Abstract

The authors analyzed data from a prospective, community-based cohort to assess the risk of incident type 2 diabetes mellitus associated with coffee and sweetened beverage consumption. They included 12,204 nondiabetic, middle-aged men and women in the Atherosclerosis Risk in Communities (ARIC) Study (1987–1999). Consumption of each beverage was assessed by food frequency questionnaire and classified into categories of cups per day. They found an inverse association, after adjusting for potential confounders, between increased coffee consumption and risk of type 2 diabetes mellitus in men (for ≥4 cups (≥0.95 liter)/day compared with almost never: hazard ratio = 0.77, ptrend = 0.02) with no significant association in women (hazard ratio = 0.89, ptrend = 0.32) using a combination of self-report of physician-diagnosed diabetes, diabetes treatment, and a fasting or nonfasting blood glucose test. When self-reported diabetes or diabetes treatment alone was used, a stronger and significant inverse association was seen in men and women. Sweetened beverage consumption (men: hazard ratio = 1.03, ptrend = 0.94; women: hazard ratio = 1.01, ptrend = 0.58) showed no consistent association with the incidence of type 2 diabetes mellitus. In summary, increased coffee consumption was significantly associated with a decreased risk of diagnosed type 2 diabetes mellitus in community-based US adults.

Of the approximately 18 million people with diabetes in the United States, the majority have type 2 diabetes mellitus (1). The identification of modifiable risk factors is of paramount importance. To date, obesity and physical activity are the only well-established modifiable risk factors (2–4). Nutrition and diet have received considerable attention with mixed results. One potential dietary component is nonalcoholic beverage consumption, including coffee and sweetened beverages. If these nonalcoholic beverages were found to be related to the development of type 2 diabetes, the potential effect of dietary modification on diabetes risk could be quite large because of the widespread consumption of these beverages in the general population.

Previous prospective studies of coffee consumption have shown mixed results (5–13). A series of studies showing an inverse relation (5, 7, 9–13) have been reported, paralleled by those showing no association (6, 8), with a recent meta-analysis showing an overall inverse association (14). The studies have varied in consumption levels, as well as in assessment of incident type 2 diabetes and the measurement of important covariates, such as body mass index. There have also been substantial variations by geography, race, and gender in the populations studied. Overall, studies with reported physician diagnosis- or medical record-based outcomes have tended to find a stronger association than those with a screening test. To date, no large population-based, multiethnic US study has been conducted.

Sweetened beverage consumption has been shown to be associated with obesity incidence in children (15, 16), obesity prevalence cross-sectionally in young adults (17), and temporally with national obesity rates (18). Moreover, the association between sweetened beverage consumption (defined as sugar-sweetened soft drinks) and type 2 diabetes risk, as well as a relation between change in consumption and change in body mass index, was recently reported in the Nurses' Health Study (19).

We examined the association between coffee and sweetened beverage consumption in a US community-based cohort of middle-aged Black and White men and women. Our study addresses the question of whether the incidence of type 2 diabetes is related to consumption of the two beverages, whether these associations are independent of obesity and other potential confounders, and whether the relation is influenced by the method of ascertainment of diabetic status.

MATERIALS AND METHODS

Study population

The Atherosclerosis Risk in Communities (ARIC) Study is an ongoing, prospective study that examines clinical and subclinical atherosclerotic diseases in a cohort of 15,792 persons, aged 45–64 years at baseline examination, selected by probability sampling from four US communities: Forsyth County, North Carolina; Jackson, Mississippi; the northwest suburbs of Minneapolis, Minnesota; and Washington County, Maryland (20). Our analysis was based on a maximum of 9 years of follow-up, which included three clinic visits scheduled 3 years apart. For this analysis, we excluded participants who reported ethnicity other than Black or White (n = 48), had diabetes at baseline (n = 1,860), had missing exposure or diabetes information (n = 134), and had missing information on any of the potential confounders (n = 683), as well as individuals who did not return after the baseline visit (n = 814). Participants who did not return after the baseline visit were slightly more likely to drink 1 cup (0.24 liter) of coffee per day and more than 2 cups (0.48 liter) of sweetened beverages. After exclusions, 12,204 participants remained for this analysis.

Exposure assessment

An interviewer-administered, modified version of the 61-item food frequency questionnaire developed by Willett et al. (21) was used to assess food and beverage consumption during the previous 12 months. Questions about intakes of beer, wine, and spirits were included separately. Participants were asked how often they had consumed 1 cup (8 ounces or 0.24 liter) of each of the beverages, and they could choose from nine responses (almost never, 1–3 per month, 1 per week, 2–4 per week, 5–6 per week, 1 per day, 2–3 per day, 4–5 per day, and ≥6 per day). Consumption of cups of coffee (caffeinated only) was analyzed using the following collapsed categories: almost never, <1 per day, 1 per day, 2–3 per day, and 4 or more per day. To create the sweetened beverage category, we combined the consumptions of fruit punch, nondiet soda, and orange or grapefruit juice using the midpoint daily consumption for the nine possible response categories.

Covariate assessment

Information on age, gender, race, family history of diabetes, smoking status, alcohol intake, and education was obtained from interviews conducted at the baseline visit (22). Body mass index (weight (kg)/height (m)2) and the waist:hip ratio were determined by the anthropometric measurements taken at the baseline clinic visit. Measurements were made with the participants wearing light-weight, nonconstricting clothing and no shoes following standard protocol. Physical activity was assessed using a modified interviewer-administered version of Baecke's questionnaire (23) and characterized as a leisure activity index with scores ranging from 1 to 5 (1 indicating least active). Total caloric intake and percent of diet composed of fiber were derived from the same food frequency questionnaire used in exposure assessment (21).

Outcome assessment

Type 2 diabetes mellitus was defined as the presence of any one of the following: 1) fasting glucose of ≥126 mg/dl, 2) nonfasting glucose of ≥200 mg/dl, 3) current use of hypoglycemic medication, or 4) a positive response to the question, “Has a doctor ever told you that you had diabetes (sugar in the blood)?” Those with diabetes or unknown diabetes status at baseline were excluded from the prospective analyses. Individuals without diabetes at baseline who met any of these conditions during follow-up were considered incident cases. All incident cases were classified as type 2 diabetes, because the age of onset in this cohort was between 45 and 70 years.

Alternate outcome definition

An alternate definition of type 2 diabetes was also developed for comparison with studies where diabetes was measured by self-report. For this purpose, type 2 diabetes was defined as either a positive response to the question, “Has a doctor ever told you that you had diabetes (sugar in the blood)?”, or current use of hypoglycemic medication. This criterion was then applied to each visit to generate the study population for the alternate analysis. Individuals with no self-report of diagnosed diabetes but with a high serum glucose level were considered nondiabetic until they reported a diagnosis or medication use. The population available for this analysis consisted of 12,747 people with 913 cases of incident self-reported diagnoses of type 2 diabetes.

