Larsson SC, Virtanen MJ, Mars M, Männistö S, Pietinen P, Albanes D, Virtamo J. Magnesium, Calcium, Potassium, and Sodium Intakes and Risk of Stroke in Male Smokers. Arch Intern Med. 2008;168(5):459-465. doi:10.1001/archinte.168.5.459
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
A high intake of magnesium, calcium, and potassium and a low intake of sodium have been hypothesized to reduce the risk of stroke. However, prospective data relating intake of these minerals to risk of stroke are inconsistent.
We examined the relationship of dietary magnesium, calcium, potassium, and sodium intake with risk of stroke in a cohort of 26 556 Finnish male smokers, aged 50 to 69 years, who were free from stroke at baseline. Dietary intake was assessed at baseline using a detailed and validated food frequency questionnaire. During a mean follow-up of 13.6 years (1985-2004), 2702 cerebral infarctions, 383 intracerebral hemorrhages, and 196 subarachnoid hemorrhages were identified in the national registries.
After adjustment for age and cardiovascular risk factors, a high magnesium intake was associated with a statistically significant lower risk of cerebral infarction but not with intracerebral or subarachnoid hemorrhages. The multivariate relative risk of cerebral infarction was 0.85 (95% confidence interval, 0.76-0.97; P for trend = .004) for men in the highest quintile of magnesium intake compared with those in the lowest quintile. The inverse association between magnesium intake and cerebral infarction was stronger in men younger than 60 years (relative risk, 0.76; 95% confidence interval, 0.64-0.89; P for interaction = .02). Calcium, potassium, and sodium intake was not significantly associated with risk of any subtype of stroke (P for trend > .05).
These findings in male smokers suggest that a high magnesium intake may play a role in the primary prevention of cerebral infarction.
Stroke is an important cause of mortality in developed countries. Even among stroke survivors, quality of life could be reduced substantially in many who are left with permanent disability. Therefore, identification of risk factors that could be modified could have a marked impact on reducing stroke morbidity and mortality.
Recent studies have shown that dietary modifications are an important means of preventing stroke.1 Because hypertension is a strong risk factor for stroke,2 dietary factors that influence blood pressure levels may affect the risk of stroke. Magnesium, calcium, and potassium intake has been inversely related to blood pressure and hypertension in several observational studies,3- 6 and some, though not all, randomized controlled trials have demonstrated reduction in blood pressure levels in persons receiving supplementation with these minerals alone or in combination.7- 10 Sodium intake, however, has been positively associated with blood pressure, and there is evidence that reduced sodium intake leads to a modest reduction in blood pressure levels.11,12 Nevertheless, the question of whether magnesium, calcium, potassium, and sodium intake is associated with risk of stroke remains controversial. We used prospective data from the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study13 to examine the association of magnesium, calcium, potassium, and sodium intake with risk of stroke in Finnish male smokers.
The ATBC Study was a randomized, double-blind, placebo-controlled, primary prevention trial that was designed to test whether the use of α-tocopherol (50 mg/d) or beta carotene (20 mg/d) could reduce lung cancer incidence in male smokers who were recruited from southwestern Finland between 1985 and 1988.13 The cohort consisted of 29 133 men, aged 50 to 69 years, who smoked 5 or more cigarettes per day at baseline. The trial continued until April 1993, with the cohort followed up through national registers thereafter. Men were excluded from the trial if they (1) had a history of cancer (other than nonmelanoma skin cancer or carcinoma in situ) or other serious disease that might limit long-term participation; (2) received anticoagulant therapy; or (3) used vitamin E, vitamin A, or beta carotene supplements in excess of predefined doses. For the present analyses, we also excluded men with incomplete dietary data and those with a self-reported history of stroke at baseline, leaving 26 556 men for the analyses.
Written informed consent was obtained from each participant before randomization. The study was approved by the institutional review boards of the National Public Health Institute of Finland, Helsinki, and the National Cancer Institute, National Institutes of Health, Bethesda, Maryland.
At baseline, participants completed a questionnaire on background characteristics and medical, smoking, and physical activity histories.13 Height, weight, and blood pressure levels were measured, and a blood sample was drawn and stored at −70°C. Serum levels of total cholesterol and high-density lipoprotein (HDL) cholesterol were determined enzymatically (CHOD-PAP method; Boehringer Mannheim, Mannheim, Germany).
