European Journal of Epidemiology

, Volume 27, Issue 8, pp 581–591

Circulating vitamin D, calcium and risk of cerebrovascular disease: a systematic review and meta-analysis


  • Rajiv Chowdhury
    • Department of Public Health and Primary CareUniversity of Cambridge
  • Sarah Stevens
    • NHS Midlands and East
  • Heather Ward
    • Department of Epidemiology School of Public HealthImperial College London
  • Susmita Chowdhury
    • Foundation for Genomics and Population Health
  • Ayesha Sajjad
    • Department of Epidemiology, Erasmus MCUniversity Medical Center Rotterdam
    • Department of Public Health and Primary CareUniversity of Cambridge
    • Department of Epidemiology, Erasmus MCUniversity Medical Center Rotterdam

DOI: 10.1007/s10654-012-9729-z

Cite this article as:
Chowdhury, R., Stevens, S., Ward, H. et al. Eur J Epidemiol (2012) 27: 581. doi:10.1007/s10654-012-9729-z


Available literature suggests that both vitamin D and calcium may be associated with a wide range of non-skeletal outcomes. However, epidemiological evidence supporting their individual associations with incident cerebrovascular disease is scarce. We conducted a systematic review and meta-analysis of prospective cohort studies, published before February 2012 and sought from MEDLINE, EMBASE, BIOSIS and the Science Citation Index databases, and reported cerebrovascular disease (defined as any fatal or non-fatal ischemic stroke, hemorrhagic stroke, cerebrovascular accident or transient ischemic attack) by circulating vitamin D (25-hydroxy vitamin D [25(OH)D] as active metabolite) and calcium levels. Two independent investigators abstracted information on 25(OH)D and calcium, cerebrovascular outcomes and other characteristics from selected studies. Relative risks (RRs) were pooled by both random and fixed effects meta-analyses and were further examined under different study-level characteristics. Publication bias was assessed with funnel plots and Egger’s asymmetry test. From 5,778 initial references, nine unique prospective cohort studies met our inclusion criteria. Seven studies (involving 47,809 participants and 926 cerebrovascular events) focused on circulating 25(OH)D and 3 reported on circulating calcium (22,577 participants and 727 events). For 25(OH)D, in a comparison of individuals in the top third versus those in the bottom third at baseline, the combined RR for cerebrovascular disease, adjusted for several conventional risk factors, was 0.60 (95 % CI 0.48, 0.72). The corresponding RR in the prospective studies that reported on baseline circulating calcium levels for cerebrovascular disease was 1.40 (95 % CI 1.19, 1.64). There was no apparent evidence of heterogeneity or publication bias among included studies. Available data indicate that higher circulating level of vitamin D is associated with a decreased risk of cerebrovascular disease. Conversely, higher circulating calcium concentration is associated with an increased risk of cerebrovascular disease.


Vitamin DCalciumCerebrovascular diseaseMeta-analysis


Vitamin D, a fat-soluble vitamin, functions as a hormone in the body to regulate metabolism of calcium (a key element for bone mineralization and other metabolic functions) by promoting its intestinal absorption, resorption from bone, and retention in the kidneys [1]. A growing body of recent literature [28] suggests that vitamin D and calcium homeostasis could be associated with several chronic illnesses, including coronary heart disease, cardiovascular disease (CVD) [6, 7], diabetes [8] and all-cause mortality [9]. This is of considerable public health significance as vitamin D insufficiency represents a global problem [10]. For example, only about a quarter of the US adult population [11] may have an optimum circulating level of vitamin D (defined as 30 nanogram per milliliter or greater) [10], while similar or lower estimates have been reported across the majority European countries [12].

Despite great interest in vitamin D and calcium for their supposed links to a wide range of non-skeletal outcomes, specific associations of these markers of mineral metabolism with cerebrovascular disease remain unclear. Interpretation of the findings from available randomised control trials (RCTs) is difficult as the majority of these studies assessed combined effects of vitamin D and calcium supplements, were based largely on osteoporotic populations, had insufficient events, and were not adjudicated to study cerebrovascular outcomes. In the absence of reliable experimental data [13], findings from prospective cohort studies can potentially help understand any aetiological relevance of these markers to cerebrovascular risk and inform policy. However, many of these studies were based on dietary assessments which are prone to misclassification for using self-report tools [14] and play a less significant role in determining true body levels of these markers. By contrast, circulating biomarkers can reflect overall status more objectively than self-reported dietary studies [15, 16]. Although few observational studies have reported on circulating vitamin D and calcium in relation to cerebrovascular disease, they vary considerably, and have never been systematically assessed.