Statistical analysis

All analyses were stratified by gender both for comparison with previous studies and because the confounding effect of body mass index was different for the two groups. Incidence rate ratios were calculated for each consumption group by use of a person-years approach. For the standard type 2 diabetes definition, the date of diabetes incidence was estimated by linear interpolation using glucose values at the ascertaining visit and the previous one, as previously described (24). For the self-reported definition, incident cases were assigned to the midpoints between the last visit where they reported no physician-diagnosed diabetes or medication use and the first visit in which they reported either of the two.

Multivariate analysis was done using Cox proportional hazards models, and hazard ratios and corresponding 95 percent confidences intervals were estimated. Tests for trend were done using consumption categories as ordinal variables. For coffee, three models were assessed: 1) adjusted for study center, age, race, education, and family history of diabetes; 2) model 1 additionally adjusted for body mass index by quartile, waist:hip ratio quartile, total caloric intake quartiles, dietary fiber quartiles, smoking, alcohol consumption (never, former, current), leisure activity, and hypertension; and 3) model 2 without adjustment for body mass index or waist:hip ratio. An additional model for coffee consumption added serum magnesium, for its putative mediation role, to the fully adjusted model 2 above. For sweetened beverages, because body mass index and the waist:hip ratio were thought to mediate rather than confound the relation with incident type 2 diabetes, the following three models were assessed: 1) same as model 1 for coffee; 2) model 1 additionally adjusted for total caloric intake, dietary fiber, smoking, alcohol consumption, leisure activity, and hypertension; and 3) model 2 additionally adjusted for body mass index and waist:hip ratio quartiles.

All analyses were also performed using the alternate outcome of self-reported diabetes. All analyses were done using STATA, version 8, software (25).

RESULTS

Baseline characteristics and beverage consumption patterns

A total of 5,414 men and 6,790 women from the ARIC Study were included in the main analyses. The average age at baseline of both men and women was 54 years. Approximately 18 percent of the men and 25 percent of the women were African American. Both men and women had a mean body mass index of 27.2 kg/m2, with a mean waist:hip ratio of 0.96 in the men and 0.89 in the women. Of the men, 29 percent reported never smoking at baseline compared with 53 percent never smokers in the women. As shown in table 1, drinking patterns differed both by beverage and across gender. For both coffee and sweetened beverages, there were more men in the highest consumption category and more women in the lowest. The median coffee consumption was 1 cup (0.24 liter) per day for both men and women.

TABLE 1

Incidence of type 2 diabetes mellitus by consumption level and type of beverage, follow-up in person-years, and adjusted hazard ratio* and 95% confidence interval by gender, Atherosclerosis Risk in Communities Study, 1987–1999

 No. Cases Person-years Adjusted hazard ratio* 95% confidence interval ptrend 
Coffee (cupsper day
Men        
    Almost never 1,183 22 167 8,795 Referent 0.09 
    <1 780 14 109 5,866 1.00 0.78, 1.27  
    1 958 18 152 7,073 1.12 0.90, 1.39  
    2–3 1,372 25 165 10,378 0.88 0.71, 1.09  
    ≥4 1,121 21 125 8,414 0.84 0.66, 1.06  
Women        
    Almost never 1,968 29 238 14,909 Referent 0.24 
    <1 908 13 90 6,977 0.89 0.70, 1.14  
    1 1,276 19 151 9,526 0.94 0.77, 1.15  
    2–3 1,554 23 150 12,082 0.89 0.72, 1.09  
    ≥4 1,084 16 90 8,546 0.88 0.68, 1.13  
Sweet beverages (cups per day
Men        
    <1 2,557 47 331 19,205 Referent 0.68 
    1 504 67 3,706 1.03 0.80, 1.35  
    1.1–1.9 1,415 26 182 10,665 0.96 0.80, 1.15  
    ≥2.0 938 17 138 6,892 1.09 0.89, 1.33  
Women        
    <1 3,510 52 320 27,438 Referent 0.05 
896 13 103 6,815 1.21 0.97, 1.51  
    1.1–1.9 1,490 22 182 11,255 1.20 1.00, 1.44  
    ≥2.0 894 13 114 6,533 1.17 0.94, 1.46  
 No. Cases Person-years Adjusted hazard ratio* 95% confidence interval ptrend 
Coffee (cupsper day
Men        
    Almost never 1,183 22 167 8,795 Referent 0.09 
    <1 780 14 109 5,866 1.00 0.78, 1.27  
    1 958 18 152 7,073 1.12 0.90, 1.39  
    2–3 1,372 25 165 10,378 0.88 0.71, 1.09  
    ≥4 1,121 21 125 8,414 0.84 0.66, 1.06  
Women        
    Almost never 1,968 29 238 14,909 Referent 0.24 
    <1 908 13 90 6,977 0.89 0.70, 1.14  
    1 1,276 19 151 9,526 0.94 0.77, 1.15  
    2–3 1,554 23 150 12,082 0.89 0.72, 1.09  
    ≥4 1,084 16 90 8,546 0.88 0.68, 1.13  
Sweet beverages (cups per day
Men        
    <1 2,557 47 331 19,205 Referent 0.68 
    1 504 67 3,706 1.03 0.80, 1.35  
    1.1–1.9 1,415 26 182 10,665 0.96 0.80, 1.15  
    ≥2.0 938 17 138 6,892 1.09 0.89, 1.33  
Women        
    <1 3,510 52 320 27,438 Referent 0.05 
896 13 103 6,815 1.21 0.97, 1.51  
    1.1–1.9 1,490 22 182 11,255 1.20 1.00, 1.44  
    ≥2.0 894 13 114 6,533 1.17 0.94, 1.46  
*

Hazard ratios are adjusted for race and age.

One cup = 0.24 liter.

Relation of beverage consumption to type 2 diabetes risk

Over the follow-up interval, 1,437 participants developed incident diabetes (718 men and 719 women). Type 2 diabetes incidence rates differed among the six coffee consumption groups in both men and women (table 1). Although no statistically significant trend was seen in the age- and race-adjusted hazard ratio, in both men and women the groups with highest consumption had the lowest risk. For men, the crude incidence rates per 1,000 person-years from lowest coffee consumption to highest were 18.99, 18.46, 21.36, 15.84, and 14.78. For women, the crude incidence rates per 1,000 person-years were 15.86, 12.84, 15.70, 12.35, and 10.51, respectively. Sweetened beverage consumption showed a trend of increased risk with increased consumption in women and no association in men. Removing juice from the sweetened beverage definition did not affect the results.

