点击显示 收起
1 From the Department of Internal Medicine (SB, MD, SG, M Carello, BS, M Cassader, and GP), the Unit of Cancer Epidemiology (CS and RR), University of Turin, Turin, Italy, and the Diabetic Clinic, Hospital of Asti, Italy (LG)
2 Supported by a grant from Regione Piemonte. 3 Reprints not available. Address correspondence to S Bo, Department of Internal Medicine, University of Turin, Corso Dogliotti 14, 10126 Turin, Italy. E-mail: sbo{at}molinette.piemonte.it.
ABSTRACT
Background: Type 2 diabetes (DM), metabolic syndrome (MetS), and inflammation are linked to reduced magnesium and fiber intakes; these associations are attenuated by adjustment for each of these nutrients.
Objective: We investigated the association among magnesium and fiber intakes, metabolic variables, and high-sensitivity C-reactive protein (hs-CRP) values.
Design: Cross-sectional analyses were performed in a representative cohort of 1653 adults and in a subgroup with normal body mass index without dysmetabolisms (n = 205). A validated semiquantitative food-frequency questionnaire was used; magnesium intake was computed by multiplying its content in each food by the frequency of food consumption.
Results: The prevalence of DM, MetS, and hs-CRP 3 mg/L significantly decreased from the lowest to the highest tertile of magnesium and fiber intakes. Subjects within the lowest tertiles of magnesium and fiber intakes were 3–4 times as likely to have DM, MetS, and hs-CRP 3 mg/L, after multiple adjustments. After the analysis was additionally controlled for fiber intake, associations with hs-CRP 3 mg/L proved to be significant (odds ratio: 2.05; 95% CI: 1.30, 3.25), whereas reduced magnesium intake and DM and MetS were no longer significant. The lowest tertile of fiber intake remained associated with DM, hs-CRP 3 mg/L, and MetS after adjustments for multiple confounders and magnesium intake. In the lean, healthy subject subgroup, hs-CRP values were inversely associated with magnesium and fiber intakes in a multivariate model (P < 0.001).
Conclusions: Reduced fiber intake was significantly associated with metabolic abnormalities; the magnesium effect might be confounded by fiber being in foods that also provided magnesium. Lower magnesium and fiber intakes were linked to hs-CRP 3 mg/L in both the entire cohort and healthy persons.
Key Words: Body mass index • fiber • high-sensitivity C-reactive protein • magnesium • metabolic syndrome • type 2 diabetes
INTRODUCTION
Growing evidence from the literature suggests that reduced magnesium intake and serum concentrations are associated, both cross-sectionally and prospectively, with type 2 diabetes and insulin resistance (1-4), hypertension (4-6), dyslipidemia (7), the metabolic syndrome (MetS) (8), and cardiovascular diseases (CVDs) (4, 9, 10). Furthermore, magnesium supplementation prevents the development of diabetes in rats (11) and improves insulin sensitivity in patients with type 2 diabetes (12, 13). Magnesium is an essential cofactor in multiple enzymatic reactions and a direct antagonist of intracellular calcium, and it could affect insulin action and carbohydrate metabolism. Insulin is, itself, an important regulator of intracellular magnesium (6).
Some recent studies (8, 14, 15) have reported an inverse association between dietary and serum concentrations of magnesium and C-reactive protein (CRP), both a marker of systemic inflammation and a well-known predictor of diabetes, MetS, and atherosclerotic diseases in healthy subjects (16). Those studies were performed on US or Mexican cohorts and included at least a portion of overweight or obese persons. Values of serum insulin and CRP could be affected by dysmetabolic disorders, such as overweight, obesity, or hypertension, or other metabolic abnormalities; otherwise these patients could be on particular diets as a consequence of their disorders. Furthermore, adjustments for fiber intake have given highly discordant results (8, 15). Dietary fibers have repeatedly shown to be inversely correlated with metabolic and inflammatory markers (17-20) and are strongly related to dietary magnesium, because both are contained in the same foods. Previous studies have found that associations between metabolic abnormalities and high-fiber foods are attenuated by adjusting for magnesium intakes (18-20).