Therefore, we have conducted a systematic review and meta-analysis of available prospective cohort studies to: (1) quantify individual associations of circulating blood levels of vitamin D and calcium with incident fatal or non-fatal cerebrovascular disease, and (2) examine these associations in several clinically relevant study-level characteristics.


This review was conducted using a predefined protocol and in accordance to the PRISMA guidelines (Appendix 1).

Information sources, search strategy and eligibility criteria

Between October 2011 and January 31, 2012 (date last searched) we comprehensively searched the following electronic databases: Medline, EMBASE, BIOSIS and the Science Citation Index. Studies were sought that had reported on associations of circulating vitamin D [measured as 25-hydroxy vitamin D or 25(OH)D] or circulating calcium levels with any incident cerebrovascular disease outcomes (defined as any fatal or non-fatal ischemic stroke, hemorrhagic stroke, cerebrovascular accident or transient ischemic attack). The computer-based searches combined search terms related to the exposure (e.g, calcium, vitamin D) and outcomes of interest (e.g, stroke, cerebrovascular disorders), without any language restriction. Further details on the search strategy are presented in Appendix 2.

Study selection

Observational studies were eligible for inclusion if they were prospective in design, had at least 1 year of prospective follow up, assessed individual association of 25(OH)D or calcium with cerebrovascular disease, and involved any of the following participants: (1) approximately general populations (i.e, participants not selected on the basis of pre-existing disease at baseline); or (2) people selected on the basis of elevated CVD risk factors. Two independent reviewers working in pairs (RC and HW, SC) screened the titles and abstracts of the initially identified studies to determine whether they would satisfy the selection criteria. Any disagreements about selection were resolved through consensus or consultation with a third author. Full text articles were retrieved for the selected titles. The retrieved studies were assessed by two independent authors (RC, HW) to ensure adherence to selection criteria and their reference lists were searched to identify additional publications.

Data collection process and data items

A data abstraction form was designed prior to the implementation of the search strategy. This form was used by two independent reviewers (RC, SS) to extract the relevant information from the selected studies. The data abstraction form included questions on study size; study design; baseline population (defined as healthy, and pre-existing or high risk of CVD); geographical location (defined as Europe, North America and Asia–Pacific); year of baseline survey; age range of participants at baseline; mean duration of follow-up; reported degree of adjustment for potential confounders (defined as ‘+’ when RRs were adjusted for age and sex, and ‘++’ following further adjustment for established vascular risk factors); average baseline level, storage temperature, number of disease outcome and reported risk estimate. In instances of multiple publications, the most up-to-date or comprehensive information was used.

Quality evaluation assessment

Two independent reviewers (SS, OhF) evaluated the quality of included studies using a modified scoring system, which was designed in line with the MOOSE, QUATSO, and STROBE guidelines [1719]. This allowed a total score between 0 and 6 points, with 6 being the highest quality score and allocated 1 point if a study: (1) provided a justification to conduct the investigation; (2) used appropriate inclusion and exclusion criteria; (3) assessed exposure with a validated assay method (prospective blood-based studies), (4) collected outcomes not solely based on self-report; (5) reported findings adjusted for at least age, sex and smoking status, and (6) additionally for other covariates (e.g, blood pressure, diabetes status, lipids, diabetes, body mass index, physical activity).

Statistical methods

Analyses involved only within-study comparisons (i.e, cases and controls were only directly compared within each cohort) to limit potential biases. To enable a consistent approach to analysis in this review, risk estimates for each prospective study were transformed to involve comparisons between the top third and bottom third of the population’s baseline distribution of biomarker values, using methods previously described [20]. Briefly, log risk estimates were transformed assuming a normal distribution, with the comparison between top and bottom thirds being equivalent to 2.18 times the log risk ratio for a 1 standard deviation increase (or equivalently, as 2.18/2.54 times the log risk ratio for a comparison of extreme quarters). Standard errors of the log risk ratios were calculated using published confidence limits and were transformed in the same way. Hazard ratios and odds ratios were assumed to approximate the same measure of relative risk (RR). Summary RRs and corresponding 95 % confidence intervals (95 % CIs) were calculated by pooling the study-specific estimates using a random-effects model that included between-study heterogeneity. Where studies reported RRs with differing degrees of adjustment for other risk factors, the maximum adjusted estimate was used. Consistency of findings across studies was assessed by standard χ2 tests and the I2 statistic [21]. Heterogeneity was assessed by comparing results from studies grouped according to pre-specified study-level characteristics using meta-regression. Evidence of publication bias was assessed using funnel plots and Egger test [22]. Meta-regression analyses were carried out as formal statistical test to determine differences between overall estimates yielded. All statistical tests were two-sided and used a significance level of P < 0.05. All analyses were performed using Stata release 11 (StataCorp, College Station, Texas).


