Serum and Dietary Magnesium and the Risk for Type 2 Diabetes Mellitus: The Atherosclerosis Risk in Communities Study

Background  Experimental studies in animals and cross-sectional studies in humans have suggested that low serum magnesium levels might lead to type 2 diabetes; however, this association has not been examined prospectively.

Methods  We assessed the risk for type 2 diabetes associated with low serum magnesium level and low dietary magnesium intake in a cohort of nondiabetic middle-aged adults (N=12,128) from the Atherosclerosis Risk in Communities Study during 6 years of follow-up. Fasting serum magnesium level, categorized into 6 levels, and dietary magnesium intake, categorized into quartiles, were measured at the baseline examination. Incident type 2 diabetes was defined by self-report of physician diagnosis, use of diabetic medication, fasting glucose level of at least 7.0 mmol/L (126 mg/dL), or nonfasting glucose level of at least 11.1 mmol/L (200 mg/dL).

Results  Among white participants, a graded inverse relationship between serum magnesium levels and incident type 2 diabetes was observed. From the highest to the lowest serum magnesium levels, there was an approximate 2-fold increase in incidence rate (11.1, 12.2, 13.6, 12.8, 15.8, and 22.8 per 1000 person-years; P=.001). This graded association remained significant after simultaneous adjustment for potential confounders, including diuretic use. Compared with individuals with serum magnesium levels of 0.95 mmol/L (1.90 mEq/L) or greater, the adjusted relative odds of incident type 2 diabetes rose progressively across the following lower magnesium categories: 1.13 (95% CI, 0.79-1.61), 1.20 (95% CI, 0.86-1.68), 1.11 (95% CI, 0.80-1.56), 1.24 (95% CI, 0.86-1.78), and 1.76 (95% CI, 1.18-2.61) (for trend, P=.01) In contrast, little or no association was observed in black participants. No association was detected between dietary magnesium intake and the risk for incident type 2 diabetes in black or white participants.

Conclusions  Among white participants, low serum magnesium level is a strong, independent predictor of incident type 2 diabetes. That low dietary magnesium intake does not confer risk for type 2 diabetes implies that compartmentalization and renal handling of magnesium may be important in the relationship between low serum magnesium levels and the risk for type 2 diabetes.

TYPE 2 DIABETES mellitus imposes a substantial public health burden in the United States. It affects approximately 15 million Americans and leads to an excess risk for blindness, renal failure, lower-extremity amputation, and cardiovascular disease.^1,2 The pathogenesis of the disease is complex. To date, only obesity and physical inactivity have been well established as modifiable risk factors for type 2 diabetes.^3,4 Several lines of evidence suggest a possible novel risk factor: magnesium deficiency.^5-8 First, several large observational studies, including the Atherosclerosis Risk in Communities (ARIC) Study, have demonstrated strong cross-sectional associations between low serum magnesium levels and type 2 diabetes.^7,9-14 Second, in vitro studies have shown an effect of magnesium on the secretion of insulin by the pancreas and on the responsiveness to insulin by peripheral tissues.^15,16 Third, magnesium supplementation prevents the development of diabetes in a rat model of spontaneous type 2 diabetes.¹⁷ Finally, clinical intervention studies demonstrated that daily magnesium supplementation can improve short-term insulin response and glucose handling in diabetic individuals.^18-22 The establishment of an association between serum magnesium level and the risk for diabetes might suggest dietary or pharmacological measures to prevent type 2 diabetes. To our knowledge, the association of magnesium intake or magnesium body content, as reflected by serum magnesium level, with type 2 diabetes has only been studied cross-sectionally in humans. We therefore conducted a prospective cohort study to examine the association between serum magnesium level and dietary magnesium intake and the subsequent risk for incident type 2 diabetes in a community-based cohort of middle-aged adults. Cross-sectional analyses were also performed for contrast with the prospective analyses. In light of national survey data indicating that black individuals have lower levels of serum magnesium and are more likely to have type 2 diabetes compared with white individuals,²³ we were particularly interested in this relationship in black participants.