A cross-sectional analysis was, then, performed to investigate associations among dietary magnesium and fiber intakes and metabolic variables and CRP values in 1) population-based Southern-European middle-aged healthy subjects and 2) a healthy subgroup of these subjects with normal body mass index (BMI; in kg/m2) and with no dysmetabolic disorders, to minimize the effects of the metabolic abnormalities on CRP serum values.
SUBJECTS AND METHODS
All the patients between the ages of 45 and 64 y (n = 1877) of 6 family physicians from the province of Asti (northwestern Italy), whose patients were representative of the local health units, were invited to participate in a metabolic screening; 1658 (88.3%) decided to participate. All the participants were questioned about personal habits, answered a food questionnaire, and were tested for several clinical and laboratory measurements, after having given their informed consent. Both the participants and nonparticipants were similar to the resident population of the corresponding age group for percentage of men, education level, prevalence of known diabetes, and percentage of subjects living in a rural area (21). Five subjects were excluded, because they were taking magnesium supplements; 1653 persons were thus evaluated. All procedures were in accordance with the Declaration of Helsinki.
The semiquantitative food-frequency questionnaire used in the European Prospective Investigation into Cancer and Nutrition studies was used for all subjects. It assessed average frequency and portion intake of 148 foods consumed during the 12 mo before examination. Frequency of intake was measured with the use of 10 categories, ranging from "never" to "5 times per day or more"; photographs for standard portion size were used. Nutrient intake was calculated by multiplying the frequency of each food consumption by the nutrient content of the specified portion size. Although the reproducibility and validity of this tool for assessing food intake in the population was previously validated (22), dietary magnesium intake was not one of the factors included. Dietary magnesium intake was computed by multiplying magnesium content of the specific serving of each food item [according to the tables of food composition of the Italian National Institute of Nutrition (23)] by the frequency of its daily consumption and summing all items. The average glycemic index, a method of ranking foods on the basis of the blood glucose response to a given amount of carbohydrate from that food, was calculated according to a white bread standard for each participant, as previously described (17, 24). Each nutrient was adjusted for total energy with the use of the residual method (25).
Dietary energy-adjusted magnesium and fiber intakes of the cohort were divided into tertiles. Cutoff points, respectively, for men and women were <269.1, 269.1–337.4, >337.4 mg/d and < 278.1, 278.1–338.6, >338.6 mg/d for magnesium and < 16.4, 16.4–23.0, >23.0 g/d and < 16.1, 16.1–22.6, >22.6 g/d for fiber.
Weight and height were measured after an overnight fast. Waist circumference was measured with a plastic tape meter at the umbilicus level. Blood pressure was measured twice, after a 10-min rest interval, in a sitting position by using a standard mercury sphygmomanometer.
In the morning a venous blood sample was drawn to measure fasting concentrations of glucose, insulin, total and HDL cholesterol, triacylglycerol, uric acid, liver enzymes, prealbumin, and CRP. If the fasting serum glucose concentration was 110 mg/dL, a second fasting glucose measurement was then performed.
Serum glucose was measured by the glucose oxidase method; plasma triacylglycerol, HDL-cholesterol, uric acid, and -glutamyl transferase (GGT) concentrations were measured by enzymatic colorimetric assay (HITACHI 911 Analyzer; Sentinel Ch, Milan, Italy); and serum insulin was measured by immunoradiometric assay (Radim SpA, Pomezia, Italy; intraassay CV: 1.6–2.2%, interassay CV: 6.1–6.5%). Alanine aminotranferase and aspartate aminotransferase were measured with a kinetic determination (HITACHI 911 Analyzer; Sentinel Ch) and serum prealbumin by an immunochemical nephelometric reaction (Nephelometer 100 Analyzer; Dade Behring, Marburg, Germany).
Serum CRP values were measured with a high-sensitivity (hs-CRP) latex agglutination method on HITACHI 911 Analyzer (Sentinel Ch). The intraassay and interassay CVs were 0.8–1.3% and 1.0–1.5%. Hs-CRP cutoff of 3 mg/L was used to differentiate high- and low-risk groups for future cardiovascular events, consistent with recent recommendations (16).
Subjects were considered inactive if their level of physical activity during leisure was light (inactive, <4 h physical activity/wk). Otherwise, they were considered active if exercise was moderate (4 h/wk) or heavy (>4 h/wk, regularly) (26).
