Intake of Dairy Products, Calcium, and Vitamin D and Risk of Breast Cancer

Abstract

Background: Laboratory data suggest that calcium and vitamin D, found at high levels in dairy products, might reduce breast carcinogenesis. However, epidemiologic studies regarding dairy products and breast cancer have yielded inconsistent results. We examined data from a large, long-term cohort study to evaluate whether high intake of dairy products, calcium, or vitamin D is associated with reduced risk of breast cancer. Methods: We followed 88 691 women in the Nurses' Health Study cohort from the date of return of their food-frequency questionnaire in 1980 until May 31, 1996. Dietary information was collected in 1980 and updated in 1984, 1986, 1990, and 1994. We identified 3482 women (premenopausal = 827, postmenopausal = 2345, and uncertain menopausal status = 310) with incident invasive breast cancer. We used pooled logistic regression to estimate multivariable relative risks (RRs) using 2-year time increments. The RRs and 95% confidence intervals (CIs) were calculated for each category of intake compared with the lowest intake group. All statistical tests were two-sided. Results: Intakes of dairy products, calcium, or vitamin D were not statistically significantly associated with breast cancer risk in postmenopausal women. In premenopausal women, however, consumption of dairy products, especially of low-fat dairy foods and skim/low-fat milk, was inversely associated with risk of breast cancer. The multivariable RRs comparing highest (>1 serving/day) and lowest (≤3 servings/month) intake categories were 0.68 (95% CI = 0.55 to 0.86) for low-fat dairy foods and 0.72 (95% CI = 0.56 to 0.91) for skim/low-fat milk. Dairy calcium (>800 mg/day versus ≤200 mg/day; RR = 0.69, 95% CI = 0.48 to 0.98), total vitamin D (>500 IU/day versus ≤150 IU/day; RR = 0.72, 95% CI = 0.55 to 0.94), and lactose (quintile 5 versus quintile 1; RR = 0.68, 95% CI = 0.54 to 0.86] also had inverse associations with premenopausal breast cancer risk. By taking into account supplemental calcium and vitamin D intake, we found that association with calcium was due mainly to dairy sources whereas the association with vitamin D may be independent of dairy intake. Conclusions: We found no association between intake of dairy products and breast cancer in postmenopausal women. Among premenopausal women, high intake of low-fat dairy foods, especially skim/low-fat milk, was associated with reduced risk of breast cancer. Similar inverse associations were seen with components (calcium and vitamin D) of dairy foods, but their independent associations with breast cancer are difficult to distinguish.

Milk and dairy products have been suspected to increase the risk of breast cancer. International comparative studies have shown a positive association between milk and breast cancer mortality (1,2). The association between milk and breast cancer, however, has been inconsistent in case–control studies, with reports of no association (3–5), inverse association (6–10), and positive association (11–15). Cohort studies of dairy foods and breast cancer have yielded similarly conflicting findings (16,17), with inverse associations between milk and breast cancer found in several recent studies (18–20). In some studies, different associations between milk and breast cancer were found, depending on the type of milk (7,17,21). In a 1993 meta-analysis, a small increase in breast cancer risk with higher milk intake was observed (22), but the more recent findings showing inverse associations were not part of that meta-analysis.

Components in milk that might be anticarcinogenic include calcium and vitamin D, and some studies have suggested protective effects of calcium and vitamin D against colon cancer (23,24). The hypothesized effect of calcium on colon cancer is intraluminal binding with bile acids and fatty acids, thus reducing the proliferative stimulus of these compounds. With breast cancer, calcium has been proposed to reduce fat-induced cell proliferation by maintaining intracellular calcium concentrations (25,26). Vitamin D modulates calcium metabolism (27) and has calcium-independent antiproliferative actions (26). Some suggest that only vitamin D, not calcium, inhibits mammary tumorigenesis (28), but an independent anticancer effect of higher calcium intake in rats has also been reported (29).

Epidemiologic evidence relating calcium and vitamin D intakes to breast cancer risk is limited. Breast cancer rates in white women are highest in areas with the least winter sunlight and longest winters, which has been interpreted to support an association with vitamin D (30), although we did not find this association in the Nurses' Health Study (NHS) (31). A few epidemiologic studies have reported a statistically significant inverse association between calcium intake and breast cancer (19,32–34), while others reported no association (35,36). However, intake of calcium and dairy foods are strongly correlated, so the observed associations with calcium are difficult to separate from those with milk, dairy products and, consequently, other components of milk and dairy products. To our knowledge, there is no epidemiologic study of dietary vitamin D and breast cancer.

We examined data from a large, long-term cohort study to evaluate the hypotheses that higher intakes of dairy products, calcium, or vitamin D are associated with reduced risk of breast cancer. By taking into account supplemental intake of calcium and vitamin D, we hoped to dissociate the effects of calcium and vitamin D from those of milk and dairy products.

Methods

The NHS cohort was established in 1976, when 121 700 registered nurses from 11 states in the United States answered a mailed questionnaire on risk factors for cancer and cardiovascular disease. This investigation was approved by the Human Research Committee of the Brigham and Women's Hospital. Every 2 years, follow-up questionnaires have been sent to update information on potential risk factors. In 1980, a 61-item food-frequency questionnaire designed to assess dietary intake was added. In 1984, 1986, 1990, and 1994, an expanded (approximately 130 foods) food-frequency questionnaire was used. The participants included in this analysis were the 88 691 women who completed the 1980 dietary questionnaire, did not have implausible values for total energy intake (<500 kcal or >3500 kcal), had not left more than 10 food items blank, and had not been diagnosed with cancer by 1980.

Semiquantitative Food-Frequency Questionnaires and Calculation of Nutrient Intake

The food-frequency questionnaires have been described in detail, and their validity and reproducibility have been documented previously (37,38). A commonly used portion size was specified for each food (e.g., one 8-oz glass of milk). Participants were asked how often, on average, they had consumed that amount of each food over the past year. The nine prespecified frequency categories ranged from “never” to “6 or more times per day.”

The dairy food group included skim/low-fat milk, whole milk, cream, sour cream, sherbet, ice cream, yogurt, cottage cheese, cream cheese, other (hard) cheese, and butter. Total dairy food intake was calculated by summing up the daily servings of all the foods in the dairy group, except butter because it was composed almost entirely of fat and was therefore unlike the other dairy foods. When we calculated dairy fat intake, however, we took into account all the foods in the dairy group, as well as dairy ingredients from other foods. Low-fat dairy food intake was calculated by summing the daily servings of skim/low-fat milk, sherbet, yogurt, and cottage cheese. High-fat dairy food intake was calculated by summing the daily servings of whole milk, cream, sour cream, ice cream, cream cheese, other cheese, and butter. Total milk intake was calculated by summing the daily servings of skim/low-fat milk and whole milk. Total fermented milk intake was calculated by summing the daily intakes of sour cream, yogurt, cottage cheese, cream cheese, and other cheese.

Nutrient intakes were computed by multiplying the frequency of consumption of each unit of food and the nutrient content of the specified portions. Nutrient values in foods were obtained from U.S. Department of Agriculture sources (39). Nutrient intakes were energy-adjusted by using the residuals from the regression of nutrient intake on total caloric intake (40). Residuals were adjusted to 1600 kcal/day, the approximate median caloric intake of all the participants with acceptable diet data, to have meaningful nutrient values.

The information on current use, brand, and dosage of multivitamin supplements was collected on each biennial questionnaire. Current use and dosage of individual calcium and vitamin D supplements were also collected biennially from the 1982 questionnaire.