The ARIC Study is an ongoing prospective study that examines clinical and subclinical atherosclerotic diseases in a cohort of 15,792 persons, aged 45 to 64 years at baseline examination, selected by probability sampling from the following 4 US communities: Forsyth County, North Carolina; Jackson, Miss; the northwest suburbs of Minneapolis, Minn; and Washington County, Maryland. The sampling procedure and methods used in the ARIC Study have been described in detail elsewhere.²⁴ The prospective analysis was based on information obtained after 6 years of follow-up, which included 2 clinic visits scheduled 3 years apart. For this analysis, we excluded participants who reported ethnicity other than black or white American (n=48), had diabetes at baseline (n=1867), had missing exposure or outcome information (n=578), were nonfasting (<8 hours) at baseline (n=279), or were unavailable for follow-up (n=746) or dead (n=146) before visit 2. After these exclusions, we were left with 12,128 participants. Between the second and third visits, additional participants were unavailable for follow-up (n=1093) or dead (n=164), leaving 10,871 participants who were observed for the full 6 years. Thus, total follow-up of our cohort was 67,212 person-years. For the dietary magnesium intake analysis, an additional 232 individuals were excluded due to missing exposure information or unacceptable dietary data. For the cross-sectional analysis, only nonwhite and nonblack individuals (n=48) or those who had missing data (n=1208) were excluded, leaving 14,536 individuals for analysis. Individuals who were excluded from the present analyses did not differ from those who were included.

Baseline examination included home interviews, clinic examinations, and clinic questionnaires. Participants were asked to fast for at least 12 hours before the actual blood collection. All analyses were performed at the Central Chemistry Laboratory at University of Minnesota, Minneapolis.²⁵

Serum magnesium level was measured using the metallochromic dye calmagite (1-[1-hydroxy-4-methyl-2-phenylazo]-2-naphthol-4-sulfonic acid), based on the procedure of Gindler and Heth.²⁶ The coefficient of variation, based on specimens sent 1 week apart blindly to the laboratory, was 3%.¹⁴ Serum glucose level was assessed using a modified hexokinase–glucose-6-phosphate dehydrogenase procedure, a Centers for Disease Control and Prevention national glucose reference method.²⁷ Standard radioimmunoassay was used to determine serum insulin level (iodine 125–labeled Insulin [732] Kit; Cambridge Medical Diagnostics, Inc, Billerica, Mass). Serum calcium level was determined by means of orthocresolphthalein complexone, a modification of the method of Connerty and Briggs.²⁸ Serum potassium level was determined using a direct electrochemical measurement on undiluted serum.²⁵

Information on age, sex, race, family history of diabetes, education, and use of diuretics was obtained from the home and clinic interviews conducted at the baseline visit. A positive family history of diabetes was defined as having any biological parent with diabetes, whereas a negative family history was defined as having both parents without diabetes or having 1 parent without diabetes and the other of unknown status. Body mass index (BMI; calculated as weight in kilograms divided by the square of height in meters) and waist-to-hip ratio were determined using the anthropometric measurements taken at the baseline clinic visit. Physical activity was assessed by using a modified interviewer-administered version of the questionnaire developed by Baecke et al.²⁹ Dietary intake was also assessed by an interviewer-administered, modified version of the 61-item food frequency questionnaire developed by Willett et al.³⁰ Although the validity and reproducibility of the questionnaire have been studied elsewhere, dietary intake of magnesium was not one of the factors included in those studies.^30,31 Dietary magnesium intake was computed by multiplying the magnesium content of the specified serving of each food item by the frequency of its daily consumption and summing all items.²⁵ Dietary magnesium intake was characterized as dietary magnesium in milligrams per 4.2 kJ daily energy intake. Since the use of additional magnesium supplements (n=3) or other multivitamins that may or may not contain magnesium (n=108) was negligible, and since there was not enough information to quantify this additional source of magnesium well, only dietary intake of magnesium was used as the exposure variable. Moreover, similar results were obtained from additional analyses that excluded these individuals.