Diet was assessed at baseline using a validated self-administered food frequency questionnaire that included 276 food items and mixed dishes commonly consumed in Finland.14 The questionnaire was used with a portion-size color picture booklet of 122 photographs of foods, each with 3 to 5 different portion sizes. Participants were asked to report their average consumption and portion size for each food during the past year. Frequencies were reported as the number of times per month, week, or day. Nutrient intake was calculated by multiplying the frequency of consumption of each food by the average nutrient content of the specified portion size. Values for the amounts of nutrients in the foods were based on the food composition database at the National Public Health Institute. The amount of salt added at the table was not collected in the ATBC Study. The salt used in cooking was included in the average recipes of mixed dishes as an ingredient, and the recipe file was used in all the nutrient calculations. This method of calculating total sodium intake has been shown to give a valid estimate of total sodium intake in population groups.15,16 The validity of the food frequency questionnaire has been evaluated previously.14 The energy-adjusted correlation between the first questionnaire and the food records was 0.6 for magnesium, 0.7 for calcium, 0.6 for potassium, and 0.6 for sodium.14
The end point of the study was first-ever stroke that occurred between the date of randomization and December 31, 2004. The strokes were further divided into cerebral infarction, intracerebral hemorrhage, subarachnoid hemorrhage, and unspecified stroke. The end points were identified by record linkage with the National Hospital Discharge Register and the National Register of Causes of Death. Both registers used the codes of the International Classification of Diseases (ICD): the 8th edition was used until the end of 1986, the 9th edition through the end of 1996, and the 10th edition thereafter. The end points comprised ICD-8 codes 430 through 434 and 436; ICD-9 codes 430, 431, 433, 434, and 436; and ICD-10 codes I60, I61, I63, and I64, excluding ICD-8 codes 431.01 and 431.91 denoting subdural hemorrhage and ICD-9 codes 4330X, 4331X, 4339X, and 4349X representing occlusion or cerebral or precerebral artery stenosis without cerebral infarction. In a reviewed sample, the diagnoses of cerebral infarction, subarachnoid hemorrhage, and intracerebral hemorrhage proved correct by strict preset criteria in 90%, 79%, and 82% of the discharge diagnoses and in 92%, 95%, and 91% of the causes of death, respectively.17
Each participant contributed person-time of follow-up from the date of randomization until the date of occurrence of first stroke, death, or December 31, 2004, whichever came first. When the risk estimates for a stroke subtype were analyzed, the other subtypes were treated as censored. All nutrients were energy adjusted using the residual method.18 Magnesium, calcium, potassium, and sodium intake was categorized into quintiles based on the distribution among the study population. Because calcium from supplements contributed only 2% to total calcium intake in this population,19 we focused on dietary calcium. Cox proportional hazards models20 were used to estimate relative risks (RRs) with 95% confidence intervals (CIs). All models were adjusted for age at baseline and supplementation group (α-tocopherol, beta carotene, or both or placebo). Multivariate models were further controlled for cardiovascular risk factors (alcohol intake, number of cigarettes smoked daily, body mass index, systolic and diastolic blood pressure levels, serum total and HDL cholesterol levels, histories of diabetes and coronary heart disease, and leisure-time physical activity) and total energy intake. Additional adjustment for dietary variables, including folate, vitamin C, vitamin E, fat, carbohydrate, protein, and fiber, did not change the results appreciably; therefore, these variables were not included in the main multivariate model. Tests based on Schoenfeld residuals and graphical methods using Kaplan-Meier curves21 showed no evidence that the proportional hazards assumption was violated.
Tests for trend were conducted by assigning the medians of mineral intake in quintiles treated as a continuous variable. Effect modification was examined in stratified analyses and was statistically tested by including the cross-product term of the mineral variable (modeled as a continuous variable) and the effect modifier (as a dichotomous variable). Statistical analyses were performed using Stata software, version 9.2 (StataCorp, College Station, Texas). All P values are 2-sided; P < .05 was considered statistically significant.
