Vitamin D Supplementation, Serum 25(OH)D Concentrations and Cardiovascular Disease Risk Factors: A Systematic Review and Meta-Analysis

Background: Cardiovascular disease (CVD) risk factors are associated with low serum 25 hydroxyvitamin D (25(OH)D) concentrations in observational studies; however, clinical trial findings are inconsistent. Objective: We assessed the effect of vitamin D supplementation and increased serum 25(OH)D concentrations on CVD risk factors in a systemic review and meta-analysis of randomized controlled trials (RCTs). Design: MEDLINE, CINAHL, EMBASE, and Google Scholar were searched for RCTs that evaluated vitamin D supplementation and cardiovascular outcomes [blood pressure, parathyroid hormone (PTH), serum high-sensitivity C-reactive protein (hs-CRP), total cholesterol, high and low density lipoprotein (HDL and LDL, respectively), triglycerides, peak wave velocity (PWV) and Augmentation Index (AI)] from 1992 through 2017. Meta-analysis was based on a random-effects model and inverse variance method to calculate standardized mean difference (SMD) as effect sizes, followed by a leave-one-out method for sensitivity analysis. Risk of publication bias was assessed using Cochrane checklist and Begg funnel plots. The systematic review is registered as CRD42015025346. Results: We identified 2341 studies from which 81 met inclusion criteria. The meta-analysis indicated a significant reduction in systolic blood pressure (SMD = −0.102 ± 0.04 mmHg, 95% confidence interval (CI), −0.20 to −0.03), diastolic blood pressure (SMD = −0.07 ± 0.03 mmHg, 95% CI, −0.14 to −0.006), serum PTH (SMD = −0.66 ± 0.08 ng/L, 95% CI, −0.82 to −0.49), hs-CRP (SMD = −0.20 ± 0.07 mg/L, 95% CI, −0.34 to −0.06), total cholesterol (SMD = −0.15 ± 0.06 mmol/L, 95% CI, −0.25 to −0.04), LDL (SMD = −0.10 ± 0.05 mmol/L, 95% CI, −0.20 to −0.003), triglycerides (SMD = −0.12 ± 0.06 mmol/L, 95% CI, −0.23 to −0.003) and a significant increase in HDL (SMD = 0.09 ± 0.04 mmol/L, 95% CI, 0.00 to 0.17) with vitamin D supplementation. These findings remained significant in sensitivity analyses for blood pressure, lipid profile, serum PTH, and serum hs-CRP. There was no significant effect of vitamin D supplementation on PWV (SMD = −0.20 ± 0.13 m/s, 95% CI, −0.46 to 0.06, p = 0.14) and AI (SMD = −0.09 ± 0.14%, 95% CI, −0.37 to 0.19, p = 0.52) for vitamin D supplemented groups. Conclusion: These findings suggest that vitamin D supplementation may act to protect against CVD through improving risk factors, including high blood pressure, elevated PTH, dyslipidemia, and inflammation.

Design: MEDLINE, CINAHL, EMBASE, and Google Scholar were searched for RCTs that evaluated vitamin D supplementation and cardiovascular outcomes [blood pressure, parathyroid hormone (PTH), serum high-sensitivity C-reactive protein (hs-CRP), total cholesterol, high and low density lipoprotein (HDL and LDL, respectively), triglycerides, peak wave velocity (PWV) and Augmentation Index (AI)] from 1992 through 2017. Meta-analysis was based on a random-effects model and inverse variance method to calculate standardized mean difference (SMD) as effect sizes, followed by a leave-one-out method for sensitivity analysis. Risk of publication bias was assessed using Cochrane checklist and Begg funnel plots. The systematic review is registered as CRD42015025346.