Diabetes mellitus, at baseline and during follow-up, was defined as the presence of any of the following: (1) fasting glucose level of at least 7.0 mmol/L (126 mg/dL), (2) nonfasting glucose level of at least 11.1 mmol/L (200 mg/dL), (3) current use of diabetic medication, or (4) a positive response to the question "Has a doctor ever told you that you had diabetes (sugar in the blood)?" Individuals with diabetes at baseline were excluded from the prospective analyses. Individuals without diabetes at baseline who met any of these conditions at visit 2 or visit 3 were considered to have incident cases of diabetes. All incident cases of diabetes were classified as type 2 diabetes because the age of onset in this middle-aged cohort was 45 to 70 years. Persons who met the criteria for diabetes at visit 2 but not at visit 3 were nonetheless considered to have type 2 diabetes. Analyses where these individuals were excluded, treated as having incident diabetes, or treated as nondiabetic individuals showed similar results as those from the present analyses.

All analyses were stratified by race because of a significant interaction that was observed between race and the association of serum magnesium level and incident type 2 diabetes and because of our a priori interest in this association among black participants. Tests for interaction with magnesium level were also conducted with the variables sex, BMI, and education level; none was statistically significant. Serum magnesium concentration was categorized into 6 symmetrical and clinically meaningful groups. The lowest level of serum magnesium consisted of values from 0.25 to 0.70 mmol/L (0.50-1.40 mEq/L); hypomagnesemia is commonly defined as values below 0.75 mmol/L (1.50 mEq/L).¹³ The highest level consisted of values of at least 0.95 mmol/L (1.90 mEq/L), which is also a common clinical definition of hypermagnesemia.³² The group with the highest serum magnesium level was used as the reference in all regression analyses. The distribution of dietary magnesium intake per 4.2 kJ consumed appeared approximately normal and was categorized into quartiles. The highest quartile was used as the reference group.

Means and frequencies of potential confounders assessed at baseline were calculated for each level of serum magnesium, and analysis of variance and χ² analysis were used to assess the statistical significance of the differences across magnesium levels. Incidence rates were calculated for each level of the serum magnesium using a person-years approach, and χ² analysis was used to determine if the incidence rates for the levels of serum magnesium differed significantly. Since outcome was assessed every 3 years, the midpoint between 2 visits was used as the end point for those who had the event or were censored.

The association between baseline serum magnesium level and subsequent incident diabetes was examined using contingency tables and logistic regression models. Logistic regression, rather than the Cox proportional hazards model, was used and presented because of concerns about the lack of precise time of events; however, the relative hazards obtained from the Cox proportional hazards model were similar to the relative odds obtained from logistic regression. Based on the underlying model of the pathogenesis of type 2 diabetes, groups of potential confounders related to each other were formed. The first group of potential confounders included demographic characteristics, family history of diabetes, and variables that were related to obesity. The second group consisted of other clinically important cations, ie, serum calcium and potassium levels, and the use of alcohol and diuretics. The third group included fasting serum insulin and glucose levels. Results from the regressions are expressed as the relative odds of incident diabetes for each category of magnesium level and their 95% confidence intervals (CIs). Parallel analyses of the association between dietary magnesium intake and incident diabetes were performed. All statistical analyses were performed using the SAS statistical package.³³

Serum magnesium concentrations were distributed approximately normally in black and white participants, but the distribution of the level of serum magnesium in black participants was significantly shifted to the left compared with that of the white ones. The mean±SD serum magnesium level in the black participants was 0.80±0.08 mmol/L (1.60±0.16 mEq/L) compared with 0.83±0.07 mmol/L (1.66±0.14 mEq/L) in the white participants (P<.001). Although a difference of 0.03 mmol/L (0.06 mEq/L) in serum magnesium concentration is clinically insignificant at the individual level, this difference at the population level has great public health implications.

Table 1 and Table 2 display the differences in the baseline characteristics among the 6 levels of serum magnesium in both race groups. In black and white participants, those in the group with the lowest level of serum magnesium reported the use of diuretics more frequently, and they also had higher mean BMI and fasting insulin levels and lower serum calcium and potassium levels. Lower education level and higher waist-to-hip ratio were associated with low serum magnesium level in black participants only. Among white participants, those in the low serum magnesium group were more likely to be female and to have higher fasting glucose levels and lower dietary magnesium intake. There was no apparent association between dietary magnesium intake and serum magnesium concentration in black participants.