Total calcium and total vitamin D intakes were calculated from all sources of calcium and vitamin D, combining diet and supplements. Calcium supplement users were defined as users of either individual calcium supplements or multivitamins containing calcium. Vitamin D supplement users were defined as users of either individual vitamin D supplements (tablets) or multivitamins containing vitamin D. We took into account the multivitamin brand and type information supplied by the cohort participants when calculating the supplement intake amount. Dietary calcium and dietary vitamin D intakes were calculated from all dietary sources without supplements. Dietary calcium was further divided according to the food sources, i.e., calcium from dairy sources (dairy calcium) and calcium from nondairy sources (nondairy calcium). Dairy calcium intake was calculated from all dairy foods as well as dairy ingredients from recipes of other foods, for example, cheese from pizza and milk from mashed potatoes. Nondairy calcium was calculated by subtracting dairy calcium from dietary calcium intake. Total nondairy calcium was the combination of nondairy calcium and supplemental intake. We did not divide dietary vitamin D further because the sources are limited to relatively few foods. Calcium supplement users (18.7% of premenopausal and 35% of postmenopausal person-years) were excluded in the primary analyses of dietary calcium and dairy calcium; vitamin D supplement users (31% of premenopausal and 33% of postmenopausal person-years) were excluded in the primary analysis of dietary vitamin D because the supplements could obscure the effects of diet. We also conducted alternative analyses that included the supplement users.

Among 173 members of the NHS cohort, Pearson correlation coefficients (corrected for within-person day-to-day variation) between the means of four 1-week diet records and the 1980 food-frequency questionnaire were 0.69 for skim/low-fat milk and 0.56 for whole milk (38). Correlation coefficients between intakes of calcium from the questionnaire used in 1984 and intakes from four 1-week diet records measured in 1980 among 150 Boston-area women were 0.56 with supplements and 0.51 without supplements (37). Correlation coefficients between intake and plasma concentration of vitamin D in 57 men and 82 women were 0.35 with supplements and 0.25 without supplements (41).

Identification of Breast Cancer Cases

In each biennial questionnaire, participants were asked whether they had been diagnosed as having breast cancer in the previous 2 years. Deaths were identified by a report from a family member, the postal service, or the National Death Index. Medical records were obtained for breast cancer cases identified by either self-report or vital records, and more than 99% of these records confirmed the self-report. Cases of carcinoma in situ were excluded.

Statistical Analysis

Each participant accumulated person-time beginning with the return of the 1980 questionnaire and ending with her cancer diagnosis, death, or May 31, 1996, whichever came first. To avoid recall bias, we started the follow-up of each variable from the time it was collected, even when it was related to previous experience. For example, high school dietary information was collected in 1986, and its analysis included person-time from 1986.

To reduce within-person variation and best represent long-term dietary intake of participants, we assessed the incidence of breast cancer in relation to the cumulative average of dietary intake from all available dietary questionnaires up to the start of each 2-year follow-up interval (42). For example, the incidence of breast cancer from 1980 through 1984 was related to the dietary information from the 1980 questionnaire, and the incidence of breast cancer from 1984 through 1986 was related to the average intake from the 1980 and 1984 questionnaires. When dietary information was missing at the start of the interval, the cumulative average from the previous period was carried forward. To take into account a longer latency of exposure, the 1980 baseline diet also was related to the breast cancer incidence during the entire follow-up period. To consider intermediate latency, we used a 4-year time lag between diet assessment and the start of a follow-up interval. We also analyzed premenopausal diet, that is, the cumulative average diet until menopause, to assess whether postmenopausal breast cancer risk was related to diet in the premenopausal period. A subcohort of premenopausal women at baseline was used for this analysis, and their follow-up was started when they became postmenopausal. To assess the influence of consistency in dairy food intake, we conducted restricted analyses using data from women who had not greatly changed their milk intake during the 10 years prior to baseline in 1980.

All dietary variables were categorized according to quintiles based on their distribution in each questionnaire cycle. Consequently, the cutoff points of the quintiles slightly varied by questionnaire cycles. Categories for dairy foods and nutrients based on their absolute intakes were also created to have common categories throughout the whole follow-up cycle. The relative risks (RRs) and 95% confidence intervals (CIs) were calculated for each category of intake compared with the lowest intake group. We used pooled logistic regression to estimate multivariable RRs using 2-year time increments (43). We simultaneously adjusted for age in 5-year categories, time period, physical activity in metabolic equivalent-hours (METs, with activity at rest = 1.0) values, history of benign breast disease, family history of breast cancer, height, weight change since age 18 years, body mass index (BMI; kg/m²) at age 18 years, age at menarche, parity, age at first birth, alcohol intake, total energy intake, total fat intake, glycemic index value (= [glycemic index × carbohydrate intake for each food]/total carbohydrate intake), β-carotene intake, and total vitamin E intake. For postmenopausal women, we added age at menopause and postmenopausal hormone use to the multivariable model. For vitamin D analyses, we added history of outdoor sun exposure and participant's residential area. These covariates were either known or suspected risk factors for breast cancer or had been found to be associated with dairy/calcium/vitamin D intake and with breast cancer risk in the NHS data. RRs were adjusted for glycemic index value because blood glucose level could affect the active cellular uptake of calcium. We calculated two-sided P values for the test for trend by using the Wald statistic. Median values of each intake category were used in the test for trend. Interactions were evaluated by using the likelihood ratio test.

All nondietary covariates included in the multivariable analyses were updated every 2 years, except age at menarche, height, weight at 18 years old, and outdoor sun exposure (yes, if time spent outdoors was ≥8 hours/day in summer), which were determined at baseline. Activity in METs values was calculated in 1980, 1986, 1988, 1992, and 1994; information on family history of breast cancer (mother, yes/no, or sister, yes/no) was sought in 1976, 1982, 1988, and 1992; and site of residence was recorded in 1976 and updated in 1986, 1988, 1990, 1992, and 1994 and, at each time, was recoded into four regions that included a total of 11 states (California, Northeast, Midwest, and South). Participants who moved out of the 11 original states during the follow-up were categorized as living in an ‘other’ region. A woman was classified as postmenopausal from the time she returned a questionnaire reporting natural menopause or hysterectomy with bilateral oophorectomy. Women who reported hysterectomy without bilateral oophorectomy were classified as being of uncertain menopausal status until they reached the age at which natural menopause had occurred in 90% of the cohort (54 years for current cigarette smokers and 56 years for nonsmokers), at which time they were classified as postmenopausal. Because we observed differences in the associations of major study variables with breast cancer risk between pre- and postmenopausal women, we report our results from pre- and postmenopausal women separately. Women with uncertain menopausal status were excluded from the analysis.

Results

During 16 years of follow-up, 827 premenopausal women and 2345 postmenopausal women of the 88 691 women in the cohort were diagnosed with breast cancer. An additional 310 women who developed breast cancer were of uncertain menopausal status and were excluded from analyses.

At baseline, most of the known breast cancer risk factors did not vary appreciably across categories of total milk, total calcium, and total vitamin D intake (Table 1). Calcium, vitamin D, and multivitamin supplement users were more frequent in the high total calcium and vitamin D intake groups. Postmenopausal hormone use and history of osteoporosis were not strongly associated with total milk intake or total calcium intake. Women who consumed more total milk, total calcium, or total vitamin D exercised more, smoked less, and drank less alcohol.