INTRODUCTION
The main physiological role of vitamin D has long been regarded as regulation of calcium and phosphorous homeostasis and proper bone mineralization. In more recent years, however, inadequate vitamin D status has been linked to a number of nonskeletal chronic health conditions such as diabetes, cancer, and cardiovascular disease (CVD) (1)(2)(3). The prevalence of vitamin D deficiency is high in populations across the globe and an additional 30-50% are at risk of being vitamin D deficient (4,5). Aging is also associated with decreased vitamin D synthesis in the body, putting individuals already vulnerable at an increased risk of these conditions (6).
Observational studies have consistently found an association between low serum 25-hydroxyvitamin D (25(OH)D) concentrations and presence of CVD risk factors, including blood pressure, dyslipidemia, and inflammation (7)(8)(9)(10)(11). A review of prospective studies found that serum 25(OH)D concentrations <25 or 37 nmol/L (10 or 15 ng/mL) were associated with an increased risk of CVD disease or mortality (12). This is supported by a recent meta-analysis that revealed a significant association between low 25(OH)D concentration and increased cardiovascular mortality, a consistent finding across countries, sexes, age groups, and season of blood testing (13).
Current evidence suggests a role for several different vitamins in the protection of proper heart function, especially those with antioxidant potency, and thus multiple vitamin deficiencies may contribute to development of CVD. Antioxidant vitamins such as vitamin C and vitamin E might diminish the rate of oxidative stress which is a crucial component in the pathogenesis of atherosclerosis and CVD. B vitamins, which play a role in ATP energy production, and vitamin D all induce cardioprotective effects and maintain cardiovascular health (14). B vitamins might inhibit homocysteine mediated superoxide production and attenuate the atherogenicity of homocysteine (15), and improve endothelial function through decreasing homocysteine levels leading to increased flow-mediated vasodilation (16). The presence of vitamin D receptor expression in endothelial cells, vascular smooth muscle cells, and cardiac myocytes provides biological support for these observations (17); vitamin D has also been associated with the improvement of endothelial function and glucose homeostasis, reduction of oxidative stress, inflammatory response, and thrombogenesis, as well as the modulation of calcium and lipoprotein metabolism (18). Secondary hyperparathyroidism, excess parathyroid hormone, resulting from chronic vitamin D deficiency has been associated with CVD potentially through several different pathological pathways, including: (1) increased insulin resistance and pancreatic β cell dysfunction, leading to metabolic syndrome and diabetes, (2) activation of renin-angiotensin-aldosterone system (RAAS), increasing blood pressure, leading to apoptosis and fibrosis, and (3) stimulation of systemic and vascular inflammation leading to atherogenesis (4,19). Current evidence suggests vitamin D deficiency is an important new cardiovascular risk factor that may play a causal role in the development of cardiovascular disease (20).
Several published meta-analyses and systematic reviews have found no beneficial effect of vitamin D supplementation on CVD risk factors (21)(22)(23)(24)(25)(26). Ford (24), for example, suggested that there is insufficient evidence to support vitamin D supplementation for the reduction of cardiovascular events, although these authors did raise the possibility that vitamin D supplementation might have an effect on heart failure. Several meta-analyses and systematic reviews have similarly failed to find an association. In their systematic review, Wang et al. (26) showed a statistically nonsignificant reduction in cardiovascular disease with moderate doses of vitamin D (approximately 1,000 IU/d). Mao et al. (25) showed that neither vitamin D supplementation nor calcium supplementation had an effect on major cardiovascular events, myocardial infarction, or stroke. However, a meta-analysis is only as good as the quality of studies included.
The quality of the RCTs included in these meta-analyses has been criticized (27). Many RCTs do not have the ability to detect any effect due to an effect size that is simply too narrow (28). Several RCTs provided vitamin D doses that are far too low to measure a detectable increase in serum 25(OH)D concentration (29)(30)(31) and/or are too short in duration (e.g., weeks rather than months or years) to expect a change in health outcomes (32,33). Further, most of the RCTs were grossly underpowered to detect changes in secondary outcomes (28). Several of the studies do not report baseline and/or follow-up serum 25(OH)D concentrations making it impossible to determine whether a change in vitamin D status occurred and thus whether it can be implicated in observed outcomes.
Given these uncertainties, the question of whether vitamin D supplementation improves cardiovascular risk factors and reduces subsequent disease remains without a convincing answer. The current meta-analysis investigates the role of vitamin D supplementation on cardiovascular outcomes by imposing a stringent set of inclusion criteria for studies by aggregating trials that properly take into account the biology of vitamin

Inclusion Criteria
Only studies that met the following criteria were considered for this systematic review and meta-analysis: (1) studies included participants with any baseline 25(OH)D level; (2) studies recorded changes in blood pressure, PTH, hs-CRP, lipid profile, peak wave velocity, and/or augmentation index; (3) a minimum of 3 months of supplementation/therapy to ensure that the intervention had sufficient time to produce an effect on serum 25(OH)D concentrations; (4) studies with daily, weekly, or monthly frequency of dosage; (5) studies reported pre-and postserum 25(OH)D levels (or when it was supplied by authors following request); (6) studies with control groups using a placebo and those receiving placebo plus a co-intervention (if both arms of the study received the co-intervention); and (7) studies using vitamin D 3 or cholecalciferol.