Table 3 summarizes the incidence of type 2 diabetes during 6 years of follow-up. Type 2 diabetes incidence rates differed significantly between black and white participants; overall the rates in black participants were substantially higher than those in white ones. Among black participants, there was no graded association between the crude incidence rate of type 2 diabetes and serum magnesium concentration at baseline. However, because of the smaller number of black ARIC participants, we cannot rule out a possible U-shaped relationship. In contrast, a strong, graded, inverse relationship between incidence rates of type 2 diabetes and serum magnesium levels was observed in the white participants. The magnesium level–specific rates rose from 11.1 per 1000 person-years in the highest category of serum magnesium level to 22.8 per 1000 person-years in the lowest category (P=.001).

The absence of a linear relationship between serum magnesium level and incident type 2 diabetes in black participants persisted after simultaneous adjustment for potential confounders using logistic regression (Table 4). In contrast, a strong, graded, inverse association between serum magnesium level and incident type 2 diabetes was observed among white participants in the unadjusted model (model 1), and although attenuated, it remained statistically significant even after simultaneous adjustment for age, sex, education, family history of diabetes, BMI, waist-to-hip ratio, physical activity, alcohol consumption, diuretic use, and serum calcium and potassium levels (models 2 and 3).

To this point, we had not adjusted for fasting serum insulin and glucose levels, under the assumption that insulin resistance and hyperglycemia may be in the causal pathway of low serum magnesium levels, leading to type 2 diabetes. However, since the longitudinal association observed between low serum magnesium levels and the risk for type 2 diabetes may merely be a reflection of the cross-sectional association between them, ie, individuals with a higher glucose level at baseline may have already had a lower serum magnesium level due to osmotic diuresis, we further adjusted for fasting serum insulin and glucose levels in model 4 (Table 4). Although the relationship was somewhat attenuated, low serum magnesium level remained an independent predictor of incident type 2 diabetes even after adjustment for fasting serum insulin and glucose levels in white participants. Type 2 diabetes was still about 55% more likely to develop in individuals with serum magnesium concentration from 0.25 to 0.70 mmol/L (0.50-1.40 mEq/L) compared with those with serum magnesium concentration of at least 0.95 mmol/L (1.90 mEq/L), holding serum insulin level and other covariates constant.

At baseline, there were 567 diabetic individuals among 3571 black participants (prevalence, 15.9%), and 949 diabetic individuals among 10,965 white participants (prevalence, 8.6%). After adjustment for age, sex, education, family history of diabetes, BMI, waist-to-hip ratio, physical activity, alcohol consumption, diuretic use, and serum calcium and potassium levels, low serum magnesium level was very strongly associated with prevalent type 2 diabetes. Compared with individuals in the group with the highest levels of serum magnesium, the relative odds of having type 2 diabetes among individuals with lower serum magnesium levels were 5.00 (95% CI, 3.44-7.27), 2.54 (95% CI, 1.77-3.66), 1.99 (95% CI, 1.40-2.83), 1.20 (95% CI, 0.83-1.73), and 0.96 (95% CI, 0.64-1.43) among white participants and 3.95 (95% CI, 2.20-7.07), 1.84 (95% CI, 1.02-3.33), 1.45 (95% CI, 0.80-2.62), 1.38 (95% CI, 0.75-2.52), and 0.72 (95% CI, 0.36-1.44) among black participants. Figure 1 illustrates the comparison of the cross-sectional and prospective associations of low serum magnesium levels with type 2 diabetes. In black and white participants, the cross-sectional association was much stronger than the prospective association, especially at the lowest levels of serum magnesium.

