Magnesium, vitamin D status and mortality: results from US National Health and Nutrition Examination Survey (NHANES) 2001 to 2006 and NHANES III

Background Magnesium plays an essential role in the synthesis and metabolism of vitamin D and magnesium supplementation substantially reversed the resistance to vitamin D treatment in patients with magnesium-dependent vitamin-D-resistant rickets. We hypothesized that dietary magnesium alone, particularly its interaction with vitamin D intake, contributes to serum 25-hydroxyvitamin D (25(OH)D) levels, and the associations between serum 25(OH)D and risk of mortality may be modified by magnesium intake level. Methods We tested these novel hypotheses utilizing data from the National Health and Nutrition Examination Survey (NHANES) 2001 to 2006, a population-based cross-sectional study, and the NHANES III cohort, a population-based cohort study. Serum 25(OH)D was used to define vitamin D status. Mortality outcomes in the NHANES III cohort were determined by using probabilistic linkage with the National Death Index (NDI). Results High intake of total, dietary or supplemental magnesium was independently associated with significantly reduced risks of vitamin D deficiency and insufficiency respectively. Intake of magnesium significantly interacted with intake of vitamin D in relation to risk of both vitamin D deficiency and insufficiency. Additionally, the inverse association between total magnesium intake and vitamin D insufficiency primarily appeared among populations at high risk of vitamin D insufficiency. Furthermore, the associations of serum 25(OH)D with mortality, particularly due to cardiovascular disease (CVD) and colorectal cancer, were modified by magnesium intake, and the inverse associations were primarily present among those with magnesium intake above the median. Conclusions Our preliminary findings indicate it is possible that magnesium intake alone or its interaction with vitamin D intake may contribute to vitamin D status. The associations between serum 25(OH)D and risk of mortality may be modified by the intake level of magnesium. Future studies, including cohort studies and clinical trials, are necessary to confirm the findings.

Magnesium, the second most abundant intracellular cation, plays a critical role in the synthesis and metabolism of parathyroid hormone (PTH) and vitamin D [21][22][23]. Previous studies have shown that the activities of three major enzymes determining 25(OH)D level [22][23][24][25] and vitamin D binding protein [23] are magnesium dependent (Figure 1). Magnesium deficiency, which leads to reduced 1,25(OH) 2 vitamin D and impaired PTH response [23], has been implicated in 'magnesium-dependent vitamin-D-resistant rickets' [21]. Magnesium supplementation substantially reversed the resistance to vitamin D treatment [21]. Interestingly, a study conducted among osteoporosis patients showed much higher prevalence rates of magnesium deficiency or insufficiency among people with insufficient 25(OH)D than those with sufficient 25(OH)D [26]. Two small clinical trials of magnesium-deficient patients [23,27] found that magnesium infusion alone led to a nonsignificant increase in 1,25(OH) 2 D and 25(OH)D [23] whereas magnesium infusion plus oral vitamin D substantially increased both serum 25(OH)D and 1,25 (OH) 2 D [27]. These findings suggest a potential interaction between vitamin D and magnesium treatments We hypothesize that magnesium intake alone and particularly its interaction with vitamin D intake contribute to serum 25(OH)D status and tested this novel hypothesis utilizing data from the National Health and Nutrition Examination Survey (NHANES) 2001 to 2006. Furthermore, previous studies reported that serum 25(OHD) concentrations were associated with reduced risks of total mortality, particularly mortality due to colorectal cancer [28] and CVD [4,29]. We hypothesize that the inverse associations between serum 25(OH)D and risk of mortality are modified by intake level of magnesium and tested this hypothesis using the NHANES III cohort.

Outcomes
We have utilized the cut-off points as defined in the recently issued Institute of Medicine (IOM) report to define vitamin D deficiency and insufficiency [33]. We classified the participants in the NHANES 2001 to 2006 into the following three categories: (1) participants who had IOM recommended vitamin D level (serum 25(OH) D ≥20 ng/ml (50 nmol/l)); (2) participants who had insufficient vitamin D (serum 25(OH)D ≥12 ng/ml (30 nmol/l) but below 20 ng/ml (50 nmol/l)); and (3) participants who had vitamin D deficiency (serum 25 (OH)D <12 ng/ml (30 nmol/l)). In addition to Institute of Medicine's recommendation, the other reason we used 20 ng/ml as the cut-off point is that the median value of serum 25(OH)D for the NHANES 2001 to 2006 was 21.0 ng/mg, which is very close to 20 ng/ml.
