Dairy, Calcium, and Vitamin D Intake and Postmenopausal Breast Cancer Risk in the Cancer Prevention Study II Nutrition Cohort

Dairy products have been hypothesized to have opposing effects on breast cancer risk because they contain both potentially protective and deleterious compounds. Dairy foods are the major source of calcium and vitamin D, metabolically interrelated and highly correlated dietary factors that may influence breast carcinogenesis through a variety of mechanisms. Both nutrients have direct effects on cell proliferation and differentiation of several cancer cell lines in vitro (1, 2). In a rodent model, calcium reduces fat-induced cell proliferation by maintaining intracellular calcium concentrations (3). Vitamin D acts through the vitamin D receptor, a nuclear transcription regulating factor that signals the synthesis of proteins involved in cell cycle regulation (1, 4). Clinically, breast tumors with greater vitamin D receptor expression correlate positively with time between first diagnosis and relapse (5). The vitamin D receptor ligand 1,25,dihydroxyvitamin D [1,25(OH)₂D] can be directly synthesized in breast tissue (6) from 25-hydroxyvitamin D (7). 1,25(OH)₂D was shown to preferentially induce differentiation in estrogen receptor–positive (ER+) tumors (5) and down-regulate the ER (8, 9). Other potential anticarcinogenic mechanisms of vitamin D compounds and the vitamin D receptor include blocking the mitogenic activity of insulin and insulin-like growth factor I (10) and induction of BRCA1 gene expression by transcriptional activation (11). Dairy foods are also rich in conjugated linoleic acid, a fatty acid with anticarcinogenic effects in experimental studies (12).

Milk also contains several hormones and growth factors (13, 14), including insulin-like growth factor I, a potent mitogen proposed to increase cancer development in the breast, colon, and prostate. In cross-sectional studies of adult men (15) and women (16), dairy products were positively associated with higher circulating insulin-like growth factor I. Thus, the net effect of dairy product consumption on breast cancer risk may reflect a balance of beneficial and detrimental factors.

Over 36 individual case-control and 10 prospective cohort studies have evaluated the relationship of dairy products and breast cancer with inconsistent findings (summarized in ref. 17). A recent pooled analysis of eight North American and Western European cohort studies on meat and dairy consumption found no association with dairy fluids or solids in postmenopausal breast cancer (18). Fewer studies have specifically evaluated dietary calcium and vitamin D in relation with breast cancer risk (19-28), particularly among postmenopausal women (21, 29). Furthermore, no studies have evaluated these dietary factors in relation to ER status.

We examined whether recent intake of dairy products, calcium, or vitamin D was related to the risk of postmenopausal breast cancer among over 68,000 women from the Cancer Prevention Study (CPS) II Nutrition Cohort. In a secondary analysis, we assessed whether these relationships were modified by ER status of the tumor.

Women in this study were participants in the CPS II Nutrition Cohort, a prospective study of cancer incidence and mortality among 86,404 men and 97,786 women (30). The Nutrition Cohort was begun by the American Cancer Society in 1992 and is a subgroup of the ∼1.2 million participants in CPS II, a prospective study of cancer mortality established in 1982. In brief, Nutrition Cohort participants were recruited from members of the CPS II cohort who resided in 21 states with population-based cancer registries as described in detail elsewhere (30). Participants were 50 to 74 years of age at enrollment in 1992 or 1993, when they completed a 10-page confidential, self-administered mailed questionnaire that included questions on demographic, medical, lifestyle, and dietary factors. Follow-up questionnaires were sent to cohort members in 1997, 1999, and 2001 to update exposure information and to ascertain newly diagnosed cancers. The response rate among living cohort members was 91% through 2001. The Emory University School of Medicine Institutional Review Board has approved all aspects of the CPS II Nutrition Cohort.

Incident cancer was defined as a new diagnosis of either nonfatal or fatal breast cancer that occurred after enrollment in 1992 or 1993 and before August 31, 2001. We included all interval reported breast cancers because previous pilot work in this cohort found that the sensitivity of self-reported breast cancer was 91% (31). Ascertainment of breast cancer as a contributing cause of death among cohort members was accomplished through linkage with the National Death Index (32). We documented ER and progesterone receptor (PR) status of the tumors through medical record reports and through state cancer registries when available. ER status was verified in 53% of the cases: 1,283 tumors were ER+ and 227 were ER−.

