Vitamin D deficiency causes rickets in children and osteomalacia in adults and may contribute to the causation of CVD. Cholecalciferol (vitamin D3) is made by the action of UVB light on the skin but is also provided in the diet from foods of animal origin (eggs, oily fish and meat). Ergocalciferol (vitamin D2) is present in fungi that have been exposed to UVB irradiation. Both forms of vitamin D undergo 25-hydroxylation in the liver to form 25(OH) metabolites, most of which circulate in plasma bound to the vitamin D binding protein (VDBP). In the kidney, 25 hydroxyvitamin D (25(OH)D) is converted to the biologically active metabolite 1,25 dihydroxyvitamin D (1,25(OH)2D) by the action of 1α-hydroxylase (Fig. 1). This then binds to the vitamin D receptor, which is a high-affinity nuclear hormone receptor that regulates gene expression( Reference Norman 1 ). Measurement of serum 25(OH)D to assess vitamin D status has predominantly used immunoassays, which show lower sensitivity (about 30 %) to D2 metabolites than gold-standard methods such as HPLC/tandem MS. Formerly, a lack of standardisation of methods and appropriate quality control made comparisons between studies difficult( Reference Binkley, Krueger and Cowgill 2 ). The situation has been remedied by the availability( Reference Phinney 3 ) of standard reference serum (SRM 972, National Institute of Standards and Technology) and a quality control programme in the UK organised by the Vitamin D External Quality Assessment Scheme( 4 ). It has been suggested that the free, unbound form of 25(OH)D may provide a more accurate index of vitamin D status( Reference Chun, Peercy and Orwoll 5 , Reference Glendenning, Chew and Inderjeeth 6 ) than the total 25(OH)D( Reference Bikle and Gee 7 , Reference Safadi, Thornton and Magiera 8 ). Powe et al. found that black Americans had lower total 25(OH)D concentrations compared with whites, but because they also had lower concentrations of VDBP, the amount of bioavailable 25(OH)D for both ethnic groups was similar( Reference Powe, Evans and Wenger 9 ). The same group found free 25(OH)D concentrations to be better correlated with bone mineral density than the total 25(OH)D concentrations in young adults( Reference Powe, Ricciardi and Berg 10 ). Prospective cohort studies have provided evidence that indicates that low concentrations of total 25(OH)D are associated with an increased risk of CVD( Reference Anderson, May and Horne 11 , Reference Kendrick, Targher and Smits 12 ). The discovery that 1α-hydroxylase is expressed in extra-renal tissues, including the vascular endothelium( Reference Zehnder, Bland and Chana 13 – Reference Merke, Milde and Lewicka 16 ) may help provide a mechanistic basis for an effect of vitamin D status on the processes associated with the development of CVD.
Metabolism of vitamin D in the body. Vitamin D from exposure of the skin to UVB radiation and dietary sources is transported in the circulation to the liver where 25 hydroxyvitamin D (25(OH)D) is formed by the action of 25-hydroxylase (25-OHase). 25(OH)D is then transported to the kidneys where the enzyme 25-hydroxyvitamin D-1α-hydroxylase (1α-OHase) converts it to the active form, 1,25 dihyroxyvitamin D (1,25(OH)2D). When serum calcium levels are low, parathyroid hormone (PTH) is released which stimulates a greater production of 1,25(OH)2D. This has the effect of increasing serum calcium (Ca2+) concentrations by increasing bone Ca2+ re-absorption, increasing the intestinal absorption of Ca2+ and phosphate, and decreasing the renal excretion of calcium. 24-hydroxylase (24-OHase) converts both 25(OH)D and 1,25(OH)2D to metabolites which can be excreted from the body. 1,25(OH)2D has also been shown to have effects on pancreatic β cells, immune cells and vascular smooth muscle cells (VSMC).
Sunlight exposure v. dietary vitamin D
Exposure to UVB irradiation increases plasma 25(OH)D concentrations( Reference Ashwell, Stone and Stolte 17 ) and explains the seasonal variations observed in vitamin D status. Skin pigmentation may reduce the capacity to synthesise vitamin D as melanin skin pigmentation competes for and absorbs UVB photons that are responsible for the photolysis of 7-dehydrocholesterol to previtamin D. It has been suggested that an individual with heavily melanised skin may need greater sunlight exposure than one with white skin to make the equivalent amount of vitamin D( Reference Clemens, Adams and Henderson 18 ). Migrants with darker skin moving to northern latitudes are particularly at risk of developing vitamin D deficiency( Reference Solanki, Hyatt and Kemm 19 ). The use of a sunscreen or clothing/headscarves which cover the skin can also reduce skin synthesis of vitamin D( Reference Holick, Binkley and Bischoff-Ferrari 20 – Reference Pearce and Cheetham 22 ). Older age may lead to a decrease in 25(OH)D concentrations as ageing has been shown to be associated with decreased concentrations of 7-dehydrocholesterol in the skin( Reference MacLaughlin and Holick 21 ) and older people may be less likely to spend time outdoors, particularly if they are institutionalised.