To determine whether the association between magnesium and incident type 2 diabetes was related to dietary intake, we next examined the association between dietary magnesium intake and incident type 2 diabetes (Table 5). Univariate analysis showed that in black and white participants, individuals in the highest quartile of dietary magnesium intake generally appeared to be at lower risk for development of type 2 diabetes than those in the lower 3 quartiles, although this relationship was not statistically significant and showed no evidence of gradedness. In black participants, simultaneous adjustment for demographic factors, adiposity, physical activity, medication use, and dietary mineral intake did not alter this pattern. In white participants, the unadjusted association had a significant trend, but demographic factors, adiposity, and physical activity largely accounted for this association in the multivariate model.

To assess the robustness of our results, we conducted the following 2 additional analyses: (1) exclusion of individuals with impaired glucose tolerance at baseline (fasting glucose level from 6.1 to 6.9 mmol/L [110-125 mg/dL]) for fear that a proportion of them may truly have prevalent diabetes and the resulting low serum magnesium level, thus driving the association seen in the prospective analysis, and (2) adjustment for baseline hypertension status, since cross-sectional association between high blood pressure and low serum magnesium levels has also been well established. The relative odds of diabetes, comparing the lowest level of serum magnesium to the highest level in white participants after adjusting for other confounders (not including fasting serum glucose and insulin levels), was still 1.77 (95% CI, 1.15-0.55; for trend, P=.03). Both analyses yielded results similar to those from the main analysis. Next, to assess possible geographic confounding related to uneven distribution of black participants across study sites, we performed analyses stratified by study site among black participants. There was little or no association among black participants from Forsyth County, the site with the second greatest number of black participants (n=301), or Jackson, the site with the greatest number (n=2281). Furthermore, to minimize the possibility of confounding by geographic area, race-stratified analysis of Forsyth County participants only (the only site with a reasonable number of black and white participants) was performed. The results showed that the inverse relationship between serum magnesium levels and risk for type 2 diabetes was only present in the white participants from that county. Finally, to determine if the association between serum magnesium level and type 2 diabetes was dependent on follow-up time, we performed analyses stratified by time at type 2 diabetes onset (0-3 vs 4-6 years after baseline visit) and found that the association between serum magnesium level and risk for development of type 2 diabetes 0 to 3 years after the measurement was approximately the same as the association between serum magnesium level and the risk for development of type 2 diabetes 4 to 6 years after the measurement (data not shown).

Our results support 4 main conclusions. First, there was a graded, inverse, independent relationship between serum magnesium level and the subsequent development of incident type 2 diabetes in white middle-aged adults. Second, although black participants were at higher risk for development of incident type 2 diabetes than their white counterparts, there was little or no association between serum magnesium level and the risk for type 2 diabetes in black participants. Third, the prospective association observed was markedly attenuated compared with the cross-sectional association in black and white participants, suggesting that much of the cross-sectional relationship was explained by the effect of diabetes on serum magnesium level (probably via osmotic diuresis) rather than by the effect of magnesium on glucose metabolism.³⁴ Fourth, there was no association between dietary magnesium intake and incident type 2 diabetes in black or white participants. The results from this study are consistent with previous observational and experimental studies that have reported an association between magnesium and type 2 diabetes^{9,12,14,20,35,36}; however, this study is the first to establish an association between low serum magnesium level and incident diabetes prospectively in humans. Strengths of our study include a population-based sampling method, extensive data on potential confounders, and a large sample size that increased precision and permitted multiple statistical adjustments.