Association between coffee consumption and potential confounders and mediators of type 2 diabetes

In men, coffee consumption was strongly related to many characteristics previously found to be associated with incident type 2 diabetes in the ARIC Study cohort (table 2). Race, smoking status, educational level, alcohol consumption, hypertension, and energy intake were significantly associated with coffee consumption, with White participants and smokers drinking more coffee. In men, lower intake of fiber and younger age were also significantly associated with higher coffee consumption, while body mass index, obesity, and the waist:hip ratio were not. Serum magnesium levels also increased with increased coffee consumption, as expected. Higher coffee intake was also associated with lower serum insulin levels but showed no association with serum glucose.

TABLE 2

Selected baseline characteristics (mean and standard deviation or %) of men (n = 5,414) and women (n = 6,790) by coffee consumption, Atherosclerosis Risk in Communities Study, 1987–1989

 Cups† of coffee per day 
 Men Women 
 Almost never <1 2–3 ≥4 ptrend Almost never <1 2–3 ≥4 ptrend 
No. 1,183 780 958 1,372 1,121  1,968 908 1,276 1,554 1,084  
Age, years 54.4 (5.8) 54.7 (5.7) 55.1 (5.8) 54.3 (5.7) 53.6 (5.4) <0.01 53.3 (5.8) 53.4 (5.6) 54.0 (5.9) 53.7 (5.5) 53.2 (5.4) 0.36 
Black, % 27 21 28 13 <0.01 32 23 37 17 <0.01 
Currently smoking, % 20 19 24 26 40 <0.01 17 18 21 26 44 <0.01 
Body mass index, kg/m2 27.3 (4.2) 27.4 (3.9) 27.3 (4.1) 27.1 (3.9) 27.1 (3.9) 0.06 27.8 (6.1) 27.3 (5.6) 27.9 (6.0) 26.6 (5.4) 25.8 (4.7) <0.01 
Obesity (≥30 kg/m2), % 21 23 22 19 21 0.30 30 27 31 22 17 <0.01 
Waist:hip ratio 0.957 (0.055) 0.962 (0.051) 0.957 (0.054) 0.960 (0.052) 0.963 (0.054) 0.06 0.890 (0.080) 0.887 (0.078) 0.896 (0.080) 0.883 (0.077) 0.876 (0.080) <0.01 
Less than high school education, % 23 19 29 17 16 <0.01 20 18 25 17 17 0.03 
Family history of diabetes, % 19 20 23 23 21 0.09 24 22 23 24 24 0.73 
Consuming alcohol, % 58 68 64 73 73 <0.01 43 55 46 60 68 <0.01 
Hypertensive, % 35 32 37 27 22 <0.01 35 31 35 27 20 <0.01 
Leisure activity index 2.39 (0.56) 2.37 (0.55) 2.31 (0.57) 2.38 (0.54) 2.35 (0.52) 0.32 2.39 (0.59) 2.47 (0.58) 2.36 (0.58) 2.44 (0.56) 2.47 (0.57) <0.01 
Energy intake, kcal/day 1,711 (636) 1,723 (602) 1,779 (631) 1,772 (631) 1,854 (668) <0.01 1,470 (537) 1,461 (529) 1,498 (530) 1,496 (524) 1,586 (546) <0.01 
Fiber, g/day 18.0 (8.8) 17.6 (8.0) 17.7 (8.1) 17.1 (8.1) 17.0 (7.9) <0.01 17.3 (8.7) 16.9 (7.5) 17.1 (8.2) 17.0 (7.6) 17.0 (7.7) 0.29 
Serum magnesium, milliequivalent/liter 1.64 (0.16) 1.65 (0.16) 1.64 (0.15) 1.66 (0.14) 1.67 (0.14) <0.01 1.63 (0.15) 1.64 (0.16) 1.64 (0.15) 1.64 (0.15) 1.66 (0.14) <0.01 
Serum insulin, pmol/liter† 82 (61) 81 (54) 86 (66) 78 (52) 73 (49) <0.01 81 (68) 78 (56) 86 (69) 69 (46) 61 (43) <0.01 
Blood glucose, mmol/liter† 5.58 (0.50) 5.58 (0.51) 5.61 (0.53) 5.58 (0.49) 5.58 (0.49) 0.72 5.40 (0.52) 5.38 (0.50) 5.42 (0.52) 5.38 (0.48) 5.35 (0.48) 0.03 
 Cups† of coffee per day 
 Men Women 
 Almost never <1 2–3 ≥4 ptrend Almost never <1 2–3 ≥4 ptrend 
No. 1,183 780 958 1,372 1,121  1,968 908 1,276 1,554 1,084  
Age, years 54.4 (5.8) 54.7 (5.7) 55.1 (5.8) 54.3 (5.7) 53.6 (5.4) <0.01 53.3 (5.8) 53.4 (5.6) 54.0 (5.9) 53.7 (5.5) 53.2 (5.4) 0.36 
Black, % 27 21 28 13 <0.01 32 23 37 17 <0.01 
Currently smoking, % 20 19 24 26 40 <0.01 17 18 21 26 44 <0.01 
Body mass index, kg/m2 27.3 (4.2) 27.4 (3.9) 27.3 (4.1) 27.1 (3.9) 27.1 (3.9) 0.06 27.8 (6.1) 27.3 (5.6) 27.9 (6.0) 26.6 (5.4) 25.8 (4.7) <0.01 
Obesity (≥30 kg/m2), % 21 23 22 19 21 0.30 30 27 31 22 17 <0.01 
Waist:hip ratio 0.957 (0.055) 0.962 (0.051) 0.957 (0.054) 0.960 (0.052) 0.963 (0.054) 0.06 0.890 (0.080) 0.887 (0.078) 0.896 (0.080) 0.883 (0.077) 0.876 (0.080) <0.01 
Less than high school education, % 23 19 29 17 16 <0.01 20 18 25 17 17 0.03 
Family history of diabetes, % 19 20 23 23 21 0.09 24 22 23 24 24 0.73 
Consuming alcohol, % 58 68 64 73 73 <0.01 43 55 46 60 68 <0.01 
Hypertensive, % 35 32 37 27 22 <0.01 35 31 35 27 20 <0.01 
Leisure activity index 2.39 (0.56) 2.37 (0.55) 2.31 (0.57) 2.38 (0.54) 2.35 (0.52) 0.32 2.39 (0.59) 2.47 (0.58) 2.36 (0.58) 2.44 (0.56) 2.47 (0.57) <0.01 
Energy intake, kcal/day 1,711 (636) 1,723 (602) 1,779 (631) 1,772 (631) 1,854 (668) <0.01 1,470 (537) 1,461 (529) 1,498 (530) 1,496 (524) 1,586 (546) <0.01 
Fiber, g/day 18.0 (8.8) 17.6 (8.0) 17.7 (8.1) 17.1 (8.1) 17.0 (7.9) <0.01 17.3 (8.7) 16.9 (7.5) 17.1 (8.2) 17.0 (7.6) 17.0 (7.7) 0.29 
Serum magnesium, milliequivalent/liter 1.64 (0.16) 1.65 (0.16) 1.64 (0.15) 1.66 (0.14) 1.67 (0.14) <0.01 1.63 (0.15) 1.64 (0.16) 1.64 (0.15) 1.64 (0.15) 1.66 (0.14) <0.01 
Serum insulin, pmol/liter† 82 (61) 81 (54) 86 (66) 78 (52) 73 (49) <0.01 81 (68) 78 (56) 86 (69) 69 (46) 61 (43) <0.01 
Blood glucose, mmol/liter† 5.58 (0.50) 5.58 (0.51) 5.61 (0.53) 5.58 (0.49) 5.58 (0.49) 0.72 5.40 (0.52) 5.38 (0.50) 5.42 (0.52) 5.38 (0.48) 5.35 (0.48) 0.03 