Exclusion Criteria
Studies were excluded if: (1) they were nonclinical studies, observational studies, case-control, or cross-sectional studies; (2) they were methodological reports, editorials, narrative reviews, comments, and letters; (3) participants were younger than 18 years old; (4) intervention periods were less than 3 months; (5) dosage was less frequent than monthly or if a bolus dose was used; and (6) studies provided inadequate information on outcomes or serum 25(OH)D levels; (7) studies showed on improvement in vitamin D status (serum 25(OH)D change over time ≤0).
Two authors (NM, JR) independently reviewed each reference title and abstract to determine whether the studies met the inclusion criteria. Any disagreements with study selection were resolved through the discussion with the third author (SMK). Full-text articles were retrieved for the selected abstracts. Full articles were again assessed by the two independent authors (NM, JR) to ensure that they were eligible to be included in metaanalysis and any disagreements were finalized by the third author (SMK).

Data Extraction
Two independent authors (NM, JR) extracted the following data from the included trials: first author and year of publication; number, age, and sex of participants; study population characteristics; latitude of residence, dosage details of vitamin D including frequency, duration and IU; any cointervention; pre-and post-serum 25(OH)D levels; pre-and post-measures for blood pressure, PTH, hs-CRP, total cholesterol, LDL, HDL, triglycerides, PWV, and AI. NM or JR also contacted several authors to provide missing data or to clarify data within the primary report. All data was then reviewed by the third author (SMK).

Risk for Bias Assessment
We assessed each included study for risk of bias by using fields from the Cochrane checklist (34) to determine the following variables: quality of random allocation concealment, blinding of outcomes assessors, treatment and control group comparability, clear definition of inclusion and exclusion criteria, participant blinding to allocation, selective reporting, if intention-to-treat analysis applied, and description of withdrawals and dropouts. Each criterion was marked as (+) with adequate information, (−) with inadequate information and (?) with unclear information (Table S1).
We generated Begg funnel plots to visually examine possible publication bias. These plots were supplemented by formal statistical testing using the Egger weighted regression tests (35). The analysis for the effects of publication bias was adjusted using the Duval and Tweedie trim-and-fill method (36).

Strategy for Data Synthesis and Statistical Analysis
We performed the meta-analysis at the trial level using Comprehensive Meta-Analysis V3 (Biostat 2014, Englewood, NJ) (37). To calculate the effect size, the mean change in concentrations, calculated as measure at the end of intervention minus measure at baseline, and the standard deviation of the outcomes were used for both treatment groups (38) and the effect size was expressed as standardized mean difference between vitamin D intervention and placebo groups, with a 95% confidence interval. For all treatment effects, a negative value denoted a reduction in the outcomes within the vitamin D group compared with placebo. We used the I 2 index to evaluate heterogeneity among the included studies and with a value ≥50%, random-effects model (using the DerSimonian-Laird and generic inverse variance method) was applied (39,40).
We conducted a sensitivity analysis using the leave-one-out method to assess the effect of each study on the overall effect size (41). For studies with more than one vitamin D supplemented group (e.g., different daily doses given), the trial with the higher dose was selected and compared with the placebo group.

Subgroup Analysis
To further assess interactions among subgroup treatments and also to address heterogeneity among included studies, we defined a priori subgroups as followed: participant's age (<55 vs. ≥55 years old), which was the median of study population's age and the central value of data providing an equal distribution of information for comparison; vitamin D supplementation dose (<4,000 vs. ≥4,000 IU/day), which is the average dose required to provide optimal serum physiological levels of vitamin D (100-130 nmol/L) (30,42); serum 25(OH)D concentration at the end of the intervention (<86 vs. ≥86 nmol/L), which was selected as the median value of serum 25(OH)D levels; duration of intervention (<6 months vs. ≥6 months), which was selected based on the half-life of serum 25(OH)D and the time required to maintain a steady serum levels and potentially influence other biomarkers (43)(44)(45); obesity (BMI < 30 vs. BMI ≥ 30 kg/m 2 ) as defined by WHO (46) and based on and the fact that obese individuals need 2-3 times the amount of vitamin D to achieve the same serum 25(OH)D as normal weight individuals (47); vitamin D deficiency at the beginning of the intervention [serum 25(OH)D<50 vs. ≥50 nmol/L] as defined by IOM (48) and based on evidence demonstrating a strong association between vitamin D deficiency and higher incidence of CVD risk factors (12); and, calcium co-intervention.