Milk and Dairy Products

In postmenopausal women, none of the dairy products had any appreciable association with breast cancer risk (Table 2). However, in premenopausal women, a statistically significant inverse association between low-fat dairy food intake and breast cancer risk was observed (Table 2). The multivariable RR comparing the highest with the lowest categories of consumption was 0.68 (95% CI = 0.55 to 0.86; P_trend = .003). Skim/low-fat milk intake accounted for 49% of low-fat dairy food intake in the 1984 data and was the dairy food most strongly related to breast cancer risk (>1 serving/day versus never; RR = 0.72, 95% CI = 0.56 to 0.91; P_trend = .007). High-fat dairy food and whole milk, which accounted for 12% of high-fat dairy food intake in 1984, showed statistically nonsignificant inverse associations with premenopausal breast cancer risk (for high-fat dairy food, >2.5 servings/day versus ≤3 servings/week; RR = 0.83, 95% CI = 0.62 to 1.10). When we included the intakes of skim/low-fat milk and whole milk simultaneously in the same model, the RRs were 0.68 (95% CI = 0.53 to 0.88) for skim/low-fat milk and 0.71 (95% CI = 0.49 to 1.05) for whole milk. Total milk intake showed a statistically significant linear inverse association with premenopausal breast cancer (>1 serving/day versus ≤3 servings/month; RR = 0.69, 95% CI = 0.54 to 0.87; P_trend<.001).

High milk consumption during high school also had a statistically nonsignificant association with lower risk of premenopausal breast cancer (RR = 0.76, 95% CI = 0.48 to 1.21) (Table 2). Most of the milk consumed in high school was consumed as whole milk (among total milk consumers in high school, 87% consumed whole milk and 11% consumed skim milk). Yogurt, other cheese, and total fermented dairy foods did not have any association with risk. Additional adjustment for total calcium or total vitamin D slightly attenuated the associations of dairy foods with premenopausal breast cancer.

There was no qualitative difference between age-adjusted and multivariable RRs. For example, the age-adjusted RR for highest versus lowest intake of low-fat dairy food was 0.77 (95% CI = 0.63 to 0.95) and that for skim/low-fat milk was 0.82 (95% CI = 0.66 to 1.01). Analyses using diet at baseline (1980), diet with a 4-year time lag, and diet among consistent milk consumers yielded RRs that were qualitatively similar but slightly attenuated compared with those in Table 2. Also, premenopausal intakes of dairy foods were not associated with postmenopausal breast cancer risk.

Calcium and Vitamin D

Calcium and vitamin D intakes were not associated with postmenopausal breast cancer risk (Table 3). Dairy and total nondairy calcium also had no association with risk in postmenopausal women. In premenopausal women, most of the calcium- and vitamin D-related variables were inversely associated with breast cancer incidence (Table 3). The multivariable RR for highest versus lowest total calcium intake was 0.80 (95% CI = 0.58 to 1.12; P_trend = .05) and that for dietary calcium intake was 0.67 (95% CI = 0.49 to 0.92; P_trend = .02). When we further divided dietary calcium into dairy and nondairy calcium, dairy calcium was inversely associated with risk (>800 mg/day versus ≤200 mg/day; RR = 0.69, 95% CI = 0.48 to 0.98; P_trend = .01); nondairy dietary calcium had no association, although the range in intake was much smaller (>350 mg/day versus ≤275 mg/day; RR = 1.12, 95% CI = 0.78 to 1.60). Total nondairy calcium also had no appreciable association with risk. Simultaneous inclusion of dairy and nondairy calcium in the model did not change the results appreciably.

Total vitamin D intake was associated with lower risk of premenopausal breast cancer (highest versus lowest intake; RR = 0.72, 95% CI = 0.55 to 0.94; P_trend = .01), while dietary vitamin D intake was associated with a similar but marginally statistically significant risk reduction (highest versus lowest intake; RR = 0.66, 95% CI = 0.43 to 1.00. Adjusting for residential area and history of outdoor sun exposure did not affect the associations of total and dietary vitamin D with premenopausal breast cancer, nor were these variables themselves associated with breast cancer incidence.

No qualitative difference was observed between age-adjusted and multivariable RRs for the associations of vitamin D and calcium with premenopausal breast cancer risk. The age-adjusted RR for highest versus lowest intake of dietary calcium was 0.74 (95% CI = 0.55 to 0.98) and that for dietary vitamin D was 0.78 (95% CI = 0.52 to 1.18). When we used 1980 baseline intakes of calcium from various sources in the analyses, inverse associations with premenopausal breast cancer were still seen, but the magnitude and statistical significance of trends were attenuated compared with the use of cumulative average intakes (Table 3). Baseline total and dietary vitamin D intakes were not associated with premenopausal breast cancer risk. Similar attenuation was observed when a 4-year time lag was applied between calcium and vitamin D intakes and premenopausal breast cancer. The RR for dairy calcium (>800 mg/day) was 0.78 (95% CI = 0.57 to 1.07; P_trend = .06). When we included supplement users in the analyses, the associations of calcium and vitamin D intakes with premenopausal breast cancer were slightly attenuated but were not changed substantially. Again, the premenopausal intakes of calcium and vitamin D were not associated with postmenopausal breast cancer risk.

Supplemental calcium had no apparent linear association with breast cancer risk in either premenopausal or postmenopausal women (Tables 4 and 5). In the stratified analysis by tertiles of dietary calcium among premenopausal women, women who used high-dose calcium supplements (≥900 mg/day) had a lower risk only if they were in the high dietary calcium intake group (RR = 0.44, 95% CI = 0.19 to 1.01) although the number of case subjects in this analysis was very small. Supplemental vitamin D had a weak, statistically nonsignificant inverse association with premenopausal breast cancer risk, and this association was more prominent among women in the low dietary vitamin D intake group (for ≥400 IU/day; RR = 0.80, 95% CI = 0.57 to 1.14). Most of the supplemental vitamin D use was determined by multivitamin use. Postmenopausal hormone use did not modify the associations between calcium or vitamin D supplement intake and breast cancer.

Other Constituents of Dairy Foods and Their Correlations With Dairy Food Intake

The intake of other constituents of dairy foods had associations similar to those with dairy calcium. Lactose and phosphorus intakes, which were highly correlated with intake of low-fat dairy foods (r = 0.75 and 0.71, respectively), were inversely associated with breast cancer risk in premenopausal women (quintile 5 versus quintile 1; RR = 0.68, 95% CI = 0.54 to 0.86; P_trend<.001 for lactose and RR = 0.73, 95% CI = 0.56 to 0.96; P_trend = .01 for phosphorus). Dairy fat intake was associated with lower risk in the highest quintile relative to the lowest, but the trend was not statistically significant (quintile 5 versus quintile 1; RR = 0.78, 95% CI = 0.62 to 0.98; P_trend = .13).

These inverse associations between the constituents of dairy food intake and premenopausal breast cancer disappeared rapidly after menopause. In early menopause, i.e., within 10 years after menopause, the RRs for dairy calcium, lactose, phosphorus, and dairy fat were already attenuated and statistically nonsignificant. In late menopause, i.e., beyond 10 years after menopause, the RRs were virtually null. For example, for dairy calcium, the RR for the highest versus lowest quintiles was 0.67 (95% CI = 0.52 to 0.88) in premenopausal women, 0.85 (95% CI = 0.65 to 1.11) during early menopause, and 1.05 (95% CI = 0.83 to 1.34) in late menopause. Also, the associations between the constituents of dairy foods with premenopausal breast cancer were more pronounced in women of even younger ages. For example, for dairy calcium (highest versus lowest quintile), the RR of premenopausal breast cancer occurring 6 years or more before menopause was 0.66 (95% CI = 0.38 to 1.12), and within 5 years of menopause, the RR was 0.86 (95% CI = 0.68 to 1.09).

The inverse association between low-fat dairy food intake and premenopausal breast cancer risk was attenuated when it was further adjusted for lactose, dairy calcium, total vitamin D, and phosphorus intakes, but only the adjustment for lactose was strong enough to make the association not statistically significant (highest versus lowest intake of low-fat dairy products; RR = 0.81, 95% CI = 0.61 to 1.08).