Diabetes was diagnosed when 2 glucose values were 126 mg/dL or if known diabetes was recorded by the family physician, according to published recommendations (27). The MetS was defined in accordance with criteria from the National Cholesterol Education Program (Adult Treatment Panel III) (28). Persons reporting a history of hypertension and current blood pressure medication were defined as having hypertension regardless of the blood pressure values measured. Insulin resistance was calculated from the homeostasis model assessment-insulin resistance (HOMA-IR), according to the published algorithm (29).
Diagnosis of CVD was based on resting electrocardiogram performed in all subjects and interpreted according to the Minnesota Code criteria, Rose questionnaire, and history of documented events recorded by the family physician. These events included angina, previous myocardial infarction, coronary artery bypass graft or other invasive procedures to treat coronary artery disease, transient ischemic attacks, strokes, gangrene, amputation, vascular surgery, intermittent claudication, absent foot pulses, and abnormal brachial and posterior tibial blood pressure by using Doppler scanning techniques.
Later, within this cohort, we identified 205 healthy subjects with normal BMI (<25) and with no MetS component and no known disorder (ie, CVD, impaired renal function, or liver or gastrointestinal diseases), or drug use. To avoid considering as normoglycemic persons who could be classified as hyperglycemic after the stimulatory test, these 205 subjects were submitted to the oral glucose tolerance test (75 g glucose) (27). All of them showed normal glucose tolerance.
Because the distributions of GGT, hs-CRP, insulin, HOMA-IR, and triacylglycerol values were highly skewed, their values were log-transformed, thus obtaining a normal distribution. In all analyses the log-transformed values were then used. For easy interpretation, nontransformed values were reported in the tables.
Analysis of variance and the chi-square test were used to compare means for continuous variables, or frequencies for discrete variables. Linear correlation between magnesium and fiber intakes and HOMA-IR values was evaluated by Pearson partial coefficients. Logistic regression models, including variables for age, sex, smoking habits, alcohol consumption, BMI, physical activity, and intake of different nutrients, were used to estimate prevalence odds ratios (ORs) for those factors that were associated with diabetes, hs-CRP 3 mg/L, and the MetS. The likelihood ratio (LR) test was used to evaluate the interactions between tertiles of magnesium and fiber intakes on diabetes, hs-CRP 3 mg/L, and the MetS (SAS, version 8.0; SAS Institute, Cary, NC).
RESULTS
Median magnesium intake was 308 mg/d; 52.3% of men and 30.8% of women consumed less than the recommended dietary assumption for European populations (4 mg/kg; recommended range 150–500 mg/d) (30). The main sources of dietary magnesium in the entire cohort were represented by the following food categories: vegetables (31.4%), fruit (25.2%), dairy products (12.4%), bread (7.5%), and legumes (7.2%).
Subjects within the highest tertile of magnesium intake showed no different BMI and waist circumference values than did the others, but they had a significantly higher level of physical activity (Table 1). Serum glucose, triacylglycerol, insulin, HOMA-IR, hs-CRP, uric acid, and GGT values significantly decreased from the lowest to the highest tertile of magnesium intake, whereas HDL-cholesterol values increased (Table 1). No significant differences were evident for aspartate aminotransferase, alanine aminotranferase, and prealbumin values. Intakes of total calories, cholesterol, fat, fiber, potassium, calcium, zinc, vitamin C, and vitamin E were significantly higher in persons with the highest magnesium intake (Table 1).
View this table:
TABLE 1. Clinical, dietary, and laboratory characteristics according to tertiles of magnesium intake1
Prevalence of diabetes, MetS and its components (with the exception of low HDL cholesterol), hs-CRP 3 mg/L, and subjects within the highest GGT tertile significantly decreased from the lowest to the highest tertile of magnesium intake, and from the lowest to the highest tertile of fiber intake (Table 2). Dietary magnesium and fiber intakes were highly correlated (Figure 1) (correlation coefficient = 0.81). The partial Pearson coefficient (r) between HOMA-IR and fiber, adjusted for magnesium intake, was r = –0.09, P < 0.001; whereas the partial correlation between HOMA-IR and magnesium, adjusted for fiber intake, was r = –0.02, P = 0.37.