Overall, 5,778 potentially relevant citations were initially identified (Fig. 1). Following initial screening of the titles and abstracts, 5,396 were excluded, leaving 382 articles for retrieval. Full text assessment of these articles resulted in nine unique studies that were included in our analyses. Of the 373 full text articles excluded, 107 did not report the exposures of interest, 251 did not record cerebrovascular disease outcomes and the remaining 15 were otherwise inappropriate.
Fig. 1

Search strategy for the studies included in current review

Characteristics of studies included

Of the nine prospective cohort studies included in this review, seven focused on circulating 25(OH)D (involving 47,809 participants and 926 cerebrovascular events) and three studies examined circulating calcium levels (22,577 participants and 727 events) as exposures of interest (Table 1). Only one study had overlapping information on both biomarkers [23]. The age of the participants among these studies ranged from 18 to 99 years. Seven studies evaluated both men and women [2329] and two were based solely on women [30, 31]. Three studies were based in North America (USA) [24, 26, 29], 4 in Europe (Finland [23, 27] Germany) [25 28], 1 in the Asia–Pacific region (New Zealand) [30], and one study on multiple countries [31]. Five studies involved healthy populations, one study evaluated postmenopausal women with osteoporosis, two recruited individuals with elevated CVD risk and one study recruited participants with renal disease. All studies adjusted for at least age and sex, while seven studies reported additional adjustment for other conventional vascular risk factors (e.g, smoking, lipids, history of diabetes, hypertension) (eTable 1). All studies assessed vitamin D status by measuring serum vitamin D levels by either radioimmunoassay or chemiluminiscent immunoassay as 25(OH)D, the principle circulating form of Vitamin D. The average baseline values ranged from 11.8 to 28 ng/mL for circulating 25(OH)D and 9 to 10.3 mg/dL for circulating calcium concentrations, respectively. For circulating calcium, 2 studies employed baseline calcium levels corrected for albumin, while one used total calcium.
Table 1

Summary of the prospective observational studies included in the present review

Lead author, publication year

Name of the study


Population source

Baseline population

Average age, yr

Male, (%)

Storage Temperature


Assay method

Average baseline level

No. of total participants

Average duration of follow up, yr

No. of any stroke events

Kilkkinen et al. [27]



Population register




−20 °C



17.4 ng/mL




Anderson et al. [24]


United States

Healthcare register

High CVD risk






28.0 ng/mL




Michos et al. [29]


United States

Population register




−70 °C



25.1 ng/mL




Drechsler et al. [25]



Trial register

Renal disease



−70 °C



24.3 ng/mL




Bolland et al. [30]

New Zealand

Trial register




−70 °C



20.2 ng/mL




Pilz et al. [28]




High CVD risk



−70 °C



17.4 ng/mL




Marniemi et al. [23]


Population register




−20 °C





11.8 ng/mL

9.9 mg/dL




Foley et al. [26]


United States

Population register




−70 °C


o-cresolphthalein complexone

10.3 mg/dL




Slinin et al. [31]