Because vitamin D modulates calcium metabolism, effects of dairy calcium might be modified by vitamin D intake. When we stratified the data by tertiles of total vitamin D intake, dairy calcium appeared to be associated with reduced risk of premenopausal breast cancer in women at all levels of vitamin D intake (P_interaction = .85). According to the calcium–high fat hypothesis (25,26), the effect of calcium should be stronger in the high-fat-diet group. However, we observed an inverse association between dairy calcium intake and premenopausal breast cancer in all tertiles of total fat intake. Similarly, the association between total vitamin D intake and premenopausal breast cancer was not modified by total fat intake. The inverse association between dairy calcium and breast cancer also was not modified by height, current BMI, or alcohol intake.

Discussion

In this large prospective cohort study, intakes of dairy foods, calcium, or vitamin D were not associated with the risk of breast cancer among postmenopausal women. Among premenopausal women, however, most of the dairy-related variables (i.e., total dairy, low-fat dairy, total milk, skim/low-fat milk, dietary and dairy calcium, total vitamin D, lactose, total phosphorus, and dairy fat) were inversely associated with breast cancer. The associations were mostly related to the low-fat dairy variables (e.g., intake of skim/low fat milk) but it is unlikely that these variables were merely acting as markers for a low-fat diet in general, because total fat itself was not associated with breast cancer in premenopausal women. The weaker association for whole milk intake with premenopausal cancer might be due to fewer numbers of women who drank sufficiently large amounts. Calcium intake and total dairy product intake have also been associated with reduced mortality among breast cancer patients in this cohort (44).

We hypothesized that if milk intake is associated with reduced risk of breast cancer, calcium or vitamin D might be the responsible nutrient. In the present study, intake of calcium and vitamin D indeed had statistically significant inverse associations with premenopausal breast cancer. Detailed analyses for calcium, however, suggested that the inverse association with calcium was due mostly to dairy sources rather than to nondairy sources or supplements. Low-fat dairy foods remained associated with premenopausal breast cancer risk when total calcium was included in the same model, whereas total calcium was no longer associated with risk (RR for >1250 mg/day versus ≤500 mg/day of total calcium = 0.95, 95% CI = 0.66 to 1.35). This suggests that dairy calcium is not likely to be responsible for the association of low-fat dairy food and skim/low-fat milk intake with reduced risk of premenopausal breast cancer.

Vitamin D intake, by contrast, was most strongly related to risk when it was analyzed as total intake rather than as dietary or supplemental intake. Unlike calcium, vitamin D intake was largely accounted for by multivitamin supplementation (35% of total intake; 1984 data). Summation of dietary and supplement intake increases the range of intake and, hence, the power to detect an association. Although vitamin D did not contribute substantially to the association between low-fat dairy foods and premenopausal breast cancer risk when both terms were included in a multivariable model, the inverse association with vitamin D was sustained (RR for >500 IU/day versus ≤150 IU/day of total vitamin D = 0.78, 95% CI = 0.59 to 1.02). Therefore, vitamin D intake appears to offer a possible protective association apart from the “milk effect.” Residual confounding by the level of sun exposure could still be possible, because our measures of sun exposure (i.e., outdoor sun exposure and site of residence) were rather crude.

Knekt et al. (19) reported a strong inverse association between milk intake and subsequent incidence of breast cancer in a Finnish cohort study. They also found statistically significant inverse associations of breast cancer with calcium and lactose intakes. Their study was based on only 88 case subjects, but the observed decrease in risk among the highest tertile of total milk intake compared with the lowest was quite strong (RR = 0.42, 95% CI = 0.24 to 0.74). The fact that their cohort included more younger person-years (mean age at baseline = 39 years, with 25 years of follow-up) and consumed more than twice the amount of milk (531 g/day) than was consumed in our cohort (215 g/day) might account for the strength of the association. Two cohort studies conducted in Norway, where people also consume milk frequently (>1.7 glasses/day), reported both positive (17) and negative (20) associations between milk intake and breast cancer. The former study included older person-years (mean age = 43 years, with 10.4 years of follow-up) than the latter study (mean age = 40.7 years, with 6.2 years of follow-up, which was similar to our premenopausal analysis). In addition, the main type of milk consumed in the former study was whole milk, whereas in the latter study, it was skim and low-fat milk. Other cohort studies conducted in the United States, where people consume less than one glass of milk per day, reported either decreased risk with greater intake of milk (18) (age range = 35–65 years, with <5 years of follow-up) or no association (16) (median age = 55 years, with 20 years of follow-up). Most of the case–control studies with positive associations between milk intake and breast cancer risk also had the common components of older age and consumption of whole milk (12,14, 15,21).

Validity of the dietary measurement should be considered in the interpretation of these conflicting results between studies. However, intake assessments of commonly consumed beverages, such as milk or coffee, have high validity and reproducibility (38). Furthermore, by updating and averaging the repeated measurements of diet five times, we reduced within-person fluctuations and took into account changes over time, which should improve validity compared with most previous studies.

Confounding might be another reason for the conflicting results in previous studies. In the present study, we adjusted all food and nutrient items for total energy intake, so that the relative rather than the absolute amount of intake could be evaluated (40). We also controlled for a wide range of potential confounders. Adjustment for other nutrient factors changed the result meaningfully in this study. For example, the RR of premenopausal breast cancer for the highest intake of total milk relative to the lowest was 0.78 (95% CI = 0.64 to 0.96) in the age-adjusted model, 0.77 (95% CI = 0.62 to 0.95) in a multivariable model without other dietary factors, and 0.69 (95% CI = 0.54 to 0.87) in the fully adjusted model. Changes in the RRs for calcium and vitamin D were even greater than those for total milk.

The apparent association of intake of milk and its related nutrients with reduced risk of breast cancer disappeared quickly after menopause. This diminishing association with increasing age was also observed in an Italian case–control study (45). It is not clear why milk and its related nutrients might show different associations with pre- and postmenopausal breast cancer.

Components of milk other than vitamin D and calcium might be responsible for the associations we observed. Lactose can aid the absorption of dietary calcium as well as promote the growth of lactic acid-producing bacteria in the large intestine. Although lactose intake is suspected to be associated with an increased risk of ovarian cancer (46) and decreased risk of colon cancer (47), little is known about its relationship with breast cancer. Lactose has been hypothesized to increase ovarian cancer risk by direct toxicity to oocytes and by inducing premature ovarian failure (48). This effect could reduce exposure of breast tissue to estrogen. Some epidemiologic studies reported an inverse association between fermented milk (6) and yogurt (12) and risk of breast cancer, interpreting this as an effect of lactose or lactic acid.

Other components of milk have the potential to explain the apparent protective association with breast cancer. Conjugated linoleic acid (CLA), a mixture of positional and geometric isomers of linoleic acid, comes from dairy (60%) and beef (32%) products (49) and is a potent anticarcinogen in animal models (50). However, the inverse association we saw with dairy foods was probably not due to CLA because it was strongest for low-fat dairy products, which should have low CLA content.

In conclusion, high intakes of dairy products, especially low-fat dairy and skim/low-fat milk, may be associated with a modest reduction in the risk of breast cancer in premenopausal women but not in postmenopausal women. Other constituents of dairy foods, such as calcium, total vitamin D, lactose, and phosphorus, also showed inverse associations with risk of premenopausal breast cancer, but their independent associations with breast cancer are difficult to distinguish. Further study of the relationship between dairy product consumption and breast cancer is warranted, with a specific focus on premenopausal women.

Table 1.