Diet has a lesser effect on vitamin D status and median intakes from food sources in the most recent National Diet and Nutrition Survey Rolling Programme for which the data have been published on information collected between 2008 and 2012 were 2·5 μg/d in men (n 1126) and 2·1 μg/d (n 1571) in women for adults aged 19–64 years( Reference Bates, Lennox and Prentice 23 ). These intakes are well below the recommendation of 10 μg/d for those confined indoors( 24 ) which becomes relevant for most of the UK population in the winter months when there is no UVB exposure. The recommendation assumed that UVB exposure in the summer months would be sufficient to tide most people over the winter months, but up to 40 % of adults and adolescents in the UK have 25(OH)D concentrations <25 nmol/l from January to March( Reference Bates, Lennox and Prentice 23 ).
There is some controversy as to the bioequivalence of D2 and D3. Animal studies find D2 to be less potent than D3 in curing rickets( Reference Chen and Bosmann 25 ), but D3 is more toxic than D2 ( Reference Hunt, Garcia and Walsh 26 ). In the chick bioassay, vitamin D2 had only 8–11 % of the activity of D3 in preventing rickets( Reference Chen and Bosmann 25 ). However, there are large species differences and D2 has been used effectively in infant formula and to treat hypovitaminosis D in young children( Reference Gordon, Williams and Feldman 27 ) for many decades and pharmacopoeias still regard the two forms as equivalent and interchangeable( 28 , 29 ). Houghton and Vieth dispute this view arguing that D3 is more effective than D2 and have recommended that D2 should no longer be regarded as suitable for supplementation( Reference Houghton and Vieth 30 ). In support of this, a recent meta-analysis of seven randomised controlled trials (RCT) which gave vitamin D as a bolus or daily dose concluded that D2 was less effective in raising plasma 25(OH)D metabolite levels than D3 (weighted mean difference of 15·2 nmol/l(95 % CI 6·1, 24·3))( Reference Tripkovic, Lambert and Hart 31 ). However, the comparison for daily vitamin D supplementation in the meta-analysis found no significant difference between the two forms (weighted mean difference 4·8 nmol/l(95 % CI −1·0, 10·6)). There was substantial heterogeneity in the design and dose formulation in the studies included though and some did not adjust for sunlight exposure( Reference Binkley, Gemar and Engelke 32 , Reference Armas, Hollis and Heaney 33 ) or used assays that were unable to satisfactorily discriminate between 25(OH)D2 and 25(OH)D3 ( Reference Phinney 3 Reference Phinney 3 – Reference Heaney, Recker and Grote 36 ). Some studies also gave pharmacological doses of vitamin D by non-oral routes (i.e. intramuscular administration of 7500 μg( Reference Leventis and Kiely 37 )) or in tablet or capsule form (i.e. a single dose of 7500 μg( Reference Romagnoli, Mascia and Cipriani 34 ) or weekly dose of 1250 μg( Reference Heaney, Recker and Grote 36 )) which are not relevant to dietary intake. We showed that intakes of 5 and 10 μgD2/d for 4 weeks led to mean increments in 25(OH)D2 of 9·4 (se 2·5) and 17·8 (se 2·4) nmol/l, respectively, and increments in 25(OH)D3 of 15·1 (se 4·7) and 22·9 (se 4·6) nmol/l following 5 and 10 μgD3/d, respectively( Reference Fisk, Theobald and Sanders 38 ) which are equivalent to an increase of approximately 2 nmol/l per μg vitamin D. Including our results with comparable trials (Fig. 2) with daily intakes in the nutritional range (5–25 μg/d)( Reference Biancuzzo, Young and Bibuld 39 , Reference Holick, Biancuzzo and Chen 40 ) shows no major difference with regard to effects on plasma 25(OH) metabolite concentrations between D2 and D3.
(Colour online) Meta-analysis of placebo-controlled trials( Reference Fisk, Theobald and Sanders 38 – Reference Holick, Biancuzzo and Chen 40 ) comparing the effect of vitamin D doses ranging from 5 to 25 μg/d on total 25(OH)D concentrations measured by HPLC/tandem MS. A negative effect size favours vitamin D3.