In the NHANES III cohort, mortality outcomes were determined by using probabilistic linkage with the National Death Index (NDI). For mortality due to CVD or colorectal cancer, participants who died of other diseases or were not known deceased, were censored at the date of death or December 31, 2006, whichever was earlier.

Nutrient intake assessments
Daily dietary intake data were obtained from 24-h dietary recalls and 30-day supplement interviews, which are described in detail elsewhere [34]. Only one 24-h recall was conducted in NHANES III and from 2001 to 2002. To keep intake information consistent through the study period only the first dietary recall for all subjects was utilized in the present analysis. Only dietary recall data with a status of 'reliable' were used in the analysis.
Supplemental doses and intakes of magnesium, vitamin D, and calcium were obtained from the response to a dietary supplement questionnaire.

Statistical analysis
We performed statistical analyses using PROC Survey in SAS 9.2 software (SAS Institute, Cary, NC, USA) to estimate variance after incorporating the weights for the sample population in NHANES.
Covariates were compared among three groups with differing vitamin D status (Table 1). Any covariate that significantly differed among groups was considered as a potential confounding factor. In particular, those confounders that altered the estimate of association by 10% or more were retained in the final models (see footnotes to Tables 2 and 3). Other factors, such as use of phosphorus, potassium, and retinol, did not materially alter the risk estimates. Multivariate adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for vitamin D deficiency and insufficiency were calculated. Stratified analyses by potential effect modifiers and tests for multiplicative interactions using the Wald test were conducted.
Hazard ratios (HR) were estimated in Cox proportional hazard regression models to examine whether the associations between serum 25(OH)D and risk of mortality differed by intake level of total magnesium, using the data from the NHANES III cohort. The same cut-off points for serum 25(OH)D were used as those in the published report [28]. We adjusted for the same confounding factors as we did for vitamin D deficiency and insufficiency analyses. All reported P values were two-sided. P values of <0.05 were considered statistically significant in all analyses.
In Table 4, we present the associations between serum 25(OH)D and risk of mortality within two strata (<264 mg/day and ≥264 mg/day) of magnesium intake. However, this is only for data presentation purpose. In statistical modeling, we have used continuous variables for both serum 25(OH)D and intakes of magnesium in the tests for interactions between serum 25(OH)D and intakes of magnesium in relation to total mortality, mortality due to cardiovascular disease and mortality due to colorectal cancer. We chose 264 mg/day as the cut-off point presented in Table 4 based on both statistical power consideration and biological support. We have a limited sample size for colorectal cancer mortality outcome. By using a median magnesium intake of 264 mg/day, we could maximize the sample size, and, thus, the power to detect association between vitamin D levels and mortality outcomes in both strata. Furthermore, 264 mg/day is very close to the 265 mg/day, the estimated average requirements (EAR) for women and so it gave us the opportunity to evaluate the relationship to the EAR [35].

Results
Selected demographic characteristics and potential confounding factors by the three categories of serum 25 (OH)D status are shown in Table 1. Compared to the Estimate Average Requirements (EAR, intake levels for vitamin D (400 IU/day) and magnesium (330 mg/day) recommended by the US Food and Nutrition Board of the Institute of Medicine), the vitamin D normal group generally met the recommended intake levels for both vitamin D and magnesium whereas the average intake levels of these nutrients were significantly lower in the vitamin D-insufficient group and much lower among the vitamin D-deficient group.