Study Selection
We screened the titles of 2,341 studies after duplicates were removed. After excluding any irrelevant studies, 252 abstracts were retrieved for further examination. Based on the abstracts, 177 full-text articles were assessed to determine whether they satisfied the inclusion criteria. Of these, 86 were excluded from analysis because the intervention was less than 3 months intervention, the frequency of dosage was more than monthly or bolus doses were given, pre-or post-serum 25(OH)D levels were not reported, or the trials were centered on children/ adolescents.
For the other 10 trials we did not include data, they either shared similar designs and outcomes (49,50), had no post serum 25(OH)D data available (51) or, even after contacting the corresponding author, had insufficient information (52)(53)(54)(55)(56)(57)(58). We only included papers where an increase in vitamin D status followed supplementation, so the Cooper et al. (59) study was excluded. Finally, 81 studies were included in systematic review and meta-analysis. Details of the complete search process and for each outcome are given in Figure 1.

Risk of Bias Assessment
Ten of the included studies lacked information on the blinding of participants and personnel, and one study did not provide information on allocation concealment. Less than half of the studies (n = 38) used an intent-to-treat analysis. However, the vast majority of the included studies had a low risk of bias. Details of the quality of bias assessment are provided in Table S1.

Study Characteristics
The characteristics of the included trials are given in Table 1 and studies that were excluded from the meta-analysis are highlighted. The included studies were published between 1992 and 2017. The average latitude of the studies conducted in the northern hemisphere was 43.2 ± 14.7 • N (73 studies), with the maximum latitude of 70 • N, and the average latitude of the studies from the southern hemisphere was 34.6±6.8 • S (eight studies), with the maximum latitude of 43S. Sample sizes varied from 20 (61) to 484 (126); a total of 9,993 participants are included in the meta-analysis, with 5,042 in a vitamin D-supplemented group and 4,951in a placebo group. Twenty of the studies only assessed females ( Table 1), but the overall study population included 67% females and 33% males. The mean age of participants ranged from 18 (77) to 85 years (66,74,98,104), with an overall average age of 55 ± 15 years.
Participants received treatment through capsules, pills, tablets, oil drops, or as a specially fortified milk or yogurt drink. Calcium was co-administered with vitamin D and placebo in 24 of the 81 studies ( Table 1). The duration of intervention lasted 3 months to 5 years, with an average duration of 9.6 ± 9.2 months (median = 6 months). The daily dose of supplemental vitamin D ranged from 400 (66,67,91,94,104) to 12,000 IU (116), with an average of 2,967 ± 2,271 IU/day. Baseline serum 25(OH)D concentration varied widely from 16 (121) to 95 nmol/l (110), with the average of 45 ± 16 nmol/L in both vitamin D and placebo groups. The diversity of participants was considerable in these studies. Some were healthy and community dwelling populations, whereas others were institutionalized and/or had specific health conditions such as diabetes, kidney disease, women with polycystic ovary syndrome (PCOS), or included patients on hemodialysis. Forty-six of the studies reported serum PTH concentrations, 39 reported blood pressure and lipid profiles, 28 studies recorded hs-CRP concentrations, and 10/11 studies PWV and AI as their outcomes.

