A Prospective Study of Dairy Intake and the Risk of Type 2 Diabetes in Women

The Women’s Health Study (WHS) is a randomized, double-blind, placebo-controlled trial originally designed to test the efficacy of low-dose aspirin and vitamin E in the primary prevention of cardiovascular disease and cancer among 39,876 female health professionals who were without heart disease, stroke, or cancer at baseline (10). Detailed information on dietary intake was provided by 37,183 (93%) of the randomized participants without previously diagnosed diabetes at baseline and who completed a 131-item validated semiquantitative food frequency questionnaire (SFFQ) (11). The study protocol was approved by the Brigham and Women’s Hospital Institutional Review Board, and the protocol adhered to the guidelines put forth in the Helsinki declaration and Belmont Accord.

To assess dietary intake, we used a validated SFFQ that inquired about the average use of ∼131 foods and beverages during the previous year (11). Nutrient intake was computed from the reported frequency of consumption of each specified unit of food or beverage and from published data on the nutrient content of the specified portions. The average daily intakes of individual dairy items were combined to compute dairy intake: low-fat dairy products, including skim or low-fat milk, sherbet, yogurt, and cottage/ricotta cheese; high-fat dairy foods, including whole milk, cream, sour cream, ice cream, cream cheese, and other cheese; and total dairy products, including all of the above. Total milk intake included skim or low-fat milk and whole milk.

Total calcium and vitamin D intakes were calculated from both dietary and supplemental sources. Intakes of dietary calcium and vitamin D were calculated from all dietary sources without supplements. Each nutrient was adjusted for total energy using the residual method (12). In similar cohorts of health professionals, this SFFQ has demonstrated reasonably good validity as a measure of long-term average dietary intakes. Pearson’s correlation coefficients between responses from the SFFQ and those from four 1-week dietary records spaced over a year were 0.56 for total calcium and 0.51 for dietary calcium (12). With respect to milk intake, the correlation coefficients between the SFFQ and dietary records were 0.69 for skim or low-fat milk and 0.56 for whole milk (13). The validity of vitamin D intake was assessed by comparing it with the plasma concentrations of 25-hydroxyvitamin D among 323 healthy women (14). Correlation coefficients between intake and plasma concentration of vitamin D were 0.35 for total vitamin D and 0.25 for dietary vitamin D.

Details regarding ascertainment of incident type 2 diabetes in the WHS have been reported previously (15,16). Using the diagnostic criteria recommended by the American Diabetes Association (ADA), we used three complementary approaches to validate self-reported incident cases of diabetes. Method 1 involved a telephone interview by a study physician of 473 women with self-reported diabetes who participated in a previous study (17). Method 2 used a supplemental questionnaire that contained detailed information mailed to a random sample of 147 women who self-reported diabetes. Method 3 involved contacting individual physicians for medical records of those women consenting for medical record review in method 2. Self-reported diagnoses of diabetes were confirmed in 406 of 446 women who responded via a physician-led telephone interview (positive predictive value 91.0% [95% CI 88.4–93.7]). Separately, among 136 responders to the supplemental questionnaire, 124 women were classified as diabetic using the ADA criteria (positive predictive value 91.2% [86.4–95.9]). Among 113 women who gave permission for study investigators to contact their primary care physician, 97 physicians responded, of whom 90 provided adequate medical record information to apply the ADA criteria. Of these 90 women, 89 were confirmed to have diabetes on the basis of the combined information from the supplemental questionnaire and physician information. The follow-up for this cohort exceeded 99% during the study period.