Nevertheless, several limitations of our study should be kept in mind. First, our measurement of magnesium level was confined to serum, a relatively minor compartment for magnesium, accounting for roughly 1% of whole-body magnesium level, and it is known that serum magnesium level may not reflect magnesium levels accurately in other tissues.^37,38 However, it is also true that there is no consensus regarding the measurement of magnesium status. Serum magnesium level has been shown to be correlated with intracellular free magnesium level (r=0.54) in a study using phosphorus 31–labeled nuclear magnetic resonance spectroscopy.³⁹ If intracellular magnesium level is the true predictor of the development of type 2 diabetes, and if by using serum magnesium level to assess magnesium status we have randomly misclassified individuals, then we may have underestimated the true association between magnesium level and incident type 2 diabetes. In addition, since serum magnesium level was measured at only 1 occasion, the reliability of this assessment may be questionable. However, in a study conducted in healthy male subjects, the intraindividual variation of serum magnesium level was shown to be 3.2%, much smaller than the 36% found in urine magnesium level.³⁸ Although both sources of error may have contributed to random misclassification, random misclassification tends to bias the result toward the null; thus, we may have underestimated the true association between magnesium level and incident type 2 diabetes. Second, the validity and reliability of the assessment of dietary intake of magnesium are unknown. Although validity and reliability studies of the food frequency questionnaire have been conducted previously, those results may not necessarily be generalizable to the ARIC population. Third, serum glucose level measurement, rather than oral glucose tolerance test, was used to assess the outcome. Furthermore, serum glucose level was assessed at only 1 point in time. Therefore, the validity and the reliability of this assessment may not be optimal. However, the use of serum glucose level to define diabetes has been demonstrated to outperform other measurements, such as glycated hemoglobin level⁴⁰; furthermore, defining the presence of diabetes by using fasting serum glucose level of at least 7.0 mmol/L (126 mg/dL) is supported by the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus of the American Diabetes Association.⁴¹ Since glucose tolerance tests are being performed during the ongoing ARIC visit 4 examinations, this issue can be reexamined at the study's close. Finally, reliability of 1-time assessment of serum glucose level also has been shown to be better than that of the oral glucose tolerance test. When oral glucose tolerance tests were repeated in adults during a 2- to 6-week interval, the intraindividual coefficients of variation were 6.4% for the fasting serum glucose level and 16.7% for the 2-hour postload glucose level.⁴²

Previous work suggests several possible mechanisms whereby low serum magnesium levels may lead to the development of type 2 diabetes. First, as an essential cofactor in reactions involving phosphorylation, magnesium could impair the insulin signal transduction pathway.^32,43,44 Second, low serum or erythrocyte magnesium level may affect the interaction between insulin and the insulin receptor by decreasing hormone-receptor affinity or by increasing membrane microviscosity.^45,46 Finally, magnesium can also be a limiting factor in carbohydrate metabolism, since many of the enzymes in this process require magnesium as a cofactor during reactions that utilize phosphorus bonds.^{32,43,44,47,48}

On the other hand, it is possible that the association between low serum magnesium level and incident diabetes could be noncausal. Low serum magnesium level simply could be a marker for the effect of other minerals, such as serum calcium and potassium. However, no apparent association between calcium or potassium (serum level and dietary intake) and the risk for type 2 diabetes was found in the multivariate analyses. In vivo and in vitro studies have shown that insulin increases intracellular magnesium level in erythrocytes and platelets, suggesting that magnesium transport is an insulin receptor–mediated process.^15,49,50 In humans, insulin resistance has been implicated to impair the ability of insulin to stimulate magnesium or glucose uptake in diabetic individuals.¹⁰ This line of reasoning suggests that low serum magnesium level simply could be a marker of insulin resistance and hyperinsulinemia. This hypothesis, however, is contradicted by our finding that serum magnesium level predicted incident type 2 diabetes independent of serum insulin level.

The lack of association between low serum magnesium level and the risk for diabetes in black participants was unexpected. This finding is not likely to be explained by geographic confounding, since black participants from Forsyth County and Jackson had similar risk patterns, nor is it due to differential reporting of diabetes between black and white participants, since serum glucose measurements were used in determining type 2 diabetes status. Of course, confounding due to other unknown factors, residual confounding by the variables examined in our analyses, or the occurrence of a type 2 error due to the smaller number of black ARIC participants cannot be ruled out. Finally, the effect modification we observed might reflect a real difference in susceptibility to magnesium's influence in glucose metabolism in black vs white participants. We speculate that magnesium's modest effect on type 2 diabetes risk is overwhelmed by the much higher background rates of type 2 diabetes incidence in black participants.

In contrast to the Nurses' Health Study, our study did not show a significant association of low dietary magnesium intake with incident type 2 diabetes.⁵¹ The main difference between these studies lies in their participants. Individuals from the Nurses' Health Study are probably more highly motivated and more experienced in quantitative measurements than the ARIC participants, thus resulting in more precise assessment of dietary magnesium intake. Moreover, adjustment for dietary intakes of other minerals explained most of the association between magnesium intake and incident type 2 diabetes in both studies.