Vitamin D and Blood Pressure
A total of 39 studies reported on outcomes of systolic and diastolic blood pressure (Figures 2, 3). The pooled effect size (standardized mean difference) of the effect of vitamin D supplementation on systolic blood pressure was −0.102 ± 0.04 mmHg, (95% CI −0.20 to −0.01, p = 0.02, I 2 = 51%) across all studies. The pooled effect size for diastolic blood pressure was −0.072 ± 0.03 mmHg (95% CI −0.14 to −0.006, p = 0.03, I 2 = 18%) across all studies. Overall results indicated that vitamin D supplementation was significantly associated with lower blood pressure.
Six studies showed a significant reduction in systolic blood pressure (61,76,81,100,121,128) and four studies revealed significant reductions in diastolic blood pressure (81,116,125,128). Seventeen studies demonstrated a decreasing trend in systolic and/or diastolic blood pressure following vitamin D supplementation, though these changes were not statistically significant (50, 62, 82, 86, 92, 95, 108, 116, 117, 119, 123-125,  showed either a null effect or an increase in blood pressure. In the majority of these studies, blood pressure was a secondary endpoint and the studies were not designed or powered for detecting the effects of vitamin D supplementation on blood pressure. Some studies also included patients with comorbid condition like kidney (60,106) or heart failure (122), and others had all or a majority of their participants with normal blood pressure at baseline.
Only one of the included studies centered on hypertensive patients (100). After a five month intervention, this study found a significant reduction in blood pressure following vitamin D supplementation (3,000 IU/day) and improved serum 25(OH)D levels (50 nmol/L increase) compared to placebo.

Vitamin D and PTH
Forty-five papers reported serum PTH levels as a primary or secondary endpoint. The pooled effect size of vitamin D on serum PTH levels was −0.66 ± 0.08 ng/L (95% CI −0.82 to −0.50, p < 0.001, I 2 = 87%) across all studies (Figure 8). Twenty eight individual studies reported a significant reduction in PTH levels with vitamin D supplementation ( 135,138) and two studies found no change or an increase in PTH levels (110,139).

Vitamin D and Peak Wave Velocity
Eleven papers reported PWV as a primary or secondary outcome. Overall, there was no significant effect of vitamin D supplementation on PWV. The pooled effect size of vitamin D on PWV was −0.20 ± 0.13 m/s [95% CI −0.46 to 0.06, p = 0.13, I 2 = 72%) across all studies (Figure 10). Four individual studies reported a significant reduction in PWV in the group supplemented with vitamin D (77,82,114,128), two studies found a non-significant trend for reduction in PWV (88,131), and five trials did not find any significant effect of vitamin D on PWV (33, 84, 85, 100, 106).

Vitamin D and Augmentation Index
Ten studies reported AI as an outcome (Figure 11). Overall, vitamin D supplementation did not have an effect on AI. The pooled estimate (standardized mean difference) of the effect of vitamin D administration on AI was −0.09 ± 0.14% (95% CI −0.37 to 0.20%, p = 0.52, Two studies reported a significant decrease in AI (69,128), two studies reported a significant increase in AI (85,106), and six trials did not find any influence of vitamin D supplementation on AI (33,84,86,100,103,131).

Sensitivity Analysis
Using leave-one-out sensitivity analysis, the effect size of vitamin D remained significant for blood pressure, lipid profile, serum PTH, and serum hs-CRP, confirming that difference between treatment groups is not due to the effect of any single study.

Publication Bias
Visual inspection of funnel plot symmetry suggested potential publication bias for the comparison of systolic and diastolic BP, PTH, PWV, TC, HDL, and hs-CRP between vitamin Dadministered and placebo groups, though funnel plots for AI, TG, and LDL looked symmetric (Figures S1-S10). Egger's linear regressions did not indicate publication bias. After adjusting the effect size using the Trim and Fill method for potential Vitamin D and Cardiovascular Risk Factors publication bias, except for PWV, the effect size of vitamin D on lowering CVD risk parameters increased (BP, hs-CRP, PTH, AI, TC, LDL) or remained unchanged (TG, HDL) (Table S2).

Sub-Group Analysis
We investigated the effect of dose, achieved mean serum 25(OH)D, length of intervention, obesity, vitamin D deficiency, co-administration of calcium supplementation and age.

Achieved Serum 25(OH)D Concentrations at the End of the Trial
We investigated the effect of achieved serum 25(OH)D as high vs. low based on median levels ( Table 2)

Vitamin D Supplementation Dose
We investigated dose effect on outcomes by comparing trials in which subjects received ≥4,000 IU/d of vitamin D to those in which subjects received <4,000 IU/d ( Table 2)

Duration of Intervention
To investigate whether the length of the trial influenced the effect of vitamin D supplementation on outcomes we investigated trials that supplemented for ≥6 months in comparison with those <6 months in duration ( Table 2). In trials that assessed vitamin D supplementation for <6

Vitamin D Deficiency at Baseline
We investigated whether the effect of vitamin D on outcomes was dependent on vitamin D deficiency at baseline by comparing trials vitamin D deficient subjects at baseline [serum 25(OH)D

Calcium Co-administration
We investigated whether calcium co-administration influenced outcomes by comparing those trials with those that supplemented with vitamin D alone ( Table 2). Participants supplemented with both vitamin D and calcium had a significantly greater reduction in systolic BP (−0.19 ± 0.2 vs. −0.07 ±0.03 mmHg, p = 0.07) and TG levels (−0.18 ± 0.13 vs. −0.10 ±0.07 mmol/L, p = 0.11), compared with those who received vitamin D alone. There was no difference among the remaining parameters.