After adjusting for confounding factors, the ORs (95% CI) for vitamin D deficiency and insufficiency were 0.10 (0.06 to 0.17) and 0.37 (0.27 to 0.51) comparing the highest quartile intake of total vitamin D versus the lowest (P trend <0.001) ( Table 2). The corresponding ORs (95% CI) were 0.34 (0.21 to 0.56) and 0.62 (0.46 to 0.82) for total magnesium intake (P trend <0.001), respectively. In addition, we found that both dietary and supplemental intakes of vitamin D and magnesium were significantly inversely associated with risks of vitamin D deficiency and insufficiency.
In stratified analyses by intake of vitamin D and other factors related to vitamin D status (Table 3), we found intake of magnesium significantly interacted with intake of vitamin D in relation to both vitamin D deficiency (P interaction , <0.001) and insufficiency (P interaction , <0.001). The inverse association between magnesium intake and risk of vitamin D deficiency only appeared significant among those older than 50 years or with serum PTH level being in the highest or lowest tertile category. Meanwhile, the inverse associations between magnesium intake and risk of vitamin D insufficiency were statistically significant only among people at high risk of vitamin D insufficiency, such as those whose samples were collected during winter (at southern latitude), those with vitamin D intake below the median, women, non-Hispanic Blacks, obese individuals or those with the PTH levels in the highest tertile.
As reported previously, high levels of serum 25(OH)D were associated with a reduced risk of mortality due to all-cause CVD [4,29,36], and colorectal cancer [28]. We found that the inverse associations for higher serum 25 (OH)D with risks of total mortality and mortality due to CVD were only statistically significant among those with magnesium intake above the median (Table 4). Although the test for interaction was not statistically significant for total mortality, it was statistically significant for mortality due to CVD (P interaction , 0.03). Sample size was small for mortality due to colorectal cancer. None of the associations, including the main association, were statistically significant. However, the association pattern in stratified analysis (P interaction , 0.15) was very similar to that for total mortality and mortality due to CVD.

Discussion
Consistent with our hypothesis, we observed that high intake of total, dietary or supplemental magnesium was independently and significantly associated with reduced risks of both vitamin D deficiency and insufficiency. Furthermore, intake of magnesium significantly interacted with intakes of vitamin D in relation to both vitamin D deficiency and insufficiency. In the NHANES III cohort, a population-based prospective study, we found the inverse associations of serum 25(OH)D with mortality, particularly mortality due to CVD and colorectal cancer, were modified by magnesium intake, and the inverse associations were primarily present among those with magnesium intake above the median. In addition, we found the inverse association between magnesium intake and risk of vitamin D deficiency primarily occurred in those who had the highest or the lowest tertile of PTH level; while the inverse association between total magnesium intake and vitamin D insufficiency primarily appeared among populations at high risk of vitamin D insufficiency. To the best of our knowledge, this is the first study to examine the interaction between vitamin D and magnesium in association to mortality; and this is the first study to suggest a potential independent contribution of total magnesium intake and its interaction with vitamin D intake to vitamin D status in the general population. Under normal physiologic conditions, 25(OH)D is derived primarily from endogenous synthesis via exposure of skin to sunlight because few natural foods contain vitamin D except by fortification or supplementation (see Figure 1). Vitamin D 3 or D 2 is transferred to the liver via vitamin D binding protein (VDBP) and converted to 25(OH)D by 25-hydroxylase and subsequently carried to the kidney by VDBP and converted to 1,25(OH) 2 D by 1α-hydroxylase enzyme. Both 25(OH)D and 1,25(OH) 2 D can be converted by 24-hydroxylase to the 24,25(OH) 2 D or 1,24,25(OH) 3 D, respectively [37]. Therefore, 25(OH)D levels are primarily determined by VDBP, 25-hydroxylase, 1α-hydroxylase and 24-hydroxy-lase activity, a fact that has recently been substantiated by a genome-wide association study [38]. Based on previous in vitro studies, magnesium status regulates both 1α-hydroxylase and 24-hydroxylase activity [22,24]. Previous studies indicated both VDBP [23] and 25hydroxylase [25,39] might also be magnesium dependent. Therefore, magnesium would be expected to play an important role in 25(OH)D metabolism.
