Dietary Calcium and Magnesium, Major Food Sources, and Risk of Type 2 Diabetes in U.S. Black Women Free

The Black Women’s Health Study of Boston University and Howard University is a prospective cohort study of ∼59,000 women who were aged 21–69 years at baseline in 1995. Women were enrolled through questionnaires mailed mainly to subscribers of Essence magazine (a national lifestyle magazine whose readership consists almost entirely of African-American women) (15). Participants were from all parts of the U.S., and the education level was only slightly higher than reported for nationally representative data on African-American women (16). Information has been collected using biennial mailed questionnaires, and response rates have been >80% of the original cohort for each follow-up questionnaire. For the current analysis, baseline exclusions included a history of diabetes (including gestational diabetes), cancer (except nonmelanoma skin cancer), myocardial infarction, or stroke (n = 5,424); current pregnancy (n = 956); and ≥10 items on the dietary questionnaire left blank or implausible reported total energy intakes (<500 kcal/day or >3,800 kcal/day) (n = 5,967). We also excluded women who did not reach the age of 30 years during follow-up or had diabetes diagnosed before age 30 years during follow-up (as these cases may include a substantial number of women with type 1 diabetes) (n = 2,034) and those that did not complete any follow-up questionnaires (n = 268). In addition, we excluded women with missing values for BMI, calcium intake, or magnesium intake (n = 3,232). A total of 41,186 women remained for the current analysis. The study was approved by the institutional review boards of Boston University and Howard University.

Diabetes was assessed on each biennial follow-up questionnaire by asking whether the women had been diagnosed with diabetes. The accuracy of self-reports of diabetes was examined using questionnaires filled out by the physicians of a random sample of participants who reported having been diagnosed with diabetes. Of 424 women who were sent release forms asking permission to contact their physicians, 184 returned signed releases. Physician questionnaires were obtained for 141 women and indicated a diagnosis of type 2 diabetes for 134 (95%) women. Of seven other women, two had type 1 diabetes, three had metabolic syndrome, one had steroid-induced diabetes, and one did not have diabetes or related conditions.

Diet was assessed using the 68-item baseline Block food frequency questionnaire that inquired about the consumption of foods and drinks during the previous year (17). Questions included nine response categories ranging from “never to <1 per month” to “2 or more per day” for foods and from “never to <1 per month” to “6 or more per day” for drinks. We calculated whole-grain and dairy consumption by adding the daily number of servings of individual items (whole grains: “dark breads, such as wheat, rye, pumpernickel” [“2 slices”] and “high fiber, bran or granola cereals, shredded wheat” [“1 medium bowl”]; dairy: “whole milk” [“8 ounce glass”], “2% milk” [“8 ounce glass”], “skim milk, 1% milk or buttermilk” [“8 ounce glass”], “milk or cream in coffee or tea” [“1 tablespoon”], “yoghurt” [“8 ounces”], “frozen yoghurt” [“1 scoop”], “ice cream” [“1 scoop or 1/2 cup”], “butter on bread or rolls” [“2 pats”], and “cheeses and cheese spreads not including cottage cheese” [“2 slices or two ounces”]). For milk/cream in coffee, we first divided the number of daily servings by 16 to make the serving size comparable with that for other milk items. Low-fat dairy included reduced-fat milk (≤2% fat), reduced-fat yogurt, and reduced-fat frozen yogurt, and high-fat dairy included all other dairy items. Yogurt and frozen yogurt were assumed to be high-fat if women indicated that they rarely used a low-fat version of these. Nutrient intakes were calculated using the National Cancer Institute’s DIETSYS software (version 4.01, food database version 3.7c) (18). The dietary questionnaire was validated in 408 cohort members by comparison with one 3-day dietary record and three telephone 24-h recalls (19). The energy-adjusted deattenuated Pearson correlation coefficients for the food frequency data compared with diet records and 24-h recalls were 0.40 and 0.63 for saturated fat, 0.32 and 0.79 for calcium, and 0.57 and 0.67 for fiber, respectively (19).

The baseline questionnaire requested information about age, number of years of school finished, parental history of diabetes, consumption of alcoholic drinks in the past year, cigarette smoking history, average number of hours per week spent on strenuous physical activity in the past year, weight, and height. BMI was calculated as weight in kilograms divided by the square of height in meters.