Effect of Participants' Age
Age itself is a risk factor for CVD and thus we compared trials that enrolled populations ≥55 y vs. <55 y. Vitamin D supplementation in trials with populations <55 y had significantly greater reduction in systolic BP (−0.14 ± 0.07 vs.−0.05 ± 0.05 mmHg, p = 0.07). There was no significant difference in other outcomes based on participant age grouping.

DISCUSSION
As the leading cause of death and disability worldwide, cardiovascular disease (CVD) is a major public health burden (140). Much effort has been devoted to identifying modifiable risk factors to prevent CVD. Vitamins may have a role in the prevention and treatment of CVD. Antioxidant vitamins such as vitamin C, vitamin E, folic acid and vitamin B6 might decrease the rate of oxidative stress, a key component of atherosclerosis and CVD (14). Vitamin D and folic acid can inhibit inflammation with their anti-atherogenic effects. Vitamin E can inhibit platelet aggregation and B vitamins might have anti-thrombotic activity by lowering serum homocysteine levels (14,141,142). Among these, vitamin D, with its deficiency highly prevalent worldwide and having many pleiotropic effects, has been associated with CVD prevention in different community settings. Vitamin D deficiency impairs vascular function and is strongly associated with the heightened risks of various cardiovascular diseases such as hypertension, metabolic syndrome, heart failure, and stroke (3,24).
Evidence suggests that vitamin D exerts beneficial cardiovascular effects through many pathways. Improved vitamin D status reduces RAAS activity and lowers blood pressure, it has anti-inflammatory, anti-proliferative, antihypertrophic, anti-fibrotic and anti-thrombotic impacts as well (111). Following vitamin D supplementation, suppression of renin production and downregulation of RAAS directly impacts myocardium and vasculature through modulating hypertrophic stimuli (143). Vitamin D inhibits the proliferation of vascular smooth muscle cells through influx of calcium into the cells, thus preserving endothelial function (144). Antihypertensive benefits of vitamin D include suppression of RAAS, an anti-proteinuric effect, a direct effect on endothelial cells and calcium metabolism as well as preventing secondary hyperparathyroidism (145,146). Vitamin D may have both direct and indirect impacts on modifying lipid profiles. Vitamin D supplementation might decrease serum levels of triglyceride via increasing the activity of lipoprotein lipase in adipose tissue (147). Also, through improving calcium absorption, vitamin D might reduce fatty acid absorption via the formation of insoluble calcium-fatty acid complexes in the gut leading to decreased LDL cholesterol levels (148). Yet, despite these observations, evidence linking corrections to vitamin D status with improved cardiometabolic parameters is somewhat inconclusive (24,43,149).
Considering the alternate postulation, vitamin D deficiency might be a consequence of chronic conditions such as inflammation. There is a bacterial pathogenesis theory explains that intracellular bacteria commonly seen in chronic inflammation might invade different nucleated cells and affect vitamin D metabolism and its endocrine function resulting in low vitamin D status. This occurs concurrent to increased production of 1,25(OH)2D which is required for upregulating vitamin D receptors to transcribes more adenosine monophosphate. And more 25(OH)D should be metabolized in this process leading to low vitamin D status (150,151). In another study conducted on the patients recovering from knee arthroplasty, there was a significant reduction in serum 25(OH)D levels during the process of systemic inflammatory response in these patients after surgery (152). Sattar et al. (153) also mentioned that vitamin D is an acute phase reactant and declines with the increase in inflammatory cytokine in different chronic conditions. Several mechanisms including decreased vitamin D carrier proteins, increased conversion of 25(OH)D to 1,25(OH)2D and hemodilution could be responsible for this reduction (154,155). However for CVD, using Hill's criteria for causality, Weyland found that all relevant Hill criteria are satisfied suggesting low 25(OH)D level is an independent risk factor for CVD (156).