Calcium, vitamin D, and dairy product intakes were measured in 1992 to 1993 using a semiquantitative 68-item food frequency questionnaire (FFQ), which is a modification of the brief Health Habits and History Questionnaire developed by Block et al. (30, 33). The FFQ inquired about portion size (small, medium, or large). Questions on frequency of consumption ranged from “never or less than once per month” to “two or more per day” for foods and to “six or more per day” for beverages; use of low-fat foods, including ice cream and cheese, was also ascertained. The questionnaire included a section asking about frequency of use of several vitamin supplements during the past year (multivitamins and calcium among others). Individual vitamin D supplement use is not common (34) and was not included on the FFQ. Information on supplement dose was collected among participants reporting calcium supplement use (250, 500, 600, or 750 mg per tablet).

Dietary and total nutrient intakes and gram weights of foods were estimated using the Diet Analysis System version 3.8a (35). Total calcium estimates (mg/d) included contributions from diet, individual calcium supplements, and multivitamin pills (estimated at 130 mg per multivitamin pill). Vitamin D values were added to the nutrient database using U.S. Department of Agriculture sources (36). Total vitamin D (IU/d) included intake from diet and multivitamins (estimated to be 400 IU per pill). Nutrient estimates were adjusted for total energy using the residuals method (37). Gram weights for all dairy foods (whole, low-fat and skim milk, regular and low-fat cheese, yogurt, and ice cream) were converted to servings using the following weights: milk and yogurt, 240 g per serving (8 oz cup); cheese, 57 g per serving (2 oz); and ice cream, 72 g per serving (0.5 cup). We did not consider butter intake, which is not generally a recommended dairy food and is not a major contributor of calcium intake.

The FFQ was validated using four random 24-hour recalls collected over a 1-year period as the comparison measure among 441 Nutrition Cohort participants (38). Median energy-adjusted, attenuation-corrected Pearson validity correlations and 95% confidence intervals (95% CI) among women were high for dietary calcium (r = 0.66; 95% CI, 0.51-0.76) and for dairy products (r = 0.63; 95% CI, 0.45-0.77).

We excluded from this analysis women who were lost to follow-up between enrollment in 1992/1993 and August 31, 2001 (n = 3,514), or, who, at baseline, reported a history of breast cancer (n = 6,220) or other cancer (except nonmelanoma skin cancer; n = 5,805). Because we had too few women to examine menopausal status separately, we excluded women who were premenopausal or perimenopausal at baseline (n = 4,448) or with unknown menopausal status (n = 233). We also excluded women who reported extreme values of daily energy intake (<500 or >3,500 kcal/d), or FFQs with ≥15% of the 68 questions blank (n = 7,345). Women with missing or uninterpretable responses for more than four dairy products (n = 85), calcium supplements (n = 952), or multivitamin use (n = 543), or women who reported both greater than six servings of milk per day and identical responses for two types of milk (n = 42) were also excluded. Finally, women were excluded if diagnosis date was unknown (n = 32). The final analytic cohort consisted of 68,567 women.

We used Cox proportional hazards models to estimate the total cancer incidence rate ratio (RR) and 95% CI in relation with specific dietary factors (39). Age adjustment stratified on single year of age within each Cox model. Covariates included family history of breast cancer in a mother or sister (yes/no); history of breast cyst (yes/no); height (<63, 63 to <65, 65 to <67, ≥67 in, missing height), race (White versus non-White); weight gain since age 18 (−5-5, 6-20, 21-40, 41-60, 60+, missing or lost >5 lbs); education (<high school, high school, some college, college+, missing); age at menarche (<12, 12, 13, ≥14, unknown), age at menopause (<45, 45 to <50, 50 to <54, ≥54, unknown), age at first birth and number of live births (nulliparous and combinations of 1-2, 3+ live births and age <25, ≥25 or unknown number of live births or age at first birth); alcohol consumption (none, <1 drink/d, 1, or ≥1 drink/d, unknown); hormone replacement therapy [HRT; never, current estrogen replacement therapy, past estrogen replacement therapy, current combined replacement therapy (estrogen/progesterone), past combined replacement therapy, unknown/missing]; screening mammography (never, within the last year, not within the last year, unknown); and energy (in quintiles). Screening mammography was modeled as a time-varying covariate using information obtained in 1992, 1997, and 1999. We also examined energy-adjusted total fat as a potential confounder in the analysis. In situ breast cancers were included in the main analysis as cases; we also conducted sensitivity analyses excluding in situ tumors and those without information on tumor staging.