View this table:
TABLE 2. Prevalence of metabolic abnormalities according to tertiles of magnesium (upper) and fiber (lower) intakes1
FIGURE 1.. Linear correlation between magnesium and fiber intakes in the entire cohort (n = 1653). Pearson coefficient = 0.81; SE = 0.015; P < 0.001.
Subjects with the lowest magnesium intake were 3–4 times more likely to have diabetes mellitus, hs-CRP 3 mg/L, and the MetS than those within the highest tertile of magnesium intake, after controlling for demographic, clinic, and environmental characteristics (Table 3). However, after additional adjustment for fiber intake, associations with diabetes and the MetS were no longer significant, whereas the association between hs-CRP 3 mg/L and the lowest tertile of magnesium intake was significant. However, fiber intake remained inversely associated with diabetes, hs-CRP 3 mg/L, and the MetS, after adjustments for age, sex, BMI, smoking habits, alcohol intake, physical activity, and intake of total calories, total percentage of fat, and magnesium (Table 3). The interactions between categorical magnesium and fiber intakes on diabetes (LR test: 2.74; P = 0.60), hs-CRP 3 mg/L (LR test: 2.89; P = 0.58), and the MetS (LR test: 3.68; P = 0.4) were not significant.
View this table:
TABLE 3. Association of tertiles of magnesium and fiber intakes with diabetes mellitus, high-sensitivity C-reactive protein (hs-CRP) 3 mg/L, and the metabolic syndrome in a multiple logistic regression model
To avoid any possible bias because of a particular diet on magnesium consumption or drugs on metabolic or inflammatory variables, analyses were performed after excluding from the analysis subjects with known hyperglycemia, lipid abnormalities, hypertension, CVD, impaired renal function, liver or gastrointestinal diseases, or on any medical therapy (n = 456). In a multiple logistic regression model subjects within the lowest tertile of magnesium intake were 1.5–5 times more likely to have diabetes mellitus, hs-CRP 3 mg/L, and MetS. However, after adjustment for fiber intake, association with diabetes and MetS was no longer significant (data not shown), whereas association with hs-CRP 3 mg/L was of borderline significance (OR: 1.67; 95% CI: 1.00, 2.81). In the same model, the lowest tertile of fiber intake was significantly associated with diabetes, MetS, and hs-CRP 3 mg/L (data not shown). Data did not change significantly after adjustments for glycemic index and dietary intake of calcium, potassium, zinc, vitamin E, and vitamin C or after excluding persons with hs-CRP > 10 mg/L, with CVD, or with low BMI (<19; n = 13).
Finally, the association of tertiles of magnesium and fiber intakes with hs-CRP 3 mg/L was evaluated in the subgroup of healthy persons with normal BMI with no metabolic abnormality (n = 205). Hs-CRP values were inversely associated with magnesium (correlation coefficient = –0.48) and fiber intake (correlation coefficient = –0.46). In a logistic regression model, after adjustments for age, sex, BMI, smoking, alcohol intake, level of physical activity, dietary intake of total calories, percentage of total fat, and fiber, tertiles of magnesium intake were inversely associated with hs-CRP 3 mg/L (OR: 0.26; 95% CI: 0.10, 0.60; P = 0.002 from the lowest to the highest tertile of magnesium intake). Similarly, tertiles of fiber intake were inversely associated with hs-CRP 3 mg/L, after adjusting for multiple confounding variables and magnesium intake (OR: 0.15; 95% CI: 0.08, 0.31; P < 0.001 from the lowest to the highest tertiles of fiber intake). Interactions between magnesium and fiber intakes on hs-CRP 3 mg/L were not significant (LR test: 2.24; P = 0.69).
Patients with CVD from the entire group (n = 82) showed significantly lower fiber and magnesium intakes than did subjects without CVD (14.1 ± 7.0 compared with 21.0 ± 8.4 g/d, P < 0.001; 240.2 ± 74.0 compared with 324.7 ± 92.0 mg/d, P < 0.001, respectively). This was also true for those patients with unknown CVD (n = 17), established by electrocardiogram and Rose questionnaire (their fiber and magnesium intakes were 12.5 ± 5.0 g/d and 247.9 ± 47.4 mg/d, respectively).