Age-adjusted characteristics of participants in the Nurses' Health Study (1980), according to total milk (servings/day), calcium (mg/day), and vitamin D (IU/day) intake*

	Total milk intake (servings/day)			Total calcium intake (mg/day)			Total vitamin D intake (IU/day)
Characteristics	Low ≤3/mo (n = 22 180)	Medium 5/wk–1/day (n = 27 330)	High >1/day (n = 18 446)	Low ≤500 (n = 20 790)	Medium 701–800 (n = 11 010)	High >1250 (n = 5686)	Low ≤150 (n = 33 850)	Medium 250–300 (n = 5506)	High >500 (n = 16 310)
*Categories for total milk, total calcium, and total vitamin D intakes were chosen from their lowest, middle, and highest categories from Table 2 and Table 3. For example, categories for total milk intake were the first, third, and fourth categories of total milk in Table 2, and categories for total calcium intake were the first, fourth, and seventh categories for total calcium in Table 3. Therefore, the number of women among the categories presented here do not add up to the total of 88 691. IU = international units; BMI = body mass index (kg/m2); METs = metabolic equivalent-hours (with activity at rest = 1.0). One serving of milk is an 8-oz glass.
†Individual calcium supplement users in 1984.
‡Individual vitamin D supplement users in 1984.
§Current users of hormones (e.g., hormone replacement therapy) among postmenopausal women.
∥Nutrients are energy-adjusted values determined by using the residual method.
¶In 1980, the intake amount of total calcium was almost the same as the amount of dietary calcium because data on multivitamin intake was the only source of data on calcium supplement (measurement of individual calcium supplement intake began in 1982).
Number of women
Mean age, y	46.7	46.7	46.7	46.7	46.7	46.7	46.7	46.7	46.7
History of benign breast disease, %	24.3	24.6	24.2	23.5	24.8	24.9	23.9	24.6	25.8
History of osteoporosis, %	1.1	1.2	1.5	1.1	1.3	1.5	1.1	1.4	1.4
Family history of breast cancer, %	6.5	6.9	6.6	6.6	6.7	6.9	6.8	6.7	6.7
Mean BMI at age 18 years	21.4	21.4	21.4	21.1	21.5	21.7	21.4	21.4	21.4
Mean height, in.	64.4	64.5	64.6	64.3	64.5	64.6	64.4	64.5	64.6
Calcium supplement, %†	24.2	24.1	22.4	3.8	19.2	64.9	13.4	22.2	42.5
Multivitamin supplement, %	30.2	35.6	38.5	27.4	35.3	46.2	6.5	24.6	97.1
Vitamin D supplement, %‡	2.6	2.6	2.9	0.6	1.8	8.9	0.1	2.0	8.7
Physical activity, ≥4 METs/day	27.3	29.7	29.1	24.0	30.7	31.5	26.6	29.3	32.5
Current smokers, %	32.7	27.1	26.9	33.5	27.7	26.1	33.0	25.8	25.7
Nulliparous, %	7.2	6.6	6.5	7.0	6.9	7.0	6.9	6.8	7.1
First birth after age 30, %	7.6	8.3	8.7	7.5	8.3	9.2	7.8	8.4	7.9
Postmenopausal, %	36.9	36.7	36.6	37.1	36.2	37.3	36.9	36.9	36.7
Postmenopausal hormone use, %§	19.4	19.7	20.2	19.7	19.9	21.0	17.8	20.0	23.8
Change in milk intake in last 10 years, %	32.1	40.1	42.4	34.4	39.8	48.7	35.4	41.6	43.7
Mean daily intakes of
Total dairy, servings/day	0.90	1.92	3.45	0.73	1.88	3.96	1.17	2.59	2.16
Low-fat dairy, servings/day	0.24	0.95	2.09	0.21	0.84	2.67	0.39	1.45	1.26
Total milk, servings/day	0.02	0.93	2.43	0.15	0.82	2.65	0.29	1.56	1.20
Total calcium, mg∥¶	506	770	1101	397	748	1503	542	908	906
Calcium from dairy products, %	39.2	65.1	77.4	39.9	62.1	78.9	47.1	69.6	63.3
Total vitamin D, IU∥	205	312	421	174	292	619	87	274	750
Calories, kcal	1398	1593	1841	1574	1536	1563	1542	1670	1414
Total fat, g∥	73.1	68.6	65.5	74.4	68.9	61.3	72.5	67.4	67.1
Dairy fat, g∥	11.7	14.8	18.1	9.0	15.2	22.2	12.4	16.2	15.7
Glycemic index∥	8452	8680	8525	8726	8601	8251	8526	8605	8540
Beta-carotene, μg∥	4549	4418	4287	3438	4745	5351	4398	4380	4680
Vitamin E, IU∥	75.6	67.6	64.8	52.7	68.0	116.8	36.8	47.6	170.0
Lactose, g∥	3.9	15.6	29.1	4.0	13.4	39.0	6.2	21.2	19.4
Total phosphorus, mg∥	994	1181	1407	921	1163	1670	1014	1274	1287
Alcohol, g	7.7	5.8	4.9	8.2	5.9	3.7	7.7	5.0	5.4

Table 2.

Multivariable relative risks (RRs) and 95% confidence intervals (CIs) of invasive breast cancer by menopausal status according to dairy food intakes in the Nurses' Health Study (1980–1996)