Potential effect modification of findings by potential for skin synthesis of vitamin D was examined by stratifying on UV index by state of residence for the subgroup of women (95.5%) who reported the same state of residence in 1982 and 2001. UV index levels were estimated using the 1997 monthly mean UV indices for major forecast cities in 50 states (40). We also stratified by ER status of the tumor, multivitamin use (current versus not current use), dietary fat intake (tertiles), HRT (current, past, never use), body mass index (<25, 25-30, >30), and family history (yes/no). The likelihood ratio test was used to test for significance in interaction models. Two-sided tests were statistically significant if P ≤ 0.05.

Dairy products and nutrients derived largely from dairy products (e.g., dietary calcium and dietary vitamin D) were highly correlated in the analytic cohort as expected (Table 1). Total calcium and total vitamin D included supplemental nutrients and, therefore, were not as strongly correlated with other dietary variables. The largest contributors to dietary calcium included skim milk (38%), low-fat milk (32%), regular and low-fat cheese (15%), and regular and low-fat yogurt (7%). Top contributors to dietary vitamin D were similar, as follows: skim milk (42%), low-fat milk (38%), whole milk (5%), fish (5%), cereal (4%), and eggs (3%). Similarly, total dairy products were composed of skim milk (42%), low-fat milk (22%), whole milk (4%), regular and low-fat yogurt (11%), cheese and low-fat cheese (12%), and ice cream and low-fat ice cream (10%). Approximately 80% of low-fat dairy products consumed were from skim milk and low-fat milk, and the majority of milk consumed was in low fat or skim form (95%).

Table 2 provides a description of baseline characteristics and breast cancer risk factors according to intake of low-fat dairy products, dietary calcium, and total calcium, which were determined using predefined cut points. Ranges of intake were ∼3- to 6-fold from the lowest to highest intake levels. Women in the highest categories of each exposure smoked less and were more physically active. Use of HRT and calcium and multivitamin supplement was more common among women in the highest exposure groups; differences were especially pronounced among women in extreme categories of total calcium intake. Women with higher total calcium intakes also reported a somewhat greater history of breast cysts, were leaner, and gained less weight since age 18 compared with women with low total calcium intake. Reproductive history did not vary by dietary exposure category. Total fat and alcohol intakes were lower with higher exposure levels, and dairy nutrients were tracked in the expected direction.

Analyses are based on 2,855 cases of postmenopausal breast cancer identified over 536,815 person-years of follow-up from 1992 to 2001. The results with respect to nutrients are shown in Table 3. Women who consumed the most dietary calcium (>1,250 mg/d) had a 20% lower risk of postmenopausal breast cancer compared with those in the lowest category of intake (<500 mg/d; RR, 0.80; 95% CI, 0.67-0.95; P_trend = 0.02). Supplemental calcium (including both calcium supplements and multivitamins) was not associated with risk of breast cancer, with or without inclusion of dietary calcium in the model, whereas calcium supplements alone (additionally controlling for multivitamins) were marginally associated with reduced risk (RR, 0.90; 95% CI, 0.82-0.99, for ≥500 mg versus no use; not shown). Likewise, total calcium (diet plus all supplements) did not strengthen the association (for >1,750 mg total calcium/d; RR, 0.91; 95% CI, 0.79-1.06; P_trend = 0.07). However, there was a nonlinear lower risk at levels between 1,250 and 1,750 mg total calcium/d. Neither dietary nor total vitamin D intake (diet plus multivitamin sources of vitamin D) was significantly related to risk of breast cancer.