Trial register




−20 °C



9.0 mg/dL












Association of circulating 25(OH)D and calcium with cerebrovascular disease

The RRs (95 % CI) for cerebrovascular disease, adjusted typically for several conventional risk factors (e.g, gender, age, hypertension, smoking status, history of diabetes, and lipids) comparing participants in the top third with those in the bottom third of baseline circulating 25(OH)D and calcium levels are presented in Fig. 2. For circulating 25(OH)D, based on 7 available prospective cohort studies, the combined RR for total cerebrovascular disease was 0.60 (95 % CI 0.48, 0.72) using random-effects and 0.59 (0.48, 0.72) using fixed effect meta-analysis. Cause-specific cerebrovascular risk information was only available in one study [27], where adjusted RRs comparing top versus bottom thirds of baseline 25(OH)D were 0.60 (95 % CI 0.38, 0.93) and 0.61 (95 % CI 0.26, 1.46) for ischemic and hemorrhagic strokes, respectively. The corresponding RR in the 3 prospective cohort studies that reported on circulating calcium levels in relation to cerebrovascular disease was 1.40 (95 % CI 1.19, 1.64) using random-effects meta-analysis among those in the top third compared to the bottom third at baseline. When a fixed effect model was employed, the RR was identical to that of random-effects meta-analysis. There was no evidence of heterogeneity between the observational studies for either markers [I2 statistic (95 % CI) of 0 % (0, 75 %), P = 0.70 for 25(OH)D; and 0 % (0, 90 %), P = 0.90 for calcium] (Fig. 2).
Fig. 2

Relative risk (95 % CI) for stroke comparing top versus bottom thirds of circulating (1) 25-hydroxy vitamin D, and (2) calcium in available prospective observational studies

Subgroup analysis and assessment of publication bias

The overall association observed between 25(OH)D and cerebrovascular disease remained largely similar when these studies were grouped by several study-level characteristics (Fig. 3). For example, RRs (95 % CIs) comparing the top versus the bottom thirds of 25(OH)D for studies based on healthy and pre-existing/high CVD risk populations were 0.64 (0.48, 0.86) and 0.55 (0.40, 0.76), respectively (P > 0.05 in meta-regression analyses). The corresponding RRs for studies based in Europe, North America and Asia–Pacific regions were 0.60 (0.45, 0.79), 0.57 (0.39, 0.83) and 0.63 (0.36, 1.11), respectively (P > 0.05 in meta-regression analyses). Similar RRs were obtained for baseline usage of vitamin D supplement, study size, level of adjustment and control for seasonality subgroups (P > 0.05 for all) (Fig. 3). In sensitivity analysis, removing 4D study (comprising of renal patients) did not change the association of 25(OH)D with cerebrovascular disease [pooled RR: 0.59 (0.48, 0.72)]. For calcium, sensitivity analysis by removing ARIC study (with higher average levels of calcium at baseline) yielded broadly similar findings [pooled RR: 1.48 (1.02, 2.12)]. Overall RR of circulating calcium for cerebrovascular disease were also similar after excluding the MORE study based on unadjusted total baseline calcium, [pooled RR: 1.38 (1.16, 1.63)]. There was no evidence of publication bias across all prospective observational studies of 25(OH)D (P = 0.7 in Egger’s asymmetry test) (eFig. 1). As only three prospective cohort studies were available for circulating calcium levels and cerebrovascular disease, sub-group analyses were not done in these studies.
Fig. 3

Relative risk for stroke in the top versus bottom third of baseline 25-hydroxy vitamin D [25(OH) vitamin D] according to different study level characteristics


The current review has assessed the associations between circulating levels of vitamin D and calcium concentrations and subsequent risk of incident cerebrovascular disease involving 69,160 participants worldwide. Our results demonstrate that abnormalities in these biomarkers are significantly associated with future risk of cerebrovascular disease. There was 40 % lower risk of cerebrovascular disease among people who were in the top third of baseline circulating levels of 25(OH)D compared to the bottom third. In a similar comparison, circulating calcium concentrations were associated with, by contrast, a 40 % increased risk of incident cerebrovascular outcomes. Additionally, these findings were consistent across all included studies, when grouped by several clinically-relevant study characteristics.

There may be several explanations for the findings observed in the current study. Receptors for vitamin D have been found in numerous human tissue types including those involved in vascular endothelium, calcium homeostasis, cellular proliferation and differentiation [32], all of which could show a wide range of vascular effects [33]. Levels of circulating vitamin D have been associated with established vascular risk factors such as diabetes [34]; arterial hypertension [35], and inflammation [36]. Furthermore, vitamin D may exhibit specific antithrombotic and neuroprotective properties and have been shown to attenuate ischemic cortical injury in animal models [37, 38]. Therefore, an inverse association observed in the prospective observational studies in the current review lends further support to a potential link between levels of vitamin D in the bloodstream and future risk of cerebrovascular events in both healthy and other populations. Nonetheless, as primary causes of vitamin D deficiency include insufficient exposure to sunlight, poor diet, increased adiposity, and reduced synthesis or absorption [12], it is possible that an optimal vitamin D status essentially reflects a healthier lifestyle and higher socioeconomic circumstances. These individual-level factors can, in turn, influence cerebrovascular risk for their potential roles on several established cardiovascular determinants such as smoking, blood pressure, body mass index, and use of supplements [6]. The potential confounding impact of these factors in our pooled estimates, however, should be minimal as the majority of the included studies in this review controlled for these characteristics and our estimates were largely unchanged when they were further stratified by levels of multivariate adjustments employed in these studies. Potential biological roles of serum calcium level in cerebrovascular disease, on the other hand, have not been explored in greater detail. However, it has been reported that circulating calcium concentration is positively related to systolic and diastolic blood pressures, serum glucose and cholesterol concentrations [39]- all of which are known determinants of cerebrovascular disease risk. Self-reported higher calcium intake has been associated with increased vascular calcifications in patients with chronic kidney disease [40, 41]. Additionally, intervention studies, although primarily based on people at high risk of fracture, showed that calcium supplements potentiate vascular calcification and increase overall CVD risk [42].