A previous clinical study found that parenteral magnesium treatment without vitamin D replacement in 23 magnesium-deficient patients led to a 12% rise in 25 (OH)D and 30% increase in 1,25(OH) 2 D, but both changes were not statistically significant [23]. In a subsequent study of five magnesium-deficient patients, intramuscular treatment with magnesium alone also did not significantly increase 25(OH)D, but magnesium infusion together with pharmacological dose of 25(OH)D substantially increased both 25(OH)D and 1,25(OH) 2 D among patients with magnesium deficiency. One interpretation is that magnesium treatment does not affect 25(OH)D status [23,27]. However, we postulate that several factors may have contributed to the insignificant increase in 25(OH)D status. First, the subjects participating in these studies had low concentrations of 25(OH)D and 1,25(OH) 2 D as well as pre-vitamin D 3 and vitamin D 3 as a result of limited sunlight exposure, underlying disease and/or lack of oral supplementation. Therefore, concentrations of 25(OH)D and 1,25(OH) 2 D did not substantially increase during short-term magnesium repletion because pre-vitamin D3 was not available in sufficient amounts. Second, there was a modest increase in the conversion of 25(OH)D to 1,25(OH) 2 D and, thus, a reduction in 25(OH) D level was expected due to this conversion [22]. Finally, the sample size in these two studies was very small particularly if the direct effect of magnesium treatment on vitamin D status is only moderate.
We found that high magnesium intake was also associated with a reduced risk of vitamin D deficiency or insufficiency. We believe that this observation is the result of the interaction between various metabolic pathways that regulate 25(OH)D levels. Previous studies have shown that endogenously synthesized vitamin D 3 is transferred almost completely by VDBP to liver and this    transport is slow, leading to a more sustained plasma vitamin D 3, compared to that from supplementation of vitamin D, which is delivered to the liver by non-VDBP carriers in the plasma [40]. VDBP may also be an important determinant of serum 25(OH)D concentration, particularly when dietary intake of vitamin D is low. In the study by Rude et al., the concentration of VDBP was lower among 11 magnesium-deficient patients and significantly increased to normal after magnesium treatment without vitamin D supplementation [23]. Therefore, it is possible that an improvement in magnesium status leads to an increase in VDBP synthesis and, in turn, an elevated transport of vitamin D 3 to the liver and 25(OH)D to the kidney. The critical roles of magnesium in the synthesis of VDBP, PTH, 25(OH)D and 1,25(OH) 2 D may partially explain why the inverse associations between serum 25 (OH)D and risk of total mortality and mortality due to colorectal cancer and CVD primarily existed among those with magnesium intake above the median. High magnesium may increase the availability of 1,25 (OH)2D through activating the synthesis of 25(OH)D and 1,25 (OH)2D and increasing the transfer to target tissues by elevating vitamin D binding protein (VDBP) (Figure 1). This explanation is also supported by the observation in previous clinical studies that magnesium supplementation substantially reversed the resistance to vitamin D treatment among magnesium-deficient patients [21]. Previous studies found PTH level was elevated when serum 25(OH)D was under 20 ng/ml [41]. In the current study, we found magnesium intake was associated with a reduced risk of vitamin D insufficiency or deficiency only among those in the highest or the lowest tertile of PTH. This finding is possible because magnesium plays an important role in PTH regulation. This finding is also supported by observations made in a study of 30 women with osteoporosis that were investigated for magnesium deficiency using a magnesium tolerance test [26]. The subjects were divided into three groups: ten with vitamin D insufficiency (low vitamin D and raised PTH); ten with functional hypoparathyroidism (low vitamin D and low/low normal PTH); and ten who were vitamin D replete (normal vitamin D and normal PTH). All ten subjects with functional hypoparathyroidism were found to be magnesium deficient; five magnesium-deficient and five magnesium-insufficient patients were found in the subjects with vitamin D insufficiency, and only one magnesium-deficient and four magnesium-insufficient patients were found in the vitamin D replete group. Furthermore, intravenous magnesium infusion led to a significant rise in PTH in the group with functional hypoparathyroidism and a reduction in PTH in the subjects with vitamin D insufficiency.