We computed person-time of follow-up from the beginning of the trial to the date of diagnosis of type 2 diabetes, death from any cause, or the end of the study period (March 2005), whichever came first. The average daily dairy intake was categorized into quintiles of intake, and each quintile was compared with the lowest quintile. After testing the proportional hazard assumption, we used Cox proportional hazards modeling to estimate the rate ratios, or estimates of relative risk (RR), and 95% CIs for developing incident type 2 diabetes for each category of intake compared with the lowest category. The initial model was adjusted for age (continuous), treatment assignment (categorical), and total energy intake (continuous). In multivariate models, we additionally adjusted for BMI (continuous), smoking status (current, past, and never), exercise (quintiles in METs per week), alcohol intake (rarely/never, 1–3 drinks/month, 1–6 drinks/week, or ≥1 drink/day), use of postmenopausal hormone therapy (past, current, and never), high cholesterol, hypertension, and family history of diabetes (yes/no), as well as for dietary factors (all in quintiles) including intakes of dietary fiber, glycemic load, magnesium, calcium, total fat, and vitamin D.

The RR for the continuous measures indicates the change in risk associated with an average increment of one serving per day of the standard portion size. To assess possible effect modification, we conducted analyses stratified by BMI (<25 vs. ≥ 25 kg/m²), physical activity (less than median versus more than median), and family history of diabetes (yes/no), which were determined a priori. We tested the significance of the interaction using the likelihood ratio test by comparing a model with the main effects of dairy intake and the stratifying variable and interaction term with a reduced model with only the main effects. All statistical analyses were conducted using SAS (version 8.0; SAS Institute, Cary, NC).

The baseline characteristics of the cohort according to intake levels are shown in (Table 1). Women with higher dairy intake tended to have slightly lower alcohol consumption, a higher level of physical activity, lower prevalence of hypertension and hypercholesterolemia, a lower glycemic load, slightly higher caffeine intake, and higher intakes of whole grains, nuts, and fruits and vegetables. Women with higher intake of low-fat diary had similar trends, whereas women with high intake of high-fat dairy tended to have less healthy lifestyle patterns.

During the 10-year follow-up, we documented 1,603 incident cases of type 2 diabetes. After adjusting for age, treatment assignment, and total energy intake, the RR for women in the highest quintile of total dairy intake was 0.66 (95% CI 0.56–0.78; P for trend = 0.0001) compared with those in the lowest quintile (Table 2). After further adjustment for other risk factors, the association was attenuated but still statistically significant (RR 0.79 [0.67–0.94]; P for trend = 0.007). Each serving-per-day increase in dairy intake was associated with a 4% lower risk for type 2 diabetes (0.96 [0.93–1.01]).

The inverse association with type 2 diabetes appeared to be mainly attributed to low-fat dairy intake; the age-, calorie intake–, and treatment-adjusted RRs comparing the highest to the lowest quintiles were 0.64 (95% CI 0.54–0.75; P for trend = 0.0001) for low-fat dairy intake and 1.08 (0.92–1.28; P for trend = 0.20) for high-fat dairy intake (Table 3). Additional adjustment for BMI, family history of diabetes, smoking, alcohol intake, physical activity, postmenopausal hormone use, and history of hypercholesterolemia and hypertension did not appreciably change the results. The results also did not change after further adjustment for dietary intakes of calcium, vitamin D, fibers, total fat, glycemic load, and magnesium.

We also evaluated type of dairy foods in relation to risk of type 2 diabetes (Table 4). Intake of yogurt was inversely associated with risk of type 2 diabetes. Additional adjustment for other factors including vitamin D, calcium, and magnesium did not attenuate the inverse association; the multivariate-adjusted RR was 0.82 (95% CI 0.70–0.97) for women who consumed two servings/week yogurt compared with those who consumed less than one serving/day (P for trend = 0.03).

When we examined whether the association of total dairy intake and type 2 diabetes were modified by BMI, physical activity, and family history of diabetes (Table 5), significant inverse association was stronger in women who had no family history of diabetes than in those with family history (P for interaction = 0.03), but we found no significant interactions of dairy intake with BMI (P for interaction = 0.06) and physical activity (P for interaction = 0.29) in relation to type 2 diabetes risk.

In this large prospective cohort study of women, we found a moderate inverse association between dairy consumption, especially low-fat dairy consumption, and incidence of type 2 diabetes that was independent of age, BMI, family history of diabetes, history of hypertension, hypercholesterolemia, smoking history, physical activity, and some dietary factors related to type 2 diabetes.