Why should low serum magnesium levels, but not low dietary magnesium intake, predict type 2 diabetes? First, compared with serum magnesium levels, the relatively less precise measurement of dietary magnesium intake could have resulted in a type 2 error. Second, across the usual range of dietary magnesium intake, serum magnesium levels reflect compartmentalization or renal handling of whole-body magnesium level rather than dietary intake of magnesium, as suggested by the low correlation between dietary magnesium intake and serum magnesium level in the ARIC cohort (r=0.06).¹⁴ This, however, does not preclude the possibility that pharmacological doses of magnesium intake as supplement might influence serum magnesium level and/or glucose metabolism. In experimental studies of diabetic individuals, magnesium supplements of 2.0 to 4.5 g/d increased plasma magnesium concentration by 0.80 mmol/L (1.60 mEq/L) and improved short-term glucose handling and insulin response.¹⁹ In general, the doses used in experimental studies were about 10 times greater than the average dietary intake of magnesium of ARIC participants (234 mg/d for black participants and 259 mg/d for white participants).^18-21,32,52

The main implication of our results is that low serum magnesium levels confer an increased risk for type 2 diabetes in white middle-aged adults. Although type 2 diabetes is a multifactorial disease, our results nonetheless raise the possibility that increased magnesium consumption, along with modification of other risk factors for type 2 diabetes, might represent a novel means to prevent type 2 diabetes. Our findings suggest that modification of magnesium intake by dietary means alone may be inadequate to achieve such an effect. However, whether pharmacological doses of magnesium used as dietary supplements can reduce the long-term risk for type 2 diabetes remains to be investigated.

Accepted for publication February 9, 1999.

The Atherosclerosis Risk in Communities (ARIC) Study is performed as a collaborative study supported by contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022 from the National Heart, Lung, and Blood Institute, Bethesda, Md; an Established Investigator Grant from the American Heart Association, Dallas, Tex (Dr Brancati); and training grant T32HL07024-23 from the National Institutes of Health, Bethesda (Ms Kao). Computational assistance was received from the General Clinical Research Center Computing Center sponsored by grant RR00052 from the National Institutes of Health.

Presented in part at the 57th Scientific Sessions of the American Diabetes Association, Boston, Mass, June 23, 1997.

We thank the staffs at the ARIC field centers and the coordinating center. In particular, the following persons are acknowledged: Phyllis Johnson, Marilyn Knowles, Catherine Paton, Joy Rollins, Debbie Rubin-Williams, Patsy Tacker, and Lily Wang from University of North Carolina, Chapel Hill (coordinating center); Shirley Cothern, Amy Haire, Kim Jones, and Delilah Posey from University of North Carolina, Forsyth County; Virginia Overman, Stephanie Parker, Liza Sullivan, and Cora Walls from University of Mississippi Medical Center, Jackson; Gerda Nightingale, Carmen O'Donnell, Evie Onnen, and Leone Reed from University of Minnesota, Minneapolis; Dorothy Nixon, Thelma Oliver, Rodney Palmer, and Patricia Slagle from Johns Hopkins University, Baltimore, Md; Valarie Stinson, Pam Pfile, Hogan Pham, and Teri Trevino from University of Texas Medical School, Houston; Wanda Alexander, Doris Harper, Charles E. Rhodes, and Selma Soyal from Atherosclerosis Clinical Laboratory, The Methodist Hospital, Houston; and Anne Safrit, Melanie Wilder, Linda Allred, and Carolyn Bell from Bowman-Gray School of Medicine Ultrasound Reading Center, Winston-Salem, NC.

Reprints: Frederick L. Brancati, MD, MHS, Welch Center for Prevention, Epidemiology, and Clinical Research, The Johns Hopkins Medical Institutions, Baltimore, MD 21205-2223 (e-mail: fbrancat@welchlink.welch.jhu.edu).