	Premenopausal women				Postmenopausal women
	No. of case women	RR1* (95% CI)	RR2† (95% CI)	RR3‡ (95% CI)	No. of case women	RR1*§ (95% CI)
*RR1 was adjusted for age in 5-year categories, time period, physical activity in METs (metabolic equivalent-hours [with activity at rest = 1.0]), history of benign breast disease, family history of breast cancer, height, weight change since age 18, body mass index at age 18, age at menarche, parity, age at first birth, alcohol intake, total energy intake, total fat intake, glycemic index, β-carotene intake, and total active vitamin E intake.
†RR2 was adjusted for all the covariates in the RR1 model plus total calcium intake.
‡RR3 was adjusted for all the covariates in the RR1 model plus total vitamin D intake.
§For the analysis of postmenopausal women, age at menopause and postmenopausal hormone use (e.g., hormone replacement therapy) were added to the multivariable model.
∥P values for trend were calculated with the Wald statistic using median values for each category of intake. The estimator (β) is an increment of the log of odds for an increase of one serving/day, and SE is the standard error of β.
¶Milk intake during high school years was first measured in 1986, and follow-up for this variable started in 1986. Thus, there are fewer case subjects in this category when compared with the number of case subjects in other categories.
Total No. of case women	827				2345
Total dairy, servings
≤1/day	202	1.00 (referent)	1.00 (referent)	1.00 (referent)	439	1.00 (referent)
>1/day – ≤3/day	433	0.87 (0.73 to 1.04)	0.87 (0.70 to 1.08)	0.90 (0.75 to 1.09)	1239	0.95 (0.85 to 1.07)
>3/day	192	0.73 (0.58 to 0.92)	0.76 (0.56 to 1.03)	0.80 (0.63 to 1.03)	667	0.97 (0.85 to 1.12)
Ptrend [β ± SE(β)]∥		.009 [−0.11 ± 0.04]	.10 [−0.09 ± 0.05]	.10 [−0.08 ± 0.05]		.97 [−0.001 ± 0.02]
Low-fat dairy, servings
≤3/month	155	1.00 (referent)	1.00 (referent)	1.00 (referent)	226	1.00 (referent)
>3/month – ≤4/week	201	0.83 (0.67 to 1.03)	0.83 (0.67 to 1.03)	0.83 (0.67 to 1.03)	486	0.98 (0.84 to 1.16)
>4/week – ≤1/day	207	0.75 (0.60 to 0.93)	0.74 (0.58 to 0.93)	0.76 (0.61 to 0.95)	658	1.00 (0.86 to 1.18)
>1/day	264	0.68 (0.55 to 0.86)	0.70 (0.55 to 0.91)	0.73 (0.57 to 0.92)	975	1.01 (0.86 to 1.19)
Ptrend [β ± SE(β)]∥		.003 [−0.20 ± 0.06]	.02 [−0.18 ± 0.07]	.03 [−0.15 ± 0.07]		.67 [0.02 ± 0.04]
High-fat dairy, servings
≤4/week	123	1.00 (referent)	1.00 (referent)	1.00 (referent)	409	1.00 (referent)
>4/week – ≤1/day	291	1.08 (0.87 to 1.34)	1.09 (0.88 to 1.36)	1.08 (0.87 to 1.33)	832	0.95 (0.84 to 1.07)
>1/day – ≤2.5/day	313	1.05 (0.84 to 1.31)	1.08 (0.86 to 1.35)	1.05 (0.84 to 1.32)	834	0.97 (0.85 to 1.10)
>2.5/day	100	0.83 (0.62 to 1.10)	0.86 (0.65 to 1.16)	0.84 (0.63 to 1.12)	270	1.01 (0.85 to 1.20)
Ptrend [β ± SE(β)]∥		.06 [−0.07 ± 0.04]	.13 [−0.06 ± 0.04]	.09 [−0.07 ± 0.04]		.59 [0.01 ± 0.03]
Total milk, 8-oz glasses
≤3/month	169	1.00 (referent)	1.00 (referent)	1.00 (referent)	327	1.00 (referent)
>3/month – ≤4/week	216	1.00 (0.81 to 1.22)	0.99 (0.80 to 1.22)	1.00 (0.82 to 1.23)	526	1.08 (0.94 to 1.24)
>4/week – ≤1/day	253	0.84 (0.68 to 1.03)	0.81 (0.64 to 1.02)	0.85 (0.69 to 1.06)	783	1.06 (0.92 to 1.21)
>1/day	189	0.69 (0.54 to 0.87)	0.66 (0.50 to 0.88)	0.73 (0.56 to 0.94)	709	1.01 (0.87 to 1.17)
Ptrend [β ± SE(β)]∥		<.001 [−0.16 ± 0.04]	.002 [−0.17 ± 0.06]	.007 [−0.14 ± 0.05]		.59 [−0.01 ± 0.03]
Skim/low-fat milk, 8-oz glasses
Never	247	1.00 (referent)	1.00 (referent)	1.00 (referent)	431	1.00 (referent)
≤1/week	101	0.91 (0.72 to 1.16)	0.91 (0.72 to 1.16)	0.90 (0.71 to 1.15)	248	1.09 (0.93 to 1.28)
>Never – ≤6/week	241	0.90 (0.74 to 1.09)	0.90 (0.74 to 1.10)	0.91 (0.75 to 1.10)	811	1.07 (0.94 to 1.21)
>6/week – ≤1/day	112	0.80 (0.63 to 1.01)	0.81 (0.64 to 1.04)	0.83 (0.65 to 1.06)	338	1.14 (0.98 to 1.32)
>1/day	126	0.72 (0.56 to 0.91)	0.76 (0.57 to 1.00)	0.78 (0.60 to 1.01)	517	1.05 (0.91 to 1.22)
Ptrend [β ± SE(β)]∥		.007 [−0.13 ± 0.05]	.06 [−0.10 ± 0.06]	.08 [−0.09 ± 0.05]		.85 [0.005 ± 0.03]
Whole milk, 8-oz glasses
Never	461	1.00 (referent)	1.00 (referent)	1.00 (referent)	1248	1.00 (referent)
≤1/week	151	1.05 (0.87 to 1.26)	1.04 (0.86 to 1.25)	1.04 (0.86 to 1.25)	421	0.96 (0.86 to 1.08)
>1/week – ≤6/week	136	0.99 (0.81 to 1.21)	0.99 (0.81 to 1.21)	1.00 (0.82 to 1.22)	490	1.05 (0.94 to 1.17)
>6/week – ≤1/day	49	1.00 (0.74 to 1.36)	1.02 (0.75 to 1.38)	1.04 (0.77 to 1.41)	107	1.03 (0.84 to 1.26)
>1/day	30	0.80 (0.54 to 1.16)	0.86 (0.58 to 1.26)	0.87 (0.59 to 1.28)	79	0.87 (0.69 to 1.10)
Ptrend [β ± SE(β)]∥		.26 [−0.08 ± 0.07]	.48 [−0.05 ± 0.07]	.56 [−0.04 ± 0.07]		.43 [−0.03 ± 0.04]
Milk intake during high school years,   8-oz glasses¶
≤0.5/day	85	1.00 (referent)	1.00 (referent)	1.00 (referent)	383	1.00 (referent)
>0.5/day – ≤1/day	81	0.93 (0.69 to 1.27)	0.93 (0.69 to 1.27)	0.95 (0.70 to 1.29)	406	1.04 (0.91 to 1.20)
>1/day – ≤3/day	136	0.99 (0.75 to 1.31)	1.00 (0.75 to 1.32)	1.03 (0.77 to 1.37)	606	1.08 (0.94 to 1.23)
>3/day	25	0.76 (0.48 to 1.21)	0.77 (0.48 to 1.22)	0.81 (0.51 to 1.28)	114	1.02 (0.82 to 1.26)
Ptrend [β ± SE(β)]∥		.45 [−0.03 ± 0.04]	.47 [−0.03 ± 0.05]	.66 [−0.02 ± 0.05]		.47 [0.02 ± 0.02]
Yogurt, servings
Never	390	1.00 (referent)	1.00 (referent)	1.00 (referent)	1049	1.00 (referent)
≤3/month	213	0.97 (0.82 to 1.15)	0.98 (0.82 to 1.16)	0.97 (0.82 to 1.16)	565	0.93 (0.84 to 1.04)
>3/month – ≤4/week	189	0.92 (0.76 to 1.10)	0.93 (0.77 to 1.13)	0.93 (0.77 to 1.12)	609	0.96 (0.87 to 1.08)
>4/week	35	0.95 (0.66 to 1.37)	1.00 (0.69 to 1.44)	0.98 (0.68 to 1.41)	122	0.94 (0.77 to 1.14)
Ptrend [β ± SE(β)]∥		.77 [−0.06 ± 0.19]	.97 [−0.007 ± 0.19]	.90 [−0.02 ± 0.19]		.58 [−0.06 ± 0.10]
Other cheese, servings
≤1/week	201	1.00 (referent)	1.00 (referent)	1.00 (referent)	502	1.00 (referent)
>1/week – ≤4/week	328	1.00 (0.84 to 1.20)	1.02 (0.85 to 1.23)	1.00 (0.83 to 1.20)	979	0.94 (0.84 to 1.05)
>4/week	298	1.02 (0.84 to 1.24)	1.07 (0.87 to 1.30)	1.01 (0.83 to 1.23)	864	0.99 (0.88 to 1.12)
Ptrend [β ± SE(β)]∥		.52 [−0.04 ± 0.12]	.96 [0.006 ± 0.12]	.67 [−0.05 ± 0.12]		.88 [0.01 ± 0.07]

Table 3.