Influence of obesity, inflammation and genetic influences on vitamin D status
Obesity is associated with lower 25(OH)D concentrations( Reference Wortsman, Matsuoka and Chen 41 ). It has been suggested that vitamin D is sequestered in body fat compartments leading to reduced bioavailability( Reference Wortsman, Matsuoka and Chen 41 ), but animal adipose tissue is generally a poor source of the vitamin, so this may not explain the association. Increased sedentary behaviour resulting in less sunlight exposure, or inflammation induced by obesity are other possible explanations. VDBP is a negative acute phase protein (i.e. its concentration falls when there is systemic inflammation)( Reference Waldron, Ashby and Cornes 42 ). As plasma concentrations of VDBP are a major determinant of plasma 25(OH)D concentrations, disorders that are characterised by chronic systemic inflammation may lower plasma 25(OH)D concentrations. Certain drugs also increase the catabolism of 25(OH)D such as phenytoin, an anticonvulsant drug. Heredity may be important as twin studies find 40 % of the variation in 25(OH)D concentrations( Reference Hunter, De Lange and Snieder 43 ) to be attributable to additive genetic factors. Some of this variation may be due to different VDBP genotypes( Reference Fu, Yun and Oczak 44 ).
Plasma 25 hydroxyvitamin D concentrations and risk of CVD
A meta-analysis of eight large prospective cohort studies in which 2624 participants died of CVD during follow-up found that individuals in the lowest quintile of 25(OH)D concentrations had an increased risk of all-cause mortality and cardiovascular mortality with or without a history of CVD; relative risk (RR;95 % CI) 1·57 (1·36, 1·81), 1·65 (1·22, 2·22) and 1·41 (1·18, 1·68), respectively( Reference Schottker, Jorde and Peasey 45 ). An umbrella review of meta-analyses of observational studies also found significant reductions in the risk of prevalent CVD, IHD and stroke; RR (95 % CI) 0·67 (0·55, 0·82), 0·72 (0·65, 0·81) and 0·61 (0·50, 0·75), respectively( Reference Theodoratou, Tzoulaki and Zgaga 46 ). However, these observations may be confounded by other aspects of lifestyle, such as involvement in physical work outdoors, or disease processes resulting in inflammation which influence plasma 25(OH)D concentrations.
Potential mechanisms for an effect of vitamin D on CVD risk
CVD results from the inter-related processes of atherosclerosis and thrombosis. Atherosclerosis develops over several decades, whereas thrombosis is an acute process usually resulting from the rupture of an atherosclerotic plaque that causes a clinical event (heart attack or stroke). Elevated serum LDL-cholesterol and atherogenic dyslipidaemia (low HDL-cholesterol and a predominance of small dense LDL particles) result in atherosclerosis and so are key risk factors for CVD, especially CHD. Elevated blood pressure (BP) is the major risk factor for stroke but it increases the risk of CHD to a lesser extent. Vitamin D may influence the classical risk factors (BP and lipids) as well as other processes (plaque rupture, endothelial dysfunction, arterial stiffening, inflammation and thrombosis; Fig. 3).
Potential mechanisms for an effect of vitamin D on CVD. hsCRP, high-sensitivity C-reactive protein; MMP-9, matrix metalloproteinase-9; NO, nitric oxide; PTH, parathyroid hormone.
Effects of vitamin D on vascular function
1,25(OH)2D suppresses renin secretion by the kidney which results in decreased conversion of angiotensinogen to angiotensin I. Angiotensin 1 is converted in the pulmonary circulation to the vasoconstrictor angiotensin II which causes blood vessels to constrict, thus increasing BP( Reference Li 47 , Reference Li, Qiao and Uskokovic 48 ). Vitamin D may also affect BP via changes in parathyroid hormone which has been found in the vascular endothelium and smooth muscle cells( Reference Fitzpatrick, Bilezikian and Silverberg 49 ). Observational studies have found a positive association between parathyroid hormone and systolic and diastolic BP( Reference Snijder, Lips and Seidell 50 ), and parathyroid hormone intravenous infusion has been shown to increase BP in normal subjects( Reference Hulter, Melby and Peterson 51 ). Alternatively, 1, 25(OH)2D may have direct effects on the vasculature which is supported by the expression of 1α-hydroxylase in endothelial cells and evidence from experimental studies that 1,25(OH)2D may affect rat vascular smooth muscle cell growth( Reference Carthy, Yamashita and Hsu 52 ).
