Higher Magnesium Intake Reduces Risk of Impaired Glucose and Insulin Metabolism and Progression From Prediabetes to Diabetes in Middle-Aged Americans

Study Sample

The National Heart, Lung, and Blood Institute’s Framingham Heart Study (FHS) Offspring cohort is a community-based, longitudinal study of cardiovascular disease that began in 1971 whose participants are among the offspring of the original FHS cohort (19). In the fifth examination cycle (1991–1995) of the Offspring cohort, 3,799 participants underwent a standard medical examination, including laboratory and anthropometric measurements, as well as dietary assessment. Participants were followed from baseline at the fifth through seventh (1998–2001) examinations. Individuals were excluded if they had a history of diabetes or were identified as having diabetes at the baseline examination (n = 400), if they had invalid dietary data at baseline (n = 326), if they were missing necessary covariates (n = 109), if they were not present at the final follow-up examination (n = 329, 135 of whom were lost to follow-up owing to death), or if they had invalid or missing dietary data over follow-up (n = 53). The final sample size for the primary analysis was 2,582 participants.

A 2-h 75-g oral glucose tolerance test (OGTT) was administered to all participants at exam 5 and in a subset of participants at exam 7 who had undergone OGTT at exam 5, based on glucose tolerance at exam 5 (sex block-randomly selected from five quintile strata of FG). A total of 863 participants had follow-up OGTT measures available for the present analysis.

The original data collection protocols were approved by the institutional review board at Boston University Medical Center, and written informed consent was obtained from all participants. The current study protocol was reviewed by the Tufts Medical Center and Tufts University Health Sciences institutional review board.

Dietary Assessment

The Harvard semiquantitative, 126-item food frequency questionnaire (FFQ) was used to assess dietary intake at each exam (20). The FFQ included a list of foods together with a standard serving size and nine consumption frequency categories ranging from “never, or less than once per month” to “6+ per day.” Participants were asked to report consumption of each food item over the previous year. Invalid FFQs were defined as those that estimated daily caloric intake as <600 kcal/day or ≥4,000 kcal/day for women or ≥4,200 kcal/day for men or those that had ≥12 blank items. The exposure of interest, total magnesium intake, included both dietary and nondietary (i.e., supplemental) sources. Magnesium intake from nondietary sources contributed ∼5% of total intake.

The relative validity of the FFQ for energy-adjusted magnesium intake has been previously reported (20–22) and shows reasonable correlation with estimates from dietary records (r = 0.67–0.71) (20). All nutrients were adjusted for total energy using the residual method (23).

To account for long-term dietary exposure and to reduce within-person variability, intake of nutrients is presented as mean intake obtained from the dietary data of the fifth (baseline), sixth, and/or seventh examinations. For those with incident type 2 diabetes, intake of nutrients and energy was averaged across dietary data from the fifth examination up to but not including the examination at which diabetes incidence was ascertained. For those without incident diabetes, intake was averaged across all exams (5, 6, and/or 7) for which dietary data were available.

Outcome Measures and Definitions

FG and 2-h OGTT glucose were measured in fresh specimens with a hexokinase reagent kit (A-Gent glucose test; Abbot, South Pasadena, CA). At baseline (exam 5), fasting plasma insulin (FI) and 2-h OGTT insulin were measured using Coat-A-Count total insulin radioimmunoassay (RIA) (Diagnostic Products Corp., Los Angeles, CA), and at exam 7, FI and 2-h OGTT insulin were measured using a different assay, the human-specific insulin RIA (Linco Research Inc., St. Charles, MO). Owing to the different assays used to measure insulin at exams 5 and 7, a calibration study was conducted using FI in frozen plasma from 87 participants. These samples from exam 5 were reanalyzed ∼9 years later using the human-specific insulin RIA assay, and a regression equation was derived to calibrate total insulin RIA measures at exam 5 to human-specific RIA-equivalent values. The calibrated measures were used in the present analysis.

We defined metabolic impairment or type 2 diabetes based in part on impaired glucose criteria from the American Diabetes Association (ADA) (24), in addition to impaired insulin criteria (Supplementary Table 1); participants were classified as having type 2 diabetes if they had a diagnosis of type 2 diabetes, reported use of an oral hypoglycemic drug or insulin, or had FG ≥7.0 mmol/L (≥126 mg/dL) or 2-h OGTT glucose ≥11.1 mmol/L (≥200 mg/dL). Metabolic impairment was defined as having one or more of the following: IFG, IGT, IR, or hyperinsulinemia, per criteria that follow. Participants were classified as having normal FG (NFG) if they had FG <5.6 mmol/L (<100 mg/dL). IFG was classified as FG ≥5.6 to <7.0 mmol/L (≥100 to <126 mg/dL). Normal glucose tolerance (NGT) was classified as 2-h OGTT glucose <7.8 mmol/L (<140 mg/dL). IGT was classified as 2-h OGTT glucose ≥7.8 to <11.1 mmol/L (≥140 to <200 mg/dayL). Homeostasis model assessment of IR (HOMA-IR), a measure of hepatic IR, was calculated as FI (µU/mL) × FG (mmol/L)/22.5 (25). IR was defined as HOMA-IR ≥90th percentile. Hyperinsulinemia was defined as FI ≥90th percentile. Gutt’s insulin sensitivity index_0,120 (ISI), a measure of peripheral tissue insulin sensitivity, was calculated as ISI = (m/mean plasma glucose)/log(mean serum insulin), where the glucose uptake rate in peripheral tissues (m) = {75,000 mg + [FG (mg/dL) – 2-h OGTT glucose (mg/dL)] × 0.19 × weight (kg)}/120 min; mean plasma glucose = mean of FG (mmol/L) and 2-h OGTT glucose (mmol/L); and mean serum insulin = mean of serum FI (µU/L) and 2-h OGTT serum insulin (µU/L) (26).