Among the 26 556 men who were followed up for 360 187 person-years (mean, 13.6 years), we ascertained 2702 cerebral infarctions, 383 intracerebral hemorrhages, 196 subarachnoid hemorrhages, and 84 unspecified strokes. Baseline characteristics of the study population according to dietary intake of magnesium, calcium, potassium, and sodium are shown in Table 1. In general, men with higher magnesium intake had a higher body mass index, were more likely to have a history of diabetes or coronary heart disease, and were more physically active than men with lower magnesium intake. They also had a higher intake of other minerals, folate, vitamins C and E, dietary fiber, fruits, vegetables, and cereals but a lower intake of alcohol and saturated fat. Men with a higher calcium, potassium, and sodium intake showed similar characteristics to those who had a higher magnesium intake. The highest correlations among minerals were those between magnesium and potassium (r = 0.73), calcium and potassium (r = 0.41), and potassium and sodium (r = 0.35).
The RRs of stroke subtypes according to dietary intake of magnesium, calcium, potassium, and sodium are shown in Table 2. Magnesium intake was statistically significantly inversely associated with risk of cerebral infarction after age, supplementation group, cardiovascular risk factors, and energy intake were controlled for. The multivariate RR of cerebral infarction comparing the highest with the lowest quintile of magnesium intake was 0.85 (95% CI, 0.76-0.97). This association remained after additional adjustment for intake of folate, vitamin C, vitamin E, saturated fat, polyunsaturated fat, and dietary fiber (RR, 0.84; 95% CI, 0.72-0.98). There was no dose-response relationship between magnesium intake and risk of intracerebral or subarachnoid hemorrhages, but the risk of subarachnoid hemorrhage was elevated in the third and fourth quintiles. Potassium intake was statistically significantly inversely associated with the risk of cerebral infarction after adjustment for age and supplementation group only (P for trend = .002), but this association was attenuated after further adjustment for cardiovascular risk factors (P for trend = .05). When magnesium and potassium intake was included simultaneously in the multivariate model, the RRs of cerebral infarction for the highest vs the lowest quintile of intake were 0.86 (95% CI, 0.76-1.00) for magnesium and 1.02 (95% CI, 0.87-1.20) for potassium. The multivariate RR of cerebral infarction comparing men in the highest quartile of both magnesium and potassium intake with those in the lowest quartile of both minerals was 0.87 (95% CI, 0.76-1.00). Neither calcium nor sodium intake was significantly associated with risk of any stroke subtype (P for trend > .05).
We further examined calcium intake from dairy and nondairy sources in relation to risk of stroke and found an inverse relationship between nondairy calcium and cerebral infarction (highest vs lowest quintile: multivariate RR, 0.86; 95% CI, 0.76-0.96). However, this association did not persist after further adjustment for intake of folate, vitamin C, vitamin E, saturated fat, polyunsaturated fat, and dietary fiber (RR, 0.96; 95% CI, 0.84-1.10). We considered the possibility that the effect of dietary magnesium, calcium, potassium, and sodium intake on stroke risk might be mediated through blood pressure and that adjustment for blood pressure in our multivariate models might minimize potential associations. Excluding systolic and diastolic blood pressure levels from the multivariate model did not materially alter the relationship between these minerals and risk of stroke (data not shown).
The association between magnesium intake and risk of cerebral infarction was significantly modified by age (P for interaction = .02) but not by other cardiovascular risk factors (those listed in Table 1 and hypertension) or supplementation group (P for interaction >.05 for all). The multivariate RRs of cerebral infarction for the highest vs the lowest quintile of magnesium intake were 0.76 (95% CI, 0.64-0.89) among men younger than 60 years and 1.02 (95% CI, 0.84-1.23) among those 60 years and older.
To address the potential for increased exposure misclassification over time, we divided the follow-up time into 5-year periods. The multivariate RRs of cerebral infarction in the highest quintile of magnesium intake compared with the lowest were 1.08 (95% CI, 0.83-1.41) during years 0 through 5, 0.80 (95% CI, 0.64-1.00) during years 6 through 10, 0.74 (95% CI, 0.59-0.93) during years 11 through 15, and 1.04 (95% CI, 0.78-1.39) during years 16 through 20.