Person-years of exposure were calculated from the year of return of the baseline questionnaire or age 30 for participants who only reached age 30 during follow-up to the year on which the first diagnosis of diabetes was indicated on the questionnaire, the year of the last completed questionnaire, or 2003. Cox proportional hazards regression models using age as the time scale were used to calculate hazard ratios for type 2 diabetes and corresponding 95% CIs comparing categories of dietary intake with the lowest category of dietary intake. Categories for the dietary variables were defined using quintiles or the cutoff points specified on the questionnaires. Some questionnaire categories were combined to avoid small numbers and lack of precision for estimates of association. All hazard ratios were adjusted for age (years) and total energy intake (kilocalories per day). The multivariate hazard ratios were further adjusted for BMI (weight in kilograms divided by the square of height in meters), smoking status (never, past, current <15/day, or current ≥15/day), strenuous physical activity (0, <1, 1–2, 3–6, or ≥7 h/week), alcohol consumption (0, 0.1–0.4, 0.5–0.9, 1.0–1.9, or ≥2 drinks/day), parental history of diabetes (yes/no), education level (<13, 13–15, 16, or ≥17 years), coffee consumption (none, less than five cups/week, five to six cups/week, one cup/day, two to three cups/day, and four or more cups/day), sugar-sweetened soft drink consumption (none, less than five cups/week, five to six cups/week, one cup/day, or two or more cans/day), and quintiles of processed meat and other red meat consumption. We used baseline (1995) data for all exposure variables and potential confounders. To test for linear trends across categories, we modeled the median of each category of consumption as a continuous variable. Nutrient intakes were adjusted for total energy intake using the residual method (20). We calculated Pearson correlations between dietary intakes with adjustment for total energy intake. All reported P values were two tailed, and P values <0.05 were considered statistically significant. All analyses were performed using SAS software, version 8.2 (SAS Institute, Cary, NC).

Characteristics of the study population according to dietary magnesium and calcium intakes are shown in Table 1. Women with a higher magnesium intake tended to be older, more highly educated, leaner, more physically active, and nonsmokers. Women with a higher calcium intake tended to be more highly educated, more physically active, and nonsmokers. In addition, higher intakes of magnesium and calcium were associated with higher intakes of fiber, whole grains, dairy, and coffee and lower intakes of alcohol, saturated fat, linoleic acid, red meat, and sugar-sweetened soft drinks. Correlations with dietary magnesium were 0.58 for calcium, 0.25 for dairy, 0.38 for low-fat dairy, −0.08 for high-fat dairy, and 0.54 for whole grains. Correlations with dietary calcium were 0.74 for dairy, 0.73 for low-fat dairy, 0.23 for high-fat dairy, and 0.22 for whole grains.

During 264,443 person-years of follow-up, we documented 1,964 new cases of type 2 diabetes. Higher calcium intakes were associated with a lower risk of type 2 diabetes, but the association was substantially weakened after adjustment for potential confounders (Table 2). For magnesium intake, the multivariate-adjusted hazard ratio for type 2 diabetes was 0.69 (95% CI 0.59–0.81) for the highest compared with the lowest quintile (P trend <0.0001). When magnesium and calcium intakes were mutually adjusted, an inverse association with magnesium remained, whereas the association between calcium intake and risk of type 2 diabetes disappeared (Table 2). Magnesium intake was highly correlated (r = 0.82) with fiber intake, and after adjustment for fiber intake the multivariate hazard ratio was 0.75 (0.62–0.92) for the highest compared with the lowest quintile of magnesium intake (P trend = 0.008). Additional adjustment for intakes of saturated fatty acids and linoleic acid did not appreciably change the results (<10% change in regression coefficients).

Women who used calcium supplements had a lower risk of type 2 diabetes than nonusers (multivariate-adjusted hazard ratio 0.82 [95% CI 0.73–0.91]), and this association was similar for calcium supplements with vitamin D (0.82 [0.69–0.98]) and without vitamin D (0.81 [0.72–0.91]). However, among calcium supplement users neither the amount (1.34 [1.03–1.74] for >600 mg/day vs. ≤250 mg/day) nor the duration of calcium supplement use (0.86 [0.64–1.16] for ≥5 years vs. <1 year) was associated with a lower risk of type 2 diabetes. Exclusion of women who used calcium supplements did not strengthen the association between dietary calcium and risk of type 2 diabetes (0.92 [0.78–1.10] for highest vs. lowest quintile).

