests that some dietary modification strategies may be more effective in reducing breast density and hence breast cancer risk in women with a family history of breast cancer suggestive of lower-penetrance susceptibility genes, but this requires confirmation in a larger sample representing a range of genetic risks for breast cancer.

FRAP = Family Risk Assessment Program.

The authors declare that they have no competing interests.

MT initiated the research project, conducted the statistical analyses, and drafted the manuscript. CB provided input on subsequent revisions of the manuscript. KAE conducted the qualitative breast density assessments. MBD is the Principal Investigator for the FRAP, contributed to the design of this study, and provided input on the manuscript. All authors read and approved the final manuscript.