* One cup = 0.24 liter.

On fasting participants only (men: n = 5,302; women: n = 6,689).

Table 2 also shows the relation of coffee consumption to the same set of factors in women. Race, smoking status, educational level, alcohol consumption, hypertension, and energy intake are significantly associated with coffee consumption in women. Lower body mass index, waist:hip ratio, and higher leisure activity were associated with higher coffee consumption in women, while age and fiber intake were not. In women, higher coffee intake was also associated with lower levels of serum insulin and glucose at baseline.

Type 2 diabetes risk and coffee consumption

The association between type 2 diabetes and coffee was not significant in any single category and did not show in the minimal model a significant inverse trend when minimally adjusted (table 3). However, after adjustment for potential confounders, men showed a significant inverse association (ptrend = 0.02). Compared with men who did not drink coffee, those who drank 4 or more cups (0.95 liter) per day were less likely to develop diabetes (relative hazard = 0.77, 95 percent confidence interval (CI): 0.61, 0.98). The corresponding relative hazard for women was 0.89 (95 percent CI: 0.69, 1.15) after full adjustment, and the trend was nonsignificant (ptrend = 0.32).

TABLE 3

Risk of incident type 2 diabetes mellitus by level of coffee consumption at baseline, Atherosclerosis Risk in Communities Study, 1987–1999

Covariate adjustment Cups* of coffee per day ptrend 
Almost never <1 2–3 ≥4 
Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval 
Men            
    Minimal† Referent 0.99 0.78, 1.26 1.09 0.87, 1.36 0.85 0.69, 1.06 0.83 0.65, 1.05 0.06 
    Full‡ Referent 0.94 0.74, 1.20 1.04 0.83, 1.30 0.82 0.66, 1.03 0.77 0.61, 0.98 0.02 
    Full‡ – body mass index and waist:hip ratio Referent 0.98 0.78, 1.26 1.01 0.81, 1.27 0.82 0.66, 1.02 0.77 0.61, 0.99 0.02 
    Full‡ + serum magnesium Referent 0.94 0.74, 1.20 1.04 0.83, 1.30 0.83 0.66, 1.03 0.77 0.61, 0.99 0.02 
    Full‡ using diagnosis/medications§ Referent 1.02 0.76, 1.38 1.01 0.76, 1.34 0.93 0.70, 1.22 0.69 0.50, 0.96 0.04 
Women            
    Minimal† Referent 0.90 0.70, 1.14 0.94 0.76, 1.15 0.88 0.72, 1.08 0.85 0.66, 1.09 0.16 
    Full‡ Referent 0.91 0.71, 1.16 0.90 0.73, 1.10 0.92 0.75, 1.14 0.89 0.69, 1.15 0.32 
    Full‡ – body mass index and waist:hip ratio Referent 0.91 0.71, 1.16 0.94 0.77, 1.15 0.88 0.71, 1.08 0.83 0.64, 1.08 0.14 
    Full‡ + serum magnesium Referent 0.91 0.71, 1.15 0.90 0.73, 1.11 0.92 0.75, 1.14 0.91 0.70, 1.18 0.38 
    Full‡ using diagnosis/medications§ Referent 1.23 0.93, 1.62 1.06 0.83, 1.36 0.89 0.68, 1.16 0.66 0.45, 0.96 0.046 
Covariate adjustment Cups* of coffee per day ptrend 
Almost never <1 2–3 ≥4 
Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval 
Men            
    Minimal† Referent 0.99 0.78, 1.26 1.09 0.87, 1.36 0.85 0.69, 1.06 0.83 0.65, 1.05 0.06 
    Full‡ Referent 0.94 0.74, 1.20 1.04 0.83, 1.30 0.82 0.66, 1.03 0.77 0.61, 0.98 0.02 
    Full‡ – body mass index and waist:hip ratio Referent 0.98 0.78, 1.26 1.01 0.81, 1.27 0.82 0.66, 1.02 0.77 0.61, 0.99 0.02 
    Full‡ + serum magnesium Referent 0.94 0.74, 1.20 1.04 0.83, 1.30 0.83 0.66, 1.03 0.77 0.61, 0.99 0.02 
    Full‡ using diagnosis/medications§ Referent 1.02 0.76, 1.38 1.01 0.76, 1.34 0.93 0.70, 1.22 0.69 0.50, 0.96 0.04 
Women            
    Minimal† Referent 0.90 0.70, 1.14 0.94 0.76, 1.15 0.88 0.72, 1.08 0.85 0.66, 1.09 0.16 
    Full‡ Referent 0.91 0.71, 1.16 0.90 0.73, 1.10 0.92 0.75, 1.14 0.89 0.69, 1.15 0.32 
    Full‡ – body mass index and waist:hip ratio Referent 0.91 0.71, 1.16 0.94 0.77, 1.15 0.88 0.71, 1.08 0.83 0.64, 1.08 0.14 
    Full‡ + serum magnesium Referent 0.91 0.71, 1.15 0.90 0.73, 1.11 0.92 0.75, 1.14 0.91 0.70, 1.18 0.38 
    Full‡ using diagnosis/medications§ Referent 1.23 0.93, 1.62 1.06 0.83, 1.36 0.89 0.68, 1.16 0.66 0.45, 0.96 0.046 
*

One cup = 0.24 liter.

Adjusted for age, race, education, and family history of diabetes.

Adjusted for age, race, education, family history of diabetes, body mass index, waist:hip ratio, total caloric intake, dietary fiber, smoking, alcohol consumption, leisure activity, and hypertension.

§

Uses diagnosis or medication use only for outcome classification.