A number of previous studies have examined the associations between magnesium intake with risk of stroke [42,43] and coronary heart disease [43,44], however, the results have been inconsistent in these previous studies. Two meta-analyses of prospective studies found that magnesium intake was related to a significantly reduced risk of stroke [42,43]. However, the inverse association was weak (an 8% reduction in risk per 100 mg magnesium/day increment). Likewise, a previous meta-analysis found that magnesium intake was non-significantly inversely associated with coronary heart disease with a pooled RR (95% confidence intervals (CIs)) of 0.86 (0.67 to 1.10) for the highest intake versus the lowest intake category [43]. In the current study, we found that the HR (95% CIs) for risk of mortality due to cardiovascular disease was 0.88 (0.61 to 1.26) for the highest quartile intake of magnesium versus the lowest quartile intake, which is very close to that in previous studies. Several previous studies have also evaluated the associations between magnesium intake and risk of colorectal cancer and results have also not been consistent [45]. A very recent meta-analysis found that every 100 mg/day increase in magnesium intake was related to 12% reduced risk of colorectal cancer (RR: 0.88; 95% CI 0.81 to 0.97). In the current study, we found that the HR (95% CIs) for risk of mortality due to colorectal cancer was 0.76 (0.15 to 3.86) for the highest quartile intake of magnesium versus the lowest quartile intake, which is also consistent with those from previous cohort studies. The significant interaction between serum vitamin D and magnesium intake in relation to mortality for cardiovascular disease and colorectal cancer may also explain some inconsistencies in previous studies which examined the associations of magnesium intake alone with risk of cardiovascular disease and colorectal cancer. A strength of our study is that it is based on NHANES, a population-based study with nationally representative samples. As with all prevalent case-control studies, one concern is that the temporal sequence may not be clear. However, it is unlikely that serum vitamin D status led to high or low intake of magnesium or vitamin D. Furthermore, the analysis of data from the NHANES III, a population-based prospective cohort, provides additional indications that magnesium interacts with vitamin D in relation to mortality. It is possible that people who consumed a high level of magnesium may also have a healthy lifestyle (for example, more physical activity and higher proportion of dietary supplement users). Supplement users may consume a high level of vitamin D from supplements while outdoor physical activity is related to an increasing production of 25(OH)D from sunlight exposure. We have adjusted for physical activity and intake of vitamin D from both diet and supplement in all analyses. Furthermore, we have conducted stratified analyses by physical activity and use of supplement. We found, unlike intake of magnesium, physical activity and use of supplements did not significantly or marginally significantly modify the association between serum 25(OH)D and mortality. Finally, although multiple 24-h dietary recalls are used as a gold standard measure in nutritional epidemiologic studies, a one-time 24-h dietary recall may not capture long-term dietary exposure. However, similar to dietary vitamin D and supplemental vitamin D intake, we found both dietary intake of magnesium (based on 24-h dietary recall) and magnesium from supplementation intake (derived from 30-day supplement questionnaire) significantly contributed to serum vitamin D status. Moreover, total intake of magnesium significantly interacted with intake of vitamin D in relation to both magnesium deficiency and insufficiency while total intake of magnesium modified the association between serum vitamin D and mortality. Since inter-day variation in magnesium intake is random, any residual inter-day variation may lead to nondifferential misclassification, which usually biases the result to the null. Thus, the true associations of magnesium intake with vitamin D status risk may be stronger than those we observed. Our findings are not only biologically plausible, but also remarkably consistent, indicating our findings may not be solely due to chance. In the current study, we did not have data for other parameters of the vitamin D/PTH axis (that is, serum PTH and 1,25-dihydroxyvitamin D), which may indeed be affected by magnesium status. Future studies are warranted to examine if low magnesium status using dietary and body status of magnesium affected both PTH levels and 1,25-dihydroxyvitamin D because this could serve as one mechanism for the observed interaction between serum 25(OH)D and magnesium intake on mortality.

Conclusions
Our preliminary findings indicate it is possible that magnesium intake alone or its interaction with vitamin D intake may contribute to vitamin D status. The associations between serum 25(OH)D and risk of mortality may be modified by the intake level of magnesium. Future studies, including cohort studies and clinical trials, are necessary to confirm the findings.