Disruption of the functional integrity of the vascular endothelium is an important process in the development of atherosclerosis( Reference Bonetti, Lerman and Lerman 53 ). Impairment of endothelium-dependent vasodilation occurs due to a decrease in the bioavailability of vasodilators, including nitric oxide, and an increase in endothelium-derived contracting factors( Reference Griendling and FitzGerald 54 ). It has been shown in human umbilical vein endothelial cells that 1,25(OH)2D3 can cause a significant concentration-dependent increase in endothelial nitric oxide production via endothelial nitric oxide synthase activation( Reference Molinari, Uberti and Grossini 55 ). This is supported by experimental work, which has found that mice carrying a mutant, functionally inactive vitamin D receptor have a lower bioavailability of nitric oxide as a result of a reduced expression of endothelial nitric oxide synthase( Reference Andrukhova, Slavic and Zeitz 56 ).
Arterial stiffness is determined by structural and functional components of the artery( Reference Arnett, Evans and Riley 57 ). Normally the balance of collagen and elastin in the vascular wall is kept in a stable state( Reference Wagenseil and Mecham 58 ), but elastin can be degraded by proteases such as matrix-metalloproteinases (MMP) which have been shown to be inhibited by 1,25(OH)2D in experimental studies in Myobacterium tuberculosis-infected human leucocytes( Reference Coussens, Timms and Boucher 59 ) and suppressed in studies in pulmonary tuberculosis patients( Reference Anand and Selvaraj 60 ). This can result in the production of more collagen fibres which are 100–1000 times stiffer than elastic fibres, and therefore the mechanical properties of the artery are shifted into the stiffer range. Vitamin D deficiency may lead to arterial stiffening via increased vascular calcification( Reference Watson, Abrolat and Malone 61 ) whereby calcium deposits build up in the media of large arteries( Reference Wagenseil and Mecham 58 ). This may also be mediated by MMP as MMP-2 and MMP-9 knockout mice did not develop abdominal aortic calcification after treatment with CaCl2, whereas wild-type mice developed severe calcification( Reference Basalyga, Simionescu and Xiong 62 ). Furthermore, as vitamin D has been shown to suppress smooth muscle cell proliferation( Reference Chen, Law and Gardner 63 ) it may help prevent an increase in smooth muscle tone or smooth muscle cell hypertrophy which contributes to an increase in vascular stiffness( Reference Arnett, Evans and Riley 57 , Reference Zieman, Melenovsky and Kass 64 ).
Effects of vitamin D on inflammation
Experimental studies have shown that treatment with 1,25(OH)2D inhibits the production of several pro-inflammatory cytokines, including IL-6 and TNF-α in partially purified monocytes( Reference Muller, Haahr and Diamant 65 , Reference Cohen-Lahav, Douvdevani and Chaimovitz 66 ) and down-regulates IL-10 expression via the binding of the vitamin D receptor to the promoter region of the IL-10 transcription start site( Reference Matilainen, Husso and Toropainen 67 ). 1,25(OH)2D has been shown to inhibit NF-κB activation in passively sensitised human airway smooth muscle cells( Reference Song, Hong and Liu 68 ) and murine macrophage cells( Reference Cohen-Lahav, Douvdevani and Chaimovitz 66 ). Therefore it is possible that reductions in C-reactive protein (CRP) concentrations could occur via effects of 1,25(OH)2D on the NF-κB pathway, which plays an important role in regulating the inflammatory response. A proinflammatory environment also promotes the production of MMP which mediate a matrix remodelling process in which the fibrous cap in atherosclerotic lesions is degraded leading to an increased likelihood of plaque rupture( Reference Zittermann, Schleithoff and Koerfer 69 ).
Effects of vitamin D on insulin secretion
1α-hydroxylase and receptors for 1,25(OH)2D have been found in pancreatic β cells( Reference Bland, Markovic and Hills 70 ) and vitamin D appears necessary for normal insulin release( Reference Teegarden and Donkin 71 ). Insulin secretion may be affected by changes in the balance between extracellular and intracellular pools of calcium as insulin secretion is a calcium-dependent process( Reference Pittas, Lau and Hu 72 ).