To better understand this incongruence, we analyzed 81 studies that evaluated the effect of vitamin D supplementation on various cardiometabolic risk parameters, including blood pressure, serum PTH, hs-CRP, lipid profile, and arterial PWV and AI. Unlike many previous studies, we imposed several strict inclusion criteria to select only well-designed trials. Overall, vitamin D supplementation was found to improve cardiovascular risk factors. Specifically, vitamin D supplementation, with doses above 4,000 IU/d and increased serum 25(OH)D concentrations ≥86 nmol/L decreased systolic and diastolic blood pressure, serum PTH, serum hs-CRP and improved lipid profiles (total cholesterol, triglyceride, HDL and LDL). Markers of arterial stiffness (PWV and AI) may also improve with vitamin D supplementation.
Subgroup analyses revealed that the co-administration of calcium with vitamin D led to greater reductions in blood pressure. The combination of calcium with vitamin D has been suggested to improve blood pressure by facilitating calcium absorption into the blood stream and optimizing serum calcium and PTH levels (157,158). Although the greatest benefits of vitamin D supplementation can be achieved in vitamin D deficient populations, such that we observed as the lowering impact of vitamin D on serum PTH. Remarkably, we also found notable improvements in lipid profile in participants considered vitamin D sufficient prior to intervention. Further, individuals who were obese at baseline had a greater reduction in blood pressure, likely due to the higher percentage of obese individuals that are pre-hypertensive or hypertensive (159) and the higher daily doses of vitamin D provided to obese participants (61,76,81).
The results of our study compare closely with those from a number of recent meta-analyses. Jafari et al. (93), for instance, found significant reductions in the serum total cholesterol, triglyceride, and LDL levels of type 2 diabetics following vitamin D supplementation. Studying heterogeneous populations that consisted of healthy individuals, pregnant women, bedridden elderly people and those with different diseases (e.g., diabetes, heart failure, PCOS, and insulin resistant condition), both Chen et al. (160) and Rodriguez et al. (161) found that vitamin D supplementation significantly decreased inflammatory markers (i.e., hs-CRP). Chen et al. (160) additionally concluded that vitamin D supplementation led to a significantly greater reduction among those with baseline hs-CRP levels ≥5 mg/l. This is in line with increased hs-CRP levels in diabetic patients (64,69,92). Diabetes usually results in higher levels of hs-CRP and lower levels of 25(OH)D concentrations suggesting a larger effect size in subjects with this condition. Moreover, Wu et al. (162) and Witham et al. (163) found a significant modest reduction in blood pressure following vitamin D supplementation. Vitamin D supplementation may affect arterial stiffness and vascular aging through decreased synthesis of angiotensin II, following inhibition of RAAS, to increase vascular tone and arterial stiffness (164). However, limited data to assess the impact of vitamin D supplementation on the markers of arterial stiffness (PWV and AI) were inconclusive (43,165), and may be due to inappropriate study design including insufficient duration of supplementation and insufficient power (119).
In contrast, Beveridge et al. (21) found no significant reduction in blood pressure after participants whose mean SBP was ≥140 mm Hg at baseline were supplemented with vitamin D. Of note, there are significant methodological differences in our approach. Beveridge et al. (21) included trials that combined vitamin D with antihypertensive drugs, administered large bolus doses to elderly populations, included subjects with resistant HTN, and/or supplemented with very low doses of vitamin D (i.e., 600 IU). These issues could mask any effects of vitamin D supplementation or simply not lead to any observable benefits. Our strict inclusion criteria resulted in the exclusion of 14 of the 27 studies that were analyzed by Beveridge et al. We had similar concerns with the meta-analysis conducted by Wang et al. (26) who reported that vitamin D supplementation led to a statistically significant increase in LDL and included RCTs that provided very low doses of vitamin D (i.e., 300 IU) or supplemented for durations considered too short (i.e., 42 days). Of the 12 studies included, and as mentioned by the authors, none were sufficiently powered to detect changes in CVD outcomes. The current meta-analysis revealed significant impact of vitamin D supplementation on lipid profiles with increased HDL and reduced LDL and TG.