Previous studies have suggested favorable effects of dairy intake on body weight (2,3,6), hypertension (4), and glucose homeostasis (5,6). A strong inverse association between dairy intake and risk of IRS was observed in the Coronary Artery Risk Development in Young Adults (CARDIA) study of young adults who were overweight (BMI ≥25 kg/m²) but not among leaner individuals (6). In our study, however, we did not find a significant effect modification by BMI, and the overall magnitude of the association with type 2 diabetes (4% reduction per serving increase in dairy foods) was notably smaller than that with IRS shown among those who were overweight in the CARDIA study (21%). Further, the association in the CARDIA study was present with both high- and low-fat dairy products but in our study was limited to low-fat dairy intake. It is possible that a strong inverse association between dairy intake and risk of IRS in young adulthood in the CARDIA study may not directly relate to risk of type 2 diabetes later in life observed in these middle-aged or older women. Aside from the fact that type 2 diabetes and IRS are two related but different end points, age differences (at the time of dairy exposure or outcome measurement) and dilution of effect in the later phases of the causal pathway toward type 2 diabetes may potentially explain these different observations.

The exact mechanisms underlying the inverse association between dairy intake and risk of type 2 diabetes remain unclear. Milk proteins appear to have insulinotropic properties with a relatively low glycemic load (6). Milk proteins appear to induce rapid release of insulinotropic amino acids and incretin hormones (insulinotropic peptide hormones from the gut including glucose-dependent insulinotropic polypeptide and glucagon-like peptide) (18). Whey appears to be a particularly potent stimulus for the secretion of glucagonlike peptide (18). The ability of milk to enhance the secretion of insulinotropic amino acids and incretin hormone may help reduce the incidence of type 2 diabetes in those with high intakes of dairy foods. While a lower glycemic index or load associated with dairy intake may also contribute to the reduced risk, our results showed independent associations after adjusting for glycemic load. These data are consistent with the possibility that milk seems to influence glucose tolerance more through its insulinotropic effect than its relatively lower glycemic load. However, saturated fat contained in dairy foods may mitigate these potential benefits, which may explain a weaker or null association with high-dairy foods observed in our study.

Major components in dairy products such as magnesium, calcium, lactose, and dairy protein may enhance satiety and reduce the risk of overweight and obesity compared with other high-carbohydrate foods and beverages. There was no evidence in this study that BMI varied with milk intake, and adjusting for BMI did not alter the dairy food–type 2 diabetes relation. Low-fat dairy foods have been one of the main components of the Dietary Approaches to Stop Hypertension (DASH) diet that has been shown to substantially lower blood pressure (19) and were included in a randomized trial for type 2 diabetes prevention (1).

Our study findings may be affected by unmeasured confounding factors, although our study minimized the possibility of biases due to recall and/or loss of follow-up. We have also adjusted for the known risk factors of type 2 diabetes in our multivariate analyses. Thus, our results were independent of these factors that are correlated with dairy consumption at baseline. Underdiagnosis of type 2 diabetes may be a concern because the study population was not screened for glucose tolerance and the diagnosis was self-reported. However, our validation study of self-reported diabetes in the WHS indicated a high rate of agreement with medical record review. All participants in this study are health professionals, who are likely to have more robust and valid self-reported diagnostic information. The relatively high screening rate of WHS health professionals (85–90% for blood glucose screening) also reduces the likelihood of underdiagnosis. Thus, as compared with the general population, the degree of underdiagnosis was probably smaller in this cohort of health professionals with ready access to medical care.

In conclusion, higher intake of dairy foods was prospectively associated with lower risk of type 2 diabetes in women. These data indicate that a diet that incorporates higher dairy intake, especially low-fat dairy intake, may lower the risk of type 2 diabetes among middle-aged or elderly women. Further confirmation of these findings and comprehensive risk-benefit assessments, however, are required before public health measures to increase dairy consumption can be recommended for prevention of type 2 diabetes.

This study was supported by grants DK66401, CA-47988, and HL-43851 from the National Institutes of Health, Bethesda, MD.

We are indebted to the 39,876 dedicated and committed participants of the WHS.