Epidemiological and clinical trial data on the association between vitamin D and CVD risk factors
Blood pressure
Many cross-sectional and prospective studies have investigated a possible association between vitamin D status and BP. A meta-analysis of four prospective and fourteen cross-sectional studies found a pooled OR of 0·73 (95 % CI 0·63, 0·84) for hypertension in participants in the highest 25(OH)D category compared with those in the lowest category( Reference Burgaz, Orsini and Larsson 73 ). A number of RCT have measured changes in BP (primarily clinic BP) in response to vitamin D supplementation and shown mixed results( Reference Asemi, Samimi and Tabassi 74 – Reference Wood, Secombes and Thies 80 ). A recent meta-analysis of sixteen RCT found no significant treatment effects; weighted mean differences were −0·94 (95 % CI−2·98, 1·10) mm Hg and −0·52 (95 % CI−1·18, 0·14) mm Hg for systolic BP and diastolic BP, respectively( Reference Kunutsor, Burgess and Munroe 81 ). However, 24 h ambulatory BP monitoring has been shown to be better at predicting cardiovascular risk compared with clinic BP( Reference Mancia, De Backer and Dominiczak 82 , Reference Staessen, Thijs and Fagard 83 ) and the evidence has demonstrated that it is able to more accurately measure the size of reduction in BP brought about by a treatment compared with clinic BP, due to the results being more reproducible over time( Reference Coats, Radaelli and Clark 84 ), and the unlikelihood of a ‘white coat’ or placebo effect( Reference Mancia, De Backer and Dominiczak 82 , Reference Mancia, Omboni and Parati 85 ). Only a few RCT have used 24 h ambulatory BP monitors( Reference Larsen, Mose and Bech 86 – Reference Witham, Ireland and Houston 89 ) (Table 1) and these were in participants who had hypertension( Reference Larsen, Mose and Bech 86 , Reference Witham, Price and Struthers 88 , Reference Witham, Ireland and Houston 89 ) or had previously suffered from a stroke( Reference Witham, Dove and Sugden 87 ). None showed a significant difference in the change in 24 h BP between treatment and placebo groups after treatment with 75 μgD3/d for 20 weeks( Reference Larsen, Mose and Bech 86 ), a single oral dose of 2500 μg D2 with 24 h BP measured after 16 weeks( Reference Witham, Dove and Sugden 87 ), 2500 μg D3 taken every 3 months for a year( Reference Witham, Price and Struthers 88 ) or 2500 μg D3 taken every 2 months for 6 months( Reference Witham, Ireland and Houston 89 ). There is a need for further well-conducted vitamin D RCT measuring 24 h ambulatory BP in healthy older participants.
Summary of clinical trials investigating the effect of vitamin D supplementation on 24-h ambulatory blood pressure
NA, not available; SBP, systolic blood pressure; DBP, diastolic blood pressure.
Endothelial function
Endothelial function can be measured non-invasively using flow-mediated dilation of the brachial artery( Reference Yeboah, Crouse and Hsu 90 ). A couple of cross-sectional studies in healthy adults show lower serum levels of 25(OH)D to be associated with endothelial dysfunction( Reference Al Mheid, Patel and Murrow 91 , Reference Jablonski, Chonchol and Pierce 92 ). A few trials show an improvement in flow-mediated dilation with vitamin D supplementation( Reference Tarcin, Yavuz and Ozben 93 – Reference Sugden, Davies and Witham 95 ), but most show no effect( Reference Witham, Dove and Dryburgh 79 , Reference Witham, Adams and Kabir 96 – Reference Gepner, Ramamurthy and Krueger 98 ) (Table 2).
Summary of clinical trials investigating the effect of vitamin D supplementation on endothelial function measured using flow-mediated dilation (FMD) and arterial stiffness measured as pulse wave velocity (PWV) in individuals without pre-existing CVD
NA, not available.
Arterial stiffness
Arterial stiffness is most reliably measured by determining the carotid–femoral aortic pulse wave velocity ( Reference Mattace-Raso, van der Cammen and Hofman 99 , Reference Terai, Ohishi and Ito 100 ). Many cross-sectional studies have shown inverse associations between vitamin D and arterial stiffness( Reference Al Mheid, Patel and Murrow 91 , Reference Mayer, Filipovsky and Seidlerova 101 – Reference Rezai, Wallace and Sattar 107 ), but there are relatively few clinical trials that have investigated the effect of vitamin D supplementation (Table 2). Gepner et al. conducted a study in which healthy postmenopausal women were randomised to 62·5 μg vitamin D3 daily (n 57) or placebo (n 57) for 4 months. No significant differences were found between groups in the change in pulse wave velocity (P = 0·65) measured using SphygmoCor. This lack of an effect may be due to the women's baseline concentrations of total 25(OH)D being quite high: mean 78·1 (sd 26·5) nmol/l( Reference Gepner, Ramamurthy and Krueger 98 ). However, an 8-week RCT in South Asian women with mean baseline serum 25(OH)D concentrations of 27 nmol/l also found no change in pulse wave velocity measured using SphygmoCor in a group given a single oral dose of 250 μg D3/d compared with placebo (P = 0·046)( Reference Witham, Adams and Kabir 96 ). Two other trials have been conducted in subjects with type 2 diabetes mellitus( Reference Yiu, Yiu and Siu 97 , Reference Breslavsky, Frand and Matas 108 ) or hypertension( Reference Larsen, Mose and Bech 86 ), but both showed no change in arterial stiffness in response to vitamin D treatment( Reference Larsen, Mose and Bech 86 , Reference Yiu, Yiu and Siu 97 ). None of the trials have lasted longer than 20 weeks and if vitamin D prevents the age-related increase in arterial stiffness( Reference Benetos, Waeber and Izzo 109 ) due to inhibition of vascular calcification( Reference Watson, Abrolat and Malone 61 , Reference de Boer, Kestenbaum and Shoben 110 ), it is likely that longer term studies are needed to see a change.