Because body mass index and the waist:hip ratio are potential mediators as well as confounders, we also examined the full model without adjusting for body mass index and the waist:hip ratio and found it to have little change on the relative hazard estimates within each category. Stratifying by obesity status (body mass index: ≥30 kg/m2) did not show a clear pattern of modification, and tests for interaction were not significant. In addition, further adjustment for either serum magnesium or fasting serum insulin and glucose levels (data not shown) did not affect the association in both men and women. Restricting the analysis to only the Caucasian participants did not affect the results.

When only the self-reported incident cases were used in order to correspond to previous literature, a significant inverse association was observed (ptrend = 0.046), as well as a significant relative hazard of 0.66 (95 percent CI: 0.45, 0.96) for women who drank 4 or more cups (0.95 liter) per day compared with those who drank none. This increase in the strength of association was also observed in men when using the alternative criteria for diagnosed diabetes as the outcome (ptrend = 0.04), as well as a decrease in the relative hazard to 0.69 for men who drank 4 or more cups per day compared with those who drank none.

Multivariate analysis of type 2 diabetes risk and sweetened beverage consumption

Next, we examined the relation between sweetened beverage consumption and type 2 diabetes risk using the minimal and full models previously described. No pattern of association was seen in either men or women (table 4). Adjustment for body mass index and the waist:hip ratio did not affect the results, nor did use of the alternative criteria for diabetes of self-reported physician's diagnosis or medication use. Stratification by age, body mass index, or physical activity did not affect the results, nor did removing juice from the definition.

TABLE 4

Risk of incident type 2 diabetes mellitus by level of sweetened beverage consumption at baseline, Atherosclerosis Risk in Communities Study, 1987–1999

 Cups* of sweetened beverages per day ptrend 
 <1 1.1–1.9 ≥2.0 
 Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval 
Men          
    Full† – body mass index and waist:hip ratio Referent 1.03 0.79, 1.34 0.95 0.79, 1.15 1.03 0.82, 1.28 0.94 
    Full† – body mass index and waist:hip ratio using diagnosis/medications‡ Referent 1.09 0.79, 1.52 1.06 0.84, 1.34 1.02 0.76, 1.36 0.76 
Women          
    Full† – body mass index and waist:hip ratio Referent 1.13 0.91, 1.42 1.10 0.91, 1.33 1.01 0.79, 1.29 0.58 
    Full† – body mass index and waist:hip ratio using diagnosis/medications‡ Referent 0.77 0.56, 1.06 1.05 0.83, 1.32 1.07 0.79, 1.43 0.63 
 Cups* of sweetened beverages per day ptrend 
 <1 1.1–1.9 ≥2.0 
 Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval Hazard ratio 95% confidence interval 
Men          
    Full† – body mass index and waist:hip ratio Referent 1.03 0.79, 1.34 0.95 0.79, 1.15 1.03 0.82, 1.28 0.94 
    Full† – body mass index and waist:hip ratio using diagnosis/medications‡ Referent 1.09 0.79, 1.52 1.06 0.84, 1.34 1.02 0.76, 1.36 0.76 
Women          
    Full† – body mass index and waist:hip ratio Referent 1.13 0.91, 1.42 1.10 0.91, 1.33 1.01 0.79, 1.29 0.58 
    Full† – body mass index and waist:hip ratio using diagnosis/medications‡ Referent 0.77 0.56, 1.06 1.05 0.83, 1.32 1.07 0.79, 1.43 0.63 
*

One cup = 0.24 liter.

Adjusted for age, race, education, family history of diabetes, total caloric intake, dietary fiber, smoking, alcohol consumption, leisure activity, and hypertension.

Uses diagnosis or medication use only for outcome classification.

DISCUSSION

This study examined the risk of type 2 diabetes associated with nonalcoholic beverage consumption in a US community-based group of middle-aged African-American and Caucasian adults. This cohort reflects the distribution of consumption patterns in four communities with moderate consumption of coffee and sweetened beverages. High coffee intake was inversely associated with type 2 diabetes risk after adjustment for multiple potential confounders and mediators, including serum magnesium level. On the other hand, there was no association between sweetened beverage consumption and diabetes risk. Of interest, the inverse association between coffee consumption and diabetes risk was most evident for diabetes defined by self-report of a physician's diagnosis or current use of hypoglycemic medications. For both men and women, individuals who drank more than 4 cups (0.95 liter) of coffee per day had a risk of developing type 2 diabetes that was about 67 percent of that of their counterparts who drank no coffee.

Previous prospective studies of coffee consumption and the risk of type 2 diabetes have shown mixed results (5–10). Some differences in results may be attributable to differences in diabetes ascertainment, assessment of covariates, and study populations. As shown in our results, the definition of diabetes has a substantial impact on the strength of the association, especially in women. Differences in the measurement of and adjustment for obesity-related traits could also have led to differences in results, as suggested by the effect of adjustment for body mass index in our study. Finally, different coffee consumption patterns may create different associations with diabetes among studies.

In light of these differences, results from this and previous prospective studies have, in fact, been fairly consistent regarding the association between coffee consumption and type 2 diabetes risk, as summarized in table 5. Although the recent meta-analysis by van Dam and Hu (14) presented an overall picture of the research, we focus instead on placing our study into the context of the literature. Although variation in outcome ascertainment and reference groups makes direct comparisons difficult, two qualitative trends in the studies emerge: 1) higher consumption of coffee seems to associate with decreased hazard of diabetes; and 2) measurement of disease using clinically diagnosed diabetes or medication use is associated with stronger associations. Our highest category (≥4 cups (≥0.95 liter) per day), using the self-reported criteria, showed similar hazard ratios to those in comparable groups in other populations.

TABLE 5

Summary of Atherosclerosis Risk in Communities Study (1987–1999) results in the context of prospective studies to date