In this cohort study of middle-aged male smokers, we found that a high magnesium intake was associated with a significant reduced risk of cerebral infarction that was not accounted for by other potential risk factors. Intake of calcium, potassium, and sodium was not significantly associated with risk of any stroke subtype after potential confounders were controlled for.
An inverse association between magnesium intake and cerebral infarction is biologically plausible. A recent meta-analysis of 12 randomized clinical trials showed that magnesium supplementation may slightly reduce diastolic blood pressure by 2.2 mm Hg.9 Therefore, a potential hypotensive effect of magnesium intake is small and could only partially explain the inverse association of magnesium intake with cerebral infarction. Furthermore, adjustment for baseline blood pressure had little effect on the estimated RR relating magnesium intake to cerebral infarction. Besides a hypotensive effect, magnesium supplementation had favorable effects on plasma glucose, triglyceride, HDL, low-density lipoprotein, very-low-density lipoprotein, and total cholesterol levels in rats with chronic diabetes.22 There are also reports showing that magnesium deficiency increases the susceptibility of lipoproteins to peroxidation in animals.23 In cross-sectional studies, dietary magnesium intake has been found to be inversely associated with markers of systematic inflammation and endothelial dysfunction, carotid artery thickness, fasting insulin and glucose concentrations, and the metabolic syndrome.24- 26 Also, a meta-analysis of cohort studies showed that a high magnesium intake may lower the risk of type 2 diabetes mellitus,27 which in a recent large cohort study was associated with an increased risk of ischemic stroke but not with hemorrhagic stroke.28 In our study, magnesium intake was inversely associated with risk of cerebral infarction but not with hemorrhagic stroke. In this cohort, we previously found that the risk factor profiles of stroke subtypes differ.29 For example, serum total cholesterol concentrations were positively associated with risk of cerebral infarction only, and serum HDL cholesterol concentrations were inversely related to risk of cerebral infarction and subarachnoid hemorrhage but not to intracerebral hemorrhage.29 Therefore, if magnesium reduces stroke risk by influencing cholesterol concentrations or insulin resistance, the beneficial effect of high magnesium intake may be limited to cerebral infarction.
There are limited prospective data on magnesium intake in relation to risk of stroke. In the Health Professionals Follow-up Study of 43 738 US men, 328 strokes were documented during 8 years of follow-up.30 In that study, magnesium intake was significantly inversely associated with risk of total stroke among men with hypertension.30 No significant association was found between magnesium intake and risk of total stroke or ischemic stroke in the Nurses' Health Study (including 690 strokes)31 or in the Women's Health Study (368 strokes).32
Potassium intake was also inversely associated with risk of cerebral infarction in our study, but this association was substantially weakened and did not remain in a multivariate model that simultaneously included magnesium and potassium. However, the strong positive correlation between these minerals and the inevitable measurement error in dietary assessment reduced the ability of the multivariate analysis to discriminate between them. Potassium intake was not significantly associated with risk of stroke in the Nurses' Health Study after adjustment for cardiovascular risk factors and calcium intake.31 In the Health Professionals Follow-up Study, potassium intake was inversely associated with stroke risk among hypertensive men.30 A low potassium intake was related to increased risk of stroke in the National Health and Nutrition Examination Survey Epidemiologic Follow-up Study33,34 and among nonusers of diuretics in a cohort of 5600 older men and women.35
As in our study, no association between calcium intake and stroke was found in the Health Professionals Follow-up Study.30 The Nurses' Health Study31 showed an inverse association between intake of calcium, especially dairy calcium, and stroke risk. Likewise, dairy calcium intake was inversely associated with stroke mortality (n = 566 deaths) in a cohort of Japanese men and women with low total calcium intake (median daily intake of 449 mg among men and 462 mg among women).36 In a recent randomized trial including 36 282 postmenopausal women, combined calcium and vitamin D supplementation neither decreased nor increased the risk of stroke over a 7-year use period.37
Findings from previous studies that have examined the relationship between sodium intake and stroke risk have been inconsistent. Specifically, one prospective study from Finland38 and one Japanese cohort39 reported no association. However, another Japanese cohort study showed a significant increased risk of mortality from stroke among men and women in the highest tertile of sodium intake (median daily intake of 7194 mg among men and 6478 mg among women),40 and the National Health and Nutrition Examination Survey Epidemiologic Follow-up Study found a positive association among overweight persons.41
The dietary intake of magnesium, calcium, and potassium was high in our study population compared with other populations,30- 32,34- 36 most likely because of both methodological and cultural differences. The relatively long food frequency questionnaire (276 items) that was used to assess dietary intake in our population probably resulted in a slight overestimation of energy intake, whereas a shorter food frequency questionnaire and a 24-hour recall (eg, which were used in the National Health and Nutrition Examination Survey34) tend to result in an underestimation. Also, in our population, high consumption of whole-grain cereals explains the high magnesium intake, and high consumption of milk, coffee, and potatoes explain the high potassium intake. Differences in intake levels of minerals between populations may explain the inconsistent results. The prevalence of diabetes and coronary heart disease was higher among men in the highest quintile of mineral intake than among those in the lowest quintile (Table 1). This difference probably reflects dietary changes after diagnosis of these diseases.