Inflammation
Inflammation plays an important pathogenic role in all stages of atherosclerosis. Currently, the best available inflammatory biomarker is CRP, an acute phase reactant that is a reliable, independent predictor of the risk of myocardial infarction, stroke and cardiovascular death( Reference Torres and Ridker 111 ). Clinical trials of vitamin D supplementation show inconsistent results regarding an effect on CRP concentrations: most report no effect( Reference Muldowney, Lucey and Hill 77 , Reference Wood, Secombes and Thies 80 , Reference Yiu, Yiu and Siu 97 , Reference Asemi, Hashemi and Karamali 112 – Reference von Hurst, Stonehouse and Coad 116 ), whereas some show reductions in CRP concentrations( Reference Asemi, Samimi and Tabassi 74 , Reference Timms, Mannan and Hitman 117 , Reference Witham, Dove and Khan 118 ). Some trials have measured high-sensitivity CRP (hsCRP), which can detect low-grade inflammation. Witham et al. gave patients with a history of myocardial infarction 2500 μg D3 or placebo at baseline, 2 months and 4 months and found that after 6 months CRP had decreased significantly in the D3 group compared with placebo (−1·3 v. 2·0 mg/l, P = 0·03)( Reference Witham, Dove and Khan 118 ). Similarly, a significant decrease in hsCRP concentrations was observed after 9 weeks of 10 μg/d D3 (mean −1·41(se 0·55) mg/l), compared with placebo (mean 1·50(se 0·94) mg/l; P = 0·01) in forty-eight pregnant women aged 18–40 years at week 25 of gestation( Reference Asemi, Samimi and Tabassi 74 ). In contrast, two larger placebo-controlled RCT that supplemented type 2 diabetic patients with 125 μg/d D3 for 12 weeks( Reference Yiu, Yiu and Siu 97 ), or overweight and obese subjects with 1000 or 500 μg D3 per week for 1 year( Reference Jorde, Sneve and Torjesen 114 ), found no significant difference in the change in hsCRP between groups. Furthermore, after smaller vitamin D3 doses of 10 or 25 μg/d for 1 year( Reference Wood, Secombes and Thies 80 ), or 5, 10 or 15 μg/d for 22 weeks in healthy young and older adults( Reference Muldowney, Lucey and Hill 77 ), hsCRP levels were unaffected. However, a meta-analysis, including ten RCT in 924 participants published between 2009 and 2014 concluded that vitamin D supplementation (median dose 100 μg/d with a range of 10–179 μg/d) significantly reduces hsCRP by 1·08 mg/l (95 % CI −2·13, −0·03) compared with the control( Reference Chen, Wan and Han 119 ) but there was significant heterogeneity between trials.