Covariates

Potential confounders of the relationship between diet and progression to metabolic impairment or diabetes were considered as covariates. Covariates were assessed at baseline as follows. Age was measured in years. BMI was calculated as weight in kilograms divided by height in meters squared (kg/m²). Waist circumference (cm) was measured at the umbilicus with the participant standing. Parental history of diabetes was based on self-reported history in one or both natural parents. Blood pressure (BP) was measured twice by a physician and averaged to calculate the systolic and diastolic BP (mmHg). Hypertension (yes/no) was defined as BP ≥130/85 mmHg or undergoing treatment for hypertension. Information on regular smoking during the year prior to the examination (yes/no) was assessed via questionnaire. Physical activity was quantified as a continuous score based on activity levels as well as intensities of these activities, as previously described (27).

Statistical Analyses

We generated energy-adjusted quintile categories of averaged magnesium intake. Participant characteristics adjusted for age, sex, and energy (in the case of foods and nutrients) are presented across quintile categories. Tests for linear trend across increasing categories of intake were performed by assigning the median value of intake within each category and treating these as a continuous variable.

Because we sought to characterize magnesium’s associations with progression from normal to metabolic impairment, we assessed the association of magnesium intake with 1) incident metabolic impairment (defined as having IFG, IGT, IR, or hyperinsulinemia), among participants with normal status (NFG, NGT, no IR, and normoinsulinemia) at baseline, and 2) incident type 2 diabetes among participants who had baseline metabolic impairment, as defined above. Because there were few cases of incident diabetes among those with normal baseline status (n = 25), these cases were incorporated into our definition of incident metabolic impairment. In secondary sensitivity analyses, we redefined metabolic impairment by use of the ADA prediabetes criteria of IFG and IGT only, as these are frequently used in other clinical and research contexts. This redefinition also allowed us to examine whether excluding IR and hyperinsulinemia from the definition of metabolic impairment impacts magnesium’s associations with incident disorder. Relative risks (RRs) and 95% CIs across quintile categories of magnesium intake were estimated from multivariable logistic regression analyses for incident metabolic impairment or diabetes. P for trend was estimated using the median value in each category of intake.

In secondary analyses in participants without incident diabetes, we assessed the association between magnesium intake and change in continuous measures of FG, FI, HOMA-IR, 2-h OGTT glucose and insulin, and ISI over an average 7-year period. Change was modeled in each case as the final measure adjusted for the baseline measure. For these continuous outcomes, we estimated least squares adjusted means of values in each quintile category of energy-adjusted magnesium intake. P for trend was estimated using the median value in each category of intake. Natural-logged values were used for FI, HOMA-IR, and 2-h OGTT insulin, which were back transformed to geometric means for presentation.

For all outcomes, the initial analysis was adjusted for age, sex, and energy intake (model 1). Model 2 was adjusted as for model 1, plus parental history of diabetes, BMI, physical activity, smoking status, alcohol intake, and hypertension. In model 3, we further adjusted for dietary fiber. Additional adjustment for caffeine did not change the results, and therefore we do not include those results. Dietary fiber and caffeine were initially chosen because they represent surrogates of nonmagnesium constituents of commonly consumed magnesium-containing foods (i.e., whole grains and coffee), which themselves have been associated with lower risk of type 2 diabetes (6,28–30). Adjusting for these dietary variables allows us to at least partially distinguish the associations of magnesium from the associations of the foods themselves, their constituents (e.g., phytochemicals), or health behaviors associated with these nutrients (e.g., higher fiber may also be a surrogate for a healthy lifestyle).

In post hoc analyses, we modeled energy-adjusted dietary magnesium intake, adjusted for magnesium from supplements, to assess whether dietary intake specifically accounted for the observed associations. The results of dietary magnesium paralleled those of total magnesium, and without an a priori hypothesis regarding the mechanism of magnesium from dietary versus supplemental sources, we present the results from the original analyses of total magnesium intake described above.

Finally, we separately tested for statistical interaction between magnesium and age, sex, and BMI in the final models using cross-product terms. No interaction was statistically significant (all P > 0.1). Substituting waist circumference for BMI, or including waist circumference or change in weight between baseline and follow-up, did not substantively alter results.

All analyses were conducted in SAS (version 9.3; SAS Institute, Cary, NC). Statistical significance was set at the 0.05 level. All tests were two tailed.