Many observational studies support an association between cardiovascular risk factors and low vitamin D status. Perhaps most importantly, what constitutes vitamin D deficiency and repletion is somewhat debateable and, at times, contentious. The Institute of Medicine issues dietary recommendations, such as the Recommended Dietary Allowance (RDA), at the request of the U.S. and Canadian governments. In 2010, the Institute of Medicine set the RDA for vitamin D at 600 IU per day for individuals between the age of 1 and 70 (48). This RDA is assumed to achieve serum 25(OH)D levels of ≥50 nmol/l in 97.5% of the population. The methodology used to calculate this RDA, however, has been deemed erroneous (166) and estimates of much higher magnitude have been calculated by others -3,875 IU/day (167) to 8,895 IU/day (168).
Similarly, the definition of what is an "optimal" serum 25(OH)D concentration is also controversial. Serum 25(OH)D concentrations >75 nmol/l (12) and >80 nmol/l (28) have been suggested as necessary for lipid and cardiovascular health. Serum 25(OH)D concentrations of 100-150 are defined as optimal by the U.S. Endocrine Society with values below 75 nmol/l deemed insufficient (169,170). The results of the present meta-analysis suggest that serum 25(OH)D concentrations ≥86 nmol/L are optimal for reductions in blood pressure, markers of arterial stiffness, and reductions in hs-CRP. It is important to note that these serum 25(OH)D concentrations were achieved with vitamin D supplement doses ≥4,000 IU/d-the current tolerable upper level of intake. Using the standards of the U.S. Endocrine Society, 27 of the 81 included studies in our metaanalyses reached optimal 25(OH)D levels post-supplementation, and only 16 had post-supplement 25(OH)D levels that were insufficient.
The duration of supplementation is an important factor in assessments of vitamin D. With a half-life of 2 months, to achieve and maintain a steady serum 25(OH)D concentration requires a follow-up period of at least 3 months. Here, we included trials that ranged from 3 months to 5 years of intervention in the meta-analysis. Somewhat surprisingly, we found better improvement in some outcomes (blood pressure and lipids) in trials that were less than 6 months, although this is likely related to higher compliance in short-term interventions (171). Improvements in blood pressure and lipid profile were also witnessed in a short, 3 month vitamin D intervention in obese PCOS patients (116). The women received 12,000 IU/d of vitamin D for an increase in their serum 25(OH)D of 50-168 nmol/L. In contrast, a 5 year trial of obese and vitamin D-insufficient prediabetics provided 2,800 IU/d of vitamin D found no change in blood pressure (50). It is known that overweight and obese individuals require two to three times the amount of vitamin D to increase serum 25(OH)D concentrations to the same extent as those with a normal BMI (31,47).
The present study has several strengths and limitations. Even after enforcing a strict inclusion criteria, the included studies varied with regard to participant age, serum 25(OH)D concentrations at baseline, concurrent use of other nutrients or medication, and overall health status. We used a randomeffect model and performed sensitivity analyses to mitigate these limitations. For some of the studies, cardiovascular outcomes of interest were secondary outcomes or the trial was not of sufficient power to detect a change in these outcomes. Many of the studies also did not describe dietary intakes, season of treatment, or sun exposure. Further, some included trials assessed relatively small populations (10-13 participants per intervention group), but taken together offer support to the larger trials. Strengths include the large sample and the consideration of a wide variety of CVD risk parameters from at least 28 clinical trials for each CVD outcome (with the exception of arterial PWV and AI).

CONCLUSION
Vitamin D deficiency is a highly prevalent condition and is independently associated with most CVD risk factors. The present meta-analysis demonstrated that vitamin D supplementation improved serum 25(OH)D concentrations significantly lowered blood pressure, serum PTH, hs-CRP, TC, LDL, and TG and increased HDL. Vitamin D supplementation also appears to improve arterial stiffness (PWV), but large and well-designed RCTs are required to confirm these findings. The present analysis suggests that for improvements in CVD risk factors vitamin D supplementation ≥4,000 IU/d and achieved serum 25(OH)D concentrations ≥86 nmol/L are required.

AUTHOR CONTRIBUTIONS
NM, JR, and SK designed the study, NM and JR searched databases and performed the selection of studies. NM, JR, and SK wrote the manuscript. NM analyzed the data. SK and JR critically evaluated the review, commented on it, and approved the last version. All authors reviewed and approved the final manuscript. SK is the guarantor of this study.

ACKNOWLEDGMENTS
The authors would like to thank Ken Fyie for preparing funnel plots and Brian Rankin for reviewing this manuscript.