Insulin sensitivity and secretion
Several large epidemiological studies have found 25(OH)D concentrations to be associated with improved insulin resistance and sensitivity( Reference Chiu, Chu and Go 120 – Reference Liu, Meigs and Pittas 122 ), but RCT have found mixed results; some have found a significant beneficial effect of supplementation( Reference Nagpal, Pande and Bhartia 78 , Reference von Hurst, Stonehouse and Coad 116 , Reference Nikooyeh, Neyestani and Farvid 123 ), while others have shown no effect( Reference Jorde, Sneve and Torjesen 76 , Reference Patel, Poretsky and Liao 124 ). An RCT in 100 apparently healthy, but centrally obese men observed an improvement in postprandial insulin sensitivity measured by the oral glucose insulin sensitivity index after supplementation with 3000 μg D3 once a fortnight for 6 weeks: mean difference in change in oral glucose insulin sensitivity between placebo and the D3 group was 41·1 (se 15·5), P = 0·01( Reference Nagpal, Pande and Bhartia 78 ). Similarly, Nikooyeh et al. ( Reference Nikooyeh, Neyestani and Farvid 123 ) conducted a 12-week D3 supplementation trial in ninety diabetic subjects and showed a significant decrease in the homeostasis model assessment of insulin resistance after 25 μg/d D3 compared with placebo (P < 0·001). In contrast to these findings, an RCT of 438 overweight or obese subjects found no significant difference in insulin sensitivity in the fasting state from baseline to 1 year between the placebo group and two D3 groups that received either 500 μg or 1000 μg per week( Reference Jorde, Sneve and Torjesen 76 ). Furthermore, 10 or 30 μg/d of D3 did not lead to significant improvements in insulin sensitivity in a 4-month RCT of patients with type 2 diabetes( Reference Patel, Poretsky and Liao 124 ). A meta-analysis has been published to summarise the effects of vitamin D supplementation on glycaemic control and insulin resistance( Reference George, Pearson and Witham 125 ). For insulin resistance, ten studies were included, four of which were in individuals with normal fasting glucose and six in individuals with abnormal glucose tolerance. When all ten studies were combined there was no effect of vitamin D supplementation on insulin resistance; the standard mean difference favouring vitamin D supplementation was −0·07 (95 % CI −0·20, 0·06). This was also the case when looking at only patients with normal fasting glucose, but in patients with diabetes or impaired glucose tolerance, there was a small improvement in insulin resistance (standard mean difference favouring vitamin D supplementation −0·25(95 % CI −0·48, −0·03))( Reference George, Pearson and Witham 125 ). In conclusion, it would appear that vitamin D does not affect insulin sensitivity and that the association of plasma 25(OH)D with insulin resistance may be because VDBP is a negative acute phase reactant.
Randomised controlled trials of the effect of vitamin D supplementation on mortality/clinical endpoints for CVD
A meta-analysis of fifty-six RCT found that vitamin D supplementation slightly decreased mortality risk (RR 0·97 (95 % CI 0·94, 0·99))( Reference Bjelakovic, Gluud and Nikolova 126 ), but other meta-analyses suggest the protective effects against all-cause mortality if supplementation was given for longer than 3 years (RR 0·95 (95 % CI 0·90, 0·98))( Reference Zheng, Zhu and Zhou 127 ), and in D3 trials, but not D2 trials; RR were 0·89 (95 % CI 0·80, 0·99) and 1·04 (95 % CI 0·97, 1·11), respectively( Reference Chowdhury, Kunutsor and Vitezova 128 ). No significant effects on myocardial infarction or IHD, or stroke or CVD were found in a meta-analysis of RCT by Bolland et al.; RR were 1·02 (95 % CI 0·93, 1·13) for the former outcome (nine trials, 48 647 patients) and 1·01 (95 % CI 0·90, 1·13) for the latter outcome (eight trials 46 431 patients)( Reference Bolland, Grey and Gamble 129 ). However, some of the trials included had compared calcium and vitamin D with a placebo or control and it is not always possible to disaggregate the calcium/vitamin D effects. If calcium has a negative effect on CVD, it could mask a potential protective effect of vitamin D( Reference Bolland, Grey and Avenell 130 ). Ford et al.( Reference Ford, MacLennan and Avenell 131 ) also found no significant effects in a meta-analysis of twenty-one vitamin D supplementation RCT in participants aged ≥60 years which attempted to adjust for calcium use: hazard ratios for cardiac failure, myocardial infarction and stroke were 0·82 (95 % CI 0·58, 1·15), 0·96 (95 % CI 0·83, 1·10) and 1·07 (95 % CI 0·91, 1·29), respectively. The findings of these meta-analyses are inconclusive and the potential effect size is much smaller than indicated by the observational studies. However, the trials were not designed and powered to assess the effects of vitamin D on CVD as a primary outcome. There are a few large trials in progress mostly using high doses of vitamin D3 which will report in a few years time (Finnish Vitamin D Trial NCT01463813, Vitamin D and Omega-3 Trial NCT01169259 and Vitamin D Assessment Trial ACTRN12611000402943). As high doses of vitamin D have potential for toxicity, there is also a need for trials using more physiological doses of vitamin D. Trials involving appropriate sun exposure would ensure that toxic levels of vitamin D are not reached as any excess previtamin D3 or vitamin D3 from UVB exposure is destroyed by sunlight, but concerns over the risk of skin cancer would make these difficult to implement safely.