First author, year (reference) Population Measure of type 2 diabetes Highest/lowest category of coffee consumption (cups*/day) Hazard ratio for highest to lowest categories in study Hazard ratio for closest categories to ≥4 cups/day in study 
Current study Men and women, ARIC† Study, United States Fasting blood glucose or self-report of physician's diagnosis or medication use ≥4/almost never Men: fasting blood glucose and self-report; hazard ratio = 0.81 Women: fasting blood glucose and self-report; hazard ratio = 0.92 
  Self-report of physician's diagnosis or medication use ≥4/almost never Men: self-report only; hazard ratio = 0.69 Women: self-report only; hazard ratio = 0.66 
van Dam, 2006 (13Women, Nurses' Health Study II, United States Self-report of physician's diagnosis with follow-up questionnaire ≥4/none Hazard ratio = 0.61 Hazard ratio = 0.61 
van Dam, 2004 (11Men and women, Hoorn Study, Netherlands Oral glucose tolerance test, blood glucose ≥7/≤2 Hazard ratio = 0.69 3–4 cups/day—hazard ratio = 0.94 5–6 cups/day—hazard ratio = 0.92 
Carlsson, 2004 (5Men and women, Finnish Twin Cohort Drug registry and hospital discharge ≥7/≤2 Hazard ratio = 0.65 3–4 cups/day—hazard ratio = 0.70 5–6 cups/day—hazard ratio = 0.71 
Tuomilehto, 2004 (9Men and women, Finland Drug registry and hospital discharge ≥10/≤2 Men: hazard ratio = 0.45 Women: hazard ratio = 0.21 3–4 cups/day Men: hazard ratio = 0.73 Women: hazard ratio = 0.71 
     5–6 cups/day Men: hazard ratio = 0.70 Women: hazard ratio = 0.39 
Salazar-Martinez, 2004 (7Men and women, Nurses' Health Study and Health Professionals Follow-up Study, United States Self-report of physician's diagnosis with follow-up questionnaire ≥6/none Men: hazard ratio = 0.46 Women: hazard ratio = 0.71 4–5 cups/day Men: hazard ratio = 0.71 Women: hazard ratio = 0.70 
Rosengren, 2004 (12Women (BEDA‡), Göteborg, Sweden Blood or urine glucose, self-report of physician's diagnosis, hospital discharge >6/≤2 Hazard ratio = 0.57 3–4 cups/day—hazard ratio = 0.56 5–6 cups/day—hazard ratio = 0.45 
Saremi, 2003 (8Men and women, Pima Indians, United States Oral glucose tolerance test “Occasionally heavy”/none Hazard ratio = 0.81 Hazard ratio = 0.81 
Reunanen, 2003 (6Men and women, Finland Drug registry ≥7/≤2 Hazard ratio = 0.92 3–4 cups/day—hazard ratio = 1.01 5–6 cups/day—hazard ratio = 0.98 
van Dam, 2002 (10Men and women, Netherlands Self-report of diagnosis or drug use ≥7/≤2 Hazard ratio = 0.50 3–4 cups/day—hazard ratio = 0.79 5–6 cups/day—hazard ratio = 0.73 
First author, year (reference) Population Measure of type 2 diabetes Highest/lowest category of coffee consumption (cups*/day) Hazard ratio for highest to lowest categories in study Hazard ratio for closest categories to ≥4 cups/day in study 
Current study Men and women, ARIC† Study, United States Fasting blood glucose or self-report of physician's diagnosis or medication use ≥4/almost never Men: fasting blood glucose and self-report; hazard ratio = 0.81 Women: fasting blood glucose and self-report; hazard ratio = 0.92 
  Self-report of physician's diagnosis or medication use ≥4/almost never Men: self-report only; hazard ratio = 0.69 Women: self-report only; hazard ratio = 0.66 
van Dam, 2006 (13Women, Nurses' Health Study II, United States Self-report of physician's diagnosis with follow-up questionnaire ≥4/none Hazard ratio = 0.61 Hazard ratio = 0.61 
van Dam, 2004 (11Men and women, Hoorn Study, Netherlands Oral glucose tolerance test, blood glucose ≥7/≤2 Hazard ratio = 0.69 3–4 cups/day—hazard ratio = 0.94 5–6 cups/day—hazard ratio = 0.92 
Carlsson, 2004 (5Men and women, Finnish Twin Cohort Drug registry and hospital discharge ≥7/≤2 Hazard ratio = 0.65 3–4 cups/day—hazard ratio = 0.70 5–6 cups/day—hazard ratio = 0.71 
Tuomilehto, 2004 (9Men and women, Finland Drug registry and hospital discharge ≥10/≤2 Men: hazard ratio = 0.45 Women: hazard ratio = 0.21 3–4 cups/day Men: hazard ratio = 0.73 Women: hazard ratio = 0.71 
     5–6 cups/day Men: hazard ratio = 0.70 Women: hazard ratio = 0.39 
Salazar-Martinez, 2004 (7Men and women, Nurses' Health Study and Health Professionals Follow-up Study, United States Self-report of physician's diagnosis with follow-up questionnaire ≥6/none Men: hazard ratio = 0.46 Women: hazard ratio = 0.71 4–5 cups/day Men: hazard ratio = 0.71 Women: hazard ratio = 0.70 
Rosengren, 2004 (12Women (BEDA‡), Göteborg, Sweden Blood or urine glucose, self-report of physician's diagnosis, hospital discharge >6/≤2 Hazard ratio = 0.57 3–4 cups/day—hazard ratio = 0.56 5–6 cups/day—hazard ratio = 0.45 
Saremi, 2003 (8Men and women, Pima Indians, United States Oral glucose tolerance test “Occasionally heavy”/none Hazard ratio = 0.81 Hazard ratio = 0.81 
Reunanen, 2003 (6Men and women, Finland Drug registry ≥7/≤2 Hazard ratio = 0.92 3–4 cups/day—hazard ratio = 1.01 5–6 cups/day—hazard ratio = 0.98 
van Dam, 2002 (10Men and women, Netherlands Self-report of diagnosis or drug use ≥7/≤2 Hazard ratio = 0.50 3–4 cups/day—hazard ratio = 0.79 5–6 cups/day—hazard ratio = 0.73 
*

One cup = 0.24 liter.

ARIC, Atherosclerosis Risk in Communities.

“BEDA” is not an acronym but the name of the archetypal Göteborg woman—tough, sharp witted, and with a generous heart.

There are several possible explanations for the difference in the strength of the association between coffee and incident diabetes according to the different diabetes definitions. If there is a true causal association between coffee and diabetes, physician-diagnosed or medically treated diabetes might be a more specific measure that results in less misclassification of diabetes (fewer individuals with false positive diabetes), thus leading to stronger associations. Moreover, if caffeine consumption leads to lower blood glucose levels, then regular coffee drinkers in the ARIC Study who fasted and did not have coffee on the day of the study visit could have experienced an increase in their tested blood glucose as a result of caffeine withdrawal. This would lead to a greater number of false positive diabetic cases, thus weakening the association between coffee consumption and diabetes risk. However, the effect of increased blood glucose as a result of caffeine withdrawal has not been shown in other settings (26). Alternatively, coffee consumption may not be causally associated with the risk of developing diabetes at all. In this case, coffee drinkers could appear to be less likely to have diagnosed diabetes because coffee consumption might either be associated with less health-seeking behavior or mask some symptoms of diabetes, both of which would lead to less diagnosis.