In analyses by foods (Table 4), consumption of at least two servings of dairy products daily (versus ≤0.5 servings/d) was inversely associated with breast cancer risk (RR, 0.81; 95% CI, 0.69-0.95; P_trend = 0.002). When we examined low-fat dairy products separately from high-fat dairy, associations were slightly stronger for low-fat dairy (controlled for other types). Total milk consumption (3 servings/d versus <1/2 cup/d) was nonsignificantly, inversely related to risk. When milk types were examined separately (compared with non–milk drinkers), consumption of low-fat and skim milk ≥3 servings/d (combined as “low fat”) was marginally associated with lower risk. In these analyses, we considered individuals who reported cereal use but no milk consumption as “non–milk drinkers.” In analyses including these individuals as milk drinkers (with reference group including very small intake levels), results were not materially different.

In analyses that stratified by ER status of the tumor (Table 5), women in the highest dietary calcium category had 33% lower risk of ER+ tumors compared with women in the lowest category (RR, 0.67; 95% CI, 0.51-0.88; P_trend = 0.004). Dietary vitamin D was also significantly related to lower risk of ER+ (RR, 0.74; 95% CI, 0.59-0.93; P_trend = 0.006) but not ER− tumors (RR, 1.03; 95% CI, 0.61-1.73; P_trend = 0.84). Consumption of ≥2 servings/d of dairy products was associated with 29% lower risk of ER+ tumors than consumption of <0.5 cups/d (RR, 0.71; 95% CI, 0.56-0.91; P_trend <0.0003). Results were similar when we examined ER/PR status combined (e.g., ER+/PR+ was similar to ER+ alone and ER−/PR− was similar to ER− alone; there were too few cases to examine other combinations).

There was no significant interaction between the dietary exposures and multivitamin use, body mass index, HRT, family history of breast cancer, or total fat intake. Sensitivity analyses excluding women with in situ tumors were not materially different. However, women from states with lower UV exposure (UV index ≤ median of 4.2) had a lower risk of breast cancer with greater dietary vitamin D consumption >300 IU/d (RR, 0.81; 95% CI, 0.67-0.97) compared with women from these states who consumed ≤100 IU/d. Conversely, women from states with UV index >4.3 did not have lower risk with vitamin D consumption >300 versus ≤100 IU/d (RR, 1.05; 95% CI, 0.82-1.35; P_interaction = 0.05). Similar findings were observed by UV index for >2 versus <0.5 servings/d of low-fat dairy (RR, 0.77; 95% CI, 0.66-0.90 within low UV index states and RR, 1.05; 95% CI, 0.85-1.29 within high UV index states, P_interaction = 0.03) and >2 versus <0.5 total dairy servings/d (RR, 0.76; 95% CI, 0.64-0.91 within low UV index states and RR, 0.93; 95% CI, 0.73-1.17 within high UV index states, P_interaction = 0.02).

In this large prospective U.S. cohort, high intake of dietary calcium and low-fat dairy products was associated with a moderately lower risk of developing postmenopausal breast cancer compared with women with lowest intake levels. Supplemental calcium or higher levels of total calcium (diet plus supplements) and vitamin D were not related to overall breast cancer risk. However, vitamin D, calcium, and dairy products were inversely associated with risk of ER+ breast tumors, but not ER− tumors.

Our overall findings suggest that dairy products, composed mainly of low-fat sources or some component within these foods, are associated with a small but significantly lower risk of breast cancer in postmenopausal women. Three prospective studies of low-fat dairy products or dietary calcium in relation with postmenopausal breast cancer risk had mixed results (18, 29, 41). In the Netherlands Cohort Study, consumption of skim milk and skim milk products was unrelated to postmenopausal breast cancer risk (RR, 1.04, for median intake of 203 g/d or 1 cup/d versus none; ref. 41). In our study, risk was not lower until intake reached ≥480 g/d (2 cups/d), implying that greater intakes may be necessary to lower risk. In a large pooled analysis of U.S. and European cohorts (18), dairy products were not related to breast cancer risk but milk fortification policies may be different in European countries. Shin et al. (29) observed a significant inverse association between low-fat dairy products, dietary calcium, and premenopausal breast cancer risk in the Nurses' Health Study using repeated FFQs to cumulatively update diet over a 16-year follow-up period. However, these dietary factors were not related to postmenopausal breast cancer, except in analyses limited to baseline exposure when dietary calcium was associated with a 15% statistically significantly lower risk (29).