Several strengths and potential limitations of this study deserve comment. An advantage is the prospective design, which precluded the possibility for recall bias in the dietary assessment. Furthermore, the extensive information on cardiovascular risk factors allowed comprehensive adjustment for potential confounders, and the large number of cases provided high statistical power to detect associations.
Measurement error in the assessment of dietary intake is inevitable and will have led to some misclassification of the studied dietary exposures. We had dietary intake assessed only at baseline, which may have contributed to misclassification because of dietary changes during follow-up. Given the prospective design, this misclassification was unlikely to be associated with the studied outcomes and therefore probably led to underestimation of the associations. The intake of calcium from dietary supplements was negligible compared with total intake from diet19 and was therefore unlikely to have influenced the results. Using the same dietary assessment, high calcium intake was associated with reduced risk of colorectal cancer42 and increased risk of prostate cancer19 in the ATBC Study. Consequently, it is unlikely that the lack of observed association can be attributed to our inability to measure calcium intake. Although our dietary assessment provides valid estimates of dietary sodium intake on a group level, there may be misclassification on an individual level, which would lead to attenuation of the RR estimates toward the null. Another potential limitation is that the ATBC Study consisted entirely of male smokers; therefore, our findings may not be generalizable to women or to nonsmokers.
In summary, in this cohort of male smokers, a high magnesium intake was associated with a significantly decreased risk of cerebral infarction. While the biologic mechanism explaining this relationship is unclear, our findings suggest that a high consumption of magnesium-rich foods, such as whole-grain cereals, may play a role in the prevention of cerebral infarction. Whether magnesium supplementation lowers the risk of cerebral infarction needs to be assessed in large, long-term randomized trials.
Correspondence: Susanna C. Larsson, PhD, Division of Nutritional Epidemiology, National Institute of Environmental Medicine, Karolinska Institutet, Box 210, SE-171 77 Stockholm, Sweden (Susanna.Larsson@ki.se).
Accepted for Publication: October 15, 2007.
Author Contributions: Dr Larsson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Larsson, Mars, Pietinen, Albanes, and Virtamo. Acquisition of data: Albanes, Pietinen, Albanes, and Virtamo. Analysis and interpretation of data: Larsson, Virtanen, Mars, Männistö, and Virtamo. Drafting of the manuscript: Larsson and Virtanen. Critical revision of the manuscript for important intellectual content: Virtanen, Mars, Männistö, Pietinen, Albanes, and Virtamo. Statistical analysis: Larsson and Virtanen. Obtained funding: Larsson, Pietinen, Albanes, and Virtamo. Administrative, technical, or material support: Virtanen, Albanes, and Virtamo. Study supervision: Virtamo.
Financial Disclosure: None reported.
Funding/Support: The ATBC Study was supported by Public Health Service contracts N01-CN-45165, N01-RC-45035, and N01-RC-37004 from the National Cancer Institute. Dr Larsson's postdoctoral research at the National Public Health Institute was supported by a grant from the Swedish Council for Working Life and Social Research.
Role of the Sponsor: The funding sources had no role in the design, conduct, or reporting of this study.