DISCUSSION
Our cohort showed lower than recommended magnesium intake, in line with assumptions found in nationally representative US surveys (15), the main contributors being the lower intake of vegetables and fruit, which were also responsible for the reduced intake of fiber in these persons. Reduced fiber intake was significantly associated with diabetes, the MetS, and hs-CRP 3 mg/L, after adjustments for multiple confounders and magnesium intake. Inverse associations between the lower tertile of magnesium intake with metabolic variables and inflammation were greatly reduced by the introduction of fiber intake into the model. Magnesium intake was highly correlated with fiber consumption in the study population, because of the similarity in the foods providing both fiber and magnesium. However, no significant interaction between fiber and magnesium resulted from the analyses, and the association between magnesium and hs-CRP and between fiber and inflammatory and metabolic variables remained significant in the multivariate analyses, after adjustments for fiber or magnesium intakes, respectively. In contrast to other studies on magnesium intake (2, 3), the dietary pattern, with the exception of fiber, was worse in our subjects with the highest magnesium intake (higher percentage of cholesterol and fat). This suggests an even more important role for foods with high content of fiber and magnesium. However, the association between the intake of magnesium and fiber and metabolic abnormalities did not change after adjustments for intakes of other metals or antioxidant vitamins; this might imply a specific role for the nutrients in these relations. Because the cross-sectional design of the study established no definite causal or temporal relations, as well as the extreme difficulty for any observational study to tease out the independent effect from a nutrient rather than other correlated factors sharing the same main food sources, these results might suggest a primary role for reduced fiber intake on metabolic abnormalities and an implication for both fiber and magnesium deficiencies in inflammatory variables. Accordingly, the more favorable impact of a high-fiber diet from whole grain might be due to its higher magnesium content, which might be additionally beneficial. In fact, magnesium is found primarily in bran and germ, most of which are removed in the refined grain (18). Our results are in accordance with the inconsistencies in dietary intervention studies with magnesium on metabolic outcomes: only a modest or inconsistent response has been reported by many of them (10, 31-36). Furthermore, magnesium supplements do not reduce diabetes risk (3), and, generally, dietary management of a disease may be more effective when the focus is on the overall nutritional profile rather than the single-nutrient intake (36). No different results were found in clinical studies that involved antioxidant agents: antioxidant vitamins show either positive or neutral effect ("the antioxidant paradox"), because protection might derive from multiple nutrients rather than from the effect of a single antioxidant (37).
Many prospective studies have reported a protective effect of increased intake of dietary fiber, particularly whole grains, on reducing diabetes, metabolic and cardiovascular risk factors, and CRP values (17, 19, 20, 24, 38, 39). Fiber intake has shown to improve insulin sensitivity by a delayed rate of carbohydrate absorption, weight gain prevention, a longer feeling of satiety, and decrease in inflammation and oxidative stress (19, 20, 38, 39). Finally, fiber intake has been associated with a reduced incidence of CVD in many studies (40-42). Accordingly, fiber intake in our patients was inversely correlated with insulin resistance, evaluated by the HOMA-IR index.
Magnesium is essential for all energy-dependent transport systems, glycolysis, and oxidative energy metabolism and participates in intracellular signaling systems, insulin receptor activity, and phosphorylation and dephosphorylation reactions. It is considered a physiologic calcium-blocker; acts on platelet aggregation, vascular smooth muscle tone as a relaxant, and electrolyte homeostasis; and exerts antiarrhythmic effects (6).
Our results suggest a correlation between dietary fiber and magnesium and GGT values. GGT has been reported as a marker of fatty liver deposition (a condition clearly related to insulin resistance) but also to oxidative stress (43, 44). A recent study reported an inverse independent association between decreased serum magnesium values and nonalcoholic steatohepatitis in 80 obese Mexicans (45). It was hypothesized that by increasing the permeability of mitochondrial membranes, magnesium deficiency could decrease oxidation-phosphorylation coupling (46), involved in lipid peroxidation and cytokine induction, which are implicated in the pathogenesis of steatohepatitis (47). Low-grade chronic inflammation and insulin resistance might be other possible mechanisms that are both associated with hypomagnesaemia (45).