The high correlations of nutrients in dairy products make it challenging to identify the single or combined variables that may contribute to lower breast cancer risk. Dietary calcium was more strongly inversely associated with risk than other factors we examined, but the hypothesized mechanisms are not as compelling as those for vitamin D. One hypothesis for calcium centers on modification of fat-stimulated proliferation (3) but dietary fat is not consistently related to breast cancer risk (42, 43) and was not a modifier in this analysis. If calcium is truly a causal factor, a possible explanation for the weaker association with supplemental calcium is that lactose present in dairy products enhances calcium absorption. Lactose has also been hypothesized to reduce breast cancer risk by blocking gonadotropin release, which would down-regulate estrogen synthesis (29). Use of calcium supplements was also more common among women taking HRT in our cohort; thus, residual confounding by HRT use among calcium supplement users may also account for the weaker associations with supplemental calcium. Low-fat dairy products were associated with a slightly lower risk of breast cancer compared with high-fat dairy products, suggesting that conjugated linoleic acid may not be the responsible factor, although, in this study, we were not able to measure total conjugated linoleic acid intake.

Although our findings for vitamin D were not as strong as for other dietary variables, the biological plausibility for a role for vitamin D in breast carcinogenesis is strong. Both normal and neoplastic breast tissue express the nuclear receptor for 1,25(OH)₂D. Several studies report tissue-specific expression of 1,α-hydroxylase and autocrine/paracrine synthesis of 1,25(OH)₂D (6, 44). Little is known about factors that may influence 1,25(OH)₂D synthesis and degradation in breast tissue but greater substrate availability should be important (7). Others have observed inverse associations between higher 25-hydroxyvitamin D levels or sun exposure and breast cancer risk (7, 27). Vitamin D intake may be more strongly associated with risk than we were able to ascertain had we assessed sun exposure or 25-hydroxyvitamin D blood levels. Among women with expected lower UV synthesis of vitamin D (states with lower UV index), we observed stronger inverse associations with dietary vitamin D compared with women who lived in states with high UV index.

In our study, inverse associations between dietary factors and postmenopausal breast cancer risk were somewhat stronger in women with ER+ tumors. This is consistent with recent studies showing that risk factors for breast cancer vary by ER and PR status (45). Previous studies suggest that ER+ breast cancer cell lines are more sensitive to 1,25(OH)₂D-mediated growth regulation compared with ER− cell lines (4, 5, 46). Although total and dietary vitamin D was not related to overall breast cancer risk in our study, vitamin D was inversely related to risk of ER+ tumors. This suggests a role of improved vitamin D status on reducing ER+ breast cancer, which needs confirmation in other studies.

Limitations of this study include the inability to quantify vitamin D status precisely because we did not inquire about sun exposure and were not able to measure plasma 25-hydroxyvitamin D levels directly. However, our dietary instrument was validated for calcium and dairy intakes (38) and we were able to examine findings by UV index for the 21 enrollment states of our participants. Our findings are generalizable to the adult diet of mostly White, middle class postmenopausal women. We were unable to determine whether dietary exposures earlier in life influence postmenopausal breast cancer risk. We were missing data on ER status for 47% of the population. Examination of demographic and risk factor characteristics among women with and without ER tumor status revealed no important differences. The number of ER− tumors in our sample was limited (n = 227). Thus, we cannot rule out that we missed associations in this subset due to lack of statistical power.

In summary, postmenopausal women who consumed greater intakes of dietary calcium and dairy products, primarily from low-fat sources, were at lower risk of breast cancer in our cohort. The lower risk was more pronounced in ER+ breast cancer. Supplemental sources of these nutrients were not as strongly related to risk. Future studies should clarify specific components responsible and the relation between these factors and ER status.