The consistency across populations adds evidence that a causal component of the association between coffee consumption and risk of type 2 diabetes should be considered. Several biologic mechanisms for a protective effect of coffee have been proposed. Coffee consumption is hypothesized to decrease risk of diabetes through its effect on obesity and insulin sensitivity—the two most important risk factors for diabetes. A small clinical study showed that coffee consumption in healthy volunteers can stimulate thermogenesis (27). A study of the effect of caffeinated coffee on energy metabolism in women also indicated that there is a prolonged thermogenic effect of caffeine, especially among lean women (28). Coffee contains caffeine, magnesium, and chlorogenic acid, which may influence glucose metabolism and insulin sensitivity. Although a short-term study has shown that caffeine administration can lead to greater insulin sensitivity (29), others have shown the opposite (30–32). Previous experimental studies have shown that magnesium supplementation can improve glucose metabolism (33–35), while several large epidemiologic studies have shown that low serum/dietary magnesium levels can predict a higher incidence of type 2 diabetes (36, 37). Finally, the chlorogenic acid found in coffee may lead to decreased glucose absorption (38, 39). A recent case-crossover trial of coffee consumption with a 4-week follow-up found that coffee consumption increased fasting insulin, but had no effect on glucose, suggesting that further research on the biologic effects of coffee is necessary (32).

The lack of association between sweetened beverage consumption and type 2 diabetes risk is surprising. However, in the study by Schulze et al. (19), the mean body mass index at baseline was 24.6 kg/m2 and the mean age at baseline was 36.1 years, whereas in the current study the mean body mass index at baseline for women was 27.2 kg/m2 and the mean age at baseline was 53.6 years. One possibility for the differences in results is that the effect of sweetened beverage consumption is mediated largely through increases in body mass index and that, once body mass index is increased, the additional effect of continued sweetened beverage consumption is diminished. Sweetened beverage consumption may be a more important risk factor for younger or leaner people and should be studied further in that population.

Our study does have several limitations. There is potential nondifferential misclassification of beverage consumption, both because the questions were neither as specific nor as sensitive as possible and because of the one-time assessment. We would expect this misclassification to bias our results on average toward a null finding, which might have dampened our results for coffee and masked an association between incident diabetes and sweetened beverages. We also had limited ability to examine the association between coffee and incident diabetes at very high levels of coffee consumption, such as those seen in the studies from Finland and the Netherlands. However, since our cohort was population based, it reflects the actual patterns of coffee drinking in our communities. We also note that there was no measurement of milk, cream, or sugar added to coffee. The effect of these additions might be expected to be positively associated with both increased consumption (the more coffee drunk, the more additions consumed) and increased diabetes, thus acting in opposition to the pattern observed.

In conclusion, a higher consumption level of coffee was associated with a decreased risk of type 2 diabetes in middle-aged adults. Although several observational prospective studies have yielded consistent findings and although biologic explanations for decreased diabetes risk with increased coffee consumption have been postulated, further research, particularly experimental studies that can examine the long-term effect of coffee consumption on glucose metabolism, insulin sensitivity, and diabetes risk, is needed before recommendations can be made about coffee drinking with respect to the prevention of type 2 diabetes.

The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022. W. H. L. K. was supported by grant 1 K01 DK067207-01A1. N. P. P. was supported by Cardiovascular Epidemiology Training Grant T32HL07024 from the National Heart, Lung, and Blood Institute.

The authors thank the staff of the ARIC Study for their important contributions.

Conflict of interest: none declared.