Implications for inflammation
A few studies (8, 14, 15) reported an inverse association between dietary and serum values of magnesium and CRP; however, adjustments for fiber intake either do not (8) or do (15) attenuate these associations. In our study, the inverse association between the lowest magnesium tertile and hs-CRP 3 mg/L remained significant, after adjustments for fiber intake. Differences from previous studies on non-European populations might be due to the different dietary pattern (eg, the consumption of magnesium-rich foods, such as nuts or whole grain, is rare in Italy). Differently from others (8, 15) we found significant relations between dietary magnesium and hs-CRP even in patients with BMI < 25 with no metabolic abnormalities. Because of the strong association between hs-CRP with adiposity and all the MetS components, these conditions might represent confounding factors. Indeed, the inverse association between magnesium intake and hs-CRP values in subjects with no MetS components would suggest an explanation for most low-grade inflammation states in this group, which are otherwise more difficult to explain. Accordingly, a substantial protective effect of healthy foods was shown among nonobese persons (18).
Mechanisms relating low magnesium intake to elevated CRP values might be related to oxidative stress or endothelial dysfunction because magnesium deficiency inhibits endothelial growth and migration and stimulates the synthesis of nitric oxide and some inflammatory markers in vivo, thus directly modulating microvascular functions (48, 49). Finally, magnesium supplementation significantly improves endothelial function of the brachial artery in patients with CVD (50), and serum and dietary magnesium values are inversely associated with mean carotid wall thickness (4). Accordingly, our patients with CVD (both known and unknown) showed significantly lower fiber and magnesium intakes and, intriguingly, lower consumption of dietary cholesterol and fat. Low-grade inflammation might be a possible mediator in the associations found between CVD and dietary magnesium and fiber intakes (15, 38, 39, 48).
Limitations and strengths
The cross-sectional design of the study is, by its nature, hypothesis generating, thus requiring further testing in prospective analyses. Simultaneous adjustments for highly correlated nutrients might not be the best approach to identify the effect of a single nutrient independent of another, because with increasing the degree of collinearity, both the regression coefficients and their standard errors tend to be unstable, making interpretation results difficult. However, no effect modifications were found in our analyses, after adjusting for fiber and magnesium intakes, respectively, but only a reduction in the association strength. Even if the subjects studied were not obtained by a formal sampling procedure, their representativeness of the corresponding population for several characteristics is reassuring. Dietary assessments were performed with the use of a semiquantitative food-frequency questionnaire, and validity and reliability of magnesium intake assessment are unknown. Because fiber intake was determined by the algorithm used in the European Prospective Investigation into Cancer and Nutrition studies and magnesium according to specific food composition tables, 2 different types of misclassification might have been made. Separate data on types of dietary fiber were not available. With the exception of glycemia, only one determination for metabolic variables and hs-CRP was available. Measurement errors could have biased the results toward the null. Even if data have been adjusted for various potential confounders, the possibility of uncontrolled or unknown confounders cannot be ruled out. The strengths of this study include the large number of population-based subjects studied from a localized region, which limits the number of variables, the use of a central laboratory, and the high level of participation.
Conclusions
Multiple metabolic abnormalities were significantly associated with reduced fiber intake. Reduced magnesium and fiber intakes were independently associated with hs-CRP values in the entire cohort and the subsample of persons with normal metabolic variables. The association in the latter group might give a possible explanation for the low-grade inflammation state of some of these persons otherwise less easy to explain.
ACKNOWLEDGMENTS
We thank Prof Paolo Vineis for his precious help in carrying out the study.
SB conceived and designed the study, supervised the data collection, analyzed the data analysis, interpreted the study findings, and wrote and revised the manuscript; MD analyzed the data and wrote and revised the manuscript; SG and CS interpreted the study findings and wrote and revised the manuscript; M Carello collected the data, interpreted the findings, and revised the manuscript; BS, M Cassader, and LG collected the data and revised the manuscript; RR analyzed the data, interpreted the study findings, and revised the manuscript; GP conceived and designed the study, interpreted the study findings, and wrote and revised the manuscript. None of the authors had a conflict of interest.
REFERENCES