Our findings may have several implications. First, they indicate a potential etiological association between vitamin D insufficiency and cerebrovascular diseases—both of which are widely prevalent worldwide. This relationship appears to be independent of many potential confounding factors or stratification by several clinically-relevant characteristics. Second, they usefully extend previous literature on circulating vitamin D that showed inverse associations with coronary and mortality risk. Third, they demonstrate that in addition to the reported detrimental effects of calcium supplements on CVD [42], increased circulating calcium concentrations are associated with higher cerebrovascular risk—highlighting the supposed relevance of calcium metabolism pathways in vascular disease. Fourth, these associations were broadly similar in non-osteoporotic or non-renal populations, indicating potential risk predictive value of these markers in essentially healthy individuals. Finally, they stimulate further research using detailed, large-scale studies to reliably confirm the relations. This is of particular importance as direct evidence from RCTs are either limited or difficult to interpret due to limitations in the designs used (e.g. involving mainly diseased populations or insufficient power).

Potential strengths and limitations of this review merit consideration. We report a first-ever systematic synthesis of the blood concentrations of vitamin D and calcium in relation to cerebrovascular risk combining data based on nearly 70,000 unique individuals. To enhance reliability and minimise reverse causation, we selected studies which had prospective design and long duration of follow up. Where appropriate, we standardised study-level estimates to allow consistent comparisons and examined for several clinically-relevant characteristics to reduce potential heterogeneity among available studies. Nonetheless, this meta-analysis has been limited by the moderate amount of available data on cerebrovascular outcomes. For example, even in aggregate, only over 1,500 stroke events were available for this study (with <1000 events in calcium studies) and only one study reported cause-specific stroke estimates. Additional under-estimation of the observed associations may occur from the lack of repeat measurements which would reflect variability over time (i.e, “regression dilution”) [43]. Although not evident in our qualitative assessment using graphical inspection of funnel plots, publication bias, typically inherent to literature-based meta-analyses, remains a concern. Finally, as our literature-based review is based solely on prospective observational studies, we could not control for all relevant covariates consistently and were not able to investigate any causal association. Future studies involving larger, individual-level data for detailed epidemiological assessments, and population-based RCTs to test causal effects of these markers on cerebrovascular disease are, therefore, warranted.

In conclusion, available data indicate that higher circulating level of vitamin D is associated with a reduced risk of cerebrovascular disease. This association is particularly relevant as (1) global burden of both vitamin D deficiency and cerebrovascular disease remain considerably high, and (2) safe and cost-effective interventions to improve vitamin D status (e.g., supplements, sun exposure, or diet) can be scalable at the population level and may provide complementary benefits to prevent cerebrovascular, coronary and mortality events. By contrast, in a meta-analysis of largely a separate set of studies, higher circulating calcium concentrations are associated with an increased risk of cerebrovascular disease, which encourages further mechanistic and epidemiological assessments, and considerations regarding management of calcium levels among patients at accelerated risk of cerebrovascular disease.


Dr Chowdhury has been a recipient of the Gates Cambridge Ph.D. scholarship; and Prof. Franco is the recipient of a grant from Pfizer Nutrition to establish a new center on ageing research focused on nutrition and lifestyle, ErasmusAGE. Dr Sajjad works within ErasmusAGE.

Supplementary material

10654_2012_9729_MOESM1_ESM.docx (24 kb)
Supplementary material 1 (DOC 25 kb)

Copyright information

© Springer Science+Business Media B.V. 2012