Risks and benefits of obtaining vitamin D from food, fortified foods, dietary supplements and sunlight/UVB irradiation
Although exposure to UVB irradiation is the most effective means of increasing vitamin D status, it is associated with a major risk of skin cancer( Reference Ashwell, Stone and Stolte 17 ). Many authorities now advise restricting exposure to sunlight and promote the use of UVB filters in cosmetics and sun lotions, which decrease the capacity to synthesise vitamin D( Reference Matsuoka, Ide and Wortsman 132 ). For certain at-risk groups such as people with melanised skin, those confined indoors and individuals who cover up their skin when outdoors, dietary intake of vitamin D is particularly important for maintaining an adequate vitamin D status, especially during the winter months when there is negligible UVB radiation in the UK. We have shown that increasing oily fish intake can significantly increase vitamin D status. In the CRESSIDA RCT (ISRCTN92382106), a group following UK dietary guidelines, including advice to consume one to two portions of oily fish per week for 12 weeks were compared with a control group based on a conventional British dietary pattern that included less than one serving per month of oily fish. Mean vitamin D intake increased from 3·0 to 6·6 μg/d in the dietary guideline group and serum 25(OH)D concentrations at 12 weeks v. control were 9·2 nmol/l (95 % CI 4·2, 14·2; P < 0·001) greater( Reference Fisk, Reidlinger and Sanders 133 ). However, it has been suggested that intakes of 9 and 28 μg vitamin D /d would be needed to maintain winter 25(OH)D concentrations above 25 or 50 nmol/l, respectively, in 97·5 % of the population( Reference Cashman, Hill and Lucey 134 , Reference Cashman, Wallace and Horigan 135 ). Diet alone may be insufficient to meet needs in all people and consideration needs to be given to supplementation and fortification of foods. Supplementation requires remembering to take the supplement on a regular basis or having it regularly administered, but it can be an effective targeted intervention that is possible to adapt to an individual's needs. Massive dosage typically by intramuscular injection can also be a remedy for low vitamin D status in at-risk groups, although it does require medical personnel to administer the injection. Fortification of food can be a very effective means of delivering vitamin D providing the food is consumed regularly by a large proportion of the population. Although it may not result in a sufficient status, even low dose vitamin D provided in a fortified malted milk drink at 5 or 10 μg/d is able to significantly increase serum 25(OH)D concentrations( Reference Fisk, Theobald and Sanders 38 ). In the UK, foods tend to be fortified with small quantities based on the serving size, for example, breakfast cereals contain approximately 1·3 μg per 30 g portion and margarine 0·75 μg for every two teaspoons. Whilst the current levels of fortification do make a difference to vitamin D status as shown in the most recent National Diet and Nutrition Survey in adults aged 19–64 years where 19 % of vitamin D in the diet was coming from fat spreads and 13 % from cereals and cereal products, fortification at higher levels and in a product which is consumed widely in a high proportion of the population may be needed to help bring a greater proportion of the population to an adequate vitamin D status. We have demonstrated that a malted milk drink is an effective vehicle for vitamin D, but a product such as milk which is consumed extensively may be more effective. However, the potential hazards of inducing toxicity in a small minority have to be considered when fortifying foods. Although an upper level has not yet been established in the UK due to a lack of data, 25 μg/d has been considered as safe and unlikely to cause adverse effects in the general population( Reference Bassuk and Manson 136 ). This is low though compared with the tolerable upper intake level of 50 μg/d in North America and Europe( 137 , 138 ). Before policies could be introduced for fortification of a widely consumed product nationwide, it would be necessary to assess the risk and benefits in different vehicles. One approach is to use that employed by the EU Benefit-Risk Analysis of Foods project which included an estimation of the risk and benefits of universal folic acid fortification of flour/bread( Reference Verhagen, Andersen and Antoine 139 ). At present, foods are fortified with either D2 or D3, although most often D3 is used. As the forms seem to be equipotent at low doses, D2 is likely to be important for widespread fortification as it is more acceptable to vegetarians, vegans and religious groups who may not consume D3 due to it being from animal sources. Another advantage of fortifying foods with vitamin D2 is that it would allow levels of consumption to be monitored by measuring 25(OH)D2 concentrations, whereas D3 is also produced on exposure of the skin to UVB radiation. However, data are lacking on the bioavailability of vitamin D2 and D3 in different matrices and there is a need for further research in this area.
Conclusions
Although the observational evidence suggests that a low vitamin D status may increase the risk of CVD by about 40 %, this size effect is not supported by current evidence from clinical trials. Long-term, good quality, sufficiently powered and well-controlled RCT of several years duration are needed to examine the effect of vitamin D supplementation on clinical CVD endpoints in at risk populations.
Acknowledgements
The authors are grateful to Dr Hannah Theobald for her help and comments.
Financial Support
The author's research was supported by the funding from BBSRC and GlaxoSmithKline (protocol number RH01372).
Conflicts of Interest
None.
Authorship
C. M. F. drafted the present paper and T. A. B. S. discussed and modified it.