Abbreviations

    Abbreviations
  • ARIC

    Atherosclerosis Risk in Communities

  • CI

    confidence interval

References

1.
Centers for Disease Control and Prevention
National diabetes fact sheet: general information and national estimates on diabetes in the United States, 2002
2003
Atlanta, GA
Centers for Disease Control and Prevention, US Department of Health and Human Services
2.
Kenny
SJ
Aubert
RE
Geiss
LS
National Diabetes Data Group
Prevalence and incidence of non-insulin-dependent diabetes
Diabetes in America
 , 
1995
Bethesda, MD
National Institutes of Health
(pg. 
47
-
68
)
3.
Rewers
M
Hamman
RF
National Diabetes Data Group
Risk factors for non-insulin-dependent diabetes
Diabetes in America
 , 
1995
Bethesda, MD
National Institutes of Health
(pg. 
179
-
220
)
4.
Hu
FB
Manson
JE
Stampfer
MJ
, et al.  . 
Diet, lifestyle, and the risk of type 2 diabetes mellitus in women
N Engl J Med
 , 
2001
, vol. 
345
 (pg. 
790
-
7
)
5.
Carlsson
S
Hammar
N
Grill
V
, et al.  . 
Coffee consumption and risk of type 2 diabetes in Finnish twins. (Letter)
Int J Epidemiol
 , 
2004
(pg. 
33
(pg. 
616
-
17
)
6.
Reunanen
A
Heliovaara
M
Aho
K
Coffee consumption and risk of type 2 diabetes mellitus
Lancet
 , 
2003
, vol. 
361
 (pg. 
702
-
3
)
7.
Salazar-Martinez
E
Willett
WC
Ascherio
A
, et al.  . 
Coffee consumption and risk for type 2 diabetes mellitus
Ann Intern Med
 , 
2004
, vol. 
140
 (pg. 
1
-
8
)
8.
Saremi
A
Tulloch-Reid
M
Knowler
WC
Coffee consumption and the incidence of type 2 diabetes
Diabetes Care
 , 
2003
, vol. 
26
 (pg. 
2211
-
12
)
9.
Tuomilehto
J
Hu
G
Bidel
S
, et al.  . 
Coffee consumption and risk of type 2 diabetes mellitus among middle-aged Finnish men and women
JAMA
 , 
2004
, vol. 
291
 (pg. 
1213
-
19
)
10.
van Dam
RM
Feskens
EJ
Coffee consumption and risk of type 2 diabetes mellitus
Lancet
 , 
2002
, vol. 
360
 (pg. 
1477
-
8
)
11.
van Dam
RM
Dekker
JM
Nijpels
G
, et al.  . 
Coffee consumption and incidence of impaired fasting glucose, impaired glucose tolerance, and type 2 diabetes: the Hoorn Study
Diabetologia
 , 
2004
, vol. 
47
 (pg. 
2152
-
9
)
12.
Rosengren
A
Dotevall
A
Wilhelmsen
L
, et al.  . 
Coffee and incidence of diabetes in Swedish women: a prospective 18-year follow-up study
J Intern Med
 , 
2004
, vol. 
255
 (pg. 
89
-
95
)
13.
van Dam
RM
Willett
WC
Manson
JE
, et al.  . 
Coffee, caffeine, and risk of type 2 diabetes: a prospective cohort study in younger and middle-aged U.S. women
Diabetes Care
 , 
2006
, vol. 
29
 (pg. 
398
-
403
)
14.
van Dam
RM
Hu
FB
Coffee consumption and risk of type 2 diabetes: a systematic review
JAMA
 , 
2005
, vol. 
294
 (pg. 
97
-
104
)
15.
Ludwig
DS
Peterson
KE
Gortmaker
SL
Relation between consumption of sugar-sweetened drinks and childhood obesity: a prospective, observational analysis
Lancet
 , 
2001
, vol. 
357
 (pg. 
505
-
8
)
16.
Mrdjenovic
G
Levitsky
DA
Nutritional and energetic consequences of sweetened drink consumption in 6- to 13-year-old children
J Pediatr
 , 
2003
, vol. 
142
 (pg. 
604
-
10
)
17.
Liebman
M
Pelican
S
Moore
SA
, et al.  . 
Dietary intake, eating behavior, and physical activity-related determinants of high body mass index in rural communities in Wyoming, Montana, and Idaho
Int J Obes Relat Metab Disord
 , 
2003
, vol. 
27
 (pg. 
684
-
92
)
18.
Bray
GA
Nielsen
SJ
Popkin
BM
Consumption of high-fructose corn syrup in beverages may play a role in the epidemic of obesity
Am J Clin Nutr
 , 
2004
, vol. 
79
 (pg. 
537
-
43
)
19.
Schulze
MB
Manson
JE
Ludwig
DS
, et al.  . 
Sugar-sweetened beverages, weight gain, and incidence of type 2 diabetes in young and middle-aged women
JAMA
 , 
2004
, vol. 
292
 (pg. 
927
-
34
)
20.
The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives
The ARIC investigators
Am J Epidemiol
 , 
1989
, vol. 
129
 (pg. 
687
-
702
)
21.
Willett
WC
Sampson
L
Stampfer
MJ
, et al.  . 
Reproducibility and validity of a semiquantitative food frequency questionnaire
Am J Epidemiol
 , 
1985
, vol. 
122
 (pg. 
51
-
65
)
22.
National Heart, Lung, and Blood Institute
Atherosclerosis Risk in Communities (ARIC) Study. Operations manual no. 2: cohort component procedures. Version 1.0
1987
Chapel Hill, NC
ARIC Coordinating Center, School of Public Health, University of North Carolina
23.
Baecke
JA
Burema
J
Frijters
JE
A short questionnaire for the measurement of habitual physical activity in epidemiological studies
Am J Clin Nutr
 , 
1982
, vol. 
36
 (pg. 
936
-
42
)
24.
Duncan
BB
Schmidt
MI
Pankow
JS
, et al.  . 
Low-grade systemic inflammation and the development of type 2 diabetes: the Atherosclerosis Risk in Communities Study
Diabetes
 , 
2003
, vol. 
52
 (pg. 
1799
-
805
)
25.
StataCorp
Stata statistical software: release 8
2003
College Station, TX
StataCorp LP
26.
Van Soeren
MH
Graham
TE
Effect of caffeine on metabolism, exercise endurance, and catecholamine responses after withdrawal
J Appl Physiol
 , 
1998
, vol. 
85
 (pg. 
1493
-
501
)
27.
Tagliabue
A
Terracina
D
Cena
H
, et al.  . 
Coffee induced thermogenesis and skin temperature
Int J Obes Relat Metab Disord
 , 
1994
, vol. 
18
 (pg. 
537
-
41
)
28.
Bracco
D
Ferrarra
JM
Arnaud
MJ
, et al.  . 
Effects of caffeine on energy metabolism, heart rate, and methylxanthine metabolism in lean and obese women
Am J Physiol
 , 
1995
, vol. 
269
 (pg. 
E671
-
8
)
29.
Greer
F
Hudson
R
Ross
R
, et al.  . 
Caffeine ingestion decreases glucose disposal during a hyperinsulinemic-euglycemic clamp in sedentary humans
Diabetes
 , 
2001
, vol. 
50
 (pg. 
2349
-
54
)
30.
Keijzers
GB
De Galan
BE
Tack
CJ
, et al.  . 
Caffeine can decrease insulin sensitivity in humans
Diabetes Care
 , 
2002
, vol. 
25
 (pg. 
364
-
9
)
31.
Thong
FS
Derave
W
Kiens
B
, et al.  . 
Caffeine-induced impairment of insulin action but not insulin signaling in human skeletal muscle is reduced by exercise
Diabetes
 , 
2002
, vol. 
51
 (pg. 
583
-
90
)
32.
van Dam
RM
Pasman
WJ
Verhoef
P
Effects of coffee consumption on fasting blood glucose and insulin concentrations: randomized controlled trials in healthy volunteers
Diabetes Care
 , 
2004
, vol. 
27
 (pg. 
2990
-
2
)
33.
Paolisso
G
Di Maro
G
Cozzolino
D
, et al.  . 
Chronic magnesium administration enhances oxidative glucose metabolism in thiazide treated hypertensive patients
Am J Hypertens
 , 
1992
, vol. 
5
 (pg. 
681
-
6
)
34.
Paolisso
G
Sgambato
S
Pizza
G
, et al.  . 
Improved insulin response and action by chronic magnesium administration in aged NIDDM subjects
Diabetes Care
 , 
1989
, vol. 
12
 (pg. 
265
-
9
)
35.
Balon
TW
Gu
JL
Tokuyama
Y
, et al.  . 
Magnesium supplementation reduces development of diabetes in a rat model of spontaneous NIDDM
Am J Physiol
 , 
1995
, vol. 
269
 (pg. 
E745
-
52
)
36.
Kao
WH
Folsom
AR
Nieto
FJ
, et al.  . 
Serum and dietary magnesium and the risk for type 2 diabetes mellitus: the Atherosclerosis Risk in Communities Study
Arch Intern Med
 , 
1999
, vol. 
159
 (pg. 
2151
-
9
)
37.
Fung
TT
Manson
JE
Solomon
CG
, et al.  . 
The association between magnesium intake and fasting insulin concentration in healthy middle-aged women
J Am Coll Nutr
 , 
2003
, vol. 
22
 (pg. 
533
-
8
)
38.
Johnston
KL
Clifford
MN
Morgan
LM
Coffee acutely modifies gastrointestinal hormone secretion and glucose tolerance in humans: glycemic effects of chlorogenic acid and caffeine
Am J Clin Nutr
 , 
2003
, vol. 
78
 (pg. 
728
-
33
)
39.
Rodriguez de Sotillo
DV
Hadley
M
Chlorogenic acid modifies plasma and liver concentrations of: cholesterol, triacylglycerol, and minerals in (fa/fa) Zucker rats
J Nutr Biochem
 , 
2002
, vol. 
13
 (pg. 
717
-
26
)