Higher consumption of whole grains was associated with a lower risk of type 2 diabetes, and this association was independent of other risk factors (Table 3). The inverse association remained after additional adjustment for magnesium intake (hazard ratio 0.73 [95% CI 0.63–0.85] for ≥1/day vs. <1/week; P trend <0.0001), suggesting that the results for whole grains could not be explained by its contribution to magnesium intake. Dairy consumption was associated with a lower risk of type 2 diabetes, but in multivariate models only an inverse association for low-fat dairy remained (Table 3) that was further weakened after adjustment for magnesium intake (0.90 [0.78–1.03] for ≥1/day vs. <1/week; P trend = 0.13). The inverse association between magnesium intake and risk of type 2 diabetes remained after adjustment for whole grains (0.81 [0.68–0.97] for highest vs. lowest quintiles; P trend = 0.02) or low-fat dairy (0.72 [0.61–0.84]; P trend <0.0001).

In this 8-year prospective study of 41,186 U.S. black women, higher dietary magnesium intake was associated with a lower risk of type 2 diabetes. Higher calcium intake was not independently associated with risk of type 2 diabetes. Of the food sources, consumption of whole grains and low-fat dairy, but not high-fat dairy, were associated with a lower risk of type 2 diabetes.

Although African Americans are at increased risk of type 2 diabetes, few studies have examined whether diet may affect diabetes risk in this group (12,21). In the Atherosclerosis Risk in Communities study, low serum magnesium concentrations predicted type 2 diabetes in white, but not in black, participants and no association with dietary magnesium was observed in either group (12). However, serum magnesium concentrations may not accurately reflect magnesium status of other tissues (12). In a cross-sectional study of African Americans without diabetes, higher magnesium intakes were associated with higher insulin sensitivity (22). The low median magnesium intake observed in the current study agrees with that reported for a national sample of African-American women based on 24-h recalls (median 183 mg/day) and is well below the recommended dietary allowance for women aged >30 years (320 mg/day) (13). Effects of poor magnesium status on glucose homeostasis are plausible and may be mediated through oxidative stress, the role of magnesium as cofactor for enzymes involved in glucose metabolism, or the effects of intracellular ion levels on insulin sensitivity and insulin secretion (23). Some short-term intervention studies also provide support for a beneficial effect of magnesium intake on glucose metabolism, but results have not been consistent (24). In the absence of more definitive results on magnesium from larger, longer-term, intervention studies, the recommendation of magnesium-rich foods seems prudent. In adult participants of the National Health and Nutrition Examination Survey 1999–2000, the most important sources of magnesium were vegetables (12.9%), bread and cold cereals (10.7%), and milk (7.5%) (13). In addition to magnesium, many other components of whole grains including fibers, vitamin E, several B vitamins, and lignans may contribute to beneficial effects on glucose metabolism (25).

Dairy consumption was strongly associated with a lower incidence of the metabolic syndrome and hyperglycemia in both white and black participants of the CARDIA study (26). In the current study and a study of male health professionals (11), more modest associations were observed and these associations were limited to low-fat dairy. Calcium intake was inversely associated with risk of type 2 diabetes in female nurses (7). Further studies of calcium and dairy intakes in relation to type 2 diabetes are needed to establish whether effects are independent of other dietary and lifestyle factors.

We controlled for confounding by known risk factors of type 2 diabetes in detail. However, because of the nonrandomized study design, residual confounding cannot be completely excluded as a potential explanation of the observed associations. The assessment of type 2 diabetes relied on self-reports of physician diagnosis. Our validation study indicated that the specificity of self-report was high. Underascertainment of diabetes in the cohort would only have affected hazard ratios if diabetes detection was associated with the studied exposures (27). Measurement error in the assessment of dietary intakes is inevitable and will have led to some misclassification of the studied dietary exposures. We only used dietary intake assessed at baseline, which may have contributed to misclassification because of dietary changes during follow-up. Given the prospective design of the study, this misclassification was unlikely to be associated with the studied outcome and therefore probably led to underestimation of hazard ratios. Furthermore, our results for dietary magnesium and whole grains are consistent with observational data from diverse populations (3–6,8–10) and results of metabolic studies (24,28).

We observed an inverse association between magnesium intake and risk of type 2 diabetes in this cohort of African-American women, whereas no independent association for calcium intake was observed. Because coconstituents of magnesium in foods may contribute to the observed association for magnesium intake, it is not clear whether the results apply to supplemental magnesium intake. These findings indicate that higher consumption of magnesium-rich foods, particularly whole-grain products, is associated with a lower risk of type 2 diabetes in African-American women.

This work was supported by National Cancer Institute Grant CA58420 and National Institute of Diabetes and Digestive and Kidney Diseases Grant 1R01DK068738. R.M.V.D. was supported by the Netherlands Organization for Scientific Research (ZonMw VENI Grant no. 916.46.077).