Multivariate relative risks (RRs)* and 95% confidence intervals (CIs) of invasive breast cancer by menopausal status, according to the categories of calcium and vitamin D intake in the Nurses' Health Study (1980–1996)

	Premenopausal women			Postmenopausal women
	No. of case women†	Cumulative average diet model‡	1980 diet model§	No. of case women†	Cumulative average diet model‡∥	1980 diet model§
*Multivariate model included the same covariates as RR1 in Table 2 (i.e., age in 5-year categories, time period, physical activity in METs [metabolic equivalent-hours {with activity at rest = 1.0}], history of benign breast disease, family history of breast cancer, height, weight change since age 18, body mass index at age 18, age at menarche, parity, age at first birth, alcohol intake, total energy intake, total fat intake, glycemic index, β-carotene intake, and total active vitamin E intake). IU = international units.
†Number of case subjects from the cumulative average diet model.
‡Models used cumulative information on variables, updated and averaged. Supplement users were excluded in the dietary calcium, dairy calcium, total nondairy calcium, and dietary vitamin D analyses.
§Multivariate model using 1980 (baseline) nutrient values with the same covariates as in * above. Supplement users were excluded on the basis of the updated supplement information. Other nondietary variables and alcohol intake were updated for the follow-up person-years.
∥For the analysis of postmenopausal women, age at menopause and postmenopausal hormone use (e.g., hormone replacement therapy) were added to the multivariable model.
¶All nutrient categories are based on the energy-adjusted values.
#P values for trend were calculated with the Wald statistic using median values for each category of intake. The estimator (β) is an increment of the log of odds for an increase of 1 mg of calcium or 1 IU of vitamin D, and SE is the standard error of β.
**Dietary calcium, dairy calcium, and dietary vitamin D analyses did not include supplement users.
Total No. of case women	827			2345
Total calcium, mg/day¶
≤500	142	1.00 (referent)	1.00 (referent)	240	1.00 (referent)	1.00 (referent)
>500 – ≤600	106	0.88 (0.68 to 1.13)	0.82 (0.65 to 1.03)	216	0.86 (0.72 to 1.04)	0.93 (0.81 to 1.06)
>600 – ≤700	133	0.97 (0.76 to 1.24)	1.07 (0.86 to 1.33)	293	0.94 (0.79 to 1.12)	0.90 (0.78 to 1.03)
>700 – ≤800	119	0.95 (0.74 to 1.22)	0.96 (0.75 to 1.22)	292	0.92 (0.77 to 1.10)	0.94 (0.81 to 1.09)
>800 – ≤1000	161	0.82 (0.64 to 1.05)	0.89 (0.71 to 1.12)	518	0.93 (0.79 to 1.10)	0.91 (0.79 to 1.04)
>1000 – ≤1250	104	0.75 (0.57 to 0.99)	0.64 (0.48 to 0.86)	433	0.90 (0.76 to 1.07)	0.83 (0.70 to 0.98)
>1250	62	0.80 (0.58 to 1.12)	0.88 (0.64 to 1.23)	353	0.93 (0.77 to 1.12)	0.95 (0.79 to 1.15)
Ptrend [β ± SE(β)]#		.05 [−.00028 ± .00015]	.06 [−.00027 ± .00014]		.70 [−.00003 ± .000008]	.18 [−.00011 ± .00008]
Dietary calcium, mg/day**
≤500	150	1.00 (referent)	1.00 (referent)	256	1.00 (referent)	1.00 (referent)
>500 – ≤600	99	0.80 (0.62 to 1.04)	0.80 (0.63 to 1.01)	254	1.01 (0.84 to 1.20)	0.91 (0.80 to 1.04)
>600 – ≤700	129	1.01 (0.79 to 1.28)	1.10 (0.89 to 1.37)	276	1.03 (0.86 to 1.22)	0.88 (0.76 to 1.01)
>700 – ≤800	85	0.82 (0.62 to 1.09)	0.88 (0.68 to 1.13)	215	0.97 (0.81 to 1.17)	0.93 (0.80 to 1.07)
>800 – ≤1000	108	0.80 (0.61 to 1.04)	0.91 (0.72 to 1.15)	237	0.86 (0.71 to 1.03)	0.90 (0.78 to 1.03)
>1000	69	0.67 (0.49 to 0.92)	0.76 (0.59 to 0.98)	198	0.99 (0.81 to 1.21)	0.85 (0.73 to 0.98)
Ptrend [β ± SE(β)]#		.02 [−.00047 ± .00019]	.06 [−.00029 ± .00016]		.46 [−.00009 ± .00013]	.05 [−.00018 ± .00009]
Dairy calcium, mg/day**
≤200	124	1.00 (referent)	1.00 (referent)	200	1.00 (referent)	1.00 (referent)
>200 – ≤300	118	0.84 (0.65 to 1.09)	0.80 (0.63 to 1.00)	251	1.01 (0.84 to 1.22)	0.94 (0.82 to 1.07)
>300 – ≤400	110	0.76 (0.58 to 0.99)	0.88 (0.70 to 1.10)	290	1.09 (0.90 to 1.31)	0.98 (0.86 to 1.12)
>400 – ≤600	166	0.79 (0.62 to 1.00)	0.87 (0.70 to 1.07)	393	1.01 (0.85 to 1.21)	0.94 (0.83 to 1.07)
>600 – ≤800	74	0.69 (0.51 to 0.93)	0.70 (0.53 to 0.92)	162	0.85 (0.68 to 1.05)	0.90 (0.77 to 1.06)
>800	48	0.69 (0.48 to 0.98)	0.79 (0.59 to 1.05)	140	1.11 (0.88 to 1.40)	0.88 (0.74 to 1.04)
Ptrend [β ± SE(β)]#		.01 [−.00046 ± .00019]	.06 [−.00030 ± .00016]		.90 [−.00002 ± .00013]	.10 [−.00015 ± .00009]
Total nondairy calcium, mg/day
≤200	110	1.00 (referent)	1.00 (referent)	154	1.00 (referent)	1.00 (referent)
>200 – ≤300	373	1.15 (0.92 to 1.44)	1.01 (0.85 to 1.20)	738	1.00 (0.84 to 1.20)	1.04 (0.93 to 1.15)
>300 – ≤400	163	1.08 (0.82 to 1.43)	1.05 (0.82 to 1.34)	486	0.96 (0.78 to 1.17)	1.03 (0.89 to 1.19)
>400 – ≤600	103	1.11 (0.81 to 1.51)	1.14 (0.76 to 1.71)	483	0.98 (0.79 to 1.20)	0.95 (0.74 to 1.21)
>600	78	1.28 (0.91 to 1.80)	1.31 (0.68 to 2.53)	484	0.93 (0.75 to 1.15)	0.91 (0.55 to 1.48)
Ptrend [β ± SE(β)]#		.31 [.00025 ± .00025]	.34 [−.00043 ± .00045]		.25 [−.00014 ± .00012]	.67 [−.00012 ± .00029]
Total vitamin D, IU/day
≤150	268	1.00 (referent)	1.00 (referent)	526	1.00 (referent)	1.00 (referent)
>150 – ≤200	111	0.90 (0.72 to 1.13)	1.06 (0.86 to 1.32)	312	1.04 (0.90 to 1.20)	1.04 (0.91 to 1.18)
>200 – ≤250	88	0.87 (0.68 to 1.11)	0.80 (0.61 to 1.06)	272	1.04 (0.89 to 1.21)	1.04 (0.89 to 1.21)
>250 – ≤300	66	0.79 (0.60 to 1.05)	0.86 (0.63 to 1.18)	211	0.94 (0.79 to 1.11)	0.89 (0.74 to 1.08)
>300 – ≤350	55	0.76 (0.56 to 1.03)	0.92 (0.66 to 1.28)	209	1.11 (0.94 to 1.32)	0.93 (0.75 to 1.14)
>350 – ≤500	115	0.77 (0.60 to 0.99)	0.93 (0.72 to 1.22)	411	1.01 (0.87 to 1.17)	0.99 (0.85 to 1.16)
>500	124	0.72 (0.55 to 0.94)	0.89 (0.68 to 1.15)	404	0.94 (0.80 to 1.10)	0.93 (0.80 to 1.08)
Ptrend [β ± SE)β)]#		.01 [−.00059 ± .00024]	.31 [−.00023 ± .00023]		.27 [−.00015 ± .00014]	.27 [−.00015 ± .00013]
Dietary vitamin D, IU/day**
≤75	86	1.00 (referent)	1.00 (referent)	170	1.00 (referent)	1.00 (referent)
>75 – ≤150	222	1.14 (0.89 to 1.47)	1.03 (0.81 to 1.30)	485	0.96 (0.81 to 1.15)	0.92 (0.80 to 1.05)
>150 – ≤250	172	0.88 (0.67 to 1.16)	0.92 (0.72 to 1.18)	604	0.99 (0.82 to 1.18)	0.99 (0.86 to 1.15)
>250 – ≤300	49	1.05 (0.73 to 1.51)	0.77 (0.52 to 1.15)	135	0.85 (0.67 to 1.07)	0.93 (0.75 to 1.15)
>300	34	0.66 (0.43 to 1.00)	0.84 (0.59 to 1.18)	173	1.06 (0.85 to 1.34)	0.86 (0.70 to 1.05)
Ptrend [β ± SE(β)]#		.02 [−.00129 ± .00056]	.11 [−.00082 ± .00051]		.93 [.000003 ± .00032]	.32 [−.00029 ± .00029]

Table 4.

Multivariable relative risks (RRs) and 95% confidence intervals (CIs) for estimating the associations of calcium supplements with premenopausal and postmenopausal breast cancer by tertiles of dietary calcium and by the use of postmenopausal hormones (e.g., hormone replacement therapy) in the Nurses' Health Study (1980–1996)

	Premenopausal women, by calcium supplement use
	Nonuser	1–399 mg	400–899 mg	≥900 mg
Total No. of case women*	571	64	71	52
RR (95% CI)†	1.00	0.84 (0.64 to 1.10)	1.00 (0.76 to 1.31)	1.10 (0.81 to 1.50)
By dietary calcium‡
t1 (101–563, 449 mg)§	1.00	0.77 (0.47 to 1.25)	0.83 (0.51 to 1.36)	1.50 (0.97 to 2.34)
t2 (564–798, 670 mg)	1.00	0.94 (0.62 to 1.42)	1.04 (0.69 to 1.56)	1.25 (0.79 to 1.98)
t3 (799–2802, 998 mg)	1.00	0.79 (0.49 to 1.28)	1.10 (0.71 to 1.72)	0.44 (0.19 to 1.01)

	Postmenopausal women, by calcium supplement use
	Nonuser	1–399 mg	400–899 mg	≥900 mg
*The numbers of case subjects do not add up to 827 in premenopausal women and 2345 in postmenopausal women because some of the women had missing values for the amount of supplement intakes in their food-frequency questionnaires.
†Adjusted for the same covariates as RR1 in Table 2 (i.e., age in 5-year categories, time period, physical activity in METs [metabolic equivalent-hours {with activity at rest = 1.0}], history of benign breast disease, family history of breast cancer, height, weight change since age 18, body mass index at age 18, age at menarche, parity, age at first birth, alcohol intake, total energy intake, total fat intake, glycemic index, β-carotene intake, and total active vitamin E intake) plus dietary calcium intake. Nonuser RRs were the referent values (1.00).
‡From the same model as in † (of this table), except that tertiles (t1, t2, and t3) of dietary calcium intake were used as stratifying variables.
§Each tertile (t1, t2, and t3) shows the range and median values of intake in parentheses. The values given are cumulative averages and are energy-adjusted.
∥From the same model as in † (of this table), except that current use of postmenopausal hormones (e.g., hormone replacement therapy) was used as a stratifying variable.
Total No. of case women*	1122	262	359	288
RR (95% CI)†	1.00	0.94 (0.81 to 1.08)	0.97 (0.86 to 1.11)	0.93 (0.81 to 1.08)
By dietary calcium‡
t1 (101–563, 449 mg)§	1.00	0.95 (0.74 to 1.22)	0.78 (0.61 to 1.00)	0.87 (0.68 to 1.12)
t2 (564–798, 670 mg)	1.00	0.84 (0.66 to 1.06)	1.05 (0.86 to 1.28)	0.94 (0.75 to 1.17)
t3 (799–2802, 998 mg)	1.00	1.04 (0.83 to 1.31)	1.06 (0.86 to 1.31)	1.00 (0.79 to 1.25)
By hormone use∥
Current nonusers	1.00	0.94 (0.78 to 1.13)	0.87 (0.72 to 1.04)	1.01 (0.83 to 1.22)
Current users	1.00	0.95 (0.76 to 1.19)	1.03 (0.86 to 1.25)	0.84 (0.69 to 1.04)

Table 5.

Multivariable relative risks (RRs) and 95% confidence intervals (CIs) for estimating the associations of vitamin D supplements with premenopausal and postmenopausal breast cancer by tertiles of dietary vitamin D and by the use of postmenopausal hormones (e.g., hormone replacement therapy) in the Nurses' Health Study (1980–1996)

	Premenopausal women, by vitamin D supplement use			Postmenopausal women, by vitamin D supplement use
	Nonuser	1–399 IU	≥400 IU	Nonuser	1–399 IU	≥400 IU
*The numbers of case subjects do not add up to 827 in premenopausal women and 2345 in postmenopausal women because some of the women had missing values for the amount of supplement intakes in their food-frequency questionnaires. IU = international units.
†Adjusted for the same covariates as RR1 in Table 2 (i.e., age in 5-year categories, time period, physical activity in METs [metabolic equivalent-hours {with activity at rest = 1.0}], history of benign breast disease, family history of breast cancer, height, weight change since age 18, body mass index at age 18, age at menarche, parity, age at first birth, alcohol intake, total energy intake, total fat intake, glycemic index, β-carotene intake, and total active vitamin E intake) plus dietary vitamin D. Nonuser RRs were the referent values.
‡From the same model as in † (of this table), except that tertiles of dietary vitamin D intake were used as stratifying variables.
§Each tertile (t1, t2, and t3) shows the range and median values of intake in parentheses. The values given are cumulative averages and are energy-adjusted.
∥From the same model as in † (of this table), except that current use of postmenopausal hormones (e.g., hormone replacement therapy) was used as a stratifying variable.
Total No. of case women*	494	110	154	1253	234	544
RR (95% CI)†	1.00	0.91 (0.72 to 1.14)	0.88 (0.71 to 1.10)	1.00	0.95 (0.82 to 1.10)	0.96 (0.85 to 1.08)
By dietary vitamin D‡
t1 (0–107, 67 IU)§	1.00	0.85 (0.59 to 1.22)	0.80 (0.57 to 1.14)	1.00	0.72 (0.53 to 0.97)	1.04 (0.85 to 1.27)
t2 (108–192, 146 IU)	1.00	0.91 (0.63 to 1.31)	0.90 (0.64 to 1.25)	1.00	0.94 (0.74 to 1.20)	0.99 (0.82 to 1.18)
t3 (193–1299, 265 IU)	1.00	0.99 (0.67 to 1.45)	0.96 (0.68 to 1.35)	1.00	1.11 (0.89 to 1.38)	0.88 (0.74 to 1.05)
By hormone use∥
Current nonusers	—	—	—	1.00	0.93 (0.77 to 1.13)	0.89 (0.76 to 1.04)
Current users	—	—	—	1.00	0.97 (0.77 to 1.22)	1.06 (0.89 to 1.25)

Supported by Public Health Service grant CA87969 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services; American Cancer Society Internation Union Against Cancer (UICC) International Fellowships for Beginning Investigators (ACSBI) 2000; and R04-2000-00059 from the Basic Research Program of the Korean Science and Engineering Foundation.

We thank Karen Corsano, Diane Feskanich, Laura Sampson, Debbie Flynn, and Francine Laden